Human and bacterial genetic variation shape oral microbiomes and health

Kamitaki, Nolan; Handsaker, Robert E.; Hujoel, Margaux L. A.; Mukamel, Ronen E.; Usher, Christina L.; McCarroll, Steven A.; Loh, Po-Ru

doi:10.1038/s41586-025-10037-7

Download PDF

Article
Open access
Published: 28 January 2026

Human and bacterial genetic variation shape oral microbiomes and health

Nature (2026)Cite this article

54 Altmetric
Metrics details

Subjects

Abstract

Human genetic variation influences all aspects of our biology, including the oral cavity^1,2,3, through which nutrients and microbes enter the body. Yet it is largely unknown which human genetic variants shape a person’s oral microbiome and potentially promote its dysbiosis^3,4,5. We characterized the oral microbiomes of 12,519 people by re-analysing whole-genome sequencing reads from previously sequenced saliva-derived DNA. Human genetic variation at 11 loci (10 new) associated with variation in oral microbiome composition. Several of these related to carbohydrate availability; the strongest association (P = 3.0 × 10⁻¹⁸⁸) involved the common FUT2 W154X loss-of-function variant, which associated with the abundances of 58 bacterial species. Human host genetics also seemed to powerfully shape genetic variation in oral bacterial species: these 11 host genetic variants also associated with variation of gene dosages in 68 regions of bacterial genomes. Common, multi-allelic copy number variation of AMY1, which encodes salivary amylase, associated with oral microbiome composition (P = 1.5 × 10⁻⁵³) and with dentures use in UK Biobank (P = 5.9 × 10⁻³⁵, n = 418,039) but not with body mass index (P = 0.85), suggesting that salivary amylase abundance impacts health by influencing the oral microbiome. Two other microbiome composition-associated loci, FUT2 and PITX1, also significantly associated with dentures risk, collectively nominating numerous host–microbial interactions that contribute to tooth decay.

Metagenome-genome-wide association studies reveal human genetic impact on the oral microbiome

Article Open access 07 December 2021

Genome-wide association study identifies host genetic variants influencing oral microbiota diversity and metabolic health

Article Open access 26 June 2024

Impact of high altitude on composition and functional profiling of oral microbiome in Indian male population

Article Open access 10 March 2023

Main

When Antonie van Leeuwenhoek first observed bacteria as ‘animalcules’ in scrapings from his teeth in the seventeenth century, one of his first inquiries involved the extent of their variation among people⁶. Oral microbiomes are now known to vary abundantly across people^7,8,9, and twin studies have shown that some of this variation is heritable^1,2,3. However, few human genetic polymorphisms have been associated with the abundances of specific oral microbial species^3,4,5; study sizes so far (n < 3,000) have provided limited power to detect robust genetic effects. Larger genome-wide association studies (GWAS) of the gut microbiome (n = 5,959–18,340) have consistently replicated two effects of variation at the LCT and ABO loci on gut microbial abundances^10,11,12,13, and larger GWAS of oral microbiomes might yield similar discovery.

Oral pathologies, such as dental caries, result from dysbiosis of the oral microbiome¹⁴. Untreated pathologies can progress to oral infections which carried high mortality rates before modern dentistry and antibiotics¹⁵. Susceptibility to caries and other oral pathologies is also strongly influenced by genetics^16,17, and GWAS have identified 47 loci harbouring such genetic effects¹⁸. However, whether these or other genetic effects act by modulating the composition of the oral microbiome is at present unknown. Identifying such interactions could point to microbial drivers of cariogenesis⁹.

Given the effects of human hosts and resident microbes on each other’s survival and evolutionary trajectory, the human microbiome is an example of symbiosis^19,20. The stability of the gut microbiome in individuals²¹, its codiversification with humans²² and abundant structural variation of its microbial genomes²³ all suggest intricate genetic interactions between microbiomes and their human hosts, whereby microbial genomes adapt to genetic variation across people. A recently observed example of such an interaction with the gut microbiome is a structural variant in the Faecalibacterium prausnitzii genome that includes genes encoding an N-acetylgalactosamine (GalNAc)-metabolizing pathway and interacts with human ABO variation²⁴. Whether such specific co-adaptation commonly occurs in oral microbiomes remains an open question.

Oral microbiome profiles of 12,519 people

To create a dataset suitable for exploring variation in the oral microbiome and the way it is shaped by human genetic variation, we analysed DNA sequencing reads previously generated from whole-genome sequencing (WGS) of saliva samples from 12,519 participants in the Simons Foundation Powering Autism Research (SPARK) cohort²⁵ (Fig. 1a), building on previous work^26,27. WGS captured substantial non-human genomic information²⁸, with a median of 8.4% ([4.6%,14.7%], quartiles) of sequencing reads not mapping to the human reference genome (Extended Data Fig. 1a). Many of these unmapped reads instead mapped to clade-specific marker genes in microbial genomes²⁹, enabling quantification of relative microbial abundances. This produced the largest collection of oral microbiome profiles (n = 12,519) generated so far, measuring the abundances of 645 microbial species present at >1% frequency, including 439 species (spanning 13 phyla, including one fungal commensal, Malassezia restricta) commonly observed in SPARK (≥10% of participants) (Fig. 1b, Extended Data Fig. 1b and Supplementary Table 1). Comparing these profiles across individuals showed that age was a major driver of interindividual variation in oral microbiome composition, unlike autism spectrum disorder (ASD) case status, sex and genetic ancestry (Fig. 1c and Extended Data Fig. 1c–g). Across the lifespan represented in SPARK (age 0–90 years), mean species diversity sharply increased in the first few years of life (representing when the oral cavity is colonized, diet diversifies and primary teeth are acquired) and then decreased slowly with age⁸ (Fig. 1d). Individual species exhibited vastly different abundance trajectories over the lifespan, with some observed predominantly in adults and others predominantly in children (Extended Data Fig. 1h–k).

**Fig. 1: Oral microbiomes in 12,519 individuals measured by WGS of saliva samples.**

Human genetics shapes oral microbiome composition

To identify human genetic variants that influence interindividual differences in the abundances of microbial taxa, we first tested the abundances of taxa detectable in ≥10% of participants for association with common human genetic variants, accounting for family structure using a linear mixed model^30,31. Human genetic variants at seven loci associated with the abundance of at least one taxon at study-wide significance (P < 4.0 × 10⁻¹¹; Extended Data Fig. 2a), with only one locus (SLC2A9) previously identified⁴. As several loci associated with the abundances of many species (Supplementary Tables 2 and 3) and none associated with α-diversity (Extended Data Fig. 2b), we developed a statistical test to capture pleiotropic effects on many species in an interdependent microbial community^32,33,34, using principal component analysis (PCA) to enable efficient genome-wide association testing (Fig. 2a and Methods). Similar to a recent approach for GWAS on high-dimensional cell state phenotypes in single-cell RNA-seq data³⁵, this approach also reduces multiple-testing burden by testing each genetic variant only once.

**Fig. 2: Associations of host genetic variants with oral microbiome composition overlap risk loci for dentures use.**

Applying this approach to SPARK identified four additional human genomic loci (11 total) at which common genetic variation associated with oral microbiome composition (P < 5 × 10⁻⁸; Fig. 2b and Extended Data Table 1). The principal component (PC)-based test was well-calibrated (Extended Data Fig. 2c) and top signals were confirmed by multivariate distance matrix regression^33,36 (Extended Data Fig. 2d,e). The association signals tended to distribute across many microbial PCs (mPCs; Extended Data Fig. 2f–p), suggesting that human genetic variants subtly influence many axes of microbial community coordination. Among the ten new loci, eight implicated genes—and in several cases, specific variants—with readily interpretable functions that could explain their associations with microbiome composition.

Three loci contained genes encoding highly expressed salivary proteins: salivary amylase (encoded by AMY1; P = 1.5 × 10⁻⁵³, top association), submaxillary gland androgen-regulated proteins (SMR3A and SMR3B; P = 1.4 × 10⁻¹²) and basic salivary proline-rich proteins (PRB1–PRB4; P = 1.1 × 10⁻¹¹). These associations seemed to be driven mainly by genetic variants that modify gene expression or copy number (Extended Data Table 1, Extended Data Fig. 3 and Supplementary Note 1). Consistent with these results, heritability-partitioning analysis³⁷ indicated that genetic effects on oral microbiome composition are enriched at genes specifically expressed in salivary glands (P = 0.02, Extended Data Fig. 4a).
Two loci contained genes with established roles in immune function: the HLA class II genes, which encode proteins that present peptides in adaptive immunity, and TLR1, encoding Toll-like receptor 1, that binds bacterial lipoproteins in innate immunity. The strongest association at TLR1 involved a missense variant (rs5743618; P = 6.2 × 10⁻¹⁸) that produces the I602S substitution known to inhibit trafficking of TLR1 to the cell surface, reducing immune response in a recessive manner^38,39. Consistent with these reports, I602S associated recessively with microbial abundances (P = 6.7 × 10⁻²⁹; Extended Data Fig. 4b).
Two other loci, ABO and FUT2, encode glycosyltransferases that together determine expression of histo-blood group antigens on epithelial cells and secreted proteins (in addition to the well-known role of ABO in determining blood type). This broader role is important to microbial species that interact with mucosal surfaces, such that both loci are known to influence the gut microbiome^10,11,12,13, with some bacterial species using A-antigen saccharides as a carbohydrate source²⁴. The variants at ABO and FUT2 that associated most strongly with oral microbiome composition were rs2519093 (P = 9.5 × 10⁻¹⁵), which tags the A1 blood group⁴⁰, and rs601338 (P = 1.6 × 10⁻¹³¹ additively, P = 3.0 × 10⁻¹⁸⁸ recessively), the common FUT2 W154X nonsense variant that (in homozygotes) produces the non-secretor phenotype in which bodily fluids lack histo-blood group antigens⁴¹.
Associations of variants at PITX1 with oral microbiome composition colocalized with previously reported associations of these variants with dental caries and dentures use (r² = 0.99 between the top microbiome-associated variant (rs3749751; P = 3.0 × 10⁻¹¹) and the top dentures-associated single nucleotide polymorphism (SNP) at PITX1; Extended Data Fig. 4c; ref. ¹⁸). PITX1 is a developmentally expressed gene which seems to have a role in mandibular tooth morphogenesis (based on a knockout mouse model)⁴², suggesting that common genetic variation at PITX1 might influence tooth morphology and through it, oral microbiota and dental health.

The shared associations of genetic variants at PITX1 with both oral microbiome composition and dental health phenotypes suggested that other genetic influences on the oral microbiome might similarly influence dental health. To explore this, we performed GWAS of dentures use (a proxy for tooth loss and caries) in the UK Biobank (UKB) cohort (n = 75,156 cases, n = 342,883 controls)⁴³. Three loci—AMY1, FUT2 and PITX1—contained variants that associated (P < 5 × 10⁻⁸) with both oral microbiome composition and dentures use, and at each of these loci, the association patterns colocalized (Fig. 2b and Extended Data Fig. 4c,d). Moreover, at 8 of the 11 loci influencing oral microbiome composition, the most strongly associated variant also exhibited at least a nominal association (P < 0.05) with dentures risk (Extended Data Table 1), suggesting that host genetic effects on oral microbiome composition often have downstream effects on oral health.

Most of these genetic associations seemed to involve effects of human genetic variation on the abundances of several bacterial species, with 167 species–genotype pairs reaching FDR < 0.05 across the 11 loci (Supplementary Table 3). These associations were not driven by compositional effects or by ASD status (Extended Data Fig. 4e,f). The strong associations at AMY1 and FUT2 offered an opportunity for detailed investigation of how genetic variation at these loci influences oral microbiomes and oral health. FUT2 W154X associated with the abundances of 58 of the 439 species (Fig. 2c–e). FUT2 seemed to be nearly but not completely haplosufficient in these associations, with slightly weaker abundance-modifying effects observed among secretor individuals with a heterozygous W154X genotype compared to those with two wild-type alleles (Fig. 2d and Extended Data Fig. 4g). For several pairs of closely related species, FUT2 W154X associated with increased abundance of one species and decreased abundance of the other (Fig. 2e and Extended Data Fig. 5), possibly reflecting competition between closely related species for ecological niches.

Effects of complex variation at the amylase locus

AMY1 encodes salivary α-amylase, an enzyme that breaks down dietary starches into simple sugars. The dramatic copy number expansion of the amylase locus in humans and other animals⁴⁴ has attracted much interest for its theorized role in facilitating recent adaptation to starch-based diets^45,46,47, but its reported association with human body mass index (BMI)⁴⁸ and type 2 diabetes⁴⁹ has been controversial^50,51. AMY1 copy number genotypes in SPARK and UKB (estimated from WGS depth-of-coverage) showed extensive polymorphism^45,50 (2–32 copies per individual; Fig. 3a and Extended Data Fig. 6a) and high mutability (6.3 × 10⁻⁴ (3.7 × 10⁻⁴–11 × 10⁻⁴, 95% confidence interval (CI)) mutations per haplotype per generation, similar to an estimate using coalescent modelling⁴⁶; Extended Data Fig. 6b).

