Ferritinophagy was involved in long-term SiNPs exposure induced ferroptosis and liver fibrosis

Abstract SiNPs could induce liver fibrosisinvivo, but the mechanism was not completely clear. This study focused on exploring whether long-term SiNPs exposure at human-related exposure dosage could lead to ferritinophagy-mediated ferroptosis and liver fibrosis. In vivo, long-term SiNPs exposure induced liver fibrosis inrats accompanied by ferritinophagy and ferroptosis in hepatocytes. Interestingly, the progression of liver fibrosis was alleviated after exposure cessation and recovery, meanwhile ferritinophagy and ferroptosis were not further activated. In vitro, after long-term SiNPs exposure, the mitochondrial membrane ruptured, lipid peroxidation intensified, the level of redox active iron increased and the repair protein of lipid peroxidation were consumed in L-02 cells, demonstrating ferroptosis occurrence. Notably, NCOA4 knockdown inhibited ferritin degradation, alleviated the increase of intracellular ferrous iron level, reduced lipid peroxidation and the depletion of glutathione peroxidase 4 (GPX4). In conclusion, ferritinophagy mediated by NCOA4 was responsible for long-term SiNPs exposure induced hepatocytes ferroptosis and liver fibrosis, which provided a scientific basis for toxicological assessment of SiNPs and would be benefited for the safety design of SiNPs-based products.


Introduction
Silica nanoparticles (SiNPs) have been widely used in biomedicine, food, cosmetics, chemical industry and textile due to their high stability, biodegradability, adjustability and biocompatibility (Wang et al. 2018). The SiNPs-based products can directly act on human body or enter the air, soil, surface water and sediment during the product life cycle causing humans exposure risks (Wang et al. 2016). The WHO Guidelines on Protecting Workers from Potential Risks of Manufactured Nanomaterials reported SiNPs annual output ranked No.2 among the manufactured nanomaterials in the global market, reaching 1.5 million tons (WHO 2017). There have attracted the concerns of researchers and regulators over the release of SiNPs into the environment and human exposure (Nel et al. 2006;Wang et al. 2016;Klaine et al. 2012). Therefore, SiNPs has been listed as a priority nanomaterial for safety assessment by the Organization for Economic Cooperation and Development (OECD) (OECD 2019).
Inhalation has been proved to be one of the main ways of human exposure to SiNPs, affecting a wide range of people, including natural and occupational exposed population . The particles are inhaled, accumulate in the lungs, enter the circulatory system via the gas-blood exchange of the interalveolar capillaries and further access into the liver via the blood-liver molecular exchange (Wu and Tang 2018;Duan et al. 2018). Yoshida et al. found that after 7 days of nasal exposure to SiNPs at a dose of 500 lg/mouse/day, SiNPs were located not only in the nasal cavity and lungs, but also in the liver, and led to the elevated plasma alanine transferase (ALT) (Yoshida et al. 2013). Li et al. demonstrated that SiNPs exposed via the respiratory tract at a dose of 1.5, 3.0 and 6.0 mg/kgÁbw, respectively, once per week, for 3 months might affect the liver via systemic inflammation . Therefore, respiratory exposure to SiNPs for a short or long term can damage the liver by direct action or systemic inflammation.
Liver fibrosis is a common pathologic process of multiple liver diseases and a reversible injury repair process. The terminal stage of liver fibrosis is cirrhosis, and its fatal complications include liver failure, hepatocellular carcinoma and hepatic encephalopathy with poor prognosis (Jiao, Friedman, and Aloman 2009). Existing studies proved that SiNPs can cause liver fibrosis, however, most of them concerned the short-term intravenous and oral SiNPs exposure Yu et al. 2013). SiNPs exposed through the respiratory tract, one of the important exposure pathways to SiNPs, affects both workers and the general population. Moreover, reallife exposures in humans tend to be long-term exposures. However, there are few studies which explore the effects of long-term SiNPs exposure and recovery after exposure cessation on liver fibrosis progression.
Normal hepatocyte death is the core factor leading to hepatic fibrosis (Yoon, Friedman, and Lee 2016). Ferroptosis is a regulated form of cell death characterized by iron-dependent lipid peroxidation (Dixon et al. 2012). Redox active iron and glutathione peroxidase 4 (GPX4) participates in the initiation and repair of lipid peroxidation mediated by free radicals, respectively . Failure to repair lipid peroxidation further leads to membrane permeability and cell death (Dixon and Stockwell 2019). Nanoparticles have been proved to evoke ferroptosis in cancer cells and hippocampal neurons (Qin et al. 2020;Kim et al. 2016;Zhang et al. 2021). Both basic and clinical studies have illustrated that ferroptosis features prominently in many liver diseases . However, few studies have focused on ferroptosis in hepatocytes caused by nanoparticles.
Ferritinophagy is a selective autophagy of nuclear receptor coactivator 4 (NCOA4) as an autophagic cargo recognition receptor. NCOA4 recognizes ferritin and binds to its subunit ferritin heavy chain (FTH1), transporting ferritin to lysosome for degradation and releasing free ferrous iron (Mancias et al. 2014). The ferrous irons catalyze the Fenton reaction, which generate phospholipid hydroperoxides (PL-OOH), resulting in ferroptosis (Ajoolabady et al. 2021). Ferritinophagy has been observed to participate in ZnONPs-induced ferroptosis in vascular endothelial cells (Qin et al. 2021). The liver plays a central role in iron homeostasis and is the main iron storage site, which makes theliver the main target of iron overload toxicity (Brissot and Lor eal 2016;Guo et al. 2021). However, there are fewstudies on ferrintinophagy induced by nanoparticles in hepatocytes.
Our previous study explored short-term exposure to silica nanoparticles induced ferritinophagy-mediated liver injury through subacute exposure mode . Given that real-life exposures in humans tend to be long-term exposures, this study focused on ferritinophagy-mediatedliver fibrosis induced by silica nanoparticles by the chronic exposure mode at human related dosage. In particular, this study investigated the progression of liver fibrosis after SiNPs exposure cessation and recovery.

