A large genomic island allows Neisseria meningitidis to utilize propionic acid, with implications for colonization of the human nasopharynx

Neisseria meningitidis is an important human pathogen that is capable of killing within hours of infection. Its normal habitat is the nasopharynx of adult humans. Here we identify a genomic island (the prp gene cluster) in N. meningitidis that enables this species to utilize propionic acid as a supplementary carbon source during growth, particularly under nutrient poor growth conditions. The prp gene cluster encodes enzymes for a methylcitrate cycle. Novel aspects of the methylcitrate cycle in N. meningitidis include a propionate kinase which was purified and characterized, and a putative propionate transporter. This genomic island is absent from the close relative of N. meningitidis, the commensal Neisseria lactamica, which chiefly colonizes infants not adults. We reason that the possession of the prp genes provides a metabolic advantage to N. meningitidis in the adult oral cavity, which is rich in propionic acid-generating bacteria. Data from classical microbiological and sequence-based microbiome studies provide several lines of supporting evidence that N. meningitidis colonization is correlated with propionic acid generating bacteria, with a strong correlation between prp-containing Neisseria and propionic acid generating bacteria from the genus Porphyromonas, and that this may explain adolescent/adult colonization by N. meningitidis.


Introduction
Bacterial pathogens, such as Neisseria meningitidis (a.k.a. the meningococcus), are adapted to suit the host environ-ment. In addition to the host-pathogen interaction, the microbe is part of a diverse and complex microbiota, in which it cooperates and competes for resources with its neighbours. N. meningitidis is notorious for its ability to cause severe sepsis and/or meningitis within a matter of hours following its acquisition and the movement of the bacterium into the bloodstream (van Deuren et al., 2000). However, disease is a rare outcome of N. meningitidis acquisition. Typically the bacterium lives harmlessly in the human nasopharynx of 5-40% of the population at any given time (Cartwright et al., 1987;Caugant et al., 1994;Sim et al., 2000;Yazdankhah and Caugant, 2004). N. meningitidis is most closely related to the sexually transmitted pathogen Neisseria gonorrhoeae, but other members of the Neisseria genus are non-pathogens. Neisseria lactamica is the non-pathogen that is most closely related to the pathogenic Neisseria species, and is of particular clinical relevance since it contributes to the development of immunity to meningococcal disease in childhood (Gold et al., 1978) and following experimental inoculation with N. lactamica in adults (Evans et al., 2011). N. meningitidis and N. lactamica both colonize the human nasopharynx as their natural habitat, but N. lactamica colonization is largely confined to infants, whereas N. meningitidis colonization increases with age from birth, reaching a peak in late teenage and declining in older adults (Gold et al., 1978;Cartwright et al., 1987;Olsen et al., 1991;Christensen et al., 2010). The incidence of meningococcal disease is highest among infants (with a naïve immune system) but a secondary age related peak in disease is seen among adolescents and young adults, which correlates with meningococcal carriage frequency (Jones and Mallard, 1993). Ultimately, the reservoir of adult meningococcal colonization is crucial for the occurrence of epidemic meningococcal disease (in both adults and children), and thus it is important to understand what enables N. meningitidis, but not N. lactamica, to effectively colonize adults.
An analysis of the genomic differences between N. lactamica and the two pathogenic Neisseria species reveals nine genomic islands that are found in all N. meningitidis strains, but are absent from all four N. lactamica strains currently available in NCBI database (ST640, N5195, ATCC23970 and Y92-1009) (Fig. 1). The largest of these DNA regions is a 10200 base pair region which contains genes NMB0430-NMB0435 (of the serogroup B strain N. meningitidis MC58). This region is also conserved in all N. gonorrhoeae strains. Homologous DNA sequences flanking this region (NMB0428 and NMB0436) are adjacent to one another (separated by less than 100 base pairs) in N. lactamica (Fig. 1) and other closely related Neisseria species N. polysaccharea and N. cinerea indicating that the NMB0430-0435 gene cluster is a recent horizontally transferred acquisition. More distantly related Neisseria species (for a Neisseria genus phylogeny, see e.g. Marri et al., 2010), N. subflava, N. flavescens, N. macaceae, N. sicca and N. mucosa contain homologues of the genes from this island, bearing very high levels of identity (typically 97-99% identical), but the flanking genes are not the same as found in N. meningitidis. All this is consistent with the NMB0430-0435 gene cluster being a horizontally transferred acquisition either from more distantly related members of the Neisseria genus or from outside the genus.
Here we demonstrate for the first time that N. meningitidis can utilize propionic acid as a carbon source, and that this is dependent on possession of the NMB0430-0435 gene cluster. We propose that the ability of N. meningitidis to utilize propionic acid provides it with a selective advantage in the adult nasopharynx which is rich in anaerobes, many of which produce propionic acid as an endproduct of fermentation.

