Microbial genes, brain & behaviour – epigenetic regulation of the gut–brain axis



To date, there is rapidly increasing evidence for host–microbe interaction at virtually all levels of complexity, ranging from direct cell-to-cell communication to extensive systemic signalling, and involving various organs and organ systems, including the central nervous system. As such, the discovery that differential microbial composition is associated with alterations in behaviour and cognition has significantly contributed to establishing the microbiota–gut–brain axis as an extension of the well-accepted gut–brain axis concept. Many efforts have been focused on delineating a role for this axis in health and disease, ranging from stress-related disorders such as depression, anxiety and irritable bowel syndrome to neurodevelopmental disorders such as autism. There is also a growing appreciation of the role of epigenetic mechanisms in shaping brain and behaviour. However, the role of epigenetics in informing host–microbe interactions has received little attention to date. This is despite the fact that there are many plausible routes of interaction between epigenetic mechanisms and the host-microbiota dialogue. From this new perspective we put forward novel, yet testable, hypotheses. Firstly, we suggest that gut-microbial products can affect chromatin plasticity within their host's brain that in turn leads to changes in neuronal transcription and eventually alters host behaviour. Secondly, we argue that the microbiota is an important mediator of gene-environment interactions. Finally, we reason that the microbiota itself may be viewed as an epigenetic entity. In conclusion, the fields of (neuro)epigenetics and microbiology are converging at many levels and more interdisciplinary studies are necessary to unravel the full range of this interaction.

Since their emergence, the evolution of multicellular eukaryotic organisms has taken place in the presence of prokaryotes and a plethora of diverse micro-organisms now colonize virtually all body surfaces of animal hosts, residing as beneficial symbionts, harmless commensals or pathogenic parasites (Dave et al. 2012; Schloissnig et al. 2013; Turnbaugh et al. 2007) most prominently within the gastrointestinal tract. An understanding of the importance of these interactions is undergoing a renaissance with large-scale scientific projects like the Human Microbiome Project (HMP; Human Microbiome Project Consortium 2012; Turnbaugh et al. 2007) designed to sample, determine and quantify all human-associated microbiotic life. In parallel the European HMP-counterpart MetaHIT focuses on the intestinal-tract microbiota in general (Qin et al. 2010) with the Eldermet project centering on the elderly (Claesson et al. 2012).

An estimated 90% of cells found in the human body are not human after all but of mostly prokaryotic origin, derived from at least 40 000 bacterial strains in 1800 genera (Forsythe & Kunze 2013; Frank & Pace 2008; Luckey 1972). Though considerably smaller in size, these approximately 100 trillion cells add up to a mass of almost 1–2 kg in an adult individual (Forsythe & Kunze 2013) – approximately the weight of a full-grown human brain (ca. 1.5 kg, Parent & Carpenter 1996).

There is a rapidly increasing amount of evidence implicating host–microbe interactions at virtually all levels of complexity, ranging from direct cell-to-cell communication to extensive systemic signalling, and involving various organs and organ systems, including the central nervous system (CNS). The microbiome (see Box 1) critically supports host metabolism and yields a source of metabolites, many of which would otherwise not be available to host cells. This is achieved by the huge diversity of genetic material that constitutes the microbiome. It is estimated that the human gut harbours more than 3.3 million non-human genes (Zhu et al. 2010), making the 23 285 human protein-coding genes currently annotated in the ENSEMBL database (http://www.ensembl.org) appear almost negligible. Thus, the sole presence of micro-organisms as well as the specific composition of this microbiota has multiple, critical consequences for host physiological and metabolic processes ranging from postnatal development and immunomodulation to, perhaps most surprisingly, behaviour and cognition (Sommer & Bäckhed 2013) which forms the basis of this review.

BOX 1. Glossary

  • Microbiota: The microbiota is the sum of all micro-organisms associated with a given host individual. Micro-organisms can be found on all body surfaces (including the lumen of gastro-intestinal organs, which strictly speaking belong to the outside world) and cavities, such as the skin, nose, ears and genitals. The microbiota is not only limited to the bacterial domain of life but also includes archaea as well as eukaryotes such as protozoa, fungi and (mostly parasitic) nematodes. Though not considered as organisms, one could even include viruses (host associated as well as bacteria associated phages). However, the mammalian non-pathogenic virome and the complexity of phages are largely unexplored and will not be addressed in this review.
  • Microbiome: The genome of a given organism is defined as the sum of its chromosomal genes and also includes the extra-chromosomal genetic information found in other organelles or endosymbionts (e.g. plasmids, chloroplasts, mitochondria). Analogously, the microbiome is defined as the entirety of all genes present in the micro-organisms colonizing a given host. The microbiome is also referred to as metagenome. Together, the genome and the associated metagenome make up an organisms hologenome (Brucker & Bordenstein 2013), a term coined by Rosenberg et al. in 2007 (Rosenberg et al. 2007 2009).
  • Enterotype: The concept of enterotypes was introduced in 2011 to define inter-individual variation in gut-microbiota species composition. More specifically, three distinct microbiomic clusters were identified that were mainly separated not only by abundance of certain species but also by abundance of genes with shared molecular function (Arumugam et al. 2011).
  • Germ-free (GF): Also referred to as axenic (from the Greek ‘free from foreign[er]’), though axenic can have other connotations in microbiology. GF animals are defined to be free from any microbial colonization and are kept in isolators under strictly sterile conditions. Also see Box 2.
  • Gnotobiotic: A system, in which all organisms are either defined by or known to the investigator, is referred to as gnotobiotic (from the Greek ‘known life’). Thus, a gnotobiotic animal is inhabited only by certain micro-organisms. The status includes the absence of any colonization as in GF animals, since it can also be viewed as a known status. Often the term is used to describe formerly GF animals that have be colonized by a defined set of micro-organisms such as the ‘Schaedler Flora’ (Schaedler et al. 1965) or mono-association with just one strain of bacteria. For problems with defining the gnotobiotic status see Box 2.
  • Probiotic: Probiotics are living organisms that contribute to a host-beneficial microbial flora. Thus probiotic organisms can be viewed as symbionts.
  • Prebiotic: Meant to contrast the term antibiotic, prebiotics are chemical compounds that influence the microbial flora in a host-beneficial way.
  • Psychobiotic: A concept recently introduced by Dinan, Stanton and Cryan defines psychobiotics – analogous to probiotics – as live organisms that produce positive effects on mental health (Dinan et al. 2013). It can also be argued that psychobiotics may exhibit benefits to healthy individuals, for example as memory enhancers (Misra & Medhi 2013).
  • Dysbiosis and probiosis: A dysbiotic state is marked by disadvantageous alterations in microbial composition. Probiosis, on the other hand – following the definition of probiotics – rather describes a beneficial microbial status that supports normal host function, e.g. stress resilience.
  • Nucleomodulins: Several bacteria can influence their host's transcriptome by secreting protein effectors directly targeting the epigenetic machinery. So far these effectors have only been found in intracellular parasitic bacteria and viruses. Bierne and Cossart proposed to classify non-eukaryotic epigenetic effector proteins collectively as nucleomodulins (Bierne & Cossart 2012). We propose to apply this concept to potential transcriptional regulators that affect neuronal gene expression and alter host behaviour and could hence be termed neuro-nucleomodulins.

