The TiO2 NPs were purchased from Shanghai Macklin Reagent Co. Ltd., China. The size and shape of the particles were characterized by scanning electron microscopy (SEM, Nova, Tecnai F30, FEI Company, Oregon, USA). Energy dispersive X-ray spectroscopy (EDS, Nova_NanoSEM430, FEI Company, Oregon, USA) was used to measure the ratio of Ti to O atoms. The purity of the particles was analyzed by detecting the content of Ti element using inductively coupled plasma mass spectrometry (ICP-MS, IRIS Advantag, TJA, Franklin, MA, USA). The crystal structure of the particles was identified by X-ray powder diffractometry (XRD, PANalytical’s X’Pert PRO, X’Celerator, EA Almelo, Netherlands). The specific surface area (SSA) of the particles was measured according to the Brunauer–Emmett–Teller (BET) method (Quantachrome, Autosorb 1, Boynton, FL, USA).
The artificial gastric juice (AGJ, pH = 1.2) was prepared using 10 g/L pepsin (3800 units/mg) and 45 mmol/L HCl. The artificial intestinal juice (AIJ, pH = 6.8) was made with 10 g/L trypsin (2500 units/mg) and 6.8 g/L KH2PO4. The pH was adjusted to 6.8 using 0.1 mol/L NaOH. After TiO2 NPs were dispersed in ultrapure water (H2O), AGJ or AIJ to obtain a final concentration of 1 mg/mL TiO2 NPs, the suspensions were supersonicated for 15 min to break up aggregates. The particle hydrodynamic diameters and Zeta potentials were tested using the ZetaSizer Nano ZS90 (Malvern Instruments Ltd., Malvern, UK).
Animals and experimental design
Three-week-old healthy Sprague-Dawley (SD) rats were bred and supplied by the Department of Laboratory Animal Science, Peking University Health Science Center. The rats were fed a commercial pellet diet and deionized water ad libitum and were kept in plastic cages at 20 °C ± 2 °C and 50 to 70% relative humidity with a 12:12-h light-dark cycle. After 1 week of acclimation, the rats were weighed and randomized into experimental and control groups, with six male rats in each treatment group.
All experimental rats were provided humane care. The study was conducted in accordance with the guidelines of European Union Directive 2010/63/EU for animal experiments, and received approval from the Peking University Institutional Review Board (Approval number: LA2017073). The design of animal experiment refers to OECD Guidelines for the Testing of Chemicals No. 408 Repeated Dose 90-Day Oral Toxicity Study in Rodents.
The TiO2 NPs were dispersed in ultrapure water and sonicated for 15 min. To obtain homogenized suspension, the particle suspension was vortexed before every use. Suspensions of TiO2 NPs (0, 2, 10, 50 mg/kg BW) were administered to rats via oral gavage in a volume of 1 mL daily for 90 consecutive days. The intragastric doses of TiO2 NPs for rats were selected based on the oral intake of TiO2 NPs for children [4, 10], using 100 as safety factor.
The symptoms and mortality were observed and recorded daily throughout the entire duration of exposure up to 90 days. The body weight of rats was assessed every 7 days, and the food intake of rats was recorded every 3 to 4 days. During the experiments, no significant changes in the body weight and food intake of the exposed rats were found (Additional file 1: Figure S1, S2), and no mortality was observed. After 90 days, rat feces were collected and quickly transferred and stored in an − 80 °C refrigerator. Then, animals were weighed and euthanized. The blood samples were collected from the abdominal aortic vasculature. Serum was obtained by centrifuging blood at 3000 rpm (1500 g) for 10 min. The liver tissues were harvested and weighed.
Measurement of blood biochemical parameters and histopathological analysis related to hepatic damage
The serum levels of alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL), total protein (TP), albumin (ALB), and globulin (GLB) were assayed to evaluate hepatic damage. All biochemical assays were performed using a clinical automatic chemistry analyzer (Type AU400, Olympus, Japan).
