Fluorescence in situ Hybridization method using Peptide Nucleic Acid probes for rapid detection of Lactobacillus and Gardnerella spp.

Background Bacterial vaginosis (BV) is a common vaginal infection occurring in women of reproductive age. It is widely accepted that the microbial switch from normal microflora to BV is characterized by a decrease in vaginal colonization by Lactobacillus species together with an increase of Gardnerella vaginalis and other anaerobes. Our goal was to develop and optimize a novel Peptide Nucleic Acid (PNA) Fluorescence in situ Hybridization assay (PNA FISH) for the detection of Lactobacillus spp. and G. vaginalis in mixed samples. Results Therefore, we evaluated and validated two specific PNA probes by using 36 representative Lactobacillus strains, 22 representative G. vaginalis strains and 27 other taxonomically related or pathogenic bacterial strains commonly found in vaginal samples. The probes were also tested at different concentrations of G. vaginalis and Lactobacillus species in vitro, in the presence of a HeLa cell line. Specificity and sensitivity of the PNA probes were found to be 98.0% (95% confidence interval (CI), from 87.8 to 99.9%) and 100% (95% CI, from 88.0 to 100.0%), for Lactobacillus spp.; and 100% (95% CI, from 92.8 to 100%) and 100% (95% CI, from 81.5 to 100.0%) for G. vaginalis. Moreover, the probes were evaluated in mixed samples mimicking women with BV or normal vaginal microflora, demonstrating efficiency and applicability of our PNA FISH. Conclusions This quick method accurately detects Lactobacillus spp. and G. vaginalis species in mixed samples, thus enabling efficient evaluation of the two bacterial groups, most frequently encountered in the vagina.


Background
Bacterial vaginosis (BV) is worldwide the most important perturbation of the normal vaginal condition in women [1]. BV affects pregnant women or women in reproductive age leading to high public health costs and associated complications, such as pelvic inflammatory disease, preterm birth, postoperative infections and an increased risk of acquisition and transmission of sexually transmitted diseases, such as human immunodeficiency virus (HIV) and human papillomavirus (HPV) [1,2]. Several studies have associated this condition with an imbalance in the vaginal microflora although BV etiology is still unclear [3][4][5]. BV has been described as a complex interaction of multiple factors related with several components of the vaginal microbial ecosystem and their human host, although many of these factors remain uncharacterized [2,6]. When BV is established, a decrease in the beneficial bacteria number, specifically Lactobacillus spp., and an increase in the numbers of anaerobic bacteria, such as Gardnerella vaginalis, Atopobium vaginae and Mobiluncus spp., are observed in the vaginal epithelium [7,8]. The disruption of the normal microflora and overgrowth by anaerobes are responsible for the BV signs and symptoms, namely the increase in vaginal pH (pH ≥ 4.5), the formation of vaginal biofilms on vaginal epithelia, observable as clue cells [9], fishy odor and milky vaginal discharge in the absence of an inflammatory response [9,10]. Although BV is often considered a polymicrobial condition, the predominant bacterial species identified is G. vaginalis [2]. In 2005, Swidsinski and colleagues characterized clinical vaginal swabs and found that a multispecies biofilm was formed, which was mainly composed of G. vaginalis and Atopobium vaginae. They hypothesized that G. vaginalis is the initial colonizing species and that its adherence is required before other BVassociated anaerobes are able to interact with the vaginal epithelium [10]. Due to G. vaginalis resistance against Lactobacillus spp. antimicrobial products, such as hydrogen peroxide and lactic acid [11,12], biofilm forming G. vaginalis might compete in initial adhesion against Lactobacillus spp. and may enable other anaerobes to incorporate and grow inside the biofilm [13]. Therefore, the development of an optimized and rapid diagnostic tool that allows the simultaneous visualization of G. vaginalis increase and Lactobacillus species reduction in vaginal samples could be of great use for further study of the previous hypothesis and as a diagnostic tool.
Fluorescence in situ Hybridization (FISH) is a molecular method used to identify and quantify microorganisms in a wide range of samples. This technique combines the simplicity of microscopic observation and the specificity of DNA/rRNA hybridization, allowing detection of selected bacterial species and morphologic visualization [14,15]. Nowadays, Peptide Nucleic Acid (PNA) probes are used instead of natural nucleic acids to improve FISH efficiency [16][17][18][19], because they enable more rapid and more specific hybridization [19][20][21][22][23]. The main goal of our work was to evaluate the PNA-FISH performance on mixed samples using a multiplex approach to detect Lactobacillus spp. and G. vaginalis. To validate the PNA probes, we determined, both in silico and in vitro, their specificity and sensitivity, using a broad diversity of representative Lactobacillus and Gardnerella strains, as well as other taxonomically related or pathogenic bacterial strains commonly found in vaginal samples. To confirm the usefulness of our methodology, the efficiency and specificity of the probes was also tested at different concentrations of Lactobacillus and G. vaginalis strains in the presence of a monolayer of HeLa cells.

