Mounting evidence suggests a pivotal role of gut microbiota in the aetiology of psychiatric symptoms in stress-related diseases such as anxiety disorders and depression(Reference Rogers, Keating and Young1,Reference Valles-Colomer, Falony and Darzi2) . The mechanisms underlying this microbiota–gut–brain communication are beginning to be unravelled (see(Reference Cryan and Dinan3–Reference Mayer5) for reviews). In particular, certain gut bacteria can have a beneficial effect on mood and emotional behaviour and, as such, have been proposed for potential therapeutic interventions in psychiatry (concept of psychobiotics)(Reference Dinan, Stanton and Cryan6,Reference Sherwin, Rea and Dinan7) . The bidirectional interplay between gut and brain is illustrated in population survey studies revealing a strong correlation between anxiety, depression and functional gastrointestinal (GI) disorders. Furthermore, psychological distress can predict later onset of a functional GI disorder and the converse is also true(Reference Koloski, Jones and Kalantar8).
Early postnatal life is a critical period during which both brain and gut undergo important maturation(Reference Borre, O'Keeffe and Clarke9,Reference Sharon, Sampson and Geschwind10) . Moreover, this maturation is greatly influenced by gut microbiota colonisation and diversification during the lactating period. Exposure to stressful events during childhood has been repeatedly associated with increased vulnerability to both psychiatric and GI disorders such as the irritable bowel syndrome (IBS)(Reference Chitkara, van Tilburg and Blois-Martin11–Reference O'Mahony, Clarke and Dinan13). IBS is defined as a disorder of the gut–brain interaction. According to Rome IV classification, it is characterised by abdominal pain and altered bowel habits(Reference Drossman14), but also increased intestinal permeability and gut dysbiosis. Chronic disruption of the mother–infant relationship in rodents, best known as maternal separation (MS), is a useful preclinical tool since it models the co-morbidity between IBS and psychiatric disorders. Indeed, it induces a wide range of brain and gut alterations in offspring(Reference O'Mahony, Hyland and Dinan15). In the following, we concisely overview the adverse consequences of MS, which is the most used model of early adversity in gut–brain axis research. We then discuss the effects of gut-directed interventions on the microbiota–gut–brain axis, with a particular focus on stress-related behaviours.
The maternal separation model
Pioneering work from Harlow in non-human primates and Levine, Denenberg, Meaney and Plotsky in rodents has shown that the early environment, in particular the quality of maternal care, shapes emotional behaviour as well as stress responsivity in adult life(Reference Denenberg, Ottinger and Stephens16–Reference Seay, Hansen and Harlow20). The work of Hofer also revealed the deleterious impact of early weaning on offspring physiology, including intestinal physiology(Reference Ackerman, Hofer and Weiner21). Since then, a vast body of literature has documented the effects of early mother–infant separations in rats during the first weeks of life (1–3 weeks). The most common MS paradigm consists in daily 3 h separations between postnatal days (PND) 2 and 14(Reference Lippmann, Bress and Nemeroff22). However, there are other models using different separation durations (3–8 h daily) or an acute 24 h separation(Reference Barna, Bálint and Baranyi23–Reference Viveros, Llorente and López-Gallardo26). MS results in different degrees of perceived stress in dams and pups according to the protocol used (litter isolated in the homecage without the mother or litter isolated in a novel environment; pups individually separated or not; undisturbed control or ‘handling’ i.e. short separation episode (15 min)). The different models and their respective effects are reviewed in(Reference Korosi and Baram27–Reference Vetulani29). In any case, pups are deprived of maternal care during the separation period. Importantly, the absence of the dam implies that the pups cannot benefit from dams' heat and milk. Temperature issues can be easily corrected by maintaining the room at 28–29°C during separation sessions. However, the lack of milk intake likely contributes to the short and long-term effects of 24 h MS(Reference Suchecki, Rosenfeld and Levine30,Reference Van Oers, de Kloet and Whelan31) . Mother–infant separation-based models have also been developed in other rodents (e.g. guinea pigs and mice) and in primates (rhesus macaques)(Reference Cirulli, Francia and Berry32), but the largest literature still involves rats, with mice being more and more used; we will focus on these rodent species in the present review. It appears that mice are less sensitive to early-life stress than rats(Reference Tan, Ho and Song33) (see(Reference Millstein and Holmes34) for review). This might be attributable to species specificities in neurodevelopment and maternal care patterns. Another possible reason is that mouse studies more often involve inbred strains (while outbred strains are used in rats) as well as transgenic strains that exhibit different levels of sensitivity to stress(Reference Millstein and Holmes34). To produce significant behavioural alterations in mice, MS is often combined with others stressors such as unpredictable stress in dams(Reference Franklin, Russig and Weiss35,Reference Gapp, Soldado-Magraner and Alvarez-Sánchez36) , early weaning(Reference Tchenio, Lecca and Valentinova37) or a combination of perinatal stressors(Reference Rincel, Aubert and Chevalier38).
Maternal separation and emotional vulnerability
Long-term psychoneuroendocrine alterations
The long-term consequences of MS on emotional behaviour have been extensively documented. Available tools to evaluate emotionality are mostly limited to tests with good predictive validity (i.e. sensitive to anxiolytics or antidepressants) such as the elevated plus maze, open-field or light–dark tests for anxiety and the forced swimming test or tail suspension test for depression. These tests have however a poor construct validity contrary to other tests such as sucrose preference or female urine sniffing tests used to assess reward deficiency as index of anhedonia (see(Reference Cryan and Holmes39–Reference York, Blevins and Baynard41)).
Typically, MS leads to increased anxiety- and depressive-like behaviours. Indeed, animals exposed to MS during early-life display reduced exploration of the open areas in the elevated plus maze, light–dark box and open-field tests compared with non-separated controls(Reference Amini-Khoei, Haghani-Samani and Beigi42–Reference Yang, Cheng and Tang72). Moreover, it has been shown that exposure to a novel stress at adulthood aggravates these anxiety-like behaviours(Reference Eiland and McEwen73,Reference Marais, van Rensburg and van Zyl74) . Numerous studies also report increased depressive-like behaviours in the forced swimming test or tail suspension test. Indeed, adult MS rodents show greater immobility time in these tests compared with controls(Reference Amini-Khoei, Haghani-Samani and Beigi42,Reference Gracia-Rubio, Moscoso-Castro and Pozo53,Reference Lambás-Señas, Mnie-Filali and Certin58,Reference Lee, Kim and Kim59,Reference Liu, Hao and Zhu61,Reference Maniam and Morris62,Reference Portero-Tresserra, Gracia-Rubio and Cantacorps66,Reference Amini-Khoei, Mohammadi-Asl and Amiri75–Reference Yamawaki, Nishida and Harada91) . MS has been associated with decreased sucrose preference(Reference Shu, Xiao and Tang69,Reference Wang, Dong and Wang71,Reference Amini-Khoei, Mohammadi-Asl and Amiri75–Reference Bai, Zhu and Zhang77,Reference Hui, Zhang and Liu79–Reference Li, Yang and Yao81,Reference Masrour, Peeri and Azarbayjani83,Reference Sadeghi, Peeri and Hosseini87,Reference Dallé, Daniels and Mabandla92–Reference Yang, Xu and Zhang96) and decreased social behaviour with a conspecific(Reference Rincel, Lépinay and Delage67,Reference Yang, Cheng and Tang72,Reference Kundakovic, Lim and Gudsnuk93,Reference Farrell, Holland and Shansky97–Reference Zimmerberg and Sageser99) . The effects of MS are not limited to the above alterations of emotional behaviours; numerous studies also report that MS exacerbates motivation for alcohol and drugs of abuse (see(Reference Moffett, Vicentic and Kozel100) for review).
