Statistical Analyses of Complex Denaturing Gradient Gel Electrophoresis Profiles
http://www.100md.com
微生物临床杂志 2005年第8期
Division of Microbial Diseases
Biostatistics Unit, Eastman Dental Institute, UCL, 256 Gray's Inn Road, London WC1X 8LD, United Kingdom
ABSTRACT
Studies using molecular techniques have demonstrated that a culture-based approach can severely underestimate the bacterial diversity in most environments. One of the molecular techniques that has been applied in microbial ecology is denaturing gradient gel electrophoresis (DGGE). The purpose of this study was to investigate differences in the microbiota of plaque, using a number of analysis techniques, from children without gingivitis (n = 30) and from those with gingivitis (n = 30). Extracted DNA from gingival margin plaque was subjected to PCR targeting the 16S rRNA gene using universal primers. DGGE profiles were analyzed in three ways. (i) Bacterial diversity was compared between cohorts by using the Shannon-Wiener index (also known as the Shannon-Weaver index). (ii) A hierarchical cluster analysis of the banding patterns was calculated and expressed as a dendrogram. (iii) Individual DGGE bands and their intensities for both cohorts were compared using a logistic regression analysis. The Shannon-Wiener indices demonstrated a greater bacterial diversity associated with no-gingivitis plaque (P = 0.009). Dendrograms demonstrated that seven clades associated with gingivitis and five clades associated with no gingivitis. The logistic regression demonstrated that one band was significantly associated with no gingivitis (P = 0.001), while two bands were significantly associated with gingivitis (P = 0.005 and P = 0.042). In conclusion, this study demonstrates that the development of gingivitis might be accompanied by a decrease in bacterial diversity. Furthermore, we have demonstrated that logistic regression is a good statistical method for analyzing and characterizing DGGE profiles.
INTRODUCTION
The accumulation of dental plaque as a result of poor or impaired oral hygiene may result in gingivitis. This inflammation occurs primarily in response to bacteria present in dental plaque (8, 35, 42). Gingivitis is a reversible condition that can precede periodontitis, although not all cases of gingivitis will progress to periodontitis (9). A number of studies of the microbiota associated with gingivitis have been carried out using culture-based techniques (28, 40). Studies employing molecular techniques have elucidated that culture-based approaches may severely underestimate the bacterial diversity in most environments (1). The use of rRNA gene sequences for the analysis of microbial communities has been widely used in microbial ecology and is replacing the more-conventional cultural methods (36). Various molecular approaches for the analysis and comparison of the microbiota of dental plaque in healthy and diseased individuals have been used. One such approach involves the use of DNA probes and checkerboard DNA-DNA hybridization assays (43) that have been previously used to compare the microbiota of supra- and subgingival plaque in healthy individuals and in individuals with periodontitis (57). This approach is hampered by the use of 40 species-specific DNA-DNA hybridization probes, as it only partially considers the microbial diversity in dental plaque, given that approximately 500 species live in the oral cavity (36). Furthermore, most conventional methods used for analyzing bacterial communities do not take the unculturable proportion into consideration (1). This may be a large oversight, as unculturable bacteria are believed to constitute up to 50% of dental plaque communities (46). Earlier workers (23) have demonstrated that a greater microbial diversity could be ascertained by amplifying rRNA genes directly from subgingival plaque than by cultivating bacteria from the same specimen. Currently the best model for exploring microbial diversity involves PCR cloning of the 16S rRNA gene (4, 44). Cloning and sequencing of multiple plaque samples are extremely laborious and expensive tasks to perform thoroughly due to the huge diversity involved (36). For this purpose, genetic fingerprinting techniques may be better suited.
Genetic fingerprinting techniques provide a rapid and relatively easy alternative to the analysis of microbial communities. The different types of genetic fingerprinting techniques that can be used for this purpose, together with their advantages and disadvantages, have already been described (31). One such fingerprinting technique is the electrophoretic separation of low-molecular-weight rRNA molecules (5S rRNA and tRNA) extracted from natural samples (19). Denaturing gradient gel electrophoresis (DGGE) is a more recent fingerprinting technique by which PCR-amplified DNA fragments are separated according to their sequence information (33). The basis of this technique is that DNA fragments of the same size but with differing base pair sequences can be separated (32). This separation by DGGE relies on the electrophoretic mobility of partially denatured DNA molecules in a polyacrylamide gel, which is encumbered in comparison to the completely helical form of the molecule (32). DGGE has been applied in environmental microbiology (6, 11, 49) and food microbiology (2, 5) and in the analysis of microbial communities in the human body (10, 14, 53, 54). Recently DGGE has also been applied to analyze the bacterial diversity of human subgingival plaque (17, 58) as well as laboratory-grown dental plaque microcosms (27). This paper presents a fingerprint analysis of dental plaque sampled from the gingival crevice of prepubertal children with and without gingivitis. Any marked differences observed in the microbial composition of plaque associated with healthy dental sites and gingivitis would further advance our understanding of how specific bacteria are associated with, and respond to, inflammation and indeed the progression to periodontal disease.
Analysis of DGGE profiles often involves the use of principal component analysis (34, 56) and multidimensional scaling (7, 21) but generally utilizes hierarchical cluster analysis to demonstrate similarities in the data, and the results are invariably presented in the form of dendrograms (6, 7, 13, 51). This study analyzes the DGGE profiles of the dental plaque sampled from both cohorts by using hierarchical cluster analyses as well as measuring community diversity by means of the Shannon-Wiener index (also known as the Shannon-Weaver index). Furthermore, we have applied a statistical test for the analysis of the electrophoretic fingerprints that appears to be used infrequently. This involves a logistic regression analysis that looks at single bands and computes a regression relationship between the presence and absence of gingivitis and band presence and intensity (independent or explanatory variable). This type of analysis can pinpoint individual bands that when present/absent will affect the outcome of interest, in this case the outcome being whether the subject had gingivitis or not.
MATERIALS AND METHODS
Patients. Ethical approval was obtained from the Eastman Dental Hospital NHS Trust, London, United Kingdom. Each parent was given an information sheet to read and asked for written consent. Each child was asked for verbal consent.
Subjects. Healthy children aged between 5 and 9 years attending the Department of Pediatric Dentistry and the School of Dental Therapy at the Eastman Dental Hospital were recruited. Children with chronic medical disorders or who had been treated with antibiotics within the preceding 3 months were excluded. Two groups of children (a total of 60) were selected: group 1, 30 children with discernible plaque but no gingivitis associated with either the lower left or lower right first permanent molar tooth; group 2, 30 children with discernible plaque and gingivitis associated with either the lower left or lower right first permanent molar tooth. Plaque and gingivitis scores were recorded for each child using a modified version of the method of O'Leary (15). Four gingivally related quadrisections of each tooth (mesiobuccal, distobuccal, mesiolingual, and distolingual) were visually examined, and each was given a score of 0 for no discernible plaque or a score of 1 for discernible dental plaque deposits. Similarly, each tooth quadrisection associated with no gingival inflammation was given a score of 0, and each quadrisection with gingival inflammation was given a score of 1. Plaque was sampled from the buccal and lingual gingival margin of either the lower left or lower right first permanent molar tooth using a sterile wooden toothpick (47, 55). When both lower permanent first molars had erupted, one tooth (either the right or left tooth) was randomly selected. The toothpick was immediately placed in a sterile container with 1 ml of reduced transport fluid (45) and five glass beads. The plaque samples were dispersed in the reduced transport fluid by vortexing for 10 s, and DNA was immediately extracted using the Puregene DNA isolation kit for yeast and gram-positive bacteria (Gentra Systems, Minneapolis, MN). The DNA was then stored at –20°C.
DGGE profiling. DNA from plaque sampled from 60 different patients were used as templates (group 1, 30 patients with discernible plaque and no gingivitis; group 2, 30 patients with discernible plaque and gingivitis). The V2-V3 region of the 16S rRNA gene corresponding to positions 339 to 539 of Escherichia coli was partially amplified using PCR primer (Genosys, Cambridgeshire, United Kingdom) F357GC (5'-CGCCCGCCGCGCCCCGCGCCCGGCCCGCCGCCCCCGCCCCCCTACGGGAGGCAGCAG-3'), which contains a GC-rich clamp, and PCR primer R518 (5'-ATTACCGCGGCTGCTGG-3'). The final volume of each PCR mixture was 100 μl. The amplification reaction mixture contained 10x PCR buffer, 2.5 mM MgCl2, 0.2 μM of both primers F357GC and R518 and 5 U of Taq polymerase (Bioloine, London, United Kingdom). All deoxynucleoside triphosphates (Promega, Southampton, United Kingdom) were used at a final concentration of 0.2 mM. The cycling parameters for touchdown PCR were carried out on a Primus thermal cycler (MWG Biotech, Milton Keynes, United Kingdom) and were as follows. After preincubation at 94°C for 5 min, 30 cycles were performed at 94°C for 1 min, the annealing temperature (TA) for 1 min, and 72°C for 1 min. The TA decreased stepwise by 1°C every two cycles from 65°C in the first cycle to 56°C in the 20th cycle. The TA for the last 10 cycles was 55°C. Cycling was followed by 5-min incubation at 72°C (60).
The 16S rRNA gene amplicons were cleaned and concentrated into a final volume of 30 μl by using a QIAquick PCR purification kit (QIAGEN Ltd., Crawley, United Kingdom). In addition, two positive controls consisting of genomic DNA extracted from Tannerella forsythensis ATCC 43037 (formerly Bacteroides forsythus) and Prevotella intermedia (wild type) as well as a negative control consisting of sterile deionized water. The two positive controls were used to align multiple gel images (Fig. 1).
