1Saskatoon Research Centre, Agriculture and Agri-Food Canada, 2Department of Veterinary Microbiology, University of Saskatchewan, 3Plant Biotechnology Institute, National Research Council of Canada
This article is a part ofJoVE Immunology and Infection. If you think this article would be useful for your research, please recommend JoVE to your institution's librarian.Recommend JoVE to Your Librarian
Current Access Through Your IP Address
Current Access Through Your Registered Email Address
Dumonceaux, T. J., Town, J. R., Hill, J. E., Chaban, B. L., Hemmingsen, S. M. Multiplex Detection of Bacteria in Complex Clinical and Environmental Samples using Oligonucleotide-coupled Fluorescent Microspheres. J. Vis. Exp. (56), e3344, doi:10.3791/3344 (2011).
Bacterial vaginosis (BV) is a recurring polymicrobial syndrome that is characterized by a change in the "normal" microbiota from Lactobacillus-dominated to a microbiota dominated by a number of bacterial species, including Gardnerella vaginalis, Atopobium vaginae, and others1-3. This condition is associated with a range of negative health outcomes, including HIV acquisition4, and it can be difficult to manage clinically5. Furthermore, diagnosis of BV has relied on the use of Gram stains of vaginal swab smears that are scored on various numerical criteria6,7. While this diagnostic is simple, inexpensive, and well suited to resource-limited settings, it can suffer from problems related to subjective interpretations and it does not give a detailed profile of the composition of the vaginal microbiota8. Recent deep sequencing efforts have revealed a rich, diverse vaginal microbiota with clear differences between samples taken from individuals that are diagnosed with BV compared to those individuals that are considered normal9,10, which has resulted in the identification of a number of potential targets for molecular diagnosis of BV11,12. These studies have provided a wealth of useful information, but deep sequencing is not yet practical as a diagnostic method in a clinical setting. We have recently described a method for rapidly profiling the vaginal microbiota in a multiplex format using oligonucleotide-coupled fluorescent beads with detection on a Luminex platform13. This method, like current Gram stain-based methods, is rapid and simple but adds the additional advantage of exploiting molecular knowledge arising from sequencing studies in probe design. This method therefore provides a way to profile the major microorganisms that are present in a vaginal swab that can be used to diagnose BV with high specificity and sensitivity compared to Gram stain while providing additional information on species presence and abundance in a semi-quantitative and rapid manner. This multiplex method is expandable well beyond the range of current quantitative PCR assays for particular organisms, which is currently limited to 5 or 6 different assays in a single sample14. Importantly, the method is not limited to the detection of bacteria in vaginal swabs and can be easily adapted to rapidly profile nearly any microbial community of interest. For example, we have recently begun to apply this methodology to the development of diagnostic tools for use in wastewater treatment plants.
This method was used in the research reported in Dumonceaux et al. J. Clin. Microbiol. 47, 4067-4077, doi:10.1128/jcm.00112-09 (2009).
A schematic diagram depicting the overall procedure is presented in Figure 1.
1. Bead coupling
This describes the methods to be used for coupling oligonucleotide probes to polystyrene Luminex beads (see Table 2). Volumes are adapted slightly for evaluation of new capture probes on a trial basis; these volumes are indicated in parentheses.
2. Chaperonin 60 universal target (cpn60 UT) amplicon production and generation of single strands.
3. Hybridization of single-stranded PCR product to oligonucleotide-coupled polystyrene beads.
4. Representative results:
One of our goals in developing and implementing this assay was to make it as simple and streamlined as possible. We therefore optimized the critical post-amplification steps, including T7 exonuclease treatment time and hybridization time in order to develop an assay that can be completed in a reasonable time frame. As shown in Figure 2A, T7 treatment of the amplicon is essential for signal generation, as most probes had little or no signal with short or no T7 treatment. The signal increased linearly until approximately 40 minutes, at which time the signal increase slowed. No degradation of the signal was observed even at 2 hours of T7 treatment time, indicating that the phosphorothioate modification of the primers is highly effective in preventing target strand degradation as described15. We chose 40 minutes for T7 treatment time to minimize the overall protocol time, but it is clear from Figure 2A that T7 treatment can go on for much longer. We also determined the effect of hybridization time on signal generation (Figure 2B) and found that 10 minutes was sufficient for maximum signal, as no further increase in signal was observed even after 4 hours of hybridization. Therefore, a 10 minute hybridization step was chosen, again to minimize the overall assay time. With these results in mind, and with rapid DNA extraction techniques such as InstaGene (Bio-Rad), the overall assay including DNA extraction, PCR, and Luminex analysis can be completed in 4-5 hours.
