| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Institute of Medical Microbiology, Otto-von-Guericke-University Magdeburg, 39120 Magdeburg, Germany
Correspondence
Frank L. Thies
frank.thies{at}medizin.uni-magdeburg.de
Received 8 February 2006
Accepted 31 January 2007
Abbreviations: BV, bacterial vaginosis; BVAB, bacterial vaginosis-associated bacterium; 6-FAM, 6-carboxyfluorescein; T-RF, terminal restriction fragment; T-RFLP, terminal restriction fragment length polymorphism.
 |
INTRODUCTION |
|---|
 TOP INTRODUCTION METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Bacterial vaginosis (BV) is a common disease with a reported prevalence ranging from 4 to 40 % in diverse populations and is associated with a number of severe gynaecological and obstetric complications (Sobel, 2000). From the microbiological point of view, BV is a polymicrobial syndrome characterized by disturbance of the normal vaginal microbiota and appears to be the result of an aetiologically unknown process in which the physiological microbial flora is replaced by a still ill-defined set of mostly anaerobic bacteria.
Based on conventional culture identification, an association has been found between BV and the detection of Gardnerella vaginalis, Mobiluncus mulieris and anaerobic Gram-negative bacteria, especially Prevotella spp. Recent findings, mostly based on molecular methods, have emphasized the association between Atopobium vaginae and BV (Ferris et al., 2004). However, the alleged pivotal role of these bacteria has been questioned, as they were also found in the vaginal fluid of healthy women. Therefore, recent research has focused on the development of culture-independent methods for in-depth analysis of vaginal microbiota. Detailed information on the complex microbial communities in BV as well as under normal conditions has been obtained by cloning and sequencing of 16S rRNA gene libraries (Verhelst et al., 2004; Zhou et al., 2004; Hyman et al., 2005; Fredricks et al., 2005). As a result of this work, in addition to Atopobium vaginae, relatively unknown bacterial species such as ‘Leptotrichia amnionii’ (Shukla et al., 2002), Eggerthella sp., Sneathia sanguinegens and Megasphaera sp. have been implicated in the pathogenesis of BV.
Diagnosis of BV is based on the presence of clinical signs as well as microbiological findings. Among the many laboratory methods used for the diagnosis of BV, the Gram-stain criteria as defined by Nugent et al. (1991) are currently regarded as the standard procedure. Although this method has been repeatedly refined (Pereira et al., 2005; Verhelst et al., 2005), morphological assessment alone may not reflect the microbiological variations among BV individuals and should be extended by molecular methods (Fredricks & Marrazzo, 2005).
Terminal RFLP (T-RFLP) is a promising molecular approach for the analysis of microbial ecosystems that is increasingly being used by microbiologists for the study of complex human microbiota in health and disease (Rogers et al., 2004; Sakamoto et al., 2004). Unlike competing methods used to analyse microbial communities, such as denaturing gradient gel electrophoresis and single-strand conformation polymorphism, T-RFLP offers the advantage that appropriate computer software can generate in silico predictions of terminal restriction fragments (T-RFs), thereby permitting straightforward identification of bacterial species from fingerprinting data. As analysis of clone libraries is time-consuming as well as cost-intensive, it cannot be applied in microbiological laboratories for routine purposes. Therefore, we attempted to establish T-RFLP profiling as a rapid and relatively inexpensive method for community analysis of the vaginal fluid. This will allow the application of T-RFLP as a molecular tool whenever an in-depth analysis of the vaginal flora is desirable, for example within a clinical studies framework.
|
METHODS |
|---|
|
TOPINTRODUCTIONMETHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
DNA isolation. For DNA isolation, a QIAamp DNA mini kit (Qiagen) was used according to the manufacturer's instructions, with minor modifications. Swab specimens were swirled in 180 µl lysis buffer containing 1 % (w/v) Triton X-100, 0.5 % (w/v) Tween 20 and 1 mmol EDTA in 10 mM Tris/HCl (pH 8.0). Mutanolysin (2 µl containing 50 U; Sigma) was added to the lysate and incubated for 30 min at 37 °C. After mixing samples with 200 µl buffer AL (Qiagen) and 20 µl proteinase K (20 mg ml–1), samples were incubated for 30 min at 56 °C followed by 15 min at 95 °C. After adding 200 µl ethanol, samples were loaded onto a QIAmp spin column, washed and eluted in 200 µl buffer AE (Qiagen), as described by the manufacturer. DNA samples were stored at –20 °C until further use.
