Distinct Gut Microbiota Profiles Reflect Severity in Chronic Insomnia Disorder

Insomnia disorder is the most prevalent sleep disorder, characterized by difficulty initiating or maintaining sleep despite having sufficient sleep opportunities, leading to daytime functional impairment (Perlis et al. 2022). The global prevalence of insomnia disorder in adults ranges from 10% to 20%, with approximately 50% of cases following a chronic course (Morin et al. 2006; Morin and Benca 2012). Despite the high prevalence and significant disease burden, insomnia is often under‐recognized and inadequately treated (Morin et al. 2006; Morin and Benca 2012), which constitutes a pressing clinical challenge.

Currently, the diagnosis of insomnia disorder primarily relies on patients’ subjective reports rather than objective sleep measurements (Morin et al. 2015), potentially leading to clinical assessment biases. A 10‐year prospective study found that the remission rate for individuals with severe insomnia symptoms was only 56% over the course of a decade (Janson et al. 2001). The severity of insomnia may serve as an important predictor of the risk for disease progression, with more severe insomnia symptoms being associated with an increased risk of persistent sleep disturbances (Morin et al. 2020). Therefore, timely and accurate identification of patients with severe insomnia is crucial for optimizing treatment management and improving outcomes. Furthermore, despite advances in both pharmacological and non‐pharmacological treatments for insomnia, significant individual variability in clinical practice remains, resulting in the absence of standardized treatment protocols for different patients (Yue et al. 2023), which increases the uncertainty in treatment decision‐making. This highlights the urgent need for the exploration of novel, personalized therapeutic approaches.

Since the introduction of the microbiome‐gut‐brain axis (MGBA), gut microbiota have been recognized for their significant role in maintaining human health (Gilbert et al. 2016; Qin et al. 2010; Sonnenburg and Bäckhed 2016) and are closely linked to an increased risk of various diseases in the host (Martin et al. 2018). Recent years have seen growing recognition of the bidirectional nature of the MGBA, which may serve as the basis for the interaction between sleep and the gut microbiome (Wang et al. 2022). Animal studies have shown that chronic sleep fragmentation exacerbates gut microbiota dysbiosis in mice (Poroyko et al. 2016), while depletion of the gut microbiome negatively affects sleep structure (Ogawa et al. 2020). Clinical studies have found that sleep loss significantly affects the composition of the human gut microbiome, with changes in the Firmicutes to Bacteroidetes (F/B) ratio and the abundances of several gut microbiota species (Benedict et al. 2016). Critically, alterations in gut microbiota have also been linked to insomnia disorder: Liu et al. first reported that patients with chronic insomnia disorder (CID), defined as those with a disease duration of at least three months, have significant deviations in the structure and function of their gut microbiota compared to healthy individuals (Liu et al. 2019). Another study similarly found gut microbiota changes in both chronic and acute insomnia disorder (defined as those with a disease duration of less than three months), with specific bacterial signatures emerging as important biomarkers for identifying insomnia disorder and showing a significant correlation with sleep quality (Li et al. 2020). Additionally, in elderly patients with insomnia disorder, objective sleep efficiency was a significant predictor of variations in microbiota composition, with sleep quality being associated with the abundance of specific bacterial taxa (Haimov et al. 2022). A cohort study also found differences in gut microbiota characteristics between general populations with different sleep quality, with a more stable bacterial interaction pattern observed in those with good sleep quality, suggesting a more consistent co‐occurrence in this group (Seong et al. 2024). Collectively, these findings establish a clear association between gut microbiota dysbiosis and insomnia disorder, suggesting distinct microbial profiles may underlie clinical subtypes. This underscores the need for stratified research across different CID severity levels to identify characteristic microbiota features. Such features could serve as objective indicators for detecting severe cases, offering significant clinical value and informing personalized treatment strategies. However, despite this potential, a critical gap remains: existing studies have largely neglected stratification based on clinically defined severity. Consequently, the specific gut microbiota signatures associated with varying degrees of insomnia severity are poorly characterized. Furthermore, while microbiota‐targeted interventions show promise for sleep disorders (Li et al. 2023; Mudaliar et al. 2024), the evidence remains limited (Gil‐Hernández et al. 2023) and sometimes conflicting (Ho et al. 2021; Wu et al. 2021), particularly in severity‐stratified CID cohorts.