Parallel gels containing 10% (wt/vol) polyacrylamide (37.5:1 acrylamide:bisacrylamide) were cast using a D-code system (Bio-Rad Laboratories, Inc., Hercules, CA). The gels contained a linear gradient of the denaturants urea and formamide, increasing from 40% at the top of the gel to 80% at the bottom (with 100% denaturants corresponding to 7 M urea and 40% [vol/vol] deionized formamide). Gels were run at 35 V for 21 h (735 V hours) at a constant temperature of 60°C in 7 liters of 1x Tris-acetate-EDTA (TAE) buffer. Gels were stained for 1 h in 1x TAE containing SYBR green nucleic acid gel stain (10–4 dilution) (Molecular Probes, PoortGebouw, The Netherlands) and photographed under a UV light transilluminator (AlphaImager, San Leandro, CA).
Statistical analysis of DGGE banding patterns. All gels were aligned using the two positive controls. The DNA bands that migrated to the same position within each gel were ascribed a number. Band strengths were estimated visually; weak bands were assigned a value of 1, intermediate bands a value of 2, and strong bands a value of 3 (Fig. 1). Thus, if the odds ratio were unity, the odds of a subject having gingivitis did not differ from the odds of a subject having no gingivitis, as the strength of the band increased to unity. If the odds ratio for a band were greater than 1, the odds of having gingivitis were greater than the odds of not having gingivitis as the intensity of the band increased. In the same way, if the estimated odds ratio for a band was less than 1, then the odds of not having gingivitis was greater than that of having gingivitis as the band intensity increased. These data were then analyzed using the following three methods.
(i) Shannon-Wiener diversity index. The Shannon-Wiener index of diversity (H') (7, 12, 22) was used to determine the diversity of taxa present in plaque sampled from children with and without gingivitis. This index was calculated by the following equation:
where s is the number of species in the sample and Pi is the proportion of species i in the sample. Gaussian distribution tests were applied to the Shannon-Wiener diversity index values for the 30 subjects in both plaque groups. Generally, the data were nonuniformly distributed; therefore, a nonparametric analysis using a Mann-Whitney U test was performed where a P of <0.05 was interpreted to be statistically significant. The nonparametric statistical analysis was performed using SPSS version 12 (Chicago, Ill.).
(ii) Cluster analysis. Similarities between the banding patterns generated by PCR-DGGE of the plaque samples were analyzed using the Pearson correlation coefficient (41). Similarities were displayed graphically as a dendrogram. The clustering algorithms used to calculate the dendrograms was an unweighted pair group method with arithmetic averages (UPGMA) (16, 39). The cluster analysis and dendrogram generation were carried out by using the Phoretix 1D Advanced and Phoretix 1D Database software (Phoretix International, Newcastle upon Tyne, United Kingdom).
(iii) Logistic regression. The data were first analyzed using univariable chi-square tests with a cutoff for significance of 5% to reduce the number of bands included in the multivariate analysis. The bands demonstrating significance were then included as explanatory variables (each band was recorded as present [coded as 1] or absent [coded as 0]) in a multivariate logistic regression analysis, for which the codes, no gingivitis (coded as 0) and gingivitis (coded as1) were used for the dependent variable. The estimated odds ratio with a 95% confidence interval was calculated for each band in the logistic regression model. The predicted probabilities for a subject belonging to a particular group (in this case, gingivitis and no gingivitis) were also calculated. Furthermore, a Hosmer and Lemeshow test for goodness of fit of the logistic model was performed. The logistic regression analysis was conducted using SPSS version 12.
RESULTS
Shannon-Wiener diversity index. The Shannon-Wiener indices for subjects with and without gingivitis were compared using the Mann-Whitney U test. The results revealed that there was a significantly greater biological diversity in the sample group with no gingivitis (for the no-gingivitis group, the median value was 3.4 and the values ranged from 2.5 to 4.3; for the gingivitis group, the median value was 3.2 and the values ranged from 1.6 to 4.1; P = 0.009).
Cluster analysis. The cluster analysis generated an UPGMA dendrogram (Fig. 2) which demonstrates that seven clades associated only with the gingivitis group and five clades associated with the no-gingivitis group.
Hosmer and Lemeshow test. The Hosmer and Lemeshow test for goodness of fit showed that the model was a good fit (chi-square = 2.903, degrees of freedom = 8, P = 0.940). The expected probability of a subject having gingivitis or not having gingivitis, as determined by the logistic regression model, is represented as both a classification table (Table 1) and a classification plot (Fig. 3). The classification table shows that the model works well and approximately 84% of the individuals were correctly classified as having the appropriate disease status. The classification plot further demonstrates that the model works well, in that most subjects with no gingivitis (coded as 0) are predicted as not having gingivitis (i.e., on the left-hand side of the horizontal axis). Likewise, most of the subjects with gingivitis (coded as 1) are predicted as having gingivitis (i.e., on the right-hand side of the horizontal axis). Logistic regression analysis also determined that three bands were significantly associated with the presence or absence of gingivitis in these subjects. From this regression analysis, the estimated odds ratio of any band could be interpreted as the ratio of the odds of having gingivitis in those subjects with the band to the odds of having gingivitis in those subjects without the band. Table 2 presents the results of the analysis and demonstrates that band 2 was significantly associated with no gingivitis (P = 0.001), while bands 1 and 3 were significantly associated with gingivitis (P = 0.042 and P = 0.005, respectively). The positions of these three bands on a DGG are shown in Fig. 4.
DISCUSSION
One of the major challenges in microbial ecology is the assessment of the bacterial diversity or community structure present in a defined environment. It is important to understand the changes in diversity and structure during the inception and progression of polymicrobial diseases. Recent advances in molecular biology techniques have allowed us to better characterize and understand these changes. While the molecular biology techniques are now becoming available, the associated data analysis techniques have lagged behind somewhat. A case in point is the present study, which investigates plaque sampled from the gingival crevice of children both with and without gingivitis in order to detect any differences between the two groups. Such a study could potentially be used to understand the differences between a healthy dental microbiota and its progression to periodontal disease. We chose to use a culture-independent technique (DGGE) to characterize the population differences between the two cohorts. This approach would take into account the uncultivable portion of the oral microbiota in the analysis. The advantages of using DGGE as a means of studying microbial ecology are well established in that it provides a more accurate means of visualizing whole microbial communities (since there should be a reduced bias in the detection of unculturable species). Additionally, it is less labor-intensive than the more-conventional culture-independent techniques of PCR cloning and sequencing.
Once the data were generated for all the subjects in both cohorts, we used a number of different analysis techniques to interrogate these data. In the first instance, this included a "by-eye" analysis as is common in the literature (27, 48). It is relatively easy to comment on three or four patterns and select bands of interest on the basis of visual analysis. However, the banding patterns generated were quite complex, and comparing 60 patterns proved difficult. In the first instance, we analyzed the data using a similar approach commonly used by other workers, e.g., diversity indices and cluster analysis.
There are a variety of ecological diversity measures, but their suitability for use with highly diverse microbial communities is unclear (18). The Shannon-Wiener index of diversity (38) has been used previously in microbial ecology (12, 18, 30, 34). There is some uncertainty surrounding this index, particularly as it is referred to by different names. The full title of the index is the Shannon-Wiener function, after Claude Shannon and Norbert Wiener (38). It is more commonly known as the Shannon-Weaver index after Shannon's coauthor, Wallace Weaver. In the present study, the Shannon-Wiener index of diversity was applied to 16S rRNA genes from dental plaque communities that were separated on denaturing gels according to their sequence heterogeneity. Using both the total number of DGGE bands and their relative intensities, it was possible to calculate the bacterial diversity index (H'). This would in theory reflect on sample diversity without the need for cultivation. It is a fair assumption to expect that certain target rRNA genes within the plaque samples would be present in lower concentrations and thus would not be amplified sufficiently to be visualized as bands. Therefore, the bands generated in the present study would reflect the most abundant rRNA gene types from dental plaque or perhaps those strains most applicable to amplification (37). Thus, the diversity indices calculated from the DGGE pattern were regarded as relative. Previous workers (13) have demonstrated that the diversity index H' can be applied to complex microbial communities and is well suited for comparing large sets of samples.
The results demonstrated that there was a greater biological diversity in the sample group with no gingivitis than the gingivitis group (P = 0.009). This might indicate that a decrease in bacterial diversity may be associated with the shift from health to gingivitis. According to the ecological plaque hypothesis, a change in the environment leads to a shift in the community structure; however, this shift is only a change in the proportions of certain taxa within the community (25, 26). Therefore, perhaps the perceived decrease in diversity is a result of certain taxa proliferating as a consequence of inflammation. As a further consequence, this may "mask" the presence of bacteria that do not necessarily increase in number as a consequence of inflammation. Nevertheless, when considering ecological diversity and community structure, it is believed that species diversity is an important feature in maintaining a degree of stability within that community. Thus, it is expected that stable and resilient microbial communities must contain a certain level of diversity (3). As has been shown, the present results demonstrate a greater diversity in the DGGE profiles of dental plaque associated with no gingivitis. This may indicate that the dental plaque community associated with no gingivitis is probably more stable than the dental plaque community associated with gingivitis. While the overall precepts of ecology support this (3), workers using culture-dependent techniques to study experimental gingivitis have come to the opposite conclusion (29). There are no studies that are experimentally comparable to this present study; therefore, a detailed comparison cannot be made. However, a factor that may go some way to reconciling these apparently conflicting results is that the DGGE technique takes into account the uncultivable microbiota present in the sample (in this case, the oral cavity where it is thought that around 50% of the species are uncultivable).
The use of diversity indices for DGGE analysis in this case uses very little of the data present and as stated provides only a measure of diversity. Therefore, it is entirely possible for two communities to be equally diverse (i.e., equal H' values) but have, in the most extreme case, a completely different species structure (3) (Fig. 5). We believe that the DGGE data contain more information than simply this.