Finally, several studies have also shown deleterious effects of MS on cognition (see(Reference Kosten, Kim and Lee101) for review). Briefly, these effects include impaired hippocampal-dependent spatial learning and memory(Reference Dandi, Kalamari and Touloumi47,Reference Rincel, Lépinay and Delage67,Reference Hui, Zhang and Liu79,Reference Couto, Batalha and Valadas102–Reference Wang, Jiao and Dulawa107) , altered non-spatial memory(Reference Banqueri, Méndez and Arias44,Reference Reshetnikov, Kovner and Lepeshko105,Reference Wang, Jiao and Dulawa107–Reference Wang, Nie and Li115) and impairments in prefrontal cortex (PFC)-dependent tasks (working memory, extinction, cognitive flexibility)(Reference Feifel, Shair and Schmauss50,Reference Yang, Cheng and Tang72,Reference Wang, Jiao and Dulawa107,Reference Baudin, Blot and Verney116–Reference Thomas, Caporale and Wu120) . In contrast, amygdala-dependent aversive memory (e.g. fear conditioning) seems to be enhanced by MS(Reference Wilber, Southwood and Wellman121–Reference Xiong, Yang and Cao127).
Endocrine response and neurobiological correlates
MS exerts long-lasting effects on hypothalamic–pituitary–adrenal (HPA) axis function, leading in most of the studies to endocrine hyper-responsivity to a novel stress(Reference Plotsky and Meaney19,Reference Dandi, Kalamari and Touloumi47,Reference Aisa, Tordera and Lasheras108,Reference Biagini, Pich and Carani128–Reference Slotten, Kalinichev and Hagan135) . Within the central nervous system, this HPA axis hyper-reactivity is associated with an up-regulation of corticotrophin-releasing hormone (CRF) expression in the paraventricular nucleus (PVN) of the hypothalamus and amygdala but also with high CRF concentration and increased CRF receptor density in the locus coeruleus and raphe nucleus(Reference Plotsky and Meaney19,Reference Ladd, Owens and Nemeroff130,Reference Ladd, Huot and Thrivikraman131,Reference de Almeida Magalhães, Correia and de Carvalho136) (see(Reference Rivarola and Renard137)) as well as altered oxytocin and vasopressin expression (either up- or down-regulated) in the PVN (see(Reference Veenema138) for review). MS also decreases glucocorticoid receptor (GR) expression in the hippocampus and PFC(Reference Wilber, Southwood and Wellman121,Reference Rivarola and Suárez139) , two main brain areas involved in HPA axis negative feedback. Numerous neurotransmission systems are affected by MS. MS decreases the number of type A γ-aminobutyric acid (GABA-A) receptors in noradrenergic neurons of the locus coeruleus and in the nucleus tractus solitarius(Reference Caldji, Francis and Sharma46) and hippocampus(Reference Yang, Xu and Zhang96). The gabaergic system plays a role in CRF synthesis inhibition in the central amygdala, allowing a buffering of the noradrenergic response to stress. In addition, MS impairs glutamatergic(Reference Chen, Chen and Tang140–Reference Pickering, Gustafsson and Cebere142), serotonergic(Reference Daniels, Pietersen and Carstens48,Reference Liu, Hao and Zhu61,Reference Masrour, Peeri and Azarbayjani83,Reference Bravo, Dinan and Cryan143–Reference Wu, Ziea and Lao147) , dopaminergic(Reference Moya-Pérez, Perez-Villalba and Benítez-Páez64,Reference Kawakami, Quadros and Machado145,Reference Arborelius and Eklund148–Reference Romano-López, Méndez-Díaz and García153) , opioidergic(Reference Ploj, Roman and Nylander152,Reference Ploj, Roman and Nylander154) and endocannabinoidergic(Reference Portero-Tresserra, Gracia-Rubio and Cantacorps66) transmission. In the central nervous system, serotonin is involved in neuronal development(Reference Daubert and Condron155), emotionality and also pain modulation(Reference Kim and Camilleri156,Reference Sommer157) . Among other effects, MS reduces the expression of the serotonin transporter in the raphe nucleus(Reference Bravo, Dinan and Cryan143). Interestingly, selective serotonin reuptake inhibitor antidepressants such as paroxetine normalise HPA axis function as well as emotional behaviour in MS rats.
MS induces both functional and structural changes in several brain regions including the PFC, hippocampus, amygdala and nucleus accumbens(Reference Li, Xue and Shao150,Reference Chocyk, Bobula and Dudys158–Reference Soztutar, Colak and Ulupinar163) . More specifically, impaired synaptic long-term potentiation, dendritic atrophy as well as reduced dendritic spine density have been reported in the medial PFC and hippocampus of adolescent and adult MS rats(Reference de Melo, de David Antoniazzi and Hossain63,Reference Shin, Han and Woo68,Reference Farrell, Holland and Shansky97,Reference Guo, Liang and Liang104,Reference Baudin, Blot and Verney116,Reference Chen, Chen and Tang140,Reference Romano-López, Méndez-Díaz and García153,Reference Chocyk, Bobula and Dudys158,Reference Bock, Gruss and Becker164–Reference Sousa, Vital and Costenla171) . By contrast, MS induces dendritic hypertrophy in the amygdala(Reference Koe, Ashokan and Mitra57). A recent study reported that mice deficient for motopsin, a serine protease secreted from neuronal cells to induce filopodia, precursor structures of dendritic spines, are resistant to MS-induced increase in anxiety in the open field test(Reference Hidaka, Kashio and Uchigaki111). In addition, it has been shown that MS leads to hypomyelination in the medial PFC(Reference Yang, Cheng and Tang72).
MS is also accompanied by decreased expression of neurotrophins such as nerve growth factor and brain-derived neurotrophic factor (BDNF), that are known to play critical roles in dendrite growth and spinogenesis(Reference Lippmann, Bress and Nemeroff22,Reference Dandi, Kalamari and Touloumi47,Reference Marais, van Rensburg and van Zyl74,Reference Réus, Stringari and Ribeiro85,Reference Aisa, Elizalde and Tordera172,Reference De Lima, Presti-Torres and Vedana173) (see(Reference Park and Poo174) for review). In addition, MS leads to alterations of hippocampal neurogenesis (either decreased or increased) at adulthood(Reference Hulshof, Novati and Sgoifo112,Reference Mirescu, Peters and Gould175–Reference Lajud, Roque and Cajero177) . Interestingly, decreased hippocampal BDNF and neurogenesis are consistent observations in post-mortem brains of depressed subjects and there is mounting evidence that BDNF is involved in emotional vulnerability (see(Reference Autry and Monteggia178) for review).
Peripheral and central inflammation
There is substantial evidence that MS activates inflammatory processes both systemically and within the central nervous system, although the underlying mechanisms remain to be explored. Indeed, increased circulating levels of IL-1β(Reference Wang, Dong and Wang71) and IL-6(Reference Wieck, Andersen and Brenhouse179) have been reported in MS animals. In addition, MS offspring display neuroinflammatory marks such as increased Tnfa, Il-1b and Tlr4 expression or increased reactive oxygen species levels and decreased Il-10 expression in the hippocampus(Reference Wang, Dong and Wang71,Reference Amini-Khoei, Mohammadi-Asl and Amiri75,Reference Sadeghi, Peeri and Hosseini87,Reference Pinheiro, de Lima and Portal114) , PFC(Reference Wang, Dong and Wang71) and PVN(Reference Tang, Zhang and Ji180). Recent studies have shown a decrease in the levels of the astrocytic marker GFAP (glial fibrillary acidic protein) in the PFC of MS animals(Reference Yamawaki, Nishida and Harada91) and the opposite effect in the locus coeruleus of MS females only(Reference Nakamoto, Aizawa and Kinoshita181).