Results and discussion
The prp gene cluster enables N. meningitidis to utilize propionic acid as a carbon source To test whether the NMB0430-NMB0435 gene cluster (henceforth referred to as the prp gene cluster) is functional in N. meningitidis, we introduced an antibiotic resistance cassette into NMB0431 (prpC, the putative methylcitrate synthase). prpC does not contribute to increased growth or propionic acid resistance in N. meningitidis grown in rich media ( Fig. 3A and B). However, in chemically defined media with glucose or pyruvate as carbon source, propionic acid supplements growth of N. meningitidis ( Fig. 3C and E). The extended growth in late exponential phase induced by propionic acid is concomitant with propionic acid depletion, and is dependent on prpC ( Fig. 3C-F).
The expression of prpC is highly regulated during growth in different media (Fig. 4). Expression of prpC is considerably lower in Mueller-Hinton broth (rich media) and 10 mM propionic acid (filled inverted triangles). Propionic acid content was followed for cultures inoculated with 5 mM propionic acid and is shown as grey squares. Data are means ± standard deviation for three independent experiments.

Fig. 4. Relative expression of prpC over growth time-course in
Mueller-Hinton broth (white bars), chemically defined medium with glucose as carbon source (grey bars) and chemically defined medium with pyruvate as carbon source (black bars). Expression was assessed using quantitative real-time PCR. Expression levels are shown for cultures ± 5 mM propionic acid. Data are means ± standard deviation for at least three independent experiments. than in chemically defined media with glucose or pyruvate as the major carbon source. prp gene expression is not induced directly by propionic acid (Fig. 4). Expression of prpC increases towards stationary phase in chemically defined medium with either glucose or pyruvate as carbon source, consistent with its induction being responsive to nutrient poverty. NMB0431 has previously been shown to be regulated by a global regulator of nutrient limitation (NMB0573; Lrp) in N. meningitidis (Ren et al., 2007). The prp gene cluster has also been shown to be regulated by a small RNA (Bsn1) (Del Tordello et al., 2012), although whether this regulation is related to nutrient deprivation and/or Lrp is not currently known. Propionic acid utilization begins earlier in the growth of N. meningitidis in pyruvate medium, compared to glucose medium, corresponding with a high level of expression of prpC after 4 or 6 h of culturing, whereas propionic acid utilization in glucose media occurs rapidly, but only after around 8 h of incubation, consistent with a sharp increase in prp expression after this time (Figs 3 and 4). Overall, the transcriptional regulation of the prp gene cluster is sufficient to explain the different propionic acid utilization patterns in different media.
Genes in the prp cluster NMB0430, NMB0433 and NMB0434 are highly similar to genes required for methylcitrate cycle action in other bacteria, but NMB0432 and NMB0435 are not similar to methylcitrate cycle genes from other organisms. To determine whether these genes are involved in propionic acid utilization in N. meningitidis we constructed gene knockouts in both these genes of unknown function. Mutants deficient in NMB0432 and NMB0435 were constructed by insertion of a spectinomycin resistance cassette into these open reading frames. Mutants deficient in NMB0432 or NMB0435 are unable to utilize propionic acid, and grow identically to the wild-type in the absence of propionic acid (Fig. 5) indicating that both these genes are required for propionic acid utilization in N. meningitidis. The spectinomycin resistance cassette used for generation of the knockout mutations contains powerful transcriptional terminators (Prentki and Krisch, 1984), and is anticipated to have an effect on transcription of genes downstream. Therefore, a mutant was also constructed in which NMB0432 was disrupted with a tetracycline resistance cassette lacking transcriptional terminators. Expression of genes NMB0433 and NMB0434, immediately downstream from NMB0432, was retained in this tetracycline-resistant mutant (albeit at lower levels than in wild-type or NMB0435-deficient strains), but was severely downregulated in the NMB0432::Spec r construct (Fig. 6). This finding and further experimental evidence (Fig. S1) indicate that the prp genes are expressed in a single operon. Strain NMB0432::tet r was unable to utilize propionic acid, and grew identically to the wild-type in the absence of propionic acid (Fig. 5) indicating that NMB0432 is required for propionic acid utilization in N. meningitidis.