On a cautionary note, the level of aggregation of the polystyrene beads during a Luminex or BioPlex run can have a major impact on the efficiency of the assay. The BioPlex software will display the level of bead aggregation, which occurs when more than one bead is detected in the laser path and results in the exclusion of the hybridization signal from the aggregate. Apparent bead aggregation can also be caused by any particulate matter that is not the proper size for a single bead, and can even be caused by air bubbles. In any of these events, the signal from the bead aggregate, air bubble, or particle is discarded. In most cases, we find that bead aggregation is minimal (Figure 3A) and the targeted level of 100 counting events for each bead is easily achievable. Occasionally, however, the beads show a moderate (Figure 3B) or severe (Figure 3C) level of aggregation. In these cases, since most of the data is discarded, the instrument may have trouble reaching 100 events per bead type and the results may be questionable. The sonication step (step 3.2) is meant to minimize bead aggregation. In addition, storing the beads as a dilute microsphere master mix (step 1.20) may help. We have noticed that excluding TMAC from the hybridization buffer – adding it to the SA-PE diluent instead (step 3.5) – helps to minimize bead aggregation. Moreover, while we have not tested this, the newer magnetic beads available from Luminex or BioRad are thought to display less of a tendency to aggregate.
The results of the application of a 5-plex Luminex array targeting G. vaginalis, A. vaginae, L. iners, L. crispatus, and L. gasseri along with corresponding Gram stained vaginal swab smears are shown in Figure 4. These samples were taken from a single individual at multiple time points. At time 0, the individual was diagnosed with BV based on Gram stain (Figure 4A) and the Luminex results from this same sample (Figure 4B) show that of the organisms represented in the array, G. vaginalis was most prevalent, while A. vaginae and L. iners were also positive. Nine days later, the individual was still BV positive and the signal for G. vaginalis had increased substantially while A. vaginae and L. iners were still positive. Importantly, after this time the individual began to transition to a normal microbiota as Gram positive bacilli became detectable in the smears (Figure 4A) while the signal for G. vaginalis waned and the signal for L. iners increased (Figure 4B). By our original definition of BV with this method (samples positive for G. vaginalis and/or A. vaginae were considered BV positive)13 , this individual was BV positive at all time points, although the Gram stain failed to detect G. vaginalis in the latter two time points. With the Luminex method described here, trends can be easily compared to sequencing-based methods, and the results of the Luminex assay typically corroborate Gram stains13 while providing additional information on organism identity and abundance.
Figure 1. Schematic diagram of the Luminex protocol for determining the microbiota profile of a complex clinical or environmental sample. The protocol starts with template DNA that has been extracted from the sample of interest. (1) Generate PCR product from template DNA using strand-specific biotinylated, phosphorothioate-modified cpn60 UT PCR primers coupled with unmodified cpn60 UT PCR primers (Table 1); (2) This generates a PCR product pool representing the microbiota with the biotin-phosphorothioate modification on one strand; (3) Digest the double-stranded PCR product with T7 exonuclease, which cannot degrade the phosphorothioate-modified strand and therefore generates single-stranded DNA that is modified at the 5' end with biotin; (4) Couple polystyrene beads to species-specific cpn60 UT probes – each bead has a unique spectral address (indicated by bead color) that is discernable by the Luminex or Bio-Plex instrument; (5) Hybridize the single-stranded PCR product generated from the sample of interest to the suite of species-specific oligonucleotide-coupled beads; (6) Add streptavidin-phycoerythrin conjugate that binds to the biotinylated single-stranded PCR product and acts as an indicator of hybridization; (7) Determine the spectral address (bead identity) and hybridization signal intensity using a Luminex or Bio-Plex instrument. At least 100 beads are counted for each bead identity and the median fluorescence intensity (MFI) of the phycoerythrin signal is reported as the output. Up to 100 different bead types can be evaluated simultaneously but an assay showing the discrimination of three bead types is illustrated. (8) When the MFI of PCR replicates generated from the same sample is significantly greater than the negative control for a given bead (one-tailed student's t-test, p<0.05), the sample is considered to be positive for that organism.
Figure 2. Optimization of Luminex assay parameters. Five different probes targeted to Peptostreptococcus anaerobius were used with amplicons generated from a mixed template comprising plasmids containing the cloned cpn60 UT of 20 bacteria known to be associated with the vagina, including P. anaerobius. (A). Effect of T7 exonuclease treatment time. The protocol described above was followed but the time of treatment with T7 exonuclease was varied prior to hybridization and the median fluorescence intensity (MFI) was determined for all probes. (B). Effect of hybridization time. The protocol described above was followed with a variety of hybridization times. The same sets of probes were used with an amplicon generated from the same template as in A and the median fluorescence intensity (MFI) was determined for all probes.