SiNPs preparationandcharacterization
The st€ ober method was performed to synthesize SiNPs (St€ ober and Fink 1968). In a nutshell, anhydrous ethanol, deionized water and ethyl orthosilicate (TEOS; Fluorochem, S15425) were premixed at 100:2:5 by volume.Subsequently, we added ammonia water to the mixed solution. Then, the reaction mixture was heated for 12 h at 40 C in a water bath with constant stirring at 200 rmp/min.After washing, the synthesized SiNPs were suspended in deionized water.
TEM (JEOL, JEM-2100, Japan) was adopted to observed the morphology of SiNPs. The size distribution was calculated by Image J software (National Institutes of Health, USA). The hydrodynamic sizes, zeta potential and PDI of SiNPs in different dispersion media were detected by a Zetasizer (Nano-ZS90, Malvern, UK). To reduce aggregation, SiNPs dispersed in normal saline or Dulbecco's Modified Eagle Media (DMEM) were ultrasonicated for 20 min before being detected.

Animal treatments
The male Fisher-344 (F344) rats aged five weeks were purchased from Vital River Experimental Laboratory Animal Technology Co., Ltd. (Beijing, China). All rats were given a normal diet and kept in a barrier system of the experimental animal center of Capital Medical University, with a feeding conditions of temperature 20-24 C, humidity 40-60%, light/dark cycle of 12/12 h. We chose F344 rats as the animal models because studies had shown that F344 rats were more sensitive to liver metabolic responses (Boyce et al. 2020). The experimental protocol was in compliance with the animal research reporting of in vivo experiments 2.0 guidelines (ARRIVE 2.0) (Percie du Sert et al. 2020) and approved by the Animal Protection and Use Committee of Capital Medical University (Ethical Code: AEEI-2019-161).
After 1 week of adaptive feeding, 48 rats were randomly divided into 2 groups. The rats in SiNPs or control group were given intratracheal instillation of 2.63 mg/kg BW SiNPs suspension or the same volume of normal saline, respectively, once a week for 6 months, 24 times in total. The occupational exposure limit (5 mg/m 3 ) of precipitated silica dust (a kind of amorphous silica) in the air of workplace stipulated in the occupational health standard of the People's Republic of China (GBZ 2.1-2019) (National Health Commission 2019) and the physiological parameters of rats were adopted to calculate the exposure dose.The calculation method of exposure dose was detailed in our previously published study . Twenty-four hours after the last treatment, half of each group were sacrificed. The remaining half were observed for another 6 months and sacrificed,in which the SiNPs exposure group was referred to as the exposure cessation and recovery group in this study ( Figure S1A).

Cell treatments
The human normal liver cell line (L-02) was selected as the experimental model invitro, which was obtained fromthe Nanjing Keygen Biotech Cell Library (Nanjing, China). Although the L-02 cell line is contaminated with HeLa cells, it has been reported that the L-02 cells exhibited good liver function in vitro and in an acute liver failure model (Hu et al. 2013). The cells were cultured in DMEM with high glucose containing 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin mixture, which were placed in a cell incubator with 37 C, 5% CO 2 and the saturation humidity. According to the results of Cell Couting Kit-8 (CCK8), the cells were cultured with DMEM or 5 lg/mL SiNPs in DMEM for 24-48 h after adherence.With the above method, the cells were continuously treated for 30 passages. The P0, P10, P20 and P30 passages were selected for detection ( Figure S1B).

Cell viability assessment
The cytotoxicity of SiNPs to L-02 cells was determined by a Cell Counting kit-8 (CCK8; Dojindo, CK04). In short, L-02 cells were seeded in a 96-well plate at a density of 1.2 Â 10 4 cells per well. A total of 6 wells were set up for each group. After 24 h, the cells were treated with 1, 5, 25 and 125 lg/mL SiNPs suspended in DMEM or DMEM for 24 h. Then, the cells were incubated in 10% CCK8 agentia (DMEM) for 1 h. A microplate reader (Synergy HT; Bio Tek, Winooski, VT, USA) was employed to measure the absorbance at 450 nm.

H&E staining
H&E staining was performed to examine the histopathological changes of liver.To be specific,after the rats were sacrificed, the liver tissues were quickly separated, the maximum leaf margin of the liver was cut to an appropriate size, and fixed in 4%paraformaldehyde at 4 C for more than 24 h. The liver tissues were then dehydrated, transparent, waxed, paraffin-embedded and sectioned. After that, the paraffin sections of liver tissues were dewaxed and stained with hematoxylin and eosin (H&E) for histological examination.

Massonstaining
Masson staining was a common method to evaluate the level of tissue fibrosis, which was adopted to measure the degree of liver fibrosis in this study. In brief, the paraffin sections of liver tissues were dewaxed, stained with the potassium dichromate acetic acid solution, Regaud's hematoxylin, differentiated, stained with Ponceau Red acid magenta dye, phosphomolybdic acid and aniline blue for liverfibrosis examination. We randomly selected 3 rats per group for the Masson staining and 10 visual fields in each rat, a total of 30 visual fields per group, for the statistical analysis. The area of liver fibrosis was quantified by Image J software.

