Reduced phosphatidylcholine level in the intestinal mucus layer of prediabetic NOD mice

Type 1 diabetes (T1D) is an autoimmune disease with rising incidence. Pre‐ and manifest T1D is associated with intestinal barrier dysfunction, skewed microbiota composition, and serum dyslipidemia. The intestinal mucus layer protects against pathogens and its structure and phosphatidylcholine (PC) lipid composition may be compromised in T1D, potentially contributing to barrier dysfunction. This study compared prediabetic Non‐Obese Diabetic (NOD) mice to healthy C57BL/6 mice by analyzing the intestinal mucus PC profile by shotgun lipidomics, plasma metabolomics by mass spectrometry and nuclear magnetic resonance, intestinal mucus production by histology, and cecal microbiota composition by 16 S rRNA sequencing. Jejunal mucus PC class levels were decreased in early prediabetic NOD vs C57BL/6 mice. In colonic mucus of NOD mice, the level of several PC species was reduced throughout prediabetes. In plasma, similar reductions of PC species were observed in early prediabetic NOD mice, where also increased beta‐oxidation was prominent. No histological alterations were found in jejunal nor colonic mucus between the mouse strains. However, the β‐diversity of the cecal microbiota composition differed between prediabetic NOD and C57BL/6 mice, and the bacterial species driving this difference were related to decreased short‐chain fatty acid (SCFA)‐production in the NOD mice. This study reports reduced levels of PCs in the intestinal mucus layer and plasma of prediabetic NOD mice as well as reduced proportions of SCFA‐producing bacteria in cecal content at early prediabetes, possibly contributing to intestinal barrier dysfunction and T1D.

Type 1 Diabetes (T1D) is an autoimmune disease caused by destruction of the insulin-producing pancreatic beta cells. The incidence of T1D is rising globally [1] but the exact etiopathogenesis is not known, although both genetic predisposition and environmental factors seem to be implicated [2].
T1D is associated with a dysfunctional intestinal barrier for which increased intestinal permeability and altered microbiota are important factors [3,4]. The intestinal mucus layer serves as the first line of defense against pathogens, but only a few studies have examined the mucus layer as a barrier forming entity in T1D. 1-2% of the mucus consists of lipids, where phosphatidylcholine (PC) is the major lipid component [5,6]. PCs have been shown to increase hydrophobicity and decrease intestinal permeability after intracolonic application in Sprague-Dawley rats, implying that mucus PCs have important roles for the barrier function [7]. The clinical importance of PCs in intestinal mucus is evident from a series of studies on patients with Ulcerative Colitis (UC). Here, the UC patients showed severely reduced levels of PC in their intestinal mucus layer (ileum, colon, rectum) compared to both healthy individuals and Crohns Disease patients [6,8], while treatment with encapsulated PC released in colon significantly improved remission and clinical activity index [9]. PCs are transported paracellularly from blood to the mucus layer where it associates with Mucin 2, but the transport decreases when tight junctions are disrupted [10,11], which is the case in pre-and manifest T1D. This may reduce the PC content in intestinal mucus and thus compromise the barrier function, resulting in leakage of environmental factors such as enterovirus, beta-cell autoimmunity and destruction.
In this exploratory study, we examined the PC class and species levels in the jejunal and colonic mucus layer at different prediabetic ages in Non-Obese Diabetic (NOD) mice vs age-matched control C57BL/6 mice to evaluate if mucus PC levels are reduced in T1D, thereby potentially contributing to T1D-associated intestinal barrier dysfunction and manifest T1D. Because PCs are actively transported into the mucus layer and blood biomarkers are warranted in T1D, we also examined the plasma metabolome with focus on lipid alterations. Moreover, we analyzed intestinal mucus histology and cecal microbiota.

MATERIALS AND METHODS
Expanded materials and methods for mucus lipidomics, plasma metabolomics, and intestinal histology are reported in Appendix S1.