Characteristic profiles of microbial communities or DNA fingerprints can be produced by DGGE. These microbial fingerprints can be represented as binary vectors, which have the potential to be of a very high dimension, depending on the environment sampled, such as soil (50) or dental plaque (36). Earlier workers have used principal component analysis (34, 56) as well as multidimensional scaling (7, 21) to analyze groups of similarities in the data. In general, most researchers use hierarchical clustering algorithms based on similarity indices for binary vectors (6, 7, 13, 51). Cluster analysis, also known as "classification," has been defined as the search for a natural grouping (24). Cluster analysis, for example, the UPGMA applied in this study, can be used to identify samples that generate similar patterns (7, 20). The UPGMA dendrogram calculated demonstrated similarities in the banding profiles of plaque sampled from subjects both with and without gingivitis. There were seven clades that clustered similar band profiles from plaque sampled from gingivitis sites. Likewise, the cluster analysis demonstrated five clades of similar band profiles from plaque sampled from sites without gingivitis. Indeed, 14/30 "healthy" samples and 20/40 "gingivitis" samples clustered together using this method. This implies that specific samples whether derived from healthy subjects or subjects with gingivitis are related to each other based on the DGGE banding pattern. It also shows that more than one profile is associated with either condition and therefore that perhaps there are a number of specific microbiotas associated with health or gingivitis. Indeed, perhaps only one of the gingivitis clades will eventually lead to periodontal disease. Earlier workers have postulated the association of certain taxa with the onset and progression of periodontal disease, for example, the red complex (Porphyromonas gingivalis, Tannerella forsythensis, and Treponema denticola) suggested by Ximenez-Fyvie et al. (57). Perhaps these three key taxa are represented in only one of the gingivitis clades, and therefore, only one type of gingivitis (one bacterial community) actually leads to periodontitis. It would be interesting to follow these subjects, monitor their profile, and record which clades lead to the inception of periodontal disease. However, this analysis technique is limited by the fact that it does not take into consideration the outcome, either gingivitis or no gingivitis. It aims to look at clusters of bands to predict the outcome but does not utilize the outcome itself when comparing the data for similarities. Earlier workers (52, 59) applying similar techniques have demonstrated that individuals have their own unique fecal microbial community that remains relatively stable over time. We believe that this is also the case for the oral cavity (unpublished data). Analysis of the differences in microbial fingerprints for both gingivitis and no gingivitis by cluster analysis may be hampered by the fact that substantial differences may exist within the individual cohorts to start with.
The hierarchical cluster analysis is a good technique for longitudinal data, but with cross-sectional data, there seems to be too much variation within each group.
Both analyses have yielded valuable information, but we believe that further information can be obtained from the DGGE profiles. For example, two questions that we think can be answered by using DGGE profiles are "is there a recognizable fingerprint which can be used as a predictor to differentiate between disease states" and "are there certain bands associated with health or indeed gingivitis" A few workers have used principal component analysis and multidimensional scaling to analyze data such as these data. Furthermore, these methods may not always be appropriate, and indeed, they are very complex techniques both to understand and use. We therefore applied a different approach for data analysis, namely, logistic regression.
Unlike hierarchical cluster analysis, logistic regression analysis takes the outcome into consideration (i.e., in this case, gingivitis or no gingivitis) in addition to differences in band numbers and band migration position. This type of analysis does not take into account clusters of multiple bands in an attempt to examine DGGE profiles for similarities. Indeed, it analyzes the fingerprint with respect to the presence/absence of individual bands. Analysis of individual bands from the total fingerprint allows for the computation of a regression relationship between a clinical descriptor, such as health or gingivitis (outcome variable) and band presence/absence (explanatory variable). The results of this regression analysis (Fig. 3) have demonstrated that there were substantial differences in the DGGE profiles of both cohorts of subjects. The presence or absence of gingivitis was correctly classified in 84% of the subjects. This classification plot correctly predicts most subjects with no gingivitis (coded as 0) as not having gingivitis and most subjects with gingivitis (coded as 1) as having gingivitis on the basis of the presence/absence of bands. Moreover, the results from the regression analysis have further demonstrated that a single band (band 2) is statistically associated with the absence of gingivitis, while bands 1 and 3 were significantly associated with gingivitis.
In conclusion, we have used a number of different techniques to analyze DGGE banding profiles, each of which answer different and progressively more complex questions. The two analyses used first, i.e., the diversity index and the cluster analysis, both provide useful data but do not utilize all available information. The diversity index simply measures a change in diversity but does not specify what changes or even whether the community structure is similar. Cluster analysis, on the other hand, attempts to differentiate between cohorts that contain large differences within the cohort itself. Furthermore, cluster analysis does not specify which operational taxonomic units are important, only that differences exist. Logistic regression, however, not only successfully differentiates between the profiles of both cohorts but can also specify individual operational taxonomic units associated with these differences. Therefore, we believe that a logistic regression analysis in conjunction with a diversity measure is the method of choice to analyze and compare DGGE-generated community profiles in a cross-sectional study.
ACKNOWLEDGMENTS
We extend our thanks to Derren Ready for initial bacterial diversity analysis. We also thank Charlotte Davies for helping with the initial DGGE setup.
REFERENCES
Amann, R. I., W. Ludwig, and K. H. Schleifer. 1995. Phylogenetic identification and in situ detection of individual microbial cells without cultivation. Microbiol. Rev. 59:143-169.
Ampe, F., A. Sirvent, and N. Zakhia. 2001. Dynamics of the microbial community responsible for traditional sour cassava starch fermentation studied by denaturing gradient gel electrophoresis and quantitative rRNA hybridization. Int. J. Food Microbiol. 65:45-54.
Atlas, R. M. 1994. Diversity of microbial communities. Adv. Microb. Ecol. 7:1-47.
Becker, M. R., B. J. Paster, E. J. Leys, M. L. Moeschberger, S. G. Kenyon, J. L. Galvin, S. K. Boches, F. E. Dewhirst, and A. L. Griffen. 2002. Molecular analysis of bacterial species associated with childhood caries. J. Clin. Microbiol. 40:1001-1009.
ben Omar, N., and F. Ampe. 2000. Microbial community dynamics during production of the Mexican fermented maize dough pozol. Appl. Environ. Microbiol. 66:3664-3673.
Boon, N., C. Marle, E. M. Top, and W. Verstraete. 2000. Comparison of the spatial homogeneity of physico-chemical parameters and bacterial 16S rRNA genes in sediment samples from a dumping site for dredging sludge. Appl. Microbiol. Biotechnol. 53:742-747.
Boon, N., W. De Windt, W. Verstraete, and E. M. Top. 2002. Evaluation of nested PCR-DGGE (denaturing gradient gel electrophoresis) with group-specific 16S rRNA primers for the analysis of bacterial communities from different wastewater treatment plants. FEMS Microbiol. Ecol. 39:101-112.
Breuer, M. M., and R. S. Cosgrove. 1989. The relationship between gingivitis and plaque levels. J. Periodontol. 60:172-175.
Brown, L. J., and H. Loe. 1993. Prevalence, extent, severity and progression of periodontal disease. Periodontology 2000 2:57-71.
Donskey, C. J., A. M. Hujer, S. M. Das, N. J. Pultz, R. A. Bonomo, and L. B. Rice. 2003. Use of denaturing gradient gel electrophoresis for analysis of the stool microbiota of hospitalized patients. J. Microbiol. Methods 54:249-256.
Ebie, Y., M. Matsumura, N. Noda, S. Tsuneda, A. Hirata, and Y. Inamori. 2002. Community analysis of nitrifying bacteria in an advanced and compact Gappei-Johkasou by FISH and PCR-DGGE. Water Sci. Technol. 46:105-111.
Edwards, M. L., A. K. Lilley, T. H. Timms-Wilson, I. P. Thompson, and I. Cooper. 2001. Characterisation of the culturable heterotrophic bacterial community in a small eutrophic lake (Priest Pot). FEMS Microbiol. Ecol. 35:295-304.
Eichner, C. A., R. W. Erb, K. N. Timmis, and I. Wagner-Dobler. 1999. Thermal gradient gel electrophoresis analysis of bioprotection from pollutant shocks in the activated sludge microbial community. Appl. Environ. Microbiol. 65:102-109.
Favier, C. F., E. E. Vaughan, W. M. De Vos, and A. D. L. Akkermans. 2002. Molecular monitoring of succession of bacterial communities in human neonates. Appl. Environ. Microbiol. 68:219-226.
Franco, E., C. P. Saunders, G. J. Roberts, and A. Suwanprasit. 1996. Dental disease, caries related microflora and salivary IgA of children with severe congenital cardiac disease: an epidemiological and oral microbial survey. Pediatr. Dent. 18:228-235.
Fromin, N., J. Hamelin, S. Tarnawski, D. Roesti, K. Jourdain-Miserez, N. Forestier, S. Teyssier-Cuvelle, F. Gillet, M. Aragno, and P. Rossi. 2002. Statistical analysis of denaturing gel electrophoresis (DGE) fingerprinting patterns. Environ. Microbiol. 4:634-643.
Fujimoto, C., H. Maeda, S. Kokeguchi, S. Takashiba, F. Nishimura, H. Arai, K. Fukui, and Y. Murayama. 2003. Application of denaturing gradient gel electrophoresis (DGGE) to the analysis of microbial communities of subgingival plaque. J. Periodontal Res. 38:440-445.
Hill, T. C. J., K. A. Walsh, J. A. Harris, and B. F. Moffett. 2003. Using ecological diversity measures with bacterial communities. FEMS Microbiol. Ecol. 43:1-11.