Transmission electron microscopy (TEM) assays
The ultrastructure of the liver tissues and cells were detected by a TEM. For tissues, fresh tissues soaked in pre-cooled glutaraldehyde solution at a concentration of 2.5% were rapidly trimmed into 1 mm 3 blocks and fixed at 4 C for 3 h. For cells, the cells were washed, digested and centrifuged (1200 rmp/min, 3 min). The supernatant was discarded, and the cells were promptly fixed with pre-cooling 2.5% glutaraldehyde at 4 C for 3 h. The fully fixed tissues and cells were rinsed with phosphate buffer at a concentration of 0.1 M (PB, Coolaber, SL1326), dehydrated gradient, embeded, polymerized, solidified and then cut into ultra-thin slices about 50-60 nm thick. The ultrathin sectioning were doubledyed with 3% uranium acetate and aluminum citrate and observed by a TEM (JEOL, JEM-2100, Japan).

Hydroxyproline and MDA assay
The hydroxyproline (HYP) content in tissues were measured using hydroxyproline assay kit (Nanjing Jiancheng Bioengineering Institute, A030-2-1) to reflect the degree of liver fibrosis. The malonaldehyde (MDA), a peroxide-lipid degradation product, in tissues and cells, were measured by a MDA assay kit (Nanjing Jiancheng Bioengineering Institute, A003-1) to assess the lipid peroxidation.The experiment operation followed the corresponding manufacturer's protocols. We randomly selected 6 rats or set up 5 cell replicates per group for the statistical analysis.

Lipid ROS assay
A fluorescent molecular probe BODIPY TM 581/591 C11 (Invitrogen, D3861) was adopted to determine the lipid reactive oxygen species (lipid ROS) in L-02 cells. When lipid ROS were produced, the fluorescence of the phenylbutadiene fragment in the fluorescent probe changed from red (reduced state) to green (oxidized state). The green/red mean fluorescence intensity ratio represented the rate of lipid peroxidation.For visualization, SiNPs-exposed L-02 cells for 0, 10, 20 and 30 passages were seeded in confocal dishes. 24 h later, the cells were incubated with 5 lMBODIPY TM 581/591 C11 probe for 30 min, and then stained with hoechst for 10 min in darkness. A laser scanning confocal microscope were emloyed to visualize and photograph the stained cells. Further, to semi-quantify lipid ROS, we selected 20 fields randomly from 4 replicates. The average fluorescence intensity ratios of FITC to TRITC channel were measured using Image J software. The flow cytometry was also performed to quantify the lipid peroxidation ratio (FITC/PE-Texas red). The five cell replicates were set for each group.

Iron assay
The iron assay kit(Abcam, ab83366) was applied for measuring the ferrous and total iron in rat liver. The operation and quantitative analysis followed the manufacturer's protocols. We randomly selected 6 rats per group for the statistical analysis. The FerroOrange (DojinDo, F374) probe was employed for the intracellular ferrous ion staining in L-02 cells. The ferrous iron imaging and semi-quantitative methods were similar to the Lipid ROS. The twenty randomly selected fields from 4 replicates were analyzed.

Immunofluorescence staining
The frozen sections of liver tissues and L-02 cells were subjected to immunofluorescence staining so as to image the NCOA4, FTH1 and a-SMA proteins expression. For the frozen slices, the slices were rewarmed at room temperature for 30 min. For L-02 cells, the confocal dishes were used to inoculate the P0, P10, P20 and P30 passages L-02 cells. Afterwards, the slices and L-02 cellls were fixed with 4% paraformaldehydefor 30 min and properly cleaned with PBS. Subsequently, the slices and L-02 cells were sequentially blocked with 3% peroxidaseblocking reagent for 10 min, permeated with the 0.5% triton-X100 (PBS) for 30 min and then sealed in 10% goat serum for 2 h. Thereafter, the primary antibodies were suitably diluted with the antibody diluents (NCOA4 1:50; FTH1 1:100; a-SMA 1:200) for co-incubation with the frozen sections or L-02 cells at 4 C for 48 h. Following this, the secondary antibodies diluted properly with the antibody diluents (1:200) were coincubated with the frozen sections or L-02 cells for 3 h after rewarming at 37 C for 1 h. Then, 4,6-diamidino-2-phenylindole (DAPI; Life Technologies, D1306) was used to stain the nucleus for 10 min. A laser scanning confocal microscope was adopted to visualize the stained frozen sections or L-02 cells. We randomly selected 3 rats per group for the Immunofluorescence staining and 3 visual fields in each rat, a total of 9 visual fields, for the statistical analysis. For L-02 cells, twenty randomly selected fields from 4 replicates were analyzed.Image J software was adopted to measure the average fluorescence intensity of FITC or TRITC channels.

