Correlations of Gut Microbiome, Serum Metabolome and Immune Factors in Insomnia

Insomnia is a common sleep disorder of unclear etiology characterized by individuals experiencing the inability to sleep or the diculty staying asleep. Gut-brain interaction is being explored with the intent of discerning gut microbiota and its role in many brain-related conditions including insomnia. The number and diversity of gut microbiota that colonize the digestive tract could have a signicant association with insomnia, given the microbes that colonize the digestive tract integrate with and impact on the central nervous system, the immune system and multiple metabolic pathways. We aim to examine the diversity and to explore the functional impact of gut microbiota in insomniacs by examining fecal microbiome using 16S rRNA gene sequencing, serum metabolome using ultra-high-performance liquid chromatography coupled to tandem mass spectrometry (UHPLC-MS/MS), and various immune factors. Our results discover altered and distinct gut microbiota in insomniacs, with enriched Desulfovibrio, Lactobacillus and Streptococcus, and decreased abundance of Bidobacterium, Gardnerella, Sneathia, Aerococcus and Atopobium. Non-targeted metabolomics identify 31 aberrant metabolites and implicate metabolic pathways in insomniacs. Most importantly, correlations across gut microbiome, serum metabolome and inammatory factors are unraveled. Our study provides a better understanding of gut microbiota’s role in insomnia and new insights into potential novel etiologies for insomnia.


Introduction
Insomnia is a widespread human problem and a disorder where patients have di culties initiating or maintaining sleep and typically experience non-restorative sleep for at least three nights per week over a 3-month period 1 . It is a multifactorial condition that appears to have association, comorbidity, or cooccurrence with various diseases including diabetes, cardiovascular conditions, and obesity 2 . Although the pathogenesis of insomnia is still unclear, it appears that behavioral, cognitive and emotional factors can also play a role in the occurrence and development of insomnia. Current neurobiological and psychological viewpoints indicate that changes in brain function and genetics can also play a role.
Whereas the magnitude and pathophysiological impact of insomnia disorder has been investigated in the central nervous system and immune system 3,4 , less attention has been paid to the potential interaction effect between the insomnia condition and the potential microbial dysbiosis of gut microbiota. The gut microbiota is involved in the regulation of the phenotypic characteristics of the host 5 , such as, immunity, metabolism 6 , and circadian clock function 7 . Moreover, many complex disorders are now considered to be related to microbiota where previously these disorders were attributed to lifestyle factors [8][9][10] . Sleep disorder could also impact the metabolism of multiple substances in the human body. We hypothesized that disrupted patterns of sleep may be related to the metabolite abnormality via the changes in population of gut microbial communities. To further explore and illustrate our hypothesis, we analyzed the intestinal microbiome and metabolome of the insomniacs. We posited the non-targeted metabolomics analysis could potentially disclose the aberrant metabolites and metabolic pathways in insomniacs. The combined analysis of both metabolic groups and intestinal microbial distribution will explain some of the relationships between the gut microbiota and host metabolic phenotypes. We decided it is important to assess the gut microbiota related metabolic phenotype alterations in order to further understand the potential pathophysiological mechanism of insomnia and explore the role of gut microbiota in the development of insomnia.
In this study, the fecal and serum samples were collected from the insomniac patients and the control participants of good sleep history to explore and document the potential role of gut microbiota and the underlying metabolic mechanisms in insomnia. With this purpose in mind, we performed 16S rRNA gene sequencing for fecal and UHPLC-MS/MS analysis for serum. Meanwhile, the immune factors IL-1β, IL-6 and TNF-α were also measured using Quantitative Enzyme-linked Immunosorbent Assay (ELISA). The serum level of TNF-α and IL-1β were observed to be signi cantly different in the insomniacs. The serum metabolic parameters were evaluated by the UHPLC-MS/MS analysis and 31 potential biomarkers were screened out from the patient observations. Our results provided new insights and additional data points for the exploration of new potential mechanisms and origins of insomnia.

