OUP user menu

Validating T-RFLP as a sensitive and high-throughput approach to assess bacterial diversity patterns in human anterior nares

Amélia Camarinha-Silva , Melissa L. Wos-Oxley , Ruy Jáuregui , Karsten Becker , Dietmar H. Pieper
DOI: http://dx.doi.org/10.1111/j.1574-6941.2011.01197.x 98-108 First published online: 1 January 2012

Abstract

While recent works aimed to thoroughly characterize the bacterial community of the human anterior nares of a few candidates, this work sought to analyse a greater cross-section by sampling 100 volunteers. After optimizing and validating the method of terminal restriction fragment length polymorphism against six previously pyrosequenced samples, abundant species could be discriminated and their relative abundances measured in a high-throughput manner. The 100 volunteers could be statistically clustered into 12 groups, where two-thirds of volunteers shared more than 40% similarity in respect to their bacterial community structure, while the remaining third clustered into smaller groups being dominated by Dolosigranulum pigrum, Moraxella spp. or Staphylococcus aureus. Moraxella spp. was present predominantly in women rather than in men. The use of network analysis charting bacterial ecological co-occurrences revealed new evidence of likely positive associations between some core human nasal species. So, in the age of post ‘omics’ and ‘deep sequencing’, there is still a place for these well-tried and well-tested methods that can offer a rapid, reproducible and economical alternative, whereby also yielding valuable new information.

Keywords
  • anterior nares
  • microbiota
  • T-RFLP
  • community analysis
  • bacterial associations
  • Staphylococcus aureus

Introduction

The human anterior nares are now known to be the principle habitat of both commensals and opportunistic pathogens such as Staphylococcus aureus, Staphylococcus epidermidis, Corynebacterium spp., Propionibacterium spp., Dolosigranulum pigrum, Finegoldia magna, Peptoniphilus sp., Moraxella spp. and Anaerococcus spp., all of which constitute part of the normal and asymptomatic core bacterial community (Wos-Oxley et al., 2010). Although S. aureus colonization of the nares is asymptomatic, nasal carriage has a crucial function as a source of invasive infections in both community and hospital settings (von Eiff et al., 2001; van Belkum et al., 2009). Therefore, elimination of S. aureus nasal carriage seems to be one of the most straightforward strategies to prevent S. aureus infections (van Rijen & Kluytmans, 2008), and interest in understanding microbial community composition and species interactions in the anterior nares has recently increased.

While recent works (Costello et al., 2009; Grice et al., 2009; Lemon et al., 2010; Wos-Oxley et al., 2010) characterized the nasal communities of 7, 8, 7 and 40 healthy adult volunteers, respectively, these numbers do not cover a major proportion of the human population and thus do not fully reflect the nasal bacterial community across a large cross-section of the human community. These reports sought to characterize the nares using culture-independent methods such as 16S rRNA gene clone library construction and analysis (Grice et al., 2009; Lemon et al., 2010), hybridization of pooled 16S rRNA gene amplicons to the Affymetrix PhyloChip (Lemon et al., 2010), single-strand conformation polymorphism (SSCP) (Wos-Oxley et al., 2010) and 454-pyrosequencing (Costello et al., 2009; Wos-Oxley et al., 2010), resulting in both the elucidation of novel associations between species and in-depth characterization of the nares microbiome.

Terminal restriction fragment length polymorphism (T-RFLP) was originally developed to identify mycobacteria (Avaniss-Aghajani et al., 1996) and has since been commonly used for comparing differences in the microbial communities of a diverse range of environments (Osborn et al., 2000; Marsh & Jared, 2005; Tiquia, 2009; Joo et al., 2010; Nakano et al., 2010). The generated terminal restriction fragment abundances (peaks corresponding to species or groups of related species) are presented as the relative proportion to the total abundance of the sample, as do these other aforementioned profiling methods. Most importantly, T-RFLP is highly reproducible, giving excellent resolution, and is also a robust and high-throughput method, justifying its current use across a diverse range of environments. It can handle the analysis of hundreds of samples simultaneously (like 454-pyrosequencing) but without the price tag and laborious computational work of sequence data.

