Effects of physicochemical properties of TiO2 nanomaterials for pulmonary inflammation, acute phase response and alveolar proteinosis in intratracheally exposed mice.

Nanomaterial (NM) characteristics may affect the pulmonary toxicity and inflammatory response, including specific surface area, size, shape, crystal phase or other surface characteristics. Grouping of TiO2 in hazard assessment might be challenging because of variation in physicochemical properties. We exposed C57BL/6 J mice to a single dose of four anatase TiO2 NMs with various sizes and shapes by intratracheal instillation and assessed the pulmonary toxicity 1, 3, 28, 90 or 180 days post-exposure. The quartz DQ12 was included as benchmark particle. Pulmonary responses were evaluated by histopathology, electron microscopy, bronchoalveolar lavage (BAL) fluid cell composition and acute phase response. Genotoxicity was evaluated by DNA strand break levels in BAL cells, lung and liver in the comet assay. Multiple regression analyses were applied to identify specific TiO2 NMs properties important for the pulmonary inflammation and acute phase response. The TiO2 NMs induced similar inflammatory responses when surface area was used as dose metrics, although inflammatory and acute phase response was greatest and more persistent for the TiO2 tube. Similar histopathological changes were observed for the TiO2 tube and DQ12 including pulmonary alveolar proteinosis indicating profound effects related to the tube shape. Comparison with previously published data on rutile TiO2 NMs indicated that rutile TiO2 NMs were more inflammogenic in terms of neutrophil influx than anatase TiO2 NMs when normalized to total deposited surface area. Overall, the results suggest that specific surface area, crystal phase and shape of TiO2 NMs are important predictors for the observed pulmonary effects of TiO2 NMs.


Introduction
The global increase in production and application of titanium dioxide nanomaterials (TiO 2 NMs) in a wide range of industrial and consumer products leads to potential exposure-related adverse health effects for workers. In the workplace, exposure is most likely to occur via inhalation and thus cause lung inflammation as well as systemic effects. TiO 2 NMs are poorly soluble and have been considered as low toxicity particles. However, The International Agency for Research on Cancer (IARC) has classified TiO 2 as a Group 2B carcinogen (possibly carcinogenic to humans) based on sufficient evidence in animal experiments (IARC, 2010) as lung tumors developed in rats after two years of chronic exposure to 250 mg/m 3 of fine-sized rutile TiO 2 (Lee et al., 1985) and to 10 mg/m 3 of nano-sized TiO 2 P25 (Heinrich et al., 1995). A large number of rodent studies have reported increased pulmonary inflammation, including neutrophil influx in bronchoalveolar lavage (BAL) fluid, after acute, sub-acute and sub-chronic exposure to TiO 2 NMs by either inhalation or intratracheal (i.t.) instillation (Shi et al., 2013;Hadrup et al., 2017). Neutrophil influx in BAL fluid has been shown to correlate closely with pulmonary acute phase response in terms of increased serum amyloid A 3 (Saa3) mRNA levels in lung tissue of mice exposed to TiO 2 NMs by i.t. instillation and by inhalation (Halappanavar et al., 2011;Saber et al., 2013). It has been suggested that the SAA produced in the lungs enters systemic circulation as part of high density lipoprotein molecules, causing reversal of the cholesterol flow leading to increased formation of foam cells and plaque progression, and that this contributes significantly to the pathogenesis of cardiovascular disease (Saber et al., 2014). NM characteristics may affect the pulmonary toxicity and inflammatory response, including specific surface area, size, shape, crystal phase, chemical composition, charge or other surface characteristics. The importance of particle size has been discussed widely in the past as the smaller sized NMs and accompanying increased specific surface area entails increased toxicity and translocation, greater lung retention and slower pulmonary clearance (Ferin et al., J o u r n a l P r e -p r o o f Journal Pre-proof mice/group). The animal experiments were performed over several weeks and vehicle controls were included on each exposure day. The vehicle controls were combined for each post-exposure day; N = 15 (day 1 and 3) and N = 10 (day 28, 90, 180) per dose group. The mice had access to food (Altromin 1324) and water ad libitum. The housing conditions have been described in detail elsewhere (Hadrup et al., 2017). At 8 weeks of age the mice were anaesthetized and exposed to 50 µl anatase TiO 2 NMs, DQ12 or vehicle by single i.t. instillation. Mice for 180 days exposure only received the highest dose of 162 µg. For histology doses of 500 and 1000 µg/mouse DQ12 were included. The instillation procedure has been described in detail previously (Jackson et al., 2010;Saber et al., 2012a;Hadrup et al., 2017). In brief, the mice were placed on their backs on a 40⁰ slope. A diode light was placed touching the larynx and the trachea was intubated using a catheter.
The 50 µl NM suspension was instilled followed by 150 ml air. The catheter was removed and the mouse transferred to a vertical hanging position with the head up. This ensures that the administered material is maintained in the lung and breathing was observed to assure that airways were not blocked (Jackson et al., 2010).
Animal experiments were performed according to EC Directive 2010/63/UE in compliance with the handling guidelines established by the Danish government and permits from the Experimental Animal Inspectorate (no. 2015-15-0201-00465). Prior to the study, the experimental protocols were approved by the local Animal Ethics Council.
were snap frozen in cryotubes in liquid nitrogen and stored at 80 °C until isolation of RNA for mRNA expression analysis and sample preparation for the comet assay.