T-RFLP analysis
(i) PCR and digestion.
Universal bacterial primers 27F (5'-AGAGTTTGATCCTGGCTCAG-3') and 926R [5'-CCGTCAATTC(A/C)TTT(A/G)AGTTT-3'] were used to amplify internal fragments of 16S rRNA genes in the genomic DNA obtained from samples. Primers 27F und 926R were labelled at their 5' ends with the dyes 6-carboxyfluorescein (6-FAM) and 4,7,2',4',5',7'-hexachloro-6-carboxyfluorescein (HEX), respectively. All primers were obtained from MWG-Biotech. PCR mixtures comprised PCR buffer, 1.5 mM MgCl2, each deoxynucleoside triphosphate at a concentration of 0.2 mM, each primer at a concentration of 0.2 µM and 1 U Taq polymerase (Qiagen) in a final volume of 50 µl. The cycling program was performed using a Perkin-Elmer 2400 thermocycler with the following conditions: an initial denaturation step at 95 °C for 5 min, followed by 35 cycles of denaturation for 30 s at 95 °C, annealing for 30 s at 50 °C and elongation for 1 min at 72 °C, with a final additional elongation step for 7 min at 72 °C. PCR products were analysed on 1 % agarose gels stained with ethidium bromide. After purification using the QIAquick PCR purification kit, PCR products were digested with HinfI, HhaI and MspI (New England Biolabs). Digestion reactions were performed separately for 4 h at 37 °C, followed by treatment for 20 min at 65 °C for enzyme inactivation.
(ii) T-RF length analysis. The lengths of T-RFs were determined by electrophoresis with a model 3100 automated sequencer (Applied Biosystems Instruments), as described previously (Trotha et al., 2002). In brief, samples were prepared by combining 1 µl restriction digestion product, 21 µl Hi-Di formamide solution (Applied Biosystems), 0.5 µl Genescan 500 ROX size standard (Applied Biosystems) and 0.5 µl of a 780 bp ROX-labelled DNA fragment (12 ng µl–1). After electrophoresis, the lengths of T-RFs were determined by comparison with the internal standard, using the local Southern algorithm as the size-calling method (Southern, 1979). Only peaks with heights exceeding 50 fluorescence units were evaluated.
(iii) In silico analysis. An in-house software program (JOparin; unpublished) permitted microbial identification from T-RF data. The program performed an in silico digestion of all 16S rRNA gene sequences available in the 16S rRNA gene database (release 9.24) downloaded from the RDP-II website (http://rdp.cme.msu.edu/misc/resources.jsp) (Maidak et al., 2001). The original dataset contained 119 821 sequences, but all entries without unambiguous identification of the bacterial source were skipped, leaving 52 725 rRNA gene sequences. To this well-annotated database generated so far, we added 439 rRNA gene sequences from Hyman et al. (2005) that have been deposited in GenBank (GenBank accession nos AY958774–AY959212). Sample T-RF patterns were compared with all in silico patterns and a matching score (RF score) was assigned to each database entry. The RF score was computed from the size-weighted sum of differences between observed and predicted T-RF lengths.
T-RF sequencing. To prove experimentally the correct in silico identification of T-RFs, the 16S rRNA gene was amplified from genomic DNA with unlabelled primers 27F and 926R using the cycling conditions as described above. The PCR product was column purified and digested with an appropriate restriction enzyme (HinfI, HhaI or MspI). The restriction fragments were separated in a 2.5 % agarose gel and the fragment corresponding to the T-RF was cut from the gel and purified (QIAquick gel extraction kit; Qiagen). Cycle sequencing of this fragment with 27F or 926R as the sequencing primer was performed with the Big Dye terminator kit (Applied Biosystems) as recommended by the manufacturer. Sequences were analysed using the BIBI software tool (http://pbil.univ-lyon1.fr/bibi/query.php) (Devulder et al., 2003).