We reasoned that for a relatively simple community such as that present in the anterior nares, T-RFLP will be capable of differentiating the major community members if appropriate restriction enzymes can be chosen. Thus, we optimized the profiling approach of T-RFLP using 16S rRNA genes of species-of-interest and then validated it on six nasal samples from healthy adult volunteers against data previously collected using 454-pyrosequencing. We then used the newly validated method to further analyse 100 anterior nares samples to characterize the bacterial community composition across a greater cross-section of the human population, to further elucidate patterns both within the bacterial communities and across human communities and to expand our current knowledge of the microbial ecology of the nares.

Materials and methods

In silico terminal fragment length determination and selection of restriction enzymes

The ability of restriction enzymes to discriminate between species-of-interest was evaluated using the online program REPK (Collins & Rocap, 2007). The species-of-interest are the set of species constituting the core nasal community as previously proposed, comprising Corynebacterium accolens, S. epidermidis, Propionibacterium acnes, D. pigrum, F. magna, Peptoniphilus sp., uncultured Actinomycetales, S. aureus, Staphylococcus capitis, Moraxella spp. and Anaerococcus spp. (Table ) (Costello et al., 2009; Wos-Oxley et al., 2010). In silico analysis using previously reported partial 16S rRNA gene sequences (Wos-Oxley et al., 2010) and those of type strains revealed that the combination of AseI, TspRI and ApekI restriction enzymes consecutively applied to digest the PCR amplicons could discriminate between this list of species-of-interest. DNA extracts of some volunteers were used as template in PCR amplifications with primers 27F and 1389R (Lane, 1991; Osborn et al., 2000) (Supporting Information, Table S1) at an annealing temperature of 55 °C. PCR products were purified and cloned into the pGEM-T easy vector system following manufacturer's instructions. Randomly selected clones were sequenced on an ABI 3130xl Genetic Analyzer (Applied Biosystems, Weiterstadt, Germany) using Big-Dye™ Terminator Cycle Sequencing Reaction kit (Applied Biosystems). Sequences of clones containing 16S rRNA gene fragments derived from S. epidermidis, Anaerococcus sp., F. magna, Moraxella sp., D. pigrum, uncultured Actinomycetales bacterium, P. acnes and Peptoniphilus sp., which were used for validation purposes, were deposited in GenBank under the accession numbers JF927883JF927890. The eight clones selected above, a clone containing a 16S rRNA gene fragment derived from C. accolens available from a previous study (GU074966) and five type strains (as listed in Table ), all representing the species-of-interest, were subjected to a consecutive digestion with AseI, TspRI and ApekI.

View this table:

Predicted and observed terminal restriction fragments (T-RFs) in bp after sequential digestion with AseI, TspRI and ApekI

Source of 16S rRNA genePredicted T-RFObserved T-RFs
S. epidermidis (JF927883)318317.7
P. acnes (JF927889)18165.5; 180.9
D. pigrum (JF927887)7371.5; 77.6
F. magna (JF927885)150151.0
Peptoniphilus sp. (JF927890)175173.9
Unc. Actinomycetales (JF927888)341183.6
S. aureus (DSM3463)238238.8
S. capitis (DSM20326)201200.2
Moraxella spp. (JF927886)300301.1
Anaerococcus spp. (JF927884)357157.4; 355.0; 364.9
C. accolens (DSM44278)336116.0; 338.1; 348.0
C. accolens (GU074966)336116.0; 338.1; 348.1;
C. propinquum (DSM44285)33678.7; 114.8; 336.7; 346.7
C. pseudodiphtheriticum (DSM44287)33678.9; 114.8; 336.8; 346.8
  • Each bold T-RF indicates that this is the predominant peak generated from that strain/species. Unbold peaks are additional secondary peaks produced by some species.

  • Detected in clone library analysis.

  • These observed T-RFs have not been binned (as described in the for those T-RFs of the 100 samples).

Study population

The study population comprised 100 healthy volunteers (69 women and 31 men), aged 8–75 years from Münster, Germany. Each volunteer provided information regarding their age, gender, smoking habits and pet ownership (Table S2). Samples were obtained from the anterior nares using dry sterile cotton swabs, which were stored on dry ice and transported from Münster to Braunschweig (Germany) within 24 h and DNA-extracted immediately upon arrival. All volunteers were screened for nasal colonization of S. aureus, as previously described (Becker et al., 2006). Informed consent was obtained from all volunteers included in this study.

