Human defects in STAT3 promote oral mucosal fungal and bacterial dysbiosis

Studies in patients with primary immunodeficiencies (PIDs) have uncovered roles of specific genes and pathways in human immunity (1, 2). In this regard, patients with PIDs that present with mucocutaneous infections demonstrate essential pathways for immune-protection at barrier sites. Barrier immune surveillance is of particular significance, as these surfaces are constantly exposed to normal commensal microbiota and pathogens (3). Thus, perturbations in barrier immunity may disrupt the physiologic interplay between commensals and host, altering the composition of commensal microbial communities and often leading to infection susceptibility or microbial dysbiosis-associated immunopathology (4, 5).

Barrier surfaces are equipped with tissue-specific protective immune mechanisms largely dedicated to the maintenance of host-microbial homeostasis (6, 7). While there is significant understanding of the immunological systems of barrier compartments such as the lower gastrointestinal (GI) tract and skin, host-microbiome interactions at the oral mucosa are much less explored. However, the oral mucosa is second only to the lower GI tract as a home to a rich and diverse community of commensal microbes (8) and is the first site of microbial encounter for the immune system in the GI tract. Consequently, host-microbial interactions at this interface are important to define (9, 10).

An immunological mechanism of importance for oral mucosal barrier surveillance is that of Th17- and IL-17 cytokine–mediated immunity. Indeed, patients and mice with defects in Th17/IL-17 immunity have been documented to have severe susceptibility to oral fungal infections (11–15). IL-17–secreting cells (including Th17 cells) have been shown to play a primary role in maintenance of barrier integrity, as well as in host defense, particularly against extracellular bacteria and fungi (16–19). At the oral mucosa, Th17 and other IL-17–producing cells have been shown to mediate antifungal immunity through secretion of their signature cytokine, IL-17, which stimulates the production of antimicrobial peptides by epithelial cells (20, 21). Among the recognized PIDs with Th17 defects are patients with autosomal-dominant hyper-IgE syndrome (AD-HIES or Job’s syndrome), which harbor loss-of-function (LOF) mutations in the STAT3 gene (22, 23). AD-HIES patients have been classically characterized by the triad of eczema, skin, and lung infections, as well as extremely elevated serum IgE (24, 25). They also present with recurrent oral Candida infections, indicating a critical role for STAT3-mediated oral immunity (26, 27). STAT3 is a key transcription factor downstream of cytokine signaling from IL-6, IL-21, IL-10, and IL-23, among other immune mediators. STAT3 is also defined as a key transcription factor in the differentiation of Th17 (28, 29). Oral mucosal candidiasis in AD-HIES has been thought to be due to defective Th17 differentiation and decreased production of select antimicrobial peptides (30–32).

However, although IL-17/Th17–related defects have a dominant oral phenotype, it is not well understood whether defects in STAT3 and related Th17/IL-17 in humans affect commensal microbiome colonization and predispose to mucosal microbial dysbiosis. Interestingly, emerging literature in animal models has linked blockade of IL-17 with defective anti-microbial surveillance and mucosal dysbiosis-associated immunopathology (16, 33, 34).

Therefore, our current study aims to further our understanding of the consequences of LOF STAT3 and related Th17 defects on establishment and dynamics of the mucosal mycobiome and bacteriome.

LOF STAT3 mutations in humans are linked to recurrent oral candidiasis. To examine the consequences of LOF STAT3 on oral commensal fungal and bacterial communities, we recruited a large cohort of AD-HIES patients for clinical evaluation and research studies. AD-HIES diagnosis was based on clinical phenotype and confirmed by detection of LOF mutations in the STAT3 gene. We clinically examined our entire AD-HIES cohort (n = 36) for oral manifestations and obtained detailed medical and dental histories. We documented a significant rate of oral candidiasis, with 86% of the patients reporting recurrent episodes of oral thrush (Table 1). Most patients received antibiotic and antifungal prophylaxis. However, patients reported recurrent oral candidiasis at diagnosis, prior to commencement of antibiotic regimens. Oral mucosal fungal infections in AD-HIES have been previously linked to defects in the differentiation of Th17 cells and to reduced production of antimicrobial peptides human β-defensin 2 and several histatins (32). Consistent with decreased production of antimicrobial peptides, we detected reduced levels of S100A9 levels (a cation-binding protein, usually found as calprotectin) in oral mucosal secretions (saliva) of AD-HIES patients, compared with age- and sex-matched healthy volunteers (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.122061DS1).

