Effects of Oral Administration of the Probiotic Lactobacillus rhamnosus GG on the Proteomic Profiles of Cerebrospinal Fluid and Immunoregulatory Signaling in the Hippocampus of Adult Male Rats

In recent years, research has highlighted a potential role for the bidirectional microbiome-gut-brain axis in control of immune system function [1], metabolism [2], host behavior [3, 4], and even cognition [5‒9]. The “Old Friends” hypothesis proposes that urbanization leads to a reduction in exposures to microorganisms that help regulate the immune system [10, 11]. This hypothesis predicts that the increase in inflammatory disorders in modern societies could be mitigated by the reintroduction of “Old Friends,” i.e., microorganisms with anti-inflammatory and immunoregulatory properties [12, 13].

While the diversity and community composition of the gut microbiota shows considerable interindividual variation, dysbiosis of the gut microbiome can lead to several health conditions that involve chronic low-grade inflammation [14, 15]. Stress-related psychiatric disorders are commonly associated with chronic low-grade inflammation and altered metabolic function [16‒21]. In addition, age-related decline in the diversity and stability of the gut microbiota may be responsible for a chronic state of inflammation seen in elderly persons with dementia, including neurodegenerative disorders such as Alzheimer’s disease (AD) [22, 23]. AD is a neurodegenerative disease that is the most common cause of dementia, resulting in a progressive decline in cognitive function. AD is associated with amyloid β deposition in the central nervous system (CNS), which is thought to cause AD by formation of plaques and neurofibrillary tangles composed of a hyperphosphorylated tau protein [24‒27]. It is unknown what triggers this process in the brain; however, some researchers have proposed that a dysfunctional microbiome-gut-brain axis, leading to an increase in chronic low-grade inflammation, may contribute to the etiology and pathogenesis of AD and plaque formation [28‒30]. Currently, there are no preventative or disease altering treatments available that could prevent the neurodegenerative decline associated with AD and dementia.

One recent clinical trial found that Lactobacillus rhamnosusGG (LGG) administration is associated with improved cognitive performance in middle-aged and older adults with cognitive impairment [5]. LGG is one of the most studied probiotics that occurs naturally in the gastrointestinal tract and is able to resist high acid and bile concentrations so that it can survive through the stomach and intestines and colonize the intestinal mucosa [31‒34]. This makes LGG a natural candidate for probiotic studies especially due to its antimicrobial effects and its role in immune modulation [33, 35‒37]. However, additional work is needed to clarify neuroprotective effects in aging adults [5].

Little is known about the effects of LGG on the CNS. However, recent studies have shown that Lactobacillus-derived extracellular vesicles can cross the blood-brain barrier [38], can be found in cerebrospinal fluid (CSF), and can interact with neurons and glial cells in the CNS [39]. Lactobacillus-derived extracellular vesicles can interact directly with microglia to increase Arg1, a biomarker of microglia with an anti-inflammatory phenotype, and to increase the anti-inflammatory cytokine, interleukin 10 (IL-10), following lipopolysaccharide (LPS) stimulation [40]. Therefore, we hypothesized that LGG administration induces a biological signature in the CNS consistent with anti-inflammatory signaling. Here, we tested this hypothesis in a study utilizing oral administration of LGG in the drinking water for 21 days in adult male Sprague Dawley rats. We then collected CSF for analysis of proteomic profiles and hippocampal tissue for analysis of hippocampal mRNA expression associated with anti-inflammatory signaling. Genes associated with anti-inflammatory signaling that were analyzed included Il10, Tgfb1, Il4, and IL-4-responsive genes, Cd200, Cd200r1, and Mrc1 (Cd206). LGG administration in the drinking water altered the proteomic profile in the CSF and increased hippocampal Il10 mRNA expression, encoding the anti-inflammatory cytokine IL-10. Although probiotics containing Lactobacillus species, including LGG, are of interest for their ability to reduce systemic inflammation and enhance cognition in both rodent models and humans [41‒44], to our knowledge this is the first study to investigate the effects of LGG administration on the rat CSF proteome.