DNA extraction

DNA was extracted from the swab using the FastDNA Spin Kit for Soil (MP Biomedicals, Solon, OH). The swabs were placed in Lysing Matrix E tubes (MP Biomedicals) containing MT buffer and sodium phosphate buffer (provided by the manufacturer), and cells were lysed in a Fast Prep®-24 instrument for 30 s at an intensity of 6.0. DNA was eluted in 50 μL of DES and quantified using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Waltham, MA).

Terminal restriction fragment length polymorphism (T-RFLP)

PCR amplification was performed using a 6-carboxyfluorescein (6FAM)-labelled forward primer (27F) and Yakima Yellow–labelled reverse primer (1078R) (Table S1). Reaction mixtures (50 μL) contained 1× PCR buffer, each deoxynucleoside triphosphate at a concentration of 2.5 mM, each primer at a concentration of 0.2 μM, 1 μL of template DNA and 0.3 μL of GoTaq polymerase (5 U). An initial denaturation step of 95 °C for 5 min was followed by 35 cycles of denaturation at 95 °C for 1 min, annealing at 55 °C for 1 min and extension at 72 °C for 2 min, with a final extension step at 72 °C for 8 min. Amplification products were verified by agarose gel electrophoresis, purified using Macherey-Nagel 96-well plate purification kits (Macherey-Nagel, Düren, Germany) following manufacturer's instructions and quantified in a NanoDrop 2000 spectrophotometer. In a total volume of 40 μL, 10 U of exo-Klenow was incubated with 80 ng of amplicons in NEB Buffer 2 (New England Biolabs, Ipswich, MA) for 1 h at 20 °C followed by heat inactivation for 20 min at 75 °C. The exo-Klenow-treated amplicons (in NEB Buffer 2) were directly digested consecutively with AseI (2.5 U for 2.5 h at 37 °C), TspRI (2.5 U where 1× bovine serum albumin (100 μg mL−1) was added for 2.5 h at 65 °C) and ApekI (2.5 U for 2.5 h at 75 °C), without any purification steps in between. All the enzymes were purchased from New England Biolabs. The digested amplicons were purified using DyeEx gel filters (Qiagen), according to manufacturer's instructions. Triplicate aliquots containing 20 ng of the digested, purified amplicons were dried in a speed-vac (Eppendorf Concentrator 5301). The dried pellets were resuspended in 9.75 μL Hi-Di formamide (Applied Biosystems) with 0.25 μL of size standard GeneScan 500 LIZ (Applied Biosystems), denaturated at 95 °C for 3 min and immediately placed on ice. The lengths of the terminal restriction fragments (T-RF) were determined on an ABI 3130xl Genetic Analyzer (Applied Biosystems), and GeneMapper v3.7 software (Applied Biosystems) was used to perform peak size calling. Any fragment of a size < 50 nt was excluded from further analysis. Each fragment's abundance was standardized as a percentage of the total fragment area within each sample. The peaks of two representative replicates for all 100 volunteers were aligned and then binned if the difference was smaller than 0.4 bp in length, using the T-align program (http://inismor.ucd.ie/~talign) (Smith et al., 2005). Multiple peaks that correspond to the same taxon can be summed together to represent the total abundance of that taxon within a sample.

Real-time quantitative PCR

Real-time quantitative PCR (RTQ-PCR) was used to detect S. aureus in anterior nares communities by targeting the nuc gene that encodes the thermostable nuclease of S. aureus. Previously described primers (Table S1) were used to amplify a 269-bp fragment of the nuc gene (Brakstad et al., 1992). Each reaction was performed in duplicate in a final volume of 20 μL containing 2.5 μL of each primer (10 pmol), 10 μL QuantiTect SYBR Green PCR Master Mix (Qiagen) and 5 μL of DNA template (1/10 dilution of original extract). Amplification was carried out in a LightCycler® 480 Real-Time PCR System programmed to hold at 95 °C for 10 min and to complete 50 cycles of 94 °C for 15 s, 57 °C for 30 s and 72 °C for 30 s. The PCR results were analysed by the LightCycler® 480 Software (Roche Applied Science). Standard curves were generated from serial dilutions of known concentrations of S. aureus genomic DNA (DSM3463) (containing between 1.7 and 17 000 copies of the nuc gene μL−1) by plotting threshold cycles (CT values) vs. copy number, assuming that 1 ng of DNA contains 6 × 105 copies of the entire 2.7- to 2.8-mbp-sized S. aureus genome, where the nuc gene is present as a single copy (Hein et al., 2001).