Inconsistencies in the maternal separation literature
A number of studies did not replicate the abovementioned findings, reporting no alteration of certain emotional behaviours(Reference Tan, Ho and Song33,Reference Aya-Ramos, Contreras-Vargas and Rico43–Reference Bondar, Lepeshko and Reshetnikov45,Reference Feifel, Shair and Schmauss50,Reference de Melo, de David Antoniazzi and Hossain63,Reference Rincel, Lépinay and Delage67,Reference Eiland and McEwen73,Reference Marais, van Rensburg and van Zyl74,Reference Sadeghi, Peeri and Hosseini87,Reference Uchida, Hara and Kobayashi89,Reference Øines, Murison and Mrdalj95,Reference Reshetnikov, Kovner and Lepeshko105,Reference Hulshof, Novati and Sgoifo112,Reference de Almeida Magalhães, Correia and de Carvalho136,Reference Ershov, Bondar and Lepeshko182–Reference Zimmerberg and Kajunski192) , cognitive function(Reference Dandi, Kalamari and Touloumi47,Reference Wang, Jiao and Dulawa107,Reference Thomas, Caporale and Wu120,Reference Hill, Klug and Kiss Von Soly185,Reference Pryce, Bettschen and Nanz-Bahr193–Reference Zhu, Wang and Yao197) or HPA axis signalling(Reference Daniels, Pietersen and Carstens48,Reference Hulshof, Novati and Sgoifo112) in male or in female MS animals. In addition, others studies reported opposite effects (e.g. lower anxiety or lower depressive-like behaviour)(Reference Aya-Ramos, Contreras-Vargas and Rico43,Reference Maniam and Morris94,Reference Yang, Xu and Zhang96,Reference Tsuda, Yamaguchi and Nakata98,Reference Reshetnikov, Kovner and Lepeshko105,Reference Furukawa, Tsukahara and Tomita110,Reference Slotten, Kalinichev and Hagan135,Reference Ershov, Bondar and Lepeshko182,Reference Chocyk, Majcher-Maślanka and Przyborowska198–Reference Mourlon, Baudin and Blanc202) .
In some cases, these discrepancies may be attributed to the use of different MS protocols (number of separated pups, separation duration and control group), age of investigation, animal strain and sex, housing conditions (individual or collective cages, light–dark cycle, enrichment), but also other testing protocol issues (e.g. habituation prior testing, brightness, sucrose concentration for the sucrose preference test). Notably, the vast majority of the findings were obtained using males only. However, numerous recent studies report sex-specific behavioural alterations in MS animals.
Nevertheless, differential effects of MS have also been reported in studies using the same MS protocol, age, sex, strain or type of stressor. A recent study suggests that the effects of early adversity (maternal immune activation) depend upon the gut microbiota profile of the dams, in particular the presence of commensal segmented filamentous bacteria (which differs across animal suppliers, i.e. Jackson Laboratories and Taconic Biosciences)(Reference Kim, Kim and Yim203). Therefore, the gut microbiota profile may also influence the susceptibility to MS.
Possible early mechanisms at the origin of maternal separation programming
The mechanisms underlying the long-term effects of MS are not fully understood. Multiple, possibly synergistic effects in both dams and pups have been reported (see(Reference Korosi204) for review).
Mother–infant communication and maternal care
Maternal care is thought to play an important role in brain maturation and later vulnerability to stress. It has been established that rodent pups vocalise in response to isolation (30–90 Hz ultrasounds)(Reference Branchi, Santucci and Alleva205,Reference Hofer, Shair and Brunelli206) and MS has been shown to increase the number of these vocalisations compared with undisturbed pups in several mouse strains(Reference Feifel, Shair and Schmauss50). Because these isolation calls elicit retrieval behaviour in the mother, they are thought to serve mother–pup communication and stimulate maternal care towards their pups(Reference Brunelli, Curley and Gudsnuk207,Reference D'Amato, Scalera and Sarli208) . In the MS model, pups are deprived of maternal care during several consecutive hours, which may constitute a mechanism for the adverse effects of this early-life stress. Indeed, it has been demonstrated that the long-term behavioural effects of acute 24 h MS can be prevented by pup tactile stimulation(Reference Van Oers, de Kloet and Whelan31). Nevertheless, the role of maternal care in the long-term effects of MS remains controversial.
MS also constitutes a potent stressor for the dams. Indeed, it has been reported that this psychological stress induces anxiety and depressive-like behaviours in dams(Reference Aguggia, Suárez and Rivarola209–Reference Maniam and Morris211). As a matter of fact, several studies suggest that dam's perceived stress plays an important role in the effects of separation in the offspring. Interestingly, MS-induced HPA hyper-response to stress in the offspring can be counteracted by providing a foster litter to the dam while its own litter is being separated(Reference Huot, Gonzalez and Ladd212). Furthermore, it has been reported that the offspring of dams with an experience of separation with a previous litter exhibit MS-like fear behaviour without direct exposure to the early stress(Reference Kan, Callaghan and Richardson213).
Endocrine, immune and neurobiological effects of maternal separation in developing pups
The HPA axis is almost silenced during a short window of early postnatal development (i.e. from PND4 to 14(Reference Boersma, Bale and Casanello214–Reference Vázquez217)). This stress hypo-responsive period is characterised by extremely low basal corticosterone levels in the plasma as well as blunted adrenocorticotropic hormone and corticosterone response to stress. Nevertheless, this stress hypo-responsive period is not absolute, since a potent stressor such as MS is able to induce HPA axis activation(Reference Vázquez217–Reference Xu, Qin and Shi220). It has been proposed that stress and immune activation result in a cross-sensitisation of both systems that possibly creates a self-perpetuating cycle contributing to the emergence of the alterations in animals subjected to early stress. Bacterial translocation into the liver and the spleen has been detected after MS in juvenile PND10 rats(Reference Moussaoui, Braniste and Ait-Belgnaoui221). In addition, altered circulating pro-inflammatory IL-1β, IL-6 and TNFα were observed in MS pups(Reference Réus, Fernandes and de Moura86,Reference Pinheiro, de Lima and Portal114,Reference Do Prado, Narahari and Holland118,Reference Wieck, Andersen and Brenhouse179,Reference Roque, Ochoa-Zarzosa and Torner222,Reference Roque, Mesquita and Palha223) . Furthermore, MS juveniles display increased activated microglia in the PFC and hippocampus(Reference Gracia-Rubio, Moscoso-Castro and Pozo53) and decreased number of astrocytes in the same areas(Reference Yamawaki, Nishida and Harada91,Reference Musholt, Cirillo and Cavaliere224,Reference Saavedra, Fenton Navarro and Torner225) along with increased Il-6, Il-1b and Tnfa expression compared with controls(Reference Park, Kim and Kang65,Reference Roque, Ochoa-Zarzosa and Torner222) . Increased microglia numbers and activation patterns have also been recently reported in the nucleus of the solitary tract of MS juveniles(Reference Yamawaki, Nishida and Harada91,Reference Baldy, Fournier and Boisjoly-Villeneuve226) . Interestingly, increased cytokine expression and microglial density have also been reported in the hippocampus of juvenile mice submitted to short MS (15 min) from PND1 to PND21, which led to increased anxiety similar to prolonged MS(Reference Delpech, Wei and Hao227).