The microbiota–gut–brain axis

The discovery that differential microbial composition is associated with alterations in behaviour and cognition has significantly contributed to establish the ‘microbiota–gut–brain axis’ as an extension of the well-accepted ‘gut–brain axis’ concept. This concept is used to describe the bidirectional communication between the CNS and intestinal organs and was first introduced in terms of peripheral regulation of emotions by William James and Carl Lange in the 1880s and further challenged and refined by Walter Cannon in the 1920s to account for the primacy of the brain in regulating gastrointestinal function (see Mayer 2011). However, in the light of new and intriguing data, mostly resulting from the study of rodents, the gut–brain axis has been reviewed from a number of perspectives, focusing on different aspects ranging from basic microbiology to translational applications (e.g. see Bercik et al. 2012; Berer & Krishnamoorthy 2012; Collins et al. 2012b, 2013; Cryan & Dinan 2012; Cryan & O'Mahony 2011; Forsythe & Kunze 2013; Lyte 2011, 2013; Nicholson et al. 2012; Rhee et al. 2009; Sommer & Bäckhed 2013). In this review, we want to highlight our current understanding on the underlying mechanisms of microbiota–gut–brain interactions and associated behavioural alterations with an emphasis on a potential epigenetic contribution to these mechanisms.

A new epoch is emerging with these findings in basic research and animal studies beeing translated into the clinic. Indeed, it is becoming clear that certain pathologies, which are associated with an altered microbiome, are connected to mood, stress, behaviour and/or cognition (for review see Grenham et al. 2011; Shanahan 2012). In this regard, irritable bowel syndrome, which is highly comorbid with mood disorders such as depression, also leads to decreased cognitive performance (Berrill et al. 2013; Kennedy et al. 2013). Moreover, an important recent neuroimaging study validated rodent studies (e.g. Bravo et al. 2011) in implicating microbe-brain signalling in modulating resting brain activity in key circuits involved in pain, emotion and cognition (Tillisch et al. 2013).

Microbiota-associated phenotypes and behavioural alterations

Tables 1 and 2 summarize the growing body of research emanating from rodents that demonstrate a role for microbiota in behaviour. At the centre of many of these studies are animals that have been raised in a sterile environment and thus without microbiota, referred to as GF (see Box 2). Additionally, behavioural studies on animals with either defined infections, antibiotic treatment or administration of probiotic bacteria have been carried out (for a review see Cryan & Dinan 2012; Foster & Neufeld 2013, see Table 1). These studies showed reproducible and largely consistent effects of the various microbial states on behaviour in mice. The most commonly reported phenotype was altered anxiety-related behaviour, which can be assessed by a variety of tests (Table 1).

Table 1. Behavioural phenotypes of differential microbiota compositionThumbnail image of
Table 2. Biochemical and molecular alterations with differential microbiota compositionThumbnail image of

BOX 2. The germ-free animal

GF and other gnotobiotic animals have been generated since the early 20th century (for a historical introduction see Reyniers 1959). To achieve a GF status in a mammal, a new-born needs to be delivered by Caesarean section and hand-reared with sterilized milk in a sterile, isolated environment. Further generations can then be derived by Caesarean section exclusively, since rearing can be arranged by an already GF foster mother. Future colonies of GF animals can be maintained from interbreeding with each other within a suitable Germ-Free Unit. However, it is important to bear in mind that increasing evidence suggests that Caesarean-derived newborns are, against longstanding dogma, not sterile (for a recent review see Funkhouser & Bordenstein 2013). Though several studies on GF animals monitored contamination using culture methods, these methods are heavily biassed and bacteria, fungi, protozoans and viruses that are not easily cultureable will not be detected. The only way of definitely determining the microbial status of GF animals would be using deep sequencing. Targeted amplification of 16s rDNA by polymerase chain reaction would only be useful to determine the status of bacterial colonization, and would not give information about fungi or viruses.

Already in 1970 the first explicit behavioural observations have been made on gnotobiotic piglets (Bähr 1970). Notably, it took another 34 years until the first studies on GF mice were published, that showed alterations in brain function (Sudo et al. 2004). Since 2011 a number of studies demonstrated behavioural alterations in GF mice and thereby significantly extended the microbiota–gut–brain axis concept. Thus, this almost century-old area of research has still not lost its importance.

There is now an increasing number of studies focusing on the positive behavioural effects of various bacterial strains, mostly Bifidiobacteria and Lactobacillus species (see Table 1 and Bercik et al. 2010, 2011b) but also transient commensals such as Mycobacterium vaccae (Matthews & Jenks 2013).

Moreover, an increasing number of studies in animal models of stress, anxiety and depression also implicate a role for the microbiota in psychopathology (Bailey & Coe 1999; Bailey et al. 2011; O'Mahony et al. 2011; Park et al. 2013).