Table 1

Demographic and Clinical Information for the AD-HIES cohort

Clinical examination revealed active fungal lesions at the time of evaluation in a subset of patients, primarily presenting on the tongue dorsum and buccal mucosal areas (Supplemental Figure 1, B and C). Other significant intraoral findings included recurrent oral ulcers and severe dental caries in 36.1% and xerostomia in 19.4 % of patients (Table 1).

STAT3-mediated immunity is essential for commensalism of Candida albicans (C. albicans). We first set out to evaluate consequences of defective STAT3-mediated mucosal immunity on the oral mycobiome. For these studies, we subsampled a cohort of AD-HIES patients from our population to obtain balanced groups of patients with the presence/absence of active fungal infection for comparison (n = 18, 9 with active lesions [actively infected] and 9 without mucosal lesions [uninfected]), along with age- and sex-matched healthy controls (Supplemental Table 1). We obtained swab samples from AD-HIES patients and healthy controls at distinct mucosal sites where Candida infection typically occurs (the tongue dorsum and buccal mucosa).

All samples were subjected to internal transcribed spacer 1 (ITS1) sequencing for characterization of fungal communities. We found distinct mycobiome communities in AD-HIES patients compared with healthy controls (Figure 1A and Supplemental Figure 2A). Actively infected AD-HIES samples also significantly separated from uninfected patients, based on measures of community structure (θ YC distances) and membership (Jaccard index) both in tongue and buccal sites (Figure 1A and Supplemental Figure 2A). Evaluation of diversity of fungal communities based on the Shannon diversity index and the examination of the number of observed fungal genera revealed a decrease in fungal diversity in the actively infected AD-HIES patients (Figure 1, B and C).

Figure 1

Oral mucosal mycobiome in AD-HIES differs significantly from healthy controls. (A) Principal coordinates analysis (PCoA) plot analyzing community structure (based on θ YC distances), showing that fungal communities of patients with autosomal dominant hyper-IgE syndrome (AD-HIES) cluster apart from healthy controls (HC) in tongue and buccal surfaces. P < 0.001 as determined by AMOVA comparing HC versus all AD-HIES combined. P < 0.001 as determined by AMOVA comparing actively infected (A_HIES, dark purple) and uninfected (U_HIES, light purple) AD-HIES patients. Each circle represents 1 sample. Some data points are not visible, as they get superimposed due to tight clustering. (B) Nonparametric Shannon diversity index of tongue and buccal fungal communities of HC, U_HIES, and A_HIES patient samples. ****P < 0.0001 as determined by Kruskall-Wallis test and Dunn’s multiple comparisons test. (C) Number of observed fungal genera in tongue and buccal communities of HC, U_HIES, and A_HIES patient samples. ****P < 0.0001 and *P < 0.01 as determined by Kruskall-Wallis test and Dunn’s multiple comparisons test. (A–C) The number of samples per group included in these graphs for HC were n = 23 for tongue and n = 25 for buccal, and n = 9 for tongue and n = 8 for buccal for both U_HIES and A_HIES patient groups. Boxes extend from the 25th to 75th percentiles, and the whiskers were plotted from the minimum to maximum value. All outlying values were shown.

Fungal communities from actively infected AD-HIES patients were dominated by the genus Candida, accounting for 87.1%–100% abundance on tongue and 85%–100% on buccal lesions (Figure 2A). Species-level classification of Candida sequences revealed C. albicans as the most abundant taxa in actively infected AD-HIES communities, displaying 84.2%–100% of relative abundance in tongue samples and 57.6 %–100% in buccal samples (Figure 2A and Supplemental Figure 3), suggesting that oral candidiasis in AD-HIES represents a specific overgrowth of C. albicans. Consistent with this, C. albicans quantitation by real-time PCR was 2 logs higher in actively infected patients than in healthy controls or uninfected patients (Figure 2B). AD-HIES individuals did not harbor unique fungal taxa compared with healthy controls. Interestingly, antifungal treatment did not appear to alter mycobiome communities. Approximately half of Candida-infected and half of the uninfected patients were receiving antifungal regimens (Supplemental Table 1), yet the structure and composition of mycobiome communities did not significantly differ based on antifungal therapy (Supplemental Figure 4).