Culture of bacterial strains
The bacterial strains used in this study are listed in Table 1

PNA probe design
To identify Gardnerella genus potential oligonucleotidestarget for the probe design, we used the software Primrose [24], coupled with the 16S rRNA databases from the Ribosomal Database Project II (version 10.0; http://rdp.cme. msu.edu/) [25]. Complementarity with a low number of non-target and a high number of target sequences, as well as a higher predicted melting temperature and the absence of self-complementary sequences, were the main criteria for the PNA probe design. The selected sequences were synthesized (Panagene, Daejeon, South Korea) and the oligonucleotides N terminus was attached to an Alexa Fluor 594 molecule via a double 8-amino-3,6-dioxaoctanoic acid (AEEA) linker (PNA Probe: Gard162, Alexa Fluor 594-OO-CAGCATTACCACCCG; HPLC purified > 90%). The Gard162 probe hybridizes between positions 162 and 176 of the G. vaginalis strain 409-05 16S rRNA sequence (RDPII ID: S001872672) and was selected for probe design. For the detection of Lactobacillus spp. a previously developed probe [26], Lac663 was selected. This probe was attached to an Alexa Fluor 488 molecule, also via an AEEA linker (PNA Probe: Lac663, Alexa Fluor 488-OO-ACATGGAGTTCCACT; HPLC purified > 90%).

In silico determination of sensitivity and specificity
Theoretical specificity and sensitivity were calculated according to Almeida et al. [27]. Briefly, the theoretical specificity and sensitivity of both probes were evaluated using updated databases available at the Ribosomal Database Project II (RDP II; http://rdp.cme.msu.edu/) through the Primrose software, and then were confirmed by a BLAST search at the National Centre for Biotechnology Information (http://www.ncbi.nlm.nih.gov/BLAST/; see Table 2). Only target sequences with at least 1200 base pairs and good quality were included. Briefly, theoretical sensitivity was calculated as ts/(Tts)×100, where ts stands for the number of target strains detected by the probe and Tts for the total number of target strains present in the RDP II database (http://rdp.cme.msu.edu/probematch/, last accession date, May 2012). Theoretical specificity was calculated as nts/ (Tnt)×100, where nts stands for the number of non-target strains that did not react with the probe and Tnt for the total of non-target strains examined.

FISH hybridization procedure
Biomass from a single colony of each strain was diluted and homogenised in sterile water, and then 20 μL were