NMB0435 encodes a propionate kinase
The structure of the prp gene cluster in N. meningitidis is distinct from that of other bacteria that contain a methylcitrate cycle in two unique respects. First, the gene NMB0435, which is annotated as acetate kinase ackA1, is not present in other bacteria. N. meningitidis MC58 encodes two putative acetate kinases: ackA1 (NMB0435) and ackA2 (NMB1518). To explain the presence of the ackA1 gene in the prp gene cluster and to explain the phenotype of the NMB0435 mutant (unable to utilize pro- pionic acid) we hypothesized that ackA1 may in fact encode a propionate kinase. ackA2 is presumed to encode a genuine acetate kinase in N. meningitidis. To test this, we overexpressed NMB0435 and NMB1518 in E. coli and assayed their activities with acetate and propionate as substrates in kinase assays (Table 1; Fig. S2). NMB1518 has acetate kinase activity, but it was not possible to measure an affinity of this enzyme for propionate since the data did not approach saturation, indicating that this is the bona fide acetate kinase in N. meningitidis. NMB0435, however, is active with both acetate and propionate as substrates, indicating it is capable of fulfilling a role in activation of propionate into propionyl phosphate as part of the methyl citrate cycle in N. meningitidis. The K M of NMB0435 for propionate (16 mM) is higher, but within an order of magnitude, of that from the only other characterized propionate kinase: TdcD from Salmonella (KM = 2.3 mM) (Simanshu et al., 2005). Preparations of NMB0435 have a low activity in vitro compared to NMB1518 (Fig. S2) and other propionate kinase characterized previously (Simanshu et al., 2005). This probably relates to the large number of cysteine residues (eight) in NMB0435 and the high concentrations of reducing agent (5 mM DTT) required for purification of active NMB0435. Typically, propionate utilizing bacteria use propionyl-CoA synthase (PrpE) (Horswill and Escalante-Semerena, 1999a) to activate propionate and provide the substrate for methylcitrate synthase (PrpC). N. meningitidis generates propionyl phosphate via NMB0435, and we hypothesize that this is converted to propionyl-CoA via phosphotransacetylase (NMB0631; Pta). Indeed, NMB0631-deficient mutants are unable to utilize propionic acid as an additional carbon source during growth (Fig. 5).
The second innovation in the prp cluster in N. meningitidis is the gene NMB0432, homologues of which are not found in the methylcitrate cycle gene clusters in other organisms. It encodes a predicted membrane protein with six transmembrane spans. Mutant strains deficient in NMB0432 are unable to utilize propionic acid (despite expressing the other genes of the prp gene cluster) suggesting NMB0432 has a role in transporting propionate into the cell (although it is not possible to exclude the possibility that the phenotype on propionic acid is due to altered expression of downstream genes NMB0433-5). Pfam analysis indicates NMB0432 contains a domain (Pfam01925) similar to that of anion (sulphite) transporter TauE. Predicted membrane spans 4, 5 and 6 of NMB0432 each contain positively charged arginine or lysine residues, consistent with a role for this membrane protein in anion transport. Propionyl-CoA synthase PrpE has a much higher affinity for propionate (K M = 20 μM) (Horswill and Escalante-Semerena, 2002) than does propionate kinase AckA1 (KM = 16 mM), which may explain why an active transport system is required for propionate in N. meningitidis, but not in other propionate utilizing bacteria that possess PrpE. The KM for propionate of propionate kinase is higher than the concentration of propionate provided in the media (5 mM) and higher than the in vivo concentration of propionate in the oral cavity [up to 600 μM in saliva (Takeda et al., 2009); from 1 mM in healthy to 10 mM in diseased subgingival crevice (Niederman et al., 1997)].
prpC is not required for survival in blood NMB0431 was identified as a highly upregulated gene in response to culturing N. meningitidis in human blood (Echenique-Rivera et al., 2011). We investigated whether the possession of the prp cluster is important for survival in blood by culturing wild-type and NMB0431::spec r with whole human blood. There was no significant difference in the survival rates between the two strains indicating that Fig. 6. The relative expression of the NMB0432 (black bars), NMB0433 (pale grey bars) and NMB0434 (dark grey bars) in wild-type (WT), and mutant strains NMB0431::spec r (0431sp), NMB0432::tet r (0432tet), NMB0432::spec r (0432sp) and NMB0435::spec r (0435sp). Expression was assessed using quantitative real-time PCR. Expression was determined in cultures grown for 6 h in pyruvate medium with (denoted by +) or without (denoted by −) propionic acid. Data are expressed as relative expression compared to that of a wild-type culture grown for 6 h in MHB without propionic acid. N.D denotes not done. Error bars represent standard deviation and data are mean averages from at least three independent experiments. this pathway is not crucial for meningococcal survival in blood (Fig. S3).

