Abstract
The regulation of physiological gut functions such as peristalsis or secretion of digestive enzymes by the central nervous system via the Nervus vagus is well known. Recent investigations highlight that pathological conditions of neurological or psychiatric disorders might directly interfere with the autonomous neuronal network of the gut – the enteric nervous system, or even derive from there. By using a murine Alzheimer’s disease model, we investigated a potential influence of disease-associated changes on gastrointestinal properties. 5xFAD mice at three different ages were compared to wild type littermates in regard to metabolic parameters and enzymes of the gut by fluorimetric enzyme assay and western blotting. Overexpression of human amyloid-β protein precursor (AβPP) within the gut was assessed by qPCR and IHC; fecal microbiome analysis was conducted by 16SrRNA quantitation of selected phyla and species. While general composition of fecal samples, locomotion, and food consumption of male 5xFAD animals were not changed, we observed a reduced body weight occurring at early pathological stages. Human AβPP was not only expressed within the brain of these mice but also in gut tissue. Analysis of fecal proteins revealed a reduced trypsin amount in the 5xFAD model mice as compared to the wild type. In addition, we observed changes in fecal microbiota composition along with age. We therefore suggest that the presence of the mutated transgenes (AβPP and PS1), which are per se the basis for the genetic form of Alzheimer’s disease in humans, directly interferes with gut function as shown here for the disease model mice.
INTRODUCTION
There is growing evidence that the gut is involved in regulation of brain functions, mainly via its microbiota and its metabolic activity. Gut microorganism composition has been shown to contribute to regulation of social behavior, stress resistance, and cognitive functions [1]. For instance, it has been reported that germ-free mice display reduced anxiety-like behavior in tests such as the light/dark test [2]. Reconstitution of the microbiome early in life is able to restore this anxiolytic phenotype [3]. Additionally, a depression-like phenotype could be normalized in rats subjected to the maternal separation (MS) model of depression: MS adult rat offspring chronically treated with Bifidobacterium infantis showed normalization of the stress-altered immune response and reversal of behavioral deficits in the forced swim test [4]. In mice, chronic treatment with Lactobacillus rhamnosus induced region-dependent alterations in GABA subunit mRNA and reduced stress-induced corticosterone and anxiety- and depression-related behavior [5]. Interestingly, these effects were absent in vagotomized mice, indicating that the N. vagus plays an important role in the observed regulation of CNS function via the gut’s microbiome. Similar results were obtained in humans where administration of fermented milk led to a positive influence on emotive state [6]. The impact of the gut microbiome on brain function might be achieved by metabolic products such as 4-ethylphenylsulfate which has been described to leak into circulation in an autism mouse model [7]. Another example is given by microbial propionic acid that activates fatty acid receptor FFAR3 of the host, thereby activating hypothalamic areas which regulate energy metabolism [8]. Additionally, the enteric bacterial metabolite increases locomotor behavior in intraventricularly infused rats and leads to alteration of more than 20 phospholipid species within the brain [9]. It is not quite clear what comes first, changes in gut or functional dysregulation of brain areas. Several recent studies report, for example, an altered bacterial community in stool samples derived from children with autism [10, 11]. Changes of the microbiota such as small intestinal bacterial overgrowth, a malabsorption syndrome have also been found in Parkinson’s disease patients [12]. Such changes are accompanied by altered bacterial metabolite amounts, for example, short chain fatty acids [13]. Interestingly, in Parkinson’s disease, intranuclear alpha-synuclein depositions not only occur in the CNS but also in the gastric myenteric and submucosal plexuses [14]. For other psychiatric diseases such as schizophrenia, epilepsy, or sleep disorders, a potential correlation has yet not been deciphered in detail [15].
Alzheimer’s disease is the most prominent CNS disorder of the aging population and it has already been speculated that pathogenic microbes or altered microbiota composition might contribute to this neurodegenerative disease (summarized in [16]).