Western blot analysis
The total protein extraction kit and bicinchoninic acid (BCA) protein assay kit(KeyGEN, KGP2100; DingGuo BioTECH, BCA-02) were used to extract and quantify total protein from liver tissues and L-02 cells, respectively. According to the protein quantification results, the protein extracts of each sample were calibrated with 5 Â SDS protein loading buffer (DingGuo BioTECH, WB-0091) and lysis buffer to a uniform concentration of 4-6 lg/lL. The calibrated protein sample solution were denaturated in a preheated metal bath at 100 C for 10 min to expose the epitope. The equivalent volumes of denatured protein lysates were loaded to 10% or 15% sodium dodecyl sulfate polyacrylamide gel (SDS-PAGE) for protein separation. The isolated proteins were then transferred to a nitrocellulose membrane. The membranes loaded with proteins were blocked with 5% nonfat milk in the Tris-buffered saline (TBS) for 90 min. The blocked membranes were co-incubated with the primary antibodies diluted using antibody dilution buffer (Solatbio, P1080) such as Collagen I (Abcam, 1:1000), Collagen III (Abcam, 1:1000), a-SMA (CST, 1:1000), GPX4 (Abcam, 1:1000), COX-2 (Abcam, 2 lg/mL), LC3B (CST, 1:1000), NCOA4 (Santa, 1:100), FTH1 (CST, 1:1000) and GAPDH (Abcam, 1:1000) at 4 C overnight and then with the secondary HRP antibodies (CST, 1:2000; CST, 1:2000; Santa, 1:1000) for 90 min. The hypersensitive luminescence solution (Epizyme, SQ201) and chemiluminescence imaging system were conducted to develop and scan the protein bands. The gray values of protein bands were calculated by Image J software. We displayed the representative images and randomly selected 3 rats or cell replicates for the statistical analysis.

Lentiviral shrna generation and transduction
A 100 mm petri dish was used to inoculate 6-8 Â 10 6 293T cells, which were placed in a incubator with 5% CO 2 at 37 C for 24 h. 2 h before transfection, the medium was changed with 10 mL of the complete medium (without antibiotics). Next, the mixed reagent containing 5 lg Plvx-shRNA-Zsgreen-T2a-puro recombinant vector, 3.75 lg pspAX2 vector, 1.25 lg pMD2.G vectorand 25 lL Lipofectamine 2000 per 1 mL Opti-MEM was transferred to 293 T cell culture medium for transfection for 8 h.The cell culture medium was changed to the complete culture medium without antibiotics for 16 h and then the complete culture medium with penicillin&streptomycin for 24 h. The cell supernate containing lentivirus particles were collected and stored at 4 C. Add another 10 mL fresh complete medium to each petri dish. The cells were cultured for 24 h. We collected the supernatant again and mixed it with the supernatant collected for the first time. The mixed supernatant was centrifuged at a speed of 1000 rmp/min for 5 min. Then, the supernatant was filtered into a 50 mL round-bottom centrifuge tube with a PES filter of 0.45 lm. The filtrate was centrifuged at 50000 g/min for 2 h at 4 C, the supernatant was discarded and the virus precipitation was dried. The virus precipitates were re-suspended with PBS, and kept at room temperature for 2 h. The concentrated lentivirus were used to infect L-02 cells after the biological titer of lentivirus was determined. The polybrene was adopted to promote viral infection efficiency and puromycin was used to screen virus-infected L-02 cells to establish the stable NCOA4 knockdown L-02 cell lines.

Statistical analysis
All experimental data were statistically analyzed using GraphPad Prism 8.0 software and expressed as mean ± standard deviation (S.D.). The unpaired Student's t test was conducted to compare two groups of data, and one-way analysis of variance (ANOVA) was adopted to compare multiple groups of data. P < 0.05 was considered statistically significant.
In this study, all the results of semi-quantitative experiments such as western blot, immunofluorescence, lipid ROS and FerroOrange were represented by normalization method, that is, the value of the corresponding control group was regarded as 1.0 and the value relative to the control group was represented for the other groups. Other results, for in vivo and RNA interference experiments, were exhibited in the absolute values to provide more precise informations. For in vitro experiments, because there were many groups, in order to more clearly show the difference between the exposure groups and control groups, we exhibited the value of the exposure group relative to the corresponding control group.