A role for propionic acid metabolism in N. meningitidis colonization
The finding that the prp gene cluster possession does not affect survival in blood suggests that the function of this gene cluster may relate to survival in the normal habitat of N. meningitidis (i.e. the adult nasopharynx) rather than it having a role in virulence per se. While N. meningitidis is a common colonist of adults, N. lactamica is largely confined to infants (Gold et al., 1978;Cartwright et al., 1987;Olsen et al., 1991;Christensen et al., 2010). The finding that the methylcitrate cycle is not present in the infant-specific N. lactamica suggests that this metabolic pathway and propionate as a carbon/energy source might be important for enabling N. meningitidis to colonize older individuals. The main source of propionic acid in the oral cavity and naso-/ oro-pharynx is via strict anaerobes that generate propionic acid as an end-product of fermentation. Anaerobes are acquired from birth along with aerobes, but the oral (Kononen et al., 1999;Kononen, 2005) and nasopharyngeal (Kononen et al., 2003) microflora of infants is dominated by aerobes, and the proportion of anaerobes increases with age through childhood into adolescence. Propionibacteria (e.g. Propionibacterium acnes) are an important source of propionic acid (generated as a product of sugar fermentation) on the human body. The number of propionic acid bacteria colonizing the nares (nostrils) changes significantly with age, increasing c. 100-fold from 6 years to 13 years (Mourelatos et al., 2007). Recent culture-independent studies of microbial colonization of the nasopharynx and upper respiratory tract provide further evidence in support of a change in the microflora between infants and adults. A study of 96 infants (18 months old) revealed the dominant bacterial colonists of the nasopharynx are the aerobic and facultatively anaero-bic genera Moraxella (40%) and Haemophilus (20%) respectively (Bogaert et al., 2011). The adult nasopharynx has a more complex microbiota and is dominated by Firmicutes, Proteobacteria, Bacteroidetes and Fusobacteria (Charlson et al., 2010). Neisseria is more common in adults and clusters with colonization by anaerobes of the genera Porphyromonas and Fusobacterium (Charlson et al., 2010) both of which produce propionic acid as end-products of amino acid fermentation (Carlier et al., 1997;Takahashi et al., 2000). Notably, the proportion of the salivary microbiota belonging to genera Porphyromonas and Fusobacterium varies with age, such that there is a peak of colonization around age 18, correlating tightly with changes in nasopharyngeal meningococcal carriage with age (our analysis of data in Stahringer et al., 2012;Fig. 7). Positive correlations were observed between Neisseria and Porphyromonas colonization between individuals in the oral cavity in two separate studies of diversity in the upper respiratory tract involving 91 individuals (Human Microbiome Project, accession SRX055077) and 293 individuals (Charlson et al., 2010) and two separate studies of bacterial diversity in saliva involving 120 individuals (Nasidze et al., 2009) and 278 individuals (Stahringer et al., 2012) (Supplementary Fig. S4 and Datasets S1-S6  Cartwright et al. (1987); Porphyromonas and Fusobacterium data (B) are from Stahringer et al. (2012). The Neisseria data refer to the percentage of the population colonized by N. lactamica or N. meningitidis and are shown as bars based on age groupings 0-4, 5-9, 10-14, 15-19, 20-24 and 25-34. The Porphyromonas and Fusobacterium data refer to the percentage of total 16S rRNA reads that were identified as belonging to these genera among salivary samples from individuals aged 8, 12, 13, 17, 18 and 23.
providing a selection pressure for retention of a horizontally acquired metabolic pathway that allows utilization of propionate, as found in N. meningitidis.