One of the hallmarks of Alzheimer’s disease, the Aβ peptides that derive from proteolytic processing of the amyloid-β protein precursor (AβPP), has been demonstrated to be deposited in intestinal tissue already decades ago [17]. In AD patients as wells as healthy controls, AβPP immunoreactivity was located in enteric ganglia (from the esophagus throughout the transverse colon), while mucosa or smooth muscle were not labeled [18]. In the TgCRND8 AD mouse model, Aβ plaque-like structures in the gut have been described, accompanied by reduced thickness of longitudinal as well as circular muscle layer and decreased contractionpotential [19].
Patients already diagnosed with Alzheimer’s disease-related dementia show rapid weight loss associated with faster disease progression [20-22].This may result from forgetfulness on food intake or impairment of odor recognition by degeneration of olfactory cortex neurons which is an early event in AD (e.g., [23]). However, weight loss even before onset of clinical symptoms has been documented in several studies: a long-term study involving 1,800 Japanese-American men found that elderly patients with dementia lost on average 10% of their body weight years before being diagnosed [24]. In a cohort that included 449 elderly initially cognitively normal persons weight loss doubled one year before detection of disease in those that later-on developed AD [25]. By using data from the Rochester Epidemiology Study, Knopman and colleagues reported that weight between female cases and controls 21 to 30 years prior to the onset of dementia did not differ. However, a reduced weight occurred more than 10 years prior to the index year [26]. Aβ load has been demonstrated to build up about 15 years before first clinical signs of the disease (e.g., [27]); therefore potential effects on the gut might occur at such an early time point.
Here, we investigated if early body weight reduction observed in an aggressive AD mouse model (5xFAD, APP K670N, M671L, I716V; PS1 M146L, L286V; [28]) might be a direct consequence of AD-like pathological features interfering with gut function. Within this mouse model, Aβ deposits usually start at 6–8 weeks of age and at an age of 25 weeks result in a large numbers of hippocampal and cortical plaques, accompanied by deficits in spatial learning (starting with 17–22 weeks, [28]). We therefore chose three different stages of age for our investigation: 6 weeks (initiation of plaque deposition), 9 weeks (intermediate plaque deposition) and 18 weeks (deposition and first behavioral dysfunction).
MATERIAL AND METHODS
Animals
Male 5xFAD mice (APP K670N, M671L, I716V; PS1 M146L, L286V; Jackson Laboratory, Bar Harbor, ME, USA) were crossbred for maintenance with female C57B/6J from the animal facility of the University Medical Center of Mainz. Food (Mouse breeding extrudate, ssniff Spezialdiäten GmbH, Soest, Germany) and water were provided ad libitum and a 12 h light–dark cycle was maintained. All experimental procedures were carried out in accordance with the European Communities Council Directive regarding care and use of animals for experimental procedures and approved by local authorities. Exact n-numbers of each group are given in Supplementary Table 1.
Genotyping was performed using the primers 5xFAD_for GTAGCAGAGGAGGAAGAAGTG and 5xFAD_rev CATGACCTGGGACATTCTC as described before [29]. Sex determination of pups was performed by PCR with primers for Jarid [30] XY_for CTGAAGCTTTTGGCTTTGAG and XY_rev CCGCTGCCAAATTCTTTGG (see Supplementary Figure 1).
Body weight determination, food intake measurement, and sampling of fecal pellets
For sampling, male mice were encaged separately. Fecal pellets from a 24-h period were collected, weighed, and stored at –80°C until further use. For each mouse, samples were taken on four consecutive days at the age of 6, 9, and 18 weeks. During this time period, food intake and body weight were determined daily at a fixed time point in the morning.
Female mice were not separated during the weighing period. Body weight of pups was measured only at a single time point to avoid stress for pups and mothers.
Water content of feces was assessed by drying two pellets per sampling day per mouse and collection time point until constant weight (20 h) at 40°C. Water content was gained by calculating the weight decrease between wet and dry weight in % of wetsample.
Open field paradigm
To assess general activity, mice were subjected to an open field paradigm for 30 min (test arena: 100×100×40 cm) following the last fecal pellet sampling day. Distance moved was recorded automatically by a computerized video system (IBM-type AT computer with a video digitizer and a CCD video camera). Data acquisition and analysis was carried out using EthoVision XT release 8.0 (Noldus Information Technology, Utrecht, Netherlands).