**Fig. 3: Complex genetic variation of the salivary amylase locus affects the abundance of several oral microbial species and oral health.**

Copy number variation of AMY1 generated the strongest association of genetic variation at the amylase locus with oral microbiome composition (P = 1.5 × 10⁻⁵³; Fig. 2b) and associated with the abundances of 42 bacterial species (FDR < 0.05, 22 species at FDR < 0.01 with P = 5.1 × 10⁻²⁵ to P = 0.00047; Fig. 3b). The abundances of these species changed stepwise with AMY1 copy number, generating a long allelic series with steadily increasing or decreasing abundances (Fig. 3c), congruent with the effect of AMY1 copy number on the abundance of secreted salivary amylase^45,52. Two of these associations seemed to confirm associations previously observed in smaller candidate gene studies^49,53 (Supplementary Note 2).

AMY1 copy number also associated strongly with dentures use in UKB (P = 5.9 × 10⁻³⁵, surpassed only by the PITX1 locus; Fig. 2b). Each additional copy of AMY1 associated with a 2.1% (1.7%–2.4%) increase in the odds of having dentures, corresponding to a 1.4-fold range in odds across people with 2–16 AMY1 copies (Fig. 3d). This association replicated in the All of Us (AoU) cohort⁵⁴ (n = 230,002; P = 3.5 × 10⁻⁴ for tooth loss, P = 6.1 × 10⁻³ for caries; Extended Data Fig. 6c–e). Surprisingly, AMY1 copy number associated with decreased risk of bleeding gums in UKB (P = 1.5 × 10⁻⁶; Fig. 3e), even though gingivitis is considered a risk factor for tooth loss^55,56. However, the bleeding gums and dentures use phenotypes had little genetic overlap¹⁸ and were slightly negatively correlated (r = −0.07, s.e. 0.0015), suggesting largely independent pathology. These associations were specific to AMY1; the copy number of AMY2A and AMY2B (encoding pancreatic amylase) did not associate with dental phenotypes (Extended Data Fig. 6f,g).

Beyond the effect of AMY1 copy number, two missense variants in AMY1 carried by 1%–2% of UKB participants seemed to confer the largest increases in dentures risk of all common variants in the human genome (OR = 1.59 (1.46–1.73) per copy of AMY1 F141C (P = 2.5 × 10⁻²⁶); OR = 1.16 (1.10–1.23) per copy of AMY1 C477R (P = 8.3 × 10⁻⁸); Fig. 3d). These two paralogous sequence variants were typically carried on haplotypes containing three or four copies of AMY1 and produced the strongest conditional associations with dentures use in two stages of stepwise conditional analysis (Fig. 3f–h). The AMY1 F141C and C477R variants seemed to confer an increase in dentures risk equivalent to increasing AMY1 copy number by 22.4 (18.3–26.5) and 7.3 (4.6–9.9) copies, respectively (Fig. 3d). This apparent gain-of-function effect was surprising, as both variants were predicted to be damaging (PolyPhen-2⁵⁷ score of 1.0). Analyses of amylase protein expression did not detect effects of AMY1 F141C on enzymatic activity (Supplementary Note 3, Extended Data Fig. 7a–c and Supplementary Fig. 1). The extended allelic series of AMY1 copy number and missense variants associated with dentures use provided a set of genetic instruments for evaluating which bacterial species might causally contribute to tooth loss (Fig. 3i and Extended Data Fig. 7d). Reverse causality (that is, dentures use causing changes in the oral microbiome that associate with AMY1 variants) seemed to be unlikely based on the concordance of effect sizes in children and adults (Extended Data Fig. 7e).

The UKB and AoU datasets also enabled rigorous evaluation of whether or not AMY1 copy number influences BMI among modern humans. AMY1 copy number did not associate with BMI in UKB (n = 418,150, P = 0.85; Fig. 3j), AoU (n = 219,879, P = 0.30, Extended Data Fig. 7f) or any genetic ancestry in AoU (Extended Data Fig. 7g–i).

Genetic associations with bacterial gene dosage

To identify molecular mechanisms by which human genetic variation engages the oral microbiome, we next tested whether the microbial species for which abundances associated with each of the 11 loci might be united by shared biochemical pathway use⁵⁸ (Extended Data Fig. 8a, Supplementary Tables 4 and 5 and Supplementary Note 4). This analysis identified an adhesin gene in Haemophilus sputorum at which sequencing coverage associated particularly strongly (relative to elsewhere in the H. sputorum genome) with FUT2 W154X, suggesting that the adhesin interacts with FUT2-dependent glycosylation (Extended Data Fig. 8b and Supplementary Note 4). To search for similar molecular interactions between human and bacterial proteins, we tested the 11 lead variants associated with oral microbiome composition (Extended Data Table 1) for association with microbial gene dosages²⁴ (Fig. 4a). The key conceptual difference between testing human genetic variants for effects on microbial abundances (Fig. 2a) versus microbial gene dosages (Fig. 4a) is that the latter approach searches for effects on relative fitness of bacterial strains that do or do not contain a genomic region (rather than fitness of a bacterial species). Thus, it highlights microbial genomic regions that may contain genes whose products are involved in a host–microbe genetic interaction (Supplementary Note 5).

**Fig. 4: Numerous deletions in oral microbial genomes are selected for by human genetic variation.**

To minimize hypothesis testing burden, we searched specifically for associations of the 11 variants with measurements of normalized WGS coverage in 500 base pair (bp) bins tiled across 30 bacterial reference genomes (Fig. 4a, Supplementary Table 6, Methods and Supplementary Note 5). This analysis identified 208 associations involving 68 regions of 18 bacterial genomes in which normalized read depth associated with one or more of the 11 human genetic variants (FDR < 0.01, Fig. 4b and Supplementary Tables 7 and 8). For example, amylase-binding protein orthologues in Streptococcus parasanguinis, abpA and abpB, increased in dosage with higher host AMY1 copy number (P = 1.13 × 10⁻⁷ and 1.80 × 10⁻⁷, respectively; Extended Data Fig. 8c–g and Supplementary Note 6). Normalized read depth in nearly half of these regions (33/68) associated with secretor status (based on FUT2 W154X genotype). Eight regions associated with more than one human genetic variant; among them, five regions associated with both secretor status and ABO*A1 (Fig. 4b). Effect directions replicated for 202 of the 208 bin-level associations in 10,000 saliva-derived WGS samples from AoU (Extended Data Fig. 8h and Supplementary Table 7).

ABO A antigen selects for a glycoside hydrolase

Host ABO*A1 genotype (based on rs2519093) strongly associated with whether Prevotella strains—a prevalent oral genus involved in early biofilm formation⁵⁹—carried a gene encoding a glycoside hydrolase (Fig. 5a–d). ABO*A1 genotype associated exceptionally strongly (P = 4.8 × 10⁻¹⁹–8.3 × 10⁻²⁴¹) with normalized WGS coverage across a 3 kilobase (kb) segment of the Prevotella nanceiensis reference genome (Fig. 5a). This region is annotated as a glycoside hydrolase pseudogene due to an N-terminal truncation in the reference genome. However, assembly of unmapped sequencing reads with mates aligned to the region showed no evidence of such truncation: rather, the reads seemed to originate from a full-length gene, with 95% homology to a glycoside hydrolase found in Prevotella salivae (a species not included among the 30 reference genomes analysed).

**Fig. 5: *ABO* variation modulates selection on a glycoside hydrolase gene in *Prevotella* and also engages the oral microbiome in non-secretors.**

ABO*A1 genotype associated with sequencing coverage in this region only in secretor individuals (P = 1.7 × 10⁻³⁰⁷ in secretors; P = 0.44 in non-secretors), that is, individuals with at least one functional copy of FUT2, allowing expression of histo-blood group antigens on epithelial cells and secreted proteins (Fig. 5d). This FUT2-dependent effect of host blood group seemed to be driven specifically by A antigen presentation: the fraction of individuals for whom the glycoside hydrolase gene was detectable in saliva-derived DNA increased from 46%–48% in non-secretors and individuals with B or O blood type to 71%–77% in secretors with A or AB blood type (Fig. 5b). This association further reflected the quantity of A antigen predicted by an individual’s diploid ABO genotype: ABO*O, B, A2 and A1 alleles exhibited an allelic series of effects on normalized WGS coverage of the glycoside hydrolase gene that was consistent with the increasing abilities of the glycosyltransferases encoded by these alleles to synthesize A antigen (Fig. 5c). The B allele imposed a strong opposing effect when present in an individual heterozygous for an A1 or A2 allele (β = −0.074 [−0.12, −0.032] for B relative to O, P = 5.8 × 10⁻⁴, Fig. 5c), presumably reflecting competition between A and B transferases for available galactose residues on acceptor H antigens (Fig. 5d).

Taken together, these results indicate that the glycoside hydrolase enables Prevotella strains that express it to use type A histo-blood group antigens presented on host mucosal cell surfaces or salivary proteins (in secretors) as a carbohydrate source, similar to a recently observed effect in the gut microbiome²⁴. We hypothesize that the glycoside hydrolase binds A antigens and cleaves the α1,2-fucosyl group synthesized by FUT2 (Fig. 5d).

Host ABO genotypes showed an intriguingly different pattern of association with a genomic region of the most abundant species in SPARK, Rothia mucilaginosa (P = 1.4 × 10⁻²⁴; Fig. 5e,f). Blood groups A, B and AB all associated with absence (rather than presence) of this region of the R. mucilaginosa genome, and surprisingly, these associations were observed in non-secretors as well as secretors (Fig. 5e,f). This region contains genes that encode a protein with no annotated domains and a 3-isopropylmalate dehydrogenase functioning in leucine biosynthesis, leaving the mechanism of association unknown.

More broadly, this non-FUT2-dependent ABO association suggested the possibility that the association of ABO*A1 genotype with oral microbiome composition (Fig. 2b) might also be partially independent from secretor status. Indeed, in non-secretors the ABO*A1 association with microbiome composition remained significant (P = 0.004). This suggests that some effects of ABO variation on the oral microbiome come from cells not dependent on FUT2 for H antigen production (for example, blood and endothelial cells, which instead use FUT1). For example, bacteria can produce glycans structurally similar to A or B antigens that can then be recognized by anti-A or anti-B antibodies that are made by plasma cells and infiltrate into the mouth⁶⁰.

Secretor status selects for microbial adhesins

Conversely, many regions in oral microbial genomes associated with secretor status but not ABO*A1 genotype (Fig. 4b), and oral microbiome composition associated much more strongly with secretor status than with any variant at ABO (P = 3.0 × 10⁻¹⁸⁸ versus P = 9.4 × 10⁻¹⁵; Fig. 2b). This pattern contrasted with host genetic influences on gut microbiomes (which generate stronger associations at ABO than at FUT2; refs. ^11,12,13), leading us to wonder whether these regions might point to a molecular mechanism by which secretor status influences oral microbiomes independently of ABO.

Examining genes in bacterial genomic regions associated with secretor status identified three classes of bacterial proteins that were each implicated by several genes. Proteins with YadA-like domains were encoded by nine genes in three species: Veillonella sp. 3627 (vadA through vadF), Haemophilus sputorum (hadA and hadB) and Haemophilus parahaemolyticus (hadC) (Fig. 6a–d and Supplementary Table 8). YadA (from Yersinia pestis) is a trimeric autotransporter adhesin that aids attachment to host cells by binding components of the extracellular matrix, and some such adhesins are known to recognize host protein glycosylation^61,62, such as the glycosylation added or enabled by FUT2. Five of the seven regions containing these genes were present (that is, not deleted) more often in the oral microbiomes of secretors than non-secretors, consistent with the hypothesis that the adhesins they encode bind histo-blood group antigens on the host cell surface. This was true of hadC in H. parahaemolyticus despite this species exhibiting lower abundance in secretors (Extended Data Fig. 8i). The genome of V. sp. 3627 contained three such regions (containing vadB through vadF, where vadC through vadE fall within the same complex region, Extended Data Fig. 8j) whose presence or absence was observed largely independently in different microbiomes (Fig. 6e). Classifying individuals on the basis of which combination of regions was present in their V. sp. 3627 population showed increasing enrichment of secretors among individuals with increasing representation of vadB–vadF genes in V. sp. 3627 (Fig. 6e).

**Fig. 6: *FUT2*-dependent secretor status selects for three classes of adhesin genes across five microbial species.**

FUT2-associated genomic regions additionally implicated two other classes of proteins that seemed to have roles in adhesion to host cells. Four proteins with CshA domains (CrpD and CrpE in Streptococcus mitis, CrpF and CrpG in Streptococcus vestibularis) and six proteins with mucin-binding domains (MucBP, Muc_B2, MucBP_2) (SmdA through SmdE in S. mitis, SmdF in S. vestibularis) were encoded by genes in FUT2-associated bacterial genomic regions (Fig. 6f,g). CshA from Streptococcus gordonii binds host fibronectin⁶³, a heavily glycosylated component of the extracellular matrix. Similarly, mucins have numerous glycosylation sites and are up to 90% carbohydrate by mass⁶⁴. Interestingly, one of these proteins, CrpE in S. mitis, seems to contain both a CshA domain and multiple mucin-binding domains (Fig. 6h and Supplementary Table 9), suggesting that these might function in concert to bind the same host protein or a combination of proximal targets multivalently.