Study design and participants
This study was approved by the Institutional Review Boards of the Weihai Central Hospital (IRB# WCH2017-1201). Informed consent was obtained from all participants. Procedures were carried out in this study to conform with the ethical standards stipulated by the institutional and national research committee, as well as those stipulated in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.
Our study recruited 40 insomnia patients and 40 healthy controls of good sleep history ( Table 1). The Inclusive criteria were as follows: 1) Participants met criteria for primary insomnia de ned according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) that included: a. unsatis ed with the time or the quality of sleep, including di culty in falling asleep, prolonged sleep and early awakening. b. suffering from sleep disorders that cause serious daytime dysfunction or damage (such as emotional or cognitive disorders, work dysfunction). c. insomnia attacks at least three nights a week and sleep disorders occurring for at least 3 months. d. the presence of sleep disorders occurring even though there are adequate sleep opportunities. 2) The healthy volunteers who were recruited from among the people also receiving a physical examination at the same time. These volunteers served as control group participants. 3) Participant age was limited to those 18 years old or above. The exclusion criteria were established as follows: 1) Secondary insomnia caused by somatic or psychiatric diseases. 2) Suffering from diabetes, infectious diseases, gastrointestinal diseases and other diseases. 3) Smokers and drinkers. 4) Those who cannot communicate and cooperate normally. 5) Those with a history of radiotherapy and chemotherapy. 6) Pregnant women. 7) Anyone that had taken antibiotics in the previous three months. Sample collection and preparation The feces samples were collected by the fecal collectors that were pre-loaded with 600-800 mL absolute ethanol 11 . The samples were immediately frozen in a -80 ℃ refrigerator. Three samples were collected from each participant. One was used for the extraction of DNA; the other two samples were reserved for the further studies. All the venous blood samples were extracted between 8:30-10:00 AM using vacutainer tubes without anticoagulants (serum), and then the samples were strati ed at room temperature for 1 h.
The supernatant of the blood was saved after 3,000 rpm centrifugation at room temperature for 1 minute. The serum samples were collected from the supernatant after 12,000 rpm centrifugation at 4 ℃ for 10 minute and frozen at -80 ℃ for use.
Microbial community analysis based on 16S rRNA gene sequencing The total DNA of the gut microbiota was extracted from the fecal sample by utilizing the modi ed CTAB methods 12 . For the purpose of analyzing the taxonomic composition of the bacterial community, the V1-V2 region of the 16S rDNA gene was selected for the subsequent pyrosequencing. The speci c primer pair 27F (5'-AGAGTTTGATCMTGGCTCAG-3') and 355R (5'-GCTGCCTCCCGTAGGAGT-3') were used for the PCR ampli cation. The 8 bp Barcode sequence and 5-8 random sequences were introduced into the 5' end of the primers to distinguish samples, remove chimeric sequences and PCR redundancy. The ampli cation of each sample was performed in triplicate, then products were puri ed and quanti ed, and sequenced by Illumina HiSeq 2500 sequencing platform.
The reads were merged by FLASH 13 , the sample split and quality control were executed by the scripts of QIIME (version 1.9.1) 14 . The chimeras were removed by using the de novo-based method of Usearch61 15 . An open Operational Taxonomic Units (OTU) clustering strategy was adopted. The Greengenes database (version 13.8) was served as the reference (the similarity was set to 97%), the unaligned sequences were clustered by UCLUST. The OTU is clustered by de novo clustering strategy, and the RDP were used for the species annotation 16 . And the OTUs which abundance less than 0.005% of the total number were ltered. Then, the OTU table was rare ed to a sequencing depth of 10,000 per sample for subsequent analyses of alpha and beta diversity.

Serum metabolites analysis based on UHPLC-MS/MS
The 100 μL serum sample was thawed on the ice, and then 300 μL methanol (containing the internal standard, L-2-chlorophenylalanine 1 ug/mL) was added in to the tube, mixed by a vortex mixer for 30 s; after the ultrasonic treatment in the ice-water bath for 10 min, the protein was precipitated at -20℃ for 1 h; the samples were centrifuged at 4℃ for 15 min at 12,000 rpm. For each sample, 2 μL of the supernatant was used for the UHPLC-MS/MS analysis.
We performed chromatographic analysis on

Detection of the serum immune factors
The levels of the immune factors were measured by human IL-1β/IL-6/TNF-α ELISA kit (ABclonal Technology) according to the procedures supplied by the manufacturer. The absorbance of each sample was measured by Multiplate Reader Ascent (Thermo) at 450 nm and 630 nm, and the amount of proteins in pg/mL calculated on the basis of a standard curve. Each serum was tested in triplicate.