Statistical analysis. Basic summary statistics and hierarchical cluster analysis were carried out using the free software package R (R Development Core Team, 2005). For cluster analysis, we constructed a binary matrix with bacterial species (or respective taxonomic units) as rows and BV samples as columns considering only the presence or absence of species. Similarity, i.e. frequency of co-occurrence, between species was measured with Jaccard's coefficient, which is equal to the number of samples in which both species occur, divided by the total number of samples with either species present (Riley, 2004). Hierarchical clustering was performed by the unweighted pair group method using arithmetic averages (UPGMA).
|
RESULTS AND DISCUSSION |
|---|
|
TOPINTRODUCTIONMETHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Reproducibility and accuracy of the T-RFLP assay
Initially, we investigated the technical reproducibility of T-RFLP patterns. PCR, enzyme digestions and T-RFLP analysis were performed on three vaginal samples. After 3 weeks, the procedures were repeated using the same DNA, which had been stored at –20 °C. Comparison of corresponding T-RFLP patterns demonstrated excellent reproducibility, confirming the observations of Osborn et al. (2000). The mean difference between T-RFs was <1 bp. The peak heights displayed a higher level of variation; however, this had no impact on species identification (data not shown).
Differences between observed and predicted T-RF sizes usually were within –2 to +2 bp of the predicted T-RF size. However, the differences grew larger for T-RFs >500 bp (data not shown). This T-RF drift has been reported previously and appears to be affected by subtle differences in molecular mass, from either purine content or dye label (Kaplan & Kitts, 2003). We observed 109 distinct T-RFs from 50 to 500 bp, which could be attributed to a total of 23 identifiable T-RF patterns. The mean difference between maximum and minimum values of all T-RFs was 1.29 bp, demonstrating the very low variability among observed T-RF sizes corresponding to distinct species or phylotypes from different samples (as a representative example, see Table 1
for a detailed analysis of MspI-generated T-RFs). These data suggested that T-RF length variation due to strain-to-strain differences was negligible and that overall reproducibility, defined as technical and biological reproducibility combined, could be regarded as sufficiently high.
|
Based on the concordance between observed and predicted fragment sizes, a ranked list of bacterial species was generated. The bacterial species with the highest RF score was regarded as the most probable source of the T-RF pattern. As noticed by others (Kaplan & Kitts, 2003), for some T-RF patterns there was a considerable difference between observed and predicted T-RF sizes, making in silico assignments problematic. In all cases with an equivocal assignment, which represented approximately 40 % of all patterns, we sequenced at least one T-RF from the pattern in question. T-RF sequencing provided DNA sequences that were mostly short (300–500 bp), but of sufficient quality. Together with the T-RF pattern, this usually permitted bacterial identification at least to the genus level. Otherwise, further T-RFs were sequenced. Thereafter, due to the high reproducibility of the assay, the assignment was considered accurate if the same T-RF pattern occurred in further samples.
Depending on the T-RF pattern, it was not always possible to obtain a conclusive identification at the species level. For example, Lactobacillus iners could be identified unambiguously, whereas Lactobacillus crispatus could not be differentiated from Lactobacillus suntoryeus, Lactobacillus amylovorus, Lactobacillus kalixensis or Lactobacillus ultunensis (further designated the Lactobacillus crispatus group). Similarly, Mobiluncus mulieris was indistinguishable from Mobiluncus curtisii, leading to our diagnosis as Mobiluncus sp. Assigning valid species names was particularly difficult for bacteria from the genus Prevotella. At least three distinct T-RF patterns were detected, which corresponded to Prevotella bivia, Prevotella buccalis and an uncultured Prevotella sp. (most similar to GenBank accession no. AY959212). However, T-RF sequencing showed that the first two T-RF patterns comprised more than a single phylotype (mostly uncultured Prevotella spp.) and were therefore designated P. bivia group and P. buccalis group. Due to the low variation among the 16S rRNA genes of bacteria from the family Enterobacteriaceae, it was not possible to achieve identification at the genus level with reasonable certainty. Therefore, bacteria accounting for such T-RF patterns were designated ‘enterobacteria’.