Both altered HPA axis activity and neuroinflammation during development have been shown to be deleterious for the immature brain. MS disrupts the normal course of brain development and produces functional and structural alterations including delayed GABA excitatory-to-inhibitory functional switch(Reference Furukawa, Tsukahara and Tomita110), delayed synaptic maturity(Reference Andersen and Teicher228), decreased spine density(Reference Rincel, Lépinay and Janthakhin161) and increased neuronal and glial cell death(Reference Mirescu, Peters and Gould175,Reference Kuma, Miki and Matsumoto229,Reference Zhang, Levine and Dent230) . Altered expression of neurotrophins such as BDNF and nerve growth factor in separated pups could contribute to these effects(Reference Kuma, Miki and Matsumoto229,Reference Cirulli, Alleva and Antonelli231,Reference Roceri, Cirulli and Pessina232) . In addition, MS disturbs the serotonergic system during development. Indeed, reduced expression of the serotonin receptor 5HTr1A in the hippocampus and PFC has been reported in 7-d-old pups (Reference Ohta, Miki and Warita233). A recent study demonstrates that transient juvenile, but not adult, knockdown of orthodenticle homoeobox 2 in the ventral tegmental area mimics early-life stress by increasing stress susceptibility, whereas its overexpression reverses the effects of early-life stress(Reference Peña, Kronman and Walker234). Moreover, developmental decrease of the transcription repressor Rest4 (RE-1 silencing transcription factor 4) in the PFC of pups submitted to MS may play a causal role in the long-term effects of MS(Reference Rincel, Lépinay and Delage67,Reference Uchida, Hara and Kobayashi89) . We recently demonstrated that exposure to a high-fat diet (HFD) during the perinatal period can prevent the long-term MS-associated neurobehavioural alterations, possibly via a protective effect on gene expression in the PFC(Reference Rincel, Lépinay and Delage67). Indeed, perinatal HFD prevented the MS-induced alterations of Rest4, Bdnf and 5HTr1A expression in this brain area. A recent work demonstrated that chemogenetic inhibition of MS-induced neuronal hyperactivity in the lateral habenula of mice aged 35 d attenuates depressive-like behaviours(Reference Tchenio, Lecca and Valentinova37).
Epigenetic changes in maternal-separation offspring
Epigenetic marks are dynamic and highly sensitive to environmental factors; furthermore they can last in time and even be transferred across generations(Reference Bohacek and Mansuy235). As such, they represent a potential mechanism that could underlie the long-term effects of early-life stress(Reference Heim and Binder236–Reference Silberman, Acosta and Zorrilla Zubilete239). Indeed, a number of studies have reported persistent epigenetic marks in the genome of animals submitted to MS (see(Reference Jawahar, Murgatroyd and Harrison240) for review). In particular, changes in DNA methylation of specific regulatory sites in key genes for stress processing such as Crf, Avp, GR or Bdnf in the PVN, hippocampus and PFC of maternally separated animals, have been documented(Reference Zhu, Wang and Yao197,Reference Meaney and Szyf241–Reference Wang, Cattaneo and Ryno245) . It has been shown that administration of a DNA methyltransferase inhibitor prevents the decreased prefrontal Bdnf mRNA expression induced by MS(Reference Roth, Lubin and Funk244). Moreover, DNA methylation in the offspring has been shown to be associated with the level of maternal care(Reference Weaver, Cervoni and Champagne246). Nonetheless, the group of Mansuy provided evidence for epigenetically-mediated transmission of behavioural traits induced by early-life stress across generations irrespective of cross fostering(Reference Weiss, Franklin and Vizi247).
Another major epigenetic process is histone modification, especially acetylation by histone acetyltransferases or deacetylation by histone deacetylases. Histone acetylation patterns as well as histone acetyltransferase and histone deacetylase expressions in the brain are also altered by MS(Reference Pusalkar, Suri and Kelkar248). For instance, MS leads to decreased Bdnf and GR mRNA expressions in the hippocampus, and these effects were accompanied by decreased levels of histone acetylation at their respective promoters(Reference Park, Lee and Seo249,Reference Seo, Ly and Lee250) . Furthermore, a recent study suggests that there is a cross-talk between histone acetylation and DNA methylation(Reference Zhu, Wang and Yao197). Indeed, treatment with a histone deacetylase inhibitor reversed the MS-induced increased DNA methylation in the GR promoter region.
Finally, the possible role of brain miRNA in mediating the long-term effects of MS has been addressed in a few studies. Uchida and colleagues were the first to report changes in expression of several miRNA in the PFC of MS rats(Reference Uchida, Hara and Kobayashi89). Another MS study reported an increase in miR-16 in the hippocampus that was negatively correlated with Bdnf expression in the same brain area and also negatively correlated with sucrose preference(Reference Bai, Zhu and Zhang77).
Maternal separation as a model of irritable bowel syndrome: impact on the gastrointestinal tract
As mentioned earlier, MS is also widely used as a model of IBS (see(Reference O'Mahony, Hyland and Dinan15,Reference Barreau, Ferrier and Fioramonti251,Reference Moloney, Johnson and O'Mahony252) for reviews). In addition to its effects on stress vulnerability, it leads to several GI dysfunctions, in particular increased visceral sensitivity to painful stimuli, and increases the vulnerability to experimental colitis.
Effects of maternal separation on the enteric nervous system, visceral sensitivity and motility
MS induces dynamic structural and functional changes in the enteric nervous system(Reference Barreau, Salvador-Cartier and Houdeau253,Reference Tominaga, Fujikawa and Tanaka254) . For instance, MS increases nerve density and synaptogenesis in juveniles, but these effects are no longer present at adulthood(Reference Barreau, Salvador-Cartier and Houdeau253). In contrast, the levels of the neuronal marker PGP 9·5 (anti-protein gene product 9·5) in the colon are increased in adult MS animals but not in juveniles. Interestingly, early-life adversity has been shown to affect enteric nervous system development in a sex-dependent manner, with females being more sensitive than males(Reference Million and Larauche255). MS also produces increased intestinal motility in response to stress, as evidenced by reduced total transit time and increased number of faecal pellets(Reference Li, Yang and Yao81,Reference Hyland, O'Mahony and O'Malley256–Reference Yi, Zhang and Sun260) . It has been extensively reported that MS rats display visceral hyperalgesia during colorectal distension(Reference Felice, Gibney and Gosselin51,Reference Rincel, Lépinay and Delage67,Reference Chen, Chen and Tang140,Reference Chen, Huang and Xu144,Reference O'Mahony, Chua and Quigley146,Reference Wu, Ziea and Lao147,Reference Tang, Zhang and Ji180,Reference Xu, Qin and Shi220,Reference Hyland, O'Mahony and O'Malley256,Reference Moloney, Stilling and Dinan257,Reference Schwetz, McRoberts and Coutinho259–Reference Zhang, Li and Leung283) . A recent study demonstrated that MS-induced visceral hypersensitivity is dependent on Paneth cell defects and associated Escherichia coli expansion in the gut(Reference Riba, Olier and Lacroix-Lamandé284). MS-induced visceral hypersensitivity is lost in mice deficient for Toll-like receptor 4 (TLR4)(Reference Tang, Zhang and Ji180). This study suggests that TLR4 signalling in the PVN mediates increased CRF immunostaining and visceral hypersensitivity associated with MS. Interestingly, multiple MS-induced intestinal phenotypes, including visceral hyperalgesia and gut leakiness, can be prevented by CRF receptor antagonist administration(Reference Tang, Zhang and Ji180,Reference Schwetz, McRoberts and Coutinho259,Reference Barreau, Cartier and Leveque285–Reference Million, Wang and Wang287) . GR antagonists or agonists of the metabotropic glutamate receptor type 7 (mGluR7) also prevent stress-induced visceral hyperalgesia(Reference Shao, Liu and Xiao278,Reference Myers and Greenwood-Van Meerveld288–Reference Zhou, Sha and Huang290) .