Histology and immunohistochemistry (IHC)
Full-thickness samples from duodenum and distal colon were dissected from 1-month-old male 5xFAD and wild type mice, washed with PBS, and drop-fixed in 4% formaldehyde for 24 h. Tissue was embedded at 58°C in paraffine. 2 μm serial sections were cut using a rotary microtome (Leica RM 2245) and prepared for IHC stainings. After mounting on histological slides (Superfrost plus, Menzel), sections were stained following standard protocols using the primary antibody 6E10 (diluted 1 : 500 in Antigen Retrieval Buffer 1; Medac, Wedel, Germany) and 3,3’ diaminobenzidine (DAB). Morphological features were assessed on routine hematoxylin and eosin staining. Samples were evaluated at 400x magnification, using a light microscope equipped with a digital camera (EVOS XL, Life Technologies).
RealTime RT PCR
Tissue was collected, mechanically purified, and washed in ice-cold isotonic NaCl. Tissue stabilized by RNAlater (QIAGEN, Hilden, Germany) was stored at –80°C. Total RNA was prepared using RNeasy kit (QIAGEN, Hilden, Germany). 100 ng RNA per sample were subjected in duplicate to AβPP mRNA quantitation using a StepOne Plus cycler (Applied Biosystems, Waltham, MA, USA) and One Step RT SYBR Green kit (QIAGEN, Hilden, Germany) with the AβPP primers Hs_APP_2_SG (QIAGEN, Hilden, Germany). 18SrRNA was used as a house keeping gene (Mm_Rrn18s_1_SG, QIAGEN, Hilden, Germany). Relative RNA quantities were calculated according to standard curves.
Western blotting
Fecal samples (two pellets from each mouse per collection day) were incubated on ice for 10 min in 500 μl extraction buffer (50 mM EDTA, 20% v/v 5xPBS, Protease inhibitor complete mini (Roche, Mannheim, Germany)) and homogenized using a hand-held homogenizer (XENOX MHX/E, Fähren) for 2 min. Samples were centrifuged at 4°C 18.188 g for 30 min, protein concentration of supernatant determined by the method of Bradford [31], and samples diluted with NuPAGE buffer (Thermo Fisher, Waltham, MA, USA) to yield 0.5 μg/μl. 10 μg protein were subjected to 10% SDS polyacrylamide gels and transferred to a nitrocellulose membrane. Blots were blocked in 5% milk powder buffer and incubated with the antibodies as follows: anti-carbonic anhydase VI (CAVI), anti-trypsin, anti-major urinary protein (MUP) (all Santa Cruz, 1 : 200 in milk powder buffer, for a specificity control see Supplementary Figure 2); antibody 6E10 (1 : 1000, Covance, Princeton, NJ, USA) was used after blocking with I block buffer (0.2%, Thermo Fisher, Waltham, MA, USA). Appropriate secondary antibodies labeled with HRP (1 : 3000; Thermo Fisher, Waltham, MA, USA) were used for detection of epitope-derived signals with ECL Super signal West Femto (Thermo Fisher, Waltham, MA, USA) and a CCD camera (Raytest). Densitometric analyses were performed with software Aida 3.5 (Raytest, Straubenhardt, Germany).
Silver staining
Detection of trypsin (bovine/ porcine, Sigma-Aldrich, Steinheim, Germany) from in vitro binding assays with Aβ peptides was performed using the RotiBlackP Kit (Roth, Karlsruhe, Germany) as recommended by the vendor.
Aβ fibrillization
Human Aβ42 was prepared and purified as described previously [17]. Peptides were dissolved in 10 mM NH4OH and protein concentration calculated via absorbance at 280 nm (Nanodrop, Pierce). Aliquots of the peptide were stored at –80°C. For fibrillization, peptides were incubated in 1xPBS for 16 h at 24°C, 300 rpm.
Trypsin enzymatic assay
For assessing in vitro impact of Aβ monomers or fibrils on trypsin catalytic activity, the enzyme (porcine, 25U per reaction) was incubated at 37°C for 10 min 300 rpm with the indicated amount of Aβ preparations. Enzymatic activity was measured by absorbance at 253 nm (Biowave photometer, Biochrom, Berlin, Germany) over a 5-min incubation period with 74 μg/ml BAEE (Nα-benzoyl-DL-Arginine-β-naphthylamide, Sigma-Aldrich, Steinheim, Germany) in 67 mM NaH2PO4 pH 7.6.
