Loss of the disease-associated glycosyltransferase Galnt3 alters Muc10 glycosylation and the composition of the oral microbiome

Galnt3 is the most abundant Galnt in the adult submandibular gland

Previous studies have shown that Galnt3 is abundantly expressed in other tissues throughout the body, including the salivary glands (23, 38). To begin to interrogate the relative abundance and temporal expression of Galnt3 in the submandibular glands (SMGs), we investigated expression levels of each Galnt family member at three embryonic stages (embryonic day 13 (E13), E15, and E17), two early postnatal stages (postnatal day 2 (P2) and 1 week), and three adult stages (4, 8, and 12 weeks) by performing qPCR (Fig. 1 and Fig. S1). At E13, Galnt1 is the most abundantly expressed isoform, followed by Galnt2 and Galnt11 (Fig. 1A). At E15 and E17, Galnt2 becomes the most abundantly expressed isoform, followed by Galnt1 at E15 and Galnt4 and Galnt1 at E17 (Fig. 1A). Interestingly, a dramatic change in Galnt expression occurs after birth, as Galnt3 and Galnt12 are the most abundantly expressed isoforms at P2 and 1 week of age (Fig. 1B and Fig. S1A). Moreover, Galnt3 expression continues to increase into adulthood, where it is by far the most abundant isoform at 4, 8, and 12 weeks of age (Fig. 1B and Fig. S1A), in both males and females (Fig. S1B). In situ hybridization revealed that Galnt3 expression is confined to the acinar cells (detected with Aqp5) of the SMGs, the cells responsible for producing the secreted components of the saliva (Fig. 1C).

Previous work has shown that one of the most abundantly expressed Galnts during embryonic stages (Galnt1) plays key roles during SMG growth and development (28). This led us to examine the possibility that Galnt3, the most abundant isoform in adult SMGs, may play a role in adult SMG function. To investigate this, we crossed heterozygous Galnt3-deficient (Galnt3^+/−) animals to generate homozygous Galnt3-deficient (Galnt3^−/−) animals and WT littermate controls. Galnt3^−/− mice have been previously generated as a mouse model for HFTC and were found to have hyperphosphatemia, inappropriately normal levels of 1,25-dihydroxyvitamin D, and decreased levels of circulating intact Fgf23, similar to what is seen in HFTC patients (36). As it is well-documented that rodent SMGs become morphologically distinct between males and females beginning at ∼6 weeks of age (with males developing many more granular convoluted tubules than females) (39), all future experiments analyzed male and female animals separately. Upon examination of SMGs at 8 weeks of age, we found no significant differences in gland weight (Fig. S2, A and B) or morphology (Fig. S2C) between WT and Galnt3^−/− mice for either sex. Previous studies from our laboratory investigating the role of another member of this family in the SMG found an induction of ER stress upon loss of Galnt1 (28). To determine whether the loss of Galnt3 also induced ER stress and the unfolded protein response (similar to that seen in Galnt1^−/− SMGs (28)), we next examined Xbp1 mRNA splicing. Quantification of the ratio of spliced Xbp1 (Xbp1s) to unspliced Xbp1 (Xbp1u) revealed no statistically significant differences between WT and Galnt3^−/− SMGs for either sex (Fig. S2, D–G), indicating that the loss of Galnt3 does not result in ER stress. Additionally, we performed a number of analyses of adult SMG function. Ex vivo analysis of SMG salivary flow rate, volume of saliva secreted, and ion content of saliva revealed no significant differences between WT and Galnt3^−/− animals for either sex (Fig. S3, A–F). These results indicate that the loss of Galnt3 does not result in overt differences in SMG size, morphology, or function via these assays.