BAL cell differential counting
The separated BAL cells were re-suspended in 100 µl HAMF12 medium with 10% fetal bovine serum. 40 µl of the cell suspension was mixed with 160 µl medium containing 10% dimethyl sulfoxide (DMSO) and stored at -80 °C for later analysis in the comet assay. The total number of cells and of dead cells was determined from 20 µl diluted cell suspension by Allerød,Denmark). 40 µl of the fresh cell suspension was collected on microscope slides by centrifugation at 60 x g for 4 min and then the cells were fixed with 96% ethanol and stained with May-Grünwald-Giemsa stain. Differential counts of macrophages, neutrophils, lymphocytes, eosinophils, and epithelial cells were determined by counting 200 cells/sample under light microscope (100x magnification).

mRNA expression of Saa
Pulmonary and hepatic acute phase response was assessed by the measurement of Saa3 and Saa1 mRNA expression levels, respectively, as described previously (Saber et al., 2009;Wallin et al., 2017). In brief, total RNA was isolated from frozen liver and lung tissue using Maxwell® 16 LEV simply RNA Tissue Kit (AS1280, Promega, USA) according to the manufacturer's protocol.
Complementary DNA (cDNA) was prepared using TaqMan® reverse transcription reagents (Applied Biosystems, USA) as described by manufacturer's protocol. Total RNA and cDNA concentrations were measured on NanoDrop 2000c (ThermoFisher, USA). The Saa3 and Saa1 gene expression was determined by real-time reverse transcriptase polymerase chain reaction (RT-PCR) using TaqMan pre-developed reagents (Applied Biosystems, USA) and 18S as a reference gene.
The samples were run in triplicates using ViiA7 Real-Time PCR detector (Applied Biosystems, J o u r n a l P r e -p r o o f USA). The relative expression was measured by the comparative C T method. Negative controls were included in each run of the analysis. The day-to-day variation for the plate control in the analyses was 17.5 % and 4 % for Saa3 and Saa1, respectively.

Comet assay
DNA strand breaks were determined by the comet assay on BAL cells, lung and liver tissue as described in (Jackson et al., 2013b;Modrzynska et al., 2018). In brief, single-cell suspensions were obtained by homogenization of frozen liver and lung pieces in ice-cold Merchant's buffer through a stainless steel mesh. The BAL cells were thawed in a 37 °C water bath before diluting with Merchant's buffer. The cell suspensions were embedded in agarose (0.7%) on microscope Trevigen 20-Well CometSlides™. The slides were immersed in cold lysing solution and stored overnight at 4 °C. Samples were treated with alkaline buffer and electrophoresis with circulating ice-cold electrophoresis buffer was performed (25 min, 38 V, 70-77A). Thereafter, slides were washed in neutralization buffer (0.4 M Tris, pH 7.5), fixed with 96% ethanol and stained with SYBRGreen®.
DNA strand breaks were quantified as percentage of DNA in the comet tail (%TDNA) and as the comet tail length (TL). The comets were scored by the fully automated PathFinder™ system (IMSTAR, France). In order to control the day-to-day variation both negative (A549 human lung epithelial cell line treated with PBS for 30 min at 4 °C) and positive (A549 human lung epithelial cell line treated with 60 μM H 2 O 2 for 30 min at 4 °C) controls were included. The samples were run over eight experiment days. The day-to-day variation for the negative and positive controls was 9 % and 18 % for TL, respectively.

Histopathology
At 28, 90 or 180 days post-exposure mice were weighed and anesthetized. Lungs were filled slowly with 4% formalin under 30 cm water column pressure. A knot was made around the trachea to J o u r n a l P r e -p r o o f secure formaldehyde in lungs to fixate tissue in "inflated state". Lungs were then removed and placed in 4% neutral buffered formaldehyde solution for 24 hours as previously described (Poulsen et al., 2016b). After fixation the samples were trimmed, dehydrated on a Leica ASP300S (Leica Systems) and embedded in paraffin. Sections were cut at 3 µm on a Microm HM 355S Microtome (Thermo Scientific™). Sections for light microscopic examinations were stained with Haematoxylin and Eosin (H&E-staining) or Sirius Red staining. The sections were examined by light microscopy using a Leica DM 4000B microscope equipped with a Leica DFC 480 camera and an Olympus BX43 microscope with a Nikon DS-Fi2 camera. Lymphocytic and macrophage infiltrates were defined as areas of tissue where the density of lymphocytes and macrophages were higher than the background. The minimum requirement was that they should contain 20 or more lymphocytes or 5 or more macrophages. The proteinaceous debris in the alveolar spaces was compatible with that seen in pulmonary alveolar proteinosis (PAP). The number refers to percentage of afflicted alveoli in the lobe. Granuloma formations refer to foreign body type granulomas with material surrounded by at least a few adherent macrophages and exclude giant cells.

