Nanomaterials induce different levels of oxidative stress, depending on the used model system: Comparison of in vitro and in vivo effects

Background The immense variety and constant development of nanomaterials (NMs) raise the demand for a facilitated risk assessment, for which knowledge on NMs mode of actions (MoAs) is required. For this purpose, a comprehensive data basis is of paramountcy that can be obtained using omics. Furthermore, the establishment of suitable in vitro test systems is indispensable to follow the 3R concept and to master the high number of NMs. In the present study, we aimed at comparing NM effects in vitro and in vivo using a multi-omics approach. We applied an integrated data evaluation strategy based on proteomics and metabolomics to four silica NMs and one titanium dioxide-based NM. For in vitro investigations, alveolar epithelial cells and alveolar macrophages were treated with different doses of NMs, and the results were compared to effects on rat lungs after short-term inhalations and instillations at varying doses with and without a recovery period. Results Since the production of reactive oxygen species (ROS) is described to be a critical biological effect of NMs, and enrichment analyses conrmed oxidative stress as a signicant effect upon NM treatment in vitro in the present study, we focused on different levels of oxidative stress. Thus, we found opposite changes for proteins and metabolites that are related to the production of reduced glutathione in alveolar epithelial cells and alveolar macrophages, illustrating that NMs MoAs depend on the used model system. Interestingly, in vivo, pathways related to inammation were affected to a greater extent than oxidative stress responses. Hence, the assignment of the observed effects to the levels of oxidative stress was different in vitro and in vivo as well. However, the overall classication of “active” and “passive” NMs was consistent in vitro and in vivo. Conclusions The consistent classication indicates both tested cell lines to be suitable for NM toxicity assessment even though the induced levels of oxidative stress strongly depend on the used model systems. Thus, the here presented results highlight that model systems need to be carefully revised to decipher the extent to which they can replace in vivo testing. increased GSH production and thus only tier 1 and macrophages displaying a decreased GSH production and signicant changes over all three tiers. Hence, it was not surprising that the assignment was not as clear for the in vivo results of rat lungs that consist not only of type II cells and macrophages but also of type I cells, endothelial cells and interstitial cells. However, the in vitro and in vivo results were consistent regarding the classication of the tested NMs, thus indicating that both tested cell lines are suitable to assess NM toxicity. Since the macrophages exhibited more signicant changes over all tiers of oxidative stress, they might be a more sensitive model system. Taken together, the presented results illustrate that the MoAs of NMs vary in different model systems and given that, care needs to be taken when using biomarkers for NM risk assessment. type II (STIS), and

to oxidized glutathione (GSSG). GSH is an anti-oxidant, which is oxidized to GSSG during its neutralization of ROS. Under normal conditions, the GSH/GSSG ratio is high, but it decreases with over-proportional increases of ROS, a process whose biological consequences may be described by three tiers: In tier 1 nuclear factor erythroid 2-related factor 2 (Nrf2) target genes such as anti-oxidants are formed, followed by in ammation reactions mediated by nuclear factor kappa-light-chain-enhancer of activated B cells (Nfκb) and activator protein 1 (Ap-1) in tier 2, and cytotoxicity in tier 3 [29,32,33]. Since the level of oxidative stress highly in uences the formation of adverse effects upon NM treatment [36,37], we aimed at a detailed analysis of the levels of oxidative stress by integrative omics analysis. Furthermore, we compared in vitro and in vivo results to clarify to which extent animal tests could be replaced by cellular experiments in future risk assessment.

Results
To get insights into NMs MoAs, the effects of four silica NMs (SiO2_15_Unmod, SiO2_15_Amino, SiO2_40, SiO2_7) and TiO2_NM105 were investigated in vitro and in vivo. The physico-chemical properties of these NMs have been described before [11,12,25,38,39,40], and a summary of their key physico-chemical properties is presented in Table 1.