Animals
The Danish Animal Experiments Inspectorate (license number 2016-15-0201-00841) and the local ethical committee (project number P 19-301) approved the animal experiment. The experiment adhered to the Directive 2010/63/ EU of the European Parliament and the Council of 22 nd September 2010 on the protection of animals used for scientific purposes and the Danish Animal Experimentation Act (LBK 474 15/05/2014). Furthermore, this study complied with the ARRIVE guidelines.
NOD/ShiJtL (The Jackson Laboratory, Bar Harbor, ME, USA) and C57BL/6J (The Jackson Laboratory) male and female mice arrived at 6 weeks of age and were acclimatized for 1 week on Altromin 1434 control diet 5c (Altromin, Lage, Germany) prior to breeding (for diet composition, see [12]). This diet was continued throughout the experiment. Providing the mice with the same diet preand postnatally eliminated dietary effects in the offspring. Female offspring from the 1 st generation (NOD) and the 1 st and 2 nd generation (C57BL/6) were used for analysis. Offspring were housed in the specific-pathogen-free facility at the Department of Experimental Medicine (University of Copenhagen, Copenhagen, Denmark). And 2-8 mice were housed in each ventilated cage (12-14 air exchanges/ h, temperature 22°C, humidity 55 AE 10%, 12 h light cycle). Mice were weaned at 3 weeks of age. The mice had free access to diet and bottled tap water. Only nondiabetic (blood glucose level ≤ 12 mmol/L) animals were used for experimental procedures. The sample size was chosen based on results from our previous animal studies.

Mucus, plasma and cecal content sampling
Prediabetic NOD mice and C57BL/6 mice at 4-, 8-, and 13-week of age were fasted for 4 h. Blood was collected by heart puncture during anesthesia (4% isoflurane) and analgesia (I.P. morphine 0.04 mg in 100 lL sterile water) in K2E tubes (BD, Franklin Lakes, NJ, USA) and inverted 10 times before storage on ice (<4 h). The tubes were centrifuged (10 min, 2000 g, 5°C) and plasma stored at À80°C until analysis. Intestine was cut at the duodenojejunal flexure and sectioned into proximal jejunum and colon. Intestinal content was removed by gently pressuring with a metal spatula until the tissue appeared macroscopically clean. The intestinal segments were cut longitudinally, and mucus was sampled by gentle swiping using a metal spatula excluding large enterocyte cell debris from the epithelium while maintaining shredded enterocytes as a natural component of mucus. Mucus and sampled cecal content were immediately snap frozen in liquid nitrogen and stored at À80°until analysis.

Plasma metabolomics
For the NMR analysis, plasma samples (after protein precipitation with cold methanol) were measured on a 600 MHz Bruker Avance III NMR spectrometer (Bruker BioSpin, Rheinstetten, Germany) using a Carr-Purcell-Meiboom-Gill pulse sequence with presaturation. Preprocessed and normalized data were analyzed by a multivariate approach on whole equidistantly binned spectra, and a targeted univariate analysis was based on normalized intensities of all well-resolved proton signals. Corresponding metabolites were identified in ChenomX NMR Suite 8.51 software (Chenomx Inc., Edmonton, AB, Canada) and verified in the Human Metabolome Database (http://www.hmdb.ca) using carbon and proton data extracted from 2D experiments.
For the LC-MS analysis, plasma samples (after internal standard addition and precipitation with acetonitrile/methanol mixture) were analyzed by an Agilent 1200 LC system (Agilent Technologies, CA, USA) on an Intrada Amino Acid HPLC column (150 mm 9 2 mm, 3 lm, Imtakt, Portland, USA). The high-resolution MS (micrOTOF-Q III, Bruker Daltonics, USA) was used as a detector with the electrospray ion source. Data were processed by MZmine software 2.23 V. Metabolites corresponding to the significant features detected by multivariate statistical analysis were identified by Bruker Daltonics software (USA). Predicted molecular formulas were transferred to the CompoundCrowler software to search for structures in the Human Metabolome Database, ChemSpider database (www.chemspider.com), and PubChem (pubchem.ncbi.nlm.nih.gov).