These enrichments of genes encoding proteins with YadA, CshA and mucin-binding domains were unlikely to occur by chance: the genomes of V. sp. 3627, H. sputorum and H. parahaemolyticus only contain 12, 4 and 8 genes with YadA domains, respectively (Fisher’s exact P = 7.5 × 10⁻¹², 3.5 × 10⁻⁵ and 0.021), and the genomes of S. mitis and S. vestibularis only contain two and three genes with CshA domains (P = 7.4 × 10⁻⁵ and 2.0 × 10⁻⁵) and ten and three genes with mucin-binding domains (P = 3.3 × 10⁻¹¹ and 0.0087). Most of these genes (15/19) were more commonly present in the oral microbiomes of people with functional FUT2, suggesting that they might encode bacterial lectins that depend on either fucosylation or sugar moieties added by ABO glycotransferase⁶⁵ (Fig. 6i). This convergence of bacterial genomic adaptations to host FUT2 genotype broadly suggests that commensal bacteria commonly make use of host histo-blood group antigens not only as a carbohydrate source but also for bacterial attachment to host cell surfaces.

Discussion

Analysis of the largest set of oral microbiome profiles generated to date identified many specific human genetic variants that contribute to the diversity observed across the oral microbiomes of different people^7,8. The large number of such effects suggests a larger influence of human genetics on the oral microbiome than on the gut microbiome^66,67, perhaps because host cells in the mouth interface more directly with bacteria (in contrast to cells in the gut, which are typically protected by a mucosal barrier). Some of these genetic effects on microbial abundances seem likely to mediate associations of the same human genetic variants with oral health phenotypes, nominating bacterial species that may contribute to dysbiosis. The salivary amylase gene generated the strongest such shared effect on oral microbiomes and health, driven by both AMY1 copy number variation and missense mutations in AMY1. The expansion of salivary amylase copy number in humans and domesticated animals has been hypothesized to be the result of positive selection driven by the advent of agriculture^44,45,46,47. Our observation here that AMY1 gene copy number variation associates with oral microbial phenotypes that lead to clinically relevant conditions—combined with the high mortality rate of tooth infections before modern dentistry and antibiotics¹⁵—suggests that AMY1 copy number may have been under selection as a result of effects on oral health in addition—or in response—to dietary changes.

The numerous associations that these analyses uncovered between human genetic variants and bacterial gene dosages suggest frequent intergenomic adaptation of microbial species to individual human hosts and implicate specific molecular interactions likely to drive such adaptation. Most of these associations involved genes in bacterial species whose overall abundances were unaffected by the same human genetic variants, similar to recent observations of associations of BMI with gut microbial sequence variation⁶⁸, suggesting that genomic adaptations enable many bacterial species to survive equally well across variable host genetic environments. By contrast, an association with relative species abundance could imply that the microbial genome is unable to adapt to a particular human variation. The variable gene regions we identified showed some breakpoint heterogeneity (Extended Data Fig. 9a,b) and could either reflect gene dosage variation among circulating strains or recurrent mutations, such as in Helicobacter pylori⁶⁹. The large number of such effects suggests that analyses of bacterial gene dosage may be a powerful way to identify host genetic influences on microbiomes, perhaps because analysing the balance between members of the same species with and without a variable gene controls for strong environmental influences on species abundance.

We note a need for care in conducting GWAS of microbial-abundance phenotypes. We initially observed a strong association (P = 2.5 × 10⁻⁷⁰) of oral microbiome composition with variant calls in the ribosomal RNA gene region of the p-arm of chromosome 21; however, these variant calls (which later failed a mappability filter) actually reflected the presence of orthologous bovine rDNA sequences and associated with the abundances of bacterial species used in dairy fermentation, suggesting DNA co-acquisition from recently eaten dairy foods (Supplementary Note 7). Our analytical approach for identifying host–microbe genetic interactions had several limitations that should be ameliorated with larger cohorts and improved microbial reference genomes (Extended Data Fig. 9c–h and Supplementary Note 8). Future datasets will also provide increased power to resolve possible pleiotropy and reverse causality with oral health phenotypes, either through cohorts with human genetic, microbiome and oral health phenotypes⁷⁰ or by Mendelian randomization approaches powered by even larger saliva sequencing datasets—which we have shown here provide rich information about how oral microbiomes are shaped by human genetics.

Methods

Ethics

This research complies with all relevant ethical regulations. The study protocol was determined to be not human subjects research by the Broad Institute Office of Research Subject Protection as all data analysed were previously collected and de-identified.

Quantification of microbial relative abundances from saliva WGS

We analysed saliva-derived WGS data previously generated for 12,519 individuals from the SPARK cohort of the Simons Foundation Autism Research Initiative (SFARI)²⁵. In brief, DNA extracted from saliva samples was prepared with PCR-free methods for 150 bp paired-end sequencing on Illumina NovaSeq 6000 machines. Reads were aligned to human reference build GRCh38 by the New York Genome Center using Centers for Common Disease Genomics project standards. Details of saliva sample collection, DNA extraction and sequencing were described in ref. ²⁶ (which analysed data from sequencing waves WGS1–3 of the SPARK integrated WGS (iWGS) v.1.1 dataset; here we analysed WGS1–5, which included additional samples included in subsequent sequencing waves).

From the CRAM files previously aligned to GRCh38, we extracted all unmapped reads for subsequent generation and analysis of oral microbiome phenotypes. This retrieved a median of 67.5 million unmapped reads per sample ([35.9 million,126.1 million], quartiles), comparable with the total number of reads used for previous metagenomic characterization of human microbiomes⁷¹ and consistent with previous analyses of SPARK samples for oral microbiome profiling^26,27. Unmapped reads were converted to compressed FASTQ with samtools (v.1.15.1) and then used as input for microbiome profiling using MetaPhlAn (v.4.0.6) with the vOct22 reference database. To evaluate robustness of these oral microbiome profiles, relative abundances of species in SPARK samples were compared with those from samples in the Human Microbiome Project⁷² by first subsetting to those profiled in both cohorts and then performing principal coordinate analysis using Bray–Curtis distance (Extended Data Fig. 1b).

From the relative abundance phenotype generated for each species by MetaPhlAn, we estimated the fractions of variance explained by covariates (age, sex, ASD and genetic ancestry PCs; Fig. 1c) using analysis of variance (finding the sum-of-squares for each covariate and dividing by the total sum-of-squares).

Generation of mPCs

Relative abundance measures of all microbes were first filtered to 439 entries corresponding to microbial species found in at least 10% of SPARK DNA samples. Rank-based inverse normal transformation across individuals was then performed for the abundance of each species. This transformation did not always produce values with mean 0 and variance 1 (due to large fractions of samples with zero abundance), so these were then scaled and centred for use as input to PC analysis to obtain 439 orthogonal microbial abundance principal components (mPCs) representing orthogonal axes of microbial variation.

Genotyping and quality control of human genetic variants in SPARK

Variant calling in SPARK was previously performed using DeepVariant (v.1.3.0) to produce sample-level VCFs from reads aligned to GRCh38 followed by GLnexus (v.1.4.1) to call variants jointly across the cohort. We performed a series of QC steps on the joint call set, starting by converting half-calls to missing and then excluding variants with >10% missingness using plink2 (v.2.00a3.6LM). Variants were further excluded if they had a minor allele frequency <1% or if they had a Hardy–Weinberg equilibrium exact test P < 1 × 10⁻⁶ with mid-P adjustment for excessive heterozygosity, leaving 12,525,098 common variants. Genetic ancestry PCs were generated by LD-pruning variants in 500 kb windows with r² > 0.1 and then running plink2 --pca approx. No individuals were filtered for outlier heterozygosity after inspection in each genetic ancestry group. Variants were then filtered to those present in the TOPMed-r3 imputation panel to exclude those in regions of poor mappability to produce a final set of 9,618,621 common variants to test for association with oral microbiome phenotypes.

mPC-based GWAS of oral microbiome composition

A straightforward way to search for host genetic effects on microbiomes is to test human genetic variants for association with the abundance of each microbial taxon in turn^10,11,12,13. However, we reasoned that a statistical test designed to aggregate evidence of pleiotropic genetic effects on many species in a microbial community could considerably increase statistical power^32,33,34. To perform such a test in a scalable manner (efficient enough to test millions of human genetic variants), we made use of the decomposition of the microbial abundance matrix into 439 orthogonal mPCs. Specifically, we tested each genetic variant for association with each rank-based inverse normal transformed mPC, after which we summed the 439 test statistics obtained per variant to compute a single, combined association test for each variant (Fig. 2a; details in next section). Beyond increasing power to detect pleiotropic effects, the approach reduces multiple-testing burden by testing each genetic variant only once. We evaluated applying this approach to a subset of top axes of microbial variation (rather than all 439 mPCs) but did not observe a further increase in power, consistent with many axes of variation contributing association signal in this dataset (Extended Data Fig. 2f–p).

Details of GWAS of oral microbiome composition

We performed GWAS on each mPC phenotype using the linear mixed model implemented in BOLT-LMM to account for the family structure of the SPARK cohort^30,31,73. Specifically, we ran BOLT-LMM using the --lmmInfOnly flag (as the non-infinitesimal mixed model provided a negligible increase in statistical power) with the following covariates: sequencing batch, age, age squared, square root of age, sex, percentage of mapped reads and the top ten genetic ancestry PCs. A single father without a recorded age was assigned the average age of other fathers in the dataset. AMY1 and PRB1 copy numbers were rescaled to a range of [0,2] and encoded as dosages for association testing.

To test a genetic variant for association with an effect on overall oral microbiome composition, we summed chi-square statistics across the 439 orthogonal mPCs and computed the P value based on a chi-squared distribution with 439 degrees of freedom. We computed P values using a one-sided test, analogous to how in linear regression, one-sided chi-squared test statistics are computed (corresponding to two-sided tests of z-statistics).

MDMR of oral microbiomes with selected genetic variants

We compared our test for genetic effects on oral microbiome composition with multivariate distance matrix regression (MDMR)³³ as implemented in the MDMR R package³⁶ (v.0.5.2), which finds significant predictors of multivariate outcomes by estimating the attributable amount of dissimilarity between samples. Rank-based inverse normal transformed relative abundances of the 439 most prevalent species (with or without initial centred log-ratio transformation) were used to generate the Euclidean distance matrix. MDMR was then run with the following covariates: sequencing batch, age, age squared, square root of age, sex and percentage of mapped reads. As genetic ancestry PCs frequently produced a singular matrix as a result of multicollinearity with individual variants, the top ten genetic PCs were first regressed from each tested variant (rather than including genetic PCs as covariates). The 11 loci identified from our mPC-based GWAS of oral microbiome composition were tested alongside 1,000 randomly selected variants on chromosome 1.

Stratified LD score regression for estimating enrichment of heritability at genes with tissue-specific expression

We observed that the same mathematical framework that enables partitioning of heritability by means of stratified LD score regression on summary statistics from GWAS of a single trait⁷⁴ could be extended to analyse test statistics for association with oral microbiome composition (based on summing chi-squared test statistics across 439 mPCs). Starting from the representation of expected marginal chi-square association statistic for SNP i based on linkage disequilibrium with variants in categories C_k,

$$E[{\,\chi }_{i}^{2}]=1+{Na}+N\sum _{k}{\tau }_{k}l(i,k)$$

averaging across the 439 chi-square statistics for each variant gives

$$E[\bar{{\chi }_{i}^{2}}]=1+N\bar{a}+N\sum _{k}\bar{{\tau }_{k}}l(i,k)$$

such that providing $\bar{{\chi }_{i}^{2}}$ as input to S-LDSC generates enrichments corresponding to $\bar{{\tau }_{k}}$. Averaged chi-square statistics per variant were used as input to munge_sumstats.py. LDSC was then run with baseline v.1.2, weights_hm3_no_hla as weights and previously described tissue-specific expression bins derived from Genotype-Tissue Expression (GTEx) project samples³⁷.

GWAS of abundances of individual taxa

To avoid test statistic inflation from zero inflation and outlier values, relative abundance measures for each taxon were rank-based inverse normal transformed. Abundances of 1,262 taxa (of any phylogenetic level: species, genus, family and so on) observed in >10% of SPARK samples were then tested for association with host genotypes using BOLT-LMM. The top 20 PCs from PCA on 439 species observed at >10% prevalence (that is, the top 20 mPCs) were used as covariates to control for the largest axes of variation across samples, along with sequencing batch, age, age squared, square root of age, sex, percentage of mapped reads and the top ten genetic ancestry PCs. To test loci that might be associated as dominant/recessive rather than additive, BOLT-LMM was rerun using the --domRecHetTest flag.

To evaluate whether some of these associations could reflect compositional effects rather than being specific to the associated taxa, we computed an alternative set of taxon abundance phenotypes in which we took the centred log-ratio transform of relative abundances in each sample⁷⁵ (after replacing zero values observed for a given taxon with the minimum non-zero value for that taxon, to allow computing geometric means). Centred log-ratio transformed values for each taxa were then tested for association with host genotypes using BOLT-LMM with the same covariates as above (Extended Data Fig. 4e).

For estimating effect sizes of specific genotype values such as FUT2 W154X genotypes (Fig. 2d), AMY1 copy numbers (Fig. 3c) or PRB1 copy numbers (Extended Data Fig. 3c), we used linear regression with the same covariates as above, encoding each genotype value (rounded if necessary) as a separate factor. The standard errors estimated by these regressions are slightly underestimated because they do not account for relatedness among SPARK participants, but we determined that this underestimation of standard errors was mild (~7% based on a ~14% inflation of chi-square test statistics computed using linear regression versus a linear mixed model for the 11 genome-wide significant loci (Supplementary Fig. 2)). To compare effect sizes in adults versus unrelated children, one child was randomly selected from each family in SPARK. BOLT-LMM was used to run linear regression on each of these subsets separately.