For an experimental timeline, see Figure 1. Adult male Sprague Dawley^® rats (Hsd:Sprague Dawley^® SD^®; Envigo, Indianapolis, IN, USA; N = 16) weighing 250–265 g upon arrival were pair-housed in Allentown micro-isolator filter-topped caging (259 mm (W) × 476 mm (L) × 209 mm (H); cage model #PC10198HT, cage top: #MBT1019HT; Allentown, NJ, USA) containing an approximately 2.5 cm-deep layer of bedding (Cat. No. 7090; Teklad Sani-Chips; Harlan Laboratories (now Inotiv), Indianapolis, IN, USA). All rats were kept under standard laboratory conditions (12 h light/dark cycle, lights on at 07:00 h, 22°C) and had free access to a standard rat diet (Harlan Teklad 2918 Irradiated Rodent Chow, Envigo (now Inotiv), Huntingdon, United Kingdom) and reverse-osmosis water. The reverse-osmosis system at the vivarium consists of Boulder city water that is processed through reverse osmosis, then chlorinated to a final concentration of 0.5‐1 ppm chlorine via addition of household bleach. The water is chlorinated to prevent microbial contamination, which is standard practice in many laboratory animal institutions. Water bottles were filled approximately 1 week prior to delivery to each animal facility, and the vast majority of chlorine is dissipated by the time the water reaches the animals. Animals had access to probiotic-enriched or control water ad libitum. Cages were changed once per week. The reporting described here was conducted in compliance with The ARRIVE 2.0 Guidelines for Reporting Animal Research [45], and all studies were consistent with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, Eighth Edition (National Research Council, 2011) [46]. The Institutional Animal Care and Use Committee at the University of Colorado Boulder approved all procedures. All efforts were made to limit the number of animals used and their suffering.

LGG was obtained from the American Type Culture Collection (ATCC, L. rhamnosus GG, Cat. No. ATCC 53103) and grown in Lactobacilli MRS Broth (Cat. No. 288130, BD Biosciences, San Jose, CA, USA) at 37°C. The concentration of bacteria was determined by counting colony-forming units (CFU) in total volume of media. Bacteria were spun at 10,000 × g for 10 min, media was removed, and 150 mL of reverse-osmosis water was added. Concentrated LGG solution was kept at 4°C for the duration of the experiment.

Each cage of experimental rats received either a water bottle containing live LGG solution diluted in reverse-osmosis water or standard reverse-osmosis water as a control. The goal was to administer approximately 1 × 10⁹ CFU LGG/rat/day; after estimating that each animal freely drank about 35 mL per day, we adjusted the final concentration of LGG in water bottles to 3.34 × 10⁷ CFU/mL. Water bottles containing LGG in drinking water or drinking water alone were administered daily for 21 days; water bottles were replaced daily 1 h before lights-off. Daily consumption of liquid in bottles was monitored by volume, as shown in online supplementary Table S1 (for all online suppl. material, see https://doi.org/10.1159/000544842), each cage averaging 69.8 ± 4.3 (SEM) mL probiotic-enriched water (N = 4) – approximately 1.17 × 10⁹ CFU of LGG per animal – or 72.5 ± 2.5 (SEM) mL control water (N = 4) per day; liquid consumption was not significantly different between the groups (p > 0.05; t = 0.59). Animal weights were also assessed to monitor the effects of LGG treatment on body weight gain, but there were no significant effects found (online suppl. Table S1). LGG-treated animals (n = 8) averaged an increase in body weight of 62.9 ± 3.7 (SEM) g whereas control animals (n = 8) averaged an increase in body weight of 60.0 ± 4.6 (SEM) g across the duration of the experiment (p > 0.05; t = −0.48).

Rats were euthanized on day 0, after 21 days of LGG administration, using an overdose of sodium pentobarbital (Fatal Plus^®, Vortech Pharmaceuticals Ltd., Dearborn, MI, USA; 200 mg/kg, i.p.). Immediately after euthanasia, CSF was collected, as described previously [16, 47]. Briefly, the dorsal aspect of the skull was shaved and swabbed with 70% ethanol. A sterile 26-gauge needle, attached via PE50 tubing to a sterile 1 mL syringe, was inserted into the cisterna magna, and 0.2 mL of clear CSF was drawn into the syringe. CSF was then spun down at 1,000 × g for 10 min, and the supernatant was collected and stored at −80°C. Due to loss of samples during CSF collection, final sample sizes for CSF analysis were: (1) water control, n = 6; and (2) LGG, n = 7. Rats were then perfused with ice-cold 0.9% saline to remove peripheral leukocytes from CNS vasculature. After perfusion, whole hippocampus was dissected from both left and right hemispheres, stored separately, and immediately placed in liquid nitrogen; hippocampi were stored at −80°C.