Microbiota analysis
Cecal contents were thawed on ice prior to genomic DNA extraction using the Bead-Beat Micro AX Gravity Kit (A&A Biotechnology, Gdynia, Poland) according to the manufactures protocol. The concentration and purity of extracted DNA was measured using a NanoDrop ND-1000 spectrophotometer (Saveen and Werner AB, Sweden). MiniON (Oxford Nanopore Technologies (ONT), Oxford, UK) 16 S rRNA gene amplicon sequencing was used to define the prokaryotic community. A dilution of 10 ng/lL extracted DNA was used for library preparation. Near full-length 16 s rRNA gene amplicons were amplified and sequenced with ONT targeting V1-V8 hypervariable regions as recently described [17]. Samples that did not reach 10 000 operational taxonomic unit (OTU) reads were removed from analysis.

Intestinal histology
Mice were fasted and water-deprived for 6 h prior to killing by heart puncture. The gastrointestinal tract was cut at the duodenojejunal flexure, whereafter proximal jejunum and colon (including feces content) were fixated in methacarn prior to tissue processing and paraffin embedding. And 4 lm thick sections were cut in areas with visible lumen and histochemically stained using a standard Alcian Blue and Periodic Acid Schiff staining protocol. Slides were scanned using a Nanozoomer S360 (Hamamatsu, Hamamatsu City, Japan) at 209. All measurements were performed blinded using NDP.view2 (Hamamatsu). Jejunal goblet cell count was performed on 6-15 longitudinal cut villi per mouse in single layered areas. The thickness (10-20 regions) and the area (5-10 regions) of the inner mucus layer was examined in distal and proximal colon around a fecal pellet for consistency [18].

Statistics
Unless otherwise stated, statistics were performed in GraphPad Prism version 9.2.0 (La Jolla, CA, USA). p < 0.05 were considered as statistically significant and p < 0.10 were considered a trend. For populations with normal distribution and equal variance, two-tailed unpaired t-test was performed. Otherwise, two-tailed Mann-Whitney U test was performed. Two-tailed multiple unpaired t-tests were performed with 5% FDR using the two-stage step-up method (Benjamini, Krieger, and Yekutieli) to examine differences in abundance of PC species (lipid species with zero abundance was set to 0.0001%, which was <1/10 of the lowest detected lipid species). Statistical analysis for pre-processed NMR and MS data was performed using MATLAB R2021 (The MathWorks Inc., Natick, USA) as well as MetaboAnalyst 3.0 and 4.0 software (Xia Lab, Montreal, Canada, [19]). First, the normalized NMR and MS data were analyzed by principal component analysis (PCA) to detect trends in the sample grouping and to reveal possible outliers (Fig. S1). Subsequently, the univariate approach was applied to describe changes in individual metabolites. Confirmation of data normality for univariate statistics was based on the results of the Lilliefors test. Significance of differences in metabolite levels was evaluated by two-tailed unpaired t-test, for MS data performed with Bonferroni correction. For microbiota analysis, the species-level abundance table was subsampled with 10 000 reads (contained within the most indigent sample) using the GUniFrac package v1.3 [20]. Distance-based redundancy analysis (db-RDA) was performed on Canberra-distances, differences between mouse groups were evaluated with ANOVA-like permutational test, and calculation of diversity richness indexes were carried out using Vegan package v2.5-7 [21]. Changes in relative distribution were determined with MaAsLin2 [22] using log transformation. All microbiota analyses were carried out using R-base v.4.0.4.

