Age-associated Impairment of the Mucus Barrier Function is Associated with Profound Changes in Microbiota and Immunity

Aging significantly increases the vulnerability to gastrointestinal (GI) disorders but there are few studies investigating the key factors in aging that affect the GI tract. To address this knowledge gap, we used 10-week- and 19-month-old litter-mate mice to investigate microbiota and host gene expression changes in association with ageing. In aged mice the thickness of the colonic mucus layer was reduced about 6-fold relative to young mice, and more easily penetrable by luminal bacteria. This was linked to increased apoptosis of goblet cells in the upper part of the crypts. The barrier function of the small intestinal mucus was also compromised and the microbiota were frequently observed in contact with the villus epithelium. Antimicrobial Paneth cell factors Ang4 and lysozyme were expressed in significantly reduced amounts. These barrier defects were accompanied by major changes in the faecal microbiota and significantly decreased abundance of Akkermansia muciniphila which is strongly and negatively affected by old age in humans. Transcriptomics revealed age-associated decreases in the expression of immunity and other genes in intestinal mucosal tissue, including decreased T cell-specific transcripts and T cell signalling pathways. The physiological and immunological changes we observed in the intestine in old age, could have major consequences beyond the gut.

differences in the H&E stained colon (Fig. 1A,B) and ileum (Fig. 1C,D) sections of 19-month-old and 10-weekold mice. However, in 19-month-old mice the villi were significantly (P < 0.001) longer (Fig. 1E) and there were fewer Paneth cells than in younger mice (Fig. 1F,G). Solitary intestinal lymphoid tissue (SILT) structures 42 were more frequently observed in the colonic mucosa of 19-month-old mice than 10-week-old mice. However, SILT structures are known to develop postnatally at various locations throughout the colon so it is not possible to draw firm conclusions without use of a technique which examines the whole intestine. These lymphoid structures comprised mainly of CD45-positive haematopoietic cells and some CD3-positive T-cells (Fig. 2). Immunostaining for cleaved caspase 3 and mucus-containing goblet cells revealed increased apoptosis of cells in the upper part of the colonic crypts of 19-month-old mice than 10-week-old mice, but no obvious reduction in total goblet cell numbers in the colon (Fig. 3A,B). A proportion of goblet cells in the in colonic crypts of 19-month-old mice but not in 10-week-old mice stained positive for cleaved caspase 3 which may account for the reduced thickness of secreted mucus in old mice (Fig. 3C) was observed to counter balance the apoptosis of goblet cells (Fig. 3G,F). These findings suggested that aged mice would produce less secreted mucus than young mice due to the presence of fewer vital goblet cells. The 19-monthold mice had significantly fewer goblet cells in the ileum than 10-week-old mice (Fig. 3D,E,H). The colonic mucus layer is thinner or absent in aged mice resulting in a failure to spatially compartmentalise the microbiota to the intestinal lumen. Alcian blue-staining was used to identify acidic carbohydrates like Muc2, and Periodic Acid Schiff (PAS) for neutral carbohydrates, both of which occur on the Muc2 glycoprotein. Previous studies showed that the inner mucus layer of the proximal colon of young wild-type mice forms a stratified layer of Muc2 mucin, physically separating bacteria from the epithelium. The colonic mucus layer was absent or much thinner (1-5 µm) in 19-month-old mice than 10-week-old mice   ( Fig. 4A-G). Additionally, we observed that mucus layer was absent over the epithelium covering the SILT which was more abundant in the colon of old mice than young mice. This was due to the lack of mucus-secreted goblet cells in this area (Fig. 4H). The effect of ageing on the spatial compartmentalisation of bacteria in both small intestine and colon was investigated by fluorescent in situ hybridization (FISH). In aged mice the thinner mucus layer was associated with increased bacterial penetrability and contact with the epithelium. For example, in small intestine and colon of young mice we observed a clear "gap" of about 50 μm in between the microbiota and the epithelium (Fig. 5A,C). This "gap" corresponded to the thick mucus layer between the surface of the epithelium and the luminal content (Fig. 4). However, at many locations in the small intestine and colon of 19-month-old mice the microbiota was frequently observed in direct contact with epithelial surfaces, due to lacunas in the mucus layer, which was never observed in young mice (Fig. 5B,D).