Samples were processed for liquid chromatography-tandem mass spectrometry (LC-MS/MS) at the Central Analytical Mass Spectrometry Facility, University of Colorado Boulder. CSF (10 µg protein in 50 µL from each animal) was diluted in 150 µL of 0.1 m ammonium bicarbonate, 0.01% sodium deoxycholate, 5 mm tris (2-carboxyethyl) phosphine (TCEP), 20 mm chloroacetamide, and incubated at 80°C for 15 min in darkness. Samples were digested with 0.2 µg (2% w/w) of Lys-C/trypsin mixture (Promega, Madison, WI, USA) at 37°C overnight. An equal volume of ethyl acetate containing 2% formic acid was added to the sample and mixed vigorously, and the mixture was loaded onto in-house-packed Stop-and-Go-Extraction (STAGE) tips (styrene divinylbenzene-reverse phase sulfonate [SDB-RPS] membrane) [48]. Samples were washed twice with ethyl acetate containing 2% formic acid and twice with Buffer A (0.1% formic acid in water), and eluted using Buffer X (80% acetonitrile, 5% ammonium hydroxide). The desalted peptides were dried using vacuum centrifugation. Tandem mass tag (TMT)-labeling (6-plex) of peptides was performed as described in Zecha et al. [49].

The TMT-labeled, tryptic peptides were resolved using a Waters nanoACQUITY Ultra-Performance Liquid Chromatography (UPLC) system in a single pump trap mode. The peptides were loaded onto a nanoACQUITY 2G-V/MTrap 5 µm Symmetry C18 column (180 µm × 20 mm) with 95% Buffer A and 5% Buffer B (0.1% formic acid in acetonitrile) at 15 µL/min for 7 min. The trapped peptides were eluted and resolved on a BEH C18 column (130 Å, 1.7 µm × 75 µm × 250 mm) using gradients of 5%–30% B (0–145 min) and 30%–60% B (145–150 min) at 0.3 µL/min. MS/MS was performed on a linear trap quadrupole Orbitrap Velos mass spectrometer, scanning precursor ions between 400 and 1,800 m/z (1 × 10⁶ ions, 60,000 resolution) and selecting the 6 most intense ions for MS/MS with 180 s dynamic exclusion, 10 ppm exclusion width, repeat count = 1, and 30 s repeat duration. Ions with unassigned charge states and MH⁺¹ were excluded from the MS/MS. Higher energy collisional dissociation MS/MS (5 × 10⁴ ions, 7,500 resolution) was performed with a default charge state = 5, isolation width = 2 m/z, normalized collision energy = 40, and activation time = 1 ms. Maximal ion injection times were 500 ms.

MaxQuant/Andromeda (version 1.6.2.10) was used to process raw files from linear trap quadrupole-orbitrap with a database consisting of UniProt Rattus norvegicus proteome (total 36,098 entries, downloaded on 2/15/2019) [50‒52]. All protein names and gene abbreviations were gathered from the UniProt database [52]. A preset “6-plex TMT” was used. The search allowed trypsin specificity with a maximum of two missed-cleavage, setting carbamidomethyl modification on cysteine as a fixed modification and protein N-terminal acetylation and oxidation on methionine as variable modifications. MaxQuant used 4.5 ppm main search tolerance for precursor ions, 20 ppm higher energy collisional dissociation MS/MS match tolerance, searching the top 12 peaks per 100 Da. False discovery rates for both protein and peptide were 0.01 with a minimum peptide length of seven amino acids. “Reporter ion corrected” values in proteinGroups.txt were used for downstream statistical analysis.

Real-time reverse transcription polymerase chain reaction (RT-PCR) was performed as described previously [53]. Total RNA was isolated from the left whole hippocampus utilizing a standard method of phenol:chloroform extraction [54]. cDNA sequences were obtained from Genbank at the National Center for Biotechnology Information (NCBI; www.ncbi.nlm.nih.gov) [55]. Primer sequences were designed using the Operon Oligo Analysis Tool (http://www.operon.com/technical/toolkit.aspx) tested for sequence specificity using the Basic Local Alignment Search Tool (BLAST) at NCBI [56]. Primers were obtained from Invitrogen. Primer specificity was verified by melt curve analyses. All primers were designed to span exon/exon boundaries and thus exclude amplification of genomic DNA (see online suppl. Table S2 for primer sequences). PCR amplification of cDNA was performed using the QuantiTect SYBR Green PCR Kit (Cat. No. 204056, Qiagen, Hilden, Germany). Formation of PCR product was monitored in real time using the MyiQ Single-Color Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). Relative gene expression was determined using the delta delta Ct method, by taking the expression ratio of the gene of interest to Actb, encoding β-actin. Genes associated with anti-inflammatory signaling that were analyzed included Il10, Tgfb1, Il4, and IL-4-responsive genes, Cd200, Cd200r1, and Mrc1 (Cd206). Three animals in the water control group had undetectable levels of Il10 mRNA expression in the dorsal hippocampus and were subsequently removed from the analysis as missing data to prevent biases induced by assigning values to non-detects [57].

