The microbiota, the ecological community of commensal, symbiotic, and pathogenic microorganisms literally sharing our body space, includes more than 10 times the number of host cells to human cells.¹ The majority of the microbiome lives in the gastrointestinal (GI) tract and is composed of 10 100 trillion microorganisms, containing 100 times as many genes as our genome.² Symbiosis of the gut microbiota (GM) can maintain a normal physiology in the host, while dysbiosis of the GM can shift the balance and may induce diseases.

Recent studies have found a role for the GM in the gut-brain axis, which can alter minds and behaviors through the central nervous system (CNS).^3-8 Maintaining GM symbiosis is important for retaining healthy CNS functions. Sudo’s study was the first linking the GM and CNS, showing an increased hypothalamic-pituitary-adrenal (HPA) stress response and decreased brain-derived neurotrophic factor (BDNF) levels in the hippocampus of germ-free (GF) mice.⁹ Recently, researchers have found a relationship between the GM and CNS-related disorders in humans, such as Parkinson’s disease and autism: overall diversity and individual genus abundances were associated with their symptoms.^10-12

Probiotics are defined as “live micro-organisms which, when administered in adequate amounts, confer a health benefit on the host”.^13,14 Some types of probiotics have been used to treat gastrointestinal disorders, such as irritable bowel syndrome (IBS).^15-18 Also, probiotics have been studied in relation to altering visceral pain responses.¹⁹ Recently they have been reported to have an influence in the CNS by altering the GM composition.^3,8 Studies using probiotics to change CNS functions have increased over the last 10 years. The CNS functions mostly reported being altered by probiotics are psychiatric disorders and memory abilities. Studies in animals and humans have found, and reviews have summarized, potential mechanisms underlining these probiotic effects (Fig. 1). First, probiotics may directly alter CNS biochemistry, such as by affecting levels of BDNF, γ-aminobutyric acid (GABA), serotonin (5 hydroxytryptamine; 5 HT), and dopamine (DA), thus influencing mind and behavior.^20-24 Both the vagus and the enteric nerves are involved in this gut-brain interaction and can be affected by certain probiotics.^22,23 The HPA stress response, which regulates mood and emotion, has frequently been shown to be attenuated by probiotics, decreasing corticosteroid (CORT) levels.²⁵ The immune system can be influenced by probiotics, limiting pro-inflammatory cytokine production and inflammation, which, in turn, can affect the endocrine and nervous systems.^26,27 Probiotics manipulate GM by increasing microbiota diversity and beneficial bacteria compositions.^28-30 Improved GM changes metabolites, such as short-chain fatty acids and tryptophan, which can indirectly improve CNS function.^26,31

While there are reviews describing effects of probiotics on CNS function, there has been no previous systematic review that analyzes all the current animal and human studies in the field and describes the most effective probiotic interventions. Furthermore, animal studies in this area outnumber human studies, because behavioral experiments in animals are better established and standardized, while clinical studies in humans on this topic started to increase a few years ago and need translation from preclinical studies. However, from preclinical animal models to clinical trials in humans, there is no direct translation. To bridge the gap between preclinical and clinical studies, a systematic review is needed to summarize effective probiotic interventions on CNS function. We first sought to describe CNS functions that can be influenced by probiotics. Second we provide information about probiotic interventions including strain, dosage, and duration. Furthermore, we analyze and discuss the potential translation of animal models to human studies.

Search Strategy

This systematic review was conducted according to guidelines of the “Cochrane Handbook for Systematic Reviews of Interventions”, following the Preferred Items for Systematic Reviews and Meta-analysis guidelines.^32,33 Relevant studies were found by searches in the Pubmed, Medline, Embase, and Cochrane Library databases. Articles from 1950 to April 2016 were initially searched using the search terms “(probiotic OR gut microbiota) AND (behavior OR central nervous system)”. Additional citations were sought using references in articles retrieved from searches. We only included articles written in English. The first search was undertaken by analyzing text words contained in the title and abstract, and of the key words describing the articles. The second search was conducted according to the citations from all identified reports and relevant review articles.

Study Selection

We included all animal and human studies using different strains of probiotics. In human studies reports of both healthy volunteers and patients were considered.