Data analysis

Statistical analyses involving nonparametric multivariate data sets were carried out using primer (v.6.1.6, PRIMER-E; Plymouth Marine Laboratory, UK). All following routines were computed on standardized abundance data. A sample-resemblance matrix was generated using the Bray–Curtis coefficient (Bray & Curtis, 1957), and the bacterial community structures were explored by both ordination using nonmetric multidimensional scaling (nMDS) (50 random restarts) and clustering using group-average hierarchical clustering. Then, either bubbles or clusters were superimposed over the nMDS plot, where bubbles show the relative abundance of a species-of-interest in each volunteer's anterior nares community and clusters represent the percentage similarity within a group of samples (arbitrarily chosen from the hierarchical cluster dendrogram). The bias of arbitrarily choosing a particular similarity level to group samples can be eliminated using the similarity profile permutation test (SIMPROF), which seeks for statistically significant evidence of genuine clusters within communities in the absence of a priori groups (Clarke et al., 2008). This method was used to delineate groups of volunteers with distinct nasal community structures at the 99% significance level (using 999 simulations to generate the dendrogram and 1000 permutations to generate a permutation distribution for comparison with the observed test statistic).

Significant differences in the anterior nares communities between a priori predefined groups of samples (such as gender, smoking habits, age and pet ownership) were evaluated using analysis of similarity (anosim) (999 permutations). Groups of samples (nasal communities) were considered significantly different if the P-value falls < 0.05. The accompanying R statistic measures the degree of separation between groups and ranges from −1 to 1, in which the higher its value (closer to 1), the more distinct the groups (Clarke & Warwick, 2001).

A Mantel-type correlation using the RELATE routine in Primer 6 (Clarke & Warwick, 2001) was used to quantify the pattern match between the two resemblance matrices generated from the T-RFLP and 454-pyrosequencing data matrices. The correlation using the Spearman coefficient (ρs) and 9999 permutations was between Bray–Curtis association measures, which were also used to generate nMDS ordinations for visual comparison of both the T-RFLP and 454-pyrosequencing data sets.

Differences in the abundance of species-of-interest between predefined groups were evaluated using the unpaired Welch's t-test that can handle unequal variances, unequal sample sizes and nonparametric data (Welch, 1947) using Prism 5 (GraphPad Software, CA). Groups were considered significantly different if the P-value falls < 0.05.

Species–species association patterns were explored using network of bacterial ecological co-occurrence analysis (Freilich et al., 2010) where the degree to which pairs of species are distributed across all 100 samples, taking into consideration their proportional abundances, was assessed. A species-resemblance matrix was generated using the Bray–Curtis coefficient and the corresponding values used to measure the occurrence of a pair of species sharing the same niche in relative proportions. A network of bacterial ecological co-occurrences was constructed using the freely available program ‘neato’ from the Graphviz software (http://www.graphviz.org/) (Gansner et al., 2005). Bray–Curtis values were first transformed into inversely proportional numbers to represent distances between nodes, because high Bray–Curtis values indicate higher likelihood that pairs of species aggregate together in relative proportions, thus corresponding to smaller distances between nodes in the network. The neato algorithm uses a global energy minimization function to set the node layout. The network is a depiction of the likelihood that core species (nodes) of the anterior nares share the same niche (using the experimental T-RFLP data from 100 nares). The edges represent the strongest 5% positive and the strongest 5% negative associations of all possible pairs.

Results

T-RFLP validation

Different restriction enzyme combinations targeting 16S rRNA gene amplicons can be used in T-RFLP analysis to discriminate between species-of-interest. To use T-RFLP to characterize the bacterial community of the anterior nares, we evaluated new enzyme combinations to resolve major species known to be present in the anterior nares, including S. aureus, Corynebacterium spp., Moraxella spp., Peptoniphilus sp., F. magna, D. pigrum, P. acnes, S. capitis, S. epidermidis and Anaerococcus spp. In silico analysis of 16S rRNA gene amplicons, generated using the 27F forward primer, indicated that of the 189 commercially available enzymes in the database (Collins & Rocap, 2007), only ApeKI was capable of digesting amplicons generated from S. aureus (238 bp) differentially compared to those of other Staphylococcus spp. previously reported as anterior nares inhabitants. This enzyme has also the capacity to differentiate between 16S rRNA gene amplicons generated from P. acnes (181 bp), C. accolens (336 bp) and Anaerococcus spp. (357 bp), while AseI discriminates F. magna (150 bp) and Peptoniphilus sp. (175 bp), and TspRI distinguishes S. epidermidis (318 bp), S. capitis (201 bp), Moraxella spp. (300 bp) and D. pigrum (73 bp). Overall, in silico analysis suggested a combination of three restriction enzymes to be adequate to generate distinguishable T-RFs from most species previously identified as important anterior nares microbial community members (Table ).