Conclusion
In this article we provide the first report that N. meningitidis contains a functioning methylcitrate cycle and is able to utilize the enzymes in this cycle to catalyse the breakdown of propionate. The utilization of propionate helps support growth of N. meningitidis, particularly under conditions of nutrient limitation. The genes encoding the methylcitrate cycle are located on the largest genomic island in N. meningitidis, an island that is present in the other pathogenic Neisseria (N. gonorrhoeae) but absent from closely related non-pathogenic Neisseria species N. lactamica, N. cinerea and N. polysaccharea. Genomic islands in pathogens are typically referred to as pathogenicity islands. In this case, it appears that the prp gene cluster is not a pathogenicity island per se, but confers a metabolic capacity relevant to the natural habitat of the meningococcus. That the prp gene cluster is not central to virulence is in keeping with the fact that it appears to be universally maintained in N. meningitidis genomes, not just the hypervirulent lineages associated with most human disease. As such, the metabolic capacity to use propionic acid may be crucial in underpinning colonization of the N. meningitidis natural habitat (human adult upper respiratory tract), distinguishing it from the N. lactamica habitat (human infant upper respiratory tract). The capacity of N. meningitidis to colonize adolescents/ adults is important for its transmission and disease epidemiology (Christensen et al., 2013). This increase in carriage in young adulthood is frequently attributed to increased social interaction and contact in this age group (see e.g. Christensen et al., 2010). While this is no doubt true, here we present for the first time a mechanistic explanation for why N. meningitidis in particular has increased carriage with age, based on the genetic properties of the meningococcus and co-colonizing microbiota in the host.

Strains, media and growth conditions
Neisseria meningitidis strain MC58 was used as wild-type control and as source of genomic DNA for mutant strain constructs. Its complete genome has been sequenced, published and annotated (Tettelin et al., 2000). N. meningitidis strains NMB0431::Spec r , NMB0432::Spec r , NMB0432::Tet r and NMB0435::Spec r were generated by allelic exchange of an interposon disrupted version of the gene with the wild-type, using similar methods to those described previously (Anjum et al., 2002). Genes to be mutated were amplified by PCR. NMB0431 was amplified with primers 5′-CCAAGCTTG TGTCGAAGCC and 5′-TTTCAGACGGCCTTTCCAATAAGG. NMB0432 was amplified with primers 5′-TTCTCTGCC GTTTCCTACCAA and 5′-ATGTCGGTTCTCCTGTGGAT. NMB0435 was amplified with primers 5′-TTGACGTAGCAT GGGTTTGC and 5′-ACGCCCGAAATTCAAAATCC. In each case, the resultant product was cloned into pCR®-Blunt II-TOPO® vector (Invitrogen). The resultant plasmids were digested with PsiI (NMB0431), ClaI and AclI (NMB0432) and ClaI and BsmI (NMB0435) prior to insertion of the spectinomycin resistance gene from pHP45Ω (Prentki and Krisch, 1984) or the tetracycline resistance gene from pCMT18 (Heurlier et al., 2008). Resulting plasmids were transformed into N. meningitidis MC58 as described previously (Heurlier et al., 2008). N. meningitidis deficient in phosphotransacetylase (pta; NMB0631) was a kind gift from Dr Jon Shaw, University of Sheffield.
Neisseria meningitidis strains were streaked on Columbia Blood Agar Base plates (CBA) and incubated at 37°C overnight in a 5% CO2 atmosphere (Heurlier et al., 2008). Liquid cultures of N. meningitidis were grown in Mueller-Hinton broth (MHB) (Oxoid) or in chemically defined medium (CDM) modified from the method described by Catlin (Catlin and Schloer, 1962). CDM was prepared as described in Table S1. Both media where supplemented with 10 mM NaHCO3 prior to incubation. Liquid cultures were routinely grown from a starting OD600 reading of 0.05. Cultures were grown in triplicate in 30 ml polystyrene universal tubes (Sterilin®) in a total of 15 ml, and were shaken at 200 r.p.m. at 37°C in a microbial C25KC incubator shaker (New Brunswick Scientific). Growth in the presence of propionic acid was enabled by the addition of propionic acid (Sigma-Aldrich®) to a final concentration of 5 mM. Antibiotics were included as appropriate: spectinomycin, 50 μg ml −1 ; tetracycline 2.5 μg ml −1 .

Growth of N. meningitidis in whole blood
Human venous blood was collected from seven healthy adult volunteers (three males and four females) and an anticoagulant agent, heparin, was instantly mixed in the blood at a concentration of 17 U ml −1 . Samples were obtained from Wayne Burrill, University of Bradford, the morning subsequent to collection. N. meningitidis MC58 and NMB0431::Spec r were grown on plates overnight. A few colonies for each strain were inoculated the following day into MHB + 10 mM NaHCO3 until they reached an OD600 of 0.3, which corresponded to approximately 3 × 10 8 bacteria·ml −1 . Bacteria were subsequently diluted to 1 × 10 6 or 2 × 10 6 cfu ml −1 into MHB + 10 mM NaHCO3 and 10 μl of these suspensions were inoculated in triplicate into 190 μl of 100% human whole blood, resulting in a starting experimental concentration of between 5 × 10 4 and 1 × 10 5 cfu ml −1 . Whole blood infected with bacteria was then incubated over a period of 2 h in a 96-Well Optical Reaction Plate (Applied Biosystems) at 37°C with shaking at 110 r.p.m. to avoid red blood cell sedimentation. At each time point (0, 30, 60, 90 and 120 min), 20 μl of sample was removed and spread onto CBA plates. The number of viable bacteria was determined by cfu counts by plating serial dilutions onto CBA plates. Bacterial survival was determined by comparison of the viable count at the different time points with the control time 0, where survival rate corresponded to 100%.