Fusobacteria nucleatum
Clinical isolate -- The PNA Probe (Lac663 and Gar162) efficiencies were tested in triplicate experiments for each strain, with the following hybridization PNA Germany), 0.2% (wt/vol) ficoll (Sigma, Germany), 5 mM disodium EDTA (Sigma, Germany), 0.1% (vol/vol) triton X-100 (Sigma), 50 mM Tris-HCl (at pH 7.5; Sigma, Germany) and 200 nM of the PNA probe. Subsequently, the samples on glass slides were covered with coverslips and incubated in moist chambers at the hybridization temperature under analysis (from 50°C to 72°C) during a range of hybridization times (from 230 to 180 min). Next, the coverslips were removed and a washing step was performed by immersing the slides in a pre-warmed washing solution for 30 min at the same temperature of the hybridization step. This solution consisted of 5 mM Trisbase (Fisher Scientific, United Kingdom), 15 mM NaCl (Sigma, Germany) and 0.1% (vol/vol) triton X-100 (at pH 10; Sigma, Germany). Finally, the glass slides were allowed to air dry. A FISH procedure in suspension was developed and optimized according to the previous work of Almeida and colleagues [27,37] and to the results obtained for the FISH procedure on glass slides described above. Hybridization was performed at 60°C for 90 min and for washing (60°C for 30 min) and a fresh solution was prepared less than 24 h before use. The suspension samples were stored at 4°C in the dark for a maximum of 24 h before microscopic observation/visualization. Both hybridization procedures (in glass slides and in suspension) are able to detect lactobacilli and G. vaginalis strains. While glass slide hybridization is the more commonly used technique in analytical laboratories [27], hybridization in suspension is frequently used to avoid autofluorescence background in complex matrix samples, besides being the hybridization technique used in flow cytometry [27,37].

Microscopic visualization
Prior to microscopy, one drop of non-fluorescent immersion oil (Merck, Germany) was added to either slides or filters and covered with coverslips. Microscopic visualization was performed using an Olympus BX51 (Olympus Portugal SA, Porto, Portugal) epifluorescence microscope equipped with a CCD camera (DP72; Olympus) and filters capable of detecting the two PNA probes (BP 470-490, FT500, LP 516 sensitive to the Alexa Fluor 488 molecule attached to the Lac663 probe and BP 530-550, FT 570, LP 591 sensitive to the Alexa Fluor 594 molecule attached to the Gard162 probe).
Other filters (such as BP 365-370, FT 400, LP 421) present in the microscope, that are not capable of detecting the probe fluorescent signal were used to confirm the absence of autofluorescence. In each experimental assay, a negative control was performed simultaneously in which all the steps described above were carried out, but where no probe was added in the hybridization step. All images were acquired using Olympus CellB software using a total magnification of × 1000.

Experimental assessment of probe specificity and sensitivity
After the hybridization optimization, the specificity and sensitivity of the PNA Lac663 and Gard162 probes were tested using 36 representative strains from the genus Lactobacillus, 22 representative strains from Gardnerella vaginalis (the only species of the genus Gardnerella [4]) and 27 representative strains from other related genera (see Table 1), of which 16 belonged to the order Lactobacillales and the other are common pathogens usually found in  Detection of Lactobacillus spp. and G. vaginalis adhered to HeLa cell line The application of cellular lines is a standard procedure that has already been used to mimic vaginal epithelium at several in vitro studies [41][42][43]. So, HeLa epithelial cells (from American Tissue Culture Collection, ATCC) were cultured at 37°C, in 5% CO 2 (vol/vol), in Dulbecco's modified Eagle's medium (DMEM; Quality Biological, USA) supplemented with 10% FBS (vol/vol) and 1 IU penicillin/ streptomycin ml −1 (MediaTech, Germany). Aliquots of 1ml from HeLa epithelial cells were seeded into 24-well tissue culture plates (Frilabo, Portugal) containing glass slides (12 mm) at a density of 2×10 5 cells per well, and incubated at 37°C and 5% CO 2 (vol/vol) until the formation of a cell monolayer. The cultures were fed with fresh media every 48 hours. Simultaneously, several Lactobacillus (L. crispatus and L. iners) strains and G. vaginalis strain 5-1 were grown in MRS broth and BHI broth as described above. Prior to the adhesion assay, these broth cultures were harvested by centrifugation (4,000 g, 12 min, at room temperature) and washed twice with sterile phosphate buffer saline (PBS). Several standard concentrations of the bacteria were prepared in eukaryotic cell media (DMEM) and the optical density at 600 nm was adjusted using a microplate reader (Tecan, Portugal). When a HeLa cell monolayer was obtained, the cells were washed twice with 500 μl of sterile PBS to remove non adhered cells and culture media. Next, aliquots of 250 μl of cell culture media with a known concentration of a Lactobacillus strain and G. vaginalis 5-1 strain (1×10 3 to 1×10 9 CFU/ ml; see Table 4) were added to each well with the washed cell monolayer from the 24-well tissue culture plate. Then the 24-well tissue culture plate was incubated for 30 min at 37°C in anaerobic conditions and 120 rpm. Finally, each well of the incubated plate was carefully washed twice with 500 μl of sterile PBS to remove non-adherent bacteria. The glass slides containing the adhered bacteria and eukaryotic cells were fixed and hybridized with both PNA probes and observed in fluorescence microscopy, as referred above. An additional 4',6-diamidino-2-phenylindole (DAPI; Sigma, Portugal) staining step was done at the end of the hybridization procedure, covering each of the glass slides with 10 μl of DAPI for 5 min at room temperature in the dark, followed by immediate observation in the fluorescence microscope. All these assays were repeated three times, on separate days, with three fields of view assessed each time.

