Objective
The objective was to examine the use of a tailor-made DNA microarray containing probes representing the vaginal microbiota to examine bacterial vaginosis.
Study Design
One hundred one women attending a health center for HIV testing in South Africa were enrolled. Stained, liquid-based cytology slides were scored for bacterial vaginosis. An inventory of organisms was obtained using microarray technology, probing genera associated with bacterial vaginosis in more detail, namely Gardnerella, Atopobium, Dialister, Leptotrichia, Megasphaera, Mobiluncus, Peptostreptococcus, Prevotella , and Sneathia .
Results
Of 101 women, 34 were diagnosed positive for bacterial vaginosis. This condition was associated with an increased microbial diversity. It is no longer useful to base the diagnosis of bacterial vaginosis on Gardnerella alone. Rather, its presence with Leptotrichia and Prevotella species, and especially Atopobium was more indicative of an aberrant state of the vaginal flora.
Conclusion
To understand the vaginal microbiota in more detail, microarray-based identification can be used after microscopic scoring.
The vaginal microbiota play a pivotal role in maintaining health. When this predominantly Lactobacillus community is disrupted, decreased in abundance and replaced by Gardnerella, Atopobium, Dialister, Leptotrichia species , Megasphaera, Mobiluncus, Peptostreptococcus, Prevotella , and Sneathia , bacterial vaginosis (BV) results. This diverse microbiota, detected by microscopy and clinical means is extremely common among HIV positive patients, with 1 study of 2292 pregnant HIV positive women in Tanzania showing a prevalence of 47.8%.
Recently, bacteria associated with BV have been detected by polymerase chain reaction (PCR) and PCR-based microarrays, methods which enable the analysis of complex bacterial systems that are often difficult to culture. In mainstream medical practice, and particularly in developing countries, diagnosis and treatment of BV is undertaken with mostly clinical signs of discharge and malodor and the finding of a pH elevated above 4.5. BV can be established light microscopically on stained conventional smears (Nugent score), or in so called liquid-based cytology (LBC). The biofilms of adhesive material of which the bacteria are caught are visible in the stained LBC slide. The biofilms covering the epithelial cells are indicated as “glue” cells. This allows us to adjust the Nugent score to LBC slide. The inability to determine the actual organisms present and their abundance means that medical intervention of BV is suboptimal. We hypothesized that a tailor-made DNA microarray, containing probes that represent BV-associated genera, could be developed with a long-term view of providing a more insightful diagnosis of this aberrant condition. The objective of the current study was to assess the usefulness of a microarray on fixed vaginal samples from South African women.
Material and Methods
Subjects
Women attending the Ndlovu Medical Centre in Elandsdoorn in Moutse District, South Africa, for HIV testing were offered enrollment. Information was provided to the study participants and oral consent was obtained by the doctors for a vaginal sample to be examined for microbiota. Pregnant women or ≤16 years were excluded. None of the HIV-positive women were receiving antiretroviral therapy at the time of sample collection. After collection of a vaginal sample with a standard sampling brush, the physician placed the brush head into a vial with coagulant fixative, BoonFix (Denteck, Zoetermeer, The Netherlands), which consists of ethyl alcohol, low molecular weight PEG, and acetic acid. This substance has no detrimental effects on DNA preservation. The fixed samples, coded by a method that did not disclose any subject names, were sent to the Leiden Cytology and Pathology Laboratory (LCPL) in The Netherlands.