GWAS in UK Biobank

Starting from 488,377 individuals in the UKB SNP-array dataset⁴³, individuals were excluded on the basis of the following criteria: 36,008 were removed to drop one relative in pairs of close relatives with kinship coefficient >0.0884, preferentially keeping individuals if they (1) reported having dentures or (2) reported not having dentures (that is, had a non-missing dentures phenotype); 28,701 were removed for not having European genetic ancestry⁷⁶; 1,469 were removed for not having available TOPMed-imputed genotypes (including for chromosome X); 2,601 were removed for not having available WGS data; and 53 were removed for having withdrawn, leaving 419,545 available individuals for GWAS. For the binary oral health phenotypes (dentures use and bleeding gums), 418,039 had non-missing values. For the quantitative BMI z-score phenotype⁷⁷, 418,150 had non-missing values.

TOPMed-imputed variants for these individuals were filtered to require minor allele frequency >0.001 and INFO >0.3. BOLT-LMM was run in linear regression mode on these samples and variants with the following covariates: age, age squared, sex, genotype array, assessment centre and top 20 genetic ancestry PCs. For estimating effect sizes of specific copy numbers of AMY1 (Fig. 3d,e,j), AMY2A (Extended Data Fig. 6f) or AMY2B (Extended Data Fig. 6g), we performed logistic regression (for oral health phenotypes) or linear regression (for BMI) with the same covariates, encoding each copy number (rounded to the nearest integer) as a separate factor and using the modal copy number as the reference level.

Phyletic stratification of genetic associations

The phylogenetic tree of all species in the MetaPhlAn 4 database used (v.Oct22), mpa_vOct22_CHOCOPhlAnSGB_202212.nwk, was first subsetted to the tree spanning nodes with primary label among the 439 species seen at >10% prevalence in the SPARK cohort. This tree was used with graphlan (v.1.1.3) for depiction of phylogenetic trees (Figs. 1b and 2e and Extended Data Fig. 5f). For comparisons among the effect sizes of a human genetic variant associated with relative abundances of many species, phylogenetic distances between pairs of species were first computed as a cophenetic distance matrix from this tree. For a given index species A, phylogenetic distances between A and other species B were then compared with either (1) absolute values of effect sizes for species B (that is, |β_B|, Extended Data Fig. 5b,c,g,h) or (2) effect sizes for species B oriented relative to the effect direction for species A (that is, sign(β_A) × β_B, Extended Data Fig. 5d,e,i,j).

Estimation of AMY1 copy number in all cohorts

In the SPARK (Extended Data Fig. 6a) and AoU v7 cohorts (Extended Data Fig. 6c), AMY1 copy number was estimated by counting WGS reads that mapped to the duplicated regions that include AMY1A (chr. 1: 103638545–103666411), AMY1B (chr. 1: 103685558–103713427) and AMY1C (chr. 1: 103732687–103760549) in GRCh38 and normalizing against the total number of reads that aligned in either the 0.5 Mb upstream of AMY2B or the 0.5 Mb downstream of AMY1C. For the UKB cohort (n = 490,415 (ref. ⁷⁸), Fig. 3a), we applied a more comprehensive read-depth normalization pipeline that incorporated sample-specific GC-bias correction inferred from genome-wide alignments (similar to Genome STRiP⁷⁹) before normalizing against read depth in the 0.5 Mb regions flanking the amylase locus. We corrected for slight miscalibration of these diploid copy number estimates by fitting a linear model to identify coefficients that centred peaks of copy number estimates at integers.

Among the UKB participants with WGS available, we identified 5,149 siblings that shared both amylase haplotypes IBD2 (based on at most three mismatching SNP-array genotypes in a 2 Mb window flanking the amylase locus, computed using plink1.9 --genome). Among these IBD2 sibling pairs, 13 pairs were identified as copy number discordant (and likely to reflect a copy number mutation in the past generation) based on (1) having AMY1 copy number estimates that differed by >1.0 and (2) having AMY2A copy number estimates consistent with a duplication or deletion of a commonly variable amylase gene cassette (that is, ±1 AMY2A copies for AMY1 copy number discordances of odd parity and no difference in AMY2A copy number for AMY1 copy number discordances of even parity). We estimated AMY1 copy number genotyping accuracy by computing the correlation across IBD2 sibling pairs excluding these 13 copy number discordant pairs.

Testing AMY1 copy number for association with dental phenotypes and BMI in All of Us

We defined the binary tooth loss phenotype as 1 for individuals with at least one recorded instance of ‘acquired absence of all teeth’ (OMOP concept ID: 40481327, code: 441935006) and 0 otherwise. Likewise, the caries phenotype was derived from ‘dental caries’ (OMOP concept ID: 133228, code: 80967001). For BMI, we generated a normalized z-score phenotype from BMI (OMOP concept ID: 903124, code: bmi) by first adjusting for age (in months, determined from time at weight and height measurement) and age squared and then applying inverse rank normal transformation in each sex separately. The BMI z-scores for males and females were then merged together.

Individuals with WGS available for genotyping AMY1 copy number were first filtered to an unrelated subset of samples (iteratively dropping one individual per related pair with kinship score >0.1, from relatedness_flagged_samples.tsv). For oral health phenotypes, we performed logistic regression against AMY1 copy number including age, age squared, sex and the top 16 genetic ancestry PCs (from ancestry_preds.tsv) as covariates. For BMI, we performed linear regression using only genetic ancestry PCs as covariates as age and sex had already been residualized out. For estimating effect sizes of specific copy numbers of AMY1 (Extended Data Fig. 6d,e and Extended Data Fig. 7f–i), we performed logistic regression (for oral health phenotypes) or linear regression (for BMI) with the same covariates, encoding each copy number (rounded to the nearest integer) as a separate factor and using the modal copy number of 6 as the reference level (that is, computing the effect size of each copy number relative to copy number 6).

Paralogous sequence variation in AMY1

To identify and genotype paralogous sequence variants (PSVs) from UKB WGS data, we used a read-counting approach similar to our previous work⁸⁰. In brief, reads from each sample that had been aligned to any of the three 27.6 kb regions in GRCh38 corresponding to AMY1A (chr. 1: 103638695–103666261), AMY1B (chr. 1: 103685708–103713277) and AMY1C (chr. 1: 103732837–103760399) were realigned with bwa (v.0.7.17) to the reference sequence of AMY1A after filtering out reads with any of the last four SAM flags (-F 0xF00). Read counts supporting each base at each position were tabulated with htsbox (r345) pileup, filtering alignments <50 bp and base calls with quality score <20. Individuals were called heterozygous for a PSV allele (having at least one copy of AMY1 with each of two alleles) if at least five reads supported the variant allele in that sample and at least five reads did not. PSVs were then filtered to those with heterozygosity >0.002 (resulting in 892 PSVs passing filters). To estimate diploid copy number genotypes for a PSV, we multiplied each individual’s diploid AMY1 copy number by the allelic fraction of the PSV in that individual. In association tests using linear regression with BOLT-LMM, we rounded copy number estimates to integer genotypes.

For follow-up analyses of effect sizes of specific AMY1 F141C and C477R copy number genotypes, we optimized the assignments of integer copy number genotypes based on manual inspection of histograms of allelic depth-derived PSV copy number estimates (Supplementary Fig. 3). Specifically, we assigned copy numbers for each of F141C and C477R using the thresholds [0,0.25) = CN0, [0.25,1.75) = CN1, [1.75,2.7) = CN2, [2.7,3.5) = CN3 and [3.5,5) = CN4. In UKB, F141C had 416,381 (99.2%), 3,124 (0.74%) and 40 (0.0095%) individuals with 0, 1 and 2 copies, respectively, and C477R had 412,450 (98.3%), 6,527 (1.6%), 547 (0.13%), 19 (0.0045%) and 2 (0.0005%) individuals with 0, 1, 2, 3 and 4 copies, respectively. In SPARK, F141C had 12,459 (99.5%) and 60 (0.48%) individuals with 0 and 1 copies, respectively, and C477R had 12,343 (98.6%), 172 (1.4%) and 4 (0.032%) individuals with 0, 1 and 2 copies, respectively. These threshold-based copy numbers of the alternate alleles were used in logistic regression along with copy numbers of the reference alleles (F141 and C477) and covariates as above.

Protein expression of AMY1

Plasmid pCAGEN⁸¹ was a gift from C. Cepko (Addgene plasmid no. 11160; http://n2t.net/addgene:11160; RRID Addgene_11160). Codon-optimized sequences encoding reference, F141C and C477R AMY1 alleles were synthesized and ordered as gBlocks from Integrated DNA Technologies for cloning into pCAGEN downstream of the CAG promoter. Clones were screened for sequence errors before plasmid preparation using Plasmid Plus Midi Kit (Qiagen, catalogue no. 12943). Plasmid pUC19 (New England Biolabs, catalogue no. N3041S) was used as a negative control. A total of 15 µg of each plasmid was lipofected into separate 10 cm plates of HEK293T (Takara, catalogue no. 632180) cells at ~70% confluence using 30 µl of Lipofectamine 3000 (Invitrogen, catalogue no. L3000015). Authentication of HEK293T was done by morphological match for type and verification of SV40T antigen with PCR assay. Lack of mycoplasma contamination was confirmed by Takara as well as inhouse with MycoAlert Mycoplasma Detection Kit (Lonza, catalogue no. LT07-318). Medium was switched to serum-free after 24 h before collection of both supernatant and lysate at 72 h post-lipofection. A total of 10 ml of supernatant was spun at 1,000g for 10 min to remove cells and debris before the addition of 100 µl of Halt Protease Inhibitor Cocktail (100×, Thermo Scientific, catalogue no. 78438) and 100 µl of EDTA (0.5 M). Cells were washed with ice-cold PBS before the addition of cold 1 ml of RIPA Lysis and Extraction Buffer (Thermo Scientific, catalogue no. 89900) with 10 µl of Halt Protease Inhibitor Cocktail (100×) and 10 µl of EDTA (0.5 M). After sufficient solubilization of the cells had occurred, 1 µl of Benzonase Nuclease (250 U per µl, Milipore, catalogue no. E1014) and 10 µl of MgCl₂ (1 M) were added before incubation at 37 °C, 500 rpm for 30 min. Lysate was then spun at 10,000g for 10 min to remove insoluble precipitate. Both supernatant and lysate were stored at −80 °C until further use.

Western blot of AMY1 in cell culture supernatant and lysate

A total of 7.5 µl of supernatant or purified lysate was first run denatured and reduced in a 10% Mini-PROTEAN TGX Precast Protein Gel (Bio-Rad, catalogue no. 4561036) before wet transfer to nitrocellulose membrane (120 V, 2 h). After evaluation of equal loading by Ponceau S Staining Solution (Thermo Scientific, catalogue no. A40000279), the membrane was blocked for 1 h at room temperature with TBS, 0.1% Tween-20 and 5% w/v non-fat dry milk before washing three times for 5 min each with TBS-T (TBS, 0.1% Tween-20). Amylase antibody (G-10, Santa Cruz Biotechnology, catalogue no. sc-46657, lot no. G0324) was used as the primary antibody at a 1:200 dilution in TBS-T with 5% w/v milk for incubation overnight at 4 °C with rotation. Membrane was washed three times for 5 min each with TBS-T before addition of anti-mouse IgG, HRP-linked antibody (Cell Signaling Technology, catalogue no. 7076, lot no. 39) as secondary at a 1:2,000 dilution in TBS-T with 5% w/v milk for 1 h incubation at room temperature with rotation. Membrane was washed three times for 5 min each with TBS-T before detection with Amersham ECL Prime Western Blotting Detection Reagent (Cytiva, catalogue no. RPN2236) (Extended Data Fig. 7a).

Purification of AMY1 reference and F141C protein

Purification of amylase from supernatant was done using glycogen, adapted from refs. ^82,83. All following steps were conducted on ice or at 4 °C (rotation and centrifugation). In brief, 10 ml of supernatant was initially concentrated using prewet 15 ml Amicon Ultra Centrifugal Filter, 30 kDa MWCO (Milipore, catalogue no. UFC9030) and put up to a total volume of 900 µl with PBS. A total 600 µl of cold ethanol was added slowly to make 40% ethanol v/v. This was then centrifuged at 10,000g for 10 min to remove any insoluble precipitate. To the supernatant, 75 µl of 0.2 M sodium phosphate buffer (pH 8), 75 µl of glycogen (20 mg ml⁻¹, Roche, catalogue no. 10901393001) and 100 µl of ethanol were added in that order. This was then incubated for 5 min with end-over-end mixing before centrifugation at 5,000g for 6 min. The glycogen pellets were then washed twice with 10 mM sodium phosphate buffer (pH 8) containing ethanol (40% v/v) before resuspension in 100 µl of 50 mM MOPS buffer (pH 7, Thermo Scientific, catalogue no. J61821-AK) with 5 mM CaCl₂. These were then incubated at 37 °C, 500 rpm for 30 min to allow for glycogen digestion. The above purification procedure was repeated as necessary to arrive at purified amylase as evaluated on 10% Mini-PROTEAN TGX Precast Protein Gel with InstantBlue Coomassie Protein Stain (Abcam, catalogue no. ab119211) (Extended Data Fig. 7b).

