Frontiers

titanium dioxide yeast
2022 04-02 05:31
分享

Frederick National Laboratory for Cancer Research, National Cancer Institute at Frederick (NIH), United States

Faculty of Pharmacy, University of Belgrade, Serbia

China Agricultural University, China

University of Texas MD Anderson Cancer Center, United States

The editor and reviewers affiliations are the latest provided on their Loop research profiles and may not reflect their situation at the time of review.

Impact of the Food Additive Titanium Dioxide (E171) on Gut Microbiota-Host Interaction

The Charles Perkins Centre, The University of Sydney, Sydney, NSW, Australia

Faculty of Medicine and Health, School of Medical Sciences, The University of Sydney, Sydney, NSW, Australia

Sydney Nano Institute, The University of Sydney, Sydney, NSW, Australia

Human Health, Nuclear Science Technology and Landmark Infrastructure (NSTLI), Australian Nuclear Science and Technology Organisation, Sydney, NSW, Australia

School of Medical Sciences, University of New South Wales, Sydney, NSW, Australia

Department of Cardiology, Charles Perkins Centre, Royal Prince Alfred Hospital, Heart Research Institute, University of Sydney, Sydney, NSW, Australia

Sydney Pharmacy School, The University of Sydney, Sydney, NSW, Australia

School of Science, RMIT University, Bundoora, VIC, Australia

School of Health, Medical and Applied Sciences, Central Queensland University, Rockhampton, QLD, Australia

The interaction between gut microbiota and host plays a central role in health. Dysbiosis, detrimental changes in gut microbiota and inflammation have been reported in non-communicable diseases. While diet has a profound impact on gut microbiota composition and function, the role of food additives such as titanium dioxide (TiO2), prevalent in processed food, is less established. In this project, we investigated the impact of food grade TiO2on gut microbiota of mice when orally administered via drinking water. While TiO2had minimal impact on the composition of the microbiota in the small intestine and colon, we found that TiO2treatment could alter the release of bacterial metabolitesin vivoand affect the spatial distribution of commensal bacteriain vitroby promoting biofilm formation. We also found reduced expression of the colonic mucin 2 gene, a key component of the intestinal mucus layer, and increased expression of the beta defensin gene, indicating that TiO2significantly impacts gut homeostasis. These changes were associated with colonic inflammation, as shown by decreased crypt length, infiltration of CD8T cells, increased macrophages as well as increased expression of inflammatory cytokines. These findings collectively show that TiO2is not inert, but rather impairs gut homeostasis which may in turn prime the host for disease development.

Bacterial species that inhabit the colon interact with the host, promoting the development and function of immune cells locally and systemically. These interactions are mediated by bacterially derived metabolites such as short-chain fatty acids (SCFAs), which have been identified as critical inducers of immune subsets (13) key for protecting mice from disease development (25), emphasizing the role of the microbiota in gut homeostasis and host health.

The colonic epithelium acts as a physical barrier between the host and the gut microbiota. The secretion of mucus by goblet cells provides a barrier to microbial infiltration. Further, Paneth cells release antimicrobial peptides that protect against pathogen invasion as well as regulate gut microbiota composition (6). Expression of tight junction proteins by enterocytes also limits bacterial penetration. Epithelial function can be regulated by the gut microbiota via SCFAs, by stimulating mucus production (7) and tight junction assembly (8). In contrast, dysbiosis, marked by detrimental changes in gut microbiota composition, triggers increased gut permeability and gut inflammation (9). Alterations in antimicrobial peptide production, mucus layer thickness and/or epithelial permeability have been implicated in the development of a broad range of diseases such as colitis and colorectal cancer (10). These diseases have also been linked to abnormal interactions between the host epithelium and the gut microbiota through the formation of biofilm. Biofilms consist of aggregates of adherent and planktonic bacteria protected by an extracellular matrix and have been observed in the proximal colon of patients diagnosed with such diseases (11). The mechanisms behind the formation and the role of biofilm in the gut are not fully understood, but biofilm formation has been shown to impact both disease development and resolution. Both in a colitis rat model and in humans, biofilm in the colon has been shown to facilitate pathobiont adherence to the epithelium and translocation to the host (1213). In human inflammatory bowel disease, biofilm formation at the site of epithelial wound healing has been shown to negatively affect healing by impairing epithelialization and tissue repair (14). Finally, a recent study has shown that inoculation of germ-free mice with biofilm positive human colon inocula was carcinogenic (15).

