Abstract
Introduction: An increasing body of evidence suggests a strong relationship between gut health and mental state. Lately, a connection between butyrate-producing bacteria and sleep quality has been discussed. The PROVIT study, as a randomized, double-blind, 4-week, multispecies probiotic intervention study, aims at elucidating the potential interconnection between the gut’s metabolome and the molecular clock in individuals with major depressive disorder (MDD). Methods: The aim of the PROVIT-CLOCK study was to analyze changes in core clock gene expression during treatment with probiotic intervention versus placebo in fasting blood and the connection with the serum- and stool-metabolome in patients with MDD (n = 53). In addition to clinical assessments in the PROVIT study, metabolomics analyses with 1H nuclear magnetic resonance spectroscopy (stool and serum) and gene expression (RT-qPCR) analysis of the core clock genes ARNTL, PER3, CLOCK, TIMELESS, NR1D1 in peripheral blood mononuclear cells of fasting blood were performed. Results: The gene expression levels of the clock gene CLOCK were significantly altered only in individuals receiving probiotic add-on treatment. TIMELESS and ARNTL gene expression changed significantly over the 4-week intervention period in both groups. Various positive and negative correlations between metabolites in serum/stool and core clock gene expression levels were observed. Conclusion: Changing the gut microbiome by probiotic treatment potentially influences CLOCK gene expression. The preliminary results of the PROVIT-CLOCK study indicate a possible interconnection between the gut microbiome and circadian rhythm potentially orchestrated by metabolites.
Introduction
Major depressive disorder (MDD) is a highly prevalent psychiatric disorder. Especially during the COVID-19 pandemic, the numbers increased dramatically with an estimation of 53.2 million additional MDD cases worldwide, which underlines the urgent necessity to explore new add-on therapies to treat the widespread MDD [1].
Even though MDD is a common, devastating disorder, the underlying pathological mechanisms are still not completely uncovered. It is common scientific knowledge that a polygenic orchestra builds up the genetic heritability, including hundreds of gene variants, leading to an individual’s genetic predisposition for MDD [2]. In addition to the genetic predisposition, a concatenation of chronic and acute psychosocial stress, gene-environment interactions (e.g., the microbiome in the gut and the circadian rhythms) are influencing the complex pathophysiology of MDD [3]. Recently, the gut-brain axis and their impact on disturbed circadian rhythm in individuals with MDD got in the spotlight of attention [4‒6]. Today we know that psychiatric disorders are associated with alterations in the gut microbiome composition. Individuals with MDD exhibit a depletion of butyrate-producing bacteria such as Faecalibacterium, Coprococcus spp., as well as Dialister and an increased abundance of proinflammatory bacteria like Eggerthella [7, 8].
Nevertheless, how do gut bacteria influence depression? The gut bacteria can affect the human body and brain via several pathways. For example, bacteria can release small metabolites, which facilitate bidirectional communication between the gastrointestinal tract and the central nervous system (CNS) via the gut-brain axis. Of importance is the production of short-chain fatty acids by fermenting and metabolizing dietary fiber by gut microbiota, which can have far-reaching effects [9, 10]. Short-chain fatty acids (e.g., butyrate) can pass the blood-brain barrier and can even affect our 24-h clock and sleep [11, 12]. Supplementation of butyrate in animal models demonstrated increases in non-rapid eye movement, sleep duration, and promoting sleep [13]. Gut microbial metabolites influence central and hepatic clock gene expression, sleep duration, and body composition [14]. Disturbed sleep has been clinically associated with depression for decades and recent studies underline this concatenation [15].
Hypothesis-driven gene association and gene expression studies already discovered, in relatively small cohorts, nominal associations between clock gene variants and mood disorders (MDD and bipolar affective disorder) [16‒25]. The genome-wide association study (GWAS) era brought new insights and provided evidence for the association between disturbed circadian rhythms based on a polygenic heritability and mood instability on a genetic level [26]. This is not surprising because the molecular clock regulates mood by activating the transcription of monoamine oxidase A (MAO-A). More precisely, the clock gene ARNTL encodes for a transcription factor, which activates together with the gene product of NPAS2 as a heterodimer, the transcription of the neurotransmitter degrading enzyme MAO-A [27, 28]. The molecular clock and mood show therefore a strong concatenation. Nevertheless, the circadian clock in the suprachiasmatic nucleus of the hypothalamus in the CNS also plays a crucial role in this equation by managing various endocrinological and vegetative vital pathways (cortisol and melatonin release, body temperature, heartbeat and metabolism) as it activates over 200 clock gene controlled target genes [11]. As previously mentioned, there is a strong cross-link between the gut-brain axis, circadian rhythm, and mood, which needs to be further elucidated [11].
The PROVIT study, as a randomized double-blind placebo-controlled probiotic intervention study, aimed at exploring the effects of a 4-week probiotic intake in inpatients with MDD on a clinical and molecular level. In addition, the PROVIT study is one of the first worldwide studies that analyzed microbiome, metabolome, gene expression, and cognition after 28 days of placebo-controlled, double blind, multispecies probiotic intake. One prior finding of the PROVIT study was a shift in the stools’ metabolome of individuals with MDD receiving probiotic add-on treatment for 28 days toward higher normalized concentrations of butyrate [29]. As butyrate is suspected to influence the circadian rhythm, the current PROVIT-CLOCK investigation was designed as a further step to gain knowledge about the interplay between gut-microbiota metabolites and the molecular clock [11, 12].