Amylase activity assay

Protein was normalized between reference and F141C isoforms by densitometric measurement against linear dilution series (r² = 0.99 with input). The concentration chosen for the assay was determined by having maximal linear activity over the observation window. For each technical replicate, 10 µl of reference or F141C amylase enzyme diluted in 50 mM MOPS buffer (pH 7), 5 mM CaCl₂ and 0.02% w/v BSA (New England Biolabs, catalogue no. B9000) was quickly mixed with 10 µl of BODIPY FL conjugated-starch substrate from EnzChek Ultra Amylase Assay Kit (Invitrogen, catalogue no. E33651) in 50 mM MOPS buffer (pH 7) with 5 mM CaCl₂. The reaction was then maintained at 20 °C for 2 h in a Bio-Rad CFX384 Real-Time PCR Detection System with fluorescence reading taken every minute using FAM fluorophore settings with CFX Manager (v.3.1) software. Fluorescence at 30 min relative to initial reading was used as input to a linear model with allele and plate to regress out any run-to-run effects during comparison across technical replicates (Extended Data Fig. 7c).

Bacterial genome reference panel for analysing bacterial gene dosages

To reduce hypothesis testing burden in association analyses of human genetic variants with bacterial gene dosage phenotypes, we restricted analyses to a set of 30 bacterial genomes representing highly abundant species and species whose abundances we had found to associate with human genetic variants. Specifically, we selected 30 bacterial species with genomes available in GenBank by including: (1) the five most abundant species in SPARK oral microbiomes, (2) species that associated strongly (P < 4 × 10⁻¹¹) with at least one human genetic variant and (3) the top two associated species for each locus if not already included. We substituted Stomatobaculum longum for the related Stomatobaculum SGB5266 which would have been included under (2) but lacked a GenBank assembly. We selected a single genome for each of the 30 species using the following criteria. The SGB centroid was prioritized over other genomes if it corresponded to a GenBank assembly. For cases in which the centroid was not a GenBank assembly and multiple genomes corresponded to GenBank assemblies, the reference genome was prioritized if available, or otherwise the highest ranked genome among those listed. A list of these species and GenBank assemblies is included in Supplementary Table 6. A bowtie2 index was then built from these merged genomes.

Measuring bacterial gene dosages using read-depth phenotypes

We computed WGS coverage-derived phenotypes informative of gene dosage across each of the 30 bacterial reference genomes by first realigning unmapped reads from SPARK saliva WGS to the 30 reference genomes using bowtie2 (v.2.5.1) with the --very-sensitive flag. These alignments were then position-sorted within contigs with samtools (v.1.15.1). For each WGS sample, we quantified read depth in 500 bp bins tiling each of the 30 bacterial reference genomes using mosdepth (v.0.3.6), excluding reads with mapping quality <5. For each sample, for each of the 30 bacterial reference genomes, we then median-normalized the bin-level read-depth measurements across the 500 bp bins of that reference genome to control for species abundance (such that normalized read-depth measurements had a median of 1 among bins corresponding to each species). If a sample had <0.5× median coverage across bins corresponding to a given species, we set that sample’s normalized read-depth measurements for that species to missing to focus downstream analyses on samples with less-noisy measurements. Finally, we truncated median-normalized read-depth values to the interval [0,1], both to focus on deletions in bacterial genomes and to reduce the influence of outlier measurements that might reflect mismapped reads (potentially derived from either duplicated genomic regions or homologous sequences in microbial species not represented among the 30 reference genomes). We reasoned that these bin-level measurements would capture kilobase-scale deletions of bacterial genomes, circumventing the need to predefine a set of structural variant regions (which was difficult because of the limited sequencing coverage of most species). Additional details are provided in Supplementary Note 5.

Testing bacterial gene dosage phenotypes for association with host genotypes

We used linear regression to test each of the 11 human genetic variants we had found to associate with oral microbiome composition (Extended Data Table 1) for association with normalized read depth (truncated to [0,1]) in each 500 bp bin of each of the 30 bacterial reference genomes. We used an additive model for all variants except FUT2 W154X and TLR1 I602S, for which we used a recessive model (corresponding to secretor/non-secretor status for FUT2). We took two precautions to avoid potential confounders. First, for each of the 30 species, we included as covariates the top 20 PCs of the normalized, truncated read-depth matrix for that species (running PCA after centring and scaling each bin to have a mean of 0 and s.d. of 1 across samples) to control for linked gene dosages (for example, differences across strains) that could potentially generate non-causal associations in a manner analogous to population structure in GWAS (Supplementary Note 5). Second, we applied a form of genomic control⁸⁴ (applied across the 500 bp bins of each reference genome) to adjust for remaining test statistic inflation. Specifically, for each pairing of a species and a human genetic variant, we computed the adjustment factor

$$\frac{{\rm{median}}\,({\chi }_{1}^{2})}{{F}^{-1}(0.5)}$$

across the ${\chi }_{1}^{2}$ test statistics for the 500 bp bins of that reference genome, where F⁻¹(x) is the inverse cumulative distribution function for a ${\chi }_{1}^{2}$ random variable. The ${\chi }_{1}^{2}$ test statistics were then divided by this factor. This yielded 208 read-depth bins in 18 species that significantly associated with at least one of the 11 human genetic variants (FDR < 0.01, Supplementary Tables 7 and 8) and resolved to 68 unique microbial regions after merging bins within 1.5 kb of another significantly associated bin (Fig. 4b). We verified that the truncated normalized read-depth phenotypes involved in these associations were broadly reasonably distributed (with most bimodal at 0 and 1 and mean between 0.05 and 0.95; Supplementary Table 7), such that testing these phenotypes for association with common variants (MAF = 0.08–0.45) using linear regression in a cohort of size 12,519 was expected to produce robust test statistics⁷⁷.

For combinatorial analysis of deletions of the five vad genes in V. sp. 3627 that associated in the same direction with FUT2 genotype (vadB through vadF) (Fig. 6e), we first selected individuals whose oral microbiomes had evidence of near-complete presence (>0.8 median-normalized coverage) or near-complete absence (<0.2 median-normalized coverage) of each vad gene (n = 3,081, representing roughly half of the SPARK samples with coverage of the V. sp. 3627 genome reaching the >0.5× threshold for analysis). For each common combination (>2.5% of selected individuals) of presence/absence status of vadB through vadF, the number of individuals with each FUT2 W154X genotype were then counted.

Replication of bacterial gene dosage associations in the All of Us cohort

To evaluate the generalizability of the associations we identified in SPARK between human genetic variants and bacterial gene dosages, we applied the same computational pipeline described above to 10,000 randomly selected saliva-derived WGS samples from the AoU v.8 data release. We then attempted to replicate the 208 associations (between human genetic variants and normalized read-depth measurements in 500 bp bins of bacterial genomes) that had reached significance (FDR < 0.01) in SPARK.

AlphaFold3 prediction of protein structures

To predict protein structures for Streptococcus parasanguinis AbpA or AbpB (bound to human AMY1), the Veillonella sp. 3627 VadD trimer and Streptococcus mitis CrpE, we used AlphaFold3 (ref. ⁸⁵) for multimer prediction with default reference databases and max template date of 2021-09-29. The TonB domain of AbpB and the signal peptide of VadD were excluded for visualization. To minimize the model size, CrpE was truncated to the region spanning the CshA NR2 domain to the sixth mucin-binding domain (residues 572–2701). Structures were visualized with ChimeraX (v.1.9)⁸⁶ and pLDDT, pTM and ipTM values can be found in Supplementary Table 9.

Genetically derived blood typing in SPARK

We assigned blood types to SPARK participants on the basis of WGS-derived SNP and indel genotypes using a procedure similar to previous work¹². The genotype of the rs8176746 missense SNP was first used to determine an individual’s dosage (that is, allele count) of type B alleles (T allele count) and non-type B dosage (G allele count).

The rs8176719 indel was next used to determine type O1 dosage (deletion allele count), which was subtracted from the non-type B dosage to yield non-type B/O1 dosage, as the rs8176719 deletion allele typically occurs on haplotypes that would otherwise be type A alleles. Although this is true in European, East Asian and American ancestry haplotypes in 1KGP populations, in a small fraction of African (3.2%) and South Asian (0.2%) ancestry haplotypes, the rs8176719 deletion occurs in cis with the type B missense allele. As we found 36 SPARK participants who had O1 dosage exceeding non-type B dosage, we subtracted this excess from their type B dosages.

The rs41302905 missense SNP was next used to determine type O2 dosage (T allele count) and subtracted from the non-type B/O1 dosage to yield type A dosage, as it seems to be in cis with type A alleles in all 1KGP populations. O1 and O2 dosages were then merged to compute type O dosage.

The rs56392308 indel was next used to determine the type A2 dosage (deletion allele count) and subtracted from the type A dosage to yield type A1 dosage. For seven individuals in which the type A2 dosage exceeded type A dosage, five seemed to be on type O alleles and two on either type O or type B alleles, so this excess was subtracted from their type A2 dosage.

Enrichment of conserved domains in bacterial genes associated with secretor status

For each of the bacterial species with adhesin genes in dosage variable regions that associated with FUT2 loss of function (Fig. 6a–c,f,g), protein IDs (WP numbers) for the species were extracted from its RefSeq general feature format (GFF) file. Conserved domains (from National Center for Biotechnology Information (NCBI) conserved domain database) were identified for each protein using a modified version of the provided bwrpsb.pl script (applied to up to 250 proteins at a time). A one-sided Fisher’s exact test was used to identify domains enriched among proteins encoded by genes within read-depth bins that associated with FUT2 genotype.

GWAS of read depth in selected microbial genomic regions

We used BOLT-LMM to perform GWAS on the five normalized read-depth phenotypes for which we observed the most significant associations with human genetic variants (in our targeted analysis of 11 human genetic variants). Specifically, these phenotypes measured WGS read depth in the following 500 bp bins: H. sputorum QEQH01000003.1: 197,000–197,500, P. nanceiensis KB904333.1: 123,500–124,000, S. mitis MUYN01000003.1: 100,000–100,500, S. vestibularis AEKO01000011.1: 186,000–186,500, V. sp. 3627 RQVG01000009.1: 13,500–14,000. In each GWAS, we included as covariates the top 20 PCs of the normalized, truncated read-depth matrix for the species under consideration, along with sequencing batch, age, age squared, square root of age, sex, percentage of mapped reads and the top ten human genetic PCs.

We also used BOLT-LMM to perform GWAS on normalized read-depth measurements in 500 bp bins spanning the genome of R. mucilaginosa (the most prevalent species observed in SPARK). To avoid test statistic inflation due to non-normality, we first excluded bins for which <10% of samples had non-modal read-depth values, leaving 3,441 bins. We then rank-based inverse normal transformed these bins (with random tie-breaking) to further normalize the phenotypes. We ran BOLT-LMM using the same covariates as above.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The following data resources are available by application: UKB (http://www.ukbiobank.ac.uk/), All of Us Research Program (https://allofus.nih.gov/) and SFARI SPARK (https://www.sfari.org/resource/spark/). To protect participant confidentiality, approved researchers can obtain access to the SPARK population dataset described in this study (SPARK integrated WGS (iWGS) v.1.1) by applying at https://base.sfari.org. Quantifications of microbial abundances in SPARK generated in this study can also be obtained from SFARI Base (Dataset DS0000116). Summary statistics from GWAS of microbial abundances in SPARK are available from the GWAS Catalog under accessions GCST90709872 to GCST90711133. Summary statistics from mPC-based GWAS of oral microbiome composition are available at Zenodo (https://doi.org/10.5281/zenodo.14559457)⁸⁷. The following data resources are publicly available: Human Microbiome Project (https://hmpdacc.org/), human reference genome build GRCh38 (https://ftp.1000genomes.ebi.ac.uk/vol1/ftp/technical/reference/GRCh38_reference_genome/), MetaPhlAn v.Oct22 reference database (http://cmprod1.cibio.unitn.it/biobakery4/metaphlan_databases/), TOPMed-r3 imputation panel variant list (https://imputation.biodatacatalyst.nhlbi.nih.gov/), LD score resources https://alkesgroup.broadinstitute.org/LDSCORE/), NCBI GenBank (https://www.ncbi.nlm.nih.gov/genbank/) and NCBI Conserved Domain Database (https://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml).

Code availability

The following publicly available software resources were used: MetaPhlAn (v.4.0.6, http://segatalab.cibio.unitn.it/tools/metaphlan/index.html), DeepVariant (v.1.3.0, https://github.com/google/deepvariant), GLnexus (v.1.4.1, https://github.com/dnanexus-rnd/GLnexus), HUMAnN (v.3.8, https://huttenhower.sph.harvard.edu/humann), GraPhlAn (v.1.1.3, http://segatalab.cibio.unitn.it/tools/graphlan/index.html), mosdepth (v.0.3.6, https://github.com/brentp/mosdepth), bowtie (v.2.5.1, https://bowtie-bio.sourceforge.net/bowtie2/index.shtml), bcftools (v.1.14, http://www.htslib.org/), samtools (v.1.15.1, http://www.htslib.org/), plink (v.1.90b6.26 and v.2.00a3.7, https://www.cog-genomics.org/plink/), BOLT-LMM (v.2.4.1, https://alkesgroup.broadinstitute.org/BOLT-LMM/), qqman (v.0.1.8, https://cran.r-project.org/web/packages/qqman/index.html), MDMR (v.0.5.2, https://cran.r-project.org/web/packages/MDMR/index.html), bedtools (v.2.27.1, https://bed-tools.readthedocs.io/en/latest/), AlphaFold3 (https://alphafoldserver.com/) and ChimeraX (v.1.9, https://www.cgl.ucsf.edu/chimerax/). Custom code used to generate results in this study is available at Zenodo (https://doi.org/10.5281/zenodo.14559457)⁸⁷.