The identification of environmental factors that can affect gut homeostasis is thus a critical first step in preventing the development of so-called western lifestyle diseases, encompassing autoimmune, allergic and metabolic diseases. A broad range of environmental factors can affect gut homeostasis, with diet composition being the major driver (16). Western-like diets enriched in fat and simple carbohydrates and deficient in dietary fiber have been shown to trigger dysbiosis, increases gut permeability and inflammation (16). While the impact of these macronutrients on gut homeostasis has been extensively studied (17), the role of food additives prevalent in processed food remains poorly defined. Food additives are used to improve the texture, preservation and aesthetics of food. Food grade titanium dioxide (TiO2) or E171, is a whitening agent present in over 900 commonly consumed food products. The average adult consumes between 0.7 and 5.9 mg of TiO2per kg of body weight (BW) per day throughout their life and children are the most exposed, consuming up to 32.4 mg TiO2/kg BW/day in maximally exposed individuals (18). Despite the fact that regulatory bodies do not define strict guidelines around its consumption, new evidence from animal studies has emerged, highlighting that TiO2may potentiate cancer development (19) and exacerbate inflammatory bowel disease (20).

The effect of TiO2on gut homeostasis is poorly understood yet evidence suggests that TiO2interacts with gut epithelial cells.In vivoandin vitrostudies have demonstrated the accumulation of TiO2in the mucus layer (21) and its uptake by colonic epithelial cells (2223). A study in rats has shown that TiO2affects immune cells in the Peyers patches associated with a decreased regulatory T cell proportion (19). However, the impact of TiO2on colonic immune cells, the site where microbiota is the densest, has never been investigated. While the impact of TiO2on the colonic microbiota has been previously investigated in a short term study (2.5 mg TiO2/kg BW/day for 1 week) (24) and using a high dose (100 mg TiO2/kg BW/day) for up to 4 weeks (25), the impact of TiO2on the small intestine microbiota is unknown.

The aim of the present study is to establish the effects of food grade TiO2on gut homeostasisin vivo. We investigated the impact of physiological doses (2 and 10 mg TiO2/kg BW/day) and a high dose of TiO2(50 mg TiO2/kg BW/day) on mouse colonic and small intestine microbiota composition and function, epithelial function and mucosal inflammation after 34 weeks of treatment via drinking water.

Food grade TiO2was purchased from All Color Supplies PTY. Average hydrodynamic diameter, polydispersity index and zeta potential of the TiO2nanoparticles dispersed in drinking water were determined with a Malvern Zetasizer Nano ZS at 25C. The dispersion was measured 3 times for both size and zeta potential. The size distribution and shape of the TiO2nanoparticles dispersed in mice drinking water were determined using a NanoSight NS300 (equipped with a sCMOS camera) at 25C. The dispersion was measured 5 times (1 min per measurement). The size distribution and shape of the TiO2nanoparticles dispersed in drinking water were further investigated using a Zeiss Ultra Plus scanning electron microscope operated at an accelerating voltage of 10 kV. A drop of the nanoparticle dispersion was allowed to dry on a stub, after which 20 of platinum metal was sputter coated onto the stub under vacuum to prevent charging.

A D8 Advance Bruker diffractometer was used to conduct the X-ray powder diffraction (XRD) analysis in a flat plate geometry using Ni-filtered Cu K radiation and a Bruker Lynx eye detector. The XRD patterns were acquired from 10 to 100 2 with a step size of 0.02 and a count time of 0.1 s. Elemental composition was determined using X-ray photoelectron spectroscopy (XPS) with an Al K monochromator X-ray source. A survey scan was acquired at 100 eV pass energy between 0 and 1,400 eV. High resolution spectra for individual elements were collected at 100 Ca 0.05 Ga. Elemental composition was calculated from the high-resolution spectra using CasaXPS with measurements done in triplicate.