Darkfield microscopy
Standard darkfield and Cytoviva enhanced darkfield hyperspectral system (Auburn, AL, USA) was used to detect particles in lung tissue by scanning histological sections at 10x and 40x magnification. Images were acquired at 10x on a Leitz Laborlux K microscope with a Leica MC170 HD camera and a 100x using an Olympus BX 43 microscope with a Qimaging Retiga4000R camera.

Electron microscopy
J o u r n a l P r e -p r o o f Samples were fixed in 2.5% glutaraldehyde and post fixed in 1% osmium tetroxide, dehydrated and embedded in LX-112. Thin sections were collected on carbon coated copper grids, stained with uranyl acetate and lead citrate. Grids were analyzed with a transmission electron microscope operated at 100kV (JEM-1220, Jeol ltd, Japan, Tokyo) and photographed with Veleta TEM CCD camera (Olympus Soft Imaging Solutions GmbH, Germany). Preparation of TEM samples were done in Electron Microscopy Unit, University of Helsinki.

Statistical analysis
The data sets of BAL cells differential counting, mRNA expression, DNA strand breaks and histopathology were analyzed using the software package Graph Pad Prism 8.1.2. (Graph Pad Software Inc., La Jolla, CA, USA). All data are expressed as mean ± standard deviation. Data were tested for normality using the Shapiro-Wilks test and for variance homogeneity using the Bartlett's test. It was not possible to fulfill the normality and variance homogeneity criteria for most data by neither logarithmic nor cube root transformations. Therefore, the data were analyzed by the nonparametric Kruskal-Wallis test followed by Dunn's multiple comparison method as post hoc to test the differences between the test groups. P-value ≤0.05 was considered significant.
To identify important TiO 2 NMs properties multiple regression analyses were conducted on the following endpoints: neutrophil influx in the BAL fluid and pulmonary Saa3 expression. The four anatase and the five rutile TiO 2 NMs were all included in the analyses. Selected physicochemical properties were: BET surface area, anatase phase, tube shape, and surface modification (NRCWE-002 and UV-Titan L181). Initially, a Pearson Correlation analysis was used to investigate the pairwise associations between physicochemical parameters (BET surface area, Anatase, Tube and Modified).
No correlations between parameters were observed (results not shown), and all parameters were included as independent variables in the following multiple regression analyses. Multiple regression J o u r n a l P r e -p r o o f analyses investigating the relationship between physicochemical properties (BET surface area, Anatase, Tube and Modified) and neutrophil influx in the BAL fluid (all time points) or pulmonary Saa3 expression (day 1 and 3 only), were performed. BET surface area was transformed using log(BET)/log(1.25), so the estimated effect corresponded to a 25% increase in BET. Statistical significance was determined at the 0.01 level in the multiple regression analyses, since no other correction for mass-significance was performed. The Pearson Correlations and all multiple regression analyses were performed in SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