Figure 2

Actively infected AD-HIES oral mucosal fungal communities are dominated by Candida albicans. (A) Relative abundance plot depicts the major fungal taxa across communities in healthy controls (HC) and patients with AD-HIES. Patient groups shown are HC (n = 23 for tongue, n = 25 for buccal), uninfected AD-HIES (U_HIES, n = 9 for tongue and n = 8 for buccal), and actively infected HIES (A_HIES, n = 9 for tongue and n = 8 for buccal). Malassezia and Candida genera were further classified to species level due to abundance and clinical relevance. Each bar represents 1 subject. Empty bars represent missing samples. (B) Candida albicans genome copies per μg of DNA in tongue and buccal fungal communities from HC (n = 12 for tongue and n = 16 for buccal), uninfected (U_HIES, n = 8 for tongue and n = 7 for buccal), and actively infected (A_HIES, n = 7 for tongue and n = 7 for buccal) AD-HIES patients, quantified by real-time PCR. Candida albicans biomass values are expressed as log₁₀ of genome copies. ****P < 0.0001 as determined by Kruskall-Wallis test and Dunn’s multiple comparisons test. Boxes extend from the 25th to 75th percentiles, and the whiskers were plotted from the minimum to maximum value. All outlying values were shown.

Fungal communities from uninfected AD-HIES patients were dominated by the genus Malassezia with representation of C. dubliniensis (Figure 2). Highlighting the commensal nature of C. albicans and the specificity of its opportunism in AD-HIES is the observation that the mycobiomes of a small number of healthy individuals and 1 uninfected AD-HIES patient were also dominated by C. albicans (Figure 2), despite absence of active Candida infection and despite the fact that C. albicans was detected by real-time PCR in healthy volunteers.

Our evaluation also afforded a detailed view of the oral mycobiome in 2 distinct oral mucosal surfaces in healthy subjects (Figure 2A). We observed high variability in fungal composition of mycobiome communities in healthy controls, with a predominance of the genus Malassezia in both tongue and buccal surfaces. Within the Candida genus, C. albicans was the most common species observed in health-associated communities. C. parapsilosis, Boletus, and Penicillium were also among the highly abundant fungal taxa present in communities from healthy controls.

Collectively, our characterization of fungal communities in AD-HIES and healthy controls demonstrates that oral mucosal fungal infections in AD-HIES are dominated by the opportunistic yeast C. albicans, revealing a specific role for STAT3/Th17-mediated immunity in containing C. albicans as a commensal organism.

Defects in STAT3-mediated immunity are associated with shifts in bacterial communities. We next aimed to characterize the consequences of STAT3 deficiency on oral mucosal bacterial communities. For this, we performed 16S rRNA gene sequencing on the identical samples analyzed for mycobiome. Evaluating samples from Candida-infected and uninfected AD-HIES individuals allowed us to characterize the consequences of STAT3 deficiency on the oral microbiome alone and in the presence of active candidiasis. We found that bacterial communities in AD-HIES separated from those of healthy controls based on measures of community structure (θ YC distances) and composition (Jaccard index) (Figure 3A and Supplemental Figure 2B). AD-HIES microbial communities displayed decreased diversity with a reduced number of bacterial species, particularly in the setting of candidiasis (Figure 3, B and C).

Figure 3

AD-HIES oral mucosal bacterial communities differ significantly from healthy controls. (A) Principal coordinates analysis (PCoA) plot analyzing community structure (based on θ YC distances) of the bacterial microbiome in HC and AD-HIES. For the tongue plot, P < 0.001 comparing HC vs. all AD-HIES combined and P = 0.103 (ns) comparing actively infected HIES (A_HIES) and uninfected HIES (U_HIES) samples as determined by AMOVA. For buccal plot, P < 0.05 comparing HC vs. all HIES combined and P = 0.118 (ns) comparing A_HIES versus U_HIES, as determined by AMOVA. Each circle represents 1 sample. (B) Nonparametric Shannon diversity index of tongue and buccal bacterial communities of HC and AD-HIES patients. ****P < 0.0001 as determined by Kruskall-Wallis test and Dunn’s multiple comparisons test. (C) Number of observed bacterial species in tongue and buccal communities of HC and AD-HIES patients. ****P < 0.0001 as determined by Kruskall-Wallis test and Dunn’s multiple comparisons test. (A–C) The number of samples per group included in these graphs for HC were n = 25 for tongue and buccal, for U_HIES n = 9 for tongue and buccal, and for A_HIES n = 9 for tongue and n = 8 for buccal. Boxes extend from the 25th to 75th percentiles, and the whiskers were plotted from the minimum to maximum value. All outlying values were shown.