Tryptic fecal activity was determined by pulverization of one frozen pellet for each collection day of the respective age and animal using a mortar and pestle. 10 mg grinded material was mixed with 1 ml ice cold homogenization buffer (150 mM NaCl, 20 mM Tris HCL pH 8.3) and incubated on ice for 5 min. Subsequently, fecal samples were homogenized in a grind mill (Tissue Lyzer, QIAGEN, Hilden, Germany) for 5 min at 50 Hz. After centrifugation at 18.188 g, 4°C for 5 min supernatant was aspirated and used for enzymatic assay. For pancreatic tissue, 50 mg were homogenized in 2000 μl buffer (500 mM NaCl, 20 mM CaCl2, 200 mM Tris HCL pH 8.0) for 10 min at 50 Hz, followed by a centrifugation at 18.188 g, 4°C for 5 min. An aliquot diluted with water (1 : 4) served for measuring trypsin activity derived from already activated enzyme. Whole trypsin amount (trypsinogen + trypsin) was determined by incubating pancreatic tissue lysate supernatant for 2 h with 10 U enteropeptidase (NEB, Frankfurt am Main, Germany) at 37°C and subsequent enzymatic assay. Heat-inactivated trypsin or trypsin pretreated for 5 min with trypsin inhibitor (30 μg/ml, Sigma-Aldrich, Steinheim, Germany) served as negative controls.
Analysis of the fecal microbiota
Fecal samples were sent for microbiological analysis to the MVZ Institut für Mikroökologie (Herborn, Germany). Microbial DNA was extracted from 200 mg feces using the QIAsymphony DSP Virus/Pathogen Mini-Kit on the QIAsymphony SP (QIAGEN, Hilden, Germany) by automated isolation and pipetting (QIAsymphony SP/AS instrument. Primers [11, 33] were selected to recognize either the whole bacterial phyla (Firmicutes, Bacteroidetes) or in the case of the phyla Verrucomicrobia, Actinobacteria, and Proteobacteria main representatives within the murine microbiota (Akkermansia muciniphila, the genus Bifidobacterium, Escherichia coli). Additionally, the Clostridium leptum group as the dominant group of the Firmicutes was quantified.
PCR amplification and detection was performed using an ABI PRISM 7900HT sequence detection system (Applied Biosystems, Darmstadt, Germany). Each reaction mixture (25 μl) was composed of 15 ng fecal DNA, 20 pmol primer mix and the QuantiTect SYBR Green PCR Master Mix (QIAGEN, Hilden, Germany). A standard curve was produced using the appropriate reference organism to quantify qPCR values into number of bacteria per gram wet weight. Real-time PCRs were performed in triplicate. Mean values obtained from the four collection days per mouse were normalized to mean of total numbers of sequences obtained.
Statistical analysis
Testing of statistical significance was performed using one-way ANOVA followed by Bonferroni post-test or by unpaired Student’s t test as indicated (GraphPadPrism 6.0, Graph Pad Software, La Jolla, CA, USA). Values of p < 0.05 were considered statistically significant.
RESULTS
Transient body weight reduction in male 5xFAD mice
Onset of Aβ production has been described in 5xFAD mice as early as at the age of 1.5 months [28]. Therefore, first pathological effects can be expected at this young adult stage. When measuring body weight of 6-week-old male 5xFAD mice, we observed a reduction of 18% in comparison to wild type littermates (p = 0.003; Fig. 1A). This decreased body weight persisted to an age of nine weeks (6% difference to wild type, p = 0.023) and disappeared at 18 weeks of age (p = 0.606). In female mice, we did not detect this transient phenotypic difference between the 5xFAD model mice and wild type littermates (e.g., p = 0.467 for 6 weeks, Fig. 1B). This might be based on hormonal variations even if we did not obtain higher standard deviation values within the female groups as compared to the males. By weighing pups aged six days, we ascertained that the difference in body weight was not due to developmental disability caused by the transgenes. Pups of both sexes were indistinguishable by their weight from respective wild type littermates (Fig. 1C, p = 0.535 for male wild type versus 5xFAD, all other p > 0.999).