Alterations in innate and adaptive immune responses are observed in the large and small intestine of ageing mice. In the large intestine, a transcriptomics approach was used to gain more insight into the potential pathways and mechanisms that might be modulated by the dysfunctional barrier functions observed in aged mice (above). Applying the following criteria, fold-changes >1.2, p-value < 0.05 and signal intensity >20 in at least one of the arrays, we found 1503 differentially expressed genes (759 up-regulated and 744 down-regulated) in old mice versus young mice (data not shown).
A Heatmap was generated for immunity-related genes, including those that encode Pattern Recognition Receptors (PRRs), cytokines, chemokines, immunoglobulins, antimicrobial (poly)peptides, T cell markers (CD3ε), and T-helper (CD4 or CD8) and Tregs (Foxp3) subsets that were differentially expressed in the proximal colon of 19-month-old mice (Fig. 6A). The down-regulated Cd3ε, Cd4, and Cd8 suggest a decreased abundance of T cells. The strong down-regulation of chemokine genes and immunoglobulin expression suggest a decrease in immune responsiveness and B cell activity.
In the ileum 930 genes were found differentially expressed (428 up-regulated and 502 down-regulated) in old versus young mice (data not shown). The Gene Ontology Biological Processes annotations of the differentially expressed genes showed that predominantly processes related to adaptive immunity were strongly down-regulated in the small intestine. The most down-regulated pathways were CTLA4 Signalling in Cytotoxic  Fig. 1). The antimicrobial protein angiogenin 4 (Ang4) that is produced by Paneth cells was strongly down-regulated in the ileum of old mice. A similar Heatmap as for colon was generated for immunity-related genes including those that encode Pattern Recognition Receptors (TLRs etc.), cytokines, chemokines, immunoglogulins, antimicrobial (poly)peptides, T cell markers (CD3ε) and T-helper (CD4 or CD8) and Tregs (Foxp3) subsets that were differentially expressed in the ileum of 19-month-old males (Fig. 6B). As in colon, genes related to innate and adaptive immunity were strongly down-regulated in old mice compared to young mice.
Ageing is associated with altered intestinal microbiota. To investigate the impact of ageing on the colonization pattern of the colon, 16S rRNA microbiota profiles of faeces from 10-week-, 8-month-, 13-month-, 15-month-, and 19-month-old C57BL/6 mice were determined using the MITChip microarray 43 . The faecal content of 15 -and 19 months-old mice displayed both significantly increased richness and α-diversity compared to young mice (P = 0.036 and P = 0.0014, respectively), (Fig. 7). The intermediate ages (8 and 13 months) displayed similar diversity and richness as the young mice, although diversity was statistically higher at 8 months compared to 10-week-and 13-month-old mice (Fig. 7). Nevertheless, redundancy analysis (RDA) clearly established that at 8, 13, 15, and 19 months of age the microbiota composition was clearly distinct compared to 10 week-old mice, with the exception of one mouse in the 19-month-old group, causing a partial overlap of the microbiota cluster with 8-month-old mice (Fig. 8). Akkermansia muciniphila, Porphyromonas asaccharolytica et rel., Collinsella, Corynebacterium et rel., and Lactobacillus gasseri et rel., were statistically less abundant (P < 0.05 with correction for multiple testing) in old mice compared to young ones (Fig. 8). Lachnospira pectinoschiza et rel., and Butyrivibrio crossotus et rel., were significantly more abundant in old mice (P < 0.05 with correction for multiple testing), (Fig. 8).