Characterization of SiNPs
SiNPs synthesized by the St€ ober method has been comprehensively characterized in our previous study . Briefly, the synthesized SiNPs was approximately spherical and exhibited relatively good monodispersity detected by a TEM. The particle size distribution of SiNPs presented an approximate normal distribution, with an average particle size of 51.74 ± 6.10 nm. The data were from 500 particles. The hydrodynamic sizes, zeta potential and polydispersion index (PDI) of SiNPs dispersed in the deionized water, normal saline and DMEM at different time points (0, 1, 3, 6, 12, 24 h) were measured to characterize the dispersion stability of SiNPs in different mediums. The results showed that the hydrodynamic sizes were 67.49-80.23 nm, the zeta potentials were 16.12-41.91 mV and the PDI were 0.11-0.2. The synthesized SiNPs had good monodispersity and relative stability.
Long-term SiNPs exposure induced liver fibrosis and the progression of liver fibrosis was alleviated after exposure cessation and recovery in vivo The H&E staining was performed to evaluate the histopathological effects of SiNPs on rat livers. The images (Figure 1(A)) exhibited that the liver lobule structure was intact and liver cells were arranged neatly in the control group. A few inflammatory cells were observed near the central venous in the livers of rats exposed to SiNPs for a long term. The granulomas were still observed in the vicinity of the central venous after exposure ceased and resumed. The Masson staining results (Figure 1(B)) demonstrated that, compared with the corresponding control group, liver fibrosis appeared near the portal area and the fibrosis area was significantly increased after long-term SiNPs exposure or exposure cessation and recovery (p < 0.05) (Figure 1(D)). a-SMA was a marker of hepatic stellate cell activation, and the development of fibrosis was closely related to its level. In this study, the expression level of a-SMA in rat livers was observed by the immunofluorescence method. The immunofluorescence images (Figure 1(C,E)) revealed that, compared with the corresponding control group, a-SMA in the rat liver tissues were obviously upregulated after long-term SiNPs exposure or exposure cessation and recovery (p < 0.05).
After qualitative observation of liver fibrosis in rats, we also quantitatively analyzed the content of hydroxyproline (HYP) in liver tissues, which indicated the degree of liver fibrosis. The quantitative analysis suggested that (Figure 1(F)), compared with the control group, the concentration of HYP in the rat liver tissues of long-term SiNPs exposure group was notably increased (p < 0.05). However, there was no difference in HYP content between the group of exposure stopped/resumed and its control (p > 0.05).
The development of fibrosis was closely related to the expression level of a-SMA. Liver Collagen deposition was an important indicator of liver fibrosis, especially collagen I and collagen III. Therefore, the levels of liver fibrosis key proteins were detected by western blot. The results illustrated that a-SMA in the liver tissues of rats after longterm exposure to SiNPs or exposure cessation and recovery were remarkably upregulated (p < 0.05), compared with the corresponding control group, which was consistent with the immunofluorescence results (Figure 1(G,H)). Compared with the corresponding control group, the levels of collagen I and collagen III in the liver tissues of rats exposed to SiNPs for a long term were greatly upregulated (p < 0.05), while the changes of collagen I and collagen III expressions were not statistically significant after exposure cessation and recovery (p > 0.05) (Figure 1(G,I,J)).
The above evidence indicated that long-term exposure to SiNPs for 6 months at human exposure relevant dosage could induce liver fibrosis in rats, and the liver fibrosis progression was alleviated after exposure cessation and recovery for 6 months.
Long-term SiNPs exposure triggered hepatocellularferroptosis and ferroptosis was not further activated after exposure cessation and recovery in vivo In order to test whether SiNPs induced ferroptosis in vivo after long-term SiNPs exposure or exposure cessation and recovery, TEM was performed to observe the ultrastructure of liver tissues. In the control group, the morphology of mitochondria was observed to be normal and the bilayer membrane was examined to be intact (Figure 2(A)). After longterm SiNPs exposure or exposure cessation and recovery, mitochondrial membrane rupture was detected, which was the morphological feature of ferroptosis. MDA was one of the final products of lipid peroxidation, therefore we mesasured the content of MDA in rat liver tissues. The results demonstrated that MDA was significantly increased after long-term SiNPs exposure (p < 0.05), but there was no obvious difference in MDA between the groups of recovery for 6 months and its control (p > 0.05) (Figure 2(B)). Since it has been proved that Fe 2þ is the catalyst of Fenton reaction, the total iron The change of MDA content in rat liver tissues after long-term SiNPs exposure or exposure cessation and recovery (n ¼ 6). (C) The change of Fe 2þ and total iron content in rat liver tissues after long-term SiNPs exposure or exposure cessation and recovery (n ¼ 6). (E) The representative western blot pictures of GPX4 and COX-2 in rat liver tissues. The western blot pictures of (F) GPX4 and (G) COX-2 were quantitatively analyzed by Image J software (n ¼ 3). All data were expressed as mean ± S.D. Ã p < 0.05 compared with the control groups. content reflects the potential to produce Fe 2þ . To investigate whether ferrous iron stimulated Fenton reaction and led to lipid peroxidation, an iron assay kit was conducted to determine the contents of ferrous and total iron in rat liver tissues. The results showed that Fe 2þ cumulated significantly in the liver tissues of rats after long-term SiNPs exposure (p < 0.05) (Figure 2(C)). However, there was no difference in Fe 2þ between the group of exposure cessation/recovery and its control (p > 0.05). The total iron in the long-term SiNPs exposure group or exposure cessation and recovery did not vary, compared to their corresponding control (p > 0.05) (Figure 2(D)).
Glutathione peroxidase 4 (GPX4) is an enzyme that effectively reduced the toxic lipid peroxides to nontoxic phosphatidyl alcohols using two molecules of the reduced glutathione with sulfhydryl group (GSH). The Cyclooxygenase-2 (COX-2) was a marker of lipid peroxidation. Western blot was used to detect the GPX4 and COX-2 expression levels in liver tissues of rats after long-term SiNPs exposure or exposure cessation and recovery, to characterize SiNPs-induced ferroptosis in hepatocytes. The data revealed that GPX4 and COX-2 were clearly down-regulated and up-regulated, respectively, in the long-term SiNPs exposure group compared with the corresponding control group (p < 0.05). However, no significant difference in the expression of GPX4 and COX-2 was detected in the liver tissues of rats after exposure cessation and recovery (p > 0.05) (Figure 2(E,F,G)).
The above proofs suggested that long-term SiNPs exposure triggered ferroptosis in rat hepatocytes, and ferroptosis was not further activated after exposure cessation and recovery, which might be one of the reasons why the fibrosis progress was alleviated after exposure cessation and recovery.

Long-term SiNPs exposure activated hepatocellularferritinophagy and ferritinophagy was not further activated after exposure cessation and recovery in vivo
For the sake of further exploring the reasons for increased ferrous iron in rat liver, immunofluorescence and western blot were employed to examine the key proteins of ferritinophagy in rat liver tissues. Immunofluorescence and western blot data were consistent, demonstrating that NCOA4 was up-regulated and the FTH1 down-regulated in the liver tissues of rats after long-term SiNPs exposure (p < 0.05), and no statistical significance were found in the changes of the NCOA4 and FTH1 expression after exposure cessation and recovery (p > 0.05) (Figure 3(A-F)).
The above analysis indicated that long-term SiNPs exposure activated ferritinophagy in rat hepatocytes, which was not further triggered after the cessation of exposure and recovery, which might be one of the causes explaining why ferroptosis in rat hepatocytes was not further activated after the cessation of exposure and recovery.