Light microscopic scores and the diagnosis of BV
All microscopy scores were performed on stained LBC slides. These were scored by 2 experienced microscopists. The score for Gardnerella was based on the presence of cells staining dark blue due to the thick biofilm containing coccoid bacteria, adhering to squamous cells. These glue cells are also known as Blue Mountain cells ( Figure 1 ). This is determined as follows: in 10 random fields of view, at low magnification. The mean for the 10 fields of view determines the Gardnerella score: a score of G0 is given when zero percent of the squamous cells are glue cells: a score of G1 = 1-5%, a score G2 = 6-20%, and a score G3 for >20%. Biofilms adhering to squamous cells which contain presumptive lactobacilli rods can be seen. Dislodged biofilms with lactobacilli can also be noted, because the biofilms containing lactobacilli are less prominent. The scoring was performed at high magnification, with the mean calculated from ten fields of view. A Lactobacillus score of L0 is given for absence of long, thin bacteria indicative of lactobacilli: a score of L3 is large numbers of lactobacilli (>20%), scores of L1 (1-5%), and L2 (6-20%) are in between scores.
The cases diagnosed as BV positive include G2+L0, G3+L0, and G3+L1. The cases diagnosed as BV intermediate include G1+L0, G1+L1, G2+L1, G2+L2, and G3+L2. The cases diagnosed as BV negative include all remaining combinations ( Tables 1 and 2 ). This methodology of defining BV light microscopically on thin layer slides, is similar but not identical to the Nugent scoring used for conventional smears, therefore, we refer to this scoring system as the “Nugent-Boon” method. The latter was used in our cytology practice for over 5 years and proved to correlate well with quantitative PCR of L crispatus and G vaginalis.
| Gardnerella code | Lactobacillus code | |||
|---|---|---|---|---|
| L0 | L1 | L2 | L3 | |
| G0 | 4 | 19 | 15 | 3 |
| G1 | 9 | 12 | 1 | 0 |
| G2 | 14 | 3 | 1 | 0 |
| G3 | 19 | 1 | 0 | 0 |
| HIV | BV | ||
|---|---|---|---|
| Negative | Intermediate + positive | Sums | |
| Positive | 12 | 40 | 52 |
| Negative | 30 | 19 | 49 |
| Sums | 42 | 59 | 101 |
The microarray
The results of massive parallel sequencing, performed according to described procedures, of a number of vaginal microbiota samples as well as literature research have laid the foundation for the probe design. Taxonomic selection was confirmed and expanded based on DGGE analysis (data not shown). For each bacterial genus represented on the microarray 1 or more unique short oligonucleotide sequences from within the 16s rDNA gene were selected. Criteria for sequence selection, apart from being unique, included length and melting temperature. Short oligonucleotide sequences (approximately 20 nt) were used with a melting temperature of 60°C according to the Wallace rule for which a 1 nucleotide mismatch already resulted in an absence (or very strong decrease) of signal after hybridization. Although the emphasis in probe selection was on unique species specific probes sequences representing groups of organisms (at higher taxonomic levels than species) were also selected. These latter sequences mostly represent a large part of a higher taxonomic unit (genus, family, order). These sequences were selected by ARB software.
Construction of a vaginal microbiota-representing microarray
Microarrays were constructed according to previously described methods. In brief, oligonucleotides with a 5′ NH2-C6 extension for improved attachment were dissolved in 50 mM phosphate buffer (pH 9) at 25 μM and spotted on CodeLink slides (GE Healthcare, Little Chalfont, Buckinghamshire, UK) through TeleChem SMP3 quill pins in a SDDC-2 Eurogridder (ESI, Mississauga, Ontario, Canada). After spotting, microarray slides were blocked according to the manufacturer’s instructions and stored at room temperature under nitrogen.
DNA isolation for microarray analysis
DNA was isolated from the BoonFixed vaginal samples. For this, the samples were mixed with 150 μL lysis buffer BL, 350 μL zirconium beads (0.1 mm), and 200 μL phenol and introduced into a BeadBeater (BioSpec Products Inc, Bartlesville, OK) for 2 minutes. Samples were then allowed to cool, and centrifuged for 10 minutes at 4000 rpm. The water phase was transferred to a new tube and mixed with 2 volumes binding buffer and 10 μL magnetic beads (both Agowa). After mixing, magnetic beads were captured with a magnetic separator. Supernatant was removed and beads were washed twice with buffer. After the last washing step, elution buffer was added for 10 minutes incubation at 55°C, before beads were captured with a magnetic separator and the supernatant containing the eluted DNA was stored at –20°C until further use. The concentration of the nucleic acids was determined visually after electrophoresis on an ethidium bromide stained agarose gel by comparison with standards of known concentration.
