TEM of the particles in cell culture medium revealed single particle sizes of 10.7 nm (mean), 10.0 nm (median), and 9.0 nm (mode) for NP1, 12.1 nm (mean), 11.0 nm (median), and 10.0 (mode) for NP2, 23.1 nm (mean), 17.0 nm (median), 16.0 nm (mode) for NP3, and 138.9 nm (mean), 134.0 nm (median), and 78.0 nm (mode) for NP4 (Figure 2a–d). Agglomerates were observed in NP1, NP2, and NP4, while no agglomeration could be detected in NP3. By static light scattering SLS analysis (Figure 2e–h) we could demonstrate that the agglomerates of NP4 dispersed, while NP1 and NP2 showed a wide size distribution of the particles in a range from 40 nm to 57 µm with peaks at 300 nm and 4.5 µm for NP1 and peaks at 400 nm, 2.5 µm and 30 µM for NP2. NP3 and NP4 showed a unimodal size distribution with a mean agglomeration size of 96 nm for NP3 and 390 nm for NP4.

Based on our previous study [9], and others [12,13,14,15], we reasoned that TiO₂ particles are endocytosed by Caco-2 cells. To investigate this possibility, we used laser scanning confocal microscopy identification of 10 nm TiO₂ NPs within the Caco-2 cells (Figure 3a–c). We chose 10 nm particles as these caused an inflammatory response in Caco-2 cells, as shown in our previous study [9]. Three sections along the z-axis ((a) apical side; (b) within; (c) basal side) of the cells provide information about TiO₂ NPs localization. Confocal images show nuclei (blue) surrounded by actin filaments (red). Aggregates of 10 nm TiO₂ NPs (green) are located at the apical side of the cells (Figure 3a), in the cell membrane, and within the cell (Figure 3b,c).

To further prove that NPs are endocytosed by Caco-2 cells, we used inhibitors of molecules which are involved in endocytosis (see Figure 1) and measured the NF-κB response of Caco-2 cells after administration of TiO₂ NPs. NP1 and NP2 induced NF-κB activity in Caco-2^nfkb-RE cells by a factor of 7.0 ± 0.6 and 7.7 ± 0.7, respectively, NP3 and NP4 showed no effect on NF-κB activity (1.0 ± 0.0 and 1.4 ± 0.1, respectively) (Figure 4a). Nystatin (50 µg mL⁻¹), a caveolae-mediated endocytosis inhibitor, had no effect on NP-induced NF-κB activation. Treatment with 10 µM chlorpromazine (a clathrin-mediated endocytosis inhibitor) reduced NP-induced luciferase activity by 74% and 75% in NP1 and NP2, respectively. Dynasore (100 µM) (a dynamin inhibitor) reduced activation of NF-κB by 68% and 69% in NP1 and NP2, respectively. As actin filaments and microtubules play an essential role in all types of endocytosis, we investigated the effects of cytochalasin D (0.1 µg mL⁻¹) and nocodazole (2 µM) on NP-induced NF-κB response. Treatment with cytochalasin D reduced NF-κB activity by 64% in both, NP1 and NP2 treated cells. Nocodazole reduced NP-induced NF-κB activity by 45% (NP1) and 42% (NP2). Since NP2 showed pronounced effects, we chose this NP for all further experiments. Expression of IL8 mRNA, chemokine (C–C motif) ligand 20 (CCL20) mRNA and C–X–C motif chemokine 10 (CXCL10) mRNA as indicators for inflammatory response of the cell was increased by 6.7 ± 1.1-fold, 9.6 ± 1.4-fold, and 9.2 ± 2.9-fold 3 h after NP2 exposure compared to untreated control cells (Figure 4b). Co-treatment with dynasore reduced the response by 76.2%, 66.7%, and 84.3% (IL8, CCL20, CXCL10), cytochalasin D altered the expression by 77.6%, 76.3%, and 87.1% and nocodazol lowered the reduced the response cells by 59.2%, 77.3%, and 89.5%. Treatment with nystatin did not change the NP-induced increase in mRNA expression of IL8, CCL20, or CXCL20, while chlorpromazine only reduced CCL20 expression by 47.3%.