Proteomics analyses were performed using R Statistical Programming (version 3.6.1 for Windows) as previously described [58]. Abundance of each protein was measured by reporter ion-corrected values. To correct for potential noise in the MS/MS detection, reporter ion-corrected values were removed for any proteins that had a reporter intensity count less than two. Corrected peak intensities for each protein were then normalized to the total of all peak intensities for that sample. For fold change analysis, normalized peak intensities of each protein were averaged for each treatment group, and ratios of treatment to control for each protein were created. To control for batch effects of the 6-plex TMT, generalized linear model (GLM) analysis was performed on log2-transformed normalized peak intensity for each protein, with treatment and batch as factors. Two-tailed significance was set at α = 0.1, and an adjusted p value was not calculated due to low power [59]. Statistical analysis was not performed on any proteins if a group sample size was less than 50% of the full sample size for that treatment group. Proteins that were found to be significantly different between treatment and control as well as proteins with a fold change (up or downregulated) greater than two were used in downstream pathway analysis via the Database for Annotation, Visualization and Integrated Discovery (DAVID, version 6.8) [60, 61] using UniProt accession IDs [52] for gene labeling and R. norvegicus as species. For protein interaction analysis, STRING database (version 11) was used [62].

Statistical analysis for hippocampal mRNA expression was performed using the software package IBM Statistical Package for the Social Sciences (version 26.0, SPSS Inc., Chicago, IL, USA). Data were analyzed for extreme outliers using Grubbs’ test for single outliers [63], and extreme outliers were removed from data sets prior to graphical representation of the data and statistical analysis. Normality was assessed using the Shapiro-Wilks test for normality; if the data were found to be non-normal, the data were log-transformed and normality reassessed for confirmation of statistical assumptions. The independent-samples t tests were run on normalized data, and homogeneity of variance was analyzed by the Levene’s Test for Equality of Variance. If variance was unequal, the adjusted p value (obtained from Welch’s t-test) from the SPSS output was used.

Among 224 proteins identified in CSF, our preliminary average fold change analysis revealed that LGG treatment did not alter the abundance of any proteins by two-fold (Fig. 2). However, GLM analysis found significant differences in abundances of seven out of 224 proteins between LGG- and water-treated animals. LGG administration increased the abundance of cochlin (Coch, p = 0.035), neuronal pentraxin receptor (NPTXR, p = 0.035), reelin (p = 0.087), seizure-related 6 homolog-like (Sez6l, p = 0.087), and vacuolar protein sorting 13 homolog C (VPS13C, p = 0.082) (Fig. 2). LGG administration decreased carboxypeptidase Q (CPQ, p = 0.099) and insulin-like growth factor binding protein 7 (IGFBP-7, p = 0.027) (Fig. 2). The top 10% of proteins (22 proteins) of interest in the CSF, determined by p value comparing rats that received LGG administration or drinking water alone, can be found in Table S3.

In order to better understand how LGG administration might be altering different biological functions in the CSF, we performed: (1) protein-protein interaction analysis; and (2) functional enrichment gene analyses (including pathway analysis and gene ontology (GO) analyses for biological process, cellular component, and molecular function). Analyses were performed collectively on both proteins upregulated and downregulated by LGG administration (N = 7: Coch, CPQ, IGFBP-7, NPTXR, reelin, Sez6l, VPS13C) using STRING (version 11; protein-protein interaction analysis) and DAVID bioinformatics resource (version 6.8; functional enrichment gene analyses) [60, 61].

There were no established interactions between any of the proteins altered in the CSF, not by curated database, experimentally determined, textmining, nor coexpression, as computed by STRING database (data not shown). Likewise, there were no pathways enriched by the set of seven genes, as assessed by DAVID analysis (Table 1). Functional enrichment analysis for GO, biological process, revealed the top two hits to be “cell adhesion” and “response to organic cyclic compound” (Table 1). Functional enrichment analysis for GO, cellular component, indicated the top four hits to be “extracellular space,” “extracellular exosome,” “proteinaceous extracellular matrix,” and “extracellular matrix” (Table 1). Functional enrichment analysis for GO, molecular function, found one hit: “metal ion binding” (Table 1). Functional enrichment analysis for the top 10% of proteins (22 proteins) of interest in the CSF, as determined by and ordered by p value, can be found in online supplementary Table S4.