Five to six week-old male C57BL/6JAusb mice from Australian Bio Resources were maintained under specific-pathogen-free conditions. All experimental procedures involving animals were approved by the University of Sydney Animal Ethics Committee under protocol number 2014/696. Mice were cohoused with water and food (AIN93G; Specialty Feeds) accessad libitum. Titanium dioxide (E171) was added to water and sonicated daily. TiO2was administered in drinking water at doses of 0, 2, 10, and 50 mg TiO2/kg BW/day, which was calculated based on the water intake measured per cage. At week 4, mice were euthanized using CO2asphyxiation.

Pieces of colon were incubated at 37C for 40 min in Hanks Balanced Salt Solution (HBSS; Gibco) with 5 mM EDTA, 5% FBS (Gibco) and 15 mM HEPES (Gibco). Intraepithelial lymphocytes were discarded and the remaining tissue was incubated at 37C for 1 h in HBSS (Gibco) with 6.7 mg/ml collagenase type IV (Gibco), 10% FBS (Gibco) and 15 mM HEPES (Gibco). Cells were passed through a 70 m mesh and lymphocytes enriched via percoll gradient of 80% and 40% (GE Life Sciences). The list of antibodies used for flow cytometry is in theSupplementary Methods. Viability was determined using the LIVE/DEAD Fixable Blue Dead Cell stain kit (Invitrogen). Flow cytometry was performed on a LSRII flow cytometer (BD Biosciences) and data analysis with FlowJo software (Treestar Inc., Ashland, OR, USA).

Total tissue RNA was extracted using TRI Reagent (Sigma) and converted into cDNA using iScript RT Supermix (BioRad) according to both manufacturers instructions. qPCR was performed on a LightCycler 480 (Roche) using SYBR Green (Biorad) with primer sequences listed inSupplementary Table 1.

Quantitative measurements of acetate and TMA in plasma were determined by nuclear magnetic resonance spectroscopy (NMR). Briefly, plasma was filtered through a 3 kDa membrane filter (Merck Millipore) and polar metabolites extracted from the aqueous phase of a water:chloroform:methanol mixture. Samples, containing 4,4-dimethyl-4-silapentane-1-sulfonic acid as an internal standard, were analyzed on a Bruker 600 MHz NMR.

A hydrophilic interaction chromatography LC-MS/MS method was used for choline detection in plasma as described previously (26). The LC was connected to an AB Sciex Triple Quad 5,500 mass spectrometer run in positive ion mode. Data analysis was done on software Multi-Quant 3.0 for MRM Q1/Q3 peak integration.

Escherichia coliK-12 MG1655 (E. coli) orEnterococcus faecalisNCTC 775 (E. faecalis) were incubated for 7 h at 37C, 5% CO2with Luria-Bertani (LB) broth containing E171 at indicated concentrations and then fixed in 3% formalin overnight. Cells were resuspended in PBS and visualized using a Nanolive 3D cell explorer. False colors were applied to images based on refractive index using STEVE software.

Thein vitrobiofilm formation assay was based on a previously published protocol (13). Overnight culture in quadruplicates ofE. coli(low salt LB broth; Beckton Dickinson),E. faecalis(tryptone soya broth supplemented with 0.25% glucose; Sigma Aldrich) orStaphylococcus epidermidisNCTC 6512 (LB broth) was adjusted to OD of 0.5 at 600 nm and 100 l of each bacterial culture was plated on separate round bottom 96-well tissue culture plates. A further 100 l of appropriate media supplemented with TiO2was added to achieve the indicated final concentrations. TiO2at the different final concentrations in media alone was used as background controls. Plates were incubated at 37C aerobically on a shaker (Ratek, 70 rpm) for either 24, 48, or 72 h.