Therefore, to address this need, our study aimed to (1) characterize gut microbiota structural variations across CID patients with severity‐based stratification using 16S rRNA gene sequencing. (2) Infer microbial metabolic pathways associated with CID severity through PICRUSt2 analysis (Douglas et al. 2020). (3) Evaluate the discriminative capacity of microbiota features for clinical severity subgroups. These investigations seek to provide mechanistic insights for future therapeutic strategies in CID management.

This study investigated the relationship between sleep quality and gut microbial structure and function in patients with CID. We found that patients with poorer sleep quality exhibited more significant dysbiosis in terms of microbiota diversity, composition, and function. Furthermore, the random forest model demonstrated moderate discriminative capacity for more severe CID cases, suggesting microbial features may complement existing clinical assessments in severity stratification. Notably, this study employed a cross‐sectional design. Due to the inherent limitations of this approach, the findings demonstrate associations but cannot support causal inferences. Longitudinal or interventional studies are necessary to elucidate temporal dynamics. Nevertheless, our results highlight potential avenues for future research and provide preliminary evidence to support subsequent longitudinal or intervention studies targeting the gut microbiota in CID patients.

In the present study, we conducted a comprehensive comparison of gut microbial characteristics between CID patients exhibiting severe and milder poor sleep quality. Changes in bacterial diversity and composition are critical indicators of gut microbiota dysbiosis. Previous studies have reported significant alterations in both α‐ and β‐diversity in animal models of fragmented sleep (Triplett et al. 2020; Yang et al. 2023). In CID patients, significant alterations in both α‐ and β‐diversity were also found (Liu et al. 2019), and these changes were associated with the pathological stages of insomnia, with chronic insomnia showing significantly lower α‐diversity than acute insomnia (Li et al. 2020). Our study further revealed significant differences in both α‐ and β‐diversity between the S‐CID and HC groups, as well as notable differences between the S‐CID and M‐CID groups. However, no significant differences were observed between the M‐CID and HC groups. The F/B ratio is another widely used metric for evaluating gut dysbiosis and has been linked to several diseases (Wei et al. 2021). While some studies of acute sleep deprivation have reported increased F/B ratios (Benedict et al. 2016), research specifically focusing on clinical insomnia presents a distinct pattern: rodent models of insomnia show increased Bacteroidetes and decreased Firmicutes (Ren et al. 2024), and human studies have consistently reported reduced F/B ratios in CID patients (Liu et al. 2019). Our study found that a reduced F/B ratio was present only in the S‐CID group, while no significant differences were observed between the M‐CID and HC groups. This differential pattern suggests that a pronounced reduction in the F/B ratio may be a characteristic feature of more advanced disease stages rather than a universal marker of CID. Collectively, these results indicated that CID patients with poorer sleep quality exhibit more pronounced gut microbiota dysbiosis, implying that poorer sleep quality may indicate a more advanced stage of the disease, whereas milder sleep disturbances may represent a transitional phase from health to severe insomnia.