In vitro investigations
First, the effects of these ve NMs were investigated in vitro in alveolar type II cells and alveolar macrophages at three different doses after an exposure time of 24 h. Changes of protein (Additional le 2: Figure S1 - Figure S4) and metabolite (Additional le 2: Figure S5 - Figure S8) abundances were considered relative to untreated controls and the summary of the percentages of signi cantly (p.adj ≤ 0.05) altered proteins and metabolites ( Figure 1a) shows that signi cant changes appeared for metabolites only at the highest investigated dose in type II cells. Interestingly, the pyrogenic SiO2_40 and SiO2_7 were the only NMs, which signi cantly affected metabolites also at a dose of 10 µg/cm 2 in type II cells and macrophages, where none of the other tested NMs resulted in signi cant alterations. TiO2_NM105 showed signi cantly changed metabolites at a dose of 10 µg/cm 2 in type II cells but not in macrophages. In contrast, almost no signi cantly altered proteins were observable in type II cells, while dose-dependent increases of the portion of signi cantly altered proteins were noticeable for SiO2_40 and SiO2_7 after treatment of macrophages. Thus, in accordance to our previous studies, SiO2_40, SiO2_7, and TiO2_NM105 were considered "active" NMs [11,12], while SiO2_15_Unmod and SiO2_15_Amino were classi ed as "passive".
Additionally, exposure times of 6 h and 48 h were investigated for type II cells. After 6 h of NM treatment, a maximum of 72 signi cantly altered proteins was observable for 10 µg/cm 2 SiO2_7, compared to 220 and 462 proteins under the same conditions but after 24 h and 48 h incubation, respectively (Table E3). Thus, the 6 h time point was excluded from further analyses.
An enrichment analysis was applied to the signi cantly (p.adj ≤ 0.05) altered proteins and metabolites, which showed mainly enrichment of pathways that are related to oxidative stress, e.g. mitochondrial dysfunction, Nrf2-mediated oxidative stress response and oxidative phosphorylation (Additional le 2: Figure  S9). Thus, we focused on the investigation of the levels of oxidative stress that were reached under the tested conditions. For this purpose, we used only results after 24 h treatment because only minor time-dependent changes were observable between 24 h and 48 h (Additional le 2: Figure S9 and Figure S10).

GSH/GSSG signaling
Since the level of GSH is highly connected to the formation of oxidative stress, proteins and metabolites that are linked to GSH/GSSG signaling [41,42,43,44,45,46,47] were examined rst ( Figure 1b). In type II cells (Figure 1b, left), SiO2_40, SiO2_7 and TiO2_NM105 were the only NMs with signi cant changes after treatment with 10 µg/cm 2 , which is the dose we used in previous screening experiments with twelve NMs in type II cells as well [11]. Interestingly, SiO2_15_Unmod and SiO2_15_Amino showed the same trends, with SiO2_15_Unmod leading to higher elevations (expressed as fold changes, FCs) than SiO2_15_Amino. Overall, the NM treatment of type II cells led to increased abundances of proteins and metabolites that are part of the GSH/GSSG signaling, thus indicating an increased production of GSH, especially in case of SiO2_40, SiO2_7, and TiO2_NM105, con rming that not only SiO2_40 and SiO2_7 are "active" NMs but also TiO2_NM105, which is consistent with our previous study [11].
Interestingly, the opposite effects were observed for macrophages (Figure 1b, right) compared to type II cells (Figure 1b, left). Most signi cant changes were obtained for SiO2_7 and SiO2_40. SiO2_15_Unmod showed the same trends, even though with less signi cance. The decreased abundances of proteins and metabolites suggest a decreased GSH production for these three NMs in macrophages. This in turn may have led to an insu cient neutralization of appearing ROS, thus inducing further oxidative stress reactions. The opposite changes in type II cells and macrophages occurred especially in the treatment with 10 µg/cm 2 SiO2_7 (Figure 1c).