As receptor mediated endocytosis is often initiated by activating the EGFR receptor, we aimed to suppress the NP-induced inflammatory response by pretreatment of Caco-2 cells with EGFR inhibitors BIBX 1382 and CL-387785. This treatment resulted in a reduced NF-κB activity in NP2 treated cells by 51.3% and 44.6%, respectively (Figure 5a), 6h after NPs exposure. Further, 3 h after NPs exposure IL8 mRNA, CCL20 mRNA, and CXCL10 mRNA expression was reduced by 81.6%, 75.3%, and 73.7%, respectively, in BIBX 1382 treated cells and by 89.1%, 77.4%, and 73.2%, respectively, in CL-387785 treated cells (Figure 5b).

To gain evidence that TiO₂ NP2 after being taken up by endocytosis are transported though the cell within lysosomes, we used live cell imaging. Caco-2 cells treated with NP2 showed a fluorescence signal of the NPs (green, Figure 6a), whereas untreated control cells showed no signal (Figure 6d). Lysosomes labeled by LysoTracker^® Red DND-99 are present in NP-treated cells (Figure 6b) and in untreated control cells (Figure 6e). Overlay of both images indicate NPs localized to lysosomes 3 h after exposure (marked by the arrows, Figure 6c,f).

To show that after endocytosis and transport via lysosomes gene expression is mediated via the known EGFR pathway, we analyzed the downstream elements of the pathway using Western blot and EMSA analyses. NP2 treatment of Caco-2 cells induced a transient activation of ERK1/2 compared to untreated control cells since levels of p-ERK1/2 increased in a time dependent manner until 60 min after NPs exposure. Activation of ERK1/2 faded 90 min after NPs exposure (Figure 7a). To analyze the involvement of EGFR in the NP-induced ERK activation we used EGFR inhibitor BIBX 1382. As shown by Western blot analysis p-EKR1/2 was induced 30 min after NPs and EGF stimulation by a factor of 2.5 ± 0.4 and 4.2 ± 0.9 compared to untreated control cells, and was completely abolished using BIBX 1382 (Figure 7b).

The best-studied nuclear target of phosphorylated ERK1/2 is the transcription factor ELK1 [20]. Stimulation with NP2 increased the binding of ELK1 to its DNA recognition site (Figure 7c) and declined 60 min after NPs treatment. Densitometric analyses revealed a 1.8 ± 0.5-fold increase of the binding complex compared to untreated control. Expression analysis revealed an increase by factor 11.0 ± 2.6 and 12.6 ± 5.3 3 h after NPs exposure of CCL2 and CXCL3 mRNA, respectively (Figure 7d). 6 h after NPs treatment mRNA expression of CCL2 and CXCL3 was induced by 4.8 ± 0.4-fold and by 3.3 ± 0.5-fold, respectively. Pre-incubation with EGFR inhibitor BIBX 1382 decreased the NP-induced mRNA expression of CCL2 and CXCL3 after 3 h by 77% and by 70%, and after 6 h by 54% and 58%.

Titanium (IV) dioxide (anatase) particles were obtained from Alfa Aesar (Alfa Aesar GmbH and CoKG, Karlsruhe, Germany). Four differently-sized particles were used: 1. 5 nm (NP1; Stock Number 44689, Lot F11T023, specific surface area (SSA) 210 m² g⁻¹), 2. 10 nm (NP2; Stock Number 44690, Lot B19T020; SSA 120 m² g⁻¹), 3. 32 nm (NP3; Stock Number 39953, Lot F23T043, SSA 51 m² g⁻¹) and 4. 490 nm (NP4; Stock Number 36199, Lot G02S013; SSA not specified). Stock dispersions of all particles were prepared with deionized water to a final concentration of 2 mg mL⁻¹. The dispersion was sonicated at 23 KHZ and 150 W (MSE Ltd., London, UK) for 2 min and finally autoclaved. Immediately before treatment of the cells the dispersion was diluted in culture medium as indicated below.