In almost all studies the authors also reported biochemical and molecular changes (Table 2). In addition to the common finding that the microbial status is associated with altered corticosterone levels in the blood plasma or serum of stressed as well as naïve mice, gene expression changes in different brain regions could be demonstrated (Table 2). Most commonly Bdnf and Fos expression levels were analysed as a correlate for differential neuronal activation. Interestingly, Bdnf is well-known for its function in neuronal plasticity, learning and memory, and a number of psychiatric and neurodegenerative diseases (Cowansage et al. 2010; Tapia-Arancibia et al. 2008; Walker et al. 2012). In addition, alterations in neurotransmitter signalling, including neurotransmitters and associated metabolites and neurotransmitter receptors have been described. Diaz Heijtz et al. (2011) took a genome-wide approach to define the transcriptional profile of the GF mouse in five different brain regions. Further analysis showed that genes associated with the functional categories ‘citrate cycle’, ‘synaptic long-term potentiation’, ‘steroid hormone metabolism’ and ‘cyclic adenosine 5′-phosphate (adenosine monophosphate)-mediated signalling’ were enriched among the differentially regulated genes, which supports the phenotypic observations. Interestingly, while in the cerebellum and hippocampus robust changes in gene expression were found, the hypothalamus, the brain region involved in the stress-activated hypothalamuic-pituitary-adrenal axis (HPA-axis), showed almost no differential gene expression.

While certain behavioural and biochemical parameters (including anxiety, sociability, HPA-axis and tryptophan metabolism) could be reversed by recolonization with a conventional microbiota or probiotic treatment, others were unaffected by restoration of a normal microbiota (including 5-HT concentration and social cognition; see Tables 1 and 2 for details). Indeed, reversibility of the anxiolytic phenotype in GF mice is only guaranteed if recolonization happens during a critical time window during adolescence (Clarke et al. 2013; Foster & Neufeld 2013). However, since this hypothesis has not been tested systematically, it is unclear if this observation can also be generalized to other GF-associated phenotypes.

Overall, accumulating evidence suggests that there is a correlation between the microbiota composition and brain and behaviour. While intestinal probiosis is associated with increased stress resilience and decreased basal anxiety, GF animals show increased basal corticosterone levels along with an increased stress-response, hyperlocomotion and decreased brain-derived neurotrophic factor (BDNF) levels but also reduced anxiety and social activity. It may appear as a contradiction that the endocrinological stress parameters are elevated while anxiety on the behavioural level is reduced. However, these parameters do not necessarily correlate (Neufeld et al. 2011) and may be differentially influenced by the microbiota. Dysbiosis of the enteric milieu on the other hand is marked by an increase in anxiety, depressive-like behaviour and memory impairment along with decreased concentrations of key neurotrophic factors involved in plasticity such as BDNF (Fig. 1). However, the underlying molecular mechanisms leading to these behavioural and biochemical alterations are not well understood. Interestingly, there is now a growing appreciation of the role of epigenetic mechanisms in shaping brain and behaviour.

Figure 1.

The microbiota affects neurophysiological, biochemical and behavioural parameters. Accumulating evidence suggests that there is a correlation between the microbiota composition and brain and behaviour. Host-supporting species (probiosis) alter behaviour and biochemical parameters of the host in a different direction as compared to pathogenic species (dysbiosis). The phenotype of GF animals seems to represent a rather intermediate state, featuring characteristics of both conditions.

Epigenetic mechanisms and neuroepigenetics

‘Epigenetics’ is one of the most overused words in the current scientific vocabulary (Ledford 2008). Initially, it was used to describe developmental programming but was later redefined more specifically to refer to heritable changes in gene expression that do not originate from mutations of the DNA sequence (Holliday 1987; Waddington 1953). The term is now commonly used in a broader sense, although it is associated with different connotations within special scientific fields. While some disciplines focus on the aspect of sequence-independent transgenerational germ-line inheritance of a phenotypic trait, others, especially in the fields of neuroscience or biological psychiatry, emphasize early-life experiences that influence development and behaviour during later life and adulthood. However, in a more cellular view, transgenerational epigenetic traits are interpreted in the context of somatic mitotic cellular differentiation within a multicellular organism. In molecular neuroscience, most often the term epigenetics is rather used to take into account the multiple molecular events that are involved in the dynamic regulation of neuronal gene expression.

Irrespective of the interpretation one may adopt, the molecular machinery mediating these seemingly different effects are indistinguishable in all interpretations of ‘epigenetics’. In general, this molecular machinery comprises plastic changes in regulation of nuclear architecture, chromatin structure and remodelling and gene expression. This epigenetic machinery includes post-translational modifications of histone proteins that serve at least two, non-mutually exclusive functions: regulatory factor recruitment and histone–DNA interaction. Acetylation of a lysine residue in the N-terminal tail region of a histone reduces the electrostatic interaction of the positively charged lysine with the negatively charged DNA, thereby making chromatin more accessible (Korolev et al. 2007). At the same time this modification serves as a recognition signal for bromodomain-containing proteins, which in turn recruit factors of the transcriptional machinery to the now accessible genomic site (Chen et al. 2010; Hargreaves et al. 2009; LeRoy et al. 2008). Apart from acetylation, an increasing variety of post-translational modifications of histone proteins are known, including phosphorylation, methylation (three possible methylation states on lysine and arginine residues), and ubiquitylation. In addition, the DNA itself can be modified by cytosine methylation, which is generally associated with silencing of a genomic element (Jaenisch & Bird 2003). More recently also the different products during active demethylation of the DNA, especially TET1-mediated cytosine hydroxymethylation, gained interest as they seem to have regulatory function that differs from the simple absence of methylation (Kaas et al. 2013; Mellén et al. 2012 p. 2; Rudenko et al. 2013; Szulwach et al. 2011). Finally, a steadily increasing number of non-coding RNAs, which are often transcribed from DNA sites previously thought to be transcriptionally inactive (‘junk DNA’), are assigned as part of the epigenetic machinery and can act on transcriptional as well as translational activity (for some recent reviews on the diverse functions and biology of the increasingly complex RNA landscape see, e.g. Landry et al. 2013; Ng et al. 2013; O'Connor et al. 2012; Ørom & Shiekhattar 2013; Wang et al. 2012 and references therein).