To confirm in silico results, 16S rRNA gene amplicons from five type strains and from nine selected clones were subjected to a triple digestion, and their in vitro T-RF pattern was analyzed (Table ). Overall, the experimentally observed T-RFs did not deviate significantly from the predicted ones. The only discrepancy concerned the uncultured Actinomycetales where a fragment length of 341 bp was predicted but a length of 183.7 bp was observed, even though no appropriate restriction site could be observed in the respective gene region. In some cases, specifically during the analysis of T-RFs generated from Corynebacterium spp., shorter and longer restriction fragments were observed in addition to the true T-RF. This might be due to an overdigestion of the amplicons or pseudo-T-RFs (Egert et al., 2003, 20113); however, this pattern was always reproducible for all clones and type strains used.

To validate the triple digest as a rapid tool to characterize anterior nares microbial communities, the structure of the anterior nares communities of six healthy adults generated by T-RFLP was compared with that previously determined through 454-pyrosequencing (Wos-Oxley et al., 2010) (Figs S1 and S2). Not only were all dominant phylotypes detected by 454-pyrosequencing detected by the developed T-RFLP method, but also typically in a similar relative abundance (Fig. S2). Discrepancies between the two methods were only evident when phylotypes were in very low abundance. Despite these slight discrepancies in a couple of volunteers, this did not markedly change the deduced global bacterial community structure of the anterior nares of these six volunteers, where the two methods correlated strongly, yielding a rank correlation value (rho) of 0.864 (P-value = 0.0016) (as also visually illustrated in Fig. S1). This implies that both profiling methods produced highly similar profiles for each volunteer and conserved the global differences observed between the six volunteers.

T-RFLP application

The validated T-RFLP method was then applied to anterior nares samples of 100 healthy volunteers. T-RFs indicating the presence of S. epidermidis, P. acnes, uncultured Actinomycetales and Corynebacterium spp. were detected in more than 80% of the volunteers (in up to 26%, 40%, 12% and 27% relative abundance, respectively), while T-RFs demonstrating the presence of D. pigrum were prevalent in 53% of the volunteers (in up to 20% relative abundance). T-RFs indicating the presence of members of the Clostridiales family Incertae Sedis XI, in particular F. magna and Peptoniphilus sp., were present in 52% and 41% of the study population, respectively, but at a maximum relative abundance of 5%. This confirms the previous findings (Wos-Oxley et al., 2010) that this set of principal species constitutes the core nasal community. Staphylococcus aureus was observed in 26% of the volunteers (in up to 63% relative abundance). In addition to these identified species-of-interest, there were a further 14 dominant T-RFs that could not be immediately phylogenetically resolved against the principle list of representative clones (Table S3). That is, these unknown T-RFs could be novel species or could be related species/strains of the core community of the anterior nares, but with enough heterogeneity in the 16S rRNA gene that may generate slightly different fragment sizes. While we suspect that this is in fact the case, without representative clones that produce these T-RFs, it is speculative over which species these unknown peaks could represent.

T-RFLP indicated 26 of the volunteers as S. aureus carriers; however, in 11 volunteers that were identified as S. aureus carriers by culturing at the time of sampling, S. aureus could not be detected by T-RFLP. To understand the reason behind this fact, the limit of detection of S. aureus by T-RFLP was analyzed by RTQ-PCR targeting the nuc gene encoding the thermostable nuclease of S. aureus. Analysis of samples of the above-mentioned 11 volunteers and of a further six samples that were positive for S. aureus using both methods (but where the relative abundance of S. aureus as determined by T-RFLP differed greatly among the samples) revealed that T-RFLP was able to detect S. aureus when there were more than 30 copies μL−1 of S. aureus in the 50 μL DNA extract from the nasal swab (Table S4), which corresponds to about 1% of the community.