Characterizations of the particles were performed by transmission-electron microscopy (TEM) using freeze-fracture preparation technique and by static light scattering (SLS). Sonicated and autoclaved stock solutions of NPs were diluted 1:10 in cell culture medium and cryoprotected by immersion in glycerol solution. Samples were cryofixed into melting Freon 22 and liquid nitrogen. Freeze fracturing took place at −120 °C with a BAF 400 (Bal-Tec, Balzers, Liechtenstein). Freeze-fractured specimens were replicated by application of Pt/C and C by electron-gun evaporation. The replicas were cleaned in concentrated sodium hypochlorite and in acetone and examined with a Tecnai 10 (FEI Company, Hillsboro, OR, USA) transmission-electron microscope operated at 80 kV. Size measurement of single particles and particle aggregations using the TEM images was conducted with the program “Bild-Vermessen 1.0” (CAD-KAS Kassler Computersoftware GbR, Markranstädt, Germany). At least 30 single particles and 20 aggregates of the different NPs were measured. Particle size distribution was measured by SLS using a LS 230 (Beckmann Coulter, Krefeld, Germany). The particle dispersion was dosed into the instrument without special sample preparation. The volume fraction-length mean diameter was measured.

Human colon adenocarcinoma cell line Caco-2 (Toni Lindl, Munich, Germany) between passage 24 and 50 were cultured in Caco-2 medium (45% Dulbecco’s Modified Eagle Medium (DMEM), low glucose, 45% Ham’s F12, 9% fetal calf serum (FCS), 0.9% non-essential amino acids) (all PAA Laboratories GmbH, Austria), and insulin (10 µg mL⁻¹) (Biochrome AG, Berlin, Germany) at 37 °C and 5% CO₂. For all experiments cells were seeded at a density of 1× 10⁵–2 × 10⁵ cells cm⁻². TiO₂ particles were added to the cells at a final concentration of 40 µg cm⁻² cell growth surface as this concentration is in accordance with the observed exposure in humans [2] and in the range of other comparable references [3,5,18,58]. We determined the time we exposed the cells to NPs prior to the described experiments by a number time course analyses.

For mRNA expression analysis cells were seeded in 24-well or six-well culture plates (Sarstedt AG and Co., Nümbrecht, Germany). After reaching confluency, TiO₂ particles of indicated sizes or water (control) were added at a final concentration of 40 µg cm⁻² cell growth surface. At indicated time points cells were washed twice with PBS and proceeded to RNA extraction. RNA was extracted using the GeneJET RNA Purification Kit (Thermo Fisher Scientific GmbH, Dreieich, Germany) and first-strand cDNA synthesis was prepared using a RevertAid™ H Minus First Strand cDNA Synthesis Kit (Thermo Fisher Scientific GmbH, Dreieich, Germany) as described by the manufacturer. Real-time PCR was performed on a 7500 Real-Time PCR Systems (Life Technologies Inc., Carlsbad, CA, USA) using HOT FIREPol^® EvaGreen^® qPCR Mix Plus ROX (Solis BioDyne, Tartu, Estonia). Sequences of primers (used at 0.2 µM) were as follows: B2M (Beta-2 microglobulin) forward: GCAAGGACTGGTCTTTCTATCT, reverse: TAACTATCTTGGGCTGTG-ACAAA; CCL2 (chemokine (C–C motif) ligand 2) forward: CCCAAAGAAGCTGTGATCTTCA; reverse: TCTGGGGAAAGCTAGGGGAA, CCL20 (chemokine (C–C motif) ligand 20) forward: CGAATCAGAAGCAGCAAGCAA, reverse: TTGCGCACACAGACAACTTT; CXCL3 (chemokine (C–X–C motif) ligand 3) forward: CCCAAACCGAAGTCATAGCCA, reverse: ACCCTGCAGGAAG-TGTCAA; CXCL10 (chemokine (C–X–C motif) ligand 10) forward: GCCATTCTGATTTGCTGCCTT, reverse: GCTCCCCTCTGGTTTTAAGGA; GAPDH (glyceraldehyde 3-phosphate dehydrogenase) forward: AGAGCACAAGAGGAAGAGAGAG, reverse: GGTTGAGCACAGGGTACTTTATT; IL8 (interleukin 8) forward: CACCGGAAGGAACCATCTCA, reverse: TGGCAAAACTGCACCTTC-ACA. Parameters for qPCR were as follows: 95 °C for 15 min, 40 cycles of 10 s at 95 °C, 30 s at 60 °C, and 30 s at 72 °C. After cycling, melting curve analysis was performed. Expressions of the different genes was normalized to the expressions of GAPDH and B2M and compared to control.