The hyper-sensitivity to colorectal distension after MS is larger in females than in males and visceral hyperalgesia is greater when all pups are separated from the dam than when only half of littermates is removed, suggesting that sex and dam's perceived stress play a role in the long-term effects of MS on visceral sensitivity(Reference Rosztóczy, Fioramonti and Jármay277). Indeed, it has been demonstrated that MS-induced visceral hypersensitivity is transferred across generations and that this effect likely depends upon maternal care(Reference Van den Wijngaard, Stanisor and van Diest291).
Effects of maternal separation on gut microbiota composition
A growing number of studies have reported altered gut microbiota composition in MS animals. However, the use of different species, strains, sex, MS protocols, nature of the sample, microbiota analysis method and age of investigation renders between-studies comparisons difficult, and yet, there is no clear microbial pattern associated with MS.
The first study that has investigated the effects of MS on the gut microbiota was carried out by Bailey and Coe in rhesus monkeys(Reference Bailey and Coe292). The authors investigated the stability of gut microbiota 3 d after separation and found a significant decrease in faecal bacteria, in particular from the Lactobacillus genus. A few years later, O'Mahony and colleagues reported overall reduced bacterial diversity in MS rats v. controls(Reference O'Mahony, Marchesi and Scully293). This finding has been replicated in more recent studies(Reference Qian, Lu and Huang294,Reference Zhou, Li and Li295) . However, another recent study reports no change in diversity(Reference Moya-Pérez, Perez-Villalba and Benítez-Páez64). Qualitatively, MS was shown to increase the Firmicutes:Bacteroidetes ratio at the phylum level in some studies(Reference De Palma, Blennerhassett and Lu49,Reference Li, Lee and Filler286,Reference Zhou, Li and Li295,Reference El Aidy, Ramsteijn and Dini-Andreote296) , but again this finding is not consistent across studies as some report opposite(Reference Pusceddu, El Aidy and Crispie297) or no effects(Reference Zhou, Li and Li295). A consistent finding, however, is that the effects of MS on microbiota composition vary both qualitatively and quantitatively with respect to the age of investigation. Indeed, several studies comparing at least two time points show completely different patterns(Reference Moya-Pérez, Perez-Villalba and Benítez-Páez64,Reference Zhou, Li and Li295,Reference Barouei, Moussavi and Hodgson298,Reference García-Ródenas, Bergonzelli and Nutten299) . Overall, Bacteroides and Lachnospiraceae (including Clostridium XIVa) species seem to be consistently altered (either enriched or depleted) across several studies(Reference De Palma, Blennerhassett and Lu49,Reference Murakami, Kamada and Mizushima258,Reference Zhou, Li and Li295,Reference Ilchmann-Diounou, Olier and Lencina300) . Interestingly, it has been shown that changes in several bacterial taxa after MS are abrogated by adrenalectomy, suggesting that corticosterone signalling in response to stress is responsible for at least part of its effects on the microbiota(Reference Amini-Khoei, Haghani-Samani and Beigi42). More studies using global 16S-sequencing approaches are needed to better document the effects of MS on gut microbiota and potentially identify candidate species or genera associated with the behavioural effects of MS. Furthermore, considering the importance of sex differences in both stress effects and basal gut microbiota composition, more studies should be conducted in both males and females(Reference Rincel, Aubert and Chevalier38,Reference El Aidy, Ramsteijn and Dini-Andreote296) .
Effects of maternal separation on the gut mucosa
MS has been associated with alterations in the differentiation and distribution of enteroendocrine cells in the gut epithelium(Reference Estienne, Claustre and Clain-Gardechaux301) and a defect in Paneth cells(Reference Riba, Olier and Lacroix-Lamandé276,Reference Riba, Olier and Lacroix-Lamandé284) . Notably, the numbers of enterochromaffin cells in the colon are increased in MS animals compared with controls(Reference Bian, Zhang and Han264,Reference Bian, Qin and Tian265,Reference Ren, Wu and Yew275) . Accordingly, MS animals exhibit substantial increases in the levels of circulating and colonic serotonin (mainly produced by enterochromaffin cells)(Reference Chen, Huang and Xu144,Reference Wu, Ziea and Lao147,Reference Bian, Zhang and Han264,Reference Bian, Qin and Tian265,Reference Ren, Wu and Yew275,Reference Yang, Xian and Ip282) .
In addition, MS animals were shown to display colonic tissue damage including decreased crypt length and altered number of goblet cells and are more engaged in epithelial cell proliferation(Reference Barreau, Ferrier and Fioramonti262,Reference Li, Lee and Filler286,Reference Li, Zani and Lee302–Reference O'Malley, Julio-Pieper and Gibney304) . Moreover, MS rats show more colonic damage after dextran sulphate sodium or 2,4,6-trinitrobenzenesulphonic acid-induced colitis than non-stressed animals and as a result, they also lose more weight, indicating that they are more sensitive to experimental colitis(Reference Ghia, Blennerhassett and Collins305–Reference Veenema, Reber and Selch307). There is mounting evidence that MS produces long-term gut paracellular and transcellular hyper-permeability to ions and macromolecules(Reference Li, Yang and Yao81,Reference Barreau, Ferrier and Fioramonti262,Reference Miquel, Martín and Lashermes272,Reference Riba, Olier and Lacroix-Lamandé276,Reference García-Ródenas, Bergonzelli and Nutten299,Reference Li, Zani and Lee302,Reference Varghese, Verdú and Bercik306,Reference Barreau, Cartier and Ferrier308–Reference Söderholm, Yates and Gareau311) . Remarkably, stress-induced intestinal hyperpermeability appears to be glucocorticoid-dependent, as it is evoked by the synthetic glucocorticoid dexamethasone and prevented by administration of a GR antagonist, similarly to an inhibitor of the myosin light chain kinase controlling epithelial cytoskeleton contraction(Reference Moussaoui, Braniste and Ait-Belgnaoui221). In addition, exposure to a novel stress at adulthood potentiates gut hyperpermeability in maternally separated rats(Reference Øines, Murison and Mrdalj95,Reference Söderholm, Yates and Gareau311) . Furthermore, it has been shown that acute MS induces immediate passage of macromolecules across the colonic mucosa and can lead to increased number of bacterial cells penetrating the gut epithelium(Reference Moussaoui, Braniste and Ait-Belgnaoui221,Reference Barreau, Ferrier and Fioramonti262,Reference Gareau, Jury and Yang312) .