Comparison of oxidative stress levels
To clarify whether the observed effects on the GSH/GSSG signaling for the two cell lines led to a different assignment to the three tiers of oxidative stress, the changes of proteins and metabolites that are connected to these tiers were investigated. Interestingly, all the proteins that are shown in Figure 1b and Figure   1c are regulated by the transcription factor Nrf2 [48,49,50,51], a hallmark of tier 1, suggesting that the NMs led to tier 1 in the used type II cells at the highest applied dose. In macrophages, decreased abundances of analytes that are connected to GSH/GSSG signaling were observed for SiO2_7, SiO2_40, and SiO2_15_Unmod, suggesting that these three NMs led to a decreased GSH/GSSG ratio, while SiO2_15_Amino and TiO2_NM105 did not.
Besides the proteins that are part of the GSH/GSSG signaling, Gstm1, Nqo1, Hmox1, Txn, Txnrd1, Cat, Sod1, Sod2 and Lamp2 have been described to be Nrf2 targets [48,49,52] and thus appear relevant for tier 1 (Figure 2a). Tier 2 is induced by activation of Nfκb and Ap-1 target genes. Importantly, several of the already described Nrf2 target genes have also a binding position for Nfκb or Ap-1. Examples are Gclc, Idh, Pgd, Phgdh, Hmox1, Nqo1, Cat, Sod2, and Lamp2 [53,54]. Furthermore, Icam1 [55,56], Il18 [57,58], B2m [59], and Tnfaip8 [60] have been described to be target genes of Nfκb or Ap-1 and related to either in ammation or apoptosis. Ccr1 has been described to be expressed mitogen-activated protein kinase (Mapk)-dependently [61]. Bax has been described to be an Nfκb target protein [62,63], and it is a pro-apoptotic protein since it is involved in the formation of mitochondrial pores that allow for the release of pro-apoptotic molecules from the mitochondria into the cytosol [64]. Another protein that is involved in the formation of mitochondrial pores and thus in the induction of apoptosis is Vdac1 [65], which consequently should be assigned to tier 3. Furthermore, it has been shown that an increased citric acid cycle (TCA) leads to the increased formation of ROS, followed by apoptosis [66]. Besides the already mentioned proteins, Idh, Glud1, and Fh are involved in the TCA [67,68]. In addition, sphingomyelins (SMs), which belong to the class of sphingolipids and are mainly found in plasma membranes and lipoproteins, are relevant for the formation of ROS. SMs have been shown to be hydrolyzed in response to oxidative stress, thus resulting in the formation of ceramides, which act as second messengers that are involved in the induction of apoptosis. Importantly, the concrete mechanisms are not fully understood, yet [69,70]. A summary of all mentioned analytes and their assignment to the different tiers of oxidative stress can be found in Tables S1 and S2.
In type II cells, almost no signi cant changes were noticed for all these candidates up to the highest tested dose (Figure 2a), while treatment of macrophages with 10 µg/cm 2 SiO2_40 and SiO2_7 led to signi cant alterations. Importantly, SiO2_15_Unmod again showed the same trends as SiO2_40 and SiO2_7 but effects were less pronounced. Furthermore, major differences were recognizable between treated type II cells and macrophages once again, especially after treatment with 10 µg/cm 2 SiO2_7 (Figure 2b). For type II cells signi cant changes appeared only in proteins that are Nrf2 target genes, thus con rming that in type II cells only tier 1 of oxidative stress was affected under the tested conditions. The signi cantly affected proteins were Sod1, Txn, Gstm1, and Lamp2. In contrast, the used macrophages led to signi cant changes over all tiers after treatment with 10 µg/cm 2 SiO2_7. Interestingly, there are several cases in which opposite changes in protein abundance were visible in type II cells and macrophages. Examples are Sod1 and Sod2 with opposite changes among themselves and additionally, opposite changes in the two investigated cell lines. The fact that Sod1 and Sod2 resulted in different directions of alteration can be explained by their localization within the cell. While Sod1 can be found in the cytoplasm, Sod2 is responsible for neutralizing ROS in the mitochondria. Furthermore, both proteins have not only an Nrf2 binding site but also an Nfκb binding site in their promoter region. While the Nfκb binding site in Sod1 has been described to be rather insensitive, the binding site in Sod2 seems to be highly sensitive [71], thus suggesting the Sod1 expression to be predominantly induced by Nrf2, while the Sod2 expression may be induced by Nfκb. This is an additional evidence that type II cells led only to tier 1, whereas in macrophages all three tiers were initiated.