Interestingly, these processes have recently been identified to play an important role in cognitive processes during health and disease by regulating gene expression in the brain (for recent review see Day & Sweatt 2011a,2011b; Gräff & Tsai 2013a; Kosik et al. 2012; O'Connor et al. 2012). A connection between the two mechanisms of environmental information processing and epigenetics has been established, often referred to as neuroepigenetics (Day & Sweatt 2010; Sweatt 2013). Concomitantly, the term chromatin plasticity is used to acknowledge the fact that the neuronal nucleus is no exception to the general rule that virtually all components in the brain are able to undergo plastic changes (Dulac 2010). This rather new discipline tries to unravel the dynamic changes of transcriptional regulation in neurons upon stimulation or in disease. For example, dynamic regulation of gene expression in the hippocampus is critical for long-term memory consolidation and synaptic plasticity (Da Silva et al. 2008; Igaz et al. 2002, 2004) and a specific role for all three mechanisms of the epigenetic machinery has been demonstrated. Of these, histone acetylation is probably the best-understood, since the catalysing enzymes (namely histone acetyltransferases, HATs and histone deacetylases, HDACs) are well-studied and can be targeted pharmacologically, e.g. by using HDAC inhibitors, which are currently discussed as cognitive enhancers (Gräff & Tsai 2013a,2013b). Because of its activatory role, histone acetylation plays a permissive role in learning-induced transcription and the use of small-molecule HDAC inhibitors has been shown to augment memory consolidation and to parallel the beneficial effects of environmental enrichment on cognitive function mechanistically (Fischer et al. 2007). As such, HDAC inhibition has also been shown to ameliorate cognitive decline during ageing and in mouse models of neurodegenerative diseases such as Alzheimer's disease (for review see Gräff & Tsai 2013b; Stilling & Fischer 2011). The presumed mechanism of action for elevated histone acetylation to aid in procognitive processes is based on the idea that these molecules do not simply alter acetylation levels at random genomic loci but rather support a given cell's own regulatory program or facilitate it in cases where it has gone out of balance (McQuown & Wood 2011; Peleg et al. 2010; Stilling & Fischer 2011). The role of epigenetics in informing host–microbe interactions has received little attention to date. In the following sections we will elaborate on potential convergences between epigenetic mechanisms and host-microbiota dialogue.

Microbe–brain interfaces: potential sites for epigenetic regulation

Since microbes colonizing the host body are usually housed on the outer body surfaces, including the skin, mouth, lungs, gastrointestinal tract and vaginal mucosa, a large part of the interaction between the microbiota and their host will be mediated by host cells, which the microbes are in direct contact with. These are mainly epithelial cells but also cells of the immune system as well as peripheral neurons (Forsythe & Kunze 2013). However, several–mostly parasitic–bacterial species are capable of invading host tissues and can even live in intracellular vacuoles to manipulate their host cells directly (Liévin-Le Moal & Servin 2013 and see below). In addition, host dendritic cells are able to engulf living bacteria from the gut lumen and transport them through the body, though it is unclear which of the many organ barriers dendritic cells may cross and whether bacterial cells are later released or digested (Rescigno et al. 2001). Furthermore, surface-dwelling microbiota have means of signalling across the epithelial border by secretion of bioactive molecules that penetrate the outer barriers and are carried to distant effector organs, including the brain, through the blood stream and the lymphatic system (Rhee et al. 2009).

The diverse possible routes for microbiota–gut–brain signalling and diet-influenced gut–brain communication have previously been reviewed elsewhere (e.g. Collins et al. 2012b; Cryan & Dinan 2012; Forsythe & Kunze 2013; Lyte 2013; Montiel-Castro et al. 2013) and will thus only be discussed with regard to potential sites of epigenetic regulation. It should be noted that to date there is a paucity of direct evidence for the role of epigenetics in shaping host–microbiota interactions. Yet, there are plenty of indications as to how potential epigenetic mechanisms can modulate the biological interaction between hosts and micro-organisms. Interestingly, epigenetic mechanisms have been recently been put forward as a central mechanism driving host–pathogen interactions (Gómez-Díaz et al. 2012). Expanding this to non-pathogenic beneficial microbes is an important goal of this review in the context of brain and behaviour.

Interaction with the autonomous nervous system

As part of the peripheral nervous system, the autonomous nervous system is functionally subdivided into the enteric, sympathetic and parasympathetic nervous system. Since all of these systems are also involved in regulation of gastro-intestinal function, they provide the easiest target for an interaction of the microbiota with nervous tissue. Of these the vagus nerve further offers a direct link between the large intestine and the brain that makes it an interesting candidate for the study of the gut–brain axis.

As such, several of the studies investigating behavioural and neurophysiological changes investigated the contribution of the vagus nerve. Indeed, vagotomy abolished some of the effects found in studies on mice fed with probiotics or pathogens (Bercik et al. 2011b; Bravo et al. 2011, Table 1). Other experiments, however, suggested that at least some of the observed effects are functionally independent of the vagus or other autonomous pathways (Bercik et al. 2011a). Together, these findings indicate that the vagus nerve is an important, though apparently not the only, mediator of microbiota-gut–brain interaction and may depend on the bacterial strain used. The exact modalities of how the vagus interacts with the microbiota to induce such effects remain unclear. As such, increased or decreased activation of vagal pathways, for examples as a result of altered gut motility or neuroactive signals secreted by bacteria (see below), may lead to the observed effects. It will be interesting to see whether this altered vagal activation leads to sustainable epigenetic modifications in the respective cranial nucleus in the brain stem (dorsal nucleus of vagus nerve), which is further connected to the hypothalamus and the solitary tract.


The early colonization of the body with diverse micro-organisms offers a plethora of antigens, which are critical for appropriate maturation of the immune system (Cahenzli et al. 2013; Hooper et al. 2012). Consequently, GF animals exhibit severely immature immune function (Cebra 1999). Early-life establishment of the acquired immune system is heavily influenced by the presence of the microbiota and it is well-documented that its establishment and maintenance depend on epigenetic modifications that govern the expression of immune-related genes and transcriptional profiles of immune cells such as T cells (Stender & Glass 2013; Weng et al. 2012). Interestingly, gut microbiota have very recently been shown to modulate homeostasis and inflammatory response of the intestinal epithelium in an HDAC3-dependent manner (Alenghat et al. 2013), thereby establishing a direct connection between microbiota and epigenetic gene regulation. Increasing evidence shows a significant contribution of immune signalling in normal brain function as well as during ageing and in the context of neurodegenerative diseases (Collins et al. 2012a; Czirr & Wyss-Coray 2012; Lampron et al. 2013; Rostène et al. 2007; Soliman et al. 2013; Villeda et al. 2011). However, we are only beginning to fully understand this widespread interaction of the immune system with the brain.

