Metabolic-associated fatty liver disease (MAFLD), a terminology updated from non-alcoholic fatty liver disease (NAFLD) in 2020, is the hepatic manifestation of the metabolic syndrome with a broad spectrum of liver conditions ranging from simple hepatic steatosis to steatohepatitis to stage 4 fibrosis (Demirtas and Yilmaz, 2020; Eslam et al., 2020). It usually manifests clinically silent and has no approved pharmacotherapy; with time, it can gradually progress to end-stage liver diseases, such as cirrhosis and hepatocellular carcinoma(Eslam et al., 2020). Accumulating evidence has indicated that MAFLD is strongly associated with increased risks of incident diabetes, chronic kidney disease, and cardiovascular disease, with the magnitude of risk seeming to parallel the severity of MAFLD(Liang et al., 2021; Targher et al., 2021). Globally, MAFLD affects approximately a quarter of the adult population, placing an enormous burden on healthcare systems and society(Eslam et al., 2020). Thus, convenient diagnostic screening and timely strategic intervention for MAFLD are of great significance.

The oral microbiota is the second most diverse microbial ecosystem after that of the gut in the human body, consisting of over 700 bacterial species. It plays an essential role in physiological, metabolic, and immunological functions, which include nutrient digestion, metabolic regulation, immune response, and antibacterial activity (Kilian et al., 2016). In addition to oral diseases, oral microbiota dysbiosis has also been closely linked with metabolic disorders, including MAFLD (Zhao et al., 2020; Peng et al., 2022). An animal study demonstrated that the endotoxemia from Porphyromonas gingivalis was a remarkable risk factor for NAFLD pathogenesis, and the altered glucose/lipid metabolism may facilitate disease progression(Sasaki et al., 2018). Our previous work has also confirmed the community structure alterations and microbial dysbiosis changes of supragingival microbiota in patients with MAFLD(Zhao et al., 2020). Given the oral cavity being a complex microbial environment, further insights into the alterations of the oral microbial community under the pathological state of MAFLD are hence warranted.

Saliva is a heterogeneous biofluid containing various proteins, metabolites, microbes, and their genes, which are essential for maintaining oral homeostasis(Belstrøm, 2020). Spreading throughout the entire oral cavity, saliva theoretically acts as a reservoir pool of microorganisms detached from various oral niches and appears representative of the overall oral microbiome(Yoshizawa et al., 2013; Belstrøm, 2020). Besides, saliva-based microbial, immunologic, and molecular biomarkers have been progressively investigated as diagnostic tools for several diseases due to the remarkable advantages of a non-invasive, stress-free, and cost-effective sampling manner(Yoshizawa et al., 2013; Zhang et al., 2016). With the rapid development of next-generation sequencing technology, a growing number of studies have evaluated the alterations of salivary microbial communities and the diagnostic utility in various diseases, such as type 2 diabetes mellitus, schizophrenia, and rheumatic heart disease, among others(Zhang et al., 2016; Liu et al., 2021; Qing et al., 2021; Shi et al., 2021). However, there is a paucity of information about salivary microbiota’s potential role in MAFLD.

This study aimed to investigate the ecological shifts in the salivary microbiome of MAFLD patients and explore the potential function of salivary microbiota. Firstly, microbiota profiles of saliva samples were characterized by 16S rRNA gene sequencing to identify changes between MAFLD patients and control subjects. Then, the potential value of the saliva-based microbiota diagnostic model was evaluated to discriminate MAFLD patients from control subjects. Finally, the relationships between clinical variables and specific microbial genera were assessed. These findings could provide more significant insights into the salivary ecological dysbiosis associated with MAFLD.

The salivary microbiota, which serves as a reservoir of microorganisms from different sites within the oral cavity, could represent the whole oral microbiota and reflect the oral and general health status(Belstrøm, 2020). An increasing number of metabolic disorders have been reported in potential relation to the shifts in salivary microbial ecology, such as atherosclerosis, type 2 diabetes mellitus, polycystic ovary syndrome, and hepatic encephalopathy (Yoshizawa et al., 2013; Zhang et al., 2016; Liu et al., 2021). However, to the best of our knowledge, few studies have explored the microbiome-level association between salivary microbiota and MAFLD. This study discovered that abnormal metabolic levels in MAFLD patients had significantly altered the composition and structure of salivary microbiota, which had good diagnostic power in discriminating MAFLD patients from healthy controls.