For pathological studies, all histopathological examinations were performed using standard laboratory procedures. The liver tissues were embedded in paraffin blocks, then sliced into 5-μm slices and placed onto glass slides. After hematoxylin–eosin (HE) staining, the slides were observed and the photos were taken using an optical microscope (OlympusBX50, Moticam 2306, Japan). The pathologist who performed the observation and analysis was blinded of the treatment groups and dosing regimens. For TEM observation, the liver tissues were cut up into small pieces (1 mm3) and immediately fixed in 2.5% glutaraldehyde (pH 7.4) overnight. Then, the samples were treated according to the general protocols for TEM study. The ultra-thin sections (70–100 nm) were stained with lead citrate and uranyl acetate. The specimens were examined using JEOL JEM-1400 electron microscopy.
Liver metabonomic analysis
The method for liver metabonomic analysis refers to the protocol of Want et al. .
Homogenate of liver tissue and sample preparation.
30.0 mg of liver tissue was added to 900 μl pre-cooled methanol/water (1:1) solution. Then it was homogenized at 30000 rpm on ice for 30 s (Homogenized for 10 s, cooled for 30 s, repeated three times) and blended by a vortex for 20 s and stored at − 20 °C overnight. And then, the homogenate was centrifuged (16,000 g, 4 C) for 10 min and the supernatant was taken. The supernatant was dried and concentrated in a low temperature vacuum concentrator and for 4 h.
Re-suspension was carried out by 200 μl methanol/water (1:1) solvent before analysis. Meanwhile, 20 μl of each sample was taken and then divided into three parts after gentle mixing as quality control (Qc) samples, which were prepared for technical repetition to evaluate the stability and repeatability of the experimental instruments and methods.
Non-targeted metabonomics analysis using HPLC-MS
Ultra High Performance Liquid Chromatography-Q-Exactive Orbitrap-High-resolution Mass Spectrometry System (UPLC-QEMS, U3000, Thermo, USA) was used for non-targeted metabolomics analysis. The samples were randomly injected after disruption of the order to control the possible impact of instrumental stability fluctuation. Three Qc samples were analyzed before experimental samples, after half of all samples and after all samples, respectively. The QEMS was equipped with an electrospray ionization source (ESI). Fragmentation was achieved by high-energy collision dissociation (HCD). The normalized collision energies were 15, 30 and 45 eV, respectively. The results were measured by positive ion mode and negative ion mode. The mass scanning range was 50–1100 m/z, and the total scanning resolution of parent ions (MS) was 60 K.
Analysis and annotation of mass spectrometry data
The original result file obtained by instrument analysis (.raw format file, positive and negative ion mode data) is imported into Compound Discoverer 3.0 software (Thermo Fisher Scientific, USA) for peak alignment, deconvolution, noise filtering, mass-charge correction and baseline correction. The parameters are set as follows: retention time (RT) < 0.2 min; signal-to-noise ratio (SNR) > 3; DDA mode was used to analyze secondary ion mass spectrometry (MS2); signal intensity > 500,000 included in the analysis; the filling gap algorithm was used to extract and fill the peaks (More parameters for metabolite identification by Compound Discoverer software were shown in Additional file 1: Table S1).
The annotation and identification of metabolites were carried out through software-related mzCloud database and mzVault database. The peak area was used as the relative concentration for subsequent analysis. The data containing metabolite identification results and peak area were pretreated. The relative concentration of metabolites was completely clustered using Euclidean distance, and the Heatmap was drawn to show the difference of the concentrations of metabolites in each sample. Principal Component Analysis (PCA) was used to reduce the dimension of the original data and observe the difference trend and potential outlier value of samples. Qc samples were also included in PCA analysis and scoring map drawing. The stability of the instrument in the process of analysis was investigated by the aggregation of Qc samples in PCA scoring map. The closer the aggregation of Qc samples, the better stability of the instrument. Using the orthogonal projection to latent structure discriminant analysis (OPLS-DA) model to screen biomarkers that change after TiO2 NPs exposure, the OPLS-DA model score maps were drawn by the first predictive component (T score ) and the first orthogonal component (Orthogonal T score ). Through Simica-P. software (V14.1, Umetrics, Sweden), the permutation test of OPLS-DA model is performed to verify the stability of OPLS-DA model. Then V-Plot was draw according to the covariance (p1) and reliability (p(corr)1) of the first principal component of each variable in OPLS-DA model, and the metabolite with absolute value of p(corr)1 greater than 0.5 in V-Plot is selected as differential metabolite. Z Score-plot is drawn to visualize the distribution of different metabolites among groups through ggplot2 packages in R. Z-score = ((Metabolite concentration - mean of metabolite concentrations in control group) / standard deviation of metabolite concentrations in control group).