The abstracts of the retrieved papers were screened for matching the following criteria: (1) the study included a probiotic intervention and (2) the study tested CNS function. After exclusion of non-relevant studies, the remaining articles were screened for the following criteria: (1) the study was described as a randomized controlled study (RCT), (2) the study was described as double-blind if studied in human participants, (3) the study involved use of probiotics in single- and/or multi-strains and those that only used pre-biotics or antibiotics were not included, and (4) the study included measures of behavioral experiments, neuropsychological measures (eg, electroencephalography, magnetoencephalography, and functional magnetic resonance imaging). Those that only involved neurophysiological measures were not included in the qualitative analysis but only in the Discussion (eg, measuring neurochemical level, HPA axis activity) because different studies tested quite divergent aspects of lower-level CNS activity.

Data Collection and Analysis

A data extraction and assessment form from the Cochrane Collaboration was used to further exclude inappropriate studies and to extract data we needed to analyze.³³ Double extraction of data was conducted: important data included the source of participants (animal species/strains, patient types), intervention groups (probiotic, placebo, or any other intervention), sample size, duration of interventions, and outcomes (behavioral changes as the primary outcome and lower-level changes [eg, biochemical changes] as secondary outcomes).

We assessed the quality of each study included using the Quality Assessment Tool for Quantitative Studies and the Quality Assessment Tool for Quantitative Studies Dictionary developed by The Effective Public Health Practice Project (EPHPP) (National Collaborating Centre for Methods and Tools [2008]).³⁴

The inclusion of both animals and human studies resulted in great heterogeneity of participants, interventions, and CNS functions measured. The outcomes of each study were described. A qualitative synthesis of selected studies was made with the aim of coming to conclusions about which probiotic interventions, at which dose, and for how long, were more/most effective with regard to certain CNS functions. Because of the heterogeneity of designs a meta-analysis was deemed to be inappropriate.

In total, 46 potentially relevant citations were obtained through the primary search strategy, which included animal and human studies, after excluding reviews (n = 98) (systematic reviews, narrative reviews, respective reviews, and systematic reviews and meta-analyses), other articles including comments, hypotheses, viewpoints, opinions, and editorials (n = 13), studies without the use of a probiotic (n = 33), and studies not focusing on CNS function (n = 16) (Fig. 2). Many articles concerned the potential, and demonstrated mechanisms of the effects, of GM and/or probiotics on CNS function. However, we only included studies that clearly described probiotics as interventions, and excluded studies only measuring GM composition when investigating CNS function. After the full screen, 8 more studies were excluded: four studies lacked a control arm or blinding, and 4 studies did not measure CNS functions at the behavioral level. In total, 38 studies remained for the qualitative analysis. Of them, 15 were in humans and 25 were in animals (2 studies were conducted in humans and animals).

Animal Studies

Of the 25 studies, 19 provided explicit information in the Materials and Methods section on the numbers of animals allocated to each treatment group, and they accounted for all the animals in their results, while 6 studies provided specifics regarding treated animals only in the pertinent areas of the Results section. Results from the quality assessment tool for quantitative studies showed all studies were strong in global ratings for selection bias, study design, confounders, blinding, data collection method, and withdrawals and dropouts (no weak rating): 15 studies described assigning animals randomly to treatment groups or control, although these animal studies were supposed to be RCTs. Only four studies might have confounders because they used both male and female animals while the other studies only used single gender animals, excluding the effect of gender on the results. Only 13 studies reported that experimenters were blinded toward the interventions or exposure status of the participants (we included the other 12 studies that did not describe blinding, because animals were not aware of the research question or intervention, and we ignored potential effects of the experimenter). Data collection tools shown in all studies were considered valid and reliable; no study reported withdrawals or drop outs.

Although all of studies examined were on rodents (mice or rats), the selection of animals was heterogeneous in some respects: strains of animals and health conditions of the animals. Studies were also heterogeneous in the CNS functions they were looking at, the experimental models they used, the probiotics they used, and the dose and duration of the probiotic interventions. Due to the heterogeneity of the included studies, we only described the results based on the interventions and measurements of the CNS functions (Table 1).