More precisely, PROVIT-CLOCK aimed to analyze the core clock gene expression (ARNTL, PER3, CLOCK, TIMELESS, NR1D1) in peripheral blood mononuclear cells (PBMCs) of fasting blood and the metabolome (NMR) in serum and stool samples of inpatients with MDD. In detail, the primary aim of the current investigation of the PROVIT-CLOCK study was to analyze if the clock gene expression in PBMCs is altered in the group of individuals with MDD receiving probiotic add-on treatment. Second, we aimed to study a group*time interaction of clock gene expression after 4 weeks of probiotic intervention in inpatients with MDD. Third, we targeted to examine a time effect on the molecular clock. Fourth, we aimed to explore whether metabolites in stool and serum affect the clock gene expression in PBMCs of study participants with MDD to investigate the interconnection between the metabolome and the molecular clock.
Materials and Methods
Study Design of the PROVIT Trial
The detailed PROVIT study protocol was already published by Kreuzer et al. [30], Reininghaus et al. [31], and Reiter et al. [32]. In short, the PROVIT study was designed to investigate the influence of multispecies probiotic add-on treatment in individuals with MDD, in a double-blind, randomized, placebo-controlled study setting (the timeline of the PROVIT project is depicted in Fig. 1). Parameters included microbiome, metabolome, routine blood work markers and targeted gene expression, as well as scores for psychiatric symptomatology and cognition. All patients were recruited at the Department for Psychiatry and Psychotherapeutic Medicine at the Medical University of Graz in Austria. Patients were recruited via our acute outpatient clinic or referred by a psychiatrist for in-hospital psychiatric treatment.
Inclusion criteria demanded a current diagnosis of a depressive episode and age between 18 and 75 years. A psychiatrist according to the International Statistical Classification of Diseases and Related Health Problems (ICD-10) guidelines and Mini International Neuropsychiatric Interview (MINI) confirmed the MDD diagnosis of all individuals recruited. All participants were aged at least 18 years, legally insightful, and able to consent. All study participants provided written and informed consent. The probiotics study was registered at clinicaltrials.com (NCT03300440) and the local Ethics Committee gave their approval (EK 29-235 ex 16/17). In addition to the inpatient standard care, all patients with MDD received the placebo or probiotic drink (OMNi-BiOTiC Stress Repair with the bacterial strains: bifidobacteria (B. bifidum W23, B. lactis W51, B. lactis W52) and lactobacilli [L. acidophilus W22, L. casei W56, L. paracasei W20, L. plantarum W62, L. salivarius W24, L. lactis W19 provided by AllergoSan]). Both groups received biotin additionally in the study medication, due to ethical considerations. All study participants were required to receive a substance that might be beneficial for them. Therefore, both formulas consisted of 125 mg D-biotin (vitamin B7), 30 mg of common horsetail, 30 mg of fish collagen, and 30 mg of keratin in a matrix with maize starch, maltodextrin, inulin, potassium chloride, magnesium sulfate, fructooligosaccharides, enzymes (amylases), and manganese sulfate. The physician on call provided and supervised the intake of the drink every morning at 7:00–7:30 am for 28 days in a double-blind manner. As the ethical board required all patients to receive a beneficial substance, vitamin B7 (biotin) was added to the probiotics and placebo formula. Fasting blood was collected at the beginning of the trial (t0) and at the end of the intervention after 4 weeks (t2). At the beginning (t0), after 2 weeks (t1), and at the end of the intervention (t2), stool was collected. Demographic parameters (age, weight [kg], height [m], body mass index, sex, and standard medication), data on cognitive testing, and lifestyle questionnaires were obtained at the intervention start and the end. A psychiatrist according to the International Statistical Classification of Diseases and Related Health Problems (ICD-10) guidelines and Mini International Neuropsychiatric Interview (MINI) confirmed the MDD diagnosis of all individuals recruited. The depression severity was rated with the Beck Depression Inventory-II (BDI-II) and the Hamilton Rating Scale for Depression (HAMD). Further details on exclusion and inclusion criteria and details on the PROVIT study can be obtained by the former publications [29, 30, 35].
Randomization and Double-Blind Design
All participants and the study team were blinded until the end of the study. The placebo drink had the same texture, color, and taste as the probiotic drink. The doctors on call, who provided the drink, did not know whether they handed out a placebo or probiotic drink. The platform www.randomization.com was used to randomize 96 subjects into 24 blocks of 4 to assign the individuals either to the intervention or placebo group. The manufacturer assembled the test product packages. For each study participant, one package, with the relevant participant number, was provided. The person in charge of the randomization, as well as all unblinded employees of the manufacturer, was bound to secrecy toward thirds. The ratio of the probiotic group:placebo group was 1:1.
Details of the Probiotic Add-On Treatment
The provider “Institute AllergoSan” with their producer Winclove BV in Amsterdam provided the probiotic and placebo additives. The probiotic group received the commercially available multistrain probiotic “OMNi-BiOTiC® Stress Repair” supplement, which includes nine bacterial strains with ≥2.5 × 109 colony-forming units per gram. A probiotic drink consisted of at least 3 g, with a total number of 7.5 × 109 colony-forming units per bag. More precisely, the probiotic drink included bifidobacteria (B. bifidum W23, B. lactis W51, B. lactis W52) and lactobacilli (L. acidophilus W22, L. casei W56, L. paracasei W20, L. plantarum W62, L. salivarius W24, L. lactis W19). Additional ingredients included in the probiotic and placebo supplement were D-biotin (vitamin B7), common horsetail, fish collagen, and keratin. The matrix of the supplement further compromised maize starch, maltodextrin, inulin, potassium chloride, magnesium sulfate, fructooligosaccharides, enzymes (amylases), and manganese sulfate. The placebo drink had the exact same looking color, consistency, and taste texture, containing the matrix and D-biotin (vitamin B7). The detailed study procedure and ingredients of the add-on treatment were already described before [29, 30, 35].