After obtaining the differential metabolites, KEGG metabolic pathway was analyzed by Pathway Analysis function module in Metaboanalyst 4.0 website. Pathway topology analysis was conducted by using the pathway data in Rattus norvegicus pathway libraries (including 81 metabolic pathways), and hypergeometric test in over representation analysis was used to test the significance of metabolic pathway enrichment, and the out-degree centrality is used as a criterion to measure the impact of different metabolites on pathway. The significant changed pathway was determined by adjusted Holm P < 0.05 through Holm method (also known as step-down Bonferroni method), false discovery rate (FDR) < 0.05, and pathway impact greater than 0.10.
Detecting oxidative stress biomarkers and inflammatory cytokines
Oxidative damage to the liver following repeated TiO2 NPs exposure was evaluated by the levels of reduced (GSH) and oxidized (GSSG) glutathione, glutathione peroxidase (GSH-Px), lipid peroxidation products (malondialdehyde, MDA), and superoxide dismutase (SOD) in tissue homogenates, which were tested using commercial kits (Nanjing Jiancheng Bioengineering Institute, Jiangsu, China).
The inflammatory cytokines in serum from rats with repeated TiO2 NPs exposure were analyzed by the Cytometric Bead Array (CBA) Rat inflammatory cytokines Flex Set (BD Biosciences, San Jose, CA). Briefly, interleukin 1α (IL-1α), interleukin 4 (IL-4), and tumor necrosis factor (TNF) concentrations were determined using BD FACSCalibur flow cytometers according to the manufacturer’s protocol (BD Biosciences, San Jose, CA). The data were analyzed by FCAP Array Software (Soft Flow Inc., Pecs, Hungary).
16S rDNA sequencing and gut microbiota analysis
Genomic DNA of fecal samples was extracted by Cetyltrimethylammonium Ammonium Bromide (CTAB) method after water was removed by freeze-drying apparatus. After the purity and concentration of DNA were detected by agarose gel electrophoresis, DNA samples were diluted with sterile water to 1 ng/μL. PCR amplification was conducted by using the diluted genomic DNA as template, and the specific primers with Barcode, Phusion® High-Fidelity PCR Master Mix with GC Buffer and high-efficiency fidelity enzyme. The PCR amplification system was as follows: 2 × taq PCR mix: 25 μl; Primer F (10 μM): 1 μl; Primer FR (10 μM): 1 μl; gDNA: 2.5 μl; H2O: 8.0 μl. The procedure of PCR amplification was as follows: 1) 95 °C for 5 min; 2) Step a-c cycle 34 times, a) 94 °C for 1 min, b) 57 °C for 45 s, c) 72 °C for 1 min; 3) 72 °C for 10 min; 4) 16 °C for 5 min. The primer sequence was as follows (5′-3′): V4-515F, GTGCCAGCMGCCGCGGTAA; V4-806R, GGACTACHVGGGTWTCTAAT; V3 + V4-341F, CCTAYGGGRBGCASCAG; V3 + V4-806R, GGACTACNNGGGTATCTAAT; V4 + V5-515F, GTGCCAGCMGCCGCGGTAA; V4 + V5-907R, CCGTCAATTCCTTTGAGTTT. The V3-V5 region of 16S rRNA gene extracted from fecal specimens was amplified by universal primers. The PCR product was detected by electrophoresis with 2% agarose gel. According to the concentration of PCR product, the sample was mixed equally. After mixing fully, the PCR product was purified by 2% agarose gel electrophoresis with 1 × TAE, and the target band was cut and recycled. GeneJET gel recovery kit was used to recover the purified product. Sequencing libraries were generated using Ion Plus Fragment Library Kit 48 rxns library kit. The libraries were quantified by Qubit fluorescence and then single-End sequencing was performed by Ion S5™XL sequencer. Small fragment libraries were constructed for sequencing. The operation steps in the experiment were strictly in accordance with the instructions.