Long-term SiNPs exposure evoked hepatocellular ferroptosis in vitro
We adopted L-02 cells for in vitro experiments. The activity of L-02 cells treated with 0, 1, 5, 25 or 125 lg/mL SiNPs was examined by a CCK8 kit. The toxicity of SiNPs to L-02 cells was dose-dependent. L-02 cells dealed with 5 lg/mL SiNPs did not induce the cell viability decrease (Figure 4(A)). 5 lg/mL was therefore used as the long-term SiNPs exposure dosage. Furthermore, TEM was employed to observe the ultrastructural changes of P0, P10, P20 and P30 passages of L-02 cells exposed to SiNPs. The images illustrated that, in the control group, the mitochondrial membrane and crest of L-02 cells were unbroken. In L-02 cells of P10, P20 and P30 passages exposed to SiNPs, mitochondrial vacuolation and mitochondrial membrane rupture occurred (Figure 4(B)).
Next, we detected the content of MDA, one of the lipid peroxidation products, in L-02 cells exposed to SiNPs for a long term. The results showed that, compared with the control group, MDA increased in the SiNPs exposed for 30 passages group (p < 0.05) (Figure 4(C)). The qualitative (or semi-quantitative) and quantitative analysis of lipid ROS were performed using a fluorescent probe combined with the confocal microscopy and flow cytometry, respectively. The results showed that long-term SiNPs exposure induced the accumulation of lipid ROS in L-02 cells (p < 0.05) (Figure 4(D-F)). To explore the possible causes of elevated lipid ROS, FerroOrange fluorescence probe combined with confocal microscopy was used to detect the content of ferrous ions in long-term SiNPs exposed L-02 cells. The results showed that, compared with the corresponding control group, the content of ferrous ions increased in long-term SiNPs exposed group (p < 0.05) (Figure 4(G,H)). Furthermore, we examined the ferroptosis key proteins in long-term SiNPs exposed L-02 cells. Compared with the control group, GPX4 down-regulated and COX-2 upregulated in long-term SiNPs exposed L-02 cells (p < 0.05) (Figure 4(I-K)). The above evidence suggested that long-term SiNPs exposure triggered ferroptosis in vitro, which might be related to the increased ferrous iron levels in L-02 cells.

Long-term SiNPs exposure induced hepatocellular ferritinophay in vitro
Immunofluorescence detection of the key ferritinophagy proteins revealed that long-term SiNPs exposure up-regulated NCOA4, up-regulated and then down-regulated FTH1 in L-02 cells ( Figure 5(A-C)). Western blot results showed that SiNPs treatment temporarily increased FTH1, while after SiNPs treatment for P30 passages, FTH1 decreased when  NCOA4 up-regulated (p < 0.05) (Figure 5(D-G)). In addition, long-term SiNPs exposure also induced an increase in LC3B-II/LC3B-I ratio in L-02 cells ( Figure 5(D,H)).
Ferritinophagy mediated by NCOA4 was responsable for long-term SiNPs exposure activated ferroptosis NCOA4 is an autophagy cargo recognition receptor, which can identify ferritin and transport it to lysosomes for degradation. Thus, ferrous iron is released from ferritin, which mediates the Fenton reaction and ferroptosis in turn. Therefore, we used lentiviral shRNA to stably knock down NCOA4 in L-02 cells to assess the effect of NCOA4-mediated ferritinophagy on long-term SiNPs exposure-induced ferroptosis. Western blot method was adopted to detect the effect of NCOA4 knockdown on the expression level of ferritinophagy key proteins. NCOA4 knockdown inhibited the expression of NCOA4 protein, up-regulated the LC3B-II/LC3B-I and FTH1 expression (p < 0.05) (Figure 6(A-D)). In addition, the influence of NCOA4 knockdown on SiNPs-induced ferritinophagy key proteins expression was examined by immunofluorescence assay. The results showed that NCOA4 knockdown inhibited NCOA4 protein expression and upregulated FTH1 (p < 0.05) (Figure 6(E-G)). These data demonstrated that the degradation of ferritin in autophagosomes was inhibited by knockdownof NCOA4.