Results
In silico analysis of PNA probes The Lac663 probe showed a theoretical sensitivity and specificity of 91.5% and 99.7%, respectively, which corroborates the previously reported values [26]. Actually, this publication shows that these probes match the best values of the existing Lactobacillus probes. Gard162 probe presented a theoretical sensitivity of 95.0% and specificity of 100%. The theoretical specificity and sensitivity of these two probes and those developed in other studies were calculated as previously described by Almeida et al. [27] and are listed in Table 2. ProbeMatch tool, from RPDII (http://rdp.cme.msu.edu/probematch/; last accession, May 2012), was used with the following data set options: Strain -Both; Source -Both; Size -> 1200 bp; Quality -Both. For Lactobacillus probes, the specificity and sensitivity values previously determined [26], were considered.

FISH Protocol optimization and autofluorescence-related factors
FISH protocols on slides and in suspension were adapted from previous protocols developed by Almeida et al. [37], due to the crucial importance of fixation and hybridization conditions for an efficient multiplex FISH with different probes. From an initial temperature range of 50 to 72°C and an incubation time range between 30 and 180 min, the best hybridization conditions were set as a moist chamber temperature of 60°C during 90 min of incubation (data not shown). Hybridization conditions started to reveal strong signal-to-noise ratio at 59°C to 61°C from 30 min of incubation up to 120 min, reaching its peak at 60°C during 90 min of incubation. Hybridization conditions above 60°C and 90 min were also efficient, but the signal-to-noise ratio appeared to decrease beyond the selected values of time and temperature. Both hybridization protocols (on slides and in suspension) revealed the same results and pitfalls, as discussed below (some examples are shown in Figure 1).

Experimental determination of probe specificity and sensitivity
As shown in Table 1, the Lac663 probe was able to detect all Lactobacillus strains and cross hybridization was found only for Streptococcus thermophilus B, as it was previously reported [26]. Based on these results, an experimental sensitivity of 100% (95% CI, 88.0 to 100.0%) and specificity of 98.0% (95% CI, 87.8 to 99.9%) were obtained for the Lac663 PNA probe. The Gard162 probe hybridized with all G. vaginalis strains, whereas no hybridization was observed for the other species tested. Therefore, this probe revealed a sensitivity of 100% (95% CI, 81.5 to 100.0%) and a specificity of 100% (95% CI, 92.8 to 100%).
Detection of Lactobacillus spp. and G. vaginalis by Multiplex FISH Once the hybridization procedure was fully optimized, the multiplex methodology was also tested against mixed bacterial cultures (containing Lactobacillus or/ and G. vaginalis cells together with others species, see Table 3) and infected tissue cell line (Table 4). Lac663 and Gard162 probes selectively bound to Lactobacillus and G. vaginalis strains, respectively. The fluorescence signal was easily observable ( Figure 2) and no cross hybridization with other species was detected (see Table 3). Additionally, the multiplex also performed well in the presence of HeLa cells (Table 4) for all the bacterial concentrations evaluated (1×10 3 until 1×10 9 CFU/ml), confirming the in silico analysis of the PNA probes previously elaborated.