QIIME (Version 1.9.1) software was used to filter the mosaic data. Subsequently, the sequence data obtained are compared with the sequence in 16S: Gold database, and the chimera sequence is detected and removed to obtain the effective sequence (Clean reads) for subsequent analysis. Uparse (v7.0.1001) software was used to cluster all clean reads and sequences with similarity greater than 97% were clustered into an Operational Taxonomic Unit (OTU). The pseudo-OTUs caused by chimeras were discriminated and filtered. The OTUs of each sample were obtained, and the sequence with the highest frequency of each OTU was selected as the representative sequence. Mothur algorithm and SILVA SSU r132 database were used to annotate the representative sequences of OTUs and obtain taxonomic information. The community compositions of each sample were count at level of kingdom, luphym, class. Order, family, genus and species. Using MUSCLE (Version 3.8.31) software for fast multi-sequence alignment, the phylogenetic relationships of all OTUs representative sequences were obtained. Finally, according to the sequence with the least amount of data in the sample, the data of each sample were normalized. The normalized data was used in the subsequent Alpha and Beta diversity analysis to compare the different bacteria community structure among different experiment groups. KRONA was used to visualize the results of species annotation. The first 10 species with the highest abundance in each taxonomic hierarchy (phylum, class, order, family, genus, species) were selected to draw a cylindrical accumulative map of relative abundance of species generated by taxonomic tree. LEfSe (Linear Discriminant Analysis Effect Size) software was used to compare the species differences among groups. Linear Discriminant Analysis (LDA) was used to find the different intestinal bacteria among groups (LDA Score > 4). In addition, we used Tax4Fun (taxon for function) package in R software to predict the metabolic capacity based on 16S results. The principle is to align the OTUs sequence with the OTUs sequence in SILVA database, and map the OTUs sequence data to the OTUs with macrogenomic information in the corresponding SILVA database by nearest neighbors identification, so as to obtain the predicted macrogenomic data of bacteria. The data were then converted to the abundances of the coding genes of the corresponding enzymes by the prokaryotic KEGG organisms database and NCBI genome annotations. Then, according to the abundances of related enzymes, we can predict the metabolic ability of the gut microbiota to a certain metabolic pathway substance, so as to predict the metabolic pathway changes of the gut microbiota with 16S sequencing data.
Measurement of lipopolysaccharides (LPS) and short-chain fatty acids (SCFAs)
LPS content in the feces was measured using enzyme-linked immunosorbent assay (ELISA) kits (Wuhan Abebio Science Co., Ltd., China) according to the manufacturer’s instructions. The assay employed a two-site sandwich ELISA to quantitate LPS in samples. SCFAs in the feces, including acetic acid (AA), propionic acid (PA), isobutyric acid (IBA), butyric acid (BA), isovaleric acid (IVA), and hexanoic acid (HA), were assayed by targeted metabonomics using GC-MS/MS (Thermo, USA). The key parameters for GC-MS/MS analysis are shown in Additional file 1: Table S5. The software Agilent Mass Hunter was used for data processing. Quality control (QC) samples were prepared by mixing sample extracts to monitor the repeatability of the analysis process.
The methods of statistical analysis for the metabonomics and gut microbiota data were previously described. Other data were expressed as means ± standard deviations (SDs) and analyzed with SPSS 20.0. One-way analysis of variance (ANOVA) with least significant difference (LSD) or the Dunnet T3 test was applied to evaluate the statistical significance of differences between the experimental groups and the controls. A p value < 0.05 was considered to be statistically significant.