Reduced levels of PC species in colonic mucus of prediabetic NOD mice
We examined the PC lipid composition of jejunal and colonic mucus ( Figs. 1 and 2). The distribution of all lipid classes in jejunal and colonic mucus is depicted in Fig. S2A-H, respectively (for exact distribution of lipids see Table S1 for lipid classes and  Table S2 for lipid species). PC was by far the largest lipid class accounting for between 40-48 mol% and 44-46 mol% of all lipids in jejunal and colonic mucus, respectively (Figs 1A and 2A, Fig. S2A,E). We found that the lipid quantity of the whole jejunal PC class was reduced by 12% in NOD mice compared to C57BL/6 mice at 4 weeks of age ( Fig. 1A) but not at 13 weeks. Within the jejunal PC class, the most abundant PC species (mol % > 2) were PC 34:2 (10-20 mol%), PC 36:2 (8-17 mol%), PC 36:3 (2-3 mol%), PC 36:4 (3-8 mol %), and PC 38:4 (3-5 mol%) (Fig. 1B). Differences in the quantity of PC species between NOD and C57BL/6 mice were examined (Fig. 1C). Only PC 30:0 was significantly decreased by 65% in the jejunal mucus layer from NOD mice at 4 weeks, while at 13 weeks PC 40:4 and PC 42:6 were significantly decreased by 48% and 95%, respectively. These lipids were present in very low amounts (<0.074 mol%). Thus, the decrease in PC class quantity at 4 weeks of age in NOD mice was caused by a general decrease of PC species rather than a single PC species. In support of this, the fold change plot (Fig. 1C) revealed that at 4 weeks the PC species levels were generally decreased in NOD mice vs C57BL/6 mice whereas the opposite was observed at 13 weeks.
In contrast to the PC levels in jejunum at 4 weeks of age, there was no difference in the lipid quantity of the PC class in colonic mucus ( Fig. 2A) between NOD and C57BL/6 mice at 4 and 13 weeks. The most abundant PC species in colonic mucus (mol % > 2) were PC 34:1 (5-7 mol%), PC 34:2 (7-11 mol%), PC 36:2 (4-6 mol%), PC 36:3 (4-6 mol %), PC 36:4 (4-6 mol%), and PC 38:4 (3-4 mol%) (Fig. 2B). Despite the similar PC class distribution, we observed many differences between NOD and C57BL/6 mice at PC species level in colonic mucus (Fig. 2C). In total, 18 out of 39 PC species were significantly altered at 4 weeks of age and 13 out of 39 PC species at 13 weeks. Interestingly, most of these differences were due to decreased level of PC species in the colonic mucus of NOD mice compared to C57BL/6 mice at both 4 and 13 weeks of age. Of the most abundant PC species, the PC 34:1 level was increased by 19% in colonic mucus at 4 weeks, while the PC 36:4 level was decreased by 17% at this age. At 13 weeks, both the PC 36:2 level and PC 36:4 level were decreased in the colonic mucus of NOD mice by 20% and 15%, respectively, compared to C57BL/6 mice.
performed untargeted metabolomics (both NMRand MS-based) of the whole spectra on plasma samples. No outliers were detected in any of the PCA models (Fig. S1). The best group separation of NOD and C57BL/6 mice was detected using MS in positive mode at experimental age 4 and 8 weeks (Fig. S1A,B) and using NMR at 8 weeks of age (Fig. S1H). MS detected 356 peaks as potential metabolites, whereof 13 metabolites were significantly altered (Fig. 3A-I,J(lower)). NMR quantified 34 metabolites of which 19 were significantly altered between NOD and C57BL/6 mice (Fig. 3J(upper)). Several plasma lipid changes were found between NOD and C57BL/6 mice at different ages ( Fig. 3A-I). The following PC species were significantly altered between NOD and C57BL/6 mice: PC 34:1, PC 36:2, PC 36:4, and PC 38:6 ( Fig. 3E-H). PC 36:2 was increased in NOD mice at 4 weeks of age by 20%, whereas PC 34:1, PC 36:4, and PC 38:6 was decreased in plasma from NOD mice at 4 weeks by 14%, 4 and 8 weeks by 26% and 15%, and 4 weeks by 40%.