Caco-2^nfkb-RE [9] cells were seeded in 96-well plates (Corning Incorporated, New York, NY, USA). At confluency, inhibitors of endocytosis were added to the cells and incubated as indicated below. Subsequently, cells were treated with TiO₂ NPs or water (control) at a final concentration of 40 µg cm⁻² cell growth surface. After 6 h cells were harvested and luciferase activity was measured for 5 s in a CHAMELEON™ V plate reader (Hidex, Finland) using Beetle-Juice substrate (PJK GmbH, Kleinblittersdorf, Germany) according to the manufacturer’s instructions. Luciferase activity is expressed in relation to untreated controls.

Detection of lysosomes was performed by fluorescence microscopy. Caco-2 cells were seeded in eight-well on cover glass II (Sarstedt AG and Co., Nümbrecht, Germany). At confluency, prepared medium with TiO₂ NPs or water (control) was added. Simultaneously LysoTracker^® Red DND-99 (Molecular Probes, Eugene, OR, USA) was added to medium at a final concentration of 50 nM. After incubation time of 3 h cells were observed by fluorescence microscopy. Co-localization of lysosomes and NPs was detected by different filter sets. NPs were visualized by using FITC filter set and LysoTracker^® Red DND-99 by using filter set appropriate to Texas Red^® dye. To investigate intracellular distribution of the NPs, confocal laser scanning microscopy was employed. Caco-2 cells were seeded in a removable 12-well chamber (ibidi GmbH, Munich, Germany). At confluency, prepared medium with TiO₂ NPs or water (control) was added. After incubation time of 6 h cells were harvested and stained for actin with Alexa Fluor 633 Phalloidin (Molecular Probes, Eugene, OR, USA) as described by the manufacturer. Nuclei were stained with Dapi-Fluoromount-G™ clear mounting media (SouthernBiotech™, Birmingham, AL, USA). NPs were excited by a laser at 488 nm and the emission was measured from 490 nm to 501 nm. Excitation of phalloidin took place by a laser at 633 nm and emission was measured from 642 nm to 655 nm to allow a clear separation of both signals. Confocal and fluorescence images were acquired with a confocal laser scanning microscope TCS SP8 (Leica Microsystems GmbH, Wetzlar, Germany) using a 40× objective and 63× glycerol immersion objective.

Caco-2 cells were seeded in six-well plates (8.87 cm⁻² per well) (Sarstedt, Germany) at a density of 1–2 × 10⁵ cells cm⁻². At confluency, prepared medium with 10 nm TiO2 NPs was added to the cells at a final concentration of 40 µg cm⁻² cell growth surface or remained untreated. EGF at a concentration of 100 ng mL⁻¹ was used as a positive control in the EGFR/ERK signaling pathway as described [59]. Cells were washed twice with cold PBS and lysed 20 min on ice in RIPA buffer (Sigma-Aldrich, St. Louis, MO, USA) added with inhibitors (1 mM PMSF, 1 mM Na₃VO₄, 1× complete protease inhibitor cocktail (Santa Cruz Biotechnology, Santa Cruz, CA, USA), 1× phosphatase inhibitor cocktail 3 (Sigma-Aldrich, St. Louis, MO, USA). Lysates were clarified by centrifugation (12,000× g for 6 min) and protein concentration was measured using bicinchoninic acid assay (BCA) reagent. 40 µg protein from each preparation were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a nitrocellulose membrane (Carl Roth GmbH + Co.KG, Karlsruhe, Germany). Membranes were blocked in 5% (w/v) BSA/TBST (20 mM Tris–HCl (pH 7.6), 150 mM NaCl, and 0.1% (v/v) Tween 20) for 1 h at room temperature. Primary antibody incubation p-ERK1/2 1:200 (sc-7383, Santa Cruz Biotechnology, Santa Cruz, CA, USA) in 5% BSA/TBST took place at 4 °C overnight, followed by secondary antibody incubation 1:500 in TBST (anti-mouse HRP (PAB10782, Abnova, Taiwan)) for 2 h at room temperature. The antibody was detected with enhanced chemiluminescence (ECL) reagent (MBL International Corporation, Woburn, MA, USA) and visualized with Fusion Solo S (Vilber Lourmat Deutschland GmbH, Eberhardzell, Germany). Densitometry of the bands was analyzed using FusionCapt Advance Solo 4s software and the ratio of p-ERK to total protein, detected by Ponceau staining, was determined. Expressions of the treated samples were set in relation to the untreated control sample. Experiments were performed at least three times.

