Defining the healthy "core microbiome" of oral microbial communities

If you want to understand disease, start by understanding health. That's the shift the field of oral microbiology has been slowly making, and it changes everything about how we measure, compare, and ultimately diagnose. The mouth isn't just one habitat; it's a small archipelago. You have shedding islands like the cheek and tongue, and you have hard, permanent coastlines — the teeth and any dentures. Those are radically different surfaces. So, a real picture of "healthy" must be a cross-map of all those places at once, not a single swab or a guess from saliva. That's exactly what Zaura, Keijser, Huse, and Crielaard set out to do. Three healthy adults, with no antibiotics for three months, were sampled systematically across mucosal surfaces and dental sites, plus saliva. They held morning appointments. Twelve hours after brushing and a short fast was taken, just to calm the noise. The goal wasn't to chase every microbe in the city; it was to get a reliable snapshot of the neighborhoods that matter, side by side, so you could see what's shared and what's site-specific when everything is working well. On the sequencing bench, they chose a familiar workhorse of that era: 454 pyrosequencing of the V5 to V6 region of the 16S ribosomal RNA gene. Think of that region as a high-contrast badge most bacteria wear — distinctive enough to tell lineages apart, short enough to read at scale. They anchored everything to operational taxonomic units, or OTUs, clustered at a 3 percent genetic distance, which is the standard stand-in for species-level phylotypes in 16S work. Here's the conservative move that makes the rest believable: only keeping sequences that showed up at least five times anywhere in the dataset. Out went the single stray reads and most of the artefacts. Out of an initial 452,071 reads, 298,261 made the cut, forming 6,315 unique sequences and 818 OTUs at that 3 percent threshold. Depth per sample ended up around ten thousand reads. Not maximal, but careful. What came back was both rich and organized. Within each person, the mouth was a quilt of diversity. Each individual harbored more than five hundred species-level phylotypes and around ninety higher taxa — think genera and above. Zooming down to the sample level, the average was 266 phylotypes per site, with a spread that ran from about 120 up to just over 320. That's a lot of distinct lineages for a single cheek or an approximal tooth surface, and remember, this is health. It isn't sterile; it's stable. And yet, it wasn't uniform. Habitat mattered — a lot. Dental surfaces, especially the tight approximal spaces between teeth, were the most diverse. Cheek mucosa tended to be the simplest. In two of the volunteers, cheek samples were dominated by just a couple of OTUs, and even when a cheek sample carried a long list of phylotypes, many were present at low abundance. Saliva told its own story. It looked more like the mucosal neighborhoods than the hard surfaces, which makes intuitive sense — saliva is constantly washing over the mouth and sloughing cells, so it's a blended snapshot of the shedding surfaces more than a scraping of plaque. Now for the piece you might not expect if you've been told that everyone's microbiome is wildly personal: across these three unrelated, healthy adults, there was a striking shared backbone. At the level of unique sequences, just over a quarter — twenty-six percent — were present in all three. Expand to "found in at least two people," and you capture almost two-thirds. Move up a level to OTUs, and the common core jumps out: 387 out of 818 OTUs, about forty-seven percent, were shared by all three, and those alone accounted for ninety to ninety-three percent of the reads in each person. Roll it one level higher and the signal becomes overwhelming. Seventy-two percent of higher taxa were shared by all three, and those shared taxa encompassed virtually everything — ninety-nine point eight percent of reads. That's the definition of a core in practice: not just a few celebrity microbes, but an entire scaffold of lineages that consistently dominate healthy mouths. There was individuality too, and it lived in the long tail. In each person, seventeen to nineteen percent of unique sequences were exclusive to that mouth, and those exclusives contributed roughly eleven to twenty percent of that person's reads. Here's a subtle but important nuance. A large fraction of OTUs — about two-thirds — were represented by just a single sequence in a given individual. They were rare, they were real under the study's filters, and together they barely moved the needle on abundance, contributing just zero point seven percent of total reads. That's what ecological communities often look like: a steep head of abundant, shared members and a long, individualized tail of uncommon ones. If you're asking who those headliners are, the names will sound familiar. Across all sites, the big phyla were Firmicutes, Proteobacteria, Actinobacteria, Bacteroidetes, and Fusobacteria. Drill down and the recurrent cast included Streptococcus and the Veillonellaceae family within Firmicutes; Neisseria and Haemophilus among Proteobacteria; Corynebacterium, Rothia, and Actinomyces from the Actinobacteria side; Prevotella, Capnocytophaga, and Porphyromonas within Bacteroidetes; and Fusobacterium in its own phylum. Across all three people, nine specific sequences stood out