Loss of Galnt3 alters the oral microbiome

In addition to its role in hydration and lubrication of the oral cavity, saliva is believed to play a role in regulating the oral microbiota (40). This led us to investigate the effect of loss of Galnt3 on the microbial communities within the oral cavity through the characterization of the local microbiome. Microbiome studies analyze the numbers and types of distinct microbial taxa as well as their relative abundances to provide assessments of microbial community structures. We collected oral mucosal samples as described previously (41) from individual WT and Galnt3^−/− mice at both 8 and 12 weeks of age and performed 16S rRNA gene sequencing. β-diversity comparisons of the microbial community structure within the oral cavity across genotypes and age groups were assessed using the θ Yue and Clayton (θ YC) distance (which takes into account the specific bacterial species present as well as their relative abundances within the communities being compared) and visualized on two-dimensional principal coordinates analysis (PCoA) plots (Fig. 2). We also analyzed α-diversity within each genotype and age group (which measures the number of bacterial taxa and their relative proportions within a particular group), using the nonparametric version of the Shannon index (Fig. S4).

Through β-diversity analysis, we found distinct microbial communities in Galnt3^−/− male samples compared with WT at both 8 and 12 weeks of age (Fig. 2, A and B). PCoA plots demonstrate that there was significant clustering and separation of male Galnt3^−/− microbial communities from male WT communities (p < 0.05 as determined by analysis of molecular variance (AMOVA); Fig. 2, A and B). In contrast to male samples, no significant differences between the community structure of WT and Galnt3^−/− female samples were observed for either age (Fig. 2 (C and D) and Fig. S5 (A and B)). No significant differences in α-diversity were observed for either males or females (Fig. S4, A–D). These results indicate that there are significant microbial shifts in the oral microbiome community structure in males upon loss of Galnt3.

Examining the composition of the oral microbiome in more detail, we found specific changes in the relative abundance of select taxa between male WT and Galnt3^−/− animals. At 8 weeks of age, two operational taxonomic units (OTUs) from the Pasteurellaceae family (which had a high percentage of similarity with the species Muribacter muris and Rodentibacter heidelbergensis), were overrepresented in WT samples, whereas a Lactobacillus sp. (Lactobacillus faecis) and two other low-abundance OTUs (an unclassified Lachnospiraceae sp. and a Candidatus Saccharibacteria/TM7 sp.) were increased in Galnt3^−/− samples (Fig. 2E). At 12 weeks of age, a Streptococcus sp. (Streptococcus danieliae) was overrepresented in WT samples, whereas a Pasteurellaceae sp. (R. heidelbergensis) became overrepresented in Galnt3^−/− samples (Fig. 2F). Interestingly, 12-week Galnt3^−/− samples continued to show an increase in Lactobacillus sp. (L. faecis) as compared with WT (Fig. 2F). It is notable that these data come from mice of both genotypes housed across multiple cages, eliminating the possibility that the distinct microbial communities are due to cage-specific effects. We next examined the stability of the oral microbiome in individual adults of each genotype over time. Similar to previous studies in mice and humans (6, 42), the microbial community structure of WT animals was stable over time, as no significant differences were found between 8- and 12-week samples for either male or female animals (Fig. 3 (A and B) and Fig. S5 (C and D)). Likewise, no significant differences were seen in α-diversity (Fig. S4, E and F). However, significant differences in community structure were seen over time in Galnt3^−/− animals (Fig. 3, C and D). For both males and females, 8-week Galnt3^−/− oral mucosal communities significantly separated from 12-week Galnt3^−/− communities, and these differences were found to be statistically significant by AMOVA (p < 0.05) (Fig. 3, C and D). Investigating age-related differences in the relative abundance of bacterial taxa in more detail in male Galnt3^−/− samples, we found that the OTUs Pasteurellaceae sp. (M. muris) and Pasteurellaceae sp. (R. heidelbergensis) were overrepresented in 12-week samples as compared with 8 weeks (Fig. 3E). Similarly, these two OTUs belonging to the Pasteurellaceae family were also more abundant in 12-week female Galnt3^−/− samples as compared with 8 weeks, whereas a Streptococcus sp. (S. danieliae) was overrepresented at 8 weeks (Fig. 3F). Whereas there were no significant changes in α-diversity in Galnt3^−/− males over time, there was a significant increase in α-diversity in 12-week Galnt3^−/− females (Fig. S4, G and H). These results indicate that the composition of the oral microbiome is less stable over time upon loss of Galnt3 in both males and females.