Abundance of specific bacteria correlates with expression of genes in ageing mice. To identify bacteria that are correlated with gene expression changes in the colon of aged mice, we performed a multivariate integration and correlation analysis on the data from 19-month-old mice. Microbiota and transcriptomics data were pooled for individual mice to investigate direct correlation between gene expression and microbiota composition over these samples. Three bacterial clusters (A, B and D) strongly correlated either positively (red) or negatively (blue) with 4 clusters of genes (1, 2, 4 and 5; about 100 genes per cluster) (Fig. 9). The correlations between bacteria and cluster C were not included because they were relatively much weaker. The individual genes in the respective gene clusters are listed in Supplementary, Table 1. Strongest correlations were between microbiota members (cluster D on y axis, Fig. 9) and specific changes in mucosal gene expression (cluster 5 on x-axis), shown in Fig. 9. The bacteria of cluster D had higher relative abundances in the mice of 15-and 19-month-old mice compared to young mice and displayed a negative correlation with immune response genes involved in apoptosis that were differentially expressed in our transcriptomics data from 19-month vs 10-week old mice (Supplementary Data, Table 1, cluster 5 genes in bold). A positive correlation was found between bacteria in cluster D and genes listed in cluster 2, however transcriptomics data shown no significant differences between the relative expression of these genes in old and young mice.

Discussion
In this study, we showed that 19-month-old male mice have absent or greatly reduced thickness of the colon mucus layer compared to 10-week-old mice (Figs 4 and 5). A decrease of about 5 μm in the thickness of colonic mucus was reported previously in 38 week-old rats compared to animals aged 4 weeks or 8 weeks 41 , but this reduction was much smaller than we observed in very old mice (19 months). The reduction of mucus secretion in the colon of 19-month-old mice is likely to be due to the striking increase in apoptotic goblet cells, in the upper part of the colonic crypts (Fig. 2). Despite the impairment of the mucus barrier in old mice and chronic exposure of the epithelial surface to luminal bacteria (Fig. 4), there was no histological evidence of colitis (Fig. 1). However, in Muc2 −/− mice which lack a secreted mucus layer spontaneous colitis occurs around 5 to 6 weeks of age. Taken together these observations suggest that production of a small amount of intestinal mucus is sufficient to avoid colitis development, even when the microbiota is not compartmentalised to the lumen. This may be due to proposed immune regulatory effects of mucin itself and, or its crucial role in lubrication of faecal passage and the avoidance of mechanical stress on the gut wall. Another factor influencing the risk of colitis may be the age-related down-regulation of host inflammatory immune responses as evidenced by the transcriptomics study  on colon and ileal tissue. Transcription of genes related to innate and adaptive immunity such as immunoglobulins (in particular IgA) were strongly down-regulated in both the colon and ileum of old mice (Fig. 6A,B).
In the colon SILT structures were observed more frequently in 19-month-old mice than 10-week old mice. SILT are known to develop early after birth and are dependent on exposure to intestinal microbiota 44 . Thus, increased contact with intestinal microbiota induced by ageing and mucus shrinkage might explain the more frequent occurrence of colonic SILT in old mice. The SILT were mainly composed of B-cells confirming the results of a previously published study 42 .
The mucus barrier also appeared to be compromised in the ileum of 19-month-old mice and bacteria were frequently seen in contact with the epithelium (Fig. 4). As found in the colon, the transcriptome data revealed a down-regulation of innate and adaptive immune genes including immunoglobulins (in particular IgA), TLR4, T cell (CD3ε) and T-helper (CD4 or CD8) markers and the antimicrobial factor Ang4 were strongly down-regulated in both the colon and ileum of old mice. Histology showed that the Paneth cell marker lysozyme was also strongly down-regulated in the ileum. Secretory IgA and antimicrobial factors play a key role in regulating contact between the epithelium and potentially harmful antigens and microbes 31,32 and may explain our observation that the microbiota were frequently seen in contact with the villus epithelium (Fig. 5). Decreased mucus production in the ileum of old mice may also be a contributing factor. Although we did not measure any significant age-associated changes in Muc2 transcription decreased production in old mice could be due to altered post-translational processes. The only mucin gene differentially expressed in the ileum of old mice was the membrane tethered mucin Muc1, which was down-regulated. Interestingly the thickness of the secreted colonic mucus is decreased in Muc1 −/− mice, despite there not being a decrease in Muc2 transcription compared to wild-type mice 45 .
