The Component and Functional Pathways of Gut Microbiota Are Altered in Populations with Poor Sleep Quality – A Preliminary Report

A sleep disorder (SD) is characterized by difficulty in falling asleep, lack of sleep, excessive sleep, and poor sleep quality, which can affect the personal physical status and inherent mental health. Overwhelming evidence indicates that sleep plays an irreplaceable role in protecting human health, eliminating fatigue, restoring body function, enhancing immunity, protecting brain health, and others (Irwin 2015). However, previous studies show that inadequate sleep was found in 40 ~ 65% of American students (Carney et al. 2006; Lund et al. 2010; Kenney et al. 2012; Becker et al. 2014). Correspondingly, according to a recent study investigating 11,954 students from 50 universities in China, relatively severe sleep disorders were found among students attending high-level universities (Yang et al. 2018). These studies also suggested that academic and job strain and adverse lifestyle habits may be the main reason for sleep problems. One study of 1,125 participants from four countries provided novel evidence for the deterioration of problematic sleep among adolescents that at least one insomnia symptom was reported by 32.7% of adolescent workers and 35.2% of non-employed adolescents. Additionally, this study also indicated that insomnia symptoms were persistent in individuals with anxiety (73.2 ~ 76.5%) or depression (67.6 ~ 77.4%) disorders (Ohayon et al. 2000), which suggested that comorbidity led to the development and worsening of sleep disorder to a certain extent. Given the high prevalence of sleep problems in adolescents and the expanded influence on health outcomes, it has become a severe public health concern, which is needed to be solved urgently.

Recent studies have generally discovered the linkages between gut microbiota and sleep disturbance with the development of genome sequencing techniques. There is currently some evidence in the mice model that sleep disturbance can alter the composition of the gut microbiome (Poroyko et al. 2016). Subsequently, some findings that fecal microbiota transplantation could result in sleep disturbances supported the bidirectional linkages between the two (Badran et al. 2020). Additionally, a significant change in some bacterial taxa was reported by recent studies related to sleep, such as an increase in the Firmicutes to Bacteroidetes ratio (Bowers et al. 2020) and an alteration of the relative abundance of the Clostridiaceae and Lachnospiraceae (El Aidy et al. 2020) in sleep-disrupted mice. It might highlight possible directions of future research in this field. Although most mice experiments can consistently reveal the association between sleep problems and gut microbiota, some studies in humans have reported contradicting conclusions (Benedict et al. 2016; Zhang et al. 2017), suggesting that previous works needed to be repeated to verify the question. Another concern is that antibiotics and dietary prebiotics have been utilized to alter the gut microbiota and, correspondingly that sleep quality was influenced by the changes in the gut bacteria (Nakakita et al. 2016; Thompson et al. 2017). In short, the findings of previous studies, both reduction in nocturnal sleep (Lendrum et al. 2017) and metabolic disturbance of amino acids and vitamins related to neurotransmission (Ogawa et al. 2020), have been strengthened the association between gut microbiota and sleep disturbance by using antibiotics to deplete the gut microbiota of mice. The research on dietary prebiotics used to improve sleep is in its infancy, and there are still many issues to be elucidated, including safety, efficacy, and others. Even though the causal associations between the gut microbiome and sleep problems remain undetermined, the above findings indicated that gut microbiota is essential for improving sleep quality.

To this end, we conducted a study in which the differences in the diversity and composition of the gut microbiota between the individuals with poor subjective sleep quality and normal were analyzed. We tried to explore possible alteration of functional pathways and identify the significant gut bacteria for further studies.

The Basic Characteristics of the SD and HC Groups. Demographic characteristics of the SD and HC groups are presented in Table I. The PSQI score of the two groups were 5.118 ± 1.576 (HC) and 9.059 ± 1.144 (SD), with a significant difference (t = 8.342, p-value < 0.001). However, there was no significant difference in gender, age, and BMI. Both groups had similar demographic profiles.

The Quality of Fecal Samples. Rarefaction curve, Shannon-Wiener curve, rank-abundance curve, and species accumulation curve were utilized to assess the quality of the fecal samples, demonstrating that the fecal samples collected from all participants meet the requirements for experimental design and subsequent analysis (see supplementary materials).

Alpha diversity analysis. The six parameters above were compared based on the Wilcoxon rank-sum test of both groups to identify differences in alpha diversity between the SD and HC groups. Neither the species richness (observed species, Chao1 and ACE) nor diversity (Shannon index, Simpson index, and Coverage index) found significant differences between the two groups (all p-value > 0.05, two-side; Fig. 1).