Two hundred microliters of colon homogenates were cultured in quadruplicates in flat bottom 96-well-plates containing supplemented tryptic soy broth [sTSY: 30 g/L tryptic soy broth (Oxoid) with 5 g/L yeast extract, 5% L-cysteine, 50 mg/L hemin and 1 mg/L medanione (all from Sigma-Aldrich) to yield 0.05 mg/l (w/v)] for 24 h, aerobically at 37C at 70 rpm. Samples were diluted 1:100 in fresh sTSY containing TiO2at indicated doses and incubated for 5 days. After planktonic cell removal, biofilm was stained with crystal violet (CV). Briefly, plates were washed 3 times with water, air dried and stained with 1% CV (Sigma-Aldrich) for 30 min. After 4 washes in water and air drying, 95% ethanol was added for 15 min. Absorbance was recorded at 595 nm on a microplate reader (Tecan Infinite M1000).

Biofilm formation was also quantified based on Resazurin viability assay as previously described (27). Briefly, culture media was removed and wells washed once with phosphate-buffered saline (PBS). Then, media with 10% Resazurin (Sigma-Aldrich) was added to each well. The plates were incubated in the dark at 37C and fluorescence intensity measured every 15 min (excitation 570 nm, emission 585 nm). TiO2only controls were used to subtract background.

DNA from fecal samples or entire contents of small intestine lumen were extracted by mechanical disruption using a Fastprep (MP Biomedicals) using autoclaved glass beads (G8772 and G1145; Sigma-Aldrich) in lysis buffer [500 mM NaCl, 50 mM Tris-HCl (pH 8), 50 mM EDTA, 4% SDS] followed by 15 min incubation at 95C. DNA was precipitated in 10M ammonium acetate and isopropanol and washed with 80% ethanol. Protein and RNA were removed using the QIAamp DNA stool Minikit (Qiagen) following the manufacturers instructions. DNA samples were amplified across the V3-V4 region (Q5 polymerase; New England Biolabs) with these primers F: 5-ACTCCTACGGGAGGCAGCAG-3; R: 5-GGACTACHVGGGTWTCTAAT-3 and sequenced on an Illumina Miseq (2 300 bp). Data analysis was performed using QIIME 1.9.1 (28) using default parameters as described previously (29). Briefly, demultiplexed paired end data were quality filtered and paired using the Fastq-join algorithm with no errors allowed. Operational taxonomic units (OTUs) were picked using 97% similarity with UCLUST, and taxonomy was assigned with Greengenes database. The resulting OTU table was filtered by removing OTUs with 0.01% sequences and those relating to Cyanobacteria or Chloroplast. Further analysis was performed with R software (3.4.2). For statistical analysis, abundance data was transformed using the Hellinger method. Differences between treatment groups were determined by adonis (vegan 2.5-2) with 9999 permutations, alpha 0.05 and with the phyloseq package 1.25.2 (30) and Calypso 8.78 (31).

MannWhitneyU-test was used for analysis of the differences between the mean of groups and Wilcoxon paired test for paired samples. For microbiota data, significant differences in the relative abundance of genus between treatment groups were determined by one-way ANOVA withpost-hocTukeys test. Differences in overall microbial community between treatment groups were determined by adonis.p 0.05 were considered statistically significant.

We employed dynamic light scattering (DLS) to determine the hydrodynamic size of the E171 product used in this study. DLS revealed that the TiO2nanoparticles dispersed in drinking water (5 mg/ml, pH 7.8) have an average hydrodynamic diameter of 367 nm, a polydispersity index of 0.258 and a zeta potential of 23.0 mV (4.5 mV). We also employed nanoparticle tracking analysis (NTA) and scanning electron microscopy (SEM) to further investigate the size and shape of the TiO2nanoparticles dispersed in drinking water. NTA (Supplementary Figure 1A) showed that the TiO2nanoparticles are roughly spherical in shape and range in diameter from 28 to 1,158 nm. On a number basis, the particle size distribution has a mean of 202 nm and a mode of 138 nm and, on a weight basis, the particle size distribution has a mean of 363 nm and a mode of 428 nm. The average particle diameter determined by NTA on a weight basis (363 nm) is in good agreement with that determined by DLS (367 nm). SEM (Supplementary Figure 1B) confirmed that the TiO2 nanoparticles are roughly spherical in shape and revealed that they can be classified into essentially four groups (based on diameter)300, 150200, 100, and 3050 nmwhich is consistent with the particle size distribution (on a number basis) obtained by NTA. TiO2 was predominantly in anatase form as per manufacturers description. This was verified using X-ray powder diffraction (data not shown).