Discussion
SiNPs have been reported to induce or aggravate liver fibrosis, but most of them focused on the short-term digestive tract or intravenous exposure (Yu et al. 2017;Li et al. 2018). However, real-life exposures in humans tend to be long-term exposures and SiNPs exposurevia respiratory tract affected both workers and the general population. Therefore, this study focused on the effect of longterm SiNPs exposure through respiratory tract on liver fibrosis. Hepatocyte death was the core factor of liver fibrosis (Yoon, Friedman, and Lee 2016). Ferroptosis, which was a newly programmed cell death characterized by iron accumulation and lipid peroxidation (Stockwell et al. 2017), had been proved to participate in a variety of liver diseases in many published studies (Lyu et al. 2021;Zhu et al. 2021;Wang et al. 2021). Liver, as the main iron storage organ, was reported to play a key role in iron homeostasis and be more susceptible to iron overload toxicity (Guo et al. 2021;Brissot and Lor eal 2016). Existing study demonstrated that NCOA4mediated ferritinophagy released free ferrous iron by degrading ferritin, causing iron overload and triggering ferroptosis . In this regard, this study confirmed for the first time that long-term SiNPs exposure could trigger ferritinophay-cascaded ferroptosis in hepatocytes and then induce liver fibrosis. Moreover, ferritinophay-cascaded ferroptosis was not further activated after exposure cessation and recovery, and the progression of liver fibrosis was alleviated.
The results of histopathology, ultrastructure, immunofluorescence and western blot showed that the fibrosis area, hydroxyproline content, a-SMA, Collagen I and Collagen III in liver tissues increased, suggesting that long-term SiNPs exposure caused liver fibrosis in rats (Figure 1). It has been reported that nanoparticles or fine particulate matters exposure via respiratory tract triggered liver fibrosis. The male ApoE -/mice, a mouse model of metabolic syndrome, received intratracheal instillation of 6.0 mg/kgÁbw SiNPs suspension once a week for 12 times, resulting in liver fibrosis ). Wu et al demonstrated that intranasal instillation of 1 mg/kg bw graphene quantum dots for 28 days induced liver fibrosis in mice ). Zheng et al illustrated that the whole-body exposure to environmentally relevant PM 2.5 for 6 h a day, 5 days a week for 10 weeks led to liver fibrosis in normal or high-fat diet C57BL/6 male mice (Zheng et al. 2015). Our study focused on the liver fibrosis caused by long-term exposure of nanoparticles at the human exposure related dosage, which had more practical significance.
In this study, we observed that liver fibrosis area and a-SMA expression increased compared with the control group after exposure cessation and recovery. However, the content of hydroxyproline, Collagen I and Collagen III did not change significantly. These evidences demonstrated that the progression of liver fibrosis was alleviated. Liver fibrosis is different from other fibrosis in that vitreous scars are rare, which is a reversible process. It has been proved in experimental fibrotic liver and human sclerotic liver that unless progressive liver fibrosis leads to cirrhosis, removing the substances causing fibrosis will contribute to the regression of liver fibrosis (Kisseleva and Brenner 2021). Our results might be due to the fact that the liver fibrosis caused by SiNPs exposed at human relevant dosage was still in its early stage. Therefore, after exposure cessation and recovery, ECM secretion and degradation were gradually restoring balance based on the body's clearance and immune system, then the progress of liver fibrosis was alleviated.
Our results suggested that long-term SiNPs exposure through the respiratory tract could cause liver fibrosis even at human-relevant dosage. The progression of liver fibrosis could be alleviated after exposure cessation and recovery. Therefore, timely improvement of SiNPs safety design and rational disposal of SiNPs-based products might reduce the susceptibility to SiNPs-induced liver disease. It has been reported that inhaled nanoparticles are deposited in the alveolar area, and then gradually move toward the mucociliary escalator of the upper respiratory tract. Once these particles reach the bronchus, they will be transported to the pharynx and swallowed, resulting in gastrointestinal exposure. Nevertheless, the absorption of nanoparticles in the gastrointestinal tract after the respiratory tract exposure has been proved to be rare or negligible. Then, the nanoparticles present in the gastrointestinal tract will soon be cleared through the feces (Modrzynska et al. 2018). Therefore, we have reason to believe that the effect on the liver is mainly caused by SiNPs exposed through respiratory tract.
Further, we observed that long-term SiNPs exposure induced mitochondrial membrane rupture, increased MDA (one of the final products of lipid peroxidation) and COX-2 (a marker of lipid peroxidation), ferrous iron accumulation and down-regulated GPX4 (a repair substance of lipid peroxidation) in rat liver tissues, suggesting long-term SiNPs exposure might lead to ferroptosis in rat liver cells. Moreover,after the exposure cessation and recovery, MDA, COX-2, Fe 2þ and GPX4 did not change compared with the control group, indicating that the hepatocytes ferroptosis in rat liver cells was not further activated (Figure 2). Consistent with the in vivo results, long-term SiNPs exposure (the cells were continuously treated by SiNPs for 30 passage) induced mitochondrial membrane rupture, increased MDA, COX-2, lipid ROS and intracellular ferrous ion in L-02 cells, which meant that longterm SiNPs exposure evoked ferroptosis in vitro (Figure 4). Accumulating evidences suggested that nanoparticles could cause organ damage by inducing ferroptosis. Intratracheal instillation of NiONPs caused acute lung injury in mice by inducing ferroptosis in lung (Liu, Cheng, et al. 2022). Lung exposure to ZnONPs could evoke brain lesions by triggering neuronal ferroptosis in mice (Qin et al. 2020). Moreover, some studies have also revealed the ability of nanoparticles to trigger cell ferroptosis in vitro. ZnONPs induced ferroptosis in human umbilical vein endothelial and neurons cells (HUVECs and EA.hy926) (Qin et al. 2021;Qin et al. 2020). Poly(ethylene glycol)-coated silica nanoparticles evoked ferroptosis in nutrient-deficient cancer cells (Kim et al. 2016). NiONPs triggered ferroptosis in airway epithelial cells (Liu, Cheng, et al. 2022). Graphene quantum dots induced microglia and macrophages ferroptosis (Shao et al. 2021;Wu et al. 2020). Cobalt nanoparticles triggered ferroptosis in neurons (Gupta et al. 2020). CdTe QDs induced macrophages (RAW264.7) ferroptosis (Liu, Liang et al. 2022). Our study extended the effect organ of ferroptosis induced by nanoparticles to the liver and for the first time explored the ferroptosis activation after long-term nanoparticles exposure and the exposure cessation and recovery.
Next, based on the accumulation of intracellular ferrous iron detected in vivo and in vitro, we further explored the possible cause of this phenomenon. In vivo and in vitro results showed that long-term SiNPs exposure induced ferritinophagy in hepatocytes, characterized by NCOA4 up-regulation and FTH1 down-regulation. In vitro results also illustrated that long-term SiNPs exposure induced the up-regulation of LC3B-II/LC3B-I ratio in L-02 cells, which further confirmed that SiNPs activated ferritinophagy. Interestingly, in vivo, ferritinophagy was not further activated after the exposure cessation and recovery, which might be one of the reasons why ferroptosis was not further activated and the progression of liver fibrosis was alleviated (Figure 3 and 5). Therefore, ferritinophagy might be responsible for intracellular accumulation of ferrous iron after long-term SiNPs exposure. Activation of ferritinophagy by nanoparticles was rarely studied. CdTe QDs mediated local and systemic inflammation in C57BL/6 mice through macrophage ferritinophagy (Liu, Liang et al. 2022). ZnONPs mediated vascular inflammation in mice by inducing ferritinophagy in vascular endothelial cells (Qin et al. 2021). PM 2.5 mediated emphysema and airway inflammation in mice by inducing ferritinophagy in alveolar epithelial cells . Based on the fact that the liver was the primary organ for iron storage and more susceptible to iron overload toxicity, our study focused on the liver, where SiNPs activated ferritinophagy after chronic exposure, and further assessed whether ferritinophagy was continuously activated in the liver after exposure cessation and recovery.
Our previous study showed that the total iron in liver tissues of ratsincreased after short-term SiNPs exposure , which was probably due to the dehydrogenation of the silanol group on SiNPs surface or their unique internal structure leading to the adsorption or incorporation of iron (Kim et al. 2016). In this study, the total iron in rat liver tissues did not change after long-term SiNPs exposure, which may be attributed to the balance between iron absorption into the liver and metabolism at low dose. In addition,a study proved that coexposure to carbon black particles and nickel by intratracheal instillation induced autophagy and lysosome dysfunction in lung tissues of C57BL6 mice, but no persistent autophagy dysfunction was observed 30 days after exposure cessation.This was consistent with our results and may be due to hepatocytes exocytosis and macrophage clearance at low dose exposure (He et al. 2020).
To test the role of ferritinophagy mediated by NCOA4 in SiNPs-induced ferroptosis, we used lentvirus shRNA to knock down NCOA4 mRNA and observed its effect on ferritinophagy and ferroptosis. The data exhibited that NCOA4 knockdown relieved SiNPs-induced ferritin degradation, accumulated intracellular free iron, increased lipid peroxidation and impaired lipid peroxidation repair ability, demonstrating that ferritinophagy mediated by NCOA4 was responsible for long-term SiNPs exposure-triggered ferroptosis ( Figure 6). NCOA4mediated ferritinophagy was first reported in 2014 (Goodall and Thorburn 2014). A study in 2019 showed that using siRNA to knock down NCOA4 mRNA in lung epithelial cells after cigarette smoke exposure obviously decreased lipid peroxidation and recovered cell viability (Yoshida et al. 2019). A study in 2021 reported that knockdown of NCOA4 mRNA in vascular endothelial cells using siRNA could reduce lipid peroxidation and cell death induced by zinc oxide nanoparticles (Qin et al. 2021). These results were consistent with our results. Our study extend NCOA4-mediated ferritinophagy to nanoparticles or fine particulate matters induced hepatocyte dysfunction.