The levels of the saccharides sucrose, glucose, mannose, and arabinose were generally decreased in the prediabetic NOD mice (range À12% to À45%), equivalent to the measured fasting blood glucose levels at these ages (data not shown). As expected, ketone bodies were significantly increased at 4 and 8 weeks by 110% and 217% for 3hydroxybutyrate as well as 192% and 582% for . Light blue = NOD mice, 4 weeks, beige = C57BL/6 mice, 4 weeks, blue = NOD mice, 13 weeks, red = C57BL/6 mice, 13 weeks. (C) */ †q < 0.05; **/ † †q < 0.01; † † †q < 0.001 (*4 weeks, †13 weeks). Statistical analysis was performed by two-tailed Mann-Whitney U test (A (4, 13 weeks)) or multiple unpaired t-tests performed with 5% false discovery rate (FDR) using the two-stage step-up (Benjamini, Krieger, and Yekutieli) method (C (4, 13 weeks)). Data are shown as means AE SD (A, B) or log 2 (fold change) between NOD and C57BL/6 mice (C). acetone in NOD mice and also the levels of the branched chain amino acids (BCAAs) valine, isoleucine, and leucine were significantly increased in NOD mice at most of the ages (range 32.3% to 57.3%). Carnitine levels were significantly decreased at 4 and 8 weeks of age in NOD vs C57BL/6 mice by 29% and 27%, respectively. Acetyl-and valerylcarnitine levels were significantly increased in NOD mice at 4 and 8 weeks of age by 37% and 99%, respectively (Fig. 3J). In general, the metabolite differences between NOD and C57BL/6 mice alleviated at 13 weeks, except for saccharides and BCAAs.

No alterations in intestinal mucus histology between prediabetic NOD mice and C57BL/6 mice
We counted the number of goblet cells as an estimate of small intestinal mucus production (Fig. 4A). Goblet cell count was similar between the groups, suggesting equal small intestinal mucusproduction in prediabetic NOD and age matched C57BL/6 mice. The inner colonic mucus layer was quantified in proximal and distal colon. In proximal colon, the inner mucus layer was thin and no differences between NOD and C57BL/6 mice were found for the inner mucus area and the thickness   (Fig. 4B,D). There was a tendency toward a lower area of the inner mucus layer in the distal colon at 13 weeks of age in the NOD mice compared to C57BL/6 (Fig. 4C). No differences were found in the mucus thickness in the distal colon (Fig. 4E).
The cecal microbiota composition differ between prediabetic NOD mice and C57BL/6 mice The composition of cecal microbial communities were analyzed (Fig. 5A-F). The most dominant phylum in all groups (each strain at each age) was Firmicutes accounting for 85-92% of the relative distribution (Fig. 5A), whereas the remaining phyla were Actinobacteria (2.8-6.2%), Bacteriodetes (0.8-6.1%), Tenericutes (0.08-0.8%), Verrumicroba (0.6-10%), and Proteobacteria (0.2-1.0%) (Fig. 5A and Tables S3 and S4 for OTU table). Figure 5B shows family level distribution of bacteria in all groups. Bacteria species evenness (Fig. 5C) and richness (Fig. 5D) between the mouse groups at 4, 8, and 13 weeks of age were similar. However, a tendency toward a higher species richness in NOD mice at 13 weeks was found (Fig. 5D). To examine differences in microbial communities between NOD and C57BL/6 mice, a db-RDA analysis was performed revealing significantly different bacterial communities (b-diversity) between groups with significant ageinteraction, causing separation on the principal coordinates analysis plot (p = 0.01, interaction and third (CAP3) axis from distance-based redundancy analysis (db-RDA) visualizing differences in bacterial communities found significant based on Canberra taxonomic distances (p = 0.01 with ANOVA-like permutational test between groups with significant interaction (age, p = 0.032)), (F) heatmap visualizing significant changes in relative distribution between groups and with age at species taxonomy level from low relative abundance to high. For visualization purposes, all species have been row normalized to adjust all rows to a common scale. Brackets depict the percentage of the lowest and highest relative abundance for each species. (A-F) NOD 4, 8, 13 weeks n = 4, 6, 4; C57BL/6 4, 8, 13 weeks n = 5, 6, 5. White squares and white bars = NOD mice; filled circles and gray bars = C57BL/6 mice. p < 0.10 indicates a tendency. Statistical analysis was performed by two-tailed unpaired t-test (C, D (4, 8, 13 weeks)). Data are shown as means AE SD (A-D). age * mouse group p = 0.032) (Fig. 5E) with CAP1 and CAP3 giving the best 2D visualization. A heatmap showing the relative abundance of the species that caused the difference in b-diversity is shown in Fig. 5F. Species belonging to the family Coriobacteriaceae as well as Anaerotruncus spp. accounted for the differences with age, whereas the remaining bacterial species on Fig. 5F accounted for differences between mouse strains. A member of the family Clotridisceae, as well as Allobaculum spp. and Turicibacter spp. were absent in NOD mice, while Anaerotruncus spp. were absent in C57BL/6 mice.