Most of the studies (18/25) investigated the effects of a single strain of probiotics. In 18 studies using single-strain probiotics, seven used Bifidobacterium, eleven used Lactobacillus, and one used Clostridium (one used both Bifidobacterium and Lactobacillus). Almost all the studies found significant effects on measured CNS functions, except for one testing effect of Bifidobacterium infantis on depression-like behavior. The concentration of the effective probiotic interventions ranged from 10⁷ to 10¹¹ colony-forming units (CFU), with the most using 10⁹ (14/25) or 10¹⁰ (6/25) CFU per animal per day. The duration of the probiotic treatments ranged from 6 to 77 days, with the most frequent period being 14 days (7/25). Effects of different probiotics on different specific CNS functions in animals were analyzed and described in the following text.

Human Studies

In total, 15 human studies were included. All of the selected studies had strong ratings in the quality assessment tool for quantitative studies, although one of the studies did not describe the age and gender of the participants in each group.⁵¹ Among the 15 studies, 8 used a single-strain probiotic (L. casei, L. casei subsp. rhamnosus, L. casei Shirota, L. plantarum, and B. infantis), of which 2 used probiotic containing milk, and the other 7 studies used multi-strain probiotics. Eight of the 15 studies found significant effects of the probiotic interventions. Doses of the effective interventions ranged from 10⁷ to 3.63 × 10¹⁰, and the duration of the treatments ranged from 20 days to 8 weeks. Doses around 10⁹ (5/8) and 10¹⁰ (3/8) were used most often. Durations were most commonly 4 (6/15) and 8 (4/15) weeks. Due to the heterogeneity of the studies (eg, probiotic interventions and measurements of CNS functions), we only describe the results based on the different interventions (Table 2).

The number of studies using probiotics to improve central nervous system function has increased over the past 10 years, though with a focus on effects in animals. Researchers have used various strains of probiotics and studied various CNS functions. Summarizing the divergent findings motivated us to perform a systematic review of this research area. Previously, there was no systematic review or meta-analysis of the effects of probiotics on CNS function in animals and humans. To date, there are a few reviews on probiotic effects on infantile colic, which may reflect peripheral nervous system action, and one recent study reviewing only human studies.^64-66 Similarly, among the 56 RCTs so far which tested probiotics in adults with IBS that have or have not shown effects on peripheral (bowel) functions,¹⁵ none have investigated whether any CNS effect was affected or improved. We identified studies applying probiotics as single- or multiple-strain preparations in animals and humans. Because of the diversity of the interventions and the CNS functions tested in these studies, we did not perform a meta-analysis.

Effects of Probiotics

Combining all the studies in animals and humans, probiotics appear to have a positive effect in improving central nervous system function. However, a publication bias toward positive results cannot be excluded. Based on currently available studies, we can see that most of the studies used Bifidobacterium and Lactobacillus preparations, and most of them were effective in improving specific CNS functions. Again, however, negative results in relation to other functions may have been unreported, even in otherwise positive studies. Doses of 10⁹ and 10¹⁰ CFU have been used in most studies showing an effect on behaviors. An intake of the probiotic for 2 weeks in animals and 4 weeks in humans is apparently sufficient to elicit measurable effects.

B. longum, B. breve, B. infantis, L. helveticus, L. rhamnosus, L. plantarum, and L. casei were the most commonly used preparations in these studies, as single- or multi-strain preparations, all of which were able to improve anxiety, depression, and memory related behaviors, based on animal models. All of these probiotics are regarded as “good” bacteria, presumably inhibiting the growth of harmful bacteria or pathogens and/or improving the immune system.^67-69 These probiotics were also found to reduce the symptoms of gastrointestinal disease, such as irritable bowel syndrome.^15,16,58,70 Probiotics may play an important role in gut-brain axis communication, thereby benefiting both the brain and the gut.

