ABSTRACT
Restriction fragment length polymorphism (RFLP) analysis is an economic and fast technique for molecular typing but has the drawback of difficulties in accurately sizing DNA fragments and comparing banding patterns on agarose gels. We aimed to improve RFLP for typing of the important human pathogen Streptococcus pneumoniae and to compare the results with the commonly used typing techniques of pulsed-field gel electrophoresis and multilocus sequence typing. We designed primers to amplify a noncoding region adjacent to the pneumolysin gene. The PCR product was digested separately with six restriction endonucleases, and the DNA fragments were analyzed using an Agilent 2100 bioanalyzer for accurate sizing. The combined RFLP results for all enzymes allowed us to assign each of the 47 clinical isolates of S. pneumoniae tested to one of 33 RFLP types. RFLP analyzed using the bioanalyzer allowed discrimination between strains similar to that obtained by the more commonly used techniques of pulsed-field gel electrophoresis, which discriminated between 34 types, and multilocus sequence typing, which discriminated between 35 types, but more quickly and with less expense. RFLP of a noncoding region using the Agilent 2100 bioanalyzer could be a useful addition to the molecular typing techniques in current use for S. pneumoniae, especially as a first screen of a local population.
Streptococcus pneumoniae is an important human pathogen with a clonal population structure (4). Molecular typing has become increasingly important in order to understand the evolution of this pathogen and to trace clones with special traits such as antibiotic resistance. Several methods for genotypic analysis of S. pneumoniae are currently in use. The “gold standards” are pulsed-field gel electrophoresis (PFGE) and multilocus sequence typing (MLST). In PFGE, the whole genome is digested with a restriction endonuclease and run slowly through an agarose gel with an alternating current (10). The fragment banding pattern produced is used to characterize the strain. PFGE has the advantage of high discrimination between strains but is relatively laborious, time-consuming, and costly and tends to be used to compare strains within a laboratory rather than between laboratories because it involves the comparison of banding patterns rather than absolute values. MLST allows a direct comparison of genotypes by comparing sequences within several housekeeping genes (4). This technique has proven extremely useful in making comparisons between strains not only within a laboratory but also internationally by using a Web-based database. Sequencing, however, is currently expensive and time-consuming.
Restriction fragment length polymorphism (RFLP) analysis is often used on a PCR amplification product that is subjected to endonuclease digestion followed by comparison of the resulting banding patterns produced on an agarose gel. This technique has been performed on coding regions such as the capsule genes (1) and penicillin-binding protein genes (12). It has some of the shortcomings of PFGE regarding reproducibility and interlaboratory comparison of strains. However, it is the most rapid and economic of the currently employed techniques. To improve the accuracy and reproducibility of RFLP, previous studies reported the use of the Agilent 2100 bioanalyzer for sizing of the DNA fragments. Nachamkin et al. (16) analyzed the flagellin gene of Campylobacter jejuni but found problems resolving fragments differing by 8 to 20 bp using the bioanalyzer compared to an agarose gel. In contrast, Lu et al. (11) studied mutations in human mitochondrial DNA by RFLP and found that the bioanalyzer showed better reproducibility than the conventional method. In this study, we demonstrate that the Agilent 2100 bioanalyzer is a useful tool for obtaining accurate and reproducible-enough sizing of RFLP fragments for molecular typing of S. pneumoniae. A noncoding DNA region of S. pneumoniae was chosen, which, unlike DNA encoding genes, is free to mutate without affecting the viability of the bacteria and therefore might be expected to be more variable than a coding region. One of the primers binds within the gene for pneumolysin to make the PCR amplification specific for S. pneumoniae. The technique should prove to be useful for the characterization of local S. pneumoniae populations in an era of increasing vaccine selection pressure (2).