DISCUSSION
This study examined the PC mucus lipidome in jejunum and colon of prediabetic NOD mice compared to C57BL/6 mice. In NOD mice, the level of the PC class was reduced in jejunal mucus at 4 weeks of age, whereas in colon the levels of several PC species were reduced at 4 and 13 weeks. Plasma metabolomics supported the findings of dyslipidemia in intestinal mucus showing dysregulation of several PC lipid species along with signs of increased b-oxidation in NOD mice at 4 and 8 weeks of age. No differences were observed in mucus histology in either jejunum or colon, but the cecal microbiota showed altered b-diversity between the mouse strains.
The observed reduction of PC class levels in jejunal mucus of 4-week-old NOD vs C57BL/6 mice could imply a reduced local surface hydrophobicity and thus impaired barrier function at early pre-T1D [7], although further studies are needed to test this hypothesis. Aberrant mucus barrier function is likely an important part of the intestinal barrier dysfunction in T1D [3], but the research area is still in its early phase. While studies of the intestinal permeability in T1D have primarily been directed the small intestine, the colonic barrier function increasingly gains interest due to the observed diabetes-alleviating effect of short-chain fatty acid (SCFA)-inducing fiber-diets in NOD mice [3]. In colonic mucus, we observed that the level of several PC species was reduced in NOD mice at 4 and 13 weeks of age compared to C57BL/6 mice, but the effect was not observed on the PC class level. This included PC 36:4, which was previously shown to be reduced in serum of prediabetic individuals [23]. The mechanism behind the changes in the level of the PC lipid class and species is currently unknown, but the lack of difference in jejunal goblet cell count and colonic mucus area and thickness suggest that they were not caused by changes in mucus production. The hydrophobic properties of mucus PCs are well established [7,11,24], but knowledge about the biological function of specific PC species is sparse. Still, some studies have found that specific species and a cocktail of PCs exert anti-inflammatory effects, among others by an antagonistic effect to Toll Like Receptor 4 (TLR4) [25]. This is relevant as TLR4 is upregulated in islets of longstanding T1D patients vs healthy individuals [26] and inhibition of TLR4 alleviates autoimmune diabetes in NOD mice [27]. We recommend that the therapeutic potential of the specific PC species that are reduced in pre-T1D is tested in NOD mice.
To complement the observed changes in the mucus lipidomic profile, we analyzed the plasma metabolome. PC 34:1, PC 36:2, PC 36:4, and PC 38:6 were significantly altered in plasma of prediabetic NOD mice compared to C57BL/6 mice, which was also the case in colonic mucus, but only PC 36:4 and PC 38:6 showed decreased levels in both plasma and colonic mucus. Thus, the PC lipid composition in plasma and mucus does not match entirely, which might be caused by a preferred integration of specific species in the mucus layer [28]. The observed decrease of PC 36:4 in plasma from NOD mice was striking, as it was also decreased in the colonic mucus layer of NOD mice at both 4 and 13 weeks of age and is one of the most common lipids in both the jejunal and colonic mucus layer. Furthermore, PC 36:4 is the only specific lipid species that was decreased in blood in three different studies of prediabetic vs healthy individuals [23]. Further establishing a role for PC 36:4 in T1D, the Diabetes Autoimmunity Study in the Young (DAISY) study found that PC 36:4, as well as PC 34:1, PC 32:1, and PC 30:0, was positively associated with the reversal from autoantibody positive to negative in young children at risk of T1D [29]. Thus, PC 36:4 (and other PC species) potentially has a role in T1D development and accordingly, functional studies of these are warranted.