Data analysis and statistical methods
ProteoWizard software was used to convert the original mass spectrometry data into mzML format. The retention time, identify peaks, extract peaks, integrate peaks and align peaks were corrected by XCMS (version 3.2). Material identi cation was carried out by using OSI-SMMS (version 1.0) software and selfbuilt database. The Log conversion and Par formatting of data were carried out by using SIMCA software (version 14.1). The normalized data were then to perform the principal component analysis (PCA) and orthogonal projections to latent structures-discriminant analysis (OPLS-DA). The differential metabolites were screened by the variable importance in the project (VIP) in the OPLS-DA model 17 . The screened metabolites were then mapped to the KEGG, PubChem and HMDB databases. Metabo Analyst software (http://www.metaboanalyst.ca/) was used for metabolic pathway analysis 18 .
Data were presented as mean ± SD. Statistical comparisons and were analyzed by Mann-Whitney U test using the SPSS 22.0 (Chicago, USA), Chi-squared-tests were used to assess differences across genders.
P< 0.05 was considered statistically signi cant. The linear discriminant analysis effect size (LEfSe) was performed on the website (http://huttenhower.sph.harvard.edu/galaxy/), using default parameters, Alpha values of Kruskal-Wallis (KW) and rank test and Alpha values of Wilcoxon rank sum test were set to < 0.05, and LDA values were set to > 2.0.

Distinct gut microbiome observed in insomniac
The fecal microbiota composition pro les were analyzed by the 16S rRNA gene sequencing. A total of 2,480,664 high quality sequence reads were obtained and 31,008.3± 4,463.188 sequences for each sample. There were some differences in the composition of each sample at the phylum and genus levels.
At the phylum level ( Figure 1A), the relative abundance of Bacteroidetes in the gut microbiota of insomniacs and healthy controls was 54.60%±16.65% and 57.93%±20.21%, respectively; followed by Firmicutes and Proteobacteria; Firmicutes accounted for 33.64%±12.93% in insomniacs and 32.02% ±18.05% in healthy controls, while Proteobacteria accounted for 10.62%±12.14% and 8.43%±9.98% in insomnia patients and healthy controls, respectively. At the genus level ( Figure 1B), the relative abundance of Bacteroides in the gut microbiota of insomniacs and healthy controls was 31.60%±21.32% and 33.43%±22.76% respectively, followed by Prevotella, the relative abundance of Bacteroides was 19.51%±23.77% and 21.79% ±26.89% in the gut microbiota of the patients and controls, respectively. Venn diagram chart showed the core shared OTU (at least 50% of the samples in each group contain the OTU) in insomniacs and healthy controls. There were 692 core OTUs in healthy controls, 681 core OTUs in insomniacs and 611 total OTUs were founded in both two groups. However, there was no signi cant difference in alpha diversity between the two groups (P > 0.05) ( Figure S1). The principal coordinate analysis (PCoA) was used to compare bacterial community patterns between the two groups. The results ( Figure 1D) showed that there was a signi cant difference in community patterns between insomniacs and healthy controls based on the unweighted UniFrac distance (Adonis, R 2 =0.061, P =0.0001).
Linear discriminant effect size analysis (LEfSe) was performed with the purpose of screening potential gut microbiota biomarkers related to insomnia. A LDA histogram was used to represent the predominant bacteria and the structure of the microbiota in insomniac participants and healthy control participants ( Figure 2A). We screened a total of 29 biomarkers, including 2 bacterial phyla, 5 classes, 5 orders, 9 families and 8 genera, by using a Linear Discriminant Analysis (LDA) value greater than 2 as the threshold ( Table 2). These results are also illustrated in the cladogram ( Figure 2B). As shown in Figure 2B, the abundance of Desulfovibrio, Lactobacillus and Streptococcus were signi cantly enriched in the gut microbiome of insomniac participants, whereas Bi dobacterium, Gardnerella, Sneathia, Aerococcus and Atopobium were more abundant in the healthy control participants. Variation of serum in ammatory factors in insomniacs During sleep, some unique substances are produced by the human body. We wanted to study these to understand difference between observations in the control group observations obtained from the insomniac group. Certain in ammatory factors have also been shown to regulate and affect the sleep in humans 19 . However, in ammation and infection could also alter sleep architecture, whereas a lack of sleep could also impair immune function. Therefore, we measured in ammatory factors in the serum such as interleukin-1β (IL-1β), tumor necrosis factor-α (TNF-α), and interleukin 6 (IL-6) and these measurements were obtained by using ELISA. As Figure 3 indicates, the IL-1β was signi cantly elevated (2.315±2.202 pg/ml in insomniac group participants VS 0.967±0.745 pg/ml in control group participants, P 0.0001), meanwhile TNF-α was signi cantly decreased in the insomniac participants (2.055±1.619 pg/ml in insomniacs VS 3.234±1.520 pg/ml in controls, P 0.0001). However, the IL-6 level was consistent between the patients and healthy controls (1.842±1.396 pg/ml in insomniac vs 1.805±1.541 pg/ml in controls).