Measurement of propionic acid content
Gas chromatography was used for separating propionic acid from growth medium. Five hundred microlitres of samples of culture were collected at different time points, centrifuged at 12 000 g for 5 min with a Sigma 1-13 microcentrifuge and the supernatant was stored at −80°C until the last sample was collected. Each sample to be analysed was then mixed with 132 mM potassium phosphate (pH 3). Half microlitre of freshly acidified samples were drawn up into a syringe, placed into a hot injector port of the 6890 N Network GC system gas chromatograph (Agilent Technologies), and the sample was injected. The injector port was at 250°C. Helium carrier gas flowed through the column (Alltech® AT-1000 Capillary Column; 150°C) at a constant flow of 2.2 ml min −1 . The area under the curve of the propionic acid elution peak was measured and converted to the actual concentration of propionic acid with the help of a standard curve.

Measurement of gene expression by RT-PCR
Messenger RNA and cDNA were prepared and the relative expression of genes determined as described previously (Heurlier et al., 2008) at time points throughout growth. Expression of NMB0431, NMB0432, NMB0433 and NMB0434 were determined relative to MetK (NMB1799) as endogenous control. Primers for RT-PCR are in Table S2.
To determine whether the prp gene cluster is expressed as a polycistronic operon, primers were designed to amplify across all the intergenic regions within the prp gene cluster and between NMB0428 and NMB0430, and NMB0435 and NMB0436 (i.e. the genes flanking the prp cluster). The primers are indicated in Fig. S1A, and shown in Table S2.
Pellets stored at −20°C were defrosted in 37°C water baths before sonication lysis with a MISONIX sonicator model 3000 for 3 min with 3 s on 7 s off pulses at 120 W. Total pellet lysate was then centrifuged at 10 000 g for 30 min in a Sigma 3-18 centrifuge with a Sigma 19776-H rotor to pellet insoluble cell debris. Crude cell lysate was then applied to 5 ml His-trap nickel columns previously equilibrated with equilibration buffer equivalent to five column volumes. After lysate had been applied, columns were then rinsed with another five column volumes of equilibration buffer before five column volumes of elution buffer [50 mM Tris pH 7, 300 mM NaCl, 500 mM Imidazole (+5 mM DTT for AckA1)] were applied; fractions equivalent to one column volume were collected from this. Purity was checked by SDS-PAGE on completion of the purification protocol. Acetate and propionate kinase activity was assayed using the hydroxamate assay as described previously (Aceti and Ferry, 1988), at a range of concentrations of substrate.
Taxonomic analysis of 16S rRNA oral microbiome data 16S rRNA amplicon sequence data were downloaded from the NCBI Sequence Read Archive (http://www.ncbi.nlm.nih.gov/ sra/) from four separate oral microbiome studies (Nasidze et al., 2009;Charlson et al., 2010;Stahringer et al., 2012;Human Microbiome Project, accession SRX055077). Adaptor sequences were trimmed using CutAdapt version 1.2.1 with an error rate of 0.2 and minimum length after trimming of 20 base pairs. QIIME version 1.7.0 was used to assign OTUs (operational taxonomic units) to the amplicon datasets, and run at both 97% and 99% similarity. The Greengenes (May 2013 version) was used to pick OTUs (with the UCLUST clustering algorithm) and assign taxa to representative sequences for each OTU. Relative abundance data produced by QIIME were used to calculate Pearson product moment correlations, using the SciPy stats module, between the taxa across the samples in each study, with correlation coefficients of ≤ −0.1175 or ≥ 0.1175 for the data from Stahringer et al. (2012), ≤ −0.1170 or ≥ 0.1170 for the data from Charlson et al., 2010), ≤ −0.18 or ≥ 0.18 for the data from Nasidze et al., 2009) and ≤ −0.3 or ≥ 0.3 for the data from SRX055077 being considered as significantly correlated.