Comparison of in vivo and in vitro results
Next, we aimed to investigate the in vivo effects of SiO2_7 and TiO2_NM105 using STIS and of SiO2_15_Amino, SiO2_15_Unmod, SiO2_7, and SiO2_40 applying instillations. For both methods, changes in lung proteome and metabolome were investigated at several doses. Exposure groups (E) were sacri ced directly after treatment for 5 d or 3 d in STIS and instillations, respectively. In contrast, recovery groups (R) had a recovery time of 21 d after the treatment period. The percentages of signi cantly altered proteins (Additional le 2: Figure S11 - Figure S14) and metabolites (Additional le 2: Figure S15 - Figure   S18) indicated dose-dependent increases of signi cantly altered analytes mainly for STIS exposure groups (Figure 3a). The assignment to the three tiers of oxidative stress in vivo was not as clear ( Figure 3b) as for the in vitro results, even though signi cant changes were observable for all three tiers, especially in case of SiO2_7, SiO2_40, and TiO2_NM105, indicating those to be "active" in vivo. Furthermore, analytes with dose-dependent changes were detectable, e.g. Ornithine, Sod1, Sod2, and Txn.
The different results of the assignment to the three tiers in vitro and in vivo may be explained by the cellular composition of rat lungs (Figure 4a), with only 14.2% alveolar type II epithelial cells and 3% alveolar type II epithelial cells [72], and the circumstance that the two cell types investigated within the present study showed opposite effects. The comparison of analytes that are connected to the three tiers of oxidative stress after treatment with SiO2_7 in vitro and in vivo ( Figure 4b) exempli es this. For GSH/GSSG signaling, opposite effects were obtained for type II cells and macrophages, but no clear trends were observed in STIS and instillations. Analytes connected to the three tiers revealed no clear trends either, but the distributions (Figure 4b) indicated higher changes in macrophages and STIS, suggesting these model systems to be more sensitive.
Since the GSH/GSSG signaling did not re ect many of the signi cant changes in vivo (Figure 3a), Ingenuity Pathway Analysis (IPA, Qiagen) was used to identify NM effects in vivo. This analysis con rmed that oxidative stress pathways like glutathione redox reactions I, Nrf2-mediated oxidative stress response, and mitochondrial dysfunction were not among the most signi cantly enriched pathways (Additional le 2: Figure S19). Furthermore, chemokine signaling was found to be enriched mainly in case of the instillation studies with SiO2_7 and SiO2_40 (Figure 4c). Examples for pathways with higher enrichment were endocytosis signaling, phagosome maturation, leukocyte extravasion signaling, remodeling of epithelial adherens junctions, integrin signaling, and actin cytoskeleton signaling. Apoptosis signaling was again mainly enriched for SiO2_7 and SiO2_40 in the instillation studies, thus indicating that these two NMs had the strongest effects. Interestingly, most pathways showed more signi cant enrichment in the recovery groups than in the exposure groups. The complete results from the enrichment analysis can be found in the Additional le 1 (Table E25 -E30).
In summary, these results indicate the same classi cation of NMs in vitro and in vivo, although biological effects differed. SiO2_7, SiO2_40, and TiO2_NM105 were classi ed to be "active", while SiO2_15_Amino led to almost no changes, thus suggesting it to be "passive". SiO2_15_Unmod induced the same trends as SiO2_7 and SiO2_40, thus indicating that it might have adverse effects at higher doses.

Page 5/17
In the present study, we aimed to analyze the MoAs of a set of ve NMs (SiO2_15_Unmod, SiO2_15_Amino, SiO2_7, SiO2_40, and TiO2_NM105) in vitro and in vivo. Thereby, we focused on the tiers of oxidative stress [32,33,34,35] that are induced in the different model systems upon treatment with the NMs at various doses. For in vitro toxicity assessment, alveolar type II cells and alveolar macrophages were used. Both originated from rats to assure comparability to the conducted STIS and instillations. In order to obtain detailed information on the occurring effects, proteomics and metabolomics were applied.