Conclusion
Our study demonstrated that long-term SiNPs exposure activated ferritinophagy in hepatocytes characterized by NCOA4-mediated ferritin degradation and release of redox active iron, which further led to initiation of Fenton reaction, production of lipid ROS and repair of lipid peroxidation. Depletion Figure 7. Schematic diagram of the proposed molecular mechanism by which long-term SiNPs exposure induced hepatocytes ferroptosis and liver fibrosis via ferritinophagy. Long-term SiNPs exposure activated NCOA4-mediated ferritinophagy in hepatocytes, which allowed NCOA4 to bind to ferritin and transport it to lysosome for degradation, releasing ferrous iron. The released ferrous iron mediated Fenton reaction to produce lipid ROS, which resulted in the accumulation of lipid peroxidation marker protein COX-2 and depletion of lipid peroxidation repair protein GPX4. Failure of lipid peroxidation repair tiggered mitochondrial membrane/cristae fracture, which induced hepatocytes ferroptosis further. Hepatic parenchymal cell death activated hepatic stellate cells to express a-SMA, resulting in collagen fiber accumulation and liver fibrosis. of lipid peroxidation repair enzyme caused mitochondrial membrane rupture and ultimately ferroptosis in hepatocytes. Hepatic parenchymal cell death activated hepatic stellate cells to express a-SMA, resulting in collagen fiber accumulation and liver fibrosis. The progression of liver fibrosis was alleviated after the exposure cessation and recovery, which might be related to the fact that ferritinophagy and ferroptosis were not further activated after the exposure cessation and recovery (Figure 7). Our study provided a scientific basis for liver toxicological assessment of long-term SiNPs exposure, which helped to promote the safety design of SiNPs.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by the National Natural Science Foundation of China [81930091,81973077].

Data availability statement
The data that support the findings of this study are openly available in [