Correlation between serum in ammatory factors and gut microbiota
To reveal the potential correlation between the enterobacteria and the in ammatory factors in the insomniacs, the Pearson correlation analysis was used ( Figure 4). We found that 4 OTUs of Bacteroides, 2 OTUs of Bacteroides fragilis, 1 OTU of Oscillospira has a signi cant positive correlation with IL-1β. And IL-6 positively correlated with 5 OTUs of Prevotella copri, 2 OTUs Faecalibacterium prausnitzii, 2 OTUs of Bacteroides, and 1 OTU of Anaerostipes. At the same time, TNF-α revealed a signi cant positive correlation with the OTU of Prevotella copri, Parabacteroides, Oscillospira, Butyricimonas and Bacteroides.

Signi cant alteration of metabolomic pro les in insomniacs
Metabolic pro ling of the insomniacs and healthy controls were acquired by Ultra performance liquid chromatography-tandem mass spectrometer (UHPLC-MS/MS). 5831 peaks were detected, and 5736 peaks remained after using a relative standard deviation de-noising method in positive ion mode (ES+).
And in negative ion mode (ES-), 2612 peaks were detected, and 2588 peaks remained after de-noising. To discriminate the metabolic pro les between insomniacs and healthy control groups, we used principal component analysis (PCA) and orthogonal partial least-squares discriminant analysis (OPLS-DA). The results indicated a large separation of metabolome between insomniacs and healthy controls in PCA plots ( Figure 5A, B) both in ES+ (R 2 X=0.32) and ES-(R 2 X=0.317). Furthermore, individuals in insomniac group were separated from and healthy control group as further evidenced by the OPLS-DA score scatter plots both in ES+ (R 2 X=0.169, R 2 Y=0.905, Q 2 =0.707) and ES-(R 2 X=0.175, R 2 Y=0.84, Q 2 =0.532). We performed a permutation test in order to further validate the OPLS-DA model. After 200 permutations, the R 2 intercept was 0.71 and 0.78 in ES+ and ES-respectively, the Q 2 intercept values were -0.87 and -0.73 in ES+ and ES-respectively ( Figure S2A, B). In order to screen for differential metabolites, the rst principal component of variable importance in the projection (VIP) was obtained ( Figure S2C, D). The VIP values exceeding 1 were rst selected as differential metabolites, and the Student's t-test also be used. Then we found 25 metabolites in ES+ and 6 metabolites in ES-(VIP >1 P<0.05, Table S2). We also visualized it with a heatmap ( Figure 5E). We found that, in ES+, 1-palmitoylglycerophosphocholine was identi ed signi cantly increased, however, 7-hydroxy-6-methoxy-alpha-pyrufuran, acinospesigenin A, PS(22:2(13Z,16Z)/20:2(11Z,14Z)) and PA(17:1(9Z)/13:0) were signi cantly decreased in the insomniacs' serum. Meanwhile, aspartic acid, phenylalanine and phosphatidylcholine lyso 20:4 were identi ed to be signi cantly decreased in the insomniacs' serum when under ES-. All differential metabolites were then subjected to the regulatory pathways analysis to discover the metabolic pathways exhibiting high correlations with the metabolites. According to P value and in uence value, signi cant abnormalities were found in ve metabolic pathways in insomniacs: glycerophospholipid metabolism (P <0.05), glutathione metabolism (P <0.05), nitrogen metabolism (P <0.01), alanine, aspartate and glutamate metabolism (P <0.05), and aminoacyl-tRNA biosynthesis (P < 0.05) ( Figure 5F, G).