Given the well-documented anti-inflammatory effects of LGG in the periphery [64‒66], and given the ability of Lactobacillus-derived extracellular vesicles to interact directly with microglia to increase Arg1, a biomarker of microglia with an anti-inflammatory phenotype, and to increase the anti-inflammatory cytokine, IL-10, following LPS stimulation [40], we investigated the gene expression of a panel of anti-inflammatory and immunoregulatory markers in the hippocampus. We found that LGG administration increased hippocampal Il10 mRNA expression (p < 0.05) but did not affect Il4 nor transforming growth factor β1 (Tgfb1) mRNA expression in the hippocampus (Fig. 3). Consistent with these results, LGG administration had no effect on the mRNA expression of IL-4-responsive genes, including Cd200, Cd200r1, and Mrc1 (Cd206) (online suppl. Fig. S1).

Oral administration of LGG for 21 days altered the CSF proteome, increasing the abundance of five proteins (cochlin, NPTXR, reelin, Sez6l, and VPS13C) and decreasing the abundance of two proteins (CPQ, IGFBP-7) in the CSF, and increased hippocampal Il10 mRNA expression in rats. Although our protein set was not enriched for any major pathways, prior studies indicate that cochlin, NPTXR, IGFBP-7, and reelin are extracellular matrix (ECM) proteins [68, 69], consistent with the overall findings from our gene ontology enrichment analyses and suggesting an alteration in neural networks and architecture [70]. In addition to their functions as ECM proteins, NPTXR, reelin, and Sez6l proteins participate in glutamatergic signaling pathways, which may also be linked to resilience in the context of stress-related psychiatric disorders in which inflammation is considered a risk factor, such as major depressive disorder (MDD). Likewise, NPTXR [71], reelin [72, 73], and Sez6l [74, 75] are important for cognition and memory by mediating excitatory (glutamatergic) synaptic transmission of neurons in the hippocampus and cortex; decreases in the abundance or function of NPTXR [76‒78], reelin [79], or Sez6 proteins [80], are linked to various forms of cognitive impairments and dementia. LGG administration also increased the relative expression of Il10 mRNA in the hippocampus, suggesting increased anti-inflammatory signaling in the brain. Collectively, these results suggest that administration of LGG may promote a physiological resilience to neuroinflammatory events by modulating the extracellular scaffolding consistent with synaptic plasticity [69], including glutamatergic signaling, and promoting an anti-inflammatory environment in the CNS.

LGG increased the abundance of five proteins (cochlin, NPTXR, reelin, Sez6l, and VPS13C) and decreased the abundance of two proteins (CPQ, IGFBP-7) in the CSF. Although probiotics containing Lactobacillus species are of interest for their ability to reduce systemic inflammation and enhance cognition in both rodent and human models [41‒44], to our knowledge this is the first study to investigate the effects of LGG administration on the rat CSF proteome in vivo. A few published studies have already found associations between probiotic administration and alterations in the abundance of the proteins significantly altered in the CSF of our dataset. For example, previous studies report general associations between the gut microbiome composition and gene expression of cochlin [81] and reelin [82] in peripheral tissues and Sez6l in microglia [83]. Of particular interest, one study showed that cochlin mRNA expression was increased in the whole blood cells of elderly persons after 28 days of LGG administration, consistent with overall changes in gene expression related to cellular movement and cell-to-cell signaling interaction [81]. This finding is consistent with our results showing that LGG administration increased the relative abundance of cochlin, along with several other ECM-related proteins, in the CSF. Nevertheless, future studies should continue to analyze and understand the effect of Lactobacillus probiotic administration on the proteomic profiles of the host’s CNS in vivo.