Fig. 1

Beta diversity analysis. We calculated unweighted UniFrac, weighted UniFrac, Jaccard, and Bray-Curtis distances of the microbial composition to determine differences in the bacterial community structure among samples from two groups in this study. ADONIS results revealed no significant differences between Jaccard (p-value = 0.1838, Fig. 2A) and Bray-Curtis distances (p-value = 0.5401, Fig. 2B). Therefore, we utilized the PLS-DA, a supervision pattern recognition method, to assess the subtle significant differences between the two groups, and the result demonstrated that the two groups were separated obviously, as can be seen from Fig. 3.

Fig. 2

Fig. 3

Differential microbiota analysis. The linear discriminant analysis (LDA) and effect size (LEfSe) analysis, based on the nonparametric factorial Kruskal-Wallis (KW) sum rank test and Wilcoxon rank-sum test, were utilized to examine bacterial taxa with significant differential abundances between the SD and HC groups (LDA score > 2.0, p-value < 0.05, two-side). The result of LEfSe analysis showed 24 taxa with significantly different abundance, mainly focused on phyla Firmicutes and Proteobacteria, and the top ten species with the smallest p-value were shown in Fig. 4. The bacterial species with significant differences in relative abundance at each taxonomic level between the two groups were defined by the Metastats analysis under p-value < 0.05. There were eight taxa with significantly different abundance between the two groups after Bonferroni-Holm (BH) correction (q-value < 0.05). Additionally, the result of the Spearman correlation indicated that only three taxa had a significant positive correlation with PSQI score, respectively, and the correlation coefficient for genus Holdemanella reached its maximum value (r = 0.601, p-value < 0.01) (Table II). The actual correlations between them and the SD score were shown with scatter plots in Fig. 5.

Fig. 4

Fig. 5

Functional prediction of gut microbiota. As our results indicated a shift in the gut microbiome composition, we further examined whether it might relate to potential metabolic differences between the SD and HC groups, as previous studies have shown. The functional metagenome (PICRUSt) was utilized for characterizing the functions of the gut microbiota. Subsequently, Welch’s t-test was used to identify statistically significant differences between the two groups (Fig. 6). The results suggested that compared to the HC group, metabolic pathways involved in polycyclic aromatic hydrocarbons pathways degradation (p-value = 0.005, two-side) were significantly increased in the SD group. In contrast, metabolic pathways involved in styrene degradation (p-value = 0.009, two-side), butanoate metabolism (p-value = 0.018, two-side), C5‑branched dibasic acid metabolism (p-value = 0.023, two-side), propanoate metabolism (p-value = 0.043, two-side) and beta-alanine metabolism (p-value = 0.044, two-side) were significantly decreased in the SD group.

Fig. 6

Given that previous studies used mice models to verify the association between sleep disturbance and the gut microbiota and their interactions, we carried out the current study to determine whether poor sleep quality is associated with the gut microbiota in humans and, if so, whether more biomarkers can be determined for further study in the area. In the current study, a better balance of baseline characteristics between the SD and HC groups was reached after the screening of participants.

Although previous studies have suggested significant differences in the alpha diversity of gut microbiota between the sleep disorders and healthy populations, we failed to find any statistically significant differences in several parameters used to assess the alpha diversity of the gut microbiota. Moreover, the current results are consistent with those of Zhang et al. (2017), who showed that short-term sleep restriction did not significantly alter the OTU abundance, β diversity, and population shift of the gut microbiota in humans. In addition, the latest evidence has indicated that short-term sleep restriction and circadian misalignment do not appear to impair the stability of the alpha diversity of the human microbiome (Withrow et al. 2021). On the other hand, some researchers studied these relationships from a new perspective. They observed the changes in gut microbiota following short-term (two weeks) sleep extension in patients with chronic sleep deprivation, but no significant changes were found in this study (Reutrakul et al. 2020). Notably, the participants self-reported their current sleep conditions in our study, which may lead to the overestimated duration of SD in short-term partial subjects. However, recent results have shown a significant decrease in the abundance and alpha diversity of the gut microbiota in sleep disorders compared to healthy individuals (Liu et al. 2019). When there was disagreement between our results and some findings in previous studies, it is important to note the limitations of the present study and why some results should be interpreted with caution.