We also compared the microbiomes between WT males and females at each time point. Whereas a difference in structure was seen at 8 weeks of age (Fig. S6A), this difference was no longer significant by 12 weeks of age (Fig. S6B). Examination of the taxonomy data reveals that the differences at 8 weeks of age are driven by differences in the relative abundance of the major taxa rather than the presence or absence of unique taxa (Fig. S6, C and D). Likewise, there were no significant differences in α-diversity between WT males and females at either age (Fig. S4, I and J).

Galnt3 glycosylates Muc10

To investigate how the loss of Galnt3 (which encodes an O-glycosyltransferase that catalyzes the first committed step in O-glycan biosynthesis) may be influencing the composition and stability of the microbiome, we next examined the O-glycans that are normally present within the SMG using lectins that are specific to O-glycan structures. Previous studies have documented distinct gender differences in the expression of the ST3Gal1 gene (which encodes a sialyltransferase that adds sialic acid in an α2,3 linkage to glycans), with this gene being 22-fold more abundantly expressed in the SMGs of females versus males ((https://sgmap.nidcr.nih.gov/cgi-bin/sgexp11.plx?ind=1&gn=St3gal1). Indeed, female SMGs stained abundantly with the lectin Maackia amurensis agglutinin (MAA) (Fig. 4A), which detects α2,3-linked sialic acid linked to either N- or O-glycans (43). This staining was seen in the acinar cells of the SMG (labeled with the marker acinar-1) and was specifically removed upon treatment with an enzyme that removes sialic acid (neuraminidase (NM)), indicating specificity (Fig. 4A). Staining with MAA was also seen in male SMG acinar cells, but to a lesser extent (Fig. 4B). We next used the lectin peanut agglutinin (PNA) (which detects the nonsialylated core 1 structure Galβ1,3GalNAc of O-glycans) in the presence or absence of NM treatment to determine the degree to which O-glycans are sialylated in both WT females and males (Fig. 4, C and D). As shown in Fig. 4D, PNA staining is abundantly present within the acinar cells (labeled with the acinar-1 marker) of male SMGs with or without NM treatment, indicating the presence of nonsialylated core 1 O-glycans in males. However, PNA staining was only detected in female SMGs after NM treatment, indicating that all core 1 O-glycans are modified with sialic acid in females (Fig. 4C). PNA staining in both males and females could be competed away by incubation with galactose, indicating specificity (Fig. S7). Similar results were obtained on western blots of SMG extracts, where female samples showed no PNA reactivity in the absence of NM treatment, whereas male samples had PNA reactivity with or without NM treatment (Fig. 4, E and F). Taken together, these results indicate that distinct differences exist between males and females in the structure of the O-glycans present within the SMGs, with males having both sialylated and nonsialylated core 1 O-glycans and females having core 1 O-glycans that are predominantly sialylated. Therefore, all further analysis of PNA-reactive O-glycans in females was performed on samples that had been treated with NM, whereas male samples were not treated with NM.

We next set out to compare changes in glycans present in WT and Galnt3^−/− SMGs via confocal microscopy. As shown in Fig. 5 (A and B), loss of Galnt3 did not result in obvious changes in MAA staining in either males or females. However, a dramatic decrease in PNA-reactive O-glycans specifically in the acinar cells of male SMGs was seen upon loss of Galnt3 (Fig. 5C). In contrast to males, female Galnt3^−/− SMGs (treated with NM) did not show a noticeable change in PNA-reactive O-glycan staining within the acinar cells relative to WT via confocal imaging (Fig. 5D). These results suggest that the loss of Galnt3 results in a major change in PNA-reactive O-glycans normally present within the acinar cells of male SMGs.

We next set out to identify the potential protein targets of Galnt3 in the SMGs. Mucins are heavily O-glycosylated proteins known to be the major protein products produced by the salivary glands and are thought to mediate interactions with the microbiome (12, 44). Previous studies have identified mucin 10 (Muc10; also known as mouse submandibular gland mucin and proline-rich, lacrimal 1) as the major mouse SMG mucin and have shown that it localizes to the acinar cells of adult SMGs (45, 46). Here, we examined in detail the postnatal gene expression of Muc10 using qPCR (Fig. 6A). At all five stages examined (P2, 1 week, 4 weeks, 8 weeks, and 12 weeks), Muc10 was found to be abundantly expressed, with its expression increasing in adult stages (Fig. 6A). qPCR to other mucins revealed that Muc10 was by far the most abundantly expressed mucin within the SMGs (Fig. 6, A–D). Additionally, we confirmed that Muc10 is also expressed in acinar cells (detected with acinar-1) of SMGs (Fig. 6E) similar to Galnt3, suggesting that Muc10 may be a direct target of Galnt3.