The combination of increased levels of BCAAs and decreased levels of the gluconeogenic amino acid alanine, as observed in our study, have also been seen in individuals with pre-T1D [30,31] and animal models of T1D [32]. BCAA levels were previously found negatively associated with beta-cell function in T1D patients as assessed by C-peptide levels [33], suggesting that already in the prediabetic phase, the NOD mice may have beta-cell dysfunction. At 8 weeks of age, the acyl-carnitines, acetylcarnitine, and valerylcarnitine were increased compared to the C57BL/6 mice, suggesting increased b-oxidation in the prediabetic NOD mice, possibly because of insulin-deficiency and shift to an alternative energy source than glucose. This would also match the severely increased ketone body levels observed in the NOD mice.
Some of the specific bacteria that caused the difference in b-diversity between the mouse strains in our study (Fig. 3I) were previously associated with human T1D. These studies reported increased Bifidobacterium adolecentis [34], decreased Turicibacter [35], both de-and increased levels of specific Clostridiaceae species [34], and decreased Ruminococcaceae [36,37] in human pre-and manifest T1D vs healthy controls. Turicibacter levels showed positive correlation with fecal concentration of butyric acid in Wistar rats [38]. Anaerotruncus levels negatively correlated with cecum/colonic fecal propionic acid and butyric acid concentrations in piglets [39], and in a streptozotocin-induced T1D rat model the SCFA-producing genus Allobaculum was reduced compared to control rats [40]. Propionic and butyric acid are SCFAs, which are produced from bacterial fermentation of non-digestible fibers in colon. SCFAs have positive effects on the intestinal homeostasis as they function as energy sources for colonocytes, stimulate the mucus production, and regulate the immune system promoting regulatory T cells [41]. SCFA-yielding diets reduce autoimmune diabetes incidence in NOD mice [42], and the microbiome from healthy children vs that from children with T1D contains more genes related to fermentation and hence SCFA production [43]. Interestingly, in our study, Turicibacter and Allobaculum was completely absent in cecal content from the NOD mice, whereas Anaerotruncus was absent in the C57BL/6 mice. This suggests that SCFAconcentrations are decreased in the prediabetic NOD mice, but unfortunately, we were not able to detect SCFA in our plasma metabolomics analysis, preventing confirmation of the hypothesis. How the microbiota might be related to mucus lipids is currently unknown. However, given the impact of SCFA on intestinal mucus, the microbiota composition might also be regulating mucus lipid contents. Bacteria probably do not contribute significantly to the mucus PC content, as only about 10% of bacteria are able to produce PC [44]. Still, some bacteria can reside in the mucus layer and might contribute with other lipid classes than PC, thus modulating the distribution of lipids.
The strengths of this study are the state-of-theart methodologies that were used to examine the mucus lipidome, plasma metabolome, and microbiota composition. We also included different early ages for a detailed understanding of the intestinal mucus barrier in the prediabetic phase. Moreover, mice were bred on the same diet and in the same stable to exclude differences induced by variance in diets. The weaknesses include low number of samples in some of the experiments and not analyzing the origin (e.g. Muc2-associated PCs, microorganisms, or shredded enterocytes) of the observed differences in lipid levels.
In conclusion, we demonstrate that prediabetic NOD mice have reduced PC class levels in the jejunal mucus layer and reduced PC species levels in the colonic mucus and plasma. This may lead to reduced hydrophobicity and anti-inflammatory signaling, which should, however, be verified in functional studies. Particularly, the association between reduced plasma PC 36:4 levels and T1D development would be highly relevant to further explore for biomarker and therapeutic purposes. Lastly, we found altered b-diversity in the cecal microbiota between prediabetic NOD and C57BL/6 mice, thus warranting studies on the impact of microbiota on mucus lipid composition.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Appendix S1. Expanded materials and methods. Figure S1. Plasma metabolomics PCA. Figure S2. Lipid profile of intestinal mucus. Table S1. Lipid class distribution in jejunum and colonic mucus. Table S2. All detected lipid species in jejunum and colonic mucus. Table S3. OTU taxonomy table. Table S4. OTU table based on sequencing of 16S V1-V8 hypervariable regions in cecal feces in NOD and C57BL/6 mice.