DNA labeling and hybridization
PCR was performed on each of the samples in mixtures containing 0.2 mM dNTPs: 2.5 mM MgCl2, 12.5 pmol each of primers 16s-8-F (AGAGTTTGATCHTGGYTCAG) and 16s-1061-R (TCACGRCACGAGCTGACGAC), 1.25 U Goldstar Taq DNA polymerase (Eurogentec, Seraing, Belgium), 1 times Goldstar reaction buffer, and 5 ng of bacterial DNA. Primer 16s-8-F contained a 5′ phospho modification, primer 16s-1061-R a 5′-C6 Cy3 modification. PCR was performed using the following program: 94°C for 2 minutes, 30 cycles of 94°C for 30 seconds, 50°C for 40 seconds, 72°C for 80 second, 1 cycle of 72°C for 2 minutes, and a cool down to 4°C. The PCR products were purified with a Sephadex column (Autoseq G-50; GE Healthcare), according to the manufacturer’s instructions. The concentration of the PCR products was estimated by electrophoresis on an ethidium bromide stained agarose gel. The PCR products were purified with a Sephadex column (Autoseq G-50, GE Healthcare), according to the manufacturer’s instructions. The samples were dried by speed-vacuum centrifugation and a mixture of 0.5 μL Strandase enzyme (NovaGen, Merck, Madison, WI), 1 μL Strandase buffer, and 6.5 μL water was added and incubated for 30 minutes at 37°C, followed by inactivation during 10 minutes at 75°C. Next, 12 μL water was added to the solution; DNA was purified again with an Autoseq G-50 column and dried. The concentration and integrity of the PCR product (1053 bp) and single-stranded DNA were analyzed by electrophoresis on an ethidium bromide stained agarose gel.
The single-stranded sequences were dissolved in 40 μL hybridization buffer (Easyhyb; Roche, Basel, Switzerland) and the solution was heated for 2 minutes in a thermoblock at 95°C. Oligonucleotide microarray slides were covered with hybridization mix and cover slips and were placed in an incubator shaker during 4 hours (37°C and 170 rpm). After hybridization, the slides were washed with 1 × SSC, 0.2% SDS, then 0.5 × SSC at 37°C. Two additional washing steps were conducted at room temperature: 5 minutes 0.2 × SSC at 280 rpm. The slides were dried with N 2 flow and scanned with a ScanArray Express 4000 Scanner at 10 μm pixel size (Perkin-Elmer, Waltham, MA).
Data analysis
Imagene 5.6 software (BioDiscovery, Marina del Rey, CA) was used to analyze the results. Signals were quantified by calculating the mean of all pixel values of each spot and calculating the local background around each spot. For each spot, a signal-to-background ratio was calculated. For further analysis, those spots were selected that had a minimal number of observations more than 2 times above the local background. This cutoff was selected based on the observation that negative control spots never resulted in signals above this cutoff (data not shown). The minimal number of observations more than 2 times above the local background for each spot was set at 10, and this criterion was mainly used to discard data resulting from technical noise. The data matrix obtained in this way was used for hierarchical clustering and SAM analysis programs (TM4). Each microarray probe was scored by counting only positive detection, ie, having a value above the detection threshold (≥5 signal-to-background ratio). Only probes with high numbers of detections were used for further analysis. In prior publications, a wide range of bacteria using nonculture test were described ; however, from the published literature and after discussions with prominent microbiologists, we chose BV-relevant, potentially pathologic bacteria and highlighted them in more detail to clarify and strengthen the analysis of 4 Prevotella species, 3 Sneathia / Leptotrichia species, 2 Mobiluncus species, and 5 other bacteria ( Table 3 ).