Correlation between serum metabolome and gut microbiota
To explore the functional relationship of the altered gut microbiota and metabolites in insomniacs, the correlation matrixes based on Pearson's rank correlation coe cient were formulated ( Figure 6). Correlations between the varied metabolites and the gut microbiota were identi ed.

Discussion
Sleep is a complex rhythmic physiological process, considerable evidence showing that the gut microbiome not only affects the immune functions of the host but also regulates sleep through the microbiome-gut-brain axis. Early research shown that gut microbiota exhibits diurnal oscillations in composition and function, it also can control metabolic homeostasis 20 . At present, research on the effects of the gut microbiome on the pathogenesis of insomnia is incomplete. Liu et al. demonstrated that the composition, diversity and metabolic function of the gut microbiota are signi cantly changed between insomnia patients and healthy controls 21 . But the assessment of metabolic function is only based on PICRUSt function prediction. Smith et al. also con rmed the association between gut microbiota diversity and sleep physiology 22 . In addition, they found that positive correlations between total microbiome diversity and IL-6. To our knowledge, our study is the rst to comprehensively examine the relationship of gut microbiota and immunity with insomnia. It may lead to a better understanding of the bidirectional communication between the host and the gut microbiome and may produce novel strategies for insomnia treatment and intervention.
The shift in populations of gut microbiota has been associated with various diseases, in this study, we demonstrated that α-and β-diversity of the gut microbiota in insomnia patients is signi cantly altered, which is similar to the result of Liu et al. 21 . We found primarily the beta diversity of the gut microbiota in insomniacs is different, alpha diversity rarely varied. What's more, we found Desulfovibrio, Lactobacillus and Streptococcus were observed as signi cantly enriched in insomnia group, while Bi dobacterium, Sneathia and Atopobium were signi cantly decreased. The Streptococcus spp. are generally regarded as the producers of D-and L-lactic acid 23 . The observation of enriched levels of Streptococcus will improve the lactic acid content, further result in a lower the colonic pH 24 , modify fecal microbiota metabolism, and alter intestinal epithelial barrier function, increasing passive intestinal permeability 25 . The high rates of colonization with Gram-positive fecal Streptococcus was observed to be related to the poor sleep.
Meanwhile poorer mood was associated with higher Lactobacillus 26 . The lower Lactobacillus level was associated with better subjective sleep quality and vigor ratings in the CFSs who were responded to the antibiotic treatment. The results suggested that the emotional responses and behavior of the human would be in uenced by the disordered commensal intestinal organisms 27 . In addition, high abundance of Lactobacillus and Streptococcus may be more detrimental to the host 26 . Desulfovibrio are the predominant sulfate-reducing bacterial in the human intestinal. Desulfovibrio generate hydrogen sul de which has a wide range of cytotoxic effects, using the sul tes and sulfates as substrates. Hydrogen sul de is an effective inhibitor of the oxidation of short-chain fatty acids (SCFAs) in cells, thus completing a vicious cycle of mutually exclusive metabolic interactions between host and bacteria 28 . The de ciency in SCFAs β-oxidation will affects the sleep quality 29 . Therefore, may be Desulfovibrio was associated to the insomnia be involved in the metabolism of SCFAs.
The level of two important immune factors IL-1β and TNF-α in the serum were observed signi cantly changed in the insomniacs, although they showed opposite trends of change. The two factors could act on the brain, and cause decreased appetite, sleep disturbance, and depression 30 . In addition, the factors might also be involved into the regulation of central receptors or activated the hypothalamic-pituitaryadrenal (HPA) axis lead to the release of stress hormones with similar peripheral effects 31 . IL-1β and TNF-α are involved in the regulation of slow wave sleep (SWS) 32,33 . IL-1β can regulate the release of NE, 5-HT and DA in hypothalamus, and participate in sleep regulation by promoting slow wave sleep 34 . In the previous studies 35,36 , IL-1β was increased across sleep deprivation young man. In addition, prolonged wakefulness will also up-regulated IL-1β in the brain 33 . We found IL-1β was also signi cantly increased in the insomniacs. The previous evidence