We first determined whether exposure to TiO2over a range of physiologically relevant doses impacted gut bacterial communitiesin vivo. To achieve this, mice were administered TiO2via drinking water at doses of either 0, 2, 10, or 50 mg TiO2/kg BW/day for 3 weeks. Sequencing of the 16S rRNA gene from fecal samples revealed that TiO2had limited effects on bacterial diversity as determined by Inverse Simpson and Shannon analyses (Figures 1A,B) nor bacterial richness (Figure 1C), evenness (Figure 1D) or Faiths diversity (Supplementary Figure 2A) at these doses. However, there was still a trend toward decrease in mice treated with physiological doses of TiO2(2 and 10 mg TiO2/kg BW/day). On the other hand, both weighted (Supplementary Figure 2B) and unweighted UniFrac (Supplementary Figure 2C) principal coordinate analysis (PCoA) showed some clustering of bacterial composition in control vs. TiO2treated mice. To test this further, we performed canonical correspondence analysis (CCA) constrained to the 4 distinct TiO2concentrations used, which revealed significant clustering in bacterial composition driven by 2 mg TiO2/kg BW/day (p= 0.0011) and 50 mg TiO2/kg BW/day (p= 0.0123) TiO2treatment (Figure 1E). We also performed CCA with TiO2as a continuous variable, which reveals a dose dependent effect of TiO2on microbiota composition (Supplementary Figure 2D). Treatment with TiO2significantly affected gut microbiota composition independently of the cage effect (with overall treatment effect:F-value = 8.2407,R2= 0.31644, Df = 3,p 0.001 and impact of treatment corrected for the cage effect:F-value = 5.8511,R2= 0.2996, Df = 3,p 0.001 both by adonis). We then determined the impact of TiO2at deeper levels and found significant changes at the genus level.Parabacteroideswere significantly elevated in TiO2treated mice, at a dose of 50 mg TiO2/kg BW/day (Figure 1F) whileLactobacillusandAllobaculumwere significantly elevated at all doses tested (Figures 1G,H). On the other hand,Adlercreutzia(Figure 1I) and UnclassifiedClostridiaceae(Figure 1J) were significantly decreased in the groups treated with TiO2at the doses of 10 and 50 mg TiO2/kg BW/dayrelative to the untreated group. These results suggest that TiO2had a minor impact on microbiota compositionin vivo, while affecting few taxa at the genus level. The gut microbiota composition in the small intestine was also analyzed to determine whether TiO2might have a greater effect here than in the colon. Bacterial diversity indices (Richness, evenness, Shannon, Inverse Simpson and Faiths diversity) were not significantly affected at doses of 10 and 50 mg TiO2/kg BW/day (Supplementary Figure 2E), although these trended toward decrease with increasing dose of TiO2. Unlike in the colon, TiO2did not significantly alter the small intestine bacterial composition (p 0.05 by adonis) and weighted and unweighted UniFrac PCoA analysis revealed no obvious clustering (Supplementary Figures 2F,G). Overall, TiO2did not appear to dramatically impact on small intestinal microbiota composition. We also performed co-occurrence analysis by examining microbial interactions from mice treated with either 0, 2, 10, or 50 mg TiO2/kg BW/day. We found that certain genera are consistently associated with each other regardless of TiO2treatment (Ruminococcus, Desulfovibrio, andOscillospiraare positively connected). Increasing TiO2intake, especially at the dose of 10 and 50 mg/kg BW/day resulted in more significant connections within the network, as well as increased number of genera with significant contributions. For example, while Akkermansia was not significantly involved in the microbial network of mice administered 0, 2, or 10 mg TiO2/kg BW/day, it is involved at a dose of 50 mg/kg involving numerous co-exclusion relationships. These co-occurrence graphs are presented inSupplementary Figures 2HK. These results were verified using the deblur pipeline (32) which resolves amplicon sequences much more accurately (Supplementary Figures 3AF).