Given the abundant expression of Muc10 within acinar cells of the SMGs, we next set out to determine whether O-glycosylation of Muc10 was specifically affected by the loss of Galnt3. We performed western blot analysis of 8-week WT and Galnt3^−/− SMG extracts, probing for PNA and Muc10 (Fig. 7). In WT male SMGs, two major PNA bands were detected (Fig. 7A). However, in Galnt3^−/− SMGs, the lower-molecular weight PNA-reactive band was greatly reduced or absent, suggesting that this protein is specifically glycosylated by Galnt3. Interestingly, probing the same western blot for Muc10 revealed that the Muc10 band overlaps with the lower PNA-reactive band in WT SMGs, suggesting that Muc10 may be this highly O-glycosylated band (Fig. 7A). In support of this, immunoprecipitation of Muc10 from WT SMGs revealed that it does contain PNA-reactive O-glycans (Fig. S8). Additionally, the Muc10 band in Galnt3^−/− SMGs no longer overlaps with PNA staining and runs at a reduced size, further suggesting the loss of O-glycans on this protein (as these sugar modifications can account for a significant change in molecular weight). These results strongly suggest that Muc10 is specifically O-glycosylated by Galnt3 in male SMGs. western blots of 8-week female SMGs did not reveal the same PNA banding pattern seen in WT males and did not result in the loss of specific PNA-reactive bands, as seen in male Galnt3^−/− SMGs (Fig. 7B). However, the loss of Galnt3 in female SMGs did result in PNA reactivity with altered mobility (Fig. 7B). Additionally, Muc10-reactive bands ran as large smears that overlapped in position with PNA staining in Galnt3^−/− female SMGs (Fig. 7B). These results suggest that Muc10 is also a substrate for Galnt3 in female SMGs and that, while the loss of Galnt3 does not result in a complete disruption of O-glycosylation (as detected by PNA), it may alter the patterns of O-glycosylation in more subtle ways in females.

To further investigate the ability of Galnt3 to glycosylate Muc10, we performed in vitro enzymatic assays using recombinant murine Galnt3 and acceptor peptides from Muc10. Standard transferase assays revealed that Galnt3 was able to glycosylate four different peptides derived from Muc10 (Fig. 7, C and D), providing further evidence that Muc10 is a substrate for Galnt3 in the SMGs. Interestingly, Galnt3 had specific site preferences among the Muc10 peptides, showing highest activity toward the Muc10-s and Muc10-264 peptides, followed by Muc10-273 and Muc10-l (all of which are derived from the repetitive domain of Muc10). Galnt3 did not show activity toward the Muc10-10 peptide (Fig. 7 (C and D) and Figs. S9 and S10). These results are in agreement with predicted site preferences previously determined for Galnt3 (47). Taken together, our results demonstrate that loss of Galnt3 leads to changes in the composition of the oral microbiome in males, possibly due to changes in the glycosylation status of the major salivary mucin Muc10. Moreover, loss of Galnt3 results in altered stability of the oral microbiome in both males and females. Given that mutations in GALNT3 cause the human disease HFTC and that Galnt3^−/− mice serve as a model for this disease, our results suggest that HFTC patients may also have alterations in their oral microbiome that could contribute to documented disease pathology in the oral cavity (Table S1).

Animal breeding and genotyping

The Galnt3-deficient mice were a kind gift from Dr. Michael J. Econs (36). Galnt3-deficient mice were backcrossed into the C57BL/6MHsd inbred mouse background for at least six generations before analysis. Heterozygous Galnt3 animals (Galnt3^+/−) were crossed to generate WT and homozygous Galnt3-deficient (Galnt3^−/−) siblings to be used for all experiments. Genomic DNA was extracted from mouse ear punches using the DNeasy Blood & Tissue Kit (Qiagen #69506), and PCR was performed as described previously (36). Experimental procedures were reviewed and approved by the Animal Care and Use Committee of the National Institutes of Health (ASP #17-833).