Decrease in fecal trypsin amount in early pathological stages of 5xFAD mice
Difficulties in gaining weight could be caused by reduced food intake as well as by hyperactivity of animals. Food intake, feces production, and water content of feces of 5xFAD male mice were not significantly different from the wild type mice (Fig. 2A-C, e.g., p > 0.999 for food consumption at an age of six weeks). Moreover, locomotion was not increased as compared to the age-matched control mice (Fig. 2D, e.g., p = 0.359, age of six weeks).
Although general daily food intake did not differ between wild type and 5xFAD mice, we wanted to further analyze molecular components of feeding behavior, namely carbonic anhydrase VI (gustin, CAVI) and the protease trypsin. A correlation between taste dysfunction and decreased levels of the salivary taste bud trophic protein CAVI has been established in humans [34] and it has also been shown to contribute in bitter taste perception in mice [35]. We measured the content of sputum-derived CAVI in fecal samples [36]. CAVI amount was not differing between the genotypes at 6 or 18 weeks (Fig. 3A and B, p = 0.676 and 0.647) and underlined unaltered appetite consistent with unchanged food intake. For trypsin, it has been shown that it reduces the secretion of the hormone cholecystokinin, which regulates meal patterns in mice (CCK, [37, 38]). The fecal trypsin amount was diminished in young 5xFADanimals to about 40–50% of control (Fig. 3A and B, p = 0.010 and 0.045). This reduced trypsin protein level was also reflected by BAEE-based chromogenic activity assay (Fig. 3C, p = 0.044): trypsin activity was decreased to about 50% in 5xFAD mice as compared to wild types. As seen with body weight, trypsin insufficiency only occurred transiently and was completely restored to wild type level at the age of 18 weeks. Trypsin activity and body weight showed a significant correlation in the male mice (data not shown, r = 0.739, p = 0.006). To make sure that the reduction of this gut enzyme does not depend on reduced synthesis per se, we analyzed trypsin activity and sum of trypsinogen/trypsin activity in pancreatic homogenates (Supplementary Figure 3). Both enzymatic activities were indistinguishable between 5xFAD male mice and wild type littermates at the age of nine weeks.
Potential role of ENS-derived Aβ in trypsin deficiency
The transgenes in the 5xFAD mice (AβPP, PS1) are under control of the Thy1 promoter [28].Enteric nervous system (ENS) neurons are described to be Thy1 positive (e.g., [39]), therefore expression of the transgenes could be assumed. We analyzed the amount of overexpressed human AβPP above background values from gut sections and from total brain of 1- and 8-month-old mice (Fig. 4A, B): while analysis of the colon revealed no statistically significant hAβPP mRNA content (for 1 month, p = 0.664; for 8 months, p = 0.091), all other gut tissue specimen showed elevated signals above background in 5xFAD mice, independently of age. Obviously, the amount of the transgene mRNA is lower in gut sections than in brain, but it has to be taken into account that whole tissue and not isolated ENS material was used and the relative amount of neural tissue in the gut is very low.
Subsequently, we aimed at detecting AβPP-derived proteolysis products in fecal samples of the 5xFAD mice. Soluble brain protein samples from transgenic mice (sAβPPα) and recombinant Aβ peptides served as controls (Fig. 4C). In murine fecal samples, no distinct band for monomeric Aβ or peptide aggregates was detectable. Bands slightly above 98 kDa and at 30 kDa occurred. These might represent sAβPPα and other cleavage products. To makesure that peptide aggregation does not hamper detection of Aβ species, we incubated fecal samples with HFIP which acts as a hydrogen-bond breaker, thereby destroying multimeric aggregates [40]. No bands unequivocally corresponding to Aβ monomers or oligomers were detected, even following HFIP treatment (Supplementary Figure 4). Presumably, the amount of Aβ within the chyme is too low, undergoes degradation, or is ingested by microbiota, hampering detection of the peptide. However, IHC staining of gut showed considerable expression of human AβPP in the 5xFAD mice (Fig. 4D, a sagittal hippocampal section was used as a positive (5xFAD) or negative (Wt) control for antibody 6E10-dependent staining). The muscular layer as well as the myenteric and submucous plexus of the 5xFAD mice showed a comparably strong diffuse signal in the duodenum with darker spots that might indicate aggregates of Aβ peptide. Additionally, the colon of the 5xFAD mice displayed a 6E10-dependent AβPP-specific staining in comparison to the Wt section. The microscopic analysis of HE stained samples meanwhile demonstrated a normal morphology of duodenum and colon (data not shown).