Gene Expression Analysis of Core Clock Genes Using qPCR
The gene expression analysis of the target clock genes in this current PROVIT investigation was conducted in the same way as previously published by Kreuzer et al. [30]. At the beginning and the end of the 28-day PROVIT intervention, fasting blood was collected in PAXgene blood RNA tubes (Nr 762165, PreAnalytix GmbH, Hombrechtikon, Switzerland) and stored in a freezer at −80°C. In the following step, RNA was isolated from PBMCs in the PAXgene blood RNA tubes according to the manufacturer’s instruction with PAXgene Blood RNA Kit (PreAnalytix GmbH, Hombrechtikon, Switzerland). The purity and quantity of the RNA were determined by using NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Reverse transcription of the RNA to cDNA was performed with the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, USA) containing RNAse inhibitor (Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s instructions. Quantitative polymerase chain reaction (qPCR) was performed on the QuantStudioTM 7 Flex Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) using the Luna Universal qPCR Master Mix (New England Biolabs, Ipswich, MA, USA). Per reaction, 2.5 ng of the cDNA and 200 nm of the primers were used. For the gene expression analysis of the targeted clock genes ARNTL, PER3, CLOCK, TIMELESS, NR1D1, QuantiTect Primer Assay (QT00000721, Qiagen, Hilden, Germany) were used. The cycling conditions included the initial denaturation at 50°C for 2 min and 95°C for 10 min, followed by 45 cycles of denaturation at 95°C for 15 s and an extension period at 60°C for 60 s. The sample’s melting curves were measured to confirm the specificity of amplified products. In general, all samples were measured as technical triplicates. For determining the transcript abundance, we used the 2−ΔΔCt method, normalized to the Ct values of housekeeping genes GAPDH and TBP. The gene expression levels of the housekeeping genes were stable over the time course of intervention and between the verum and the placebo group. Detailed information about primer sequences (online suppl. Table 1; for all online suppl. material, see https://doi.org/10.1159/000538781), gene expression levels of housekeeping genes (online suppl. Fig. 1–2), and melting curves (online suppl. Fig. 3) are given in the supplementary material. The gene expression values represent the x-fold change (FC).
Metabolomics
Reagents
From VWR International (Darmstadt, Germany), dibasic sodium phosphate (Na2HPO4), sodium hydroxide, hydrochloric acid (32% m/v), and sodium azide (NaN3), were purchased. 3(trimethylsilyl) propionic acid-2,2,3,3-d4 sodium salt (TSP) was obtained from Alfa Aesar (Karlsruhe, Germany). Deuterium oxide (D2O) was ordered from Cambridge Isotopes Laboratories (Tewksbury, MA, USA). The in-house Milli-Q Advantage Water Purification System from Millipore (Schwalbach, Germany) was used to purify deionized water. All other chemicals were used for no further purification. The phosphate NMR buffer solution was prepared by dissolving 5.56 g of anhydrous Na2HPO4, 0.4 g of TSP, and 0.2 g NaN3, in 400 mL of D2O and adjusted to pH 7.4 with 1 M NaOH and HCl. Upon addition of D2O to a final volume of 500 mL, the pH was re-adjusted to pH 7.4 with 1 M NaOH and HCl.
Metabolite Quantification Using NMR
The detailed NMR protocol was already published by Kreuzer et al. [30]. In summary, in the first step proteins were removed and the ongoing enzymatic reactions were quenched by adding a methanol-water solution in a 2:1 ratio to stool and serum samples (200 μL serum plus 400 μL methanol; 1 g stool plus 2 mL methanol). In the following step, the Precellys homogenizer was used to lyse the samples. Until further processing, the samples were stored at −20°C. Then, the samples were spun at 17,949 rcf at 4°C for 30 min. Before transferring the samples to 5 mm NMR tubes, the supernatants were lyophilized, and 500 μL of NMR buffer in D2O was added to the samples. The NMR measurements were conducted at 310 K on an AVANCE™ Neo Bruker Ultrashield 600 MHz spectrometer, equipped with a TXI probe head and processed as described previously [31]. Specifically, for the 1H 1D NMR experiments, the 1D CPMG (Carr-Purcell_Meiboom_Gill) pulse sequence (cpmgpr1d, 128 scans, 73,728 points in F1, 11,904.76 HZ spectral width, recycle delays 4 s) with water suppression by pre-saturation was used. For data acquisition, the Bruker Topspin version 4.0.2 was used. The spectra of all samples were automatically processed (exponential line broadening of 0.3 Hz), phased, and referenced using TSP at 0.0 ppm, using the Bruker Topspin 4.0.2 software (Bruker GmbH, Rheinstetten, Germany). Next, the spectra were imported to Matlab2014b; the regions around the water, TSP and remaining methanol signals were excluded. For correcting the sample metabolite dilution, a probabilistic quotient normalization was performed [32‒34]. Data analysis and preparation of figures was carried out in MetaboAnalyst 5.0 (vide infra).
Statistical Analysis
Descriptive Statistics of the PROVIT Gene Expression Sample
According to per-protocol analysis, only participants providing a full set of serum samples (t0 and t2) were included in the PROVIT-CLOCK sample. First, the normal distribution of the variables was measured with the Kolmogorov-Smirnov test, Shapiro-Wilk test, and scale level. Depending on the normal distribution, baseline differences between the groups were analyzed with a two-tailed Students’ t test, Mann-Whitney U test, or χ2 test. Descriptive data were assessed by using means, mean rank, standard deviations and percentages of the respective variables. The description of the sample concerning clinical and sociodemographic data is shown in Table 1. Table 2 shows the medication. The level of significance was set at p < 0.05. All analyses were carried out with SPSS version 27 (SPSS Inc., Chicago, IL, USA). Visualization was performed with SPSS version 27, MetaboAnalyst 5.0, and Figure 1 with the online tool Visme.