By combining two machine learning algorithms with difference analysis, our study identified seven key bacterial genera belonging to two phyla—Firmicutes and Bacteroidetes: Clostridium, Ruminococcaceae, Lachnospira, Oscillospira, and Phascolarctobacterium from Firmicutes; and Bacteroides and Parabacteroides from Bacteroidetes. Existing research has established that short‐chain fatty acids (SCFAs), specifically acetate, propionate, and butyrate produced by gut microbiota, are crucial for sustaining human health and can significantly impact sleep (Markowiak‐Kopeć and Śliżewska 2020; Wang et al. 2022). Notably, butyrate in particular has been implicated in beneficial regulatory effects on sleep and gut homeostasis (Hays et al. 2024; Leonel and Alvarez‐Leite 2012; Szentirmai et al. 2019), and a reduction in butyrate‐producing bacteria abundance has been noted in insomnia patients (Li et al. 2020; Wang et al. 2024). Among the identified genera, Clostridium is a typical butyrate‐producing bacterium that also generates acetate and propionate (Zhu et al. 2022); thus, its depletion may influence SCFA production. Supporting its relevance to insomnia, previous studies have recognized Clostridiales (the order containing Clostridium) as a key biomarker for identifying CID patients, with a significant negative correlation with PSQI scores (Liu et al. 2019). Our current study revealed a notable reduction of Clostridium in the S‐CID group, alongside a tendency of diminished abundance in the M‐CID group relative to the HC group. Further analysis revealed significant correlations between Clostridium and several subjective and objective sleep parameters, with a significant negative correlation with PSQI, ISI, and AIREM, and a significant positive correlation with R%. The correlations with R% and AIREM were particularly relevant. REM sleep‐related processes are important for subjective sleep quality (Feige et al. 2018). AIREM is a crucial index for quantifying fragmented sleep and reflects instability in REM sleep (Feige et al. 2023). Insomnia patients often exhibit increased wakefulness during REM sleep (Feige et al. 2023). As the most aroused sleep state, prolonged REM sleep may be particularly vulnerable to perception as wakefulness (Pérusse et al. 2015; Siegel 2011). Therefore, considering the reduction of Clostridium in CID groups, its established role in SCFA (especially butyrate) production, and its significant correlations with key REM sleep parameters, which are crucial for sleep quality, we hypothesize that Clostridium might influence REM sleep and sleep quality by modulating SCFA metabolism, with higher Clostridium abundance potentially playing a protective role in CID patients. Although the precise mechanisms linking SCFAs (especially butyrate) to sleep require further elucidation, accumulating evidence indirectly supports our hypothesis. For instance, a recent study demonstrated that fecal microbiota transplantation from insomnia patients into germ‐free mice induced insomnia‐like behaviors, which were accompanied by reduced serum butyrate levels and hyperactivity of hypothalamic orexin neurons. Importantly, intervention with tributyrin (a butyrate prodrug) suppressed orexin neuron activation and ameliorated sleep disturbances, suggesting that gut microbiota may contribute to sleep disorders through disrupted butyrate metabolism and impaired hypothalamic neuronal homeostasis (Wang et al. 2024). Furthermore, Szentirmai et al. reported that butyrate promotes NREM sleep in mice, potentially through sensory mechanisms located in the liver and/or portal vein system (2019). In a Parkinson's disease mouse model, butyrate supplementation was shown to restore normal sleep architecture, possibly via the BDNF‐TrkB signaling pathway (Duan et al. 2025).

Several studies have reported a significant reduction in Ruminococcaceae in patients with insomnia disorder (Benedict et al. 2016; Zhou et al. 2022), which was positively correlated with anti‐inflammatory IL‐10 and negatively correlated with PSQI and ISI scores (Zeng et al. 2024). A large cohort study found that chronic insomnia was associated with gut microbiota variations, identifying Ruminococcaceae UCG‐002 and UCG‐003 as potential genera inversely associated with chronic insomnia (Jiang et al. 2022). In our study, Ruminococcaceae was positively correlated with SE, while it was negatively correlated with WASO, AI, AIREM, etc., supporting its potential beneficial role in promoting sleep. Additionally, Lachnospira has been reported to inversely correlate with poor self‐reported sleep quality in acute insomnia patients (Li et al. 2020). Our study found significantly lower Lachnospira abundance in the M‐CID group compared to the HC group, with a significant negative correlation with ISI. Oscillospira, another butyrate‐producing bacterium, has been suggested as a beneficial microbe (Yang et al. 2021). Based on the Guangdong Gut Microbiome Project, a study demonstrated positive correlations between the abundance of Oscillospira and both microbial diversity and sleep duration (Chen et al. 2020), which were consistent with our findings. However, the association of Phascolarctobacterium with insomnia has been minimally studied, with one report showing its increased abundance following Traditional Chinese Medicine (TCM) treatment, inversely correlating with PSQI and ISI (Zeng et al. 2024). This contrasts with our findings, possibly due to the TCM‐specific patient inclusion criteria.

As for the Bacteroidetes phylum, Bacteroides has been identified as a key biomarker for recognizing CID (Liu et al. 2019), which was significantly elevated in CID patients and showed a positive correlation with PSQI (Liu et al. 2019). These findings were consistent with our current results. However, another study reported a decrease in the abundance of Bacteroides and an increase in the abundance of Clostridium in patients with insomnia disorder (Zhou et al. 2022), which contrasts with our findings. A potential reason for this discrepancy is that the study included both patients with chronic and acute insomnia disorder. In the core taxa identified in our study, Parabacteroides included only one species—Parabacteroides distasonis. Previous studies have shown that Parabacteroides distasonis was associated with a positive health status (Koh et al. 2020; Maltz et al. 2019). Parabacteroides distasonis may play an important role in mediating the sleep‐promoting effects of a prebiotic diet (Bowers et al. 2022). Our study found that the relative abundance of Parabacteroides in the M‐CID group was higher than in the S‐CID group, and it showed a significant positive correlation with TST, suggesting that Parabacteroides may be a potentially beneficial bacterium. In general, we observed that CID patients with poorer sleep quality had reduced beneficial bacteria and increased harmful bacteria, while those with milder insomnia symptoms showed minimal differences compared to the healthy individuals. Our analysis further confirmed significant associations between key bacterial taxa and both subjective and objective sleep parameters. However, it is important to note that while PSG provides objective sleep data, it is not optimal for CID diagnosis or severity assessment. Additionally, the lack of Multiple Sleep Latency Test (MSLT) data precluded an objective evaluation of daytime physiological sleepiness and hyperarousal. Future research should integrate MSLT‐derived measures (e.g., mean sleep latency) with subjective assessments (e.g., PSQI). This integration could facilitate subgroup analyses (e.g., comparing patients with longer vs. shorter mean sleep latency) and enable multidimensional severity stratification, thereby refining CID subtyping and enhancing our understanding of its associations with the gut microbiota.