References

Gomez, A. et al. Host genetic control of the oral microbiome in health and disease. Cell Host Microbe 22, 269–278 (2017).
Article CAS PubMed PubMed Central Google Scholar
Esberg, A., Haworth, S., Kuja-Halkola, R., Magnusson, P. K. E. & Johansson, I. Heritability of oral microbiota and immune responses to oral bacteria. Microorganisms 8, 1126 (2020).
Article CAS PubMed PubMed Central Google Scholar
Demmitt, B. A. et al. Genetic influences on the human oral microbiome. BMC Genomics 18, 659 (2017).
Article PubMed PubMed Central Google Scholar
Liu, X. et al. Metagenome-genome-wide association studies reveal human genetic impact on the oral microbiome. Cell Discov. 7, 117 (2021).
Article PubMed PubMed Central Google Scholar
Stankevic, E. et al. Genome-wide association study identifies host genetic variants influencing oral microbiota diversity and metabolic health. Sci. Rep. 14, 14738 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Dobell, C. Antony van Leeuwenhoek and his ‘Little Animals’: Being Some Account of the Father of Protozoology and Bacteriology and his Multifarious Discoveries in These Disciplines (Harcourt Brace, 1932).
Dewhirst, F. E. et al. The human oral microbiome. J. Bacteriol. 192, 5002–5017 (2010).
Article CAS PubMed PubMed Central Google Scholar
Odendaal, M.-L. et al. Host and environmental factors shape upper airway microbiota and respiratory health across the human lifespan. Cell 187, 4571–4585 (2024).
Article CAS PubMed Google Scholar
Baker, J. L., Mark Welch, J. L., Kauffman, K. M., McLean, J. S. & He, X. The oral microbiome: diversity, biogeography and human health. Nat. Rev. Microbiol. 22, 89–104 (2024).
Article CAS PubMed Google Scholar
Kurilshikov, A. et al. Large-scale association analyses identify host factors influencing human gut microbiome composition. Nat. Genet. 53, 156–165 (2021).
Article CAS PubMed PubMed Central Google Scholar
Rühlemann, M. C. et al. Genome-wide association study in 8,956 German individuals identifies influence of ABO histo-blood groups on gut microbiome. Nat. Genet. 53, 147–155 (2021).
Article PubMed Google Scholar
Lopera-Maya, E. A. et al. Effect of host genetics on the gut microbiome in 7,738 participants of the Dutch Microbiome Project. Nat. Genet. 54, 143–151 (2022).
Article CAS PubMed Google Scholar
Qin, Y. et al. Combined effects of host genetics and diet on human gut microbiota and incident disease in a single population cohort. Nat. Genet. 54, 134–142 (2022).
Article CAS PubMed PubMed Central Google Scholar
Pitts, N. B. et al. Dental caries. Nat. Rev. Dis. Primer 3, 17030 (2017).
Article Google Scholar
Thomas, T. T. III. Ludwig’s angina (part II): an anatomical, clinical and statistical study. Ann. Surg. 47, 335–373 (1908).
Article CAS PubMed PubMed Central Google Scholar
Wang, X. et al. Genes and their effects on dental caries may differ between primary and permanent dentitions. Caries Res. 44, 277–284 (2010).
Article CAS PubMed PubMed Central Google Scholar
Haworth, S. et al. Heritability of caries scores, trajectories, and disease subtypes. J. Dent. Res. 99, 264–270 (2020).
Article CAS PubMed Google Scholar
Shungin, D. et al. Genome-wide analysis of dental caries and periodontitis combining clinical and self-reported data. Nat. Commun. 10, 2773 (2019).
Article ADS PubMed PubMed Central Google Scholar
Bordenstein, S. R. & Theis, K. R. Host biology in light of the microbiome: ten principles of holobionts and hologenomes. PLoS Biol. 13, e1002226 (2015).
Article PubMed PubMed Central Google Scholar
Kilian, M. et al. The oral microbiome—an update for oral healthcare professionals. Br. Dent. J. 221, 657–666 (2016).
Article CAS PubMed Google Scholar
Chen, L. et al. The long-term genetic stability and individual specificity of the human gut microbiome. Cell 184, 2302–2315 (2021).
Article CAS PubMed Google Scholar
Suzuki, T. A. et al. Codiversification of gut microbiota with humans. Science 377, 1328–1332 (2022).
Article ADS CAS PubMed PubMed Central Google Scholar
Zeevi, D. et al. Structural variation in the gut microbiome associates with host health. Nature 568, 43–48 (2019).
Article ADS CAS PubMed Google Scholar
Zhernakova, D. V. et al. Host genetic regulation of human gut microbial structural variation. Nature 625, 813–821 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
The SPARK Consortium. SPARK: A US cohort of 50,000 families to accelerate autism research. Neuron 97, 488–493 (2018).
Manghi, P. et al. Large-scale metagenomic analysis of oral microbiomes reveals markers for autism spectrum disorders. Nat. Commun. 15, 9743 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Valles-Colomer, M. et al. The person-to-person transmission landscape of the gut and oral microbiomes. Nature 614, 125–135 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Samson, C. A., Whitford, W., Snell, R. G., Jacobsen, J. C. & Lehnert, K. Contaminating DNA in human saliva alters the detection of variants from whole genome sequencing. Sci. Rep. 10, 19255 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Blanco-Míguez, A. et al. Extending and improving metagenomic taxonomic profiling with uncharacterized species using MetaPhlAn 4. Nat. Biotechnol. 41, 1633–1644 (2023).
Article ADS PubMed PubMed Central Google Scholar
Yu, J. et al. A unified mixed-model method for association mapping that accounts for multiple levels of relatedness. Nat. Genet. 38, 203–208 (2006).
Article CAS PubMed Google Scholar
Loh, P.-R. et al. Efficient Bayesian mixed-model analysis increases association power in large cohorts. Nat. Genet. 47, 284–290 (2015).
Article CAS PubMed PubMed Central Google Scholar
Zhao, N. et al. Testing in microbiome-profiling studies with MiRKAT, the microbiome regression-based kernel association test. Am. J. Hum. Genet. 96, 797–807 (2015).
Article CAS PubMed PubMed Central Google Scholar
Anderson, M. J. A new method for non-parametric multivariate analysis of variance. Austral Ecol. 26, 32–46 (2001).
Google Scholar
Calle, M. L. Statistical analysis of metagenomics data. Genomics Inform. 17, e6 (2019).
Article MathSciNet PubMed PubMed Central Google Scholar
Rumker, L. et al. Identifying genetic variants that influence the abundance of cell states in single-cell data. Nat. Genet. 56, 2068–2077 (2024).
Article CAS PubMed PubMed Central Google Scholar
McArtor, D. B., Lubke, G. H. & Bergeman, C. S. Extending multivariate distance matrix regression with an effect size measure and the asymptotic null distribution of the test statistic. Psychometrika 82, 1052–1077 (2017).
Article MathSciNet PubMed Google Scholar
Finucane, H. K. et al. Heritability enrichment of specifically expressed genes identifies disease-relevant tissues and cell types. Nat. Genet. 50, 621–629 (2018).
Article CAS PubMed PubMed Central Google Scholar
Whitmore, L. C. et al. A common genetic variant in TLR1 enhances human neutrophil priming and impacts length of intensive care stay in pediatric sepsis. J. Immunol. 196, 1376–1386 (2016).
Article CAS PubMed Google Scholar
Johnson, C. M. et al. Cutting edge: a common polymorphism impairs cell surface trafficking and functional responses of TLR1 but protects against Leprosy1. J. Immunol. 178, 7520–7524 (2007).
Article CAS PubMed Google Scholar
Goumidi, L. et al. Association between ABO haplotypes and the risk of venous thrombosis: impact on disease risk estimation. Blood 137, 2394–2402 (2021).
Article CAS PubMed PubMed Central Google Scholar
Kelly, R. J., Rouquier, S., Giorgi, D., Lennon, G. G. & Lowe, J. B. Sequence and expression of a candidate for the human secretor blood group α(1,2)fucosyltransferase gene (FUT2). Homozygosity for an enzyme-inactivating nonsense mutation commonly correlates with the non-secretor phenotype. J. Biol. Chem. 270, 4640–4649 (1995).
Article CAS PubMed Google Scholar
Mitsiadis, T. A. & Drouin, J. Deletion of the Pitx1 genomic locus affects mandibular tooth morphogenesis and expression of the Barx1 and Tbx1 genes. Dev. Biol. 313, 887–896 (2008).
Article CAS PubMed Google Scholar
Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562, 203–209 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Pajic, P. et al. Independent amylase gene copy number bursts correlate with dietary preferences in mammals. eLife 8, e44628 (2019).
Article CAS PubMed PubMed Central Google Scholar
Perry, G. H. et al. Diet and the evolution of human amylase gene copy number variation. Nat. Genet. 39, 1256–1260 (2007).
Article CAS PubMed PubMed Central Google Scholar
Bolognini, D. et al. Recurrent evolution and selection shape structural diversity at the amylase locus. Nature 634, 617–625 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Yılmaz, F. et al. Reconstruction of the human amylase locus reveals ancient duplications seeding modern-day variation. Science 0, eadn0609 (2024).
Article Google Scholar
Falchi, M. et al. Low copy number of the salivary amylase gene predisposes to obesity. Nat. Genet. 46, 492–497 (2014).
Article CAS PubMed PubMed Central Google Scholar
Hasegawa, T. et al. Impact of salivary and pancreatic amylase gene copy numbers on diabetes, obesity, and functional profiles of microbiome in Northern Japanese population. Sci. Rep. 12, 7628 (2022).
Article ADS CAS PubMed PubMed Central Google Scholar
Usher, C. L. et al. Structural forms of the human amylase locus and their relationships to SNPs, haplotypes and obesity. Nat. Genet. 47, 921–925 (2015).
Article CAS PubMed PubMed Central Google Scholar
Shwan, N. A. A. & Armour, J. A. L. No evidence for association of BMI with salivary amylase gene copy number in the UK 1958 birth cohort. Obesity 27, 1533–1538 (2019).
Article CAS PubMed Google Scholar
Carpenter, D., Mitchell, L. M. & Armour, J. A. L. Copy number variation of human AMY1 is a minor contributor to variation in salivary amylase expression and activity. Hum. Genomics 11, 2 (2017).
Article PubMed PubMed Central Google Scholar
Poole, A. C. et al. Human salivary amylase gene copy number impacts oral and gut microbiomes. Cell Host Microbe 25, 553–564 (2019).
Article CAS PubMed Google Scholar
Bick, A. G. et al. Genomic data in the All of Us research program. Nature 627, 340–346 (2024).
Article Google Scholar
Chapple, I. L. C. et al. Primary prevention of periodontitis: managing gingivitis. J. Clin. Periodontol. 42, S71–S76 (2015).
Article PubMed Google Scholar
Lang, N. P., Schätzle, M. A. & Löe, H. Gingivitis as a risk factor in periodontal disease. J. Clin. Periodontol. 36, 3–8 (2009).
Article PubMed Google Scholar
Adzhubei, I. A. et al. A method and server for predicting damaging missense mutations. Nat. Methods 7, 248–249 (2010).
Article CAS PubMed PubMed Central Google Scholar
Beghini, F. et al. Integrating taxonomic, functional, and strain-level profiling of diverse microbial communities with bioBakery 3. eLife 10, e65088 (2021).
Article CAS PubMed PubMed Central Google Scholar
Tett, A., Pasolli, E., Masetti, G., Ercolini, D. & Segata, N. Prevotella diversity, niches and interactions with the human host. Nat. Rev. Microbiol. 19, 585–599 (2021).
Article CAS PubMed PubMed Central Google Scholar
Bell jr, C. E. & Fortwengler, H. P. Salivary anti-A and anti-B activity of group O males. Vox Sang. 21, 493–508 (1971).
Article CAS PubMed Google Scholar
Atack, J. M. et al. The nontypeable Haemophilus influenzae major adhesin Hia is a dual-function lectin that binds to human-specific respiratory tract sialic acid glycan receptors. mBio 11, e02714-20 (2020).
Article PubMed PubMed Central Google Scholar
Tram, G. et al. The Acinetobacter baumannii autotransporter adhesin Ata recognizes host glycans as high-affinity receptors. ACS Infect. Dis. 7, 2352–2361 (2021).
Article CAS PubMed Google Scholar
Back, C. R. et al. The Streptococcus gordonii adhesin CshA protein binds host fibronectin via a catch-clamp mechanism. J. Biol. Chem. 292, 1538 (2016).
Article PubMed PubMed Central Google Scholar
Chatterjee, M. et al. Understanding the adhesion mechanism of a mucin binding domain from Lactobacillus fermentum and its role in enteropathogen exclusion. Int. J. Biol. Macromol. 110, 598–607 (2018).
Article CAS PubMed Google Scholar
Poole, J., Day, C. J., von Itzstein, M., Paton, J. C. & Jennings, M. P. Glycointeractions in bacterial pathogenesis. Nat. Rev. Microbiol. 16, 440–452 (2018).
Article CAS PubMed Google Scholar
Rothschild, D. et al. Environment dominates over host genetics in shaping human gut microbiota. Nature 555, 210–215 (2018).
Article ADS CAS PubMed Google Scholar
Goodrich, J. K. et al. Human genetics shape the gut microbiome. Cell 159, 789–799 (2014).
Article CAS PubMed PubMed Central Google Scholar
Zahavi, L. et al. Bacterial SNPs in the human gut microbiome associate with host BMI. Nat. Med. 29, 2785–2792 (2023).
Article CAS PubMed PubMed Central Google Scholar
Suerbaum, S. & Josenhans, C. Helicobacter pylori evolution and phenotypic diversification in a changing host. Nat. Rev. Microbiol. 5, 441–452 (2007).
Article CAS PubMed Google Scholar
Divaris, K. et al. Cohort profile: ZOE 2.0—a community-based genetic epidemiologic study of early childhood oral health. Int. J. Environ. Res. Public. Health 17, 8056 (2020).
Article CAS PubMed PubMed Central Google Scholar
Lloyd-Price, J. et al. Strains, functions and dynamics in the expanded Human Microbiome Project. Nature 550, 61–66 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Huttenhower, C. et al. Structure, function and diversity of the healthy human microbiome. Nature 486, 207–214 (2012).
Article ADS CAS Google Scholar
Kang, H. M. et al. Variance component model to account for sample structure in genome-wide association studies. Nat. Genet. 42, 348–354 (2010).
Article CAS PubMed PubMed Central Google Scholar
Finucane, H. K. et al. Partitioning heritability by functional annotation using genome-wide association summary statistics. Nat. Genet. 47, 1228–1235 (2015).
Article CAS PubMed PubMed Central Google Scholar
Tsilimigras, M. C. B. & Fodor, A. A. Compositional data analysis of the microbiome: fundamentals, tools, and challenges. Ann. Epidemiol. 26, 330–335 (2016).
Article PubMed Google Scholar
Hujoel, M. L. A. et al. Insights into the causes and consequences of DNA repeat expansions from 700,000 biobank participants. Preprint at bioRxiv https://doi.org/10.1101/2024.11.25.625248 (2024).
Loh, P.-R., Kichaev, G., Gazal, S., Schoech, A. P. & Price, A. L. Mixed-model association for biobank-scale datasets. Nat. Genet. 50, 906–908 (2018).
Article CAS PubMed PubMed Central Google Scholar
The UK Biobank Whole-Genome Sequencing Consortium. Whole-genome sequencing of 490,640 UK Biobank participants. Nature 645, 692–701 (2025).
Handsaker, R. E. et al. Large multiallelic copy number variations in humans. Nat. Genet. 47, 296–303 (2015).
Article CAS PubMed PubMed Central Google Scholar
Hujoel, M. L. A. et al. Protein-altering variants at copy number-variable regions influence diverse human phenotypes. Nat. Genet. 56, 569–578 (2024).
Article CAS PubMed PubMed Central Google Scholar
Matsuda, T. & Cepko, C. L. Electroporation and RNA interference in the rodent retina in vivo and in vitro. Proc. Natl Acad. Sci. USA 101, 16–22 (2004).
Article ADS CAS PubMed Google Scholar
Peng, Y. et al. Purification and high-resolution top-down mass spectrometric characterization of human salivary α-amylase. Anal. Chem. 84, 3339–3346 (2012).
Article ADS CAS PubMed Google Scholar
Schramm, M. & Loyter, A. in Methods in Enzymology Vol. 8 (eds Neufeld, E. F. & Ginsburg, V.) 533–537 (Academic Press, 1966).
Devlin, B. & Roeder, K. Genomic control for association studies. Biometrics 55, 997–1004 (1999).
Article CAS PubMed Google Scholar
Abramson, J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493–500 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Meng, E. C. et al. UCSF ChimeraX: tools for structure building and analysis. Protein Sci. 32, e4792 (2023).
Article CAS PubMed PubMed Central Google Scholar
Kamitaki, N. et al. Code for ‘Human and bacterial genetic variation shape oral microbiomes and health’. Zenodo https://doi.org/10.5281/zenodo.14559457 (2025).
Willems, T., Gymrek, M., Poznik, G. D., Tyler-Smith, C. & Erlich, Y. Population-scale sequencing data enable precise estimates of Y-STR mutation rates. Am. J. Hum. Genet. 98, 919–933 (2016).
Article CAS PubMed PubMed Central Google Scholar
Frayling, T. M. et al. A common variant in the FTO gene is associated with body mass index and predisposes to childhood and adult obesity. Science 316, 889–894 (2007).
Article ADS CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank M. Florio, W.-S. Lee, B. Bor and C. Huttenhower for helpful discussions. We are grateful to all the families in SPARK, the SPARK clinical sites and SPARK staff. We appreciate obtaining access to SPARK genetic data on SFARI Base. This research was conducted using the UK Biobank Resource under application no. 40709. N.K. was supported by US NIH Fellowship F31 DE034283. R.E.H. and S.A.M. were supported by US NIH grant R01 HG006855. M.L.A.H. was supported by US NIH Fellowship F32 HL160061. R.E.M. was supported by US NIH grant K25 HL150334. S.A.M. was supported by the Howard Hughes Medical Institute. P.-R.L. was supported by US NIH grants R56 HG012698 and R01 HG013110 and a Burroughs Wellcome Fund Career Award at the Scientific Interfaces. The funders had no role in study design, data collection and analysis, the decision to publish or the preparation of the manuscript. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. Computational analyses were performed on the O2 High Performance Compute Cluster supported by the Research Computing Group at Harvard Medical School (http://rc.hms.harvard.edu), the UKB Research Analysis Platform and the All of Us Researcher Workbench. Molecular graphics and analyses were performed with UCSF ChimeraX, developed by the Resource for Biocomputing, Visualization and Informatics at the University of California, San Francisco, with support from National Institutes of Health R01-GM129325 and the Office of Cyber Infrastructure and Computational Biology, National Institute of Allergy and Infectious Diseases. The All of Us Research Program is supported by the NIH, Office of the Director: Regional Medical Centers: 1 OT2 OD026549; 1 OT2 OD026554; 1 OT2 OD026557; 1 OT2 OD026556; 1 OT2 OD026550; 1 OT2 OD026552; 1 OT2 OD026553; 1 OT2 OD026548; 1 OT2 OD026551; 1 OT2 OD026555; IAA no. AOD 16037; Federally Qualified Health Centers: HHSN 263201600085U; Data and Research Center: 5 U2C OD023196; Biobank: 1 U24 OD023121; The Participant Center: U24 OD023176; Participant Technology Systems Center: 1 U24 OD023163; Communications and Engagement: 3 OT2 OD023205; 3 OT2 OD023206; and Community Partners: 1 OT2 OD025277; 3 OT2 OD025315; 1 OT2 OD025337; and 1 OT2 OD025276. In addition, the All of Us Research Program would not be possible without the partnership of its participants.