Although our protein set was not enriched for any major pathways, prior studies indicate that cochlin, NPTXR, IGFBP-7, and reelin are ECM proteins [68, 69], consistent with the overall findings from our gene ontology enrichment analyses and suggesting an alteration in neural networks and architecture [70]. In the gut, ECM proteins like reelin are critical for maintaining the integrity of the intestinal barrier [84]. Meanwhile, in the brain, ECM proteins help create the extracellular scaffolding for neurons and glia, and altering the secretion of these proteins can mediate synaptic plasticity and neural connectivity [85, 86]. The neural ECM is formed by secretions from both neurons and glial cells [86] and includes a structure called the perineuronal net, which helps maintain the position of neurons in space [85]. Probiotics containing Lactobacillus species may be able to retain neural ECM integrity and prevent injury to the blood-brain barrier during inflammatory events [87‒90]. Interestingly, among the top 10% of proteins of interest (determined by p value), neuronal pentraxin-1 (NP1), the endogenous ligand for NPTXR, was also increased in CSF after LGG administration for 21 days, although this effect was not significant (online suppl. Table S3). Furthermore, the top 10% of proteins of interest revealed increases in the relative abundance of neurotrimin (Ntm), an immunoglobulin cell adhesion molecule (part of the IgLON family) involved in neurite outgrowth [91]; a disintegrin and metalloproteinase with thrombospondin motifs (ADAMST4), known for mediating extracellular matrix reorganization to promote neurite outgrowth [92], possibly by cleaving cell surface proteins such as Ntm [91]; and fibroblast growth factor receptor 1, which may be involved in assembly or maintenance of perineuronal nets [93], although these results were not significant (online suppl. Tables S3, S4). Nevertheless, an interesting trend has occurred in our study, i.e., LGG increased the abundance of several proteins with overlapping functions involved in synapse outgrowth and maintenance [86]. A parallel increase of both NP1 and NPTXR alongside increases of reelin and ADAMST4 is good evidence that LGG promotes ECM-mediated synaptic remodeling [86]. The top 10% of proteins of interest also revealed decreases in the relative abundance of galectin-3 binding protein (Gal-3BP) and fibronectin, two ECM proteins that strongly bind to one another, further supporting the idea that, among LGG-treated animals, there may be a rearrangement to the extracellular structures of brain tissue [94].

In addition to their functions as ECM proteins, NPTXR, reelin, and Sez6l proteins participate in glutamatergic signaling pathways, which may also be linked to resilience in the context of stress-related psychiatric disorders, in which inflammation is considered a risk factor, such as MDD [95‒97]. Dysregulation of glutamatergic signaling and synaptic outgrowth is thought to play an important role in the development of MDD [98], and, in fact, the rapid-acting antidepressant ketamine is an N-methyl-d-aspartate receptor antagonist [99], although additional work suggests that ketamine and its metabolites may also work though (AMPA) receptor-mediated mechanisms [100]. A recent study found that persons with MDD consistently show decreases in NPTXR protein in the CSF associated with altered glutamatergic and γ-aminobutyric acid signaling [95]. Similarly, persons with schizophrenia, bipolar disorder, and MDD show systemic reductions in reelin protein [101, 102]. While corticosterone treatment has been shown to decrease reelin in the dentate gyrus of rats, treatment with ketamine rescues this deficit [103]. Sez6 family knockout mice (a triple knockout of Sez6, Sez6l, and Sez6l2) also show increases in behavioral responses to stress and memory impairments, corresponding to decreased dendritic spines in the hippocampus [80]. Interestingly, the gut microbiome may influence glutamatergic signaling throughout the periphery and the brain of the host [104, 105]. Lactobacillus species, including L. rhamnosus, are thought to produce glutamate, which eventually influences the microbiome-gut-brain axis by interacting with the enteric nervous system [104, 105]. Although different strains of L. rhamnosus, e.g., L. rhamnosus JB-1 and LGG, can differ in genomes [106], adhesion capabilities (i.e., LGG adheres more readily to mucus than other strains) [106‒108], and biofilm formation abilities (i.e., LGG can form biofilms more effectively than other strains) [106], it is also well documented that L. rhamnosus JB-1 [41, 109] and LGG [110] probiotics promote anxiolytic and antidepressant effects on host behavior. Our data, taken together with studies of L. rhamnosus spp. on behavior, may suggest that, among undisturbed animals, administration of LGG alters proteins in the CSF that are associated with glutamatergic signaling and that, along with immunomodulatory effects, may predispose the host to a stress-resilient phenotype.

NPTXR [71], reelin [72, 73], and Sez6l [74, 75] are also important for cognition and memory by mediating excitatory (glutamatergic) synaptic transmission of neurons in the hippocampus and cortex; meanwhile, decreases in the abundance or function of NPTXR [76‒78], reelin [79], Sez6 [80], and VPS13C [111] proteins are linked to various forms of cognitive impairments and dementia. In rodents, the rostral migratory stream (RMS) is a pathway for newly generated neurons to migrate from the subventricular zone to the olfactory bulb, where they differentiate to postmitotic interneurons [112]. There is some evidence that suggest the RMS exists in the human brain as well, producing neuroblasts in adulthood [113, 114]. Reelin plays an important role in neuroblast migration through the RMS [115]. Meanwhile, Sez6l gene is upregulated in neuroblasts that have migrated from the ventricular-subventricular-zone toward the site of injury in the mouse cortex [116]. NPTXR is well-known for its influence on synaptogenesis [117], and in vitro, cochlin may regulate neural differentiation of embryonic stem cells [118].