One other mechanism for inducing immunomodulatory effects in disorders of the brain gut axis is in the context of a ‘leaky gut’ hypothesis. Indeed, chronic stress has been shown to disrupt the intestinal barrier, making it leaky and increasing the permeability to ions and bacterial peptides (Santos et al. 2001; Söderholm & Perdue 2001). These effects can be reversed by probiotic agents (Ait-Belgnaoui et al. 2012; Zareie et al. 2006). In line with these findings, human studies indirectly suggest increased bacterial translocation in stress-related psychiatric disorders such as depression (Maes et al. 2012).

Irritable bowel syndrome and visceral pain

Visceral hypersensitivity is a hallmark of pathologies of the gut–brain axis such as irritable bowel syndrome (IBS). The biological basis of IBS is poorly understood though there is evidence for a contribution of genetic risk factors as well as environmental stimuli such as early-life stress (Fukudo & Kanazawa 2011; Mayer et al. 2001). Moreover, we suggested that epigenetic mechanisms may be key to the manifestation of IBS (Dinan et al. 2010). In line with this Greenwood Van-Meerveld and colleagues looked at how epigenetic modifications in the brain may affect visceral sensitivity in rats. The authors locally injected an HDAC inhibitor Trichostatin A (TSA) into the brain after a water-avoidance stress and found an amelioration of the stress-induced increase in visceral pain sensitivity (Tran et al. 2013). Though by itself this study does not rule out a CNS-based mechanism of differential pain processing in the stressed rats, it hints to a potentially beneficial effect of HDAC inhibition in the treatment of stress-induced bowel symptoms like visceral pain. Microbiota have been shown to be altered in IBS (Jeffery et al. 2012; Shankar et al. 2013) and probiotic-based therapies have been shown to reverse visceral hypersensitivity in animal models (McKernan et al. 2010) and in human trials (Clarke et al. 2012). Future studies are needed to clarify what role microbes and microbial metabolites play in epigenetically modifying pathways relevant to visceral pain.

Stress and depression

Depression-related behaviours are also altered in mice treated with probiotics or subclinical infection (Desbonnet et al. 2010; Bravo et al. 2011, Table 1). Interestingly, depressive-like symptoms in animal models have been associated with epigenetic mechanisms such as altered HDAC activity. As such, chronic administration of the antidepressant drug imipramine induced HDAC5-mediated differential histone acetylation in the hippocampus of mice undergoing chronic defeat stress (Tsankova et al. 2006). HDAC5 activity and epigenetic regulation of gene expression in the nucleus accumbens were also specifically associated with chronic emotional stimuli such as cocaine exposure and social defeat stress (Renthal et al. 2007). Furthermore, Berton and colleagues demonstrated antidepressant-like properties of HDAC6-selective inhibitors and an HDAC6-dependent regulation of the behavioural stress response in mice (Espallergues et al. 2012; Jochems et al. 2013). However, it should be noted that in neurons HDAC6 is not localized to the nucleus but rather deacetylates cytoplasmic proteins, including alpha-tubulin and HSP90.

Autism and neurodevelopmental disorders

It is increasingly acknowledged that the development of autism-spectrum disorders (ASDs) is multifactorial with genetic as well as environmental factors contributing to their aetiology. The term ASD is used to collectively describe a group of disorders that are characterized by classical autistic symptoms, such as reduced sociability and social recognition. Recent evidence suggests, albeit in relatively small cohorts, that ASDs may be associated with alterations in microbiota composition and metabolism (Critchfield et al. 2011; de Theije et al. 2011; Douglas-Escobar et al. 2013; Gondalia et al. 2012; Louis 2012; Macfabe 2012; Ming et al. 2012; Mulle et al. 2013). In addition to these correlative findings in humans, GF mice have recently been shown to have core social deficits and increased repetitive behaviours similar to that observed in ASD (Desbonnet et al. 2013). Together, these data suggest that the microbiota is a critical determinant for the development of social behaviour and the aetiology of ASD. Interestingly, also epigenetic mechanisms have been implicated in the aetiology of ASD, as comprehensively reviewed recently (Grafodatskaya et al. 2010) and it is currently unclear if these are related to microbiota changes. Moreover, whether other neurodevelopmental disorders such as schizophrenia are associated with microbiota changes are not yet investigated neither in animal models nor human populations.

Mediators of microbe–brain interactions

The gut microbiota helps to break down certain nutrients, which subsequently can be further metabolized by host cells. Interestingly, several of these products are associated with neural function. As such, gut bacteria produce amino acids, such as GABA and tryptophan, and monoamines, such as serotonin, histamine and dopamine, that play important roles in the brain as neurotransmitters or their precursors (Lyte 2011; Lyte & Freestone 2010; Thomas et al. 2012a; Wall et al. 2014). Though they may also target the CNS by transport through the blood stream, it is likely that these neuroactive products primarily affect the neurons in the enteric part of the peripheral nervous system.

Short-chain fatty acids

Apart from direct action on neurotransmission, gut-dwelling bacteria generate a number other chemicals that display neuro-modulatory potential. For example, it was shown that some gut-dwelling bacteria produce spermidine (Noack et al. 2000), a ubiquitous polyamine that has been shown to have beneficial effects on ageing (Eisenberg et al. 2009) and age-associated memory impairment (Gupta et al. 2013), which may in part be mediated by an alteration in histone acetylation (Das & Kanungo 1979).

Moreover, it is well-known that gut bacteria are the key source of short-chain fatty acids (SCFAs) such as butyric acid, propionic acid and acetic acid. While these molecules do not belong to the classic neuroactive substances they may act on neuronal function in a more subtle way. The most well-known of these is probably butyrate, which acts as a potent inhibitor of HDACs (Candido et al. 1978; Davie 2003). Propionate and other SCFAs, as well as lactate and pyruvate, have HDAC-inhibitory functions as well, but to a much lesser degree compared to butyrate (Latham et al. 2012; Thangaraju et al. 2006; Waldecker et al. 2008). In a similar IC50-range are certain polyphenol metabolites that are produced by gut bacteria (Waldecker et al. 2008, Table 3). Acetate on the other hand serves as a substrate for acyl-CoA synthetase short-chain family member 2 (ACSS2) in the synthesis of acetyl-Coenzyme A (acetyl-CoA), the donor of acetyl groups used for acetylation of histone tail lysine residues by HATs. Acetyl-CoA can also be derived from citrate via the enzymatic activity of ATP citrate lyase (ACLY).