Author information

Authors and Affiliations

Division of Genetics, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, USA
Nolan Kamitaki, Margaux L. A. Hujoel, Ronen E. Mukamel & Po-Ru Loh
Center for Data Sciences, Brigham and Women’s Hospital, Boston, MA, USA
Nolan Kamitaki, Margaux L. A. Hujoel, Ronen E. Mukamel & Po-Ru Loh
Program in Medical and Population Genetics, Broad Institute of MIT and Harvard, Cambridge, MA, USA
Nolan Kamitaki, Robert E. Handsaker, Margaux L. A. Hujoel, Ronen E. Mukamel, Steven A. McCarroll & Po-Ru Loh
Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard, Cambridge, MA, USA
Nolan Kamitaki, Robert E. Handsaker & Steven A. McCarroll
Department of Genetics, Harvard Medical School, Boston, MA, USA
Nolan Kamitaki, Robert E. Handsaker & Steven A. McCarroll
Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA
Nolan Kamitaki
Department of Human Genetics, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, CA, USA
Margaux L. A. Hujoel
Department of Computational Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, CA, USA
Margaux L. A. Hujoel
Jerome Lipper Multiple Myeloma Center, Dana-Farber Cancer Institute, Boston, MA, USA
Christina L. Usher
Howard Hughes Medical Institute, Harvard Medical School, Boston, MA, USA
Steven A. McCarroll

Authors

Nolan Kamitaki
View author publications
Search author on:PubMed Google Scholar
Robert E. Handsaker
View author publications
Search author on:PubMed Google Scholar
Margaux L. A. Hujoel
View author publications
Search author on:PubMed Google Scholar
Ronen E. Mukamel
View author publications
Search author on:PubMed Google Scholar
Christina L. Usher
View author publications
Search author on:PubMed Google Scholar
Steven A. McCarroll
View author publications
Search author on:PubMed Google Scholar
Po-Ru Loh
View author publications
Search author on:PubMed Google Scholar

Contributions

N.K., S.A.M. and P.-R.L. conceived the study design. N.K. carried out the computational and statistical analyses. R.E.H., R.E.M. and P.-R.L. developed the read-depth normalization pipeline on UKB. M.L.A.H. helped with performing the genome-wide association for dentures use on UKB. R.E.H. and C.L.U. helped with characterizing variation at the amylase locus. N.K. carried out the in vitro experiments. N.K., S.A.M. and P.-R.L. wrote the paper with contributions from all authors.

Corresponding authors

Correspondence to Nolan Kamitaki, Steven A. McCarroll or Po-Ru Loh.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data figures and tables

Extended Data Fig. 1 Oral microbiome composition changes with host age.

a, Distribution of the percent of reads not mapping to the human reference (GRCh38) across SPARK participants (n = 12,519). The median of 8.39% is indicated by the dotted red line. b, Principal coordinate analysis of microbiome profiles from the Human Microbiome Project⁷² together with a randomly selected subset of SPARK samples (n = 250). PCoA was performed using Bray-Curtis distance on all microbial species profiled in both data sets. The SPARK saliva samples cluster with other samples from oral communities. Points in gray are samples from Human Microbiome Project sites not listed in the legend. c, SPARK samples on a UMAP (Uniform Manifold Approximation and Projection) generated from the first 20 principal components of the abundance matrix for the 439 most prevalent species fall on a gradient with respect to host age (color bar on right). d, The 439 most prevalent species on a UMAP generated from their loadings onto the first 20 principal components of the abundance matrix fall on a gradient with respect to correlation of relative species abundance with host age (color bar on right). e, Among children in SPARK (n = 5,760), a UMAP using the same 20 principal components as in b shows minimal stratification by autism spectrum disorder case status. f,g, Scatter plots of SPARK participants along axes of top genetic principal components. Individual dots are colored according to self-reported race/ethnicity for individuals who reported a single race/ethnicity. h-k, Change in abundances of Prevotella melaninogenica, Streptococcus mitis, Porphyromonas endodontalis, and Neisseria cinerea over the age range found in the SPARK cohort (truncated at 60 years old due to limited sampling of elderly individuals). Relative abundances were inverse normal rank-transformed (y-axis). Red curves indicate medians; shading indicates interquartile regions.

Extended Data Fig. 2 Associations of taxon relative abundances, α-diversity, and microbial composition principal components with human genetic variants.

a, Genome-wide associations with relative abundances of 1,262 taxa observed in >10% of SPARK samples (n = 12,519). For each genetic variant, the most significant p-value is shown (across the 1,262 tests); the red line indicates the study-wide significance threshold (p < 4.0x10⁻¹¹). Protein-altering variants and copy number variants of note are highlighted: nonsense (red squares), missense (green triangles), and multi-allelic CNVs (blue diamonds). b, Genome-wide associations with α-diversity (Shannon entropy) in SPARK. c, Quantile-quantile plot of p-values computed by our mPC-based test for associations between human genetic variants and oral microbiome composition (Fig. 2b). The genomic inflation factor λ_GC was calculated as the median chi-square statistic divided by F⁻¹(0.5), where F⁻¹(x) is the inverse cumulative distribution function for a ${\chi }_{439}^{2}$ random variable. d, Associations of 11 lead variants identified by the mPC-based test (red) and 1,000 randomly selected variants (black) with dissimilarity of relative abundances for the 439 most prevalent species using multivariate distance matrix regression (MDMR, y-axis) as compared with our mPC-based test (x-axis). e, Analogous to d, for dissimilarity after applying the centered log-ratio transform to relative abundance measurements. f, Associations of AMY1 copy number (y-axis) with each of the 439 individual microbial principal components (x-axis). g-p, Analogous to f, for the other 10 lead variants identified by the mPC-based GWAS. P-values were computed using two-sided linear mixed models (a,b,f-p), one-sided chi-squared test (c; x-axis of d,e), or one-sided multivariate distance matrix regression (y-axis of d,e).

Extended Data Fig. 3 Complex associations of human genetic variation at the PRB locus with oral microbiome phenotypes.

a, Diploid copy number of PRB1 estimated for SPARK participants from WGS read-depth (n = 12,519). b, Associations of genetic variants at the PRB locus with relative abundance of Stomatobaculum SGB5266. The association of PRB1 copy number is highlighted (red point). c, Allelic series of effect sizes of PRB1 diploid copy numbers on relative abundance of Stomatobaculum SGB5266 (in s.d. units, n = 12,517). d, Partial colocalization of associations of genetic variants at the PRB locus with Stomatobaculum SGB5266 relative abundance (-log10(p), x-axis) and oral microbiome composition (-log10(p), y-axis). PRB1 copy number (red point) and a common PRB4 loss-of-function variant (blue point) appear to deviate from a generally concordant pattern of associations, likely reflecting multiple causal effects. e, Effect sizes on Stomatobaculum SGB5266 relative abundance for each additional copy of PRB1 compared to each functional copy of PRB4 (n = 12,519). Error bars, 95% CIs in all panels. P-values were computed using two-sided linear mixed models (b,d).