Physicochemical properties
The main characteristics of the anatase TiO 2 NMs and DQ12 are summarized in Table 1. Anatase TiO 2 NMs, labeled as TiO 2 NM-1, TiO 2 NM-2, TiO 2 tube, TiO 2 cube, and DQ12 were investigated with transmission electron microscopy (TEM) to determine particle morphology and size (Table 1).
According to the supplier, TiO 2 NM-1 and TiO 2 NM-2 should have diameters of 15 and 100 nm, respectively. However, TEM of the anatase TiO 2 NMs ( Fig. 1) revealed that the average diameter of the TiO 2 NM-1 particles (Fig. 1A) were around 30 nm and displayed a more rectangular than spherical shape. Only small amounts of the TiO 2 NM-1 particles had diameters around 10 nm. In addition, tiny particles with no certain shape and sizes below 10 nm were observed. The average diameter of the TiO 2 NM-2 particles (Fig. 1B) was 16 to 32 nm, instead of 100 nm which was the size reported by the manufacturer. These particles had a rectangular to spherical shape. The largest TiO 2 NM-2 particles were about 50 nm in diameter. However, these were not spherical, but rather elongated in one direction. The BET surface area of TiO 2 NM-1 and TiO 2 NM-2 was 85 and 74 J o u r n a l P r e -p r o o f m 2 /g, respectively, indicating similar particle diameter sizes as confirmed by TEM. The TiO 2 tube NMs showed distinctive aggregates composed of tubes, while individual tubes was rarely observed (Fig. 1C). The outer diameter of the TiO 2 tube NMs varied from 6 to 11 nm and the length were up to 500 nm. The BET surface area was 154 m 2 /g. The TiO 2 cube NMs had a rectangular shape (cube and cube-like (elongated in one direction)). The average size of the shorter side was about 17 nm and the longer side up to 26 nm. The particles were crystalline (Fig. 1D) and the BET surface area was 96.9 m 2 /g. TEM of the DQ12 sample showed that the particles had different shapes and were unequally sized. The size ranged from 50 nm to 400 nm and aggregates, exceeding few micrometers, were observed. There was no presence of micropores as calculated by the t-Plotmethod for any of the anatase TiO 2 NMs. X-ray Diffraction (XRD) patterns of TiO 2 NMs (Supplementary material, Fig. S1) showed that TiO 2 NM-1 and TiO 2 NM-2 contained a small amount of rutile. The amount of rutile calculated using Spurr's formula was 11.5 wt.% and 5.6 wt.%, for NM-1 and NM-2 respectively. TiO 2 NM-2 also contained some unidentified impurities (arrows in Supplementary material, Fig. S1). For TiO 2 tube and TiO 2 cube all diffraction peaks corresponded to anatase structure (Supplementary material, Fig.   S1).
Dynamic Light Scattering (DLS) was used to determine the hydrodynamic number-based size distributions of TiO 2 NMs and DQ12 in the instillation suspensions (3.24 mg/ml in Nanopure water with 2 % mouse serum). The NM suspensions were generally well-dispersed and the size distribution showed unimodal peaks for TiO 2 NM-1, TiO 2 cube and DQ12 at 68 nm, 38 nm and 210 nm, respectively (Supplementary material, Fig. S2). TiO 2 NM-2 was bimodally distributed with a major peak at 50 nm and a minor broad peak at 18-38 nm. In addition, TiO 2 tube was bimodally distributed with a major peak at 60 nm and a narrow peak at 21 nm (Supplementary material, Fig.   S2). The intensity-based z-average size and polydispersity index (PI) are shown in the J o u r n a l P r e -p r o o f Supplementary material, Table S1. The PI was between 0.4-0.7 indicating some polydispersity in the suspensions.
The endotoxin contents were measured using the Limulus Amebocyte lysate enzyme kit (LAL) and low levels were found in all NM suspensions (Table 1).

Acellular oxidation potential
The ability to generate reactive oxygen species (ROS) was determined using the acellular 2´,7´diclorodihydrofluorescein diactetate DCFH 2 -DA assay as described previously (Jacobsen et al., 2008). All TiO 2 NMs and DQ12 generated ROS in a dose-dependent manner (Supplementary material, Fig. S3). There were no clear differences between rutile and anatase TiO 2 NMs, except that the rutile TiO 2 NM (NRCWE-002) induced a 7-fold larger ROS level than the others did.

Visualization of TiO 2 particles in the lung tissue
To assess the distribution and persistence of the NMs in the lung, histological sections of lung tissues were imaged using enhanced darkfield microscopy (

Electron microscopy
TEM revealed that the NMs in the lungs were almost exclusively present in macrophages and mostly seen in phagosomal vesicles (Figs. 5A, C, D). As also noted in bright-and darkfield microscopy the TiO 2 cube formed large aggregates (Fig. 5B). The material in the aggregates appeared to be unchanged (Fig. 5B, insert).
The acellular organic debris associated with TiO 2 tube and DQ12 exposure visualized in darkfield microscopy was quite similar when using TEM (Fig. 6). Structures consistent with tubular myelin and lamellar bodies were identified. The findings in Fig. 6 conform to that seen in association with pulmonary alveolar proteinosis after quartz exposure (Corrin and King, 1970;Seymour and Presneill, 2002).