Hofle, M. G. 1988. Identification of bacteria by low molecular weight RNA profiles: a new chemotaxonomic approach. J. Microbiol. Methods 8:235-248.
Ibekwe, A. M., S. K. Papiernik, J. Gan, S. R. Yates, C. H. Yang, and D. E. Crowley. 2001. Impact of fumigants on soil microbial communities. Appl. Environ. Microbiol. 67:3245-3257.
Iwamoto, T., K. Tani, K. Nakamura, Y. Suzuki, M. Kitagawa, M. Eguchi, and M. Nasu. 2000. Monitoring impact of in situ biostimulation treatment on groundwater bacterial community by DGGE. FEMS Microbiol. Ecol. 32:129-141.
Krebs, C. J. (ed.). 1985. Species diversity, p. 507-534. In Ecology: the experimental analysis of distribution and abundance. Harper and Row, New York, N.Y.
Kroes, I., P. W. Lepp, and D. A. Relman. 1999. Bacterial diversity within the human subgingival crevice. Proc. Natl. Acad. Sci. USA 96:14547-14552.
Marriot, F. H. C. 1974. Cluster analysis and related problems, p. 59-68. In The interpretation of multiple observations. Academic Press, London, United Kingdom.
Marsh, P. D. 1991. The significance of maintaining the stability of the natural microflora of the mouth. Br. Dent. J. 171:174-177.
Marsh, P. D. 1994. Microbial ecology of dental plaque and its significance in health and disease. Adv. Dent. Res. 8:263-271.
McBain, A. J., R. G. Bartolo, C. E. Catrenich, D. Charbonneau, R. G. Ledder, and P. Gilbert. 2003. Growth and molecular characterization of dental plaque microcosms. J. Appl. Microbiol. 94:655-664.
Moore, L. V. H., W. E. C. Moore, E. P. Cato, R. M. Smibert, J. A. Brumeister, A. M. Best, and R. R. Ranney. 1987. Bacteriology of human gingivitis. J. Dent. Res. 66:989-995.
Moore, W. E., L. V. Holdeman, R. M. Smibert, I. J. Good, J. A. Burmeister, K. G. Palcanis, and R. R. Ranney. 1982. Bacteriology of experimental gingivitis in young adult humans. Infect. Immun. 38:651-667.
Muller, T., M. Muller, U. Behrendt, and B. Stadler. 2003. Diversity of culturable phyllosphere bacteria on beech and oak: the effects of lepidopterous larvae. Microbiol. Res. 158:291-297.
Muyzer, G. 1999. Genetic fingerprinting of microbial communities: present status and future perspectives In C. R. Bell, M. Brylinsky, and P. Johnson-Green (ed.), Microbial biosystems: new frontiers. Proceedings of the 8th International Symposium on Microbial Ecology. Atlantic Canada Society for Microbial Ecology, Halifax, Canada.
Muyzer, G., E. C. de Waal, and A. G. Uitterlinden. 1993. Profiling of complex microbial populations by denaturing gradient gel electrophoresis analysis of polymerase chain reaction-amplified genes coding for 16S rRNA. Appl. Environ. Microbiol. 59:695-700.
Muyzer, G., and K. Smalla. 1998. Application of denaturing gradient gel electrophoresis (DGGE) and temperature gradient gel electrophoresis (TGGE) in microbial ecology. Antonie Leeuwenhoek 73:127-141.
Ogino, A., H. Koshikawa, T. Nakahara, and H. Uchiyama. 2001. Succession of microbial communities during a biostimulation process as evaluated by DGGE and clone library analyses. J. Appl. Microbiol. 91:625-635.
Page, R. C. 1986. Gingivitis. J. Clin. Periodontol. 13:345-359.
Paster, B. J., S. K. Boches, J. L. Galvin, R. E. Ericson, C. N. Lau, V. A. Levanos, A. Sahasrabudhe, and F. E. Dewhirst. 2001. Bacterial diversity in human subgingival plaque. J. Bacteriol. 183:3770-3783.
Polz, M. F., and C. M. Cavanaugh. 1998. Bias in template-to-product ratios in multitemplate PCR. Appl. Environ. Microbiol. 64:3724-3730.
Shannon, C. E., and W. Weaver. 1971. The mathematical theory of communication. University of Illinois Press, Urbana.
Silva, E. P., and C. A. M. Russo. 2000. Techniques and statistical data analysis in molecular population genetics. Hydrobiologia 420:119-135.
Slots, J., D. Moenbo, J. Langebaek, and A. Frandsen. 1978. Microbiota of gingivitis in man. Scand. J. Dent. Res. 86:174-181.
Smalla, K., G. Wieland, A. Buchner, A. Zock, J. Parzy, S. Kaiser, N. Roskot, H. Heuer, and G. Berg. 2001. Bulk and rhizosphere soil bacterial communities studied by denaturing gradient gel electrophoresis: plant-dependent enrichment and seasonal shifts revealed. Appl. Environ. Microbiol. 67:4742-4751.
Socransky, S. S. 1970. Relationship of bacteria to the etiology of periodontal disease. J. Dent. Res. 49:203-222.
Socransky, S. S., C. Smith, L. Martin, B. J. Paster, F. E. Dewhirst, and A. E. Levin. 1994. "Checkerboard" DNA-DNA hybridization. BioTechniques 17:788-792.
Spratt, D. A., A. J. Weightman, and W. G. Wade. 1999. Diversity of oral asaccharolytic Eubacterium species in periodontitis—identification of novel phylotypes representing uncultivated taxa. Oral Microbiol. Immunol. 14:56-59.
Syed, S. A., and W. J. Loesche. 1972. Survival of human dental plaque flora in various transport media. Appl. Microbiol. 24:638-644.
Tanner, A., M. F. Maiden, B. J. Paster, and F. E. Dewhirst. 1994. The impact of 16S ribosomal RNA-based phylogeny on the taxonomy of oral bacteria. Periodontology 2000 5:26-51.
Tanner, A. C., P. M. Milgrom, R. Kent, Jr., S. A. Mokeem, R. C. Page, S. I. Liao, C. A. Riedy, and J. B. Bruss. 2002. Similarity of the oral microbiota of pre-school children with that of their caregivers in a population-based study. Oral Microbiol. Immunol. 17:379-387.
Tannock, G. W. 2002. Analysis of the intestinal microflora using molecular methods. Eur. J. Clin. Nutr. 56(Suppl. 4):S44-S49.
Teske, A., P. Sigalevich, Y. Cohen, and G. Muyzer. 1996. Molecular identification of bacteria from a coculture by denaturing gradient gel electrophoresis of 16S ribosomal DNA fragments as a tool for isolation in pure cultures. Appl. Environ. Microbiol. 62:4210-4215.
Torsvik, V., K. Salte, R. Sorheim, and J. Goksoyr. 1990. Comparison of phenotypic diversity and DNA heterogeneity in a population of soil bacteria. Appl. Environ. Microbiol. 56:776-781.
Van Der Gucht, K., K. Sabbe, L. De Meester, N. Vloemans, G. Zwart, M. Gillis, and W. Vyverman. 2001. Contrasting bacterioplankton community composition and seasonal dynamics in two neighbouring hypertrophic freshwater lakes. Environ. Microbiol. 3:680-690.
Vanhoutte, T., G. Huys, E. De Brandt, and J. Swings. 2004. Temporal stability analysis of the microbiota in human feces by denaturing gradient gel electrophoresis using universal and group-specific 16S rRNA gene primers. FEMS Microbiol. Ecol. 48:437-446.
Walter, J., C. Hertel, G. W. Tannock, C. M. Lis, K. Munro, and W. P. Hammes. 2001. Detection of Lactobacillus, Pediococcus, Leuconostoc, and Weissella species in human feces by using group-specific PCR primers and denaturing gradient gel electrophoresis. Appl. Environ. Microbiol. 67:2578-2585.
Walter, J., G. W. Tannock, A. Tilsala-Timisjarvi, S. Rodtong, D. M. Loach, K. Munro, and T. Alatossava. 2000. Detection and identification of gastrointestinal Lactobacillus species by using denaturing gradient gel electrophoresis and species-specific PCR primers. Appl. Environ. Microbiol. 66:297-303.
Wennerholm, K., B. Lindquist, and C. G. Emilson. 1995. The toothpick method in relation to other plaque sampling techniques for evaluating mutans streptococci. Eur. J. Oral Sci. 103:36-41.
Widmer, F., A. Bach, E. Laczko, J. Schulze-Aurich, and J. Zeyer. 2001. Assessing soil biological characteristics: a comparison of bulk soil community DNA-, PLFA-, and Biolog(TM)-analyses. Soil Biol. Biochem. 33:1029-1036.
Ximenez-Fyvie, L. A., A. D. Haffajee, and S. S. Socransky. 2000. Comparison of the microbiota of supra- and subgingival plaque in health and periodontitis. J. Clin. Periodontol. 27:648-657.
Zijnge, V., H. J. Harmsen, J. W. Kleinfelder, M. E. van der Rest, J. E. Degener, and G. W. Welling. 2003. Denaturing gradient gel electrophoresis analysis to study bacterial community structure in pockets of periodontitis patients. Oral Microbiol. Immunol. 18:59-65.
Zoetendal, E. G., A. D. Akkermans, and W. M. De Vos. 1998. Temperature gradient gel electrophoresis analysis of 16S rRNA from human fecal samples reveals stable and host-specific communities of active bacteria. Appl. Environ. Microbiol. 64:3854-3859.