Impact of fibrillized Aβ on trypsin migration in SDS acrylamide gel and enzymatic activity
It has been described that fibrillary Aβ is able to complex trypsin in a SDS-stable manner [41]. This might retain formed complexes at the border of stacking and separation gel during SDS polyacrylamide gel electrophoresis and explain reduced detection of trypsin in 5xFAD mice in our experiments. We therefore pre-incubated trypsin in vitro with monomeric or fibrillized Aβ42 preparations and investigated migration behavior in SDS-polyacrylamide gel. No retention of trypsin via aggregated Aβ was observed by the silver staining procedure (Fig. 5A).Subsequently, we analyzed interference of Aβ peptides with tryptic activity in vitro. While monomeric Aβ42 had no impact on trypsin activity, fibrils or oligomers from overnight preparation inhibited conversion of the pro-chromogenic substrate by about 40% at the higher concentration (Fig. 5B, p = 0.0004). This indicates that Aβ is able to bind to and interfere with trypsin in its oligomerized or fibrillar form, at least in vitro.
Microbiota changes in AD model mice
To assess a potential influence of the AD-like genotype on the murine microbiome, we analyzed selected phyla or strains of bacteria of these mice as compared to wild type littermates in fecal samples. Several bacteria displayed no difference between 5xFAD and wild type animals: for instance, Akkermansia muciniphila was found at approximately 1 to 4% in both groups (Fig. 6A, p = 0.779, six weeks of age) which is in accordance with data observed by others [31]. The group of Bifidobacteria was also indistinguishable in the 5xFAD model mice and wild type littermates at any investigated time point (Fig. 6B, e.g., p = 0.390 at six weeks). Interestingly, within the phyla Bacteroidetes and Firmicutes, which make up the dominant bacterial divisions [42], we observed changes at the age of nine weeks (Fig. 6C): the amount of Firmicutes was increased in the 5xFAD model mice (from 34 to 49%, p = 0.003) while Bacteroidetes were decreased (from 61 to 46%, p = 0.003) in comparison to the wild type littermates. Additionally, the Clostridium leptum group was significantly increased in the 5xFAD mice at the age of nine weeks (Fig. 6C).
DISCUSSION
There is growing evidence that the gut might contribute to or be affected by diseases of the CNS. One example where depositions of pathologically aggregated proteins occur in the gut is Parkinson’s disease [14]. For other neurodegenerative diseases, such as Alzheimer’s disease, information about a potential implication of the gastrointestinal system is rather scarce.
We describe here a body weight reduction in males of the 5xFAD mouse model at a very early time point at the same time with or shortly after onset of Aβ appearance in neuronal cells [28]. We cannot exclude that at later stages of disease development other mechanisms might contribute to body weight determination such as changes in food perception, consumption, or adipogenesis. In 7-month-old 5xFAD mice for example, a 10% reduced weight was observed for both male and female animals [5], which might indicate later CNS-driven changes due to increased plaque load in the murine brain. In a triple-transgenic mouse model (3xTgAD) on the contrary, a greater body weight (34%) has been observed along with greater food intake and unchanged metabolic rate [29]. However, at a later stage (12 months), these mice also displayed reduction of body weight in comparison to littermates (–15%). Extracellular Aβ-deposits in the 3xTgAD mice became apparent at six months [43], while the aggressive 5xFAD model already developed plaques at the age of 6 to 8 weeks [28], which might explain the observed discrepancy to our study. We can only speculate why selectively male 5xFAD animals displayed an early reduced body weight in comparison to wild type littermates: on one hand expression of the transgenes has been reported to be higher in the females due to estrogen-response element of the Thy1 promoter which is used for expression of the transgenes [44]. This might shift the observed phenomenon even to an earlier time point. On the other hand, male murine microbiota shows changes along puberty [45]: while adult female microbiota was reported to stay similar to the prepubescent microbiota, postpubescent male mice revealed a deviating microbial composition of fecal content. This might result in a vulnerable ecosystem that is more open for being injured by the toxic Aβ peptides.