Extended Data Fig. 4 Additional information about associations between human genetic variants and oral microbial abundance phenotypes.

a, Stratified LD-score regression coefficients (which quantify evidence of enrichment of heritability in regions surrounding genes with tissue-specific expression) from S-LDSC analysis of chi-squared statistics from the oral microbiome composition GWAS (n = 12,519). Heritability enrichment was evaluated for genomic regions defined by 53 tissues from the Genotype-Tissue Expression Project (GTEx). b, Effect sizes (in s.d. units) on relative abundance of SGB9384 for individuals homozygous for either allele of the TLR1 I602S missense variant relative to heterozygotes (n = 12,519). c, Colocalization of genetic associations at the PITX1 locus with dentures use (-log10(p), y-axis) and oral microbiome composition (-log10(p), x-axis). d, Analogous to c, for the FUT2 locus. e, Consistency of effect sizes for 167 significant variant-species abundance associations computed with or without applying the centered log-ratio transform to relative abundance measurements. f, Consistency of effect sizes for 167 significant variant-species abundance associations computed with or without including ASD status as a covariate. g, Distribution of relative effect sizes on microbial species abundances for individuals homozygous for the FUT2 secretor allele relative to heterozygotes, across microbial species whose relative abundance associated with secretor status (FDR < 0.05). Error bars, 95% CIs in all panels. P-values were computed using one-sided chi-squared test (x-axis of c,d), two-sided linear regression (y-axis of c,d) or one-sided t-test (g).

Extended Data Fig. 5 Phyletic stratification of effects of two human genetic variants on relative abundances of oral microbial species.

a, Positive correlation of relative abundance of the Actinobacteria phylum relative to the Firmicute phylum across saliva samples from SPARK participants (n = 12,519). b, Unsigned effect sizes for associations of secretor status (based on FUT2 W154X, using a recessive model) with relative abundances of the 439 most prevalent microbial species (y-axis) versus phylogenetic distance from Haemophilus sputorum (x-axis). c, Analogous to b, for Granulicatella SGB8239. d, Signed effect size of associations of secretor status with relative abundances of microbial species (oriented relative to the effect direction for H. sputorum) (y-axis) versus phylogenetic distance from H. sputorum (x-axis). e, Analogous to d, for G. SGB8239. f, Microbial taxa whose abundance associated with the index variant at the HECTD2/PPP1R3C locus (FDR < 0.1) shown on the phylogenetic tree of 439 species (red, taxa whose relative abundances increased with the reference allele; blue, taxa whose relative abundances decreased with the reference allele). Two significantly-associated phyla (Firmicutes and Actinobacteria) are highlighted with yellow sectors. At the species level (outermost circle), dot sizes increase with statistical significance. g, Analogous to b, for Granulicatella adiacens with the HECTD2/PPP1R3C index variant. h, Analogous to g, for Streptococcus vestibularis. i, analogous to d, for G. adiacens with the HECTD2/PPP1R3C index variant. j, analogous to i, for S. vestibularis. For b-e, g-j, the red line indicates the median effect size, and the shaded region indicates the interquartile range. P-values were computed using two-sided Pearson’s product-moment correlation (a) or two-sided linear regression (b-e, g-j).

Extended Data Fig. 6 Genotyping and mutability of AMY1 copy number and replication of associations with oral health.

a, Estimated diploid copy number of AMY1 for SPARK participants (n = 12,519). b, Concordance of AMY1 copy number estimates between 5,149 sibling pairs in the UKB cohort that share both haplotypes identical-by-descent (IBD2) in the region surrounding the amylase locus. Among the 5,149 IBD2 sibling pairs, 13 had copy number-discordant calls (red points) that tended to differ by two copies (11/13), likely corresponding to de novo duplication or deletion of a copy of the common structural cassette containing two AMY1 genes in an inverted orientation to each other (Fig. 3a). Several IBD2 sibling pairs with AMY1 copy number estimates that differed by close to 1 copy appeared more likely to reflect uncertainty in copy number estimates as they lacked a corresponding AMY2A duplication or deletion that would be expected to accompany a duplication or deletion of a single copy of AMY1. This gives an estimated germline mutation rate of 6.3x10⁻⁴ mutations/meiosis ([3.5x10⁻⁴,11.1x10⁻⁴], 95% CIs, similar to recent estimates from haplotype coalescent trees⁴⁶), exceeding the mutation rate of most short tandem repeats⁸⁸. c, Analogous to a, for the AoU cohort (n = 245,377). d, Odds ratios for risk of complete tooth loss in AoU (n = 230,002) per AMY1 diploid copy number. e, Analogous to d, for having caries. f, Odds ratios for risk of dentures use in UKB (n = 418,039) per AMY2A diploid copy number. g, Analogous to f, for AMY2B diploid copy number. Error bars, 95% CIs in all panels. P-values were computed using two-sided linear regression (d,e).

Extended Data Fig. 7 Functional assays of AMY1 coding variation, consistency of microbiome associations, and lack of copy number associations with BMI in AoU.

a, Immunoblotting for amylase transgenically expressed in HEK293T cells. Amylase (~56kD) is found in the supernatant of cells transfected with the reference sequence of AMY1 (lane 1) and AMY1 F141C (lane 2), but not AMY1 C477R (lane 3) or a control plasmid (lane 4). The glycogen-purified protein is recognized by anti-amylase (lane 5). Amylase is found in the lysate of cells transfected with the reference sequence of AMY1 (lane 6), AMY1 F141C (lane 7), and less abundantly in cells transfected with AMY1 C477R (lane 8), but not a control plasmid (lane 9). Cross-reactive protein (~50kD) can be seen in the lysate of all samples and smaller AMY1 fragments in lanes containing AMY1. These results were replicated in 3 independent transformations with similar results. b, Glycogen-purified supernatant from cells transfected with the reference sequence of AMY1 (lane 1) and AMY1 F141C (lane 2), but not AMY1 C477R (lane 3) or a control plasmid (lane 4) contains a single protein band at ~56kD. These results were replicated in 3 independent transformations with similar results. c, Starch degradation is similar between equivalent mass dilutions of the reference amylase isoform and AMY1 F141C (n = 32 technical replicates for each allele). Starch degradation was measured as change in FAM relative fluorescence units over 30 min (y-axis) after addition of diluted AMY1 to quenched starch substrate. Centers, medians. d, Comparison of effect sizes for AMY1 copy number versus AMY1 C477R copy number on relative abundances of 16 bacterial species (from Fig. 3b) and on risk of dentures use (blue dot). For some species (e.g., Stomatobaculum SGB5266), the relative effect size of AMY1 copy number versus AMY1 C477R copy number on abundance is similar to this ratio for dentures use (blue line), whereas for others, it is not (e.g. Prevotella pallens). e, Concordance of effect size estimates for AMY1 copy number on relative abundances of 439 microbial species in adults (x-axis) and unrelated children (y-axis). Species whose abundances associated significantly with AMY1 copy number are indicated in darker gray (FDR < 0.05) or black (Bonferroni p < 0.05). f, Effect sizes on BMI per AMY1 copy number in AoU (n = 219,879). The line drawn is the best fit across AMY1 copy numbers. As a positive control, we confirmed that the BMI phenotype we tested (see Methods) associated strongly with the BMI-associated SNP rs1421085 at FTO⁸⁹ (p = 3.72 × 10⁻¹⁴⁰). g, Analogous to f, for the African/African American ancestry subset of AoU participants (n = 49,296, p = 0.25). h, Analogous to f, for the American Admixed/Latino ancestry subset of AoU participants (n = 38,788, p = 0.01). i, Analogous to f, for the European ancestry subset of AoU participants (n = 122,577, p = 0.66). Error bars, 95% CIs in all panels. P-values were computed using two-sided t-test (c), two-sided linear mixed models (e), or two-sided linear regression (f-i).

Extended Data Fig. 8 Pathway abundance associations and details of gene dosage associations.

a, Effect sizes of AMY1 diploid copy numbers on the relative abundance of reads mapping to genes in the tricarboxylic acid cycle (TCA) pathway across microbial species. b, Associations (-log10(p), x-axis) of the relative abundance of reads mapping to each of 2,416 Haemophilus sputorum gene families with FUT2 secretor status. Read depth measurements were inverse normal transformed across samples for each gene. Gene families generally fell into two modes, one group not associated with FUT2 secretor status and another that associated at significance similar to that of the relative abundance of H. sputorum. One outlier (indicated with the red arrow) associated much more strongly and corresponds to a gene encoding a protein with trimeric autotransporter adhesin domain annotations (UniParc ID J5HXA9). c, Associations of deletions in the genome of Streptococcus parasanguinis (as estimated by normalized coverage) with AMY1 copy number (n = 12,340). Shading indicates the two assembled contigs of the reference genome. The two significant associations (FDR < 0.01) overlap genes encoding amylase-binding proteins, abpA and abpB. d, Allelic series of effect sizes of AMY1 copy number on normalized coverage (n = 12,026) in the 500 bp bins overlapping abpA (orange) and abpB (green). e, Effect sizes of AMY1 diploid copy numbers on relative abundance of Streptococcus parasanguinis (n = 12,487). f, Protein-protein interaction between human AMY1 (purple) and S. parasanguinis AbpA (orange) predicted by AlphaFold3. AMY1 residue F141 is highlighted in red. g, Analogous to f, for S. parasanguinis AbpB (green). h, Replication of 208 associations between human genetic variants and normalized read-depth measurements in 500 bp bins of microbial genomes in the AoU cohort (comprised of individuals age 18 or older). Effect sizes estimated in AoU participants (n = 10,000, y-axis) are plotted against effect sizes estimated in SPARK (n = 12,519, x-axis). Dots correspond to the 208 associations and are colored according to the human genetic locus involved, as in Fig. 4b. i, Effect sizes of FUT2 W154X genotype on normalized coverage at bacterial genes encoding proteins with YadA-like (adhesin) domain annotations (n = 7419, Veillonella sp. 3627; n = 8153, Haemophilus sputorum; n = 7456, Haemophilus parahaemolyticus). Colors indicate the effect direction of FUT2 genotype on the relative abundance of each species (red, increasing with functional copies of FUT2; blue, decreasing with functional copies of FUT2). j, Correlation matrix of normalized coverage in the region of the V. sp. 3627 genome surrounding vadD and vadE revealed a linked deletion to vadE that contained vadC and nearly passed FDR < 0.01 (p = 3.32x10⁻⁶) in association with FUT2 W154X genotype, causing samples with vadE to tend to also contain vadC (Fig. 6d) and possibly suggestive of the event introducing vadD occurring after the one that produced vadC and vadE. Arrows indicate vad gene locations. Error bars, 95% CIs in all panels. P-values were computed using two-sided linear mixed model (a,e) or two-sided linear regression (b,c). Effect sizes were computed with two-sided linear regression (h,i).

Extended Data Fig. 9 Left breakpoints of deletions in the genome of Haemophilus sputorum that contain the hadA gene are bimodal.

a, Coverage of the QEQH01000003.1 contig in the Haemophilus sputorum genome for individuals with distinct left breakpoints of deletions containing hadA. b, Distribution of left breakpoints for the 200 individuals with highest genomic coverage of H. sputorum who had strong evidence of the deletion containing hadA (<0.1 median normalized coverage at the strongest associated bin). For each individual, mosdepth was run to measure per-base depth, and left and right breakpoints were identified as the first or last base with 10 consecutive zeros, respectively. c, Genome-wide associations for normalized read-depth in the 500 bp bin of Haemophilus sputorum QEQH01000003.1:197000-197500. d, Analogous to c, but for Prevotella nanceiensis KB904333.1:123500-124000. e, Analogous to c, but for Streptococcus mitis MUYN01000003.1:100000-100500. f, Analogous to c, but for Streptococcus vestibularis AEKO01000011.1:186000-186500. g, Analogous to c, but for Veillonella sp. 3627 RQVG01000009.1:13500-14000. h, Genome-wide associations for normalized read-depth in 500 bp bins spanning the Rothia mucilaginosa genome. For each human genetic variant, the most significant p-value is shown (across all 500 bp bins), with the red line indicating the study-wide significance threshold (p < 1.5×10⁻¹¹). P-values were computed using two-sided linear mixed models (c-h).

Extended Data Table 1 Eleven loci at which human genetic variants associate with oral microbiome composition

Full size table

Supplementary information

Supplementary Information

Supplementary Notes 1–8 and Figs. 1–3.

Reporting Summary

Supplementary Tables

Supplementary Tables 1–9.

Peer Review File

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Kamitaki, N., Handsaker, R.E., Hujoel, M.L.A. et al. Human and bacterial genetic variation shape oral microbiomes and health. Nature (2026). https://doi.org/10.1038/s41586-025-10037-7

Download citation

Received: 19 January 2025
Accepted: 10 December 2025
Published: 28 January 2026
Version of record: 28 January 2026
DOI: https://doi.org/10.1038/s41586-025-10037-7