Histopathological analysis
The TiO 2 tube caused statistically significant increased inflammatory changes in lung tissue as defined by numbers of lymphocytic and macrophage infiltrates in the same manner as DQ12 compared to the control (Figs. 7 and 8). The other anatase TiO 2 NMs caused moderate inflammatory changes in the lung tissue. Macrophage infiltrations were not observed for DQ12 at 162 µg, in contrast to increased levels for TiO 2 tube at the same dose (Fig. 8). In addition to this a proteinaceous material was observed in the alveolar spaces of TiO 2 tube-exposed mice at day 28 post-exposure which was still present but at lower levels at day 90 post-exposure (Fig. 9). The same J o u r n a l P r e -p r o o f type of proteinaceous material was observed after exposure to DQ12, however the time course was different than for TiO 2 tube. For DQ12, at the 162 µg dose, the changes appeared later and persisted longer ( Fig. 9). Increasing the dose of DQ12 to 500 µg or 1000 µg resulted in more changes at day 28 post-exposure (Fig. 9A). Proteinaceous material in the alveolar spaces in quartz-exposed animals and humans are a well-known effect and are termed pulmonary alveolar proteinosis (PAP) (Corrin and King, 1970;Seymour and Presneill, 2002). No granulomas were observed in the present study (data not shown). No signs of fibrosis were observed (data not shown). Dose-dependent neutrophil influx was observed at day 1 and 3 post-exposure for DQ12 and all the anatase TiO 2 NMs with exception of TiO 2 cube (Fig. 10). The number of neutrophils decreased over time after exposure. For the TiO 2 tube, the number of neutrophils, macrophages, lymphocytes and total BAL cells was statistically significant increased at day 28 post-exposure, which were the only observed inflammatory responses at that time-point. No statistically significant inflammatory responses were observed at the lowest dose of 18 µg or at 90 days post-exposure, except for a borderline statistically significant increase in the number of lymphocytes for TiO 2 tube 90 days post-exposure (162 µg, P-value: 0.068). However, statistically significant increased levels of neutrophils (7-fold) and lymphocytes (6-fold) was seen for TiO 2 tube 180 day post exposure to 162 J o u r n a l P r e -p r o o f µg. Overall, the TiO 2 tube generated a higher inflammatory response than the other TiO 2 NMs, whereas TiO 2 cube did not generate an inflammatory response.

Acute phase response
Pulmonary and hepatic acute phase response was assessed using Saa3 and Saa1 mRNA expression levels as biomarkers of pulmonary and hepatic acute phase response, respectively (Poulsen et al., 2017). The Saa3 gene expression was statistically significantly increased 1-day post-exposure compared to vehicle control at the highest dose (162 µg) for DQ12 and all anatase TiO 2 NMs except for TiO 2 cube (Fig. 11). The TiO 2 tube and DQ12 also induced statistically significantly increased Saa3 mRNA levels at the middle dose (54 µg) 1-day post-exposure. The significant increase of Saa3 mRNA expression was persistent at day 3 post-exposure for TiO 2 tube (54 and 162 µg) and for DQ12 (162 µg), though the level was 10 times lower than observed at day 1. A strong correlation was observed for Saa3 mRNA expression and neutrophil influx day 1 post-exposure (Supplementary material, Fig. S4).
One day post-exposure hepatic Saa1 mRNA expression was statistically significant increased only for TiO 2 tube at the highest exposure dose (70-fold) (Fig. 12). Saa1 expression levels at postexposure day 3 were not analyzed.

Comet assay
DNA strand break levels were assessed in BAL cells, lung and liver tissue by the comet assay (Supplementary material, Table S4). There were only few significant increases in DNA strand break levels. However, DNA strand break levels were also observed to be statistically significantly decreased for some of the TiO 2 NMs, especially in the lung tissue, at different time points. The observed changes were in general not dose-dependent and considered as chance findings.

J o u r n a l P r e -p r o o f
The toxicological response to anatase TiO 2 NMs was compared to previously published studies on rutile TiO 2 NMs (Supplementary material, Table S2, (Jacobsen et al., 2009;Bourdon et al., 2012;Saber et al., 2012b). Carbon black (CB), (Printex 90), was included as reference material to allow comparisons across studies. CB (162 µg/mouse) used in the previous studies on rutile TiO 2 NMs resulted in similar levels of neutrophil influx as observed in the present study on anatase TiO 2 NMs ( Fig. 13).
This comparison of anatase and rutile TiO 2 NMs was performed for acute inflammation as measured by neutrophil influx into the lung (BAL fluid) adhering to the method described by Schmid andStoeger 2016 (Schmid andStoeger, 2016). This approach calls for normalization of the neutrophil number to the total number of cells in the BAL (PMN in %) and normalization of the instilled NM surface area dose to the average weight of mouse lungs (0.18 g). We found strong logistic correlations between surface area dose and level of neutrophil influx 1-day post-exposure for anatase TiO 2 NMs without TiO 2 tube (R 2 = 0.95) and rutile TiO 2 NMs (R 2 = 0.99), Fig. 14A.

J o u r n a l P r e -p r o o f
Interestingly, also Printex 90 (P90) shows the same surface-specific inflammogenicity as rutile TiO 2 NMs (dose 30%PMN = 220 cm 2 /g).