LGG administration also increased the relative expression of Il10 mRNA in the hippocampus, suggesting an increase in anti-inflammatory signaling in the brain. We and others have previously reported on the anti-inflammatory and immunoregulatory effects of probiotics or heat-killed preparations of bacterial strains (i.e., “ghost probiotics,” “postbiotics,” “inactivated probiotics,” or “paraprobiotics” [119, 120]) associated with stress resilience [9, 12, 16, 58, 121‒131], including the L. rhamnosus strain JB-1 [41, 44]. We have shown, e.g., that subcutaneous injections of a heat-killed preparation of Mycobacterium vaccae NCTC 11659 once per week for 3 weeks can increase long-term expression of both Il4 mRNA and IL-4 protein in the hippocampus, which is associated with the prevention of stress-induced anxiety-like defensive behavioral responses in rodents [127]. IL-10 has been shown to exert protective effects in the brain by acting on microglia to suppress inflammatory signaling [132‒135]. The results from our present study are consistent with previous studies that also demonstrate anti-inflammatory effects of LGG, including increases in IL-10 signaling [64, 66, 136, 137]. Our findings imply that administration of LGG may lead to protective neuroimmunologic effects in the hippocampus that might enhance cognition and memory or may protect against inflammation-induced impairments in cognition and memory [135, 138]. For example, Richwine et al. described cognitive deficits in IL-10-deficient mice 24 h after a peripheral injection of LPS, suggesting that IL-10-deficient mice are more susceptible to inflammation-induced impairments in working memory and integrating new information with previously learned information [139]. On the other hand, Wallace et al. [138] observed that administering a combination of the antibiotic ceftriaxone and IL-10 to rats infected with E. coli increased spatial memory abilities in the Morris water maze, possibly by reducing inflammation in the hippocampus caused by the infection. More recently, it was found that intranasal treatment with IL-10 reverses stress-induced learning impairments in novel object recognition and spatial memory tasks after mice are exposed to a learned helplessness paradigm, and IL-10 treatment protects against stress-induced loss of neuronal spine density in the hippocampus [135]. The neuroprotective effects of IL-10 work presumably by dampening the microglial inflammatory response to stress [135]. Overall, these results are consistent with the finding that administration of anti-inflammatory cytokines [135], or Lactobacillus spp.-containing probiotics that promote anti-inflammatory signaling [140], can buffer against the adverse effects of hippocampal inflammation and affected cognitive pathways [141, 142].

Collectively, our results suggest that administration of LGG may promote a physiological resilience to neuroinflammatory events by modulating the extracellular scaffolding consistent with synaptic plasticity [69], including glutamatergic signaling, and promoting an anti-inflammatory environment in the CNS among undisturbed animals. Considering that disruptions in glutamatergic signaling and aberrant pro-inflammatory signaling (or impaired immunoregulatory signaling) are features common to many stress- and inflammation-related psychiatric and neurological disorders [143‒147], our findings indicate that LGG administration may counteract the underlying pathophysiology of certain stress-related psychiatric disorders and disorders associated with cognitive dysfunction [41, 44, 128]. It could be that chronic, low-grade inflammation brought on by a number of genetic and environmental factors precede and contribute to the altered glutamatergic signaling and dysregulation of immunoregulatory signaling [135, 146, 148‒150]. Preventing chronic low-grade inflammation through IL-10 induction may buffer against the manifestation of disrupted synaptic function in the hippocampus and associated cognitive dysfunction [151, 152]. Further studies should explore the potential of LGG as a prophylactic or post-diagnosis treatment for stress-related psychiatric disorders and neurological disorders in which chronic low-grade inflammation is a risk factor.