Oral microbial diversity is essential for maintaining health and varies with the physiological state of the host. Generally, elevated microbiome richness and diversity have been recognized as hallmarks of a healthy ecosystem, especially for the gut microbial ecosystem(Falony et al., 2018). However, our present study showed that the salivary microbiota in MAFLD patients exhibited increased α-diversity (ACE index and Chao1 index). Although no significant differences were observed in the Shannon and Simpson indices, both showed marginally elevated diversity among patients with MAFLD. These findings are consistent with our previous study that focused on the shifts in supragingival microbiota associated with MAFLD(Zhao et al., 2020). Following the concept proposed by Takeshita et al. (Takeshita et al., 2016), poor oral health could lead to an increased taxonomic richness in saliva. Because prolonged plaque accumulation may cause the multiplication of attached bacteria, and the gingival bleeding could provide rich nutrients for bacteria growth. These ecological shifts may promote bacterial assemblage in saliva. This phenomenon has been previously observed in patients with other metabolic diseases(Si et al., 2017). Furthermore, the PCoA and PLS-DA results revealed a significant clustering of microbial communities between the two groups. Thus, the altered metabolic level in patients with MAFLD would provoke salivary microbial shifts and increase the risk of oral diseases.

Phyla Firmicutes and Proteobacteria showed divergent abundance trends, with Firmicutes enrichment and Proteobacteria depletion from the control group to the MAFLD group, resulting in an increase in Firmicutes/Proteobacteria ratio. The increased Firmicutes/Proteobacteria ratio was also detected in the salivary microbiome of patients with primary Sjögren’s syndrome(Van der Meulen et al., 2018) and schizophrenia(Qing et al., 2021). Similar to MAFLD, these two diseases are characterized by chronic low-grade inflammation. Lau et al. also revealed a shift in gut microbiome toward decreased Firmicutes/Proteobacteria ratio in type 2 diabetic adults with mild obesity following metabolic surgery (Lau et al., 2021). A probable reason is that Proteobacteria are a major phylum of Gram-negative bacteria, mainly associated with glucose homeostasis improvement, metabolic amelioration, and inflammatory response reduction(Carvalho et al., 2012; Lau et al., 2021). In contrast, the phylum Firmicutes are mostly Gram-positive bacteria playing a pivotal role in the fermentation and metabolism of carbohydrates and lipids through chain-breaking fatty acid synthesis, which usually facilitates obesity development(Stojanov et al., 2020). Therefore, it is indicative that Firmicutes bacteria could have a competitive advantage over Proteobacteria bacteria in niche occupancy during the MAFLD state. However, whether the Firmicutes/Proteobacteria ratio could be used as a marker of ecosystem health in microbiome research still requires more studies to confirm.