Semiquantitative real-time PCR

RNA was isolated and purified using the PureLink^TM RNA Mini Kit (Invitrogen #12183018A). cDNA was synthesized using iScript^TM Reverse Transcription Supermix (Bio-Rad). Primers for Galnt and Mucin genes were designed using Beacon Designer software and are shown in Tables S2 and S3, respectively. qPCR was performed for 40 cycles (95 °C for 10 s and 62 °C for 30 s) using SYBR Green PCR Master Mix, either 1 ng (Galnt genes) or 10 ng (Mucin genes) of each cDNA sample, and the CFX96 real-time system (Bio-Rad). Gene expression was normalized to 29S rRNA. Reactions were run in triplicate and repeated using three or more animals.

Histology

8-Week mouse submandibular glands were fixed in 4% PFA/PBS overnight at 4 °C, transferred to 70% ethanol, and stored at 4 °C. Paraffin-embedded tissue sections (5 μm) were used for hematoxylin/eosin staining, in situ hybridization, and immunofluorescent staining.

In situ hybridization

Primers of Galnt3 probes were designed using Beacon Designer software. Probes were amplified by the KOD kit (TAKARA) and purified by a QIAquick spin kit (Qiagen). Digoxigenin-11-UTP–labeled single-stranded RNA probes were prepared using a DIG RNA labeling kit (Sigma-Aldrich #11175025910). Fixed, paraffin-embedded SMG sections (5 μm) were dewaxed at 60 °C for 10 min followed by three changes of xylene substitute (Sigma #A5597). Sections were rehydrated with two changes of 100% ethanol and single changes of 90, 70, and 50% ethanol. Sections were fixed by 4% PFA/PBS for 5 min, followed by 10 mg/ml Proteinase K (Sigma #4850) treatment for 30 min at 37 °C. After post-fixing with 4% PFA/PBS for 5 min, sections were dehydrated in ethanol series (50, 70, 90, and 100%). Hybridization was performed by incubating sections with 50 ng of Galnt3 RNA DIG-labeled antisense or sense probes in 25 μl of hybridization buffer for 10 min at 85 °C and then overnight at 50 °C. After washing the next day, sections were treated with 10 mg/ml RNase A in 10 mm Tris-HCl (pH 7.5), 500 mm NaCl for 15 min at 37 °C. Following washing, sections were blocked with 1% blocking reagent (Roche Applied Science #1096176) for 30 min at room temperature and then incubated with anti-DIG-POD antibody (Roche Applied Science; 1:100) and Aquaporin-5 primary antibody (Calbiochem #178615; 1:100) overnight at 4 °C. On day 3, sections were washed and then incubated with Alexa Fluor 647®-conjugated anti-rabbit IgG (Jackson ImmunoResearch; 1:400) for 1 h at room temperature. Following washing, sections were incubated with tyramide signal amplification plus Cy3 (TSA; PerkinElmer Life Sciences; 1:50 in amplification buffer) for 30 min at room temperature. After washing, sections were nuclear-counterstained with Hoechst 33342 (Invitrogen #H3570; 1:10,000) and mounted in aqueous Fluro-Gel medium (Electron Microscopy Sciences #17985-10). Sections were visualized on a Nikon A1R confocal microscope, and images were processed in Fiji (National Institutes of Health).