However, our study indeed found subtle significant differences in the structure of the gut microbiota in people with poor sleep quality compared to the sleep normal humans and finally identified eight taxa (q-value < 0.05), according to the result of Metastats analysis, which may be used to discriminate these two groups. Finding that sleep efficiency was positively correlated with the abundance of phylum Tenericutes has been mentioned in numerous previous studies (Benedict et al 2016; Reutrakul et al 2020; Withrow et al. 2020). Notably, phylum Tenericutes and class Mollicutes have been demonstrated to be resilient in response to acute sleep-wake cycle shifts (Liu et al. 2020). We also found that the relative abundance of the phylum Tenericutes (significantly, q-value < 0.05) and class Mollicutes (differences approaching significance, q-value = 0.069) in subjects with poor sleep quality were lower than in the healthy individuals. However, it remains to be determined whether this phenomenon represents that the taxa have the buffering capacity to some extent to external factors that may impair the host’s sleep. In addition, we also observed a higher abundance of the family Erysipelotrichaceae (phylum Firmicutes) in participants with poor sleep quality (no significant difference after correction), which is consistent with the previous results (Benedict et al 2016). However, the latest evidence provided by Lin et al. (2021) indicated that Lactobacillus fermentum PS150 (PS150) could improve sleep conditions in the mice model and also induces gut microbiota remodeling, including an increased abundance of class Erysipelotrichia. Our findings do not confirm these results. One possible explanation for these discrepancies is species differences between humans and mice.

The PICRUSt prediction of the gut bacterial functions showed that several metabolic pathways differed significantly from healthy controls compared with individuals with poor sleep quality. However, some of them are challenging to interpret by current and previous evidence, and their clinical significance may be questionable, for instance, polycyclic aromatic hydrocarbon degradation. However, the predicted functional compositions of butanoate metabolism, C5‑branched dibasic acid metabolism, and propanoate metabolism pathways were significantly down-regulated in the SD group. It is worth mentioning that sleep fragmentation could lead to a decreased abundance of putative butyrate-producing bacteria and further affect the butanoate metabolism (Maki et al. 2020). As the members of specific short-chain fatty acids (SCFAs) (Magzal et al. 2021), butanoate has also been demonstrated to play a role in inducing sleep in mice models (Szentirmai et al. 2019). Additionally, butanoate has some functions similar to SCFAs, such as energy metabolism and inflammation of the host (Hamer et al. 2008), and pieces of previous evidence have suggested that changes in microbiota composition induced by the circadian rhythm misalignment might have implications for inflammatory diseases (Voigt et al. 2014; Parkar et al. 2019; Mashaqi and Gozal 2020). C5‑Branched dibasic acid metabolism and propanoate metabolism in patients with obstructive sleep apnea-hypopnea syndrome has also been found with significant differences compared to healthy humans (Li et al. 2017; Ko et al. 2019). Correspondingly, the increased concentration of propanoate in human infant fecal samples was associated with the prolongation of sleep duration (Heath et al. 2020). Significant differences in the gut microbiota composition and function were observed in our study, which further demonstrated the relationship between sleep problems and the gut microbiota in humans, but the causality remains to be fully explored.

This study’s findings must be considered along with limitations. Initially, the observational characteristic of the cross-sectional study cannot provide enough evidence for supporting the potential causal links between poor sleep quality and changes in the gut microbiota. Secondly, further study on larger samples is also needed to identify the other key bacteria, reported in a previous study but failed to be found in this study. Moreover, in studies with small sample size, the change of statistical methods may need to be considered. As Weiss et al. (2017) mentioned, DESeq2 could increase sensitivity on smaller datasets (< 20 samples per group). In addition, the lack of multiple objective measures of sleep (e.g., electroencephalogram, polysomnography) could reduce the accuracy and timeliness of the data and induce recall bias due to the self-reported approach.

On the other hand, the gut microbiota composition may be influenced by diet. So, in future research, the dietary survey of the participants should be added and included as a control variable for reducing the influence on gut microbiota. Finally, as a preliminary exploration, we failed to find robust evidence from the current research to establish the association between the bacteria that changed and the metabolic pathways identified. Therefore, future studies are warranted to further examine the causal relationship between sleep and gut microbiota by considering more other factors.

In conclusion, our study demonstrated an association between the composition of the gut microbiome and poor sleep quality. Moreover, we identified several specific phyla and taxa (e.g., Elusimicrobia and Tenericutes) that may be used to distinguish sleep-disordered patients from healthy individuals. Our results also revealed the difference in functional pathways, such as butanoate metabolism and propanoate metabolism, which raise the possibility that the gut microbiota could be a potential contributor to diagnosing and treating sleep problems in the future.