We next sought to characterize the composition of bacterial communities in AD-HIES patients compared with healthy controls by evaluating the relative abundance of main bacterial taxa. Streptococcus, Prevotella, and Rothia were the most abundant bacterial constituents across all samples (Figure 4). Evaluation of bacterial taxa present in AD-HIES compared with healthy controls revealed several bacterial taxa differentially abundant between groups. Several common oral commensal bacterial species and genera — including Neisseria, Porphyromonas, and Haemophilus — had a significantly lower relative abundance in AD-HIES, while the genus Capnocytophaga and TM7 [G-1] were overrepresented in either tongue or buccal AD-HIES samples (Supplemental Figure 5).

Figure 4

Overview of most abundant bacterial genera in oral mucosal microbiome of AD-HIES compared with HC. Relative abundance plots depict the major bacterial genera in tongue and buccal samples from HC, actively infected (A_HIES), and uninfected (U_HIES) patients with AD-HIES. Each bar represents 1 subject. Empty bar represents a missing sample.

Candida infection in AD-HIES patients is associated with increased abundance of streptococci and increased dental caries. To specifically assess the effect of Candida infection on the oral mucosal microbiome in the setting of AD-HIES, we next interrogated differences in bacterial genera in infected compared with uninfected patients. Our key finding was the increase in relative abundance of the genus Streptococcus in actively infected patients. Microbial communities from infected AD-HIES patients displayed an increase in the relative abundance of the Streptococcus genus (Figure 4), and linear discriminant analysis effect size (LEfSe) analysis revealed a significantly increased relative abundance of Streptococcus in both tongue and buccal mucosa of candidiasis-infected AD-HIES patients (Figure 5A). Streptococcus biomass was also greater in buccal mucosa of Candida-infected AD-HIES patients compared with healthy controls (Figure 5B). Interestingly, investigating the bacterial species overrepresented in buccal samples from Candida-infected AD-HIES sites, we found increased abundance specifically of Streptococcus mutans (S. mutans) and Streptococcus oralis (S. oralis) (Figure 5C).

Figure 5

Increase in oral streptococci and dental caries susceptibility in AD-HIES. (A) Differentially represented bacterial genera in actively infected (A_HIES, n = 9 for tongue and n = 8 for buccal samples) and uninfected (U_HIES, n = 9 for tongue and buccal samples) patients with AD-HIES, determined via LEfSe analysis. (B) Real-time PCR quantitation of the genus Streptococcus in tongue and buccal surfaces of HC and AD-HIES patients. The number of samples included in the tongue panel were HC n = 15, U_HIES n = 8, and A_HIES n = 8; for the buccal panel, number of samples were HC n = 15, U_HIES n = 7, and A_HIES n = 8. Streptococcus biomass values are expressed as log₁₀ of 16S rRNA gene copy number. *P < 0.02 as determined by Kruskall-Wallis test and Dunn’s multiple comparisons test. (C) Differentially represented bacterial species found in actively infected and uninfected AD-HIES buccal samples, determined via LEfSe analyses. (D) Caries prevalence comparing HC (n = 22) and AD-HIES (n = 36) calculated using the Decayed, Missing, Filled Teeth (DMFT) index in dental radiographs. *P < 0.03 as determined by Mann-Whitney U test. (B and D) Boxes extend from the 25th to 75th percentiles, and the whiskers were plotted from the minimum to maximum value. All outlying values were shown.

Exploring the possible functional consequences of overgrowth of select Streptococcus spp. in the oral microbiome of AD-HIES patients, we evaluated the presence and severity of streptococci-associated oral infections in AD-HIES patients. S. mutans is recognized as the main etiologic agent of dental caries and was detected with increased relative abundance in candidiasis-affected AD-HIES patients. Interestingly, our original clinical examination had indicated high rates of dental infections in AD-HIES (Table 1).

Therefore, we quantitatively assessed dental caries in AD-HIES patients compared with age- and sex-matched healthy volunteers (AD-HIES n = 36, healthy controls n = 22). The prevalence of caries was evaluated by assessment of missing, decayed, and filled teeth (DFMT) on dental radiographs from AD-HIES and healthy controls. We found a significant increase in the prevalence of dental caries in the AD-HIES cohort (Figure 5D).