Immunofluorescent staining

Fixed, paraffin-embedded tissue sections (5 μm) were deparaffinized by incubating slides at 60 °C for 10 min, followed by three washes in xylene substitute (Sigma #A5597). Sections were rehydrated with single changes of 100, 90, 70, 50, and 30% ethanol and deionized water. Antigen retrieval was performed by incubating slides in 5% urea and 50 mm β-mercaptoethanol for 10 min at 95 °C, followed by 40 min of cooling. For female samples to be probed with PNA (except where noted), sections were treated with 20 milliunits of neuraminidase (New England BioLabs #P0720) overnight at 37 °C. Sections were blocked with 10% donkey serum, 1% BSA, and M.O.M blocking reagent (Vector Laboratories #FMK-2201). Where noted, some male samples were also treated with neuraminidase via the same procedure. Primary antibodies were diluted in PBS, 0.1% Tween 20 plus M.O.M. protein reagent and incubated overnight at 4 °C for male samples or for 2 h at room temperature for female samples. Primary antibodies used were Muc10 (Everest Biotech #EB10617; 1:200) and acinar-1 (Developmental Studies Hybridoma Bank #3.7A12; 1:20). After washing, sections were incubated for 1.5 h at room temperature with either TRITC-conjugated MAA (EY Laboratories #R-7801-2; 1:100) or FITC-conjugated PNA (Vector Laboratories #FL-1071; 1:200), Cy^TM3-conjugated anti-goat IgG (Jackson ImmunoResearch; 1:400), and Alexa Fluor 647®-conjugated anti-rat IgG (Jackson ImmunoResearch; 1:400). For sugar inhibition of PNA, 0.2 m galactose (Sigma-Aldrich #G0750) was incubated with FITC-labeled PNA for 1 h at room temperature before applying to slides. Nuclear counterstaining was done using Hoechst 33342 (Invitrogen #H3570; 1:10000) for 3 min at room temperature. Stained sections were mounted in aqueous Fluro-Gel medium (Electron Microscopy Sciences #17985-10) and visualized on a Nikon A1R confocal microscope. Images were processed in Fiji, and representative stacked section confocal images are shown in each figure. FITC-labeled PNA was pseudocolored to red, and TRITC-labeled MAA was pseudocolored to green in Fiji (Figs. 4 (A and B) and 5 and Fig. S6).

Immunoprecipitation

Immunoprecipitation was performed using TrueBlot® anti-goat IgG magnetic beads (Rockland Immunochemicals Inc. #00-1844) according to the manufacturer's instructions with modifications. 8-Week WT SMG lysates were prepared in RIPA buffer with sonication. 25 μg of SMG lysates were combined with 5 μg of Muc10 antibody (Everest Biotech #EB10617), diluted to 250 μl in RIPA buffer, and incubated overnight at 4 °C with mixing. The next day, beads were washed with RIPA buffer. 200 μl (1 mg) of prewashed beads were then combined with the preformed SMG lysate/antibody complexes and incubated overnight at 4 °C with mixing. Next, beads were collected, and the supernatant was removed, followed by washing with RIPA buffer. Beads were then resuspended in RIPA buffer and SDS-loading dye (containing a 1:25 dilution of β-mercaptoethanol) and heated at 95 °C for 10 min, followed by cooling on ice for 10 min. The supernatant (containing the target antigen) was then removed from the beads and used for SDS-PAGE and western blotting as described below.

Western blotting

PNA (Arachis hypogaea lectin, EY Laboratories #L-2301) was labeled with IRDye 680LT (LI-COR Biosciences #928-38066). 8-Week SMGs from littermate WT and Galnt3^−/− animals were lysed in RIPA buffer with sonication. SMGs from male and female animals were treated with 6.25 milliunits of neuraminidase where noted (New England BioLabs #P0720) for 1 h at 37 °C. Samples were analyzed by SDS-PAGE under denaturing and reducing conditions and transferred to nitrocellulose membranes. Membranes were blocked with 1:1 PBS/Odyssey Blocking Buffer (LI-COR Biosciences #927-4000) and probed with Muc10 (Everest Biotech #EB10617; 1:2000). After washing, membranes were probed with IRDye 680LT-conjugated PNA (1:10000) and IRDye 800CW-conjugated anti-goat IgG (LI-COR Biosciences #925-32214; 1:5000). Membranes were scanned using a LI-COR Odyssey IR imaging system.

Xbp1 mRNA splicing PCR

PCR primers used to detect splicing of Xbp1 were published previously (28). PCR was performed for 35 cycles (94 °C for 30 s, 61 °C for 30 s, 72 °C for 90 s) using 10 ng of each cDNA sample and a Veriti^TM thermal cycler (Applied Biosystems). PCR products were identified using 4–20% TBE gels and ethidium bromide. Negative images of gels are shown. NIH Fiji was used to measure band intensity, and normalized ratios of spliced Xbp1/unspliced Xbp1 were determined (ratio for WT controls set to 1).