Zwart, G., R. Huismans, M. P. van Agterveld, Y. Van de Peer, P. De Rijk, H. Eenhoorn, G. Muyzer, E. J. van Hannen, H. J. Gons, and H. J. Laanbroek. 1998. Divergent members of the bacterial division Verrucomicrobiales in a temperate freshwater lake. FEMS Microbiol. Ecol. 25:159-169.(Gavin P. Gafan, Victoria )
Biostatistics Unit, Eastman Dental Institute, UCL, 256 Gray's Inn Road, London WC1X 8LD, United Kingdom
ABSTRACT
Studies using molecular techniques have demonstrated that a culture-based approach can severely underestimate the bacterial diversity in most environments. One of the molecular techniques that has been applied in microbial ecology is denaturing gradient gel electrophoresis (DGGE). The purpose of this study was to investigate differences in the microbiota of plaque, using a number of analysis techniques, from children without gingivitis (n = 30) and from those with gingivitis (n = 30). Extracted DNA from gingival margin plaque was subjected to PCR targeting the 16S rRNA gene using universal primers. DGGE profiles were analyzed in three ways. (i) Bacterial diversity was compared between cohorts by using the Shannon-Wiener index (also known as the Shannon-Weaver index). (ii) A hierarchical cluster analysis of the banding patterns was calculated and expressed as a dendrogram. (iii) Individual DGGE bands and their intensities for both cohorts were compared using a logistic regression analysis. The Shannon-Wiener indices demonstrated a greater bacterial diversity associated with no-gingivitis plaque (P = 0.009). Dendrograms demonstrated that seven clades associated with gingivitis and five clades associated with no gingivitis. The logistic regression demonstrated that one band was significantly associated with no gingivitis (P = 0.001), while two bands were significantly associated with gingivitis (P = 0.005 and P = 0.042). In conclusion, this study demonstrates that the development of gingivitis might be accompanied by a decrease in bacterial diversity. Furthermore, we have demonstrated that logistic regression is a good statistical method for analyzing and characterizing DGGE profiles.
INTRODUCTION
The accumulation of dental plaque as a result of poor or impaired oral hygiene may result in gingivitis. This inflammation occurs primarily in response to bacteria present in dental plaque (8, 35, 42). Gingivitis is a reversible condition that can precede periodontitis, although not all cases of gingivitis will progress to periodontitis (9). A number of studies of the microbiota associated with gingivitis have been carried out using culture-based techniques (28, 40). Studies employing molecular techniques have elucidated that culture-based approaches may severely underestimate the bacterial diversity in most environments (1). The use of rRNA gene sequences for the analysis of microbial communities has been widely used in microbial ecology and is replacing the more-conventional cultural methods (36). Various molecular approaches for the analysis and comparison of the microbiota of dental plaque in healthy and diseased individuals have been used. One such approach involves the use of DNA probes and checkerboard DNA-DNA hybridization assays (43) that have been previously used to compare the microbiota of supra- and subgingival plaque in healthy individuals and in individuals with periodontitis (57). This approach is hampered by the use of 40 species-specific DNA-DNA hybridization probes, as it only partially considers the microbial diversity in dental plaque, given that approximately 500 species live in the oral cavity (36). Furthermore, most conventional methods used for analyzing bacterial communities do not take the unculturable proportion into consideration (1). This may be a large oversight, as unculturable bacteria are believed to constitute up to 50% of dental plaque communities (46). Earlier workers (23) have demonstrated that a greater microbial diversity could be ascertained by amplifying rRNA genes directly from subgingival plaque than by cultivating bacteria from the same specimen. Currently the best model for exploring microbial diversity involves PCR cloning of the 16S rRNA gene (4, 44). Cloning and sequencing of multiple plaque samples are extremely laborious and expensive tasks to perform thoroughly due to the huge diversity involved (36). For this purpose, genetic fingerprinting techniques may be better suited.
Genetic fingerprinting techniques provide a rapid and relatively easy alternative to the analysis of microbial communities. The different types of genetic fingerprinting techniques that can be used for this purpose, together with their advantages and disadvantages, have already been described (31). One such fingerprinting technique is the electrophoretic separation of low-molecular-weight rRNA molecules (5S rRNA and tRNA) extracted from natural samples (19). Denaturing gradient gel electrophoresis (DGGE) is a more recent fingerprinting technique by which PCR-amplified DNA fragments are separated according to their sequence information (33). The basis of this technique is that DNA fragments of the same size but with differing base pair sequences can be separated (32). This separation by DGGE relies on the electrophoretic mobility of partially denatured DNA molecules in a polyacrylamide gel, which is encumbered in comparison to the completely helical form of the molecule (32). DGGE has been applied in environmental microbiology (6, 11, 49) and food microbiology (2, 5) and in the analysis of microbial communities in the human body (10, 14, 53, 54). Recently DGGE has also been applied to analyze the bacterial diversity of human subgingival plaque (17, 58) as well as laboratory-grown dental plaque microcosms (27). This paper presents a fingerprint analysis of dental plaque sampled from the gingival crevice of prepubertal children with and without gingivitis. Any marked differences observed in the microbial composition of plaque associated with healthy dental sites and gingivitis would further advance our understanding of how specific bacteria are associated with, and respond to, inflammation and indeed the progression to periodontal disease.
Analysis of DGGE profiles often involves the use of principal component analysis (34, 56) and multidimensional scaling (7, 21) but generally utilizes hierarchical cluster analysis to demonstrate similarities in the data, and the results are invariably presented in the form of dendrograms (6, 7, 13, 51). This study analyzes the DGGE profiles of the dental plaque sampled from both cohorts by using hierarchical cluster analyses as well as measuring community diversity by means of the Shannon-Wiener index (also known as the Shannon-Weaver index). Furthermore, we have applied a statistical test for the analysis of the electrophoretic fingerprints that appears to be used infrequently. This involves a logistic regression analysis that looks at single bands and computes a regression relationship between the presence and absence of gingivitis and band presence and intensity (independent or explanatory variable). This type of analysis can pinpoint individual bands that when present/absent will affect the outcome of interest, in this case the outcome being whether the subject had gingivitis or not.
MATERIALS AND METHODS
Patients. Ethical approval was obtained from the Eastman Dental Hospital NHS Trust, London, United Kingdom. Each parent was given an information sheet to read and asked for written consent. Each child was asked for verbal consent.
Subjects. Healthy children aged between 5 and 9 years attending the Department of Pediatric Dentistry and the School of Dental Therapy at the Eastman Dental Hospital were recruited. Children with chronic medical disorders or who had been treated with antibiotics within the preceding 3 months were excluded. Two groups of children (a total of 60) were selected: group 1, 30 children with discernible plaque but no gingivitis associated with either the lower left or lower right first permanent molar tooth; group 2, 30 children with discernible plaque and gingivitis associated with either the lower left or lower right first permanent molar tooth. Plaque and gingivitis scores were recorded for each child using a modified version of the method of O'Leary (15). Four gingivally related quadrisections of each tooth (mesiobuccal, distobuccal, mesiolingual, and distolingual) were visually examined, and each was given a score of 0 for no discernible plaque or a score of 1 for discernible dental plaque deposits. Similarly, each tooth quadrisection associated with no gingival inflammation was given a score of 0, and each quadrisection with gingival inflammation was given a score of 1. Plaque was sampled from the buccal and lingual gingival margin of either the lower left or lower right first permanent molar tooth using a sterile wooden toothpick (47, 55). When both lower permanent first molars had erupted, one tooth (either the right or left tooth) was randomly selected. The toothpick was immediately placed in a sterile container with 1 ml of reduced transport fluid (45) and five glass beads. The plaque samples were dispersed in the reduced transport fluid by vortexing for 10 s, and DNA was immediately extracted using the Puregene DNA isolation kit for yeast and gram-positive bacteria (Gentra Systems, Minneapolis, MN). The DNA was then stored at –20°C.
DGGE profiling. DNA from plaque sampled from 60 different patients were used as templates (group 1, 30 patients with discernible plaque and no gingivitis; group 2, 30 patients with discernible plaque and gingivitis). The V2-V3 region of the 16S rRNA gene corresponding to positions 339 to 539 of Escherichia coli was partially amplified using PCR primer (Genosys, Cambridgeshire, United Kingdom) F357GC (5'-CGCCCGCCGCGCCCCGCGCCCGGCCCGCCGCCCCCGCCCCCCTACGGGAGGCAGCAG-3'), which contains a GC-rich clamp, and PCR primer R518 (5'-ATTACCGCGGCTGCTGG-3'). The final volume of each PCR mixture was 100 μl. The amplification reaction mixture contained 10x PCR buffer, 2.5 mM MgCl2, 0.2 μM of both primers F357GC and R518 and 5 U of Taq polymerase (Bioloine, London, United Kingdom). All deoxynucleoside triphosphates (Promega, Southampton, United Kingdom) were used at a final concentration of 0.2 mM. The cycling parameters for touchdown PCR were carried out on a Primus thermal cycler (MWG Biotech, Milton Keynes, United Kingdom) and were as follows. After preincubation at 94°C for 5 min, 30 cycles were performed at 94°C for 1 min, the annealing temperature (TA) for 1 min, and 72°C for 1 min. The TA decreased stepwise by 1°C every two cycles from 65°C in the first cycle to 56°C in the 20th cycle. The TA for the last 10 cycles was 55°C. Cycling was followed by 5-min incubation at 72°C (60).
The 16S rRNA gene amplicons were cleaned and concentrated into a final volume of 30 μl by using a QIAquick PCR purification kit (QIAGEN Ltd., Crawley, United Kingdom). In addition, two positive controls consisting of genomic DNA extracted from Tannerella forsythensis ATCC 43037 (formerly Bacteroides forsythus) and Prevotella intermedia (wild type) as well as a negative control consisting of sterile deionized water. The two positive controls were used to align multiple gel images (Fig. 1).