Discussion
In silico and in vitro probe specificity and sensibility Fluorescence microscopy has become a widely used technique for direct detection of bacteria in complex samples.
In fact, many authors demonstrated the efficiency of FISH methodology for the analysis of lactobacilli and G. vaginalis [6,10,32,34,[44][45][46][47]. However, the herein described multiplex approach may be the simpler to perform and still has high specificity for lactobacilli and G. vaginalis detection. As shown in Table 1, the Lac663 and Gard162 probes bound highly specific to each target strain. Only Lac663 showed cross-hybridization with S. thermophilus B. However, S. thermophilus coccus morphology allows a clear differentiation from Lactobacillus spp., which has a rodshaped morphology (with the exception of L. iners). Importantly, the Lac663 probe did not hybridize with several bacterial species from the Bacilli class and also with other common vaginal pathogenic bacteria, providing further evidence of its usefulness for Lactobacillus spp. detection in clinical samples. Furthermore, the Gard162 probe showed hybridization with all G. vaginalis strains and no cross-hybridization was observed to other species, including other related pathogenic bacteria which may be present in the vaginal microflora, such as A. vaginae, P. bivia, M. mulieris and F. nucleatum (see Table 1). It is worth to mention that in silico analysis of the Gard162 probe only identified one non-target strain as match, more precisely Bifidobacterium indicum HM534842 (RDPII ID: S002908348). However, B. indicum is not a common bacterium from vaginal microflora, as it is usually present in the gut [48]. Recently a strong association between the bacterial loads in the vagina and rectum of pregnant women was described [49]. Although some gut bacteria such as Escherichia coli [48] have been associated with vaginal infections, B. indicum has not been described as a pathogenic bacterium [50]. The FISH efficiency and hybridization quality for the Gard162 probe, either alone or together with the Lac663 probe, confirmed the applicability of these two probes together in a multiplex PNA-FISH (see Figures 1 and 2).
As shown in Table 2, sensitivity and specificity equations allowed the comparison between our PNA probes and other published ones for G. vaginalis detection. For the Lactobacillus probe, this comparison had already been performed [26] and the Lac663 theoretical performance was found to be similar to other probes reported for Lactobacillus genus detection, but with a highest specificity. Also, Lab158, LGC354 and PNA Burton et al. [31] probes were found to cross-hybridize with one strain (RDPII ID: S000536416) from G. vaginalis, which might be incompatible with a multiplex approach to be used in vaginal samples. On the other hand, it is possible that this G. vaginalis strain was a misidentified L. iners strain, because confusion between both species has been reported [51].
Gard162 theoretical performance in specificity (100 %) was found to be similar to other probes for G. vaginalis detection that have been previously reported ( Table 2). G.vag1008 is the only probe with higher sensitivity (97.5%) than our probe, being able to detect one more G. vaginalis strain. This higher sensitivity is due to the presence of a degenerate oligonucleotide in the sequence of the probe (see Table 2), allowing G.vag1008 to act as two different sequence probes. However, G.vag1008 has 24 oligonucleotides (i.e. 9 nucleotides more than our probe) and it is a DNA probe, which penetrates the cell wall less efficiently [52] and implies need for the use of long hybridization periods.
GardV probe detected species from several bacterial genera present in vaginal samples, such as Alloscardovia, Parascardovia and Scardovia spp. [53]. G.vag1008 probe hybridized with Aeriscardovia spp. that may also be found in vaginal samples [53] and therefore this represents an important pitfall for the G. vaginalis detection with such probes.
It is important to notice that our Gard162 probe is the first PNA probe specifically designed for G. vaginalis detection. Other PNA probes for the detection of lactobacilli [31,46] revealed several disadvantages when compared to Lac663 probe, as we shown before [26].