Enzyme assays

Expression of recombinant murine Galnt3 was performed as described previously (38) using COS7 cells. Assays for Galnt3 activity were performed as described previously (38) using 5 μl of media from COS7 cells transfected with murine Galnt3. Muc10 and EA2 peptide substrates were synthesized by Peptide 2.0 (Fig. 5D and Fig. S6). Reactions were run for 1.5 h at 37 °C and then quenched with 30 mm EDTA. Glycosylated products were separated from unincorporated UDP-[¹⁴C]GalNAc by anion-exchange chromatography using AG1-X8 resin columns (Bio-Rad #1401454), and product incorporation was determined by liquid scintillation counting. Background values were obtained from reactions using media from COS7 cells transfected with empty vector and subtracted from each experimental value. Assays for each peptide substrate were run in triplicate and repeated three times.

Ex vivo perfused SMG

Ex vivo SMG saliva secretions were collected by modifying the ex vivo mouse SMG perfusion technique (58). In brief, mice were anesthetized by chloral hydrate injection (400 mg/kg intraperitoneally), and all branches of the common carotid artery were ligated, except for the artery supplying blood to SMG. Intact glands were then removed and transferred to a temperature-controlled perfusion chamber at 37 °C. Then the common artery was cannulated with a 31-gauge cannula and perfused with a solution containing the muscarinic receptor agonist (CCh, 0.3 μm) plus β-adrenergic agonist (IPR, 1.0 μm). The SMG duct was placed in a calibrated glass capillary tube. The perfusion rate was kept at 1.0 ml/min with a peristaltic pump (IPC-8, ISMATEC, Wertheim, Germany). The ex vivo perfusion solution contained 4.3 mm KCl, 120 mm NaCl, 25 mm NaHCO₃, 1 mm CaCl₂, 1 mm MgCl₂, 5 mm glucose, and 10 mm Hepes, pH 7.2 (gassed with 95% O₂, 5% CO₂). Secreted saliva samples were collected and stored at −80 °C until analyzed.

Ion concentrations in SMG secretions

Sodium and potassium concentrations were analyzed by atomic absorption spectroscopy (3030 spectrophotometer, PerkinElmer Life Sciences). Chloride activity was measured using a chloride electrode (9617BNWP, Thermo Fisher Scientific) with a 4-Star Plus pH/ISE Benchtop Multiparameter Meter (Thermo Fisher Scientific).

Sample collection and DNA extraction for microbiome analysis

To allow a comprehensive sampling of oral mucosal surfaces, we collected oral mucosal samples from 8-week and 12-week mice. We utilized ultra-fine polystyrene swabs to obtain microbial samples from the murine oral cavity for 30 s, and then these samples were placed in 150 μl of TE buffer and stored at −80 °C until processing. DNA was extracted following a modified protocol of the DNeasy Blood and Tissue Kit (Qiagen) as described previously (41).

16S rRNA gene-based sequencing and bioinformatic analyses

Library generation was performed using Eppendorf liquid-handling robots. The V4 region of the 16S rDNA gene (515F-806R) was sequenced for oral mucosal samples, generating paired-end, overlapping reads on the Illumina MiSeq platform (59). Sequencing data were processed using the software mothur (60). Briefly, reads were quality-filtered and assembled into contigs, and only reads from 200–400 bp in length were kept. Then our processing continued following the MiSeq SOP pipeline as described (61). OTUs were defined at a 97% similarity, and genus-level taxonomical classification was achieved. However, we further informed our classification down to species level through blasting the reference sequence of each OTU against the NCBI 16S rRNA database using BLAST, as described previously (41). β-Diversity was assessed using the θ YC distance. α-Diversity was evaluated using the nonparametric version of the Shannon diversity index. Both diversity measurements were calculated as implemented in mothur. All sequence data from this study have been submitted to the NCBI SRA database under SRA accession number PRJNA547610.

Statistical analysis

Error bars represent S.D. for values obtained from three or more mice. Student's t test and Mann–Whitney test were used to calculate p values. For microbiome analyses, AMOVA was used to test for differences in community structure. LEfSe (62) was used to test for differences in relative abundance of taxa, considering 0.05 as the α value for statistical testing.