Figure 1. Impact of TiO2on colonic microbiota composition.(AD)Diversity of colonic microbiota composition of mice administered 0, 2, 10, or 50 mg TiO2/kg BW/day in drinking water was determined by(A)Shannon index,(B)Inverse Simpsons index,(C)richness, and(D)evenness (n= 10 mice per group from 2 cages of 5 mice).(E)Canonical correspondence analysis ordination of Bray-Curtis dissimilarity of colonic microbiota compositions of mice administered 0, 2, 10, or 50 mg TiO2/kg BW/day in drinking water. Ordination was constrained by dose of TiO2and the arrows represent the doses of TiO2driving the differences in microbiota composition observed. Composition differences between groups were significant as determined by adonis (p= 0.0012 for 0 vs. 2 mg TiO2/kg BW/day,p= 0.0006 for 0 vs. 10 mg TiO2/kg BW/day andp= 0.0105 for 0 vs. 50 mg TiO2/kg BW/day) (n= 10 mice per group from 2 cages of 5 mice).(FJ)Relative abundance of(F)Parabacteroides(G)Lactobacillus,(H)Allobaculum,(I)Adlercreutzia, and(J)UnclassifiedClostridiaceaeobserved in colonic microbiota of mice administered 0, 2, 10, or 50 mg TiO2/kg BW/day in drinking water. *p 0.05, **p 0.01, ***p 0.005 as determined by one-way ANOVA withpost-hocTukeys test on Hellinger-transformed data (n= 10 mice per group from 2 cages of 5 mice).

We and others have shown that gut bacterial metabolites such as SCFAs can have a dramatic impact on host immune function and disease development (153334). Mice treated with 50 mg TiO2/kg BW/day had a significant decrease in the SCFA, acetate, in the plasma, suggesting a possible impact of TiO2on host-bacterial interaction (Figure 2A). Such effects on bacterial metabolites were not limited to SCFAs as TMA, a bacterial product associated with development of atherosclerosis (35), was increased at doses of 10 and 50 mg TiO2/kg BW/day (Figure 2B). TMA is a product of conversion of choline, which was also found to be decreased at 50 mg TiO2/kg BW/day (Figure 2C), suggesting that increased TMA was not due to a change in the substrate availability but potentially changes in bacterial activity.

Figure 2. Impact of TiO2treatment on gut bacterial metabolites.(A,B)Concentrations of(A)the SCFA acetate and(B)TMA were determined NMR on the serum of mice administered 0, 2, 10, or 50 mg TiO2/kg BW/day in drinking water. Data is represented as median interquartile range (IQR). *p 0.05 as determined by MannWhitneyU-test (n= 10 mice per group).(C)Concentration of choline was determined by liquid-chromatography mass spectrometry in plasma of fasted mice treated with 0, 2, 10, or 50 mg TiO2/kg BW/day in drinking water. Concentration is represented as area under curve (AUC). Data are represented as median IQR. *p 0.05 as determined by MannWhitneyU-test (n= 10 mice per group). **p 0.01.

Bacteria also communicate with the host via direct interactions. Studies have shown that attachment of biofilm on the colonic epithelium was correlated with colorectal cancer, a disease in which TiO2has aggravating effects (36). To explore the possibility that TiO2might promote biofilm formation, we incubated two types of commensal bacteria,E. coliandE. faecalis, in the presence of TiO2. Nanolive imaging revealed the clustering effect of TiO2on bothE. coli(Figure 3A) andE. faecalis(Figure 3B)in vitroin a dose dependent manner. To determine whether the cluster of bacteria was due to biofilm formation, we performedin vitroculture of eitherE. faecalis or E. coliin the presence of 2, 10, or 50 g/ml of TiO2for 24 or 72 h, respectively. Using the resazurin viability assay (Figure 3C), we found that TiO2treatment significantly increased biofilm formation in both subsets of bacteria (Figures 3D,E) but not inStaphylococcus epidermidis, a strain known for its inability to form biofilm (Supplementary Figure 4). We confirmed by confocal microscopy that TiO2treatment increased biofilm formation in bothE. coliandE. faecalis(Supplementary Figure 5). To determine whether such effects were applicable to bacteria in the complex environment of the gut microbiota, we incubated commensal bacteria derived from mouse colons anaerobically for 5 days with doses of 2, 10, and 50 g/ml of TiO2. Both the doses of 10 and 50 g/ml TiO2significantly promoted biofilm formation by commensal bacteria (Figure 3F). These data highlight that TiO2can affect the spatial organization of the gut microbiota and thus its potential interaction with the host.