Description . | Intervention group (n = 24), n (%) . | Placebo group (n = 29), n (%) . | Statistics . | |
---|---|---|---|---|
. | . | χ2 . | Sig (p value) . | |
Sex (female) | 18 (75.0) | 23 (79.3) | 0.139 | 0.709 |
Smoking (yes) | 8 (33.3) | 16 (55.2) | 2.528 | 0.112 |
Description . | Intervention group (n = 24), n (%) . | Placebo group (n = 29), n (%) . | Statistics . | |
---|---|---|---|---|
. | . | χ2 . | Sig (p value) . | |
Sex (female) | 18 (75.0) | 23 (79.3) | 0.139 | 0.709 |
Smoking (yes) | 8 (33.3) | 16 (55.2) | 2.528 | 0.112 |
. | Mean (SD) . | Mean (SD) . | t . | p value . |
---|---|---|---|---|
Age, years | 44.15 (14.48) | 40.36 (10.87) | −1.088 | 0.162 |
Waist-to-hip ratio | 0.85 | 0.86 | −0.160 | 0.257 |
. | Mean (SD) . | Mean (SD) . | t . | p value . |
---|---|---|---|---|
Age, years | 44.15 (14.48) | 40.36 (10.87) | −1.088 | 0.162 |
Waist-to-hip ratio | 0.85 | 0.86 | −0.160 | 0.257 |
. | Median (mean rank) . | Median (mean rank) . | U . | p value . |
---|---|---|---|---|
Education, years | 9.00 (26.33) | 9.00 (27.55) | 332 | 0.753 |
Illness duration, years | 6.00 (25.32) | 10.00 (26.52) | 304 | 0.775 |
BMI, kg/m2 | 23.86 (27.83) | 27.18 (26.31) | 328 | 0.721 |
. | Median (mean rank) . | Median (mean rank) . | U . | p value . |
---|---|---|---|---|
Education, years | 9.00 (26.33) | 9.00 (27.55) | 332 | 0.753 |
Illness duration, years | 6.00 (25.32) | 10.00 (26.52) | 304 | 0.775 |
BMI, kg/m2 | 23.86 (27.83) | 27.18 (26.31) | 328 | 0.721 |
MDD, major depressive disorder; SD, standard deviation; BMI, body mass index.
. | Intervention group (N = 24), n (%) . | Placebo group (N = 29), n (%) . | Statistics . | |
---|---|---|---|---|
. | . | . | χ2 . | p value . |
Anticonvulsants | 3 (12.5) | 2 (6.9) | 0.483 | 0.487 |
Benzodiazepines/hypnotics | 4 (16.7) | 5 (17.2) | 0.003 | 0.956 |
Atypical antipsychotics | 8 (33.3) | 9 (31.0) | 0.032 | 0.858 |
Other antidepressants (SNRIs, etc.) | 17 (70.8) | 21 (72.4) | 0.016 | 0.899 |
Selective serotonin reuptake inhibitors (SSRIs) | 8 (33.3) | 13 (44.8) | 0.725 | 0.394 |
. | Intervention group (N = 24), n (%) . | Placebo group (N = 29), n (%) . | Statistics . | |
---|---|---|---|---|
. | . | . | χ2 . | p value . |
Anticonvulsants | 3 (12.5) | 2 (6.9) | 0.483 | 0.487 |
Benzodiazepines/hypnotics | 4 (16.7) | 5 (17.2) | 0.003 | 0.956 |
Atypical antipsychotics | 8 (33.3) | 9 (31.0) | 0.032 | 0.858 |
Other antidepressants (SNRIs, etc.) | 17 (70.8) | 21 (72.4) | 0.016 | 0.899 |
Selective serotonin reuptake inhibitors (SSRIs) | 8 (33.3) | 13 (44.8) | 0.725 | 0.394 |
N, number of study participants with MDD; MDD, major depressive disorder; SNRIs, serotonin norepinephrine reuptake inhibitors; SSRIs, selective serotonin reuptake inhibitors.
Depression Scores and Gene Expression Analyses
The mean Ct value of GAPDH and TBP was used for normalization of transcript abundance. The expression values represent the x-FC. Group (probiotics vs. placebo), time effects (before vs. after the 4-week PROVIT intervention) and group-time interaction of clock gene expression and depression scores HAMD and BDI-II were analyzed with two repeated-measures two-way analysis of variances with the independent factors time point (t0 vs. t2) and group (intervention vs. placebo). The dependent variables in the model were the FC values of ARNTL, PER3, CLOCK, TIMELESS, and NR1D1. Sphericity was checked and the homogeneity of the sample was measured with the Levene test. The Greenhouse-Geisser adjustment was used to correct for violations of sphericity for the variables TIMELESS and ARNTL. Diagrams were illustrated with SPSS 27 (SPSS Inc., Chicago, IL, USA).
Correlation Analyses between Serum Metabolome and Clock Gene Expression
Based on the PROVIT-metabolomics data, published by Kreuzer et al. [30], the association between serum metabolites and clock gene expression (ARNTL, PER3, CLOCK, TIMELESS, NR1D1) was measured in the total sample at time point t2 using Pearson correlation coefficient.
Results
Description of the Patient Population of the PROVIT-CLOCK Cohort
Forty-two patients with MDD (32 women and 10 men) were allocated to the intervention group. Forty patients with MDD (32 women and 8 men) were allocated to the placebo group. In the intervention group, 5 women and 1 man were lost to follow-up and 7 women and 1 man discontinued intervention because of non-compliance or antibiotic intake. In the placebo group, three women and two men were lost to follow-up and two women discontinued intervention. The consort flowchart was published by Reininghaus et al. [31]. In the PROVIT-CLOCK sample, 4 patients in the probiotics group and 4 patients in the placebo group were excluded from the whole PROVIT sample because of lacking serum samples or not passing the quality control of the processing steps during gene expression analysis. Study participants with lacking data were completely excluded from analysis. There was no significant difference in sex, smoking, age, waist-to-hip ratio, education in years, illness duration, and body mass index between participants subjected to the intervention (n = 24) or placebo group (n = 29) (see Table 1).