as highly abundant — each one making up roughly half a percent to nearly six percent of reads — and they were present in everyone. Two were Streptococcus, two were Veillonellaceae, and the rest lined up with Granulicatella, Corynebacterium, Rothia, Porphyromonas, and Fusobacterium. On the habitat axis, Firmicutes leaned heavily into mucosal surfaces, while a candidate division often called TM7 popped on approximal tooth sites. Those patterns aren't just trivia; they're signatures that separate shedding tissue from enamel real estate. The team didn't leave that separation to gut feeling. They pulled the OTU abundance table into a principal component analysis to see how samples cluster when you compress the complexity into a few axes. Three components captured about half the variation in the data — fifty-one percent — which is typical in ecological communities. The first component, responsible for twenty-nine point seven percent of the spread, cleanly split dental samples from mucosal ones. The second, at twelve point three percent, mostly marked out one volunteer — S3 — from the other two, which is a reminder that individuals do impart their own flavor. The third, nine point one percent, reinforced the mucosa-versus-teeth separation, with tongue samples drawing together as a tight little cluster. And that saliva-mucosa story? On these axes, saliva tucked in closer to the shedding surfaces than to dental plaque. Under the hood, their data processing leaned conservative by design, which is why the patterns are credible. They cut hard on quality, discarded any read that didn't perfectly match the target primer or that looked suspect, and used a multi-source taxonomy scheme to avoid overconfident assignments. When a sequence matched more than one reference equally well, they dropped the label down to the lowest common taxon and moved on. OTUs were built with standard distance-based clustering, and diversity estimates used Shannon's index, which blends richness — how many distinct taxa are there — with evenness — how balanced their abundances are. You don't need the formula to understand the idea: it punishes communities that are dominated by one or two taxa and rewards those that spread the diversity. There are real limits here, and the authors are straightforward about them. It's a small cohort: three healthy Caucasian men. This isn't a census; it's a deep look at a few mouths. The sequencing depth per sample, roughly ten thousand reads, is conservative for 454 at the time and means the absolute richness numbers are likely underestimates. Additionally, any 16S-based study lives with a taxonomic blur — a 3 percent genetic distance doesn't signify the same biological gap in every clade. Those caveats matter for precision. They don't erase the architecture that shows up again and again: a strong, shared core by both OTUs and higher taxa, overlaid with consistent habitat differences and a personalized, low-abundance tail. So what do you do with a map like this? First, you stop pretending that a single swab can stand in for the mouth. A healthy reference needs to be site-aware — mucosa is not teeth, and "saliva equals the mouth" is only partially true. Second, you get a baseline. As Zaura and colleagues argue, you can't call a shift a dysbiosis if you don't know what the stable state looks like across the places that matter. Their numbers give you that scaffold. Half a dozen abundant lineages show up across everyone. Roughly half the OTUs are shared and carry the vast majority of reads. Higher taxa are even more universal. Those are the landmarks you can look for when you want to say, "this site is drifting," and mean it. There's also a quieter lesson about how populations hold together. The notion that "everything is everywhere, but the environment selects" plays out here. Many of the same taxa appear across people, but whether they thrive depends on the surface. Cheek favors fast colonizers and biofilm-light lifestyles. Approximal enamel favors structured, multispecies biofilms where specialists can nestle in. That's why you see Firmicutes leaning one way, and a group like TM7 lighting up the other. It's not that one person's biology is rewriting the rules; it's that the rules differ by niche, and the personal signature mostly rides in the rare biosphere. Where this needs to go next is longitudinal and comparative. The authors call for following people over time, through perturbations — say, an antibiotic course or the onset of gingivitis — with the same cross-niche rigor. If the core is as stable and dominant as these data suggest, early warning signs of disease may appear as subtle shifts in the abundant, shared OTUs before the low-abundance tail flares. And diagnostics that respect site — a molar approximal panel, a tongue dorsum panel — could outperform one-size-fits-all saliva screens. But for today, the payoff is simple and useful. Health has a shape, and in the mouth, it's built from a common set of lineages that most of us share, expressed through the lens of the surface they live on. As Zaura and colleagues showed, you can see that shape clearly if you sample the right places, sequence deeply but carefully, and let the data tell you where the boundaries are. That's a strong foundation for thinking about oral ecology the way ecologists think about forests and reefs — not as a list of species, but as a set of stable communities, each with its own cast, sharing a common script.