Figure 3. TiO2triggers biofilm formation by commensal bacteria.(A,B)The clustering effect of TiO2on(A)E. coliand(B)E. faecalis in vitrowas visualized by Nanolive imaging in the presence of 0, 0.5, 1, or 50 g/ml TiO2after 24 h incubation. False-coloring was applied to images based on refractive index, where black represents the refractive index of TiO2and green represents bacteria.(C)Schematic representation of biofilm formation assay and resazurin viability assay to assess biofilm formationin vitro.(D,E)The ability ofE. faecalisandE. colito form biofilmin vitroin the presence of 0, 2, 10, or 50 g/ml TiO2in culture (24 and 72 h, respectively) was assessed by the resazurin viability assay (n= 8 replicates).(F)Colonic bacteria were isolated and biofilm formation assessed in the presence of 0, 2, 10, or 50 g/ml TiO2in culture after 5 days (n= 6 mice per group). Data are represented as median IQR. *p 0.05, **p 0.01; ***p 0.001 as determined by Wilcoxon paired test compared to non-treated group.

While the impact of biofilm formation on the host is unclear, impaired mucus production has been correlated with the presence of bacterial biofilms (11). To determine whether TiO2might impact the mucus layer, we examined colonicMuc2gene expression in the colon. We found that both 10 and 50 mg TiO2/kg BW/day decreasedMuc2expression, suggesting a detrimental impact of TiO2on the mucus layer (Figure 4A). While biofilm formation has been reported in colitis and colorectal cancer (36), these diseases have also been linked to increased gut permeability (37). To test whether TiO2affects gut permeability, we studied the expression ofTjp1(encoding for zonula occludens 1), which was unchanged by TiO2treatment (Figure 4B), suggesting no impact of TiO2on gut permeability. The other major mechanism of bacterial exclusion is through the release of antimicrobial peptides. Beta defensin is expressed predominantly in the colon and we found thatDefb3(encoding for beta-defensin-3) was elevated by treatment at doses of both 10 and 50 mg TiO2/kg BW/day (Figure 4C). Expressions of other antimicrobial peptides such as granzyme B (Figure 4D), cathelin-related antimicrobial peptide (CRAMP), regenerating islet-derived protein 3 gamma (REG3 gamma) and p-lysozyme (PLYz) (Supplementary Figure 6) were unchanged. Therefore, TiO2treatment impairs the expression of key colonic epithelial factors involved in gut homeostasis.

Figure 4. TiO2impairs colonic epithelial function.(AD)The impact of TiO2on colonic epithelial function was determined by comparison of gene expression of key markers(A)Muc2,(B)Tjp1,(C)Defb3, and(D)Gzmb incolonic tissue of mice administered 0, 2, 10, or 50 mg TiO2/kg BW/day in drinking water (n= 58 mice per group). Data are represented as median IQR. *p 0.05, **p 0.01 as determined by MannWhitneyU-test.

DecreasedMuc2has been correlated with inflammation and MUC2 deficiency leads to spontaneous colitis (38). To test whether TiO2might affect innate immune cells in the colon, we studied myeloid immune cell populations by flow cytometry. While neutrophils (CD45Ly6gCD11b) (Figure 5A) and dendritic cells (CD45I-abLy6gF4/80CD11c) were unchanged (Figure 5B), macrophages (CD45F4/80CD8Ly6gI-abCD11bCD103) were significantly increased by TiO2at 10 and 50 mg TiO2/kg BW/day (Figure 5C). This change was not due to an increased recruitment of total monocytes (CD45CD8Ly6GLy6CCD11bI-ab) (Figure 5D), suggesting a potentialin situproliferation of macrophages (gating strategies shown inSupplementary Figure 7). Colonic macrophages are a major source of IL-6, TNF-alpha and IL-10, cytokines, which were also upregulated in the colon of TiO2treated mice (Figures 5EG). We also observed a significant reduction in coloni