We applied KEGG‐based PICRUSt analysis to identify the potential functional changes in the microbiota. Our results revealed that patients with poorer sleep quality exhibited more DEGs compared to healthy subjects than those with milder symptoms. Functional annotation based on DEGs revealed that the major biological functions of the enriched pathways included amino acid metabolism, carbohydrate metabolism, and SCFA metabolism, with the S‐CID group showing more pronounced enrichment. These findings suggest that alterations in amino acid, carbohydrate, and short‐chain fatty acid metabolism may occur in the gut microbiota of CID patients. Amino acid metabolism is crucial for the central nervous system, as amino acids are important sources of brain neurotransmitters and regulate neurotransmitter activity (Dalangin et al. 2020). Glutamate is the major excitatory neurotransmitter in the human brain (Weigend et al. 2019), and its signaling has a significant connection with sleep (Weigend et al. 2019), as it can be further converted into GABA (Pasanta et al. 2023), which plays a role in regulating sleep/wake cycles and promoting sleep onset (Gottesmann 2002). Although exogenous GABA is usually considered incapable of crossing the blood‐brain barrier, it may indirectly influence the nervous system through its effects on the gut (Yu et al. 2020). In the present study, we found that the alanine, aspartate, and glutamate metabolism pathway was significantly enriched in both the S‐CID and M‐CID groups. Further GSEA analysis revealed that this pathway was notably downregulated in the S‐CID group compared to the M‐CID group, indicating that it may play an important role in the regulation of sleep quality in CID. Tryptophan is the precursor to serotonin, which participates in the central nervous system's regulation of sleep/wake cycles (Brown et al. 2012; Oikonomou et al. 2019; Saper et al. 2010). Existing evidence suggests that changes in the gut microbiota's tryptophan metabolism can affect the availability of peripheral tryptophan, influencing central tryptophan levels and subsequently altering central serotonin metabolism (Gao et al. 2018, Gao et al. 2019; Lukić et al. 2019). Furthermore, CID may be associated with branched‐chain amino acid (BCAA) dysregulation, as insomnia patients exhibited elevated energy metabolism products and decreased BCAA breakdown metabolites (Gehrman et al. 2018). In our current study, we observed significant enrichment of the phenylalanine, tyrosine, and tryptophan biosynthesis pathway in both the S‐CID and M‐CID groups, while changes in the valine, leucine, and isoleucine degradation pathway were noted in the M‐CID group.

Our study also identified significant enrichment in two SCFAs metabolism pathways—butanoate metabolism and propanoate metabolism—in both the S‐CID and M‐CID groups. Notably, the S‐CID group showed a significant upregulation of the butanoate metabolism pathway compared to the M‐CID group. These pathway enrichments suggest potential alterations in the levels of the corresponding metabolites, butyrate and propionate. As previously discussed, SCFAs, particularly butyrate, have beneficial effects on insomnia disorder. Furthermore, propionate may also benefit sleep, as evidenced by an association between increased propionate concentration in human infant fecal samples and prolonged sleep duration (Heath et al. 2020). Importantly, butyrate‐producing bacteria described in the human gut are typically found in the Firmicutes phylum and the Clostridiales order, predominantly belonging to the families Clostridiaceae, Eubacteriaceae, Lachnospiraceae, and Ruminococcaceae (Fu et al. 2019). Consistent with this, most of the key taxa identified in the present study belong to Clostridiaceae, Lachnospiraceae, and Ruminococcaceae families. Furthermore, these key taxa showed significant correlations with multiple objective sleep parameters, especially REM sleep indicators; for instance, Clostridium and Ruminococcaceae exhibited significant negative correlations with AIREM. These findings collectively suggest functional alterations in the gut microbiota across pathological stages of CID. However, given the inherent limitations of PICRUSt2 in predicting functional potential rather than actual metabolic activity, future research should directly quantify relevant metabolites using metabolomics for validation. Targeted metabolomics, focusing on short‐chain fatty acid quantification, is planned to validate these key pathway predictions. Furthermore, integrating metagenomics for strain‐level identification and transcriptomics for gene expression dynamics will be crucial to elucidate the mechanistic links between microbial function and CID.