Global bacterial community

Exploring the global bacterial community structure of the 100 nares using ordination revealed a dense cluster where 68% of the volunteers share more than 40% similarity in respect to their bacterial structure (Fig. a). Eighty-five per cent of these volunteers shared T-RFs corresponding to Corynebacterium spp., P. acnes and uncultured Actinomycetales and S. epidermidis. Thus, they share a similar core nasal community, and the constituents of this core community were also present at similar abundances in respect to the overall community. Other important nasal community members such as Peptoniphilus sp. and F. magna were present in 50% and 72% of the volunteers in this dense cluster, respectively (Fig. e–i).

Nonmetric multidimensional scaling plot illustrating similarities in the global bacterial community structure of the anterior nares across 100 volunteers using T-RFLP data collected from extracted DNA. Numbers 1–100 correspond to individual volunteers. T-RFLP data were standardized (%) but untransformed prior to the use of the Bray–Curtis similarity algorithm. Using this ordination (a) superimposes clusters that represent those anterior nares global communities that share more than 40% similarity as determined by group-average hierarchical clustering using Bray–Curtis similarity, (b–i) superimposes bubbles to represent the relative abundance of Staphylococcus aureus, Dolosigranulum pigrum, Moraxella spp., Peptoniphilus sp., Staphylococcus epidermidis, Corynebacterium accolens, Propionibacterium acnes, Finegoldia magna, respectively. While a 2D stress value of 0.21 indicates some stress on the plot, it is deemed acceptable considering that 100 samples are being ordinated together.

The remaining 32 volunteers clustered into smaller groups at 40% similarity. In particular, three smaller groups of volunteers were highly colonized with either S. aureus (in relative abundances between 35% and 63%), D. pigrum (in relative abundances between 10% and 20%) or Moraxella spp. (in relative abundances between 35% and 69%) (Fig. b–d, respectively). In volunteers that were colonized by these organisms, the core bacterial community was typically still present but at low abundance.

As the choice of 40% similarity is arbitrary, SIMPROF was used to determine which volunteers might naturally group together based on their nares bacterial community. Following such an approach, the 100 volunteers naturally clustered into 12 groups where different characteristic patterns of species abundance were consistently found (Fig. , Table S5). The largest group comprised 27% of the volunteers, while another closely related group comprised 19% of volunteers. These two groups resemble the majority of the volunteers that belong to the 40% similarity group (Fig. a). The other 54% of volunteers were assembled into 10 further groups, each comprising between 2% and 9% of volunteers. This does not imply that groups of people have no species in common, but merely that different characteristic patterns of relative abundance are found consistently in different groups. The dominant groups (named as D, G in Table S5) all comprised those core community members but with slightly different characteristic patterns of abundance of certain species-of-interest, while the other groups of volunteers were inhabited with less-prevalent species-of-interest.

Nonmetric multidimensional scaling plot (as shown in Fig. a) that illustrates those nares that group together into statistically determined discrete clusters as determined by (SIMPROF) where each symbol represents a distinct group (Table S5).

Incidentally, the groups being highly colonized by Moraxella spp. (I and J in Table S5) comprised only women. Overall, of the 100 volunteers, the women (n = 13) seemingly carried a greater abundance of Moraxella spp. than the few men (n = 5) that were colonized by this species, an observation being statistically significant (Welch's t-test, P = 0.012). No other similar trends were observed with any of the other species-of-interest.

It had been hypothesized that specific external factors relating to each individual volunteer may have an influence over their nasal bacterial community. To test whether this was the case, each volunteer provided information pertaining to their age, gender, smoking habits and ownership of pets. anosim was used to test for significant differences between the global bacterial communities of the 100 volunteers in respect to these individual factors. Concerning the global bacterial profiles, there were no significant differences (P > 0.05) between men and women, smokers and nonsmokers, pet ownership or different age groups. Also, the accompanying R-statistics were close to 0, indicating that these groups were indistinguishable.

Network of bacterial ecological co-occurrences

Besides host and environmental factors, species–species interactions are important for shaping bacterial community structure. Species–species association patterns were explored using network of bacterial ecological co-occurrence analysis (Freilich et al., 2010) where the proximity of the nodes on the network demonstrates groups of species that are more likely to aggregate together in relative abundances (Fig. ). The species network revealed likely positive associations between (1) S. epidermidis, C. accolens, P. acnes, members of Actinomycetales and Anaerococcus spp., (2) F. magna and Peptoniphilus sp. and (3) D. pigrum, Corynebacterium propinquum and Corynebacterium pseudodiphtheriticum. However, many species do not seem to share the same niche, or if they do, they are abundant in opposing concentrations. In particular, Moraxella spp. had a strong negative association with many other taxa such as Peptoniphilus sp. and S. capitis.