While a causative relationship between Candida infection, streptococcal overgrowth, and dental infections cannot be conclusively established based on our cross-sectional study, our data suggest that immune defects in AD-HIES are associated with C. albicans infection and microbial dysbiosis linked to streptococcal overgrowth.

Studies in experimental models and humans with defects in IL-17/Th17 have suggested that the Th17 pathway is critical in the surveillance of C. albicans (11, 20, 33, 34). Defects of Th17 differentiation have been previously documented in patients with LOF mutations of STAT3 (30, 31) and linked to oral candidiasis (26, 27). However, the consequences of defective STAT3/Th17-mediated immunity have not been thoroughly evaluated at the oral mucosal barrier in humans. Our work aimed to further our understanding of STAT3/Th17-mediated mucosal immunity in humans and its consequences on the oral mucosal mycobiome and bacteriome. Toward this goal, we recruited a large cohort of patients with LOF mutations in STAT3 (AD-HIES, n = 36) and performed clinical, immunological, and microbiome studies at the oral mucosal barrier. Clinical phenotyping of our cohort of AD-HIES patients confirmed previous reports of increased rates of oral candidiasis (26, 27), revealing recurrent candidiasis in 86% of the population. Predominant sites affected by candidiasis were the buccal mucosa and dorsum of tongue. We sampled a subset of our population to include 50% of patients with active oral candidiasis and allow for a balanced sampling of actively infected and uninfected patients for our microbiome characterization. Additional clinical oral findings included increased susceptibility to oral ulcers and dental caries (36.1%), as well as xerostomia (19%).

We found that the mycobiome and bacteriome in AD-HIES were distinct from those of age- and sex-matched healthy controls, with a significantly reduced diversity of communities, particularly in the presence of active Candida infection. These findings were unlike those in skin fungal communities in AD-HIES patients, which had higher diversity than healthy controls. These differences in oral and cutaneous microbiomes may reflect distinct regulatory pathways for microbial colonization at various epithelial sites (35, 36). Therefore, our work underscores the importance of examining tissue-specific differences in microbiome composition. Furthermore, to our knowledge, our studies provide the first characterization of the mucosal fungal and bacterial communities in AD-HIES.

By performing molecular characterization of fungal communities in AD-HIES, we observed dominance of the commensal species C. albicans. This suggested a lack of exogenous infection in AD-HIES. In fact, while C. albicans was detected by quantitative PCR (qPCR) in the majority of healthy volunteers, its biomass was much higher in AD-HIES, particularly in the setting of active infection. Our findings support that Candida infection in AD-HIES was not related to chronic antibiotic use. AD-HIES patients reported a history of oral thrush prior to diagnosis and commencement of antibiotic use. In fact, the use of antibiotics does not typically predispose to oral candidiasis, unlike vaginal candidiasis (37, 38). Interestingly, in AD-HIES, antifungal treatment did not alter the composition of fungal communities and did not predispose patients to nonalbicans Candida infections (39, 40).

Our study also provides a first characterization of fungal communities in distinct oral mucosal sites in healthy individuals. Previous studies have characterized oral mycobiome in oral rinses and saliva, summating fungal communities from all oral sites (41–43). Importantly, our work in the health-related mycobiome found important differences in the buccal mucosal and tongue fungal communities, as well as extensive intraindividual variability. Malassezia was the most common shared predominant taxa in healthy individuals and in both mucosal sites studied, similar to reports from human saliva, nares, and skin (42, 44).

We next specifically evaluated the oral microbiome in AD-HIES patients with and without active Candida infection. These analyses aim at characterizing the effects of STAT3 deficiency alone or in combination with C. albicans infection on the oral microbiome. Importantly, the setting of AD-HIES and recurrent Candida infections provide a unique disease model to evaluate oral microbiome intracommunity relationships in the context of C. albicans infection.

Bacterial communities in AD-HIES had reduced diversity and a markedly decreased abundance and presence of common oral commensal bacteria, particularly in the context of oral candidiasis.