While some studies found no significant effect of probiotic intervention, the evidence is inadequate to conclude that the interventions were ineffective because some difficulties and/or weak points exist. For example, schizophrenia as a severe mental illness, and being closely related to a genetic disposition, may be a case in which probiotics can hardly be expected to have a significant effect on changing symptoms.⁵⁴ Probiotic doses in some studies were below the supposed effective dosages (at least 10⁹ CFU), such as 10⁸ CFU in the study in rheumatoid arthritis patients and 10⁶, 5 × 10⁷, and 10⁸ CFU in IBS patients.^55-58 Also, in two studies, one in male smokers and one in children with ASD, the intervention periods were 3 weeks, shorter than the effective period, which seems to be 4 weeks, that can make a measurable effect.^59,60

Also, caution is warranted when drawing conclusions from the human studies that used psychological questionnaires and/or scales rather than behavioral or neuropsychological experiments, resulting in subjective biases. The clinical efficacy of probiotic interventions and guidelines for their administration in diseases such as diarrhea, allergies, IBS, and inflammatory bowel disease has been addressed in previous reviews.^15,71,72 More studies investigating probiotic effects in mental disorders are needed.

Mechanisms of Action of Probiotic Effects

The current state-of-the-art suggests several mechanisms: the endocrine system, immune system, action of enteric neurons, and commensal bacteria (or their metabolic activity). This evidence has come primarily from preclinical studies, while a few clinical studies have analyzed cortisol and cytokine levels in saliva and plasma. The HPA axis activity has been linked closely to mood disorders and memory abilities.^73,74 Many probiotics reduced HPA axis activity by decreasing CORT and/or ACTH levels, including most of the Lactobacillus strains tested: L. plantarum, L. helveticus, L. fermentum, L. rhamnosus, and L. casei.^{19,22,29,36,39,43,44,49} However, single strains of Bifidobacterium such as B. infantis, B. longum, and B. breve had no effect on CORT levels.^27,29,36 BDNF is the key for neurogenesis and synaptic plasticity, which structurally support CNS function.^75,76Lactobacillus (L. plantarum, L. helveticus, L. fermentum), Bifidobacterium (B. longum), and C. butyricum increased brain BDNF.^34,37,43,46 Neuronal activation can be indicated by the nuclear localization of c-Fos; the effect of c Fos in the CNS depends on its location. The combination of L. rhamnosus + L. helveticus improved c Fos expression in the hippocampus and improved memory ability, while L. casei decreased it in the paraventricular nucleus of hypothalamus region and alleviated stress responses.^45,50 Neurotransmitters 5-HT, DA, and GABA are closely related to psychiatric conditions, and were influenced directly by many strains of probiotic (L. plantarum, L. helveticus, L. fermentum, L. rhamnosus, and B. infantis). The vagus nerve has been proposed as a pathway of probiotic effects because neurochemical and behavioral changes due to L. rhamnosus and B. longum were not seen in vagotomized animals. Direct evidence for the role of the vagus nerve also comes from studies showing that gastric vagal afferent activity was enhanced by L. casei. The excitability of the enteric nervous system, which is connected to the vagus nerve, has been shown to be modulated by B. longum.²²

Probiotics also alter CNS function indirectly through several other pathways. L. helveticus, B. infantis, and B. fragilis enhanced serum tryptophan (precursor of 5-HT) levels and reduced its metabolites. Most of the probiotics tested affected the immune system by decreasing pro-inflammatory cytokines and increasing anti-inflammatory cytokines. Another important pathway through which probiotics may modulate CNS function is intestinal barrier permeability, which is essential for maintaining the immune and nervous systems. Increased intestinal barrier permeability is associated with psychiatric disorders, such as depression and autism, while it can be restored by probiotic formulations of B. longum and L. helveticus, along with improved CNS function.^31,41,77

According to the data reviewed, different probiotics exhibited several common effects; however, these effects were strain-dependent and occurred via different pathways at a lower level of the CNS. Thus, more studies are needed for clarify which probiotics target which central biochemical substances and behaviors. In clinical applications, interventions with a probiotics cocktail may have greater effects, because different probiotics may create their effects at the same time through different pathways. However, as yet, such an approach currently lacks clinical evidence.¹⁵

Translation of Animal Studies

There are many animal studies about the gut microbiome-brain interaction using germ-free, specific pathogen free (SPF), or gnotobiotic animals, colonization with specific microbiota, probiotic intake, and infections to deliberately alter the GM and to manipulate CNS function.^{9,29,30,35,37,45,78-82} In humans, we are not able to adapt most of these models for ethical reasons. Ten of the animal studies included in our review used probiotics in animals whose health state had been disturbed by various manipulations, which included antibiotic treatment, gene knockout, inflammation, infection, maternal immune activation, hyperammonemia, and diabetic induction, and depression induced by myocardial infarction. All of the manipulations were aimed at inducing changes in CNS function, including anxiety and depression-like behavior, memory impairment, or ASD-like behaviors. In humans, interfering with the participants’ healthy state is not an option. What is possible is to explore the GM composition, correlating it to certain behaviors, and using probiotics to manipulate the GM-brain interaction. It is also possible to temporarily affect single functions, such as the stress response at the central level or the GM composition by varying the food or drugs used. As yet, evidence from studies using probiotics is confined to animal studies. Validity estimates of probiotic intervention from human studies are still missing and thus need to be carried out.