Not all T-RFs could be assigned to a bacterial taxon. In such cases, no corresponding database entry may exist, or pseudo-T-RFs may be produced by incomplete digestion (Osborn et al., 2000) or by a malfunctioning of the PCR process (Egert & Friedrich, 2003). In our experience, incomplete digestion did occur, especially if a species was abundant in a bacterial community (see, for example, the peaks corresponding to Atopobium vaginae in Fig. 1a). However, the computer software we used was capable of including such information so that species identification might be further supported by partially digested terminal fragments. However, in some rare cases involving co-occurrence of very closely related species or strains, partial digests made a definite diagnosis extremely difficult.
|
Analysis of the BV-associated bacterial community
In the 50 BV samples, T-RFLP displayed a high level of bacterial diversity. Twenty-three different T-RF patterns corresponding to ‘operational taxonomic units’ (further named bacterial species or phylotypes) were identified unambiguously (Table 2). On average, in each BV sample, 6.3 T-RF patterns were found (range 2–14). Fig. 1(a)
shows a typical T-RFLP pattern. The frequency at which each of these 23 species or phylotypes occurred in the BV sample set ranged from 96 % (Atopobium vaginae, the dominating bacterium in our BV samples) to 2 % (Lactobacillus crispatus, present in a single sample). An uncultured Megasphaera sp., Lactobacillus iners and G. vaginalis were frequently detected (68, 64 and 64 %, respectively). The 16S rRNA gene sequence with the most significant RF score for the Megasphaera T-RF pattern was GenBank accession number AY959093, which showed highest similarity (95 %), among all unambiguously annotated sequences, to the Megasphaera elsdenii 16S rRNA gene. Several bacterial species or phylotypes occurred with moderate relative frequencies from 44 to 10 %: Eggerthella sp., several Clostridium-like and Prevotella spp., Peptostreptococcus (Micromonas) micros, Aerococcus christensenii, ‘Leptotrichia amnionii’, Peptoniphilus sp., Dialister sp., Mycoplasma hominis and bacteria from the family Enterobacteriaceae. The three clostridial T-RFs showed highest concordance with the sequences of GenBank accession numbers AY958888, AY959097 and AY995273, respectively. The corresponding bacteria were designated bacterial vaginosis-associated bacterium 2 (BVAB2), BVAB1 and BVAB3 by Fredricks et al. (2005), as they were only distantly related to known clostridia (16S rRNA gene identity 
90 %). In our BV sample set, BVAB2 occurred with highest frequency (36 %), followed by BVAB1 (18 %). BVAB3 was only rarely detected (4 %). In addition to the aforementioned bacteria, five further species were identified in <10 % of all specimens: Sneathia sanguinegens, Anaerococcus tetradius, Mobiluncus sp., Finegoldia magna and Lactobacillus crispatus. Although high concentrations of Mobiluncus-like morphotypes, visible on a Gram stain, are commonly regarded as indicative of BV (with Nugent scores of 9 or 10), we could identify Mobiluncus sp. in only two samples. This surprising result is in concordance with data from Fredricks et al. (2005), who detected Mobiluncus mulieris in only one expanded rRNA gene library (out of nine patients with BV) when 420 clones of this library were sequenced. It is further noteworthy that in the study of Hyman et al. (2005), in which approximately 1000 16S rRNA gene library clones from the vaginal fluids of 20 premenopausal women were sequenced, no Mobiluncus-typical sequences were found. It may be speculated that in older studies, which relied partially on microscopical examination of vaginal smears, the frequency of Mobiluncus sp. was overestimated. Fluorescent in situ hybridization technology has demonstrated that at least one uncultured BV-associated bacterium (BVAB1) has a curved rod morphology very similar to that of the Mobiluncus morphotype (Fredricks & Marrazzo, 2005).