MATERIALS AND METHODS
Bacterial strains.Clinical isolates of Streptococcus pneumoniae were selected from two nationwide surveillance programs that collect nasopharyngeal and invasive isolates (5, 8, 15). Both surveillance programs started in 1998 and are ongoing, with short interruptions between January 2000 and March 2002. Capsular serotypes were determined by the quellung reaction, as previously described (5, 15), or the “agglutination reaction” (Statens Serum Institut, Copenhagen, Denmark). A total of 47 encapsulated Streptococcus pneumoniae isolates representing 15 serotypes (serotypes 1, 4, 5, 6A, 6B, 7F, 8, 9V, 14, 15, 18C, 19F, 23F, 33, and 38) were used. For each serotype, three or four different isolates were randomly chosen from the strain collection. This included isolates from sterile body sites and from the nasopharynx.
PFGE.PFGE typing was done on all isolates as described elsewhere previously (9). SmaI was used for restriction digestion of chromosomal DNA.
MLST.Amplification, sequencing, and analysis of the seven housekeeping genes were carried out as described previously (4).
RFLP.DNA was isolated as described previously (13). The noncoding region between the pneumolysin gene and the preceding hypothetical protein gene (designated spr1738 in the R6 genome) (NCBI accession number NC_003098) was amplified by PCR using newly designed primers (forward primer NCRspanFor3 [5′-AAA GGC TGC ACG GAC ATT G-3′] and reverse primer NCRspanRev3 [5′-CCG ATT TGC CAC TAG TGC GTA AGC-3′]). Amplification was performed using the following cycling conditions: a primary denaturation step for 3 min at 94°C followed by 30 cycles of 94°C for 1 min, 45°C for 1 min, and 72°C for 3 min and ending with a final extension step for 10 min at 72°C. Each PCR mixture had a final volume of 50 μl, containing 1 unit of Taq polymerase, 0.2 mM deoxynucleoside triphosphates, 5 μl Taq buffer (all from Roche Molecular Biochemicals, Rotkreuz, Switzerland), 1 μM of each primer, and 100 ng DNA. The PCR products (approximately 1.2 kb) were quantified using the DNA 7500 kit for the Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA). Four hundred nanograms of each PCR product was digested separately with the restriction endonucleases DdeI, MseI, AluI, AfiIII, HaeIII, ApoI, TspRI, and TfiI (New England Biolabs, Ipswich, MA), according the manufacturer's instructions, in a total reaction mixture volume of 20 μl. Following digestion with DdeI, MseI, or AluI, the samples were heated to 65°C for 20 min, following digestion with the other enzymes to 80°C for 20 min, and then cooled on ice for 10 min to disrupt aggregates of DNA (7) before 1 μl of the digestion product was analyzed using the Agilent 2100 bioanalyzer according to the manufacturer's protocol.
Data analysis.PFGE patterns were analyzed with Bionumerics software (version 3.0; Applied Maths, Gent, Belgium). Patterns were clustered by the unweighted-pair group method with arithmetic means, and a dendrogram was generated from a similarity matrix calculated using the Dice similarity coefficient with an optimization of 1.0% and a tolerance of 1.5%. PFGE patterns with at least 90% similarity were given the same group number.
MLST analysis was performed using the pneumococcal MLST database (http://spneumoniae.mlst.net/ ), which is located at Imperial College London and is funded by the Wellcome Trust.
The RFLP patterns were analyzed with Bionumerics software. Bands were assigned according to the result tables and electropherograms from the bioanalyzer (threshold of 10 fluorescent units). Patterns were clustered by the unweighted-pair group method with arithmetic means, and a dendrogram was generated from a similarity matrix calculated using the Dice similarity coefficient with a position tolerance of 1%. Strains with a banding pattern of greater than 95% similarity were considered to be in the same group for the enzyme being tested.
Nucleotide sequence accession numbers.Sequences reported here have been deposited in the GenBank database and are available under accession numbers EF190945, EF190946, and EF190947.
RESULTS
Analysis of RFLP products by the Agilent 2100 bioanalyzer gives accurate and reproducible values for fragment sizes.The newly designed primers were effective in amplifying the targeted sequence for all 47 pneumococcal strains. Figure 1 demonstrates the advantages of using a bioanalyzer for the RFLP analysis. The PCR products of three strains, B102.79, B112.30, and B205.68, were digested with the restriction enzyme DdeI on different days, and the restriction fragments were analyzed by using the bioanalyzer. The simulated gel and the electropherogram results for all three strains were consistent on the different days (Fig. 1). The values for each band varied by a maximum of only 5 bp. The reproducibility of the results was significantly enhanced by the final heating step followed by the immediate transfer of the samples to ice (data not shown).