The psychopharmacological medication at admission is presented in Table 2. No significant differences between the intervention and the placebo group, in assessed depression scores HAMD and BDI, occurred in the PROVIT-CLOCK cohort. Further details on the psychopharmacological medication were already published in previous PROVIT publications [29, 30, 35].
All Individuals of PROVIT-CLOCK Cohort Improved in Depression Scores
The mean values, standard deviation, and results of repeated-measures two-way analysis of variances are depicted in Table 3. Both groups significantly improved over the time course of the intervention period of 28 days (BDI: F(1, 48) = 91.71, p < 0.001, η2 = 0.656; HAMD: F(1,51) = 61.85, p < 0.001, η2 = 0.548). There was no significant difference in the change of depression scores between the placebo or probiotic group. No significant time*group interaction occurred (see Table 3). The results did not change when adjusted for smoking.
. | Intervention group (N = 24) . | Placebo group (N = 29) . | Time . | Group . | Interaction . | |||||
---|---|---|---|---|---|---|---|---|---|---|
. | mean . | SD . | mean . | SD . | F . | p value . | F . | p value . | F . | p value . |
BDI-II t0 | 30.25 | 8.38 | 31.19 | 10.89 | 91.71 | <0.001 | 0.636 | 0.429 | 0.351 | 0.556 |
BDI-II t2 | 14.71 | 8.28 | 17.46 | 11.09 | ||||||
HAMD t0 | 14.75 | 5.56 | 14.55 | 4.7 | 61.85 | <0.001 | 0.192 | 0.663 | 0.155 | 0.695 |
HAMD t2 | 8.79 | 4.97 | 7.97 | 5.25 |
. | Intervention group (N = 24) . | Placebo group (N = 29) . | Time . | Group . | Interaction . | |||||
---|---|---|---|---|---|---|---|---|---|---|
. | mean . | SD . | mean . | SD . | F . | p value . | F . | p value . | F . | p value . |
BDI-II t0 | 30.25 | 8.38 | 31.19 | 10.89 | 91.71 | <0.001 | 0.636 | 0.429 | 0.351 | 0.556 |
BDI-II t2 | 14.71 | 8.28 | 17.46 | 11.09 | ||||||
HAMD t0 | 14.75 | 5.56 | 14.55 | 4.7 | 61.85 | <0.001 | 0.192 | 0.663 | 0.155 | 0.695 |
HAMD t2 | 8.79 | 4.97 | 7.97 | 5.25 |
N, number of study participants with MDD; MDD, major depressive disorder; SD, standard deviation.
CLOCK Gene Expression Is Altered in Individuals Receiving Probiotic Treatment
The gene expression level of CLOCK (F(1, 48) = 981.96, p = < 0.018, η2 = 0.953) was significantly upregulated in the intervention group after 28 days (see Fig. 2). ARNTL (F(1, 51) = 4.579, p = 0.037, η2 = 0.082; Fig. 3) and TIMELESS (F(1, 47) = 5.558, p = 0.023, η2 = 0.106; see Fig. 4) were significantly upregulated over the course of the intervention period in both groups (see Table 4). The gene expression of ARNTL did not differ significantly at the beginning of the study. The results did not change when adjusted for smoking.
. | Intervention group (N = 24), mean (SD) . | Placebo group (N = 29), mean (SD) . | Time . | Group . | Interaction . | |||
---|---|---|---|---|---|---|---|---|
. | . | . | F . | p value . | F . | p value . | F . | p value . |
ARNTL t0 | 0.95 (0.20) | 1.01 (0.18) | 4.58 | 0.037 | 0.048 | 0.827 | 3.523 | 0.065 |
ARNTL t2 | 1.06 (0.22) | 1.02 (0.23) | ||||||
CLOCK t0 | 0.84 (0.24) | 1.02 (0.22) | 1.00 | 0.321 | 5.97 | 0.018 | 1.04 | 0.314 |
CLOCK t2 | 0.92 (0.27) | 1.02 (0.27) | ||||||
PER3 t0 | 0.96 (0.20) | 1.02 (0.23) | 0.230 | 0.634 | 0.027 | 0.870 | 2.49 | 0.121 |
PER3 t2 | 1.02 (0.28) | 0.98 (0.22) | ||||||
TIMELESS t0 | 0.95 (019) | 0.98 (0.30) | 5.56 | 0.023 | 0.001 | 0.973 | 0.813 | 0.372 |
TIMELESS t2 | 1.07 (0.30) | 1.04 (0.34) | ||||||
NR1D1 t0 | 1.07 (0.29) | 1.03 (0.34) | 3.74 | 0.059 | 0.209 | 0.650 | 3.073 | 0.086 |
NR1D1 t2 | 0.91 (0.25) | 1.02 (0.32) |
. | Intervention group (N = 24), mean (SD) . | Placebo group (N = 29), mean (SD) . | Time . | Group . | Interaction . | |||
---|---|---|---|---|---|---|---|---|
. | . | . | F . | p value . | F . | p value . | F . | p value . |
ARNTL t0 | 0.95 (0.20) | 1.01 (0.18) | 4.58 | 0.037 | 0.048 | 0.827 | 3.523 | 0.065 |
ARNTL t2 | 1.06 (0.22) | 1.02 (0.23) | ||||||
CLOCK t0 | 0.84 (0.24) | 1.02 (0.22) | 1.00 | 0.321 | 5.97 | 0.018 | 1.04 | 0.314 |
CLOCK t2 | 0.92 (0.27) | 1.02 (0.27) | ||||||
PER3 t0 | 0.96 (0.20) | 1.02 (0.23) | 0.230 | 0.634 | 0.027 | 0.870 | 2.49 | 0.121 |
PER3 t2 | 1.02 (0.28) | 0.98 (0.22) | ||||||
TIMELESS t0 | 0.95 (019) | 0.98 (0.30) | 5.56 | 0.023 | 0.001 | 0.973 | 0.813 | 0.372 |
TIMELESS t2 | 1.07 (0.30) | 1.04 (0.34) | ||||||
NR1D1 t0 | 1.07 (0.29) | 1.03 (0.34) | 3.74 | 0.059 | 0.209 | 0.650 | 3.073 | 0.086 |
NR1D1 t2 | 0.91 (0.25) | 1.02 (0.32) |
Mean values, SD, and results of the repeated-measures ANOVA.