Multiple regression analyses
With the aim of identifying TiO 2 NMs properties important for their toxicity, the effect of dose, BET surface area, crystal phase and shape on neutrophil influx and Saa3 expression after i.t. exposure to both anatase and rutile TiO 2 NMs was analyzed by multiple regression (Table 3 and 4). Mass dose significantly predicted neutrophil influx on all post-exposure days. Increasing BET surface area significantly predicted enhanced neutrophil influx 3 and 28 days post-exposure. The anatase phase predicted lower neutrophil influx 1 and 3 days after exposure. The tube shape significantly predicted increased neutrophil influx on all post-exposure days. When tube shape was included as a variable, the effect of BET surface area disappeared, suggesting that the crystal phase anatase/rutile and shape (tube) might be more important predictors than the BET surface area.
Mass dose and BET surface area (i.e. surface area dose) significantly predicted Saa3 expression on all post-exposure days (1 and 3 days). The tube shape significantly predicted increased Saa3 expression at both post-exposure days. When tube shape was included as a variable, again the effect of BET surface area disappeared. At 3 days post-exposure the anatase phase predicted lower neutrophil influx.
There was no effect of surface modifications (rutile NRCWE-002 and rutile UV-Titan L181), data not shown.

J o u r n a l P r e -p r o o f
Inhalation exposure is the gold standard for risk assessment. However, the i.t. instillation technique used in the present study is generally well-accepted for hazard ranking of different NMs (Warheit et al., 2005;Baisch et al., 2014). The i.t. instillation technique has shown an even distribution of NMs including MWCNT across most lung-tissue sections (Mikkelsen et al., 2011;Poulsen et al., 2016).
However, compared to inhalation the i.t. instillation results in a more patchy distribution as shown using fluorescent nanoparticles (Yang et al., 2019). Recently, a study concludes that neutrophil influx from i.t.-and inhalation-exposed rats correlated with the estimated pulmonary deposited surface area across both types MWCNT and types of exposure at two different time points (Gate et al., 2019). Additionally, it has been shown that global pulmonary transcriptomic pattern following i.t. instillation and inhalation of TiO 2 NMs in mice is remarkably comparable (Halappanavar et al., 2011;Husain et al., 2013), suggesting common biological responses between administration methods. The doses used in the present study reflect occupationally relevant exposure levels and allow comparison with previously performed animal studies using the same dose levels (e.g. the rutile TiO 2 ).
The study was designed to cover a wide NM size range. Unexpectedly, the BETs for the two TiO 2 NMs with a small and large diameter size as reported by the manufacturer (TiO 2 NM-1 and TiO 2 NM-2) showed to be similar. This was confirmed by TEM. Thus, the optimal study design having a wide BET range with larger surface areas was compromised.
The greatest overall inflammatory response was seen for neutrophil influx 1 and 3 day postexposure, especially for the TiO 2 tube and DQ12. In addition, the total number of cells, lymphocytes and eosinophils were significantly increased for the TiO 2 tube. In general, the other anatase TiO 2 NMs were somewhat less responsive, including the TiO 2 cube, which induced less inflammation than TiO 2 NM-1 and TiO 2 NM-2 despite its 20-30 % larger specific surface area. The TiO 2 cube NMs had a peculiar tendency to form large aggregates in the lungs, which could be J o u r n a l P r e -p r o o f related to the relatively modest inflammatory response. DLS data for TiO 2 cube confirms a homogeneous instillation suspension without bigger aggregates, suggesting that this phenomenon happened in the lung. Persistent dose-dependent increases in the influx of inflammatory cells, including neutrophils, were observed up to 28 and 180 days post-exposure only for the TiO 2 tube suggesting that the difference in shape may be an important predictor for chronic lung inflammation.
We used Saa3 mRNA levels in the lung as biomarker of the pulmonary acute phase response and Saa1 mRNA levels as biomarker for the hepatic acute phase response. Previous studies reported significant correlations between Saa3 mRNA and protein levels in mice after MWCNTs exposure (Saber et al., 2014;Poulsen et al., 2015a;Poulsen et al., 2017). Consistent with the observed pulmonary inflammation the Saa3 mRNA levels were dose-dependently increased 1 and 3 day postexposure, again especially for the TiO 2 tube and DQ12. As in the present study, we have previously shown that pulmonary exposure to different NMs induce a pulmonary acute phase in parallel with the pulmonary inflammatory response (Bourdon et al., 2012;Saber et al., 2013;Poulsen et al., 2015a;Husain et al., 2013;Poulsen et al., 2017;Hadrup et al., 2019a;Hadrup et al., 2019b;Barfod et al., 2020).