|
In order to search systematically for the co-occurrence of phylotypes within the BV microbiota, we used hierarchical cluster analysis as an explanatory data analysis tool. The resulting dendrogram displayed one cluster formed by Atopobium vaginae, Lactobacillus iners, G. vaginalis and Megasphaera sp. (Fig. 2). However, this may simply reflect the abundance of these species in the BV samples. In 22 % of the BV-positive samples, all four species were detected by T-RFLP, whilst at least three were seen in 64 % of BV samples. Zhou et al. (2004) reported that Megasphaera sp. and Atopobium vaginae were highly coincident in vaginal 16S rRNA gene libraries, with one being found only when the other was also present, and Verhelst et al. (2004) demonstrated the co-occurrence of Atopobium vaginae and G. vaginalis in BV samples based on species-specific PCR analysis. In the current study, we observed that Megasphaera sp. was only found in BV-positive samples when Atopobium vaginae was present, although this may simply reflect the abundance of Atopobium vaginae in our BV-positive samples. However, we did not observe a strict co-existence, as Atopobium vaginae was detected without concomitant Megasphaera-typical T-RFs in 14 BV-positive samples (Table 2). This was similar for the co-existence of Atopobium vaginae and G. vaginalis, with 16 BV-positive samples shown to contain Atopobium vaginae but not G. vaginalis. These discrepancies between the current data and earlier findings may reflect differences in the sensitivity of T-RFLP and other molecular-based assays.
|
Analysis of the vaginal bacterial community not associated with BV
In 20 samples with normal vaginal microbiota, only Lactobacillus spp. could be detected using our T-RFLP approach (see Fig. 1b for a typical T-RFLP pattern). Lactobacilli from the Lactobacillus crispatus group were detected in 70 % of BV-negative samples, whilst Lactobacillus iners was seen in 55 % and bacteria from the Lactobacillus gasseri group (comprising Lactobacillus gasseri, Lactobacillus acidophilus and Lactobacillus johnsonii) in 10 % of non-BV samples. The fact that only Lactobacillus spp. were observed in the BV-negative samples provided a clear distinction from the BV-positive samples. The discrepancy of this observation (only lactobacilli seen in non-BV samples) compared with other studies is probably due to the competitive nature of the T-RFLP amplification technique. If a mixed bacterial community comprises highly abundant (or predominant) species, subdominant species or phylotypes may be below the detection limit of T-RFLP. However, such subdominant bacterial groups may readily be detectable by specific PCR or culturing on selective media.
Conclusions
We have established T-RFLP profiling as a molecular tool for the routine analysis of the vaginal microbiota. T-RFLP profiling accurately distinguished between all BV-positive and BV-negative samples, as assessed by Nugent's Gram-stain criteria. As BV is a polymicrobial disease, it is important to consider the microbial community as a whole, rather than focusing on single presumptive ‘key organisms’. T-RFLP profiling may prove particularly valuable in an attempt to characterize subgroups of women who are more prone to certain complications of BV, such as preterm birth or amniotic fluid infection. Thus T-RFLP profiling may contribute to a better understanding of the aetiology of BV and its complications.
|
REFERENCES |
|---|
|
TOPINTRODUCTIONMETHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. F. Petrosino, S. Highlander, R. A. Luna, R. A. Gibbs, and J. Versalovic Metagenomic Pyrosequencing and Microbial Identification Clin. Chem., May 1, 2009; 55(5): 856 - 866. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. A. Frank, C. I. Reich, S. Sharma, J. S. Weisbaum, B. A. Wilson, and G. J. Olsen Critical Evaluation of Two Primers Commonly Used for Amplification of Bacterial 16S rRNA Genes Appl. Envir. Microbiol., April 15, 2008; 74(8): 2461 - 2470. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. K. Kuramitsu, X. He, R. Lux, M. H. Anderson, and W. Shi Interspecies Interactions within Oral Microbial Communities Microbiol. Mol. Biol. Rev., December 1, 2007; 71(4): 653 - 670. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| INT J SYST EVOL MICROBIOL | J MED MICROBIOL | MICROBIOLOGY | J GEN VIROL | ALL SGM JOURNALS |
10 %) detected in BV samples