Oral candidiasis in AD-HIES was linked to an increased proportion of Streptococcus spp., as well as an increase in the biomass of streptococci, particularly at buccal mucosal sites. S. oralis and S. mutans were overrepresented on the buccal mucosa in active candidiasis. Previous findings had also demonstrated that oral communities dominated by Candida had decreased bacterial richness and increased oral streptococci and other acidophilic taxa (45, 46). Based on the increased representation and biomass of Streptococcus in candidiasis in AD-HIES, we suspect a synergistic relationship between C. albicans and oral streptococci (47–50). Previous in vitro data has suggested that select oral streptococci enhance C. albicans virulence by increasing its capacity to invade oral tissues (47), permitting severe oral thrush lesions in vivo (49). This cooperative relationship may be contributing to candidiasis susceptibility or secondary infections in AD-HIES. The synergy between C. albicans and S. mutans has been recognized as a key factor in dental caries, due to the metabolism of dietary sugars into glucans and acid, forming a firmly adherent biofilm on the tooth surface, leading to loss of mineralized tissue (50). C. albicans amplifies the cariogenic potential of S. mutans biofilms by increasing exopolysaccharide production, augmenting biofilm biomass, and enhancing its virulence, which leads to rampant carious lesions in rodent models (48, 51, 52). Indeed, multiple clinical studies have shown that S. mutans and C. albicans carriage is much higher in early-childhood caries and adults with increased carious lesions compared with caries-free individuals (53–57). Our data further support a cooperative interaction between C. albicans and S. mutans, as we have documented significantly increased dental caries in AD-HIES patients compared with healthy controls. A small case series had also suggested that AD-HIES patients present with severe carious lesions (58). While our current study has focused on the oral mucosal microbiome, it will be important in future work to evaluate microbiome shifts on the tooth surface where caries occur.

Our work supports a model in which STAT3/Th17 is critical in oral antifungal immunity specifically against C. albicans. In the presence of defective STAT3/Th17 immunity, C. albicans overgrows and bacterial communities become dysbiotic with a dominance of oral streptococci. Overgrowth of specific oral streptococci may be promoted by C. albicans alone and/or facilitated by defective IL-17/Th17 immunity (59, 60), leading to local secondary dental infection caries (Figure 6).

Figure 6

Proposed model in which the STAT3/Th17 axis is critical in oral antifungal immunity specifically against Candida albicans. In the presence of defective STAT3/Th17 immunity, C. albicans will overgrow and lead to shifts in bacterial communities. Consequent microbial dysbiosis (i.e., increase in specific oral streptococci) may contribute to local secondary infections.

Taken together, our studies in a population of patients with genetic defects in STAT3 reveal a critical role for STAT3-mediated immunity in the oral mucosal surveillance of C. albicans and in the establishment of oral commensal fungal and bacterial communities.

Patient characterization. Diagnosis of autosomal dominant hyper-IgE syndrome was performed using a HIES scoring system that considers the main immunologic/nonimmunologic disease manifestations, as previously described (61) and confirmed by the presence of LOF mutations in the STAT3 gene. Age- and sex-matched healthy subjects without significant medical histories were included as a healthy-control population. All demographic and relevant medical/dental clinical information was recorded.

Oral examination and microbiome sample collection. Questionnaires of dental history were completed for all patients. Data related to dental/oral findings including history of recurrent oral thrush, recurrent oral ulcers, xerostomia (subjective reporting combined with minimal or no unstimulated saliva), and severe caries susceptibility (defined as generalized carious lesions or restorations on >30% of dentition) were recorded. All AD-HIES patients and healthy volunteers received an oral examination and a panoramic radiograph. Clinical diagnosis of oral candidiasis in AD-HIES individuals was made upon detection of white lesions that rubbed off with gauze. Caries status determination was performed recording the Decayed, Missing, Filled Teeth (DMFT) index using panoramic radiographs. Patients were given instructions not to eat, not to drink anything other than water, not to chew gum 2 hours before sample collection, and to refrain from oral hygiene for 12 hours before sampling. Samples for oral mucosal microbiome analyses were collected from tongue dorsum and buccal mucosa areas by rubbing for 10 seconds using CatchAll swabs (Epicentre), which were then placed in Yeast Lysis buffer (Epicentre) and stored at –80°C until processing. Patients with xerostomia were excluded from research sampling procedures.

Saliva sample collection and S100A9 ELISA. Unstimulated saliva samples were obtained by expectoration into a polypropylene tube kept on ice for 5 minutes. Saliva samples were centrifuged at 2,600 g for 15 minutes; then, the supernatant was recovered and proteinase inhibitors (Roche Diagnostics) were added. Samples were stored at –80°C until processing. S100A9 protein levels were measured using an ELISA-based kit (R&D, catalog DY5578), following the manufacturer’s instructions.