LEfSe and ANCOM analyses revealed notable discrepancies between the two groups. Genus Filifactor was enriched in MAFLD patients, whereas genera Neisseria and Capnocytophaga were more abundant in healthy controls. Of the remaining genera, Treponema, Fretibacterium, Propionibacterium, Shuttleworthia, Eggerthia, Bulleidia, Howardella, and Desulfobulbus, presented a higher proportion in the MAFLD group. Filifactor spp. are well-known periodontal pathogens (Takeshita et al., 2016) and have been reported to exist in high abundance in the salivary microbiota of patients with hepatocellular carcinoma (Park et al., 2021). The genus Treponema consists of a diverse group of pathogenic or commensal microorganisms that not only cause syphilis, yaws, and pinta infections(Shukla et al., 2022) but are also associated with chronic liver disease(Ling et al., 2015), infective endocarditis(Hijikata et al., 2019), and Alzheimer’s disease (Riviere et al., 2002). Besides, oral Treponema denticola, together with Porphyromonas gingivalis and Tannerella forsythia, are known as the “red complex” which have been widely considered essential pathogens in periodontal disease etiology and pathogenesis. A series of animal and human studies have confirmed that the genera Fretibacterium (Yamamoto et al., 2021), Propionibacterium (Del Chierico et al., 2017), Shuttleworthia (Xie et al., 2016), Eggerthia (Yamamoto et al., 2021), and Bulleidia (Bashiardes et al., 2016)are prevalent in the oral or gut microbiota with liver disease including nonalcoholic fatty liver disease, hepatocellular carcinoma. These findings could be interpreted by the emerging oral-gut-liver axis that oral microbes could translocate into the gastrointestinal tract, where they would spread to the liver and induce hepatic diseases(Imai et al., 2021). On the other hand, Neisseria spp. (Liu et al., 2015) are generally Gram-negative microorganisms comprising mainly non-pathogenic species in the oral cavity. Previously, we also demonstrated a higher abundance of Neisseria in the supragingival plaque among healthy individuals compared to those with MAFLD (Zhao et al., 2020). Capnocytophaga is a genus of Gram-negative anaerobic bacteria reportedly linked with improved oral health and reduced caries experience (Schoilew et al., 2019). Interestingly, the genus Neisseria and Capnocytophaga have been identified as part of human oral health’s “core microbiome” (Zaura et al., 2009).

Microorganisms coexist in intricate networks of interactions, which in turn affect the species involved and may lead to disease progression. As shown in the co-occurrence networks diagram, the inter-genera interactions in MAFLD saliva exhibit higher network transitivity and lower average shortest path lengths. Transitivity measures the average connectedness of a network, with higher values indicating the presence of more tightly connected clusters (more inter-genus interactions) (Sisk-Hackworth et al., 2021). By contrast, the lower average shortest path lengths within microbial networks suggest species are interconnected through shorter paths (Loftus et al., 2021). The underlying reason may be that short path lengths and intimate connections within the salivary networks in patients with MAFLD could transmit signals rapidly between bacterial species, thereby potentially promoting shifts in community metabolism.

A timely and accurate diagnosis of MAFLD is a prerequisite for expeditious therapeutic interventions, which could inhibit disease progression. The random forest analysis showed that the salivary microbiota signature has good diagnostic power (bacterial taxa achieved an AUC of 0.82) to discriminate MAFLD patients from healthy controls. These findings reinforce the notion that saliva microbes potentially rich in diagnostic biomarkers could serve as indicators of both oral and systemic diseases(Yoshizawa et al., 2013; Zhang et al., 2016). Using 16S rRNA sequencing of fecal samples from children, Schwimmer et al. reported that prediction models combining serum ALT levels and relative abundance of encoding genes could discriminate participants in NAFLD, NASH, and severe fibrosis with AUCs of 0.95, 0,92, and 0.87, respectively(Schwimmer et al., 2019). Although gut microbiota prediction seems superior for MAFLD, given concerns about noninvasiveness and accessibility, the biochemical blood indices were not incorporated into the saliva-based biomarker model in this study. The key next step is to generalize and transform these critical microbial biomarkers into available tools for clinical practice(Mouzaki and Loomba, 2020).

With numerous studies reporting the complex etiology and pathogenesis of MAFLD, it has become clear that underlying risk factors for MAFLD development encompass IR, obesity, and dyslipidemia(Duseja and Chalasani, 2013), among which IR takes center stage (Kuraji et al., 2021). IR means the cells in the body muscles, fat, and liver cannot respond well to the normal concentration of insulin hormone that promotes glucose uptake and fat storage. Therefore, IR caused by the imbalance between energy intake and expenditure could lead to fat accumulation in the liver and promote dyslipidemia through increased circulating free fatty acids in blood(Kuraji et al., 2021). As first described by Matthews et al. in 1985, HOMA-IR has been commonly used for IR estimation in clinical studies, with a HOMA-IR value ≥2.5 indicating IR(Matthews et al., 1985).In this research, HOMA-IR in the MAFLD group exceeded the threshold value and was significantly higher than in the control group, suggesting that IR is highly prevalent among patients with MAFLD. Moreover, the elevated HbA1c and TG/HDL-C ratio, known as a simple and reliable marker of IR, further demonstrated the propensity for IR in the MAFLD group(Borai et al., 2011; Fan et al., 2019). In addition, elevated serum transaminase levels and a decreased AST/ALT ratio are markers of ongoing hepatocellular injury. They are commonly deranged in patients with progressive MAFLD(Sheka et al., 2020), implying that enrolled MAFLD patients had different degrees of hepatocellular damage.