Although this study contributes new information regarding the effects of Lactobacillus probiotics on the proteome of CSF and hippocampal mRNA expression, our clinical implications are limited by study design. First, our proteomics and gene expression results are restricted to a single time point of tissue collection (i.e., immediately after 21 days of LGG administration); this design raises interesting questions to be addressed in future studies, such as how many days of LGG treatment are required to induce changes in the CNS and how long the effects persist after the final administration of LGG. Additionally, LGG in drinking water was provided to rats ad libitum, allowing for consumption of LGG throughout the day to vary within or between animal subjects; this design may produce results that differ from alternative administration techniques such as oral gavage. Furthermore, the current study design did not address whether or not the LGG administration procedure used would alter behavioral responses, either at baseline, or following stressor exposure, and the study was limited to assessment of expression of hippocampal mRNA encoding anti-inflammatory cytokines, and did not provide insight into the expression of hippocampal mRNA encoding pro-inflammatory cytokines, which also might be affected. This experiment was conducted only in adult male rats, and future studies should include female rats to determine if LGG administration exhibits sex differences on the CSF proteome and hippocampal mRNA expression. Furthermore, LGG administration may produce different results in a population of aged animals due to age-related changes in immune and glutamatergic signaling. Considering that this study was conducted in undisturbed home cage animals, future studies should explore the effects of LGG administration in animal models of stress-related psychiatric disorders or neurodegenerative disease. Finally, effects of LGG on the plasma proteome should be analyzed in parallel with the CSF proteome to understand in part how LGG may be signaling from the periphery to the brain via the microbiome-gut-brain axis.

Disruption of glutamatergic signaling and increased neuroinflammation are classic features of AD pathology [153]. LGG increased the abundance of a number of proteins in the CSF proteome that are involved in synaptic plasticity and glutamatergic signaling, and LGG increased immunoregulatory gene expression in the hippocampus, consistent with a neuroprotective phenotype. It should also be noted here that among the top 10% of proteins altered by LGG administration, the pathway “PI3K-Akt signaling pathway” and the GO, biological process “cellular response to glucose stimulus” were significantly enriched (online suppl. Table S4). This is of particular importance in the context of AD, where impaired phosphatidylinositol-4,5-bisphosphate 3-kinase (PI3K)-Akt signaling is coupled with aberrant glucose signaling and insulin resistance [154]. Our findings suggest that LGG may influence interconnected pathways, including immune responses, cellular metabolism, and glutamatergic signaling in the CNS, the disruption of which are all hallmarks of AD etiology [155]. In addition, findings regarding glutamatergic signaling pathways may also have implications for the treatment of psychiatric conditions, including MDD. Of interest, depression has been discussed as a prodromal symptom of AD, and the cumulative incidence of MDD among those with AD is two-times higher than among those without this condition. Results from this study suggest a novel avenue for intervention for individuals with one or both if these often-comorbid conditions.

This study protocol was reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Colorado Boulder, Approval No. (2620).

C.A.L. is Cofounder and member of the Scientific Advisory Board of Mycobacteria Therapeutics Corporation, and is a member of the faculty of Clinical Care Options, LLC (CCO), Reston, VA, USA, the Integrative Psychiatry Institute, Boulder, CO, USA, the Institute for Brain Potential, Los Banos, CA, USA, and Intelligent Health Ltd, Reading, UK. In the previous 3 years, C.A.L. served on the Scientific Advisory Board of Immodulon Therapeutics Ltd., London, UK. The remaining authors have no conflicts of interest to declare.

These studies were funded by an anonymous gift through Benefunder to the Behavioral Neuroendocrinology Laboratory. L.M.D. was supported in part by the National Institutes of Health National Center for Complementary and Integrative Health (R01AT010005 (L.A.B., C.A.L., PIs). The remaining authors received no funding. Purchase of a Thermo Orbitrap Velos mass spectrometer was made possible with a grant from W.M. Keck Foundation. The funder had no role in the design, data collection, data analysis, and reporting of this study.

Study design was conceived by C.A.Z., J.D.H, and C.A.L. LGG administration was performed by K.M.L., C.A.Z., J.D.H., A.I.E., and J.E.H. Tissue collection was conducted by K.M.L, C.A.Z., H.M.D., and M.G.F. CSF sample preparation and LC-MS/MS was performed by T.L. Real-time RT-PCR was performed by H.M.D. Data analysis and statistical analysis were carried out by K.M.L. Figures and figure legends were produced by K.M.L. Manuscript preparation was conducted by K.M.L. and L.M.D. Editing and review was contributed by K.M.L, C.A.Z., T.L., J.D.H., A.I.E., J.E.H., H.M.D., M.G.F., L.A.B., S.F.M., and C.A.L.

James E. Hassell Jr.: present address: Division of Endocrinology, Metabolism and Diabetes, Department of Medicine, University of Colorado Anschutz Medical Campus, Aurora, CO, USA.

All data generated or analyzed during this study are included in this article and its online supplementary material files. Further inquiries can be directed to the corresponding author.