PCR products of three pneumococcal strains were digested with DdeI in three separate experiments performed on different days, and the DNA fragments were analyzed using the Agilent 2100 bioanalyzer. a, b, and c show the simulated gels for experiments 1, 2, and 3, respectively. d, f, and i show the electropherograms for strains B102.79, B112.30, and B205.68, respectively, for experiment 1; e, g, and j show electropherograms for experiment 2; and f, h, and k show electropherograms for experiment 3 showing the consistency of the banding patterns obtained on different days.
RFLP discriminates between different strains.Each of the 47 pneumococcal strains was digested separately with the restriction endonucleases. Figure 2 shows the fragment patterns from the bioanalyzer using the enzyme DdeI as an example along with the dendrogram produced from the similarity matrix by the Bionumerics software. A similar analysis was also performed for the other enzymes, but the results with AluI and TspRI were not reproducible and were excluded (data not shown). The results from the remaining enzymes were used to assign the groups, which are shown in Table 1. Each unique combination of group numbers was assigned an arbitrary RFLP type (Table 1).
The banding patterns for all 47 pneumococcal strains following digestion of the PCR product with DdeI are shown along with the dendrogram constructed according to the percent similarity between patterns using Bionumerics software. The group numbers were arbitrarily assigned to strains with at least 95% similarity in their banding patterns.
RFLP group for each strain after digestion with each enzyme and the consequent RFLP type
The 47 strains were divided into 33 RFLP groups, which preserved the classification according to serotype, with the exception of RFLP types 7, 8, and 12 (Table 1). In 12 of the 15 serotypes, further subgroups were differentiated. Digestion with DdeI or ApoI gave the highest number of banding patterns, which was seven each, whereas TfiI discriminated only between two RFLP groups. Digestion with the enzymes MseI, AfiIII, and HaeIII gave six, four, and four patterns, respectively.
Comparison of results obtained by RFLP, PFGE, and MLST.MLST yielded a total of 35 sequence types (Tables 2 and 3), and PFGE yielded a total of 34 groups compared to the 33 RFLP groups (Table 3). Grouping by serotype was preserved by PFGE and MLST and by RFLP, with three exceptions. MLST detected three new alleles not yet contained in the MLST database (Table 2). For strain 203.24, gdh was sequenced in both directions and was a 99% match to allele 5. However, the first 18 bp of spi could be sequenced only in one direction, but the remainder was sequenced in both directions and gave matching sequences. The sequence differed from that of allele 17 by 1 bp at position 234, which was in the region that was doubly sequenced. For strain B103.66, ddl was sequenced fully in one direction but lacked 42 bp at the 3′ end of the other. It differed at 17 bases from the most similar allele (allele 91), 14 of which were in the doubly sequenced region. Since the MLST database at http://spneumoniae.mlst.net/ accepts data only for new alleles that have been sequenced fully in both directions, the new sequences have been submitted to GenBank. For all other strains, the MLST data are available at the http://spneumoniae.mlst.net/ site.
MLST alleles and the consequent sequence types for each strain
Comparison of molecular typing by PFGE, RFLP, and MLST for each strain
Subgrouping within serotypes by RFLP matched the results from PFGE for seven serotypes (serotypes 1, 5, 6B, 8, 14, 15, and 23F) and matched the results from MLST for seven serotypes (serotypes 1, 5, 6B, 14, 18C, 23F, and 38). Subgrouping by PFGE and MLST matched for nine serotypes (serotypes 1, 4, 5, 6A, 6B, 7F, 14, 19F, and 23F). For five serotypes, all three techniques exhibited an identical discriminatory power (serotypes 1, 5, 6B, 14, and 23F). For serotype 33, there was no concordance among the techniques.
DISCUSSION
Restriction fragment length polymorphism analysis is not a new technique for the typing of S. pneumoniae (1, 3, 6, 12, 18) but has been improved here by using an Agilent 2100 bioanalyzer for the accurate sizing of the DNA fragments. In addition, a PCR product of a noncoding region was used, which is likely to vary more between strains than a gene.