MS also produces several immune alterations in the colon. Indeed, MS animals show an infiltration of immune cells (i.e. polymorphonuclear neutrophils)(Reference Barreau, Ferrier and Fioramonti262,Reference Ghia, Blennerhassett and Collins305) and an increase in mucosal mast cell density(Reference Li, Yang and Yao81,Reference Barreau, Salvador-Cartier and Houdeau253,Reference Barreau, Ferrier and Fioramonti262,Reference Hyland, Julio-Pieper and O'Mahony271,Reference Barreau, Cartier and Ferrier308) . MS also increases the expression of numerous cytokines including IL-6, IL-1β, TNFα, IFNγ, IL-4, IL-2 and IL-22 in the colonic mucosa(Reference Amini-Khoei, Haghani-Samani and Beigi42,Reference Moya-Pérez, Perez-Villalba and Benítez-Páez64,Reference Barreau, Ferrier and Fioramonti262,Reference Distrutti, Cipriani and Mencarelli268,Reference Riba, Olier and Lacroix-Lamandé276,Reference Shao, Liu and Xiao278,Reference Li, Lee and Filler286,Reference Li, Lee and Martin303,Reference Ghia, Blennerhassett and Collins305,Reference Barouei, Moussavi and Hodgson313) . Increased IFNγ and decreased IL-10 expression were prevented by mGluR7 agonist administration(Reference Shao, Liu and Xiao278) in MS animals.
It has been previously shown that MS increases IFNγ and TNF secretion by mesenteric lymph node cells(Reference Veenema, Reber and Selch307). In addition, increased mRNA expression of TLR3, 4 and 5 has been reported in the colonic mucosa of MS adult rats(Reference McKernan, Nolan and Brint314).
Impact of nutrition and microbiota-directed interventions in maternal separation offspring
An early study using the 24 h maternal deprivation paradigm suggested that feeding the pups during separation could prevent its effects on the HPA axis(Reference Van Oers, de Kloet and Whelan31). In the past decade, a growing number of studies have demonstrated that nutrition can modulate the long-term effects of early-life stress on brain and behaviour, although the underlying mechanisms remain unknown. Recent evidence suggests that the direct impact of nutrition on gut physiology and microbiota could counteract the stress-induced disruption of gut homoeostasis and promote a new state of equilibrium.
Nutritional strategies and maternal separation
Choline and vitamins
Several studies demonstrate a preventive effect of dietary choline and other vitamins in animals submitted to MS, and suggest that early nutritional interventions (before adulthood) have the strongest impact. In one study, the maternal diet was enriched with a mixture of essential C1 metabolism-associated micronutrients containing choline, betaine, methionine, folic acid, zinc, vitamins B6 and B12 during the course of MS. This treatment fully prevented the increased plasma corticosterone levels in MS pups at PND9 and further prevented later alterations of object recognition memory, but not spatial memory in adult MS mice(Reference Naninck, Oosterink and Yam315). In another study, dietary choline exposure from weaning to adulthood attenuated object recognition impairments in MS male rats(Reference Moreno Gudiño, Carías Picón and de Brugada Sauras113). In contrast, supplementation with a cocktail of methyl donors (choline, betaine, folic acid and vitamin B12) in adult maternally separated female rats failed to reverse the deleterious effect of MS on object recognition memory, but did prevent depressive-like behaviour in the forced swim test(Reference Paternain, Martisova and Campión84).
Some evidence suggests that n-3 PUFA deficiency potentiates the effects of MS. For instance, dietary n-3 PUFA deficiency acts in synergy with MS to increase sucrose consumption in adulthood, an effect prevented by desipramine(Reference Ferreira, Bernardi and Krolow184,Reference Mathieu, Denis and Lavialle316) . It was further shown that the same dietary intervention also exacerbates MS-induced anxiety in the open-field test(Reference Mathieu, Oualian and Denis317).
Conversely, it has been reported that supplementation with either n-3, folic acid or n-acetylcysteine during peri-adolescence could prevent the MS-induced depressive-like behaviour in the forced swim test, likely through antioxidant effects within the brain(Reference Réus, Maciel and Abelaira318). Interestingly, supplementation with a mixture of EPA and DHA from adolescence onwards reverses MS-induced gut-microbiota dysbiosis in adult female rats(Reference Pusceddu, El Aidy and Crispie297). However, there was no major effect of the same treatment on anxiety and depressive-like behaviours or cognition in MS animals(Reference Pusceddu, Kelly and Ariffin319), yet no effect of MS per se was observed in this study. Nevertheless, in another study, dietary supplementation with PUFA-rich tuna oil failed to affect long-term visceral hypersensitivity in MS rats, but the diet was only administered after the induction of visceral hypersensitivity by acute stress(Reference van Diest, van den Elsen and Klok320).
Previous studies have shown that palatable food consumption in adulthood can attenuate the deleterious effects of MS on anxiety and depressive-like behaviours and basal corticosterone levels(Reference Maniam and Morris62,Reference Maniam and Morris94) .
We reported that the long-term effects of MS on anxiety, social behaviour and stress endocrine response, but also visceral sensitivity, can be prevented by exposing the dams to HFD during gestation and lactation(Reference Rincel, Lépinay and Delage67). In addition to this protective effect of perinatal HFD in adult animals, we found similar beneficial effects on the developing brain(Reference Rincel, Lépinay and Janthakhin161). Indeed, maternal HFD exposure attenuated the stress-induced changes in mRNA expression of key genes involved in neuronal maturation and structural plasticity in the PFC of PND10 pups. The mechanisms underlying this protective effect of maternal HFD are elusive. We provided evidence that a comfort food effect of HFD in stressed mothers but also a modulation of the gut microbiota and/or gut barrier function by HFD in pups could contribute to its effects on brain and emotional behaviour(Reference Rincel, Lépinay and Delage67,Reference Rincel, Lépinay and Janthakhin161) .
Microbiota-directed interventions and maternal separation
The gut microbiota is highly sensitive to the environment and alterations of its composition (dysbiosis) have been described under conditions ranging from IBS and obesity to depression and autism(Reference Collins321–Reference Zhao324). In particular, early-life environment, including diet and stressful experience, shapes the gut microbiota towards health and disease later in life(Reference Dong and Gupta325). However, the mechanisms underlying the ability of stress to modulate microbiota composition remain to be unravelled. Moreover, it is unclear whether dysbiosis is a causative factor in the aetiology of the abovementioned pathologies. Interestingly, studies using different, but complementary, gut microbiota-directed interventions (germ-free (GF) rodents, antibiotics, faecal microbiota transplantation, probiotics and prebiotics) have demonstrated that gut bacteria can have a beneficial effect on emotional behaviours and, as such, psychobiotics have been proposed for potential therapeutic interventions(Reference Dinan, Stanton and Cryan6,Reference Sherwin, Rea and Dinan7) .
Germ-free animals and microbiota transplantation experiments
The study of GF (or axenic) animals served as a proof of concept for the role of gut microbiota in the regulation of brain function and behaviour. A large number of studies have explored GF-associated alterations both in the gut and the brain (see(Reference Luczynski, McVey Neufeld and Oriach326) for review).
Interestingly, many of the GF phenotypes are normalised by colonisation, although the effects largely depend upon the age of colonisation and the animal species and strain(Reference Braniste, Al-Asmakh and Kowal327–Reference Sudo, Chida and Aiba332). Accordingly, Sudo et al. reported the first evidence that colonisation during early development, but not at a later age, could attenuate the increased HPA axis response to stress in GF mice(Reference Wang and Kasper323). In line with this study, further showed that locomotor hyperactivity in GF mice could be reversed by colonisation early in life, whereas colonisation at adulthood had no effect(Reference Heijtz, Wang and Anuar333).