Binding of ELK1 was investigated by non-radioactive EMSA using a LightShift EMSA Optimization and Control Kit (Thermo Fisher Scientific Inc., Waltham, MA, USA) according to manufacturer’s protocol. In detail, Caco-2 cells were seeded in six-well plates (Sarstedt AG and Co, Nümbrecht, Germany) at a density of 1–2 × 10⁵ cells cm⁻². At confluency, cells were stimulated with 10 nm TiO₂ NPs at a final concentration of 40 µg cm⁻² cell growth surface or remained unstimulated. Cells were washed twice with ice cold PBS and lysed 20 min on ice in lysis buffer (10 mM HEPES, 1.5 mM MgCl₂, 10 mM KCl, 1 mM DTT, 1 mM PMSF, 1× complete protease inhibitor cocktail (Santa Cruz Biotechnology, Santa Cruz, CA, USA)) adapted from [60]. Cells were scraped, placed in a microliter tube, and centrifuged 20 min at 11,000× g at 4 °C to separate cytosolic and nuclear fractions. Supernatant was decanted and pellet was suspended in extraction buffer to lyse nuclei (20 mM HEPES, 1.5 mM MgCl₂, 0.42 M NaCl, 0.2 mM EDTA, 25% (v/v) glycerol, 1 mM DTT, 1 mM PMSF, 1× complete protease inhibitor cocktail, [60]). 2 µg of nucleus extraction were used for binding reaction with biotin-labeled DNA for ELK1: 5′-TTTGCAAAATGCAGGAATTGTTTTCACAGT-3′ [61]. Samples were separated by 6% native polyacrylamide gel and transferred to a nylon membrane (GE Healthcare, Little Chalfont, UK). Biotin was detected with peroxidase-coupled streptavidin and ECL (Thermo Fisher Scientific Inc., Waltham, MA, USA) and visualized and quantified with Fusion Solo S (Vilber Lourmat Deutschland GmbH, Eberhardzell, Germany). Densitometry of the binding complex was analyzed using FusionCapt Advance Solo 4s software and expressions of the treated samples were set in relation to the untreated control sample. Experiments were performed at least three times.

To investigate the mechanisms involved in the TiO₂ NPs-induced inflammatory response in Caco-2 cells, we used various inhibitors of endocytic processes (Figure 1, Table 1). Confluent monolayers of Caco-2 cells were pre-incubated with inhibitors (cytochalasin D (AppliChem, Darmstadt, Germany), nocodazole (AppliChem, Darmstadt, Germany), nystatin (AppliChem, Darmstadt, Germany), dynasore (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA), chlorpromazine (Sigma-Aldrich, St. Louis, MS, USA), BIBX 1382 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA), and CL-387785 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA)). The final concentrations that were used in medium, incubation time, and effect of inhibitors are summarized in Table 1. After pre-incubation, the medium with TiO₂ NPs or water (control) was added at a final concentration of 40 µg cm⁻² cell growth surface and incubated for 3 h or 6 h. Cytotoxic activity was tested by the ToxiLight™ bioassay kit (Lonza Group AG, Walkersville, MD, USA). No cytotoxic effect was observed for inhibitors at the used concentration (data not shown). After particle treatment, cells were washed twice with PBS and further treated with the stated methods.

Epithelial growth factor (EGF) at a concentration of 100 ng mL⁻¹ was used as a positive control in the EGFR/ERK signaling pathway, as described [61]. A confluent monolayer of Caco-2 cells was incubated for 1 h with 50 µM BIBX 1382 and subsequently stimulated with NPs, EGF, or left unstimulated. After treatment, cells were washed twice with cold PBS and processed as described above. Results of the treated cells are presented as the fold change to untreated cells.

We performed every experiment with, at least, replicate samples for each treatment with the appropriate controls carried out on three different days (i.e., at least technical duplicates with three biological replicates). Data of RT-qPCR were analyzed using the delta-delta C_t method [65]. Results are presented as means ± standard error (SEM) and expressed as the fold change to the untreated control. Statistical analysis was carried out using GraphPad Prism 4 software (GraphPad Software, Inc., La Jolla, CA, USA). After testing for Gausian distribution (Kolmogorov-Smirnoff test) and homogeneity of variance (Levene test) comparisons of the groups were either carried out by non-parametric analysis, followed by Bonferroni post-tests or by ANOVA, followed by Tukey’s multiple comparison test. Differences were considered to be statistically significant at p < 0.05.