DNA isolation, library preparation, and sequencing. DNA was extracted using the protocol previously detailed in Jo et al. (62). Briefly, oral mucosal swab samples were subjected to an enzymatic lysis step that included the addition of ReadyLyse Lysozyme solution (Epicentre) and a 1-hour incubation at 37°C in a shaking thermal block. To mechanically dissociate fungal and bacterial cells walls, 2 steel beads (5 mm, Qiagen) were added to each tube to perform a bead-beating step in a TissueLyser (Qiagen) for 2 minutes at 30 Hz. Then, samples were incubated at 65°C for 30 minutes to finalize lysis, followed by addition of the MPC Reagent (Epicentre), and their processing continued using the PureLink Genomic DNA Kit (Invitrogen) according the manufacturer’s instructions. DNA was eluted in 35 μl of PCR grade water (MoBio).

For 16S rRNA gene amplicon sequencing, the V1-V3 region was amplified using the primers V1_27F (5′-CCTATCCCCTGTGTGCCTTGGCAGTCTCAGAGAGTTTGATCCTGGCTCAG-3′) and V3_534R (5′-CAGCACGCATTACCGCGGCTGCTGG-3′) (44). For ITS1 amplicon sequencing, the ITS1 region was amplified using the primers 18SF (5′-CCTATCCCCTGTGTGCCTTGGCAGTCTCAGGTAAAAGTCGTA ACAAGGTTTC-3′) and 5.8S-1R (5′-GTTCAAAGAYTCGATGATTCAC-3′) (44). Both primer pairs (for 16S rRNA V1-V3 and ITS1) included 5′ and 3′ linker sequences, heterogeneity spacers, and index identifiers to allow dual indexing as previously described (63).

The PCR reaction for 16S rRNA and ITS1 amplification were the following: 2.5 μl 10× PCR buffer, 4 μl deoxynucleotide (dNTP) mix, 0.25 μl Taq polymerase (Takara), 1 μl of each Forward and Reverse primer, 2.5 μl of isolated microbial DNA, and 13.75 l of PCR-grade water. The thermal cycling conditions for 16S rRNA gene library amplification included an initial denaturation step of 95°C for 2 minutes and 30 cycles of denaturation at 95°C for 20 seconds, with annealing at 56°C for 30 seconds and extension at 72°C for 60 seconds. The thermal cycling conditions for ITS1 gene library amplification included an initial denaturation step of 95°C for 2 minutes, 30 cycles of denaturation at 95°C for 30 seconds, annealing at 50°C for 30 seconds, extension at 72°C for 2 minutes, and a final extension at 72°C for 5 minutes. PCR reactions were carried out in duplicate; then, PCR products were combined, purified using Agencourt AMPureXP (Beckman Coulter), and quantified via Quant-IT dsDNA Kit (Invitrogen; Life Technologies). Approximately equivalent amounts of each PCR product were then combined and purified using the MinElute PCR purification kit (Qiagen) and sequenced at the NIH Intramural Sequencing Center on an Illumina MiSeq platform.

16S rRNA gene amplicon sequence analysis pipeline. Reads were processed using the software mothur (64). Preprocessing steps included contig assembly, trimming (primers, spacers, and indices), and size filtering to include sequences with a maximum length of 510 bp and no ambiguous base calls. Sequences were then processed according to an established mothur-based pipeline (65). Chimeric sequences were detected using UCHIME (66) as implemented in mothur and subsequently removed. For taxonomical assignment, we performed a phylotype analysis using the Human Oral Microbiome Database (HOMD) version 14.5 as a reference (67). Reads were classified up to the species level by using the BLAST option and the k-nearest neighbor algorithm within mothur following the parameters established and validated by Al-Hebshi et al. (68).

ITS1 gene amplicon sequence analysis pipeline. Reads were preprocessed and chimeric sequences removed following the same procedures described for the 16S rRNA amplicon data (see above). For taxonomical classification, we used a modified version of a previously generated custom fungal ITS1 reference database (44). Modifications included a revision of fungal nomenclature, with synonym fungal genera collapsed and given a consensus genus name according to the recommendation of Dupuy et al. (42) and current literature. Additionally, sequences belonging to Aspergillus/Emericella genera were manually curated by eliminating a fraction of them that were misclassified as fungi. With this updated ITS1 database as our reference, sequences were classified down to the genus level by BLAST and the k-nearest neighbor algorithm in mothur using the parameters previously determined and validated by Findley et al. (44).