A landmark study by De Palma and colleagues using GF mice exposed to MS revealed that the microbiota is necessary for the long-term effects of MS. Indeed, early-life stress fails to induce long-term endocrine and behavioural alterations in GF mice compared with SPF (specific-pathogen-free) controls(Reference De Palma, Blennerhassett and Lu49). Interestingly, colonisation with the gut microbiota of a conventional SPF control mouse unmasked the effects of early-life stress in GF mice. However, colonisation with the microbiota of an early-stressed animal did not transfer the stress-associated behavioural phenotype in naive GF mice, suggesting that gut bacteria are necessary but not sufficient to mediate the behavioural effects of early-life stress. Although the authors did not measure intestinal permeability, gut leakiness associated with MS could also contribute to the deleterious effects of MS on behaviour. An important limitation of the GF animal model is that the GF status is not specific at all to intestinal microbes. Previous studies suggest that maternal vaginal microbiota also impacts offspring neurodevelopment(Reference Jašarević, Howard and Morrison334). Furthermore, GF animals are housed in isolators with limited handling that constitutes a stressful environment and in most of the study the control groups are not housed in similar isolators and thus are not comparable. Together, these studies suggest that gut dysbiosis may be responsible for some, but not all of the MS-associated phenotypes later in life.
The important role of gut microbiota in the regulation of behaviour was further confirmed by demonstrating the successful adoptive transfer of host behavioural phenotype between mice of different strains and with different behavioural profiles (see(Reference Collins, Kassam and Bercik335) for review). In animals, faecal transplantation can be achieved by oral gavage of fresh faecal content or by transient co-housing with the donor. The stability of the transplanted microbiota can vary depending upon several factors (strain, sex, age, housing conditions). The first evidence of gut–brain effects following faecal transplantation in animals showed a critical role of gut microbiota in host metabolism and energy balance(Reference Turnbaugh, Bäckhed and Fulton336,Reference Wang, Koonen and Hofker337) . Since then, accumulating data have demonstrated that faecal transplantation can affect brain and behaviour in rodents. For instance, social deficits in offspring from HFD-fed dams could be reversed by co-housing with offspring from dams fed a regular diet(Reference Buffington, Di Prisco and Auchtung338), an effect accompanied by restored synaptic plasticity in the brain following social interaction.
Conversely, it has recently been shown that naive rats receiving faecal microbiota from MS donors displayed MS-like intestinal hypermotility(Reference Zhao, Huang and Lu339). Interestingly, colonisation with the microbiota of IBS patients v. healthy controls recapitulated several features of IBS in GF mice, including faster GI transit, intestinal barrier dysfunction, innate immune activation, but also anxiety-like behaviour(Reference De Palma, Lynch and Lu340).
The faecal mycobiome of MS rats is altered relative to control animals(Reference Botschuijver, Roeselers and Levin341). Furthermore, fungicide treatment in adult MS rats prevented the visceral hypersensitivity induced by water avoidance stress. Strikingly, transplantation of the microbiota from MS rats could re-establish visceral hypersensitivity in the absence of water avoidance stress, an effect that was absent when the donor microbiota came from fungicide-treated rats. These findings highlight the need of considering exhaustively gut microbiota composition (i.e. bacteria but also viruses and fungi) and of better understanding the complex interactions between stress and gut microbes.
Overall, the potential clinical value of faecal transplantation for the treatment of disorders of the gut–brain axis is promising(Reference Borody and Khoruts342,Reference Brandt and Aroniadis343) and currently represents an active area of research. To date, the only indication for faecal transplantation in human subjects is the treatment of severe infections with Clostridium difficile, resulting in high success rates(Reference Khoruts344). In the recent years, two double-blind, placebo-controlled, randomised trials have investigated the impact of faecal microbiota transplantation in IBS patients(Reference Halkjær, Christensen and Lo345,Reference Johnsen, Hilpüsch and Cavanagh346) . However, evidence for clinical improvement of GI symptoms and psychiatric symptoms is unclear and has to be further established in larger studies.
The term probiotic, defined as ‘a live microbial feed supplement, which beneficially affects the host by improving its intestinal microbial balance’ was coined in 1953 by Werner Kollath to contrast with antibiotics(Reference Eberl347). The use of probiotics in animal studies has provided evidence that the gut microbiota posesses psychobiotic properties (i.e. antidepressant and/or anxiolytic-like activity) (see(Reference Dinan, Stanton and Cryan6,Reference Joseph and Law348,Reference Sarkar, Lehto and Harty349) for reviews).
Probiotic interventions are generally restricted to one or few bacterial species, thereby allowing the association between a given bug and a particular behavioural effect. The most used are members of the Bifidobacterium and Lactobacillus genera. Beneficial effects of probiotics have been reported in paradigms involving early-life stress. Several studies have shown anti-nociceptive effects of different probiotics (i.e. Faecalibacterium prausnitzii; Bifidobacterium breve or VSL#3)(Reference Barrett, Fitzgerald and Dinan263,Reference Distrutti, Cipriani and Mencarelli268,Reference Miquel, Martín and Lashermes272) . Moreover, the probiotic Bifidobacterium infantis chronically administered at adulthood (from PND50 to PND95) was reported to exert antidepressant-like effects in animals exposed to MS(Reference Desbonnet, Garrett and Clarke78). In addition, the increased peripheral levels of the proinflammatory cytokine IL-6 as well as the increased CRF mRNA levels in the amygdala in stressed animals were also normalised. Similarly, a lactobacillus strain, Lactobacillus plantarum PS128, has antidepressant-like effects in MS mice treated from weaning onwards in both the sucrose preference test and forced-swim test, but has no effect on MS anxiety(Reference Liu, Liu and Wu350). Moreover, serum increase in corticosterone (both at baseline and in response to stress), increase in IL-6 and decrease in IL-10 were all reversed by the probiotic. In addition to the beneficial effects of probiotics in adult animals, an increasing body of evidence shows that probiotics supplementation during early-life can have long-term preventive effects. Indeed, it has been shown that a mixture of Lactobacillus rhamnosus and Lactobacillus helveticus could prevent the elevation in basal plasma corticosterone observed in MS juvenile rats (PND20), in addition to mitigating the associated increased gut permeability(Reference Gareau, Jury and Perdue309). Similar findings have been reported in a mouse model of MS where mice received the probiotic Bifidobacterium pseudocatenulatum during the perinatal period(Reference Moya-Pérez, Perez-Villalba and Benítez-Páez64). Compared with their placebo-fed stressed counterparts, probiotic-fed mice exposed to early stress showed attenuated HPA axis reactivity and intestinal inflammation at weaning, as well as lower anxiety levels during adolescence. These findings were extended to other probiotic strains belonging to Bifidobacteria and Lactobacilli. Indeed, in juvenile rats, MS-induced hypercorticosteronaemia, intestinal hyper permeability and dysbiosis were all prevented by neonatal treatment with Bifidobacterium bifidum G9-1(Reference Fukui, Oshima and Tanaka351). Pretreatment with L. fermentum CECT 5716 was also able to attenuate the effects of a single 4 h-separation episode at PND10 (i.e. hypercorticosteronaemia and intestinal hyperpermeability)(Reference Vanhaecke, Aubert and Grohard352). In contrast, although maternal probiotics treatment with Bifidobacterium animalis subsp. lactis BB-12H and Propionibacterium jensenii 702 was shown to prevent the increase in plasma IFNγ in adult MS offspring(Reference Pusceddu, Kelly and Ariffin319), the same treatment increased plasma IL-6 in juveniles. In line with the latter, Barouei and collaborators showed that the maternal probiotic intervention induces MS-like dysbiosis along with increased levels of circulating corticosterone and adrenocorticotropic hormone in non-stressed developing offspring(Reference Barouei, Moussavi and Hodgson298). A recent study also reports preventive effects of maternal treatment with the probiotic Lacidofil® (L. rhamnosus R0011 and L. helveticus R0052), via the maternal drinking-water during the period of stress, on abnormal mPFC neural fear circuitry development in stressed pups(Reference Cowan, Stylianakis and Richardson353). Interestingly, the effects of neonatal probiotics in MS models are not restricted to stress response and depressive-like behaviours. It has been reported that MS disturbs puberty onset in a sex-dependent manner, but this effect is prevented by probiotic neonatal administration with Lacidofil®(Reference Cowan and Richardson354). In another study, MS rats transmitted their conditioned aversive memory to the next generation, but this effect was abolished if the F0 fathers or the F1 offspring was supplemented with Lacidofil®(Reference Callaghan, Cowan and Richardson355).