Multiplex FISH detection
Although numerous authors attempted to correlate differences between healthy and BV vaginal samples [54][55][56][57], no consensus was achieved, except that biofilm formation of G. vaginalis and a decrease in lactobacilli number could be considered as the initial stages in the pathogenesis of BV [10,58]. Swidsinski and colleagues already conducted an international follow-up study in which vaginal samples from several BV patients were analyzed by DNA-based FISH and a dense as well as active bacterial biofilm on vaginal mucosa was detected, primarily consisting of G. vaginalis [47]. Therefore, multiplex FISH to analyze G. vaginalis biofilm establishment and subsequently lactobacilli replacement appeared to be a useful molecular methodology for BV diagnosis in vaginal samples. Although several authors already developed specific probes for G. vaginalis and Lactobacillus spp. detection for FISH, our multiplex method presented new improvements on the method (see Table 2).
Due to the difficulty to obtain fresh vaginal samples diagnosed with BV, we devised an in vitro experiment mimicking the shift from healthy vaginal flora to BV HeLa cells were incubated with different concentrations of G. vaginalis and Lactobacillus strains (L. crispatus and L. iners), ranging from normal to BV vaginal microflora contents (1×10 3 to 1×10 9 CFU/ml; see Table 4). The HeLa cell line is an established tool in experimental research with lactobacilli. It has not only been used to study attachment of several Lactobacillus species, but also of other pathogens [41][42][43]. The Lactobacillus strains used here were selected because high concentrations of L. crispatus (in conjugation with low loads or absence of G. vaginalis) are usually associated to the normal vaginal microflora while high concentrations of L. iners (in conjugation with high loads of G. vaginalis) are commonly associated to the microflora of BV diagnosed women [4,7,51]. The efficiency of our multiplex PNA-FISH methodology was demonstrated by ability of the PNA probes to hybridize in a large range of Lactobacillus spp. and G. vaginalis concentrations, even in the presence of epithelial cells (see Table 4). Swidsinski and colleagues [10,47] used a multiplex FISH methodology to study BV biofilms. A drawback of their approach is that it requires pre-treatment with lysozyme before fixation and the use of urine or paraffin-embedded samples, in opposition of our methodology that do not require a pre-treatment for FISH analysis. These experimental steps increase analysis time and decrease FISH efficiency for Lactobacillus spp. and G. vaginalis strains detection, due to the lower number of cells available for hybridization. Another DNA hybridization test for vaginal infection was studied by Witt and colleagues that evaluated the Affirm VPIII Kit [59], which detected G. vaginalis, Candida spp. and Trichomonas vaginalis in clinical samples, using two distinct single-stranded nucleic acid probes for each organism, which makes the analysis more complex and vulnerable to experimental pitfalls. This validated method showed sensitivity and specificity values for G. vaginalis of 89.5% and 97.1%, respectively, both lower than our Gard162 experimental values (95.0% and 100%, respectively). Furthermore, Fredricks and colleagues developed a FISH methodology for molecular identification of unknown bacteria associated with BV [6], using DNA probes Eub338-Cy5 and G.vag198-Cy3. However, the Eub338 is an unspecific probe used to detect Lactobacillus spp., detecting all species of the order Bacillales, and G.vag198 corresponds to a twenty five oligonucleotide probe with high specificity (100%) but with low sensitivity (85.0%) when compared to our probe (see Table 2). Both these probes worked together at a hybridization temperature of 45°C, which may easily lead to the occurrence of false positive results. Moreover, previous studies have shown that probes with Cy fluorochromes present a lower fluorescence signal than those with the corresponding Alexa Fluor [60].
To conclude, our main purpose was achieved by demonstrating the in vitro applicability of the PNA multiplex methodology for detection of Lactobacillus species and G. vaginalis in the presence of the HeLa epithelial cell line and other taxonomically related or pathogenic bacterial strains commonly found in vaginal samples. These in vitro results confirmed the previous in silico analysis from Lac663 and Gard162 probes.

Conclusions
In summary, the use of the PNA multiplex FISH assay described here significantly increases the specificity and sensitivity of the detection of Lactobacillus spp. and G. vaginalis strains in mixed samples and no interference was observed in the presence of human epithelial cells. As previously referred, there are no consensual agreements regarding BV markers, except for lactobacilli number decrease and initial adherence, and consequent biofilm formation from G. vaginalis. Moreover, our approach allows a fast identification (approximately 3 hours) of the main bacteria involved in BV establishment. Further studies are necessary to detect BV biofilm formation in clinical samples and to characterize possible interactions with other unknown bacteria in the biofilm. The combination of our PNA-FISH methodology with EUB probe or other methodologies, such as electron microscopy, may help to better understand BV etiology.