Parallel gels containing 10% (wt/vol) polyacrylamide (37.5:1 acrylamide:bisacrylamide) were cast using a D-code system (Bio-Rad Laboratories, Inc., Hercules, CA). The gels contained a linear gradient of the denaturants urea and formamide, increasing from 40% at the top of the gel to 80% at the bottom (with 100% denaturants corresponding to 7 M urea and 40% [vol/vol] deionized formamide). Gels were run at 35 V for 21 h (735 V hours) at a constant temperature of 60°C in 7 liters of 1x Tris-acetate-EDTA (TAE) buffer. Gels were stained for 1 h in 1x TAE containing SYBR green nucleic acid gel stain (10–4 dilution) (Molecular Probes, PoortGebouw, The Netherlands) and photographed under a UV light transilluminator (AlphaImager, San Leandro, CA).
Statistical analysis of DGGE banding patterns. All gels were aligned using the two positive controls. The DNA bands that migrated to the same position within each gel were ascribed a number. Band strengths were estimated visually; weak bands were assigned a value of 1, intermediate bands a value of 2, and strong bands a value of 3 (Fig. 1). Thus, if the odds ratio were unity, the odds of a subject having gingivitis did not differ from the odds of a subject having no gingivitis, as the strength of the band increased to unity. If the odds ratio for a band were greater than 1, the odds of having gingivitis were greater than the odds of not having gingivitis as the intensity of the band increased. In the same way, if the estimated odds ratio for a band was less than 1, then the odds of not having gingivitis was greater than that of having gingivitis as the band intensity increased. These data were then analyzed using the following three methods.
(i) Shannon-Wiener diversity index. The Shannon-Wiener index of diversity (H') (7, 12, 22) was used to determine the diversity of taxa present in plaque sampled from children with and without gingivitis. This index was calculated by the following equation:
where s is the number of species in the sample and Pi is the proportion of species i in the sample. Gaussian distribution tests were applied to the Shannon-Wiener diversity index values for the 30 subjects in both plaque groups. Generally, the data were nonuniformly distributed; therefore, a nonparametric analysis using a Mann-Whitney U test was performed where a P of <0.05 was interpreted to be statistically significant. The nonparametric statistical analysis was performed using SPSS version 12 (Chicago, Ill.).
(ii) Cluster analysis. Similarities between the banding patterns generated by PCR-DGGE of the plaque samples were analyzed using the Pearson correlation coefficient (41). Similarities were displayed graphically as a dendrogram. The clustering algorithms used to calculate the dendrograms was an unweighted pair group method with arithmetic averages (UPGMA) (16, 39). The cluster analysis and dendrogram generation were carried out by using the Phoretix 1D Advanced and Phoretix 1D Database software (Phoretix International, Newcastle upon Tyne, United Kingdom).
(iii) Logistic regression. The data were first analyzed using univariable chi-square tests with a cutoff for significance of 5% to reduce the number of bands included in the multivariate analysis. The bands demonstrating significance were then included as explanatory variables (each band was recorded as present [coded as 1] or absent [coded as 0]) in a multivariate logistic regression analysis, for which the codes, no gingivitis (coded as 0) and gingivitis (coded as1) were used for the dependent variable. The estimated odds ratio with a 95% confidence interval was calculated for each band in the logistic regression model. The predicted probabilities for a subject belonging to a particular group (in this case, gingivitis and no gingivitis) were also calculated. Furthermore, a Hosmer and Lemeshow test for goodness of fit of the logistic model was performed. The logistic regression analysis was conducted using SPSS version 12.
RESULTS
Shannon-Wiener diversity index. The Shannon-Wiener indices for subjects with and without gingivitis were compared using the Mann-Whitney U test. The results revealed that there was a significantly greater biological diversity in the sample group with no gingivitis (for the no-gingivitis group, the median value was 3.4 and the values ranged from 2.5 to 4.3; for the gingivitis group, the median value was 3.2 and the values ranged from 1.6 to 4.1; P = 0.009).
Cluster analysis. The cluster analysis generated an UPGMA dendrogram (Fig. 2) which demonstrates that seven clades associated only with the gingivitis group and five clades associated with the no-gingivitis group.
Hosmer and Lemeshow test. The Hosmer and Lemeshow test for goodness of fit showed that the model was a good fit (chi-square = 2.903, degrees of freedom = 8, P = 0.940). The expected probability of a subject having gingivitis or not having gingivitis, as determined by the logistic regression model, is represented as both a classification table (Table 1) and a classification plot (Fig. 3). The classification table shows that the model works well and approximately 84% of the individuals were correctly classified as having the appropriate disease status. The classification plot further demonstrates that the model works well, in that most subjects with no gingivitis (coded as 0) are predicted as not having gingivitis (i.e., on the left-hand side of the horizontal axis). Likewise, most of the subjects with gingivitis (coded as 1) are predicted as having gingivitis (i.e., on the right-hand side of the horizontal axis). Logistic regression analysis also determined that three bands were significantly associated with the presence or absence of gingivitis in these subjects. From this regression analysis, the estimated odds ratio of any band could be interpreted as the ratio of the odds of having gingivitis in those subjects with the band to the odds of having gingivitis in those subjects without the band. Table 2 presents the results of the analysis and demonstrates that band 2 was significantly associated with no gingivitis (P = 0.001), while bands 1 and 3 were significantly associated with gingivitis (P = 0.042 and P = 0.005, respectively). The positions of these three bands on a DGG are shown in Fig. 4.
DISCUSSION
One of the major challenges in microbial ecology is the assessment of the bacterial diversity or community structure present in a defined environment. It is important to understand the changes in diversity and structure during the inception and progression of polymicrobial diseases. Recent advances in molecular biology techniques have allowed us to better characterize and understand these changes. While the molecular biology techniques are now becoming available, the associated data analysis techniques have lagged behind somewhat. A case in point is the present study, which investigates plaque sampled from the gingival crevice of children both with and without gingivitis in order to detect any differences between the two groups. Such a study could potentially be used to understand the differences between a healthy dental microbiota and its progression to periodontal disease. We chose to use a culture-independent technique (DGGE) to characterize the population differences between the two cohorts. This approach would take into account the uncultivable portion of the oral microbiota in the analysis. The advantages of using DGGE as a means of studying microbial ecology are well established in that it provides a more accurate means of visualizing whole microbial communities (since there should be a reduced bias in the detection of unculturable species). Additionally, it is less labor-intensive than the more-conventional culture-independent techniques of PCR cloning and sequencing.
Once the data were generated for all the subjects in both cohorts, we used a number of different analysis techniques to interrogate these data. In the first instance, this included a "by-eye" analysis as is common in the literature (27, 48). It is relatively easy to comment on three or four patterns and select bands of interest on the basis of visual analysis. However, the banding patterns generated were quite complex, and comparing 60 patterns proved difficult. In the first instance, we analyzed the data using a similar approach commonly used by other workers, e.g., diversity indices and cluster analysis.
There are a variety of ecological diversity measures, but their suitability for use with highly diverse microbial communities is unclear (18). The Shannon-Wiener index of diversity (38) has been used previously in microbial ecology (12, 18, 30, 34). There is some uncertainty surrounding this index, particularly as it is referred to by different names. The full title of the index is the Shannon-Wiener function, after Claude Shannon and Norbert Wiener (38). It is more commonly known as the Shannon-Weaver index after Shannon's coauthor, Wallace Weaver. In the present study, the Shannon-Wiener index of diversity was applied to 16S rRNA genes from dental plaque communities that were separated on denaturing gels according to their sequence heterogeneity. Using both the total number of DGGE bands and their relative intensities, it was possible to calculate the bacterial diversity index (H'). This would in theory reflect on sample diversity without the need for cultivation. It is a fair assumption to expect that certain target rRNA genes within the plaque samples would be present in lower concentrations and thus would not be amplified sufficiently to be visualized as bands. Therefore, the bands generated in the present study would reflect the most abundant rRNA gene types from dental plaque or perhaps those strains most applicable to amplification (37). Thus, the diversity indices calculated from the DGGE pattern were regarded as relative. Previous workers (13) have demonstrated that the diversity index H' can be applied to complex microbial communities and is well suited for comparing large sets of samples.
The results demonstrated that there was a greater biological diversity in the sample group with no gingivitis than the gingivitis group (P = 0.009). This might indicate that a decrease in bacterial diversity may be associated with the shift from health to gingivitis. According to the ecological plaque hypothesis, a change in the environment leads to a shift in the community structure; however, this shift is only a change in the proportions of certain taxa within the community (25, 26). Therefore, perhaps the perceived decrease in diversity is a result of certain taxa proliferating as a consequence of inflammation. As a further consequence, this may "mask" the presence of bacteria that do not necessarily increase in number as a consequence of inflammation. Nevertheless, when considering ecological diversity and community structure, it is believed that species diversity is an important feature in maintaining a degree of stability within that community. Thus, it is expected that stable and resilient microbial communities must contain a certain level of diversity (3). As has been shown, the present results demonstrate a greater diversity in the DGGE profiles of dental plaque associated with no gingivitis. This may indicate that the dental plaque community associated with no gingivitis is probably more stable than the dental plaque community associated with gingivitis. While the overall precepts of ecology support this (3), workers using culture-dependent techniques to study experimental gingivitis have come to the opposite conclusion (29). There are no studies that are experimentally comparable to this present study; therefore, a detailed comparison cannot be made. However, a factor that may go some way to reconciling these apparently conflicting results is that the DGGE technique takes into account the uncultivable microbiota present in the sample (in this case, the oral cavity where it is thought that around 50% of the species are uncultivable).
The use of diversity indices for DGGE analysis in this case uses very little of the data present and as stated provides only a measure of diversity. Therefore, it is entirely possible for two communities to be equally diverse (i.e., equal H' values) but have, in the most extreme case, a completely different species structure (3) (Fig. 5). We believe that the DGGE data contain more information than simply this.