Reg4 was of the few gene transcripts expressed in relatively higher amounts in the colon and ileum of old mice. The function of this gene is not well understood but is known to be up-regulated in cancer-initiating cells and was recently shown to be a specific marker for enteroendocrine cells in the intestine.
In previous studies, analyses of human microbiome in the elderly (>65 years) have revealed significant changes in the intestinal microflora specifically with an increase of Bacteroides ssp and distinct abundance patterns of Clostridium groups 26,27 . However, others have shown that the change in the microbiota was seen only in centenarians with increased inflammatory cytokine responses, but not in the general elderly population (average age 70 ± 3 years) 28 . To identify bacteria that might be correlated with changes in colon gene expression we used the linear multivariate method partial least squares (PLS) method 46 for each time point, as previously described  (Lange et al., in revision). We found that uncultured Clostridiales and Lachnospira pectinoschiza et rel. had higher relative abundances in the old mice of 15 and 19 months of age compared to young mice and displayed a positive correlation with immune response genes. A positive correlation was also found with stress response genes involved in apoptosis and cell proliferation, as well as immune genes. Akkermancia municiphilia was strongly decreased with ageing which has been observed in humans. Akkermansia muciniphila is a Gram negative bacterium, which in mice is the only species belonging to the phylum Verrucomicrobia 47 . Oral administration of obese and type 2 diabetic mice with viable A. muciniphila reversed high-fat diet-induced metabolic disorders, including fat-mass gain, metabolic endotoxemia, adipose tissue inflammation, and insulin resistance 48 . Furthermore, it has been shown that extracellular vesicles produced by A. muciniphila decrease expression of IL-6 in colonic epithelial cells and when orally administered to mice protect against DSS-induced colitis 49 .
In summary we showed that old mice have an impaired mucus barrier in the colon and ileum accompanied by major changes in the faecal microbiota composition and expression of immunity and other genes in intestinal mucosal tissue, leading to decreased T cell-specific transcripts and T cell signalling pathways. Given the important role of gut homeostasis and the microbiota in the aetiology of many diseases, the physiological and immunological changes we observed could have major consequences beyond the gut. An age-associated decline in antigen specific immune responses to oral vaccination has been reported to occur earlier than in the systemic immune compartment. Thus, the decline in gut barrier function may even be the trigger for the low-grade chronic inflammation or "inflammageing" typical of old age. In the future, prospective studies are needed to determine the sequence of events leading to the profound changes we observed in the intestinal mucosa of old mice as this will be important for future strategies aimed reducing the risk of infectious disease, colon cancer and potentially other diseases.

Methods
Animals and ethical approval. C57BL/6 mice (Harlan Laboratories, USA) were housed in a specific pathogen-free environment with ad libitum access to D12450B diet (10% fat) (Research Diets Services BV, Wijk bij Duurstede, the Netherlands), and acidified tap water in a 12-hour light/dark cycle. All animal experimental protocols were approved by the UMCG Animal Ethics Committee (Groningen, The Netherlands), and carried out in accordance with the approved guidelines.
Experimental set up. Groups of 8-week-old males (n = 5) were housed in individual ventilated cages, and sacrificed at 2 months and 19 months of age. Ileal and colonic tissues were fixed in Carnoy's fixative and embedded in paraffin as previously described 33 . Additionally, segments of ileum and colon were frozen in liquid nitrogen and stored at −80 °C for RNA and protein assays. Faecal samples were collected at 2, 8, 13, 15, and 19 months (sacrifice) and stored at −80 °C.