The transient loss of body weight in the male 5xFAD animals could not be explained by enhanced locomotor activity, generalized hypophagia, or fecal composition. In addition, the amount of fecal CAVI was not altered in this AD model mice. CAVI has been assigned to gustatory function [46]; moreover, it is the only protein of salivary origin found in the murine feces, indicating a general function in the whole gastrointestinal tract [36]. In contrast to CAVI, trypsin amounts and activity were diminished in young 5xFAD transgenic mice and only reached wild type levels at an age of 18 weeks. Trypsin activity in feces summarizes a multitude of single steps from enzyme production in pancreatic tissue, activation by enteropeptidase or binding to protein interaction partners. The Thy1 promoter, driver of the transgenes in the 5xFAD mouse model, has not explicitly been described to be active in the pancreas and has been acknowledged to be neuron-specific [47]. However, progenitor cells from adult rat pancreatic ducts have been shown to bear Thy1-positive subpopulations [48] and deposition of brain-derived amyloid and impact on secretory function might occur. Nevertheless, no change in trypsin activity and trypsin production in pancreatic tissue from the 5xFAD mice could be observed. This indicates that the changes in the fecal enzyme’s amount probably must be due to processes of binding and/or degradation and not to insufficiency of the pancreas. Interestingly, trypsin also negatively influences the amount of cholecystokinin which initiates satiety by acting on receptors located on vagal afferent nerve terminals in the gastrointestinal tract wall. Mice with deleted CCK1 receptor for example displayed altered mealpatterns when fed a high fat diet, although daily food was not affected [38]. An imbalanced CCK-driven feeding behavior, besides the reduced proteolytic break-down of nutritional proteins, could be involved in the observed early reduction in body weight of the 5xFAD mice. 3×TgAD mice, for example, consumed more food after a fasting period of 12 h as compared to wild type mice and showed insensitivity to exogenously added CCK [49]. This, together with our data, might hint at a defective gut-brain signaling in Alzheimer’s disease mouse models.
The gut is endowed with its own, quasi-autonomous local neuronal net, referred to as the ENS (e.g., [12]). Those neurons have been described to be at least partly Thy1-positive in rodents[39, 50]; thus, a direct contribution of ENS-derived transgene products on gut function could be hypothesized: quantitation of human AβPP-mRNA in total tissue isolates revealed an elevation above background albeit this did not reach brain levels. This is in accordance with observations of Thy1-regulated α-synuclein overexpression in mice that also resulted in a stronger presence of the respective transcript in brain tissue as compared to ascending colon [51]. The amount of α-synuclein-mRNA was not directly reflected by protein amount which was markedly lower in intensity. Our attempt to demonstrate Aβ peptides in fecal samples from AD model mice also revealed that the abundance of the overexpressed peptide seems to be rather low in the lumen of the gut: we failed to visualize Aβ monomeric or oligomeric peptides despite the occurrence of bands at about 30 and 98 kDa in samples of 5xFAD mice. Recently, an increase of Aβ peptides in aged APP/PS1 mice (13 months) in the intestinal lysate as well as in the stool was described [52]. Mice in our experiments might have been too young to display detectable levels of peptide for western blotting technique. Using IHC, we were able to detect 6E10-positive staining in duodenal tissue as well as in colon. Therefore, the 5xFAD model definitely comprises gastrointestinal expression of the transgenic human AβPP.
Furthermore, we investigated whether Aβ peptides might interfere with gut function: binding of the neurotoxic peptide to trypsin has been previously described in in vitro experiments [41]. We could not verify formation of such SDS-stable complexes; nevertheless, pre-formed fibrils in high concentration reduced tryptic activity by 40%. This has also been reported by Chauhan and colleagues who described a fibril-specific inhibition of trypsin by Aβ peptides to 50% of control [53]. An impact of Aβ on digestion and resorption of food components might be plausible because a co-localization of the peptides with chylomicron-ApoB in small intestine enterocytes of the mouse has been reported [54].