Characteristic profiles of microbial communities or DNA fingerprints can be produced by DGGE. These microbial fingerprints can be represented as binary vectors, which have the potential to be of a very high dimension, depending on the environment sampled, such as soil (50) or dental plaque (36). Earlier workers have used principal component analysis (34, 56) as well as multidimensional scaling (7, 21) to analyze groups of similarities in the data. In general, most researchers use hierarchical clustering algorithms based on similarity indices for binary vectors (6, 7, 13, 51). Cluster analysis, also known as "classification," has been defined as the search for a natural grouping (24). Cluster analysis, for example, the UPGMA applied in this study, can be used to identify samples that generate similar patterns (7, 20). The UPGMA dendrogram calculated demonstrated similarities in the banding profiles of plaque sampled from subjects both with and without gingivitis. There were seven clades that clustered similar band profiles from plaque sampled from gingivitis sites. Likewise, the cluster analysis demonstrated five clades of similar band profiles from plaque sampled from sites without gingivitis. Indeed, 14/30 "healthy" samples and 20/40 "gingivitis" samples clustered together using this method. This implies that specific samples whether derived from healthy subjects or subjects with gingivitis are related to each other based on the DGGE banding pattern. It also shows that more than one profile is associated with either condition and therefore that perhaps there are a number of specific microbiotas associated with health or gingivitis. Indeed, perhaps only one of the gingivitis clades will eventually lead to periodontal disease. Earlier workers have postulated the association of certain taxa with the onset and progression of periodontal disease, for example, the red complex (Porphyromonas gingivalis, Tannerella forsythensis, and Treponema denticola) suggested by Ximenez-Fyvie et al. (57). Perhaps these three key taxa are represented in only one of the gingivitis clades, and therefore, only one type of gingivitis (one bacterial community) actually leads to periodontitis. It would be interesting to follow these subjects, monitor their profile, and record which clades lead to the inception of periodontal disease. However, this analysis technique is limited by the fact that it does not take into consideration the outcome, either gingivitis or no gingivitis. It aims to look at clusters of bands to predict the outcome but does not utilize the outcome itself when comparing the data for similarities. Earlier workers (52, 59) applying similar techniques have demonstrated that individuals have their own unique fecal microbial community that remains relatively stable over time. We believe that this is also the case for the oral cavity (unpublished data). Analysis of the differences in microbial fingerprints for both gingivitis and no gingivitis by cluster analysis may be hampered by the fact that substantial differences may exist within the individual cohorts to start with.
The hierarchical cluster analysis is a good technique for longitudinal data, but with cross-sectional data, there seems to be too much variation within each group.
Both analyses have yielded valuable information, but we believe that further information can be obtained from the DGGE profiles. For example, two questions that we think can be answered by using DGGE profiles are "is there a recognizable fingerprint which can be used as a predictor to differentiate between disease states" and "are there certain bands associated with health or indeed gingivitis" A few workers have used principal component analysis and multidimensional scaling to analyze data such as these data. Furthermore, these methods may not always be appropriate, and indeed, they are very complex techniques both to understand and use. We therefore applied a different approach for data analysis, namely, logistic regression.
Unlike hierarchical cluster analysis, logistic regression analysis takes the outcome into consideration (i.e., in this case, gingivitis or no gingivitis) in addition to differences in band numbers and band migration position. This type of analysis does not take into account clusters of multiple bands in an attempt to examine DGGE profiles for similarities. Indeed, it analyzes the fingerprint with respect to the presence/absence of individual bands. Analysis of individual bands from the total fingerprint allows for the computation of a regression relationship between a clinical descriptor, such as health or gingivitis (outcome variable) and band presence/absence (explanatory variable). The results of this regression analysis (Fig. 3) have demonstrated that there were substantial differences in the DGGE profiles of both cohorts of subjects. The presence or absence of gingivitis was correctly classified in 84% of the subjects. This classification plot correctly predicts most subjects with no gingivitis (coded as 0) as not having gingivitis and most subjects with gingivitis (coded as 1) as having gingivitis on the basis of the presence/absence of bands. Moreover, the results from the regression analysis have further demonstrated that a single band (band 2) is statistically associated with the absence of gingivitis, while bands 1 and 3 were significantly associated with gingivitis.
In conclusion, we have used a number of different techniques to analyze DGGE banding profiles, each of which answer different and progressively more complex questions. The two analyses used first, i.e., the diversity index and the cluster analysis, both provide useful data but do not utilize all available information. The diversity index simply measures a change in diversity but does not specify what changes or even whether the community structure is similar. Cluster analysis, on the other hand, attempts to differentiate between cohorts that contain large differences within the cohort itself. Furthermore, cluster analysis does not specify which operational taxonomic units are important, only that differences exist. Logistic regression, however, not only successfully differentiates between the profiles of both cohorts but can also specify individual operational taxonomic units associated with these differences. Therefore, we believe that a logistic regression analysis in conjunction with a diversity measure is the method of choice to analyze and compare DGGE-generated community profiles in a cross-sectional study.
ACKNOWLEDGMENTS
We extend our thanks to Derren Ready for initial bacterial diversity analysis. We also thank Charlotte Davies for helping with the initial DGGE setup.
REFERENCES
Amann, R. I., W. Ludwig, and K. H. Schleifer. 1995. Phylogenetic identification and in situ detection of individual microbial cells without cultivation. Microbiol. Rev. 59:143-169.
Ampe, F., A. Sirvent, and N. Zakhia. 2001. Dynamics of the microbial community responsible for traditional sour cassava starch fermentation studied by denaturing gradient gel electrophoresis and quantitative rRNA hybridization. Int. J. Food Microbiol. 65:45-54.
Atlas, R. M. 1994. Diversity of microbial communities. Adv. Microb. Ecol. 7:1-47.
Becker, M. R., B. J. Paster, E. J. Leys, M. L. Moeschberger, S. G. Kenyon, J. L. Galvin, S. K. Boches, F. E. Dewhirst, and A. L. Griffen. 2002. Molecular analysis of bacterial species associated with childhood caries. J. Clin. Microbiol. 40:1001-1009.
ben Omar, N., and F. Ampe. 2000. Microbial community dynamics during production of the Mexican fermented maize dough pozol. Appl. Environ. Microbiol. 66:3664-3673.
Boon, N., C. Marle, E. M. Top, and W. Verstraete. 2000. Comparison of the spatial homogeneity of physico-chemical parameters and bacterial 16S rRNA genes in sediment samples from a dumping site for dredging sludge. Appl. Microbiol. Biotechnol. 53:742-747.
Boon, N., W. De Windt, W. Verstraete, and E. M. Top. 2002. Evaluation of nested PCR-DGGE (denaturing gradient gel electrophoresis) with group-specific 16S rRNA primers for the analysis of bacterial communities from different wastewater treatment plants. FEMS Microbiol. Ecol. 39:101-112.
Breuer, M. M., and R. S. Cosgrove. 1989. The relationship between gingivitis and plaque levels. J. Periodontol. 60:172-175.
Brown, L. J., and H. Loe. 1993. Prevalence, extent, severity and progression of periodontal disease. Periodontology 2000 2:57-71.
Donskey, C. J., A. M. Hujer, S. M. Das, N. J. Pultz, R. A. Bonomo, and L. B. Rice. 2003. Use of denaturing gradient gel electrophoresis for analysis of the stool microbiota of hospitalized patients. J. Microbiol. Methods 54:249-256.
Ebie, Y., M. Matsumura, N. Noda, S. Tsuneda, A. Hirata, and Y. Inamori. 2002. Community analysis of nitrifying bacteria in an advanced and compact Gappei-Johkasou by FISH and PCR-DGGE. Water Sci. Technol. 46:105-111.
Edwards, M. L., A. K. Lilley, T. H. Timms-Wilson, I. P. Thompson, and I. Cooper. 2001. Characterisation of the culturable heterotrophic bacterial community in a small eutrophic lake (Priest Pot). FEMS Microbiol. Ecol. 35:295-304.
Eichner, C. A., R. W. Erb, K. N. Timmis, and I. Wagner-Dobler. 1999. Thermal gradient gel electrophoresis analysis of bioprotection from pollutant shocks in the activated sludge microbial community. Appl. Environ. Microbiol. 65:102-109.
Favier, C. F., E. E. Vaughan, W. M. De Vos, and A. D. L. Akkermans. 2002. Molecular monitoring of succession of bacterial communities in human neonates. Appl. Environ. Microbiol. 68:219-226.
Franco, E., C. P. Saunders, G. J. Roberts, and A. Suwanprasit. 1996. Dental disease, caries related microflora and salivary IgA of children with severe congenital cardiac disease: an epidemiological and oral microbial survey. Pediatr. Dent. 18:228-235.
Fromin, N., J. Hamelin, S. Tarnawski, D. Roesti, K. Jourdain-Miserez, N. Forestier, S. Teyssier-Cuvelle, F. Gillet, M. Aragno, and P. Rossi. 2002. Statistical analysis of denaturing gel electrophoresis (DGE) fingerprinting patterns. Environ. Microbiol. 4:634-643.
Fujimoto, C., H. Maeda, S. Kokeguchi, S. Takashiba, F. Nishimura, H. Arai, K. Fukui, and Y. Murayama. 2003. Application of denaturing gradient gel electrophoresis (DGGE) to the analysis of microbial communities of subgingival plaque. J. Periodontal Res. 38:440-445.
Hill, T. C. J., K. A. Walsh, J. A. Harris, and B. F. Moffett. 2003. Using ecological diversity measures with bacterial communities. FEMS Microbiol. Ecol. 43:1-11.
Hofle, M. G. 1988. Identification of bacteria by low molecular weight RNA profiles: a new chemotaxonomic approach. J. Microbiol. Methods 8:235-248.