The random forest algorithm—a machine learning method particularly suited for high‐dimensional biological data—was employed to explore microbial signatures associated with insomnia severity. While prior studies have reported microbiota‐based classifiers for insomnia detection (Liu et al. 2019), our work extends this approach by demonstrating severity‐stratified discriminative capacity. We found that key microbial taxa showed moderate performance as potential indicators for assessing insomnia severity, and the inclusion of the F/B ratio slightly improved the model's predictive accuracy. It should be noted that, while random forest reduces overfitting risk through mechanisms such as bagging and random feature selection, this concern remains non‐negligible—particularly given the high‐dimensional yet limited sample size typical of microbiome data. The moderate AUC values observed in our model may reflect room for improvement in generalizability. To directly address overfitting concerns and enhance broader applicability, external validation in an independent cohort is necessary in future studies.

Furthermore, key confounders that may simultaneously influence both gut microbiota composition and insomnia severity—such as dietary patterns, chronic stress exposure, and circadian rhythm disruptions—were not systematically assessed in this study. Future longitudinal designs should incorporate standardized measurements of these variables (e.g., validated dietary logs, actigraphy‐derived circadian metrics, and stress biomarkers), along with covariate adjustment strategies to better isolate microbiota‐specific effects. To simultaneously address overfitting and enhance external validity, external validation in an independent cohort with matched confounder assessment is therefore warranted. Nonetheless, our findings suggest a possible biological basis for the classification of CID and support further investigation into targeted, microbiota‐based precision interventions.

In our study, we found that CID patients with more severe poor sleep quality exhibited disruptions in the gut microbiota, including alterations in microbial diversity, composition, and function. These disruptions were less pronounced in patients with only mild sleep quality issues. Furthermore, specific gut microbiota characteristics were able to effectively distinguish patients with poorer sleep quality. Overall, our findings provide evidence supporting the potential for using distinct gut microbiota features to identify more severe sleep quality problems. Future interventions targeting the modulation of gut microbiota composition and metabolic pathways may offer a promising approach to managing CID.

Our study has several limitations. First, cross‐sectional design precludes causal inference regarding gut microbiota differences across groups; longitudinal studies are needed to clarify temporal dynamics. Second, 16S rRNA gene sequencing lacks strain‐level resolution; future research should integrate metagenomics, transcriptomics, and metabolomics to gain deeper functional insights. Third, although key bacteria were linked to CID severity, their functional roles require experimental validation. Fourth, the generalizability of our findings is limited to Han Chinese populations, and although the proportion of females over 50 years did not differ significantly across groups, menopausal status—a potential confounder—was not assessed. Future studies should validate these findings in ethnically diverse cohorts and explicitly evaluate menopausal status. Fifth, the absence of standardized gastrointestinal measures may have missed subclinical disturbances; future studies should incorporate these assessments.

Yaxi Liu, Yixian Cai, Xian Shi contributed equally to this work. Yaxi Liu: conceptualization, data curation, formal analysis, methodology, supervision, validation, visualization, writing – original draft, writing – review and editing. Yixian Cai: conceptualization, data curation, formal analysis, investigation, methodology, supervision, validation, writing – original draft, writing – review and editing. Xian Shi: conceptualization, formal analysis, methodology, validation, writing – original draft, writing – review and editing. Mei Fan: data curation and investigation. Xiaotao Zhang: investigation and methodology. Jingjing Lin: writing – review and editing. Xiaoxuan Fan: writing – review and editing. Bingdong Liu: conceptualization, methodology, project administration, resources, software, supervision, validation, visualization, writing – review and editing. Jiyang Pan: conceptualization, funding acquisition, project administration, resources, supervision, writing – review and editing.

This work was supported by the National Key R&D Program of China (Grant No. 2022YFC2503902) to Jiyang Pan.

This study was approved by the Ethics Committee of the first Affiliated Hospital of Jinan University (No. KY‐2022‐167).

Informed consent was obtained from all participants.

The authors declare no conflict of interest.

Further information and requests for the data should be directed to and will be fulfilled by the lead contact, Dr. Jiyang Pan (). Jiypan@163.com

This study did not generate reagents.