The strong pulmonary inflammation of TiO 2 tube was accompanied by a hepatic acute phase response measured as increased Saa1 mRNA levels at the highest dose 1-day post-exposure (70fold increase). In previous studies, instillation of MWCNTs led to increased levels of both Saa1 mRNA and SAA1 protein (Poulsen et al., 2017), whereas CB and rutile UV-Titan L181 TiO 2 did not (Saber et al., 2009;Halappanavar et al., 2011). The hepatic acute phase response may reflect the stronger surface area-dependent pulmonary inflammation induced by the TiO 2 tube, as pulmonary exposure to NRCWE-030 (spherical, rutile TiO 2 ) with a similarly large BET surface area (139 m 2 /g) also induced hepatic Saa1 mRNA levels (Modrzynska et al., 2018). The stronger response by J o u r n a l P r e -p r o o f the TiO 2 tube might be explained by a stronger adsorption of the TiO 2 tube onto cell membranes leading to the formation of membrane-wrapped nanoparticles allowing their relocation and diffusion into systemic circulation as recently shown (Urbancic et al., 2018).
DNA damage has been suggested to be caused by particle mediated ROS production and/or as a secondary consequence of an inflammatory response (Møller et al., 2010). The standard comet assay has been used in TiO 2 genotoxicity studies in cell cultures and animal models with mixed results, which might be due to physicochemical differences (Møller et al., 2017). The few changes (both increases and decreases) in DNA damage levels in BAL cells, lung and liver observed in the present study were considered as chance findings. Contrary, the rutile NMs have been able to induce DNA strand breaks in our previous animal studies (Saber et al., 2012b;Wallin et al., 2017).
The acellular oxidation potential (DCFH 2 -DA assay) of anatase and rutile TiO 2 NMs were in general similar, and cannot solely explain the difference in genotoxicity, although the rutile NRCWE-002 with the highest level of 2´,7´-dichlorofluorescein (DCF) also generated the highest level of DNA damage (Wallin et al., 2017).
Mice exposed to TiO 2 tube and the positive control DQ12 showed increased levels of macrophage and lymphocytic infiltrations as compared to the other NMs. None of the studied particles, including DQ12, induced fibrotic changes. However, both DQ12 and TiO 2 tube exposure resulted in accumulation of alveolar matter that was compatible with pulmonary alveolar proteinosis (PAP) (Corrin and King, 1970;Seymour and Presneill, 2002). Electron microscopy confirmed that the alveolar matter for both TiO 2 tube and DQ12 exposure was similar and that it conforms to the description of PAP in the literature (Crouch et al., 1991). In humans, PAP is a rare condition of mostly unknown etiology, but hereditary forms of this condition are known to exist (Seymour and Presneill, 2002). The secondary type of PAP is known to occur after exposure to crystalline quartz but it has also been associated with exposures to other mineral particles (talc, cement and kaolin) J o u r n a l P r e -p r o o f and metal particles (aluminum, titanium and indium) (Borie et al., 2011). Interestingly at the dose of 162 µg TiO 2 tube, there was a substantial amount of proteinosis 28 days post-exposure.
However, this reaction subsided during time and had disappeared at 180 days post-exposure. In contrast, DQ12 at the same mass dose did not cause accumulation of alveolar matter until 90 days post-exposure and there was still substantial amounts 180 days post-exposure suggesting that the effect of DQ12 is more persistent than that of TiO 2 tube. The persistence of quartz-induced inflammation has been noted previously (Brown et al., 1991). However, it has to be noted that the BET surface area of TiO 2 tube is 15-fold larger than that of DQ12, i.e. for equivalent surface area dosing 2400 µg of DQ12 would have had to be dosed.
The length of the TiO 2 tube is not of the magnitude that it could be expected to cause asbestos-like effects as do some MWCNT. It is interesting that the effect of the TiO 2 tube shows similarities to DQ12 in terms of lymphocytic tissue inflammation and the ability to cause PAP. The effects of quartz have been linked to phagosomal destabilization (Hornung et al., 2008) and it would seem possible that the short fiber shape of the tube affects the lysosomes in a similar way (Købler et al., 2015). The materials in this study definitely end up in lysosome/phagosome structures of macrophages as seen in the EM pictures. We have previously observed PAP in rats i.t.-exposed to a short CNT (NM-403) of similar dimensions as the TiO 2 tube (Gate et al., 2019).
Prolonged pulmonary effects 28-day post exposure have been shown for various CNTs (Poulsen et al., 2013;Poulsen et al., 2015b;Poulsen et al., 2016a;Poulsen et al., 2017), in addition to long retention time in the lungs 28 and 90 days post-exposure (Poulsen et al., 2016a). We recently  (Modrzynska et al., 2018a, Modrzynska et al., 2018b. Whether the long-lasting inflammation is due to a longer retention time is unclear, but all four anatase TiO 2 NMs were still present in lungs 180 days post exposure. However, it is not possible to apply a quantitative approach, as it would require a very accurate sectioning to compare the exact same lung region across the samples. The crystal phase of TiO 2 has been suggested to be important for the toxicological response, where rutile TiO 2 has been considered as an inert form, whereas anatase has been considered as a more active form of TiO 2 (Johnston et al., 2009). Our results show that rutile TiO 2 NMs induces a higher inflammatory response, in terms of neutrophil influx, than similar surface area doses of anatase TiO 2 NMs. A review of in vitro and in vivo TiO 2 toxicity studies concluded that anatase TiO 2 NMs generally were more toxic in terms of cytotoxicity, cell damage, ROS production and inflammation than rutile TiO 2 NMs (Johnston et al., 2009). This was based on relatively few studies investigating the importance of crystallinity. In contrast, a recent toxicogenomic study demonstrated that the overall inflammatory and transcriptional response of mice exposed to anatase TiO 2 NMs was less compared with rutile TiO 2 NMs (Rahman et al., 2017). However, the underlying mechanisms of the different responses related to crystalline structure are unknown.