ARNTL, aryl hydrocarbon receptor nuclear translocator-like protein; CLOCK, circadian locomotor output cycles kaput protein; PER3, period circadian protein homolog 3; NR1D1, nuclear receptor subfamily 1, group D member 1.
Correlation between Metabolome and Clock Gene Expression
The comprehensive results of the correlation analyses between serum and stool metabolites (49 and 50 metabolites, respectively, were analyzed) and clock gene expression markers are depicted in Figures 5-7. ARNTL gene expression levels in serum positively correlated with lysine (r = 0.28; praw = 0.0048), valine (r = 0.24; praw 0.019), isoleucine (r = 0.24; praw 0.021), and alanine (r = 0.23; praw 0.023) in stool and with betaine (r = 0.28; praw 0.0046) and glucose (r = 0.27; praw 0.074) in serum. CLOCK gene expression levels in serum significantly correlated with formic acid (r = −0.27; praw 0.0090), fumaric acid (r = 0.26; praw 0.012), and glycerol (r = −0.23; praw 0.028) in stool. TIMELESS and NR1D1 gene expression levels in serum positively correlated with propylene glycol (r = 0.28; praw 0.0064 and r = 0.27; praw 0.011, respectively) in stool. NR1D1 gene expression levels in serum positively correlated with valeric acid (r = 0.26; praw 0.013) in stool and glycerophosphocholine (r = 0.37; praw 0.00017) and carnitine (r = 0.36; praw 0.00032) in serum. PER3 gene expression levels in serum positively correlated with butyric acid (r = 0.21; praw 0.042) in stool and levels of tryptophan (r = 0.44; praw 5 × 10−6), trimethylamine-N-oxide (r = 0.30; 0.0026), leucine (r = 0.27; praw 0.0082), and arginine (r = 0.28; praw 0.0058) in serum.
Discussion
The current analyses of the PROVIT sample, which was recruited as a randomized, placebo-controlled, multispecies probiotics study in individuals with MDD, showed significant changes in the circadian clock gene expression levels measured with RT-qPCR. The transcription of the core clock gene CLOCK changed significantly only in participants with MDD receiving the probiotics add-on therapy for 28 days and did not change in the placebo group. Gene expression levels of the other core clock genes ARNTL and TIMELESS changed significantly over the course of the intervention period in both groups. Comprehensive weak correlations between stool and serum metabolites in different pathways, e.g., amino acids, microbial derived metabolites, cofactors for epigenetic regulation, and other signaling pathways (see Fig. 5-7), and core clock gene expression markers may explain the interplay between the gut-brain axis and the molecular 24-h clock. Obviously, an orchestra of metabolites appears to be linked to the molecular 24-h clock aside from other time givers, such as light, day structure, and food intake, which in turn influence the microbiome.
From a molecular perspective, the circadian clock gene expression is regulated by a transcriptional/translational feedback loop of four key clock proteins, namely, CLOCK and ARNTL (two activators) and PER and CRY (two repressors). CLOCK and ARNTL form a heterodimeric complex; thereby, the CLOCK:ARNTL complex activates the transcription of PER and CRY. After translation, PER and CRY heterodimerize and are phosphorylated in the cytoplasm. Next, the phosphorylated PER:CRY complex translocates to the nucleus, where PER:CRY acts as an inhibitor of the transcriptional activator CLOCK:ARNTL. This complex circle generates a 24-h rhythm in the human organism, which manages crucial body functions such as heartbeat, body temperature, metabolic mechanisms, sleep, and hormone release [27, 28]. In our current PROVIT study, the gene expression levels of the core clock gene CLOCK changed significantly and the gene expression levels of ARNTL showed a trend toward significance. Linking these findings further substantiates an influence of the gut microbiome on the molecular 24-h clock. Hence, changing the gut microbiome could have favorable effects on the molecular clock and mood as the 24-h clock is directly linked to the regulation of neurotransmitter breakdown via MAO-A as described in detail in the introduction [27, 28].
However, the hospital inpatient treatment with regular day structure (regular get up and sleeping times, regular meals, and sport sessions) and the added vitamin B7 to the placebo and verum supplement could be an additional influencing factor for regulating the inner clock and the change of ARNTL and TIMELESS gene expression in both groups over time. It is known from previous literature that short-chain fatty acids, such as butyrate, released from gut microbiota have significant effects on the 24-h clock in the CNS, which might be one relevant factor for changes in clock gene expression in our current investigation [11, 12]. Results from a mechanistic murine study suggest that the sleep-inducing effects of butyrate are mediated by a sensory mechanism located in the liver and/or in the portal vein wall. Hepatoportal butyrate-sensitive mechanisms may play a role in sleep modulation by the intestinal microbiota [29]. Furthermore, butyrate may also lead to enhanced GABA production, by alleviating hypothalamic inflammation [35].
Interconnecting these findings with our PROVIT study, participants with MDD receiving probiotic add-on treatment elucidated a higher relative abundance of two butyrate-producing bacteria, namely, Coprococcus 3 and Ruminococcus gauvreauii, analyzed with 16S rRNA sequencing. This change in the gut microbiome is supported by the findings of the PROVIT-metabolomics project. The stools metabolome of depressed individuals in the probiotic group shifted toward higher normalized concentrations of butyrate, in comparison to the placebo group [30‒32]. Interestingly, Grosicki et al. [36] observed an association between the relative abundance of butyrate-producing bacteria Ruminococcus and Blautia and self-reported quality of sleep in a pilot study. Correspondingly, Ruminococcus was relatively more abundant after 28 days of probiotic add-on treatment in the PROVIT study [30‒32], which further suggests a connection between gut microbiome and sleep physiology [36].