The RDA analyses further verified the significant influences of IR (p = 0.007 for HOMA-IR; p = 0.018 for HbA1c) and obesity (p = 0.003 for BMI; p = 0.006 for WHR) over the salivary microbiota distribution. As shown in the spearman correlation heatmap (between the top 20 abundant genera and clinical variables), although some correlations did not reach statistical significance, Neisseria, Capnocytophaga, Haemophilus, and Rothia tended to be negatively associated with IR, obesity, and dyslipidemia index. These genera are members of nitrate-reducing bacteria residing in the oral cavity and exerting beneficial effects on hosts through the nitrate–nitrite–nitric oxide pathway, such as regulating glucose metabolism, lowering lipid levels, and reducing inflammation(Rosier et al., 2022). However, the specific role and related mechanisms of these oral nitrate-reducing microbes in MAFLD remain to be further delineated. The negative association results support the presumption that metabolic abnormalities (e.g., dyslipidemia, IR, and obesity) may play a role in modulating oral microbial composition and cause oral dysbiosis(Negrini et al., 2021). Genus Fusobacterium was positively correlated with TC and LDL-C levels, consistent with the concept that the presence of Fusobacterium spp. is implicated in the pathogenesis of atherosclerosis via aberrant lipid metabolism(Zhou et al., 2022).

The analysis with PICRUSt2 demonstrated that the main functional alterations between the two groups were metabolic-related pathways. The upregulation of the pathways for pyrimidine metabolism and fructose and mannose metabolism is consistent with our previous supragingival plaque study on MAFLD patients(Zhao et al., 2020), further confirming the metabolic-related pathways of oral bacteria are associated with the occurrence of MAFLD. Pathways related to amino acid metabolism (including glutathione metabolism and tyrosine metabolism) were more prevalent in healthy subjects, possibly because the MAFLD patients have disturbances in amino acid metabolism caused by hepatocellular injury. However, further metabolome studies are needed to investigate the metabolism in MAFLD and its relationship to the salivary microbiome.

A limitation of this study was the limited sample size. Although the inclusion criteria were rigorous, future investigations with a larger sample size were needed to control for potential confounding factors. Moreover, due to the limitation of 16s rRNA sequencing, metagenomics sequencing may be required to genetically annotate and validate functional information related to salivary microbiome changes in patients with MAFLD.

In conclusion, we comprehensively described salivary microbiome alterations between MAFLD patients and healthy individuals, evaluated the potential value of salivary microbiota as an auxiliary diagnostic tool to predict MAFLD, and demonstrated the role of biochemical variables, especially the IR, in microbial ecological shifts. This study provides an in-depth view of the association between the salivary microbiome and MAFLD, which may contribute to developing strategies for the prevention, diagnosis, and treatment of MAFLD.

The data presented in the study are deposited in the NCBI Sequence Read Archive (SRA) repository, accession number PRJNA929590 (https://www.ncbi.nlm.nih.gov/bioproject/929590, PRJNA929590).

The studies involving human participants were reviewed and approved by Ethics Committee of School and Stomatology Wenzhou Medical University. The patients/participants provided their written informed consent to participate in this study.

MW, C-GN and C-YQ performed the study design, data analysis, drafting, and revising of the work. L-YY and C-CZ performed the data analysis and acquisition. Y-HP and Z-HW contributed to the study design, clinical sample collection, data analysis, drafting, revising, and final approval. All authors contributed to the article and approved the submitted version.