We also analyzed potential changes of fecal microbiota due to the 5xFAD genotype. It is well known that the microbiome’s composition changes along the gastrointestinal tract (e.g., [42]) and that fecal/luminal samples might not always reflect mucosal commensals ([55, 56]). However, feces would be the source of choice for investigating human patients’ microbiome. By analyzing the fecal microbiome of the 5xFAD mice as compared to their wild type littermates, we recognized an increased Firmicutes/ Bacteroidetes ratio at the age of nine weeks, a time point where trypsin deficiency was still detectable (ratio of 34.1% :61% = 0.56 in wild type as compared to 49.2% :46% = 1.07 in 5xFAD animals, p = 0.002). Interestingly, a high Firmicutes to Bacteroidetes ratio is found in obese mice as well as humans [57, 58]. This might be interpreted as a reaction in regard to imbalance of digestive protein break-down. In consequence, this then might lead to the amelioration of reduced body weight of those animals at the age of 18 weeks. Interestingly, mice with neonatal under-nourishing showed a similar phenotype with lower proportion of Bacteroidetes and higher proportion of Firmicutes in the phase of catch-up growth upon re-feeding [59]. Alternatively, Aβ peptides have also been described to reveal antibacterial properties [60] and therefore might have a direct impact on microbiological growth. Recently, it has been shown that Aβ has protective properties against Salmonella typhimurium meningitis in mice and Candida albicans infection of C. elegans [61], indicating a rather wide antimicrobial spectrum. However, analysis of various microbiota revealed that although agglutination by amyloidogenic Aβ species was observed for all tested organisms, efficacy differed (e.g., E. coli: reduction of colony forming units of about 50%; C. albicans: reduction of about 85%, [62]). Therefore, susceptibility of microbiota to Aβ–driven effects might differ and result in changes of their composition. Besides the general changes in the two phyla, we observed a specific transient increase in the Clostridium leptum group, the dominant group of the Firmicutes. Notably, such elevated Clostridium group levels lead to resistance to allergy and intestinal inflammation in experimental mouse models [63]. If such an attenuation of the immune response might reflect a compensatory mechanism in the early phase of AD-related pathological changes of the gut has to be elucidated in the future.
In this regard it is of special note, that integrity of the epithelial barrier of the gastrointestinal tract as well as of the blood-brain barrier is reduced along with aging [64–67]. This could hint at increased entrance of inflammatory modulators or microbial products such as amyloidogenic peptides [68] from the gut into circulation and finally enhanced susceptibility of brain to these compounds and vice versa at later stages of life.
With our investigations, we are not able to answer if Aβ peptides drive the observed microbial changes or if changed digestive capacity subsequently leads to altered microbial growth. We also cannot rule out that presence of mutated presenilin— which gives rise to increase of toxic Aβ species besides the mutated AβPP itself— in this mouse model, is basis for the observed changes. Additionally, other γ-secretase substrates such as Notch1 might contribute to altered intestinal microbiota and function. However, we could show for the first time that expression of proteins with Alzheimer’s disease-linked mutations interferes with gut function on the molecular level at an early state of pathological processes. As nutritional compounds have been shown to severely affect microbiota composition of mice and also humans [69, 70], it will be interesting to analyze if diets such as Western diet or high fat diet might impact the observed changes furthermore.
SUPPLEMENTARY MATERIAL
The supplementary material is available in the electronic version of this article: http://dx.doi.org/10.3233/JAD-160926.
Footnotes
ACKNOWLEDGMENTS
Stiftung Innovation Rheinland-Pfalz funded research of KE and KHS. Funding for MG and TH was provided by Fundació la Maratò de TV3 and by JPND MindAD 1ED1508.
We thank C. Braun (IHC Core facility of the University Medical Center Mainz) for excellent support regarding tissue preparation and staining,U. Schmitt for help with the open field task, and N. Stoye for preparing RNA (both Clinic of Psychiatry and Psychotherapy, University Medical CenterMainz).