Visualization of species–species associations using network analysis charting bacterial ecological co-occurrences based on the Bray–Curtis similarity algorithm. Nodes represent the species-of-interest, and edges represent associations. Solid edges represent the 5% strongest positive associations, while dashed edges represent the 5% strongest negative associations. Abbreviations in the nodes represent each species-of-interest: DOLO, Dolosigranulum pigrum; FMAG, Finegoldia magna; PEPT, Peptoniphilus sp.; PACN, Propionibacterium acnes; UACT, uncultured Actinomycetales; SCAP, Staphylococcus capitis; SAUR, Staphylococcus aureus; MORA, Moraxella spp.; SEPD, Staphylococcus epidermidis; CPRP, Corynebacterium pseudodiphtheriticum and Corynebacterium propinquum; CACO, Corynebacterium accolens; ANAE, Anaerococcus spp., while UP01-UP14 represent those species-of-interest that have yet to be phylogenetically identified.

Discussion

Molecular profiling of bacterial communities by T-RFLP generates highly reproducible community fingerprints, which allow rapid characterization of the community composition and overall diversity for a large number of samples (Ruan et al., 2006). It has been described in former works as a significant method for bacterial characterization in diverse sources of samples (Osborn et al., 2000; Marsh & Jared, 2005; Tiquia, 2009; Joo et al., 2010; Nakano et al., 2010), and it was suggested that plausible community compositions may be derived by matching T-RFs in a profile with T-RF sizes derived in silico from the 16S rRNA gene sequences of phylotypes in databases (Dekio et al., 2007; Schütte et al., 2008). However, owing to discrepancies between true and observed T-RFs that can be caused by the purine content (Kaplan & Kitts, 2003) and the formation of pseudo-T-RFs (Egert & Friedrich, 2003; Egert & Friedrich, 2005), a careful validation is necessary before community compositions may be determined. Unknown T-RFs were observed in this work, which could be either novel species or related species/strains of the core community of the anterior nares. Future work should be directed at identifying those species generating these T-RFs through deep sequencing. Thus, the T-RFLP analysis presented here can be regarded as an iterative process.

T-RFLP was capable of specifically detecting S. aureus in complex communities and thus identifying S. aureus carriers. However, as with the limitations of all fingerprinting methods, S. aureus carriers could only be identified if this organism was in a proportion in the community higher than 1%. As typically 20–30% of the human population are characterised as permanent carriers of S. aureus (Williams, 1963; von Eiff et al., 2001; van Belkum et al., 2009), it would be valuable to assess whether such permanent carriers are typically colonized by a high proportion of S. aureus.

Previous studies characterized only a limited number of anterior nares (Costello et al., 2009; Grice et al., 2009; Lemon et al., 2010; Wos-Oxley et al., 2010) and not across a significant proportion of the human population. This work analyzed the bacterial community structure of the anterior nares of 100 independent volunteers, which statistically clustered into 12 groups, where two-thirds of the volunteers shared more than 40% similarity in respect to their bacterial community structure, while the remaining third clustered together into smaller groups being dominated by one of the three species, D. pigrum, Moraxella spp. or S. aureus. This shows that the habitat of the human anterior nares is not merely a random liberty of species, but unrelated people may share distinct attributes of their nasal community. The use of a network that charts bacterial ecological co-occurrences revealed new evidence of likely positive associations between some core nasal species. Within the previously identified core community, S. epidermidis, C. accolens, P. acnes, members of the order Actinomycetales and the genus Anaerococcus are more likely to share the same niche and at similar abundances. In addition, F. magna and Peptoniphilus sp. on the one hand and D. pigrum and C. propinquum/C. pseudodiphtheriticum on the other hand were positively associated together. In particular, Moraxella spp. had a strong negative association with many other of the taxa such as Peptoniphilus sp. and S. capitis.