In the literature, inflammation has been associated with the circadian rhythm [37]. In the PROVIT study, we additionally observed decreasing IL-6 proinflammatory cytokine levels, in individuals subjected to the probiotic group, whereas IL-6 gene expression levels increased in the placebo group. Linking these findings, a pathophysiological relationship between the proinflammatory cytokine IL-6 and a disruption in the ultradian activity rhythms in a mice model has been discussed lately [38].
However, changing the gut microbiome by probiotic supplementation could help to balance the homeostasis of the metabolome, which had clear alterations between patients with depression and healthy controls in one of the largest metabolomics MDD study [39]. Interconnecting our PROVIT-metabolomics data with the clock gene expression data, we discovered weak correlations between gene expression levels in PBMCs and various metabolites in stool and serum. The gene expression levels of ARNTL correlated with the branched-chain amino acids isoleucine, valine, the essential amino acid lysine, and the nonessential amino acid alanine in stool. Linking these results with the PROVIT-metabolomics analysis, interestingly, the amino acids lysine, alanine, isoleucine, and valine were altered in the stool’s metabolome of patients with MDD receiving the probiotic add-on treatment [30]. The transcriptional level of the core clock gene ARNTL correlated with the important methyl donor betaine in serum. Betaine is an essential metabolite in various biochemical processes, such as cellular replication, detoxification, and epigenetics [40]. In the current analysis, we observed a weak negative correlation between glycine levels and the gene expression levels of ARNTL in serum. Various studies suggest that the simplest nonessential amino acid glycine appears to be associated with sleep physiology [41‒44].
The gene expression level of the core clock gene CLOCK negatively correlated with the metabolite glycerol in the stool. Glycerol is an end product of lipolysis and is also potentially controlled by the circadian rhythm [44]. TIMELESS and NR1D1 gene expression levels correlate with propylene levels in serum. These results should be interpreted with caution as propylene glycol is a solvent of many medications [45].
The gene expression of NR1D1 shows a weak correlation with the MCFAs valeric acid in stool. Valeric acid is a microbiome-derived carboxylic acid, with protective effects on gut health. Like butyrate, valeric acid acts as a histone deacetylase inhibitor, which is a main enzyme in regulating epigenetic processes [40]. In serum, NR1D1 gene expression levels correlate with the choline derivate glycerophosphocholine and carnitine.
The more surprising correlation of the current investigation is the gene expression level of cycle length controlling PER3 with the essential amino acid tryptophan – the precursor of the sleep-regulating hormone melatonin. This result is of indispensable interest because tryptophan metabolism imbalance is a key player in the pathophysiology of neuropsychiatric disorders [46]. In serum, the metabolite trimethyl-N-oxide and the amino acids leucine and arginine correlate with the clock gene expression of PER3 in PBMCs.
As our PROVIT study showed that multispecies probiotics add-on therapy could change with the microbiome and metabolite pattern in the gut and in consequence the gene expression of important pathways (circadian rhythms, inflammation) [30‒32], we hope that in the future we will be able to improve the treatment of depression and sleeping disorders in a multi-factorial, biopsychosocial way including treatment of the gut-brain axis with probiotics. For example, probiotic treatment with strains of Lactobacillus and Bifidobacterium was associated with increased subjective sleep quality in healthy individuals, which was revealed in a recent meta-analysis [15]. Additionally, probiotic supplementation in MDD patients increased subjective sleep quality in a small pilot study after 8 weeks [47]. The molecular perspective of our probiotic intervention study underpins the findings of Wallace et al. [47] and Geoffroy et al. [15], by adding a potential interconnection of the clock gene expression in PBMCs and associated metabolites.
Nevertheless, further research is necessary to investigate the interplay between probiotics, microbiome, metabolome, and the molecular clock in MDD with new techniques and trials. For in-depth characterization of the influence of probiotics on the gut microbiome in individuals with MDD, shot gut sequencing will be necessary and has the potential to elucidate further mechanisms of the gut-brain axis, sleep, and depression. If the interplay between the gut-brain axis, the molecular clock, and mood can be replicated with other advanced techniques in the future, this may lead to the availability of even more sophisticated treatment options for MDD. Psychobiotics – tailored probiotic strains with mental health benefit – could be a promising treatment for MDD [48‒50]. Some studies already analyzed fecal microbiota transplantation and its utility in the treatment of MDD. As MDD is clearly associated with microbiome changes, this intervention becomes more and more realistic as a treatment with fewer side effects in the future [12, 51]. Probiotics are a milder form of influencing the gut-brain axis and can be easily applied as add-on therapy in clinical practice. In general, probiotics are considered safe in psychiatric settings. They are now even recommended with grade A evidence in the guidelines of the World Federation of Societies of Biological Psychiatry (WFSBP) for patients with major depression as add-on therapy [13]. Only few case studies describe a worsening of outcomes in critically and severely immunocompromised patients after probiotic interventions, mainly when Sacharomyces boulardii, a yeast, was administered to critically ill patients with weak immune status [52]. Overall, our results strengthen the idea that tailored probiotics with mental health benefits are beneficial add-on therapies in MDD with effects on the gut-brain axis and circadian rhythms. Nevertheless, further research is necessary to investigate the complex interplay between probiotics, the gut-brain axis, and the molecular clock in MDD.