Candida and Malassezia ITS1 reads were further classified to species level following the same procedures previously described for Malassezia speciation (44). Briefly, Candida sequences were extracted from our dataset through the get.lineage command in mothur. We then constructed a Candida ITS1 reference package by retrieving and aligning Candida type-strain ITS1 sequences. This manually curated Candida ITS1 database was used for phylogenetic classification of Candida reads via the pplacer software package (69), considering a likelihood score higher than 0.65.

α- And β-diversity estimates for 16S rRNA and ITS1 data. Measures of community diversity (nonparametric Shannon index), community structure (θ YC index), and membership (Jaccard index) were calculated using mothur, after subsampling at 25,000 reads per sample for 16S rRNA data and 3,000 reads for ITS1 data. These metrics were constructed based on species-level taxonomic units for 16S rRNA and genus-level for ITS1 (with the exception of Candida that was speciated). The lower subsampling cut-off for fungal data was due to the lower amount of sequences recovered in samples from healthy controls. Principal coordinates analyses (PCoA) was performed in mothur.

Quantification of Streptococcus and Candida biomass. Determination of Streptococcus load was performed via real-time PCR using 16S rRNA gene-based primers, as previously described (70). The standard curves consisted of genomic DNA of S. gordonii (ATCC, 35105), which was serially diluted to obtain a range from 1 × 10² to 1 × 10⁸ 16S rRNA gene molecules. PCR reactions were carried out using a final volume of 50 μl; each reaction consisted of 25 μl of SYBR Green PCR Master Mix (Invitrogen; Life Technologies), 500 nM of each primer, 1 μl of DNA from each sample, and PCR water (MoBio). Thermal cycle conditions comprised 95°C for 10 minutes, 40 cycles of 95°C for 15 seconds, and 60°C for 1 minute. A melting curve analysis for Streptococcus PCR products was performed and revealed a single sharp peak, confirming the specificity of the assay.

For quantification of C. albicans burden, real-time PCR reactions were run using 8 ng of total DNA mixed with 500 nM C. albicans–specific primers (that target the SAP5 gene) (71) and PerfecTa SYBR Green Fastmix (Quanta Biosciences). The cycling protocol involved an initial denaturation step at 95°C for 3 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. C. albicans genome copy values were calculated from a standard curve prepared using DNA isolated from pure C. albicans cultures. This quantitation was performed taking into consideration that the C. albicans genome harbors a single copy of the SAP5 gene, and since this yeast has a diploid genome, there are 2 copies of the SAP5 gene in 1 CFU equivalent genome, allowing direct extrapolation from CFUs to genome equivalents. Briefly, 9 × 10⁸ blastopores of C. albicans were resuspended in 300 μl of sterile TENTS (10mM Tris-HCl, pH 8.0 containing 1mM EDTA, 100mM NaCl, 2% Triton X-100 and 1% SDS; all from MilliporeSigma), and DNA isolation was done as described previously (72).

All real-time assays were performed in a 7500 Real-Time PCR system (Invitrogen; Life Technologies), and PCR reactions were carried out in triplicate.

Data access. All sequence data from this study have been submitted SRA under the to NCBI BioProject 413203 (https://www.ncbi.nlm.nih.gov/bioproject) under accession number 413203.

Statistics. Patient demographic data (age and sex) were compared via unpaired 1-taield t test and Fisher’s test, respectively. Dissimilarities in α diversity, number of taxa and real-time data among healthy controls, and uninfected and actively infected AD-HIES samples were determined using 1-way ANOVA or Kruskall-Wallis tests according to data distribution. Comparisons for S100A9 levels and DMFT scores between healthy and AD-HIES patients were assessed using Mann-Whitney U tests. Analysis of Molecular Variance (AMOVA) was used to test differences for PCoA analyses of community structure and membership. For all of these previously mentioned statistical analysis, a value of P < 0.05 was considered significant. For 16S rRNA and ITS1 sequencing data, differences in relative abundance between groups was evaluated using LEfSe (73) considering 0.01 as the α value for statistical testing. All graphs and analyses were generated using R software.

Study approval. All patient-related activities in this study were conducted under IRB approved protocols (NIDCR and NIAID from NIH). All participants provided written informed consent prior to inclusion in the study.