Ibekwe, A. M., S. K. Papiernik, J. Gan, S. R. Yates, C. H. Yang, and D. E. Crowley. 2001. Impact of fumigants on soil microbial communities. Appl. Environ. Microbiol. 67:3245-3257.
Iwamoto, T., K. Tani, K. Nakamura, Y. Suzuki, M. Kitagawa, M. Eguchi, and M. Nasu. 2000. Monitoring impact of in situ biostimulation treatment on groundwater bacterial community by DGGE. FEMS Microbiol. Ecol. 32:129-141.
Krebs, C. J. (ed.). 1985. Species diversity, p. 507-534. In Ecology: the experimental analysis of distribution and abundance. Harper and Row, New York, N.Y.
Kroes, I., P. W. Lepp, and D. A. Relman. 1999. Bacterial diversity within the human subgingival crevice. Proc. Natl. Acad. Sci. USA 96:14547-14552.
Marriot, F. H. C. 1974. Cluster analysis and related problems, p. 59-68. In The interpretation of multiple observations. Academic Press, London, United Kingdom.
Marsh, P. D. 1991. The significance of maintaining the stability of the natural microflora of the mouth. Br. Dent. J. 171:174-177.
Marsh, P. D. 1994. Microbial ecology of dental plaque and its significance in health and disease. Adv. Dent. Res. 8:263-271.
McBain, A. J., R. G. Bartolo, C. E. Catrenich, D. Charbonneau, R. G. Ledder, and P. Gilbert. 2003. Growth and molecular characterization of dental plaque microcosms. J. Appl. Microbiol. 94:655-664.
Moore, L. V. H., W. E. C. Moore, E. P. Cato, R. M. Smibert, J. A. Brumeister, A. M. Best, and R. R. Ranney. 1987. Bacteriology of human gingivitis. J. Dent. Res. 66:989-995.
Moore, W. E., L. V. Holdeman, R. M. Smibert, I. J. Good, J. A. Burmeister, K. G. Palcanis, and R. R. Ranney. 1982. Bacteriology of experimental gingivitis in young adult humans. Infect. Immun. 38:651-667.
Muller, T., M. Muller, U. Behrendt, and B. Stadler. 2003. Diversity of culturable phyllosphere bacteria on beech and oak: the effects of lepidopterous larvae. Microbiol. Res. 158:291-297.
Muyzer, G. 1999. Genetic fingerprinting of microbial communities: present status and future perspectives In C. R. Bell, M. Brylinsky, and P. Johnson-Green (ed.), Microbial biosystems: new frontiers. Proceedings of the 8th International Symposium on Microbial Ecology. Atlantic Canada Society for Microbial Ecology, Halifax, Canada.
Muyzer, G., E. C. de Waal, and A. G. Uitterlinden. 1993. Profiling of complex microbial populations by denaturing gradient gel electrophoresis analysis of polymerase chain reaction-amplified genes coding for 16S rRNA. Appl. Environ. Microbiol. 59:695-700.
Muyzer, G., and K. Smalla. 1998. Application of denaturing gradient gel electrophoresis (DGGE) and temperature gradient gel electrophoresis (TGGE) in microbial ecology. Antonie Leeuwenhoek 73:127-141.
Ogino, A., H. Koshikawa, T. Nakahara, and H. Uchiyama. 2001. Succession of microbial communities during a biostimulation process as evaluated by DGGE and clone library analyses. J. Appl. Microbiol. 91:625-635.
Page, R. C. 1986. Gingivitis. J. Clin. Periodontol. 13:345-359.
Paster, B. J., S. K. Boches, J. L. Galvin, R. E. Ericson, C. N. Lau, V. A. Levanos, A. Sahasrabudhe, and F. E. Dewhirst. 2001. Bacterial diversity in human subgingival plaque. J. Bacteriol. 183:3770-3783.
Polz, M. F., and C. M. Cavanaugh. 1998. Bias in template-to-product ratios in multitemplate PCR. Appl. Environ. Microbiol. 64:3724-3730.
Shannon, C. E., and W. Weaver. 1971. The mathematical theory of communication. University of Illinois Press, Urbana.
Silva, E. P., and C. A. M. Russo. 2000. Techniques and statistical data analysis in molecular population genetics. Hydrobiologia 420:119-135.
Slots, J., D. Moenbo, J. Langebaek, and A. Frandsen. 1978. Microbiota of gingivitis in man. Scand. J. Dent. Res. 86:174-181.
Smalla, K., G. Wieland, A. Buchner, A. Zock, J. Parzy, S. Kaiser, N. Roskot, H. Heuer, and G. Berg. 2001. Bulk and rhizosphere soil bacterial communities studied by denaturing gradient gel electrophoresis: plant-dependent enrichment and seasonal shifts revealed. Appl. Environ. Microbiol. 67:4742-4751.
Socransky, S. S. 1970. Relationship of bacteria to the etiology of periodontal disease. J. Dent. Res. 49:203-222.
Socransky, S. S., C. Smith, L. Martin, B. J. Paster, F. E. Dewhirst, and A. E. Levin. 1994. "Checkerboard" DNA-DNA hybridization. BioTechniques 17:788-792.
Spratt, D. A., A. J. Weightman, and W. G. Wade. 1999. Diversity of oral asaccharolytic Eubacterium species in periodontitis—identification of novel phylotypes representing uncultivated taxa. Oral Microbiol. Immunol. 14:56-59.
Syed, S. A., and W. J. Loesche. 1972. Survival of human dental plaque flora in various transport media. Appl. Microbiol. 24:638-644.
Tanner, A., M. F. Maiden, B. J. Paster, and F. E. Dewhirst. 1994. The impact of 16S ribosomal RNA-based phylogeny on the taxonomy of oral bacteria. Periodontology 2000 5:26-51.
Tanner, A. C., P. M. Milgrom, R. Kent, Jr., S. A. Mokeem, R. C. Page, S. I. Liao, C. A. Riedy, and J. B. Bruss. 2002. Similarity of the oral microbiota of pre-school children with that of their caregivers in a population-based study. Oral Microbiol. Immunol. 17:379-387.
Tannock, G. W. 2002. Analysis of the intestinal microflora using molecular methods. Eur. J. Clin. Nutr. 56(Suppl. 4):S44-S49.
Teske, A., P. Sigalevich, Y. Cohen, and G. Muyzer. 1996. Molecular identification of bacteria from a coculture by denaturing gradient gel electrophoresis of 16S ribosomal DNA fragments as a tool for isolation in pure cultures. Appl. Environ. Microbiol. 62:4210-4215.
Torsvik, V., K. Salte, R. Sorheim, and J. Goksoyr. 1990. Comparison of phenotypic diversity and DNA heterogeneity in a population of soil bacteria. Appl. Environ. Microbiol. 56:776-781.
Van Der Gucht, K., K. Sabbe, L. De Meester, N. Vloemans, G. Zwart, M. Gillis, and W. Vyverman. 2001. Contrasting bacterioplankton community composition and seasonal dynamics in two neighbouring hypertrophic freshwater lakes. Environ. Microbiol. 3:680-690.
Vanhoutte, T., G. Huys, E. De Brandt, and J. Swings. 2004. Temporal stability analysis of the microbiota in human feces by denaturing gradient gel electrophoresis using universal and group-specific 16S rRNA gene primers. FEMS Microbiol. Ecol. 48:437-446.
Walter, J., C. Hertel, G. W. Tannock, C. M. Lis, K. Munro, and W. P. Hammes. 2001. Detection of Lactobacillus, Pediococcus, Leuconostoc, and Weissella species in human feces by using group-specific PCR primers and denaturing gradient gel electrophoresis. Appl. Environ. Microbiol. 67:2578-2585.
Walter, J., G. W. Tannock, A. Tilsala-Timisjarvi, S. Rodtong, D. M. Loach, K. Munro, and T. Alatossava. 2000. Detection and identification of gastrointestinal Lactobacillus species by using denaturing gradient gel electrophoresis and species-specific PCR primers. Appl. Environ. Microbiol. 66:297-303.
Wennerholm, K., B. Lindquist, and C. G. Emilson. 1995. The toothpick method in relation to other plaque sampling techniques for evaluating mutans streptococci. Eur. J. Oral Sci. 103:36-41.
Widmer, F., A. Bach, E. Laczko, J. Schulze-Aurich, and J. Zeyer. 2001. Assessing soil biological characteristics: a comparison of bulk soil community DNA-, PLFA-, and Biolog(TM)-analyses. Soil Biol. Biochem. 33:1029-1036.
Ximenez-Fyvie, L. A., A. D. Haffajee, and S. S. Socransky. 2000. Comparison of the microbiota of supra- and subgingival plaque in health and periodontitis. J. Clin. Periodontol. 27:648-657.
Zijnge, V., H. J. Harmsen, J. W. Kleinfelder, M. E. van der Rest, J. E. Degener, and G. W. Welling. 2003. Denaturing gradient gel electrophoresis analysis to study bacterial community structure in pockets of periodontitis patients. Oral Microbiol. Immunol. 18:59-65.
Zoetendal, E. G., A. D. Akkermans, and W. M. De Vos. 1998. Temperature gradient gel electrophoresis analysis of 16S rRNA from human fecal samples reveals stable and host-specific communities of active bacteria. Appl. Environ. Microbiol. 64:3854-3859.
Zwart, G., R. Huismans, M. P. van Agterveld, Y. Van de Peer, P. De Rijk, H. Eenhoorn, G. Muyzer, E. J. van Hannen, H. J. Gons, and H. J. Laanbroek. 1998. Divergent members of the bacterial division Verrucomicrobiales in a temperate freshwater lake. FEMS Microbiol. Ecol. 25:159-169.(Gavin P. Gafan, Victoria )