Figure 3. Determination of bead aggregation using BioPlex software. Three examples are given of BioPlex runs in which the level of bead aggregation is (A) acceptable (5%); (B) borderline (50%); and (C) unacceptable (80%).
Figure 4. Application of a 5-plex Luminex array to BV diagnosis in sequential samples taken from the same individual. (A). Gram-stained slides showing the traditional diagnostic used to assess each sample for BV. (B). Application of a 5-plex Luminex array prepared and executed as described in this protocol to the same four samples shown in (A). Bacterial targets whose MFI signal was significantly positive by our definition (one-tailed student's t-test, p<0.05) are indicated by an asterisk (*).
The specificity of signal generation is of critical importance; you must be confident that the signal observed truly reflects the detection of amplicon generated from that organism. Software such as PrimerPlex (Premier Biosoft) can help to design probes that will hybridize efficiently, but they may or may not cross-hybridize to non-target species. When universal PCR primers are used as described in this protocol, it is important to keep in mind that the amplicon generated represents all of the organisms present in the sample. It is important, therefore, to analyze potential probes for the possibility of cross-hybridization by comparing the sequences suggested by the probe design software to the cpnDB, a database of cpn60 UT sequences16; probes that match multiple organisms are to be avoided if possible. Furthermore, when potentially specific probes are identified, we have taken an approach to probe validation in which a complex, artificial template is prepared corresponding to the cloned cpn60 UT of a number of organisms associated with the microbiota of interest13. Amplicon is then generated from the complete "mixed panel", including the target template, and the signal is compared to that generated from a second template comprised of the mixed panel that does not contain the target organism's cpn60 UT. Probes that generate a high signal-to-noise ratio (where the signal is generated from the complete panel and the noise is generated from the panel without the target template) are selected for inclusion in the final assay. We have typically analyzed 4-5 possible probes for each target organism before selecting one for inclusion in the final Luminex array.
The definition of a positive result is an issue to be considered. We have found that different probes have inherently different signal-to-noise ratios, so we decided on an approach whereby a positive result is defined by a very simple statistical test: the student's t-test (one-tailed). A positive result is defined by comparing the MFI of the sample to the "no template" control, and when the MFI is significantly higher than the control at p < 0.05, the result is considered to be positive for that organism. Normally two independently generated amplicons are each tested in duplicate. Although the PCR and T7 exonuclease steps create about 64.5 μl of single-stranded PCR product (step 2.3) and only 17 μl of this is used for hybridization (step 3.2), sufficient for 3 hybridization assays, we normally analyze duplicates from two amplifications to give an inherent technical replicate without using an excessive number of wells.
An aspect of this assay that we have investigated is its semi-quantitative nature. The MFI signal from a given probe correlates well to the qPCR-determined abundance of a given organism (Spearman rank correlation coefficient ρ = 0.4686, p < 0.0001)13, but the dynamic range is limited; the signal tends to saturate at higher target organism levels, as expected since amplicon is analyzed after end-point PCR of 40 cycles. Because it is semi-quantitative, trends in organism abundance can be monitored and patterns observed (keeping the caveat in mind regarding signal saturation at higher abundances). For example, Figure 4 shows a profile of the vaginal microbiota of an individual over time as it changes from BV to normal. As this occurs, the signal corresponding to G. vaginalis (a BV-associated organism17) increases and then wanes, while over the course of the monitoring time the signal for L. iners increases. Corresponding Gram stains, the current "gold standard" for BV diagnosis, show a pattern whereby Gram positive bacilli are initially absent but are detectable at later time points. The Luminex method both identifies the organisms present in the sample and gives some information as to their relative abundances. This type of information could potentially be used to monitor samples of any type for undesirable trends in the abundances of particular organisms and can provide this data for a large number of organisms simultaneously.
Importantly, this method is not limited to the profiling of vaginal microbiota and can reasonably be applied to any environment in which the detection and partial quantification of a range of target organisms is desirable. When applying the method to other environments, it is important to consider the possibility of PCR inhibition occurring, as inhibitors of PCR commonly co-purify with DNA from many environments18. Where PCR inhibition is an issue, the sample can either be diluted or further purified in order to generate sufficient signal for this assay.
No conflicts of interest declared.
We thank Alberto Severini and Vanessa Goleski for help with assay development and critical comments on this manuscript. This work was funded by the Public Health Agency of Canada and the Industrial Research Assistance Program (National Research Council of Canada). Additional support was obtained from the University of Saskatchewan Publication Fund.