Notably, the majority of these findings were obtained using males only. Since a consistent gender effect has been reported in the prevalence of anxiety and depression, but also IBS, with higher rates in women than men(Reference Steel, Marnane and Iranpour356), additional preclinical studies using female animals are required. Moreover, the translational potential of these findings is currently limited by methodological and technical issues. It is not clear whether probiotic strains survive under aerobic conditions and are able to efficiently colonise the gut of the recipient. Future studies should systematically assess post-treatment colonisation to draw conclusions. In this regard, comparing heat-killed v. live probiotics can also be helpful to better understand their underlying mechanisms of action. Furthermore, there is a need to improve the dosage, treatment duration and route of administration. Indeed, only a few studies in animal models addressed the dose-dependency of the effects of probiotics. It is suggested that multi-strain probiotic combinations may provide greater health benefits, but this hypothesis has not been clearly tested. The systematic comparison of the effects of probiotics with that of a clinical drug such as anxiolytics or antidepressants appears critical to quantify the benefits.
It has been proposed that probiotics might represent an adjuvant therapy in psychiatric disorders including major depressive disorder, although well-designed clinical trials are needed to make clear conclusions(Reference Vlainić, Šuran and Vlainić357). A recent study reported that pregnant women supplemented with L. rhamnosus until 6 months postpartum had significantly lower depression and anxiety scores in the postpartum period(Reference Slykerman, Hood and Wickens358). To date, antidepressant effects of probiotics have been reported in three double-blind studies conducted in subjects diagnosed with significant anxiety or depression symptoms(Reference Majeed, Nagabhushanam and Arumugam359,Reference Romijn, Rucklidge and Kuijer360) and in major depressive disorder patients(Reference Akkasheh, Kashani-Poor and Tajabadi-Ebrahimi361). However, based on preclinical data, psychotropic-like effects of probiotics on mood and anxiety in subjects exposed to early-life adversity still need to be confirmed in human trials(Reference Rackers, Thomas and Williamson362). One study has explored the effects of probiotic strains L. rhamnosus HN001 or B. animalis subsp. lactis HN019 in 11-year-old children supplemented from fetal life to age 2 years on neurodevelopment, but found no major effect of probiotics(Reference Slykerman, Kang and Van Zyl363); yet the impact on emotional behaviours and especially in early-stressed patients remain unknown.
Prebiotics and symbiotics
Prebiotics are nutrients that can be fermented by microbes in the gut and thus favour the growth of certain microbial communities(Reference Gibson and Roberfroid364). In comparison with probiotics, a much smaller number of studies have examined the effects of prebiotics on behaviour (see(Reference Kao, Harty and Burnet365) for review). These include investigations of galacto-oligosaccharides (GOS) and fructo-oligosaccharides (FOS), which are sources of nutrition for Bifidobacteria and Lactobacilli. The effects of FOS and GOS have been tested in C57BL/6J male mice in basal and chronic stress situations(Reference Burokas, Arboleya and Moloney366). GOS of FOS alone showed some levels of protective effects but to a much lower extent compared with GOS and FOS. Conversely, human-milk oligosaccharide prebiotics have been reported to impact brain development and cognitive functions(Reference Wang and Brand-Miller367,Reference Wang, McVeagh and Petocz368) . Mice supplemented with human-milk oligosaccharides in their diet (2 weeks) were protected against stress-induced hyperanxiety(Reference Tarr, Galley and Fisher369). Apart from these effects on emotional behaviours, other studies have reported improved learning and memory performance in animals supplemented with different oligosaccharides including human-milk oligosaccharides(Reference Jia, Lu and Gao370–Reference Yen, Wang and Wu373). Together, these findings suggest that combining several probiotics and/or prebiotics can improve the treatment outcome. For instance, increased intestinal permeability in adolescent MS rats was prevented by a symbiotic diet containing arachidonic acid and DHA, GOS and FOS and Lactobacillus paracasei NCC2461 (Reference García-Ródenas, Bergonzelli and Nutten299). In another study, MS rats were treated with either the prebiotics polydextrose and GOS, the probiotic L. rhamnosus GG or the symbiotic combination from weaning onwards(Reference McVey Neufeld, O'Mahony and Hoban374). Only the combination of pre- and probiotics was able to normalise anxiety in the open field test, although it impaired corticosterone negative feedback following acute restraint stress. In addition, expression of GABA receptor A2 (Gabra2) in the hippocampus was restored only by the combination of pre- and probiotics, whereas expression of GR (Nr3c1) was restored by L. rhamnosus GG alone.
MS induces a variety of long-term alterations similar to that observed in human subjects with a history of childhood adversity. In this review, we have outlined the specific effects of MS on both the brain and the gut, illustrating the validity of this model with respect to clinical data. In addition, the pivotal role played by gut microbiota in mediating the lasting imprinting by MS is highlighted in numerous studies using microbiota-directed interventions such as probiotics treatments. Preclinical studies suggest that nutritional approaches with pro- and prebiotics may constitute safe and efficient strategies to attenuate the effects of early-life stress on the gut–brain axis. However, it is still not clear whether gut dysbiosis, leakiness or inflammation precede each other and if they are the cause or consequence of stress-induced alterations within the brain. In this respect, studies are needed to understand how chronic neonatal stress disrupts gut–brain homoeostasis during development and which molecular mechanisms underlie the subsequent long-term imprinting. Moreover, despite widespread sex differences in both GI and neuropsychiatric vulnerability, there is still a gap to fill in the literature as regards the issue of sex. Meta-analyses on the impact of probiotics on anxiety and depressive-like symptoms exist, but the vast majority of the studies are conducted in healthy subjects and recent findings demonstrate that the effects of probiotics may differ between stressed and unstressed subjects(Reference Papalini, Michels and Kohn375). Future studies should develop nutritional strategies combining multiple prebiotics and probiotics, in addition to usual pharmacological strategies, to examine their impact at adult age on symptoms associated with early-life adversity using randomised placebo-control trials, with an effort to adapt these strategies according to sex(Reference Mayer and Hsiao376). Furthermore, prebiotics and probiotics effects should also be examined during development in populations exposed to stress. In human trials, it would be particularly valuable to study the potential preventive effects of prebiotics and probiotics after different stress experiences such as early-life traumas, but also parental depression, perinatal infections, premature birth or low parental socioeconomic status. Finally, early-life adversity is associated with poor diet quality at adulthood(Reference Gavrieli, Farr and Davis377). In this context, it would be crucial to improve health policies and to implement preventive interventions with nutritional advices in populations exposed to early-life adversity.
The authors were supported by the University of Bordeaux and INRA. M. R. was supported by the French ministry of research and education and Labex Brain Bordeaux Region Aquitaine Initiative for Neurosciences (PhD extension grant).
Conflict of Interest
M. R. drafted the first version of the manuscript then both authors revised and approved the manuscript.