In the present study, neutrophil influx correlated closely with surface area dose for both anatase (excluding the TiO 2 tube) and rutile TiO 2 NMs, but rutile showed 3.2-fold more neutrophil influx by surface area. CB has been included in our animal studies, as an internal reference particle, to allow comparison of results across studies. CB (162 µg/mouse) used in the previous studies on rutile TiO 2 NMs resulted in similar levels of neutrophil influx as observed in the present study on anatase TiO 2 NM, despite differences in vehicle composition. We are therefore confident in doing such comparisons. We have previously reported dose-dependent neutrophil influx that correlate with BET surface area of both fine and nano-sized TiO 2 and CB (Saber et al., 2012b). A recent J o u r n a l P r e -p r o o f paper retrospectively analyzed animal data from mice and rats on the pulmonary toxicity of i.t.
instilled NMs and concluded that the BET surface area was the biologically most effective dose metric for acute pulmonary inflammation and that so-called low solubility, low toxicity (LSLT) NMs have an equivalent 30% PMN (neutrophil) influx dose range of 175 [85 -405] cm 2 /g lung (Schmid and Stoeger, 2016). Consequently, the neutrophil data presented here indicate that both CB as well as rutile and anatase TiO 2 NMs belong to the LSLT class of NMs.
The inflammatory response of the TiO 2 tube clearly clusters more with rutile than anatase TiO 2 NMs, suggesting that the tube shape is a driver for the effect. In addition, the effect of BET surface area disappeared when the tube was included as a variable in the multiple regression analysis. In light of the substantial evidence for surface area being a strong predictor of inflammation for TiO 2 and carbonaceous NMs (Stoeger et al., 2007), this may indicate that the current data set has too limited variation in specific surface area (74 -97 m 2 /g) , with the TiO 2 tube as the only anatase NM with a substantially larger BET surface area (154 m 2 /g), and thus, more studies are needed to clarify this. However, both a previous (Rahman et al., 2017) and the present study suggest that in general anatase induces less inflammation than rutile NMs when normalized to surface area, however outliers like TiO 2 tube, which do not fit into the crystallinity paradigm, are possible.
In conclusion, anatase TiO 2 NMs with varying physicochemical properties induced pulmonary inflammation and pulmonary acute phase response, but no genotoxicity in mice after i.t. exposure.
All four anatase TiO 2 NMs induced similar inflammatory responses when surface area was used as dose metrics, although inflammatory and acute phase response was greatest and more persistent for the TiO 2 tube. Lowest response was observed for the TiO 2 cube, which might be due to the formation of large aggregates in the lungs. Histopathological changes were observed for both TiO 2 tube and DQ12 and interestingly the effect of the TiO 2 tube was more similar to DQ12 than the other anatase TiO 2 NMs in terms of persistence and the ability to cause PAP, indicating a J o u r n a l P r e -p r o o f qualitative difference related to the tube shape. Comparison with previously published data on rutile TiO 2 NMs indicated that the rutile TiO 2 NMs were more inflammogenic in terms of neutrophil influx than anatase TiO 2 NMs when normalized to total deposited surface area. BET surface area strongly correlated with neutrophil influx for both crystal phases. Multiple regression analyses indicated that BET surface area, crystalline structure and tube shape are potentially important predictors for pulmonary inflammation and acute phase response. Overall, the results suggest that specific surface area, crystal phase and shape of TiO 2 NMs are important predictors for the observed pulmonary effects of TiO 2 NMs.

J o u r n a l P r e -p r o o f
Physiochemical parameters and the influence on neutrophil influx after exposure to anatase and rutile TiO 2 NMs in multiple regression analyses.
BET surface area was transformed using log(BET)/log(1.25), so the estimated effect corresponded to a 25% increase.
Significant p-values (P<0.01) are highlighted in bold.
J o u r n a l P r e -p r o o f Physiochemical parameters and the influence on Saa3 mRNA levels after exposure to anatase and rutile TiO 2 NMs in multiple regression analyses.
BET surface area was transformed using log(BET)/log(1.25), so the estimated effect corresponded to a 25% increase.