Table 3. In vitro IC50 concentrations for selected substances with HDAC-inhibitory function additional information
HDACiIC50 (mol/l)SourceReferenceType
  • Additional reference: Shimazu et al. (2013).

  • SAHA, suberanilohydroxamic acid (Vorinostat); TSA, Trichostatin A.

  • *

    Pyruvate values greatly differ between cited publications.

TSA1.80E−09Streptomyces platensisSelleck ChemicalsAntimycotic
SAHA1.00E−08SyntheticSelleck ChemicalsHydroxamic acid
Etinostat (MS275)5.00E−07SyntheticSelleck ChemicalsCarbamate
Pyruvate*8.00E−05All domains of lifeThangaraju et al. (2006)Alpha-keto acid
Butyrate9.00E−05Gut microbiotaWaldecker et al. (2008)SCFA
p-Coumaric acid1.90E−04Gut microbiotaWaldecker et al. (2008)Polyphenol
Propionate3.60E−04Gut microbiotaWaldecker et al. (2008)SCFA
3-(4-OH-phenyl)-propionate6.20E−04Gut microbiotaWaldecker et al. (2008)Polyphenol
Caffeic acid8.50E−04Gut microbiotaWaldecker et al. (2008)Polyphenol
d-β-Hydroxybutyrate4.00E−03Mammalian cellsShimazu et al. (2013)SCFA derivative
d-Lactate1.00E−02Diet/dairy, gut microbiotaLatham et al. (2012)Organic acid
Pyruvate*3.00E−02All domains of lifeLatham et al. (2012)Alpha-keto acid
l-Lactate4.00E−02Diet/dairy, muscular tissue, gut microbiotaLatham et al. (2012)Organic acid

Thus, both processes, HDAC inhibition and increased availability of HAT substrate may lead to increased histone acetylation and thereby stimulate stimulus-driven transcription in active neurons. This has been shown to facilitate long-term memory consolidation and neuroprotection/-regeneration in a numerous in vitro studies and animal models for learning and memory and neurodegenerative diseases (Fischer et al. 2010; Govindarajan et al. 2011; Gräff & Tsai 2013b; Peleg et al. 2010). Though the effect of SCFAs that reach the CNS may be rather subtle, cumulative chronic delivery of SCFAs by the gut microbiota may result in long-lasting, stable effects on gene expression. Indeed, intracerebroventricular administration of relatively high doses of the SCFA propionic acid results in some autistic-like behaviours in rats (MacFabe et al. 2011; Thomas et al. 2012b).

Direct molecular interactions with the epigenetic machinery

Several bacteria can influence their host's transcription by secreting protein effectors that mimic host-endogenous transcriptional regulators and alter the epigenetic landscape of the host cells (Bierne 2013; Bierne & Cossart 2012; Bierne et al. 2012; Eskandarian et al. 2013; Hamon & Cossart 2008; Pennini et al. 2010; Rolando et al. 2013). Also certain influenza viruses make use of the hosts epigenetic machinery to replicate or hide within the hosts genome (Minárovits 2009) and the viral encoded histone-mimicking protein NS1 has been described to mediate transcriptional repression of the host cell's antiviral response (Marazzi et al. 2012). In fact, Bierne and Cossart (2012) have proposed to classify these non-eukaryotic effectors collectively as nucleomodulins. However, such effectors have yet only been shown to exist in intracellular parasites like Legionella pneumophilia that posses certain adapted secretion system and have a more direct contact to the intracellular milieu to interact with host signalling cascades. It is unclear whether brain-borne pathogens may have similar capabilities to alter transcription in neurons in a parasite-advantageous way, which in turn could have an effect on host behaviour (also see below).

While none of these mechanisms are explicitly described for gut-dwelling species or for parasites that live in brain tissue, it shows the potentially versatile and manifold ways that are open to microbes to interact with the host's epigenome. Undoubtedly, numerous parasites target for the brain and have various means to do so (for a review see Kristensson 2011). Just recently a well known bacterium (Staphylococcus aureus) was shown to stimulate nociceptive neurons directly (Chiu et al. 2013), which encourages us to rethink what we know about gut bacteria and whether they may stimulate neurons of the autonomous nervous system by similar mechanisms, which could have implications for pathologies associated with visceral pain. Indeed, probiotic cell-wall components were shown to activate intestinal sensory neurons (Mao et al. 2013) and decreased excitability was demonstrated for myenteric after-hyperpolarization neurons prepared from GF mice (McVey Neufeld et al. 2013). However, the exact molecular signalling pathways inducing these neurophysiological alterations remain elusive. In summary, the microbiota has multiple ways of interaction with host physiology and behaviour, some of which are potentially mediated by alterations of the epigenetic landscape of different host cells and tissues including neurons and glia. We suggest that the picture of the microbiota–gut–brain axis should be extended by these mechanisms (Fig. 2), provided experimental evidence will support our hypotheses.

Figure 2.

Scheme of epigenetic mechanisms in the microbiota–gut–brain axis. We propose to extend the classic picture of the microbiota–gut–brain interaction by effects of the intestinal micro-organisms on epigenetic processes in the brain through diverse mechanisms such as neuroactive metabolite secretion, immunomodulation and other yet unexplored neuro-nucleomodulins.

The microbiota and gene–environment interactions

Since epigenetic mechanisms are likely mediators of gene–environment interactions as well as potential mediators for interactions between microbes and host, it is important to consider how the microbiota is linked to gene–environment interactions. In the traditional view we start to become colonized with bacteria as we are delivered through the birth canal of our mother. However, it is worth noting that there is an increasing body of evidence that the sterile-womb paradigm needs a critical review (Funkhouser & Bordenstein 2013) and that transmission of certain microbes already occurs in utero. The mammalian microbiota then becomes more complex during delivery and is further established after birth. Hence, it is not surprising that the microbiota has immense effects on pre- and postnatal development and that detrimental alterations in early-life stages may lead to undesirable phenotypes during adulthood. However, the microbial composition is not fixed once and for all, as it is subject to change through various environmental factors (see Marques et al. 2010, Table 4). As such, the mode of delivery shapes microbial composition not only in the gut but also in multiple body surfaces (Dominguez-Bello et al. 2010). In fact, the percentage of new-borns delivered by Caesarean section world wide is dramatically increasing [32.8% in the USA in 2010 (http://www.cdc.gov/nchs/fastats/delivery.htm)], which may be causally related to an increase in autoimmune diseases and allergies (Neu & Rushing 2011).