Interestingly, Moraxella spp. were more prevalent and abundant in women, suggesting that gender and thus factors relating to gender may play a role in species distribution. At this point, we have no evidence as to why this may be the case. Previous studies have shown the presence of organisms related to Moraxella lacunata and/or Moraxella nonliquefaciens in the anterior nares (Devi et al., 1991; Wos-Oxley et al., 2010), which usually are of low virulence and have most commonly been associated with eye infections, pneumonitis and pulmonary abscess, and occasionally upper respiratory tract infections (Rosett et al., 1976; Laukeland et al., 2002; Woodbury et al., 2009). Among the first reports, M. lacunata has been described in 11% of women suffering from epidemic follicular conjunctivitis (Ringvold et al., 1985). Further work requires sampling across a greater number of both female and male volunteers to definitively prove that higher colonization of women is in fact not just a coincidence and if so as to the reason for such a distribution. Moreover, such work should also attempt to better discriminate closely related Moraxella species such as M. lacunata and M. nonliquefaciens (Pettersson et al., 1998). It has also been suggested that independent of the immune status of the host, Moraxella spp. may be more invasive pathogens than previously thought (Piontek & Herrmann-Czylwik, 1993; Woodbury et al., 2009).

While differences in colonization by S. aureus have been attributed to host factors such as host immunity, age, gender and/or environmental factors (García-Rodríguez & Fresnadillo Martínez, 2002), the data generated from these 100 anterior nares samples could not reveal any relationship between host factors and bacterial structure or colonization of a particular species. Nevertheless, sampling a greater cross-section of the community such as those generated with cohort studies in conjunction with detailed host data, and together with more formal investigations into the role of host/environment factors on the bacterial structure, can provide a wider scope of the interactions between these elements. As it is deemed that the nose plays an important role in invasive infections because of its role as reservoir of important opportunistic pathogens such as S. aureus among others, knowledge on community interactions will generate valuable information about the potential control of these systems. It has been postulated that a tailored antibiotic regime or treatment prior to medical intervention poses an important strategy to prevent later infections (van Rijen & Kluytmans, 2008). However, while the application of mupirocin to the anterior nares has shown to eradicate nasal S. aureus carriage in approximately 80% of surgical patients (van Rijen et al., 2008), S. aureus and even methicillin-resistant S. aureus strains have developed resistance to this antibiotic (Mody et al., 2003; Hurdle et al., 2005), indicating that alternative strategies to combat S. aureus carriage are urgently needed.

Overall, this work uncovered new patterns of species distribution in the nares and the roles that these play in controlling the whole community. The T-RFLP procedure described here should not be overlooked as an important profiling method, in case a study requires the phylogenetic high-throughput community profiling of hundreds of samples. So, in the age of post ‘omics’ and ‘deep sequencing’, there is still a place for these tried-and-tested methods that can offer a rapid, reproducible and economical alternative, whereby also yielding valuable new information and thus expanding our current knowledge on the microbial ecology of the nares.

Supporting Information

Additional Supporting Information may be found in the online version of the article:

Fig. S1. Non-metric multidimensional scaling plots illustrating similarities in the global bacterial community structure of the anterior nares of six volunteers as analyzed by T-RFLP (denoted as T) compared to 454-pyrosequencing data (denoted as P) (volunteers 3, 10, 13, 15, 17, 18 as described by Wos-Oxley et al., 2010).

Fig. S2. Comparison of the % abundance of species-of-interest in the global bacterial community structure of the anterior nares of six volunteers as analyzed by T-RFLP (black) compared to 454-pyrosequencing data (grey) (volunteers 3, 10, 13, 15, 17, 18 as described by Wos-Oxley et al., 2010).

Table S1. Primers used in this study.

Table S2. Characteristics (gender, age, smoking habits, ownership of pets) of each of the 100 volunteers used in this study.

Table S3. Observed terminal restriction fragments (in bp) after sequential digestion with AseI, TspRI and ApekI in samples collected from 100 volunteers.

Table S4. Limit of detection and quantification of S. aureus (in those volunteers that are S. aureus carriers) using RTQ-PCR.

Table S5. Summary of those groups determined by SIMPROF and their characteristic profiles.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

Acknowledgements

The authors would like to thank Iris Plumeier, Silke Kahl and Martina Schulte for technical support. This work was funded by the BMBF project ‘Susceptibility of Infection: Skin Staph’ to D.H.P. (01KI1009B) and K.B. (01KI1009A) and ‘Medical Infection Genomics’ to D.H.P. (0315832B) and K.B. (0315832A) and in part by the Helmholtz/National prospective cohort study.

Footnotes

  • Editor: Tillmann Lueders

References

View Abstract