The translation of behavioral models from animals to humans has both possibilities and limitations. The tests used to measure anxiety in animals, such as the elevate plus/Barnes maze, light/dark box, defensive burying, open field/arms, and step-down, have no equivalents in humans, and neither do tests such as the forced swim and maternal separation for inducing depression and negative mood. In human studies, measurements of anxiety and depression rely primarily on scales such as the HADS, which has accuracy issues due to subjectivity and emotional bias from the participants/patients. Moreover, it remains questionable whether the behavioral tests used in animals do, in fact, adequately reflect the assumed CNS dysfunction (anxiety, depression) in humans and, more specifically, in patients. This leads to a demand for appropriate behavioral tests of anxiety and depression not only for patients with psychiatric disorders but also for the healthy population, and for adequate behavioral measures in animals that match these functions and dysfunctions in humans.

In addition to behavioral measurements, neuroimaging methods may provide insights as to what is altered in the brain that causes behavioral changes after the consumption of probiotics. An emotional faces attention task used in the fMRI study of Tillisch et al⁶³ is one example: the brain response to emotional stimuli that may be related to psychiatric conditions was changed after a 4-week intake of probiotics.

Learning and memory abilities can be tested via numerous paradigms in humans. For spatial memory, there are computerized versions of the Morris water maze (VMWT) and the Blue Velvet Arena (BVA), which is also a variant of the Morris water maze for humans.^83-86 For non-spatial memory, object recognition tasks and fear conditioning have been used widely in humans. These memory tasks can be conducted in combination with neuroimaging experiments, such as fMRI and magnetoencephalography (MEG) experiments.

There are also several ways to experimentally induce stress in humans. The Trier Social Stress Test, developed 20 years ago, during which participants are asked to play a role in a job interview, or in performing a public speech, can effectively increase the HPA axis and sympathetic-adrenal-medullary activity.⁸⁷ Social stress can also be induced using the Cyberball paradigm, during which stress comes from social exclusion and/or ostracism.^88,89 Noise as a stressor is easy to manipulate in a laboratory environment by exposing participants to unpleasant sounds so as to induce psychological stress. Cognitive tasks can also stimulate stress responses with the advantage of being able to study the stress level by measuring task performance.⁹⁰ Other and more physical stressors include cold pressor tasks,⁹¹ heat pain,⁹² and the CO₂ challenge test, inducing stress/panic in participants by inhaling carbon dioxide-enriched air.⁹³ All of these have also been shown to be compatible with brain imaging studies.

Conclusions and Indications

We reviewed the effect of probiotics on the central nervous system in randomized controlled trials in animals and humans, and analyzed the possibility of translating animal models to human studies because few human studies have been conducted to date. According to the qualitative analyses of current studies, we can provisionally draw the conclusion that B. longum, B. breve, B. infantis, L. helveticus, L. rhamnosus, L. plantarum, and L. casei were most effective in improving CNS function, including psychiatric disease-associated functions (anxiety, depression, mood, stress response) and memory abilities. Doses between 10⁹ and 10¹⁰ CFU and durations of 2 weeks in animals and 4 weeks in humans have shown sufficient effects. Also, translations of animal studies to human studies may be applicable. Human studies can be conducted using the same probiotics and similar experimental paradigms in the emotional and neurocognitive domains. More experimental designs in humans should be developed, and more neuroimaging studies should be conducted rather than using only psychological questionnaires or scales. In addition to studies in healthy populations, clinical studies in patients with mental diseases would be worthwhile, because those with gastrointestinal disorders and psychiatric comorbidities, in general, appear to benefit from probiotic interventions.