Histology. Paraffin sections (5 µm) of ileum and colon were attached to poly-L-lysine-coated glass slides (Thermo scientific, Germany). After overnight incubation at 37 °C, slides were de-waxed and hydrated step-wise using 100% xylene followed by several solutions of distilled water containing decreasing amounts of ethanol. Sections were stained with hematoxylin and eosin (H&E) and PAS/Alcian blue 50  Bionanalyzer 2100 (Agilent, Santa Clara, CA, USA), respectively. RNA was only used to generate cDNA and perform microarray hybridisation when there was no evidence of RNA degradation (RNA Integrity Number >8). 100 ng of total RNA was labelled using the Ambion WT Expression kit (Life Technologies Ltd, Paisley, UK) together with the Affymetrix GeneChip WT Terminal Labelling kit (Affymetrix, Santa Clara, CA). Labelled samples were hybridised to Affymetrix GeneChip Mouse Gene 1.1 ST arrays. Hybridisation, washing, and scanning of the array plates were performed on an Affymetrix GeneTitan Instrument, according to the manufacturer's recommendations. Quality control (QC) of the datasets obtained from the scanned Affymetrix arrays was performed using Bioconductor 51 packages integrated in an on-line pipeline 52 . Probe sets were redefined according to Dai et al. 53 utilising current genome information. In this study, probes were reorganised based on the Entrez Gene database (remapped CDF v14.1.1). Normalised expression estimates were obtained from the raw intensity values using the Robust Multiarray Analysis (RMA) pre-processing algorithm available in the Bioconductor library affyPLM using default settings 54 .
Differentially expressed probe sets were identified using linear models, applying moderated t-statistics that implemented empirical Bayes regularization of standard errors 55 . A Bayesian hierarchical model was used to define an intensity-based moderated T-statistic (IBMT), which takes into account the degree of independence of variances relative to the degree of identity and the relationship between variance and signal intensity 56 . Only probe sets with a fold-change (FC) of at least 1.2 (up/down) and p value < 0.05 were considered to be significantly different. Biological interaction networks among regulated genes activated in response to ageing were identified using Ingenuity Pathways Analysis (IPA) (Ingenuity System). IPA utilizes a large expert-curated repository of molecule interactions, regulatory events, gene-to-phenotype associations, and chemical knowledge, mainly obtained from peer-reviewed scientific publications, that provides the building blocks for network construction. IPA annotations follow the GO annotation principle, but are based on a knowledge base of >1,000,000 protein-protein interactions. The IPA output signalling pathways with statistical assessment of the significance of their representation based on Fisher's Exact Test. Our IPA analyses compared differentially regulated genes in the ileum and colon of old males compared to young mice. The input was all differentially regulated genes (p value < 0.05, FC > 1.2 and intensity >20) of ileum and colon.
Bacterial DNA extraction and microbiota profiling. The DNA from ileal content and faeces was extracted using the repeated bead-beating-plus column method 57 . Microbiota composition was analysed by Mouse Intestinal Tract Chip (MITChip), a diagnostic 16S rRNA array that consists of 3,580 unique probes especially designed to profile mouse intestine microbiota 43 . 16S rRNA gene amplification, in vitro transcription and labelling, and hybridization were carried out as described previously 58 . The data was normalized and analysed using a set of R-based scripts in combination with a custom-designed relational database, which operates under the MySQL database management system. For the microbial profiling the Robust Probabilistic Averaging (RPA) signal intensities of 2667 specific probes for the 94 genus-level bacterial groups detected on the MITChip were used for the relative abundant measurements 59 . Diversity calculations were performed using a microbiome R-script package (https://github.com/microbiome), Shannon indices were plotted in a Box-and-whisker plot and statistical difference were tested by paired t-test between time-points. Wilcoxon signed-rank test was applied on the 94 genus-level bacterial groups to determine significant differences of individual bacterial groups between the 10-week-old and 19-month-old mice (Fig. 8B); when corrected for multiple testing all p values were higher than 0.05. Multivariate statistics, redundancy analysis (RDA), and Principal Response Curves (PRC), were performed in Canoco 5.0, and visualized in triplots or a PRC plot, and variables were tested for their significance in explaining the variation by Monte Carlo permutation test 60 .

Multivariate integration and correlation analysis. To get insight into the interactions between changes
in gene expression and microbiota composition, the datasets per time point were combined using the linear multivariate method partial least squares (PLS) 46 , as described before (Lange et al., in revision). By this integration, the two datasets were integrated per individual mouse. Both datasets were log2 transformed before analysis and the canonical correlation framework of PLS was used 61 . The correlation matrices were visualized in clustered image maps 62 . Analyses were performed in R using the library mixOmics 63 .