Limitations
The PROVIT sample is relatively small and gene expression is measured in peripheral blood in PBMCs. Other samples, such as brain tissue, are difficult to obtain from study participants. In the field of psychiatry, we are always limited to the fact that we cannot obtain brain tissue to study gene expression. To date, it remains elusive if the gene expression in the periphery exactly reflects the gene expression in the brain. However, the clock genes are oscillating in every cell in our body; therefore, it is perceivable that a change in the periphery may also occur in the brain. Additionally, it cannot be ruled out that the higher number of smokers in the placebo group had a confounding influence on the results. However, the results did not change when adjusted for smoking. Medication intake across groups did not differ significantly, yet it might be a confounding factor via interactions in pharmacodynamics and pharmacokinetic pathways of medication and probiotic supplementation. Therefore, a replication of the demonstrated results in drug-free patients with MDD would be ideal. Moreover, vitamin B7 was added to the multispecies probiotic formula due to the requirement of the Ethical Committee that all participants get a beneficial substance. Vitamin B7 might be an additional confounding factor. In future study designs, subjective or objective sleep quality assessments via self-rated questionnaires or polysomnography should be utilized to measure the potential therapeutic effect of clock gene expression, probiotic treatment, and gut microbiome metabolites on sleep. In order to get a comprehensive overview of the alteration of the whole transcriptome influenced by probiotic add-on treatment, RNA sequencing or gene expression arrays could be used in the future to increase the knowledge. Nevertheless, larger sample sizes are warranted. Based on the rather small sample, we approached the targeted design with qPCR.
In the 4 weeks study period, we could not observe a significant difference in the clinical depression scores HAMD and BDI-II between the probiotics and intervention group. This is probable due to our limited study duration of 4 weeks, which was limited to the inpatient stay in our acute hospital setting. Further studies with an expanded study duration are necessary. An influence of probiotic add-on treatment, over a longer time course than 4 weeks, on depression parameters is likely as the findings of the PROVIT study are strongly pointing in that direction.
Conclusion
The gene expression of the core clock gene CLOCK differed significantly only in individuals with MDD receiving multispecies add-on treatment for 28 days. In addition, we observed an orchestra of metabolites in serum and stool correlating with clock gene expression in PBMCs, which strengthens the idea of a potential interplay between the gut-brain axis and circadian rhythm. Hence, probiotics might be a well-tolerated add-on therapy option, in individuals with MDD.
Although our results are pioneering, they should be interpreted with caution as the intervention period was limited and the size of the study group was relatively small. Future studies are necessary to get a more precise picture of how probiotics potentially influence the pathophysiological mechanisms, such as impaired circadian rhythms and inflammation processes in patients with MDD.
Acknowledgments
Especially, we would like to thank our MTA Renate Unterweger for the many years of loyal and excellent work before retirement – not only in the PROVIT study! Moreover, we also want to thank Anna-Maria Springer, who helped a lot to prepare the samples for the metabolomics analyses. We also thank all study participants and the complete research unit for participation.
Statement of Ethics
The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Ethics Committee of the Medical University of Graz, Austria (EK 29-235 ex 16/17). Written informed consent was obtained from all subjects involved in the study.
Conflict of Interest Statement
The authors declare no conflict of interest.
Funding Sources
The gene expression and NMR data analyses were funded by Stadt Graz and the Medical University of Graz (microbiome analysis). The research was also supported by the Institute AllergoSan with the free supply of probiotic supplements OMNi-BiOTiC® SR and the financing of a part-time study nurse and part of the microbiome analysis. This research was also funded by the Swiss National Science Foundation, 32003B_159518/1, Austrian Science Fund, Grant DOI 10.55776/P28854, 10.55776/I3792, 10.55776/DOC130, and 10.55776/W1226, Austrian Research Promotion Agency (FFG) 864690 and 870454, the Integrative Metabolism Research Center Graz, Austrian Infrastructure Program 2016/2017, Styrian Government (Zukunftsfonds, doc.funds program), the City of Graz, and Biotechmed Graz (Flagship Project DYNIMO). For open access purposes, the author has applied a CC BY public copyright license to any author-accepted manuscript version arising from this submission. The manufacturer (AllergoSan) performed the randomization and AllergoSan provided the add-on treatments. The funders had no role in the recruitment of participants or analysis of the data and in the writing of the manuscript.
Author Contributions
Kathrin Kreuzer: formal analysis, resources, data curation, writing – original draft, and visualization; Anna Maria Birkl-Toeglhofer and Johannes Haybaeck: sample management, gene expression analysis, and writing – review and editing; Alexandra Reiter, Nina Dalkner, Jolana Wagner-Skacel, Eva Reininghaus, and Susanne Astrid Bengesser: conceptualization, formal analysis, resources, data curation, writing – original draft, visualization, funding acquisition, and clinical study management; Frederike T. Fellendorf, Armin Birner, Robert Queissner, Alexandra Kohlhammer-Dohr, Annamaria Painold, and Theresa Lahousen-Luxenberger: recruitment, resources, data curation, and writing – review and editing; Alexander Maget and Martina Platzer: resources, data curation, writing – review and editing, and investigation; Matthias Seidl, Lilli-Marie Mendel, Melanie Lenger, Marco Mairinger, Anna Obermayer, and Tanja Färber: resources, data curation, and writing – review and editing; Tatjana Maria Stross: data curation and writing – review and editing; Alfred Häussl: writing – review and editing; Helmut Schöggl, Carlo Hamm, Daniela Amberger-Otti, Birgitta Leitner-Afschar, Nathalie Meier-Allard, Sandra Holasek, and Sonja Lackner: resources and writing – review and editing; Sabrina Mörkl: conceptualization, formal analysis, resources, data curation, writing – original draft, visualization, and clinical study management; and Hansjörg Habisch and Tobias Madl: metabolomics analyses, visualization, formal analysis, data curation, and writing – review and editing.
Data Availability Statement
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to the lack of a suited deposition platform.