suggested that TNF-α played an important role in sleep regulation, the sleep fragmentation would be increase TNF-α level and may lead to sleepiness [37][38][39] . The TNF-α was signi cantly decreased in the insomniacs, meanwhile exogenous injection of TNF-α into animals could induce sleepiness and elicits excess sleep 33 . The results suggested additional TNF-α supplementation may help sleep in insomniacs. Accumulating studies indicated the gut microbiota affects not only gastrointestinal function but also immune system by regulating microbiota-gut-brain axis or HPA axis 40,41 . In some serious cases, the sleep disorder-related conditions may lead to serious neurasthenia and depression [42][43][44] . Zhu et al. 45 suggest that TNF-α is involved in the regulation of the uptake of 5hydroxytryptamine (5-HT). By activating mitogen-activated protein kinase, TNF-α catalyze a series of cascade reactions, which ultimately increase the re-uptake of 5-HT and reduce the concentration of 5-HT, while the decrease of 5-HT concentration will eventually induce anxiety and depression. In the present study, we also found that, the changes of immune factors were correlated with the gut microbiota.
Although IL-6 did not change signi cantly in insomnia patients, we still found a correlation between IL-6 and gut bacteria, which is different The metabolomics study suggested, the multiple serum metabolites were different in insomnia group.
Many studies have shown that lipids in the blood exhibit circadian rhythm changes 47,48 . Our research also con rms these changes. In the serum of patients with insomnia, 1palmitoylglycerophosphatecholine, phosphatidylcholine lyso 20: 4 were signi cantly decreased, while phosphatidylserine and glycerophosphatidic acid were signi cantly increased. We also found that phosphatidylcholine lyso 20:4 was positively correlated with Bacteroides. A study on difference of gut microbiota in primary insomnia with different traditional Chinese medicine found that Bacteroides were enriched in re due to yin de ciency syndrome group 49 . There is a signi cant correlation between Bacteroides and many metabolites, which indicates that Bacteroides may be a very important factor affecting sleep. Threonic acid is an ascorbate catabolic product associated with oxidative stress associated with dehydroascorbic acid. In this study, it was found that threonic acid decreased in serum in patients with insomnia, and that the antioxidant capacity of sleep patients was decreased. The level of phenylalanine in the serum of patients with sleep disorders is signi cantly reduced. Phenylalanine can be used as a precursor of L-tryptophan. Tryptophan is a very important metabolite in brain-gut axis, and many intestinal bacteria can Affecting the metabolism of tryptophan, tryptophan can reduce the time to fall asleep, so it is often used as a hypnotic drug 50 . Phosphatidylserine (PS) plays an important role in repairing brain damage and improving cognitive ability, improving Alzheimer's disease in the elderly, alleviating tension and depression, physical fatigue and anti-depression 51 . The decreased PS level may negatively affect the central nervous system of the insomniacs. Interestingly, differential bacteria in gut of insomnia patients and healthy controls are correlated with many metabolites. Streptococcus positive correlation with phospholipid metabolites (PA, PS, PC). Atopobium were negatively correlated with short chain polypeptides. This indicates that the reduction of these metabolites in patients with insomnia is related to these two bacteria, and the speci c correlation and mechanism need to be further studied.

Conclusion
In summary, our study found differences in gut microbiota, serum metabolites, and serum immune factors between insomniacs and healthy controls, and shown a novel association between them.
Moreover, we identi ed several speci c taxa and metabolites, they may be potential biomarkers related to insomnia. Of course, the mechanism of its occurrence would be studies by more cases to provide evidence in our future work. This will help us develop microbiota-based disease diagnosis, prevention and therapeutic tools.

Declarations
Ethics approval and consent to participate This study was approved by the Institutional Review Boards of the Weihai Central Hospital (IRB# WCH2017-1201). Informed consent was obtained from all participants.

Consent for publication
Not applicable.

Availability of data and material
The datasets during and/or analysed during the current study available from the corresponding author on reasonable request.

Competing interests
All authors declare that there is no con ict of interest regarding the publication of this paper.