Originally, RFLP had the disadvantage of the possibility of ambiguous bands on gels, but by using the bioanalyzer, accurate sizing and quantification of each DNA fragment are possible, allowing standardization and comparison of samples run on different days. This is enabled by the fact that in addition to a well containing a molecular weight ladder for sizing, each sample has an upper and lower molecular weight marker, allowing accurate comparisons between samples (14). Nachamkin et al. (16) used the Agilent 2100 bioanalyzer for RFLP analysis of the Campylobacter jejuni flagellin gene but found that it was difficult to resolve samples containing multiple DNA fragments differing by only 8 to 20 bp. In this study, the enzymes used produced patterns that could be resolved by the bioanalyzer. Lu et al. (11) found that variations in fragment sizes obtained among bioanalyzer chips were less than 10 bp, and this is in agreement with our finding of a variation of 5 bp or less. By using the data produced following digestion with several enzymes separately, we increased the discriminatory power of the technique. The bioanalyzer has the additional advantage over an agarose gel in that it requires only 30 min to run 12 samples. One important consideration with RFLP is that the digestion is complete and reproducible, making the choice of restriction enzyme critical.
Here, RFLP analysis was limited to one area of the genome, but the discriminatory power of the technique could be improved by amplifying several regions. However, even amplifying and digesting this one region with six restriction enzymes allowed almost as much discrimination as MLST involving the sequencing of regions of seven housekeeping genes. The region to amplify was chosen because one primer binds within pneumolysin, which should make the primer target specificity to S. pneumoniae high. The other primer binds within the preceding gene of unknown function. Because the primers bind within genes, the regions to which they bind should be quite conserved, and indeed, this seems to be the case, as the PCR was successful for all the 47 strains, of 15 different serotypes, tested. However, the majority of the PCR product corresponds to the noncoding region between the two genes, which was chosen deliberately since it is expected to vary more than coding regions between strains. There does seem to be variation within this region, because within the 47 different strains, there were 33 different RFLP types when the digestion patterns using the six restriction enzymes were considered.
MLST has the advantage of providing unambiguous results because DNA sequences, rather than banding patterns, are analyzed (4). Even a 1-bp difference between two strains will give a different allele number and therefore a different sequence type. Sequence types are easy to compare between laboratories. However, having the same multilocus sequence type does not mean that the strains are genetically identical, as described recently by Silva et al. (17), since only a tiny fraction of the genome is sequenced. Also, for every strain analyzed, seven gene fragments must be sequenced in both directions, making a total of 14 sequences. Sequencing is a relatively expensive technique and is time-consuming in terms of running the samples and checking the quality of the sequences.
RFLP analysis of a noncoding region of S. pneumoniae with accurate sizing of DNA fragments by using a bioanalyzer could be a useful addition to methods currently in use for characterizing S. pneumoniae strains by allowing a quick and relatively inexpensive comparison. It may be useful as a first analysis to determine whether two strains of the same serotype isolated from a local population are different or members of the same clone. Strains that cannot be differentiated by RFLP could then be analyzed by an alternative technique such as MLST, thereby limiting this relatively expensive method to closely related strains.
RFLP should therefore prove to be useful for characterizing regional S. pneumoniae populations in an era of increasing vaccine selection pressure. The principle of using a bioanalyzer could be applied to any other RFLP-based typing method for bacteria, fungi, or viruses.
ACKNOWLEDGMENTS
The study was supported by a grant of the Swiss National Science Foundation to K. Mühlemann, grant no. 3200-067998. A. Martynova received a grant from the Infectious Diseases Research Fellowship Program (www.isid.org ). The fellowship was jointly sponsored by the International Society for Infectious Diseases, the Swiss Society for Infectious Diseases, and the Institute for Infectious Diseases, University of Bern, Bern, Switzerland.
FOOTNOTES
- Received 24 October 2006.
- Returned for modification 11 December 2006.
- Accepted 21 December 2006.
- Copyright © 2007 American Society for Microbiology