Table 4. Environmental factors influencing the microbiota composition (Marques et al. 2010)
Mode of delivery (vaginal or caesarean section)
Status of the immune system
Pharmacological treatments (especially antibiotics)
Physical activity

In addition, the specific microbiota of an individual depends also on its behaviour. Montiel-Castro et al. (2013) have recently argued that animals, apart from co-evolutionary adaptation of the species-dependent microbiota, may have behavioural means of selective colonization. For example, social behaviour in humans and non-human primates, including kissing, grooming and sexual intercourse may serve to enhance horizontal transfer of microbes (Montiel-Castro et al. 2013).

However, not only environmental factors seem to play a role in the establishment and variation of an individual's microbiota. In fact, the microbiota also depends on the species of an individual. Though mice harbour similar phyla of bacteria in their guts, the species distribution is remarkably different in humans and a host-specific microbiota is required for proper immune system maturation (Chung et al. 2012). Moreover, differential microbial compositions have been found for closely related species, even when maintained on the same diet (Brucker & Bordenstein 2012). Hence, also genetic factors must be critically determinants of the specific microbiota in an individual. Further evidence for this hypothesis comes from the fact that the genotype of a different mouse strains is mainly responsible for variations in their gut microbiota (Kovacs et al. 2011) and adoptive transfer of phenotypic strain differences by microbiota transplantations or inter-strain cross-fostering (Bercik et al. 2011a; Collins et al. 2013; Francis et al. 2003). Moreover, ethnic affiliation is correlated with vaginal and oral microbiota composition (Mason et al. 2013; Ravel et al. 2011). Indeed, monozygotic twins seem to have a more similar gut microbiota composition than marital partners or unrelated persons (Zoetendal et al. 2001).

If genetic differences turn out to be commonly responsible for differences in microbiota composition, it would be exciting to find out what exactly these differences are. Studies on Hydra sp. (a cnidarian) suggest specialized anti-microbial peptides may be involved in the process of selective colonization (Franzenburg et al. 2013). In animals with a more sophisticated immune system other molecules, especially molecules that establish the relationship between ‘self’ and ‘non-self’ are likely to be involved. Furthermore, it would be intriguing to see whether there are also differences in the microbiome within a given species that can be attributed to host genetic variation like single-nucleotide polymorphisms (SNPs) or copy-number variations (CNVs) and thereby contribute to gene–environment interactions (G × E, see Box 3).

BOX 3. Gene–environment interactions (G×E)

The concept of G×E is often used to describe the effect of genetic risk factors on the development of certain pathologies that vary with environmental differences. In a broader sense, however, G×E can also be used to describe any effect of genetic variation on a phenotypic trait that is influenced by environmental conditions. G×E manifest when a changing environment leads to differential gene expression. This is because a genetic variation may remain undetectable as long as the gene is not expressed or silenced. Therefore, regulation of gene expression is an important mediator of G×E. As key regulators of gene expression, epigenetic processes integrate signalling of environmental cues on the level of transcriptional and translational regulation and can thereby reveal the effect of polymorphisms by exposing them to function in a given environment.

If the microbiota composition were influenced by genetic variation, in fact, interaction of the host genome and its microbiome would be an excellent demonstration of G × E. To test this, studies, designed to find genome-wide associations between genetic variation and microbiota composition would be needed. Following this logic, it may be possible to define new genetic risk factors with respect to microbial diversity/composition. Thus understanding the temporal dynamics of microbiota composition in complex psychiatric diseases, including ASDs, schizophrenia and depression are needed to give credence to this hypothesis.

The existence of incompatible or adverse genome–microbiome combinations would have significant implications for the success of faecal transplantation treatments and screening of donors and recipients of the transplant. Finally, genotype–enterotype (see Box 1) incompatibility may lie at the heart of enterotypic variation in human populations – especially with ageing since it is associated with substantial reorganizations of the transcriptional profile in virtually all organs and body systems, most prominently the immune system.

The microbiota as a discrete epigenetic entity

The microbiota has been described as a ‘virtual endocrine organ’ (Evans et al. 2013). In fact, the microbiota does not only meet the definition of an endocrine organ but, as outlined in this review, also features all characteristics of an epigenetic instrument, independent of acting through other molecular epigenetic mechanisms as suggested in this article. In the fist instance, the critical role of maternal transmission in the establishment of the offspring's microbiota argues for a view of the microbiota as an epigenetic entity (Fig. 3a).

Figure 3.

The microbiota – a discrete epigenetic entity. (a–d) Parallels between classic epigenetic mechanisms and the functions and characteristics of the microbiota argue for appreciating the microbiota as a discrete epigenetic entity. We thus suggest, that the concept of the hologenome (see Box 1), could be enhanced by the existence of the ‘holo-epigenome’.

As argued before, microbial composition is subject to environmental changes and is likely to mediate gene–environment interactions. These instances constitute another commonality between the microbiota and classic epigenetic mechanisms (Fig. 3b). Furthermore, the microbiota parallels epigenetic determination of gene expression programs in its ability to influence developmental regulation (Fig. 3c). While molecular epigenetic programs govern developmental processes like cellular identity and differentiation, the microbiotic influences on development include immune system maturation, energy metabolism and organ morphogenesis (reviewed by Sommer & Bäckhed 2013)

Finally, reversibility demonstrates another shared characteristic between classic epigenetic mechanisms and microbial colonization (Fig. 3d). As mentioned earlier, several molecular and behavioural phenotypes in studies of GF, infection and probiotic-treated animals were reversible by external intervention. Reversibility of a physiological parameter is indicative of an acquired epigenetic contribution in regulation of this parameter. Irreversibility, however, points to a hard-wired effect on the parameter that is established during development and is not likely reversed by external (e.g. pharmacological, environmental) intervention.

From this point of view, the concept of a ‘hologenome’ (Rosenberg et al. 2007 2009, see Box 1), could be enhanced by the existence of the ‘holo-epigenome’, giving credit to the fact that the diverse microbial genes act as an additional interface for environmental interaction and an dynamic and plastic resource for transgenerational phenotypic regulation, together affecting evolution and development. In fact, the holo-epigenome hypothesis could help to understand some of the fundamental questions in epigenetic and especially neuroepigenetic research (Bohacek et al. 2013; Sweatt 2013) regarding the heritability of experience-driven phenotypic changes. To test whether the brain <−> gut <−> microbe bidirectional communications are a part of a ‘soft-inheritance’ paradigm (Sweatt 2013), careful experimental design is necessary, including cross-fostering and in vitro-fertilization studies.

Friend or foe – symbionts or parasites?

Because of the long history of co-evolution it can be assumed that the intimate connection between a host and its microbiome is likely more extensive than exchanging metabolites that coincidently exhibit endocrine or neuroactive function. In fact, this co-evolutionary interdependence of microbes and their metazoan hosts poses a major challenge in classification of a given host–microbe interaction as simply commensal, parasitic or mutualistic. Yet, in the light of finding new probiotics or microorganisms with respect to psychiatric disease (psychobiotics see Box 1) this is an important question.

While it is debatable whether strict commensalism actually exists, in a truly mutualistic relationship, both species are required to benefit from each other. The benefits for micro-organisms to associate with a mammalian host are rather obvious. For example, the gut microbiota benefits from its host by a constant supply of high-quality nutrients and a relatively high, reaction-facilitating ambient temperature. The advantages of this association for the host, however are less clear. While the gut microbiota provides its host with additional enzymatic capacities to break down host-indigestible diet, this enhanced capacity may come at the price of additional, potentially harmful, side products and metabolites. Since the advantages of the relationship with certain micro-organisms may have outweighed the disadvantages during evolution, negative side effects of the association may be easily overlooked or not perceived as harmful in retrospect. Thus, it remains elusive whether behavioural alterations observed with a differential or absent microbiota are mere side products or represent another beneficial effect on the side of the microbe. For example, if we view the decreased anxiety phenotype of GF mice as the ‘standard behaviour’, and we imply that the microbiota are driven towards supporting their host's–and therefore their own–survival, it stands to reason that the microbiota direct their host to exhibit very specific behaviour. Ultimately we cannot ever fully appreciate whether a given micro-organism is rather symbiotic or parasitic until we know how evolution and development would look without it.

The curious case of behaviour-manipulating parasites

A more non-controversial case of rather unidirectional host–microbe interaction is presented by parasites that remarkably manipulate their host in their very own favour. The scientific literature describes a fascinating variety of micro- and macroparasites that significantly alter host behaviour to complete their sometimes complex life cycles by as yet elusive mechanisms (Berdoy et al. 2000; Cézilly et al. 2010; Libersat et al. 2009; Thomas et al. 2010). In appreciation of this growing field the Journal of Experimental Biology recently (January 2013) devoted a special issue to the topic of ‘neural parasitology’ (Adamo & Webster 2013). One of the best known, and for humans the most relevant, example to date is the protozoan parasite Toxoplasma gondii. Infection with T. gondii is often latent or without serious symptoms and is estimated to affect 25–30% of the world population (Flegr 2013). For replication, it is dependent on its terminal hosts, which are felines including house cats. It has been shown that rodents infected with T. gondii show a remarkable decrease in avoidance of cat urine odour, which known as the ‘fatal attraction phenomenon’ and is interpreted as a host-induced behavioural alteration that increases the chance of successful predation and ingestion by a cat (Berdoy et al. 2000; Vyas et al. 2007; Webster 2001). In humans T. gondii infection can occur by ingestion of uncooked meat or direct contact with cat faeces and is associated with an increased risk for the development of Schizophrenia and altered risk assessment (Flegr 2013; Webster et al. 2013). Interestingly, T. gondii is an intracellular parasite and is well-known for infecting brain tissue (Berenreiterová et al. 2011). Thereby, it has very intimate contact to intracellular neuronal processes and may thus achieve host behaviour alteration by manipulating the host-cell's transcriptional machinery with the use of neuro-nucleomodulins (see Box 1) in similar ways as discussed in this review.


In this review, we present a potential mechanism for host–microbe interaction through means of interacting with molecular epigenetic processes and provide multiple lines of evidence that alterations in microbiota and epigenetic modifications are at least correlated in mediating effects of the microbiota–gut–brain axis. We further suggest that some gut-microbial products can act as ‘neuro-nucleomodulins’ and thereby affect the epigenetic landscape of their host's brain cells which in turn has effects on host behaviour. Such effectors include regulators of enzymatic activity of histone-modifying enzymes by means of metabolic alterations and interactions between bacterial-secreted molecules (such as butyrate) and signalling pathways of the host's neurons that leads to a differential epigenetic landscape. However, at the moment, it is unclear if any of the phenotypes that have emerged in GF mice is due to absence of butyrate and/or other metabolites with epigenetic modifying potential. However, some parasites evidently highjack the host-cell's epigenetic machinery and it might be worthwhile looking for similar effectors in symbiotic or commensal bacteria. Further investigation in this direction is needed to elucidate potential sites for external intervention in experimental settings and in the pursuit of more translatable applications.

Moreover, we argue that the microbiota is an important mediator of gene–environment interaction and propose the existence of gene–microbiota interactions as a special case of G×E. These interactions may lead to genotype–enterotype incompatibilities, which may have critical implications for disease-associated genetic risk factors with unidentified function and may limit the success of faecal transplantation as a tool for systemic treatment in some cases. Borody and colleagues suggested faecal transplants may be extended from the treatment of Clostridium difficile infections to benefit patients suffering from multiple other pathological conditions, including neurodegenerative diseases such as Parkinson's disease (Borody et al. 2000; Smits et al. 2013). In fact, the observed effects may be attributed to epigenetic gene regulation due to altered microbiota composition since drugs targeting the epigenetic machinery are currently investigated for such diseases.

Finally, we reason that the microbiota may even be viewed as an epigenetic entity itself as it exhibits similar features in its interaction with the host as compared to classical epigenetic mechanisms such as histone modifications, DNA methylation and ncRNA-mediated regulation. Thus, the fields of (neuro)epigenetics and microbiology have the potential to converge at many levels and more interdisciplinary studies are necessary to unravel the full range of this interaction.


The authors are supported, in part, by Science Foundation Ireland in the form of a centre grant (Alimentary Pharmabiotic Centre) under Grant Number SFI/12/RC/2273 and by the Health Research Board of Ireland. T.G.D. has until recently been on the Board of Alimentary Health and both T.G.D. and J.F.C. have been on the Speakers Bureau for Mead Johnson.