Importantly, SiO2_7 and SiO2_40 were classi ed to be "active" in all four model systems in vitro and in vivo, thus con rming our previous classi cation based on in vitro data [11,12]. Since SiO2_15_Unmod generally revealed the same trends as SiO2_7 and SiO2_40, even though with less signi cance, it should be classi ed "active" as well. This is in accordance with the literature, where it has been shown that SiO2_15_Unmod is "active" in vitro [38] and in vivo [25,73]. Importantly, SiO2_15_Amino was classi ed to be "passive" based on all conducted studies, which has been described before as well [11,25,38]. This illustrates that the used surface modi cation can in uence the effects, since the only difference between SiO2_15_Unmod and SiO2_15_Amino are amino groups on the surface of latter, which is in accordance to the literature [73]. Furthermore, it suggests that the route of production may affect NM toxicity because SiO2_15_Amino and SiO2_15_Unmod, which resulted in less effects, are precipitated silica NMs, while SiO2_7 and SiO2_40 are fumed. The observation that fumed NM variants can be more reactive than precipitated ones has been described before as well [74]. Interestingly, the results of the present study indicated TiO2_NM105 to be "active" in type II cells but "passive" in macrophages, which is in contrast to its former classi cation as "active" in the same cell line of alveolar macrophages, which was based on the toxicological endpoints [38]. Furthermore, it was shown to be "active" in vivo before [25].
Importantly, our main focus was to assign the observed effects to the tiers of a previously described strati ed oxidative stress model [32,33,34,35]. This assignment revealed that the same NMs can induce different tiers of oxidative stress in dependency of the used model system. Since the induction of the tiers of oxidative stress mainly depends on the GSH/GSSG ratio, we focused on GSH/GSSG signaling rst. While the "active" NMs led to increased abundances for proteins and metabolites that are related to the GSH/GSSG signaling, decreased abundances were observable in macrophages. This suggests that upon NM treatment, the type II cells triggered the production of GSH, to be able to neutralize the appearing ROS, thus protecting themselves from too high levels of oxidative stress and further damage, which has been observed before in type II cells after treatment with different materials [75,76]. In contrast, the decreased GSH/GSSG signaling in macrophages indicated decreased GSH levels, which have been described before as well and were explained by a NM-induced release of GSH or an increased requirement for GSH [77,78]. Importantly, the results presented here rather suggest a decreased GSH production. These opposite effects on the GSH/GSSG signaling subsequently led to differences for analytes that can be assigned to the tiers of oxidative stress. Since all proteins described here to be necessary for the production of GSH and NADPH are regulated by Nrf2 [48,49,50,51], we concluded that in type II cells only tier 1 was affected by the NM treatments but not the following tiers. In contrast, for macrophages signi cant changes were recognizable for the "active" NMs over all three tiers.
While the in vitro results allowed for a clear assignment to the tiers of oxidative stress, this concept was not transferable to the in vivo results (Fig. 3). This might be due to the fact that the lung consists of various cell types (Fig. 4a) that may display different effects upon NM exposure. An enrichment analysis with proteins and metabolites was conducted (Fig. 4c) to get an overview on the cellular processes, showing that the typical pathways that are related to the tiers of oxidative stress (glutathione redox reactions, Nrf2-mediated oxidative stress response, mitochondrial dysfunction, chemokine signaling, apoptosis signaling) were not among the most signi cantly enriched pathways. In contrast, pathways related to in ammatory processes like endocytosis signaling, phagosome maturation, leukocyte extravasation signaling, remodeling of epithelial adherens junctions, integrin signaling, and actin cytoskeleton signaling played a major role. Furthermore, it was observed that the recovery groups exhibited a more signi cant enrichment of pathways than the exposure groups, indicating that NM exposure might lead to long-term effects, which needs to be analyzed in further studies.
Taken together, the integrated omics data show that depending on the in vitro model system distinct and even opposite biological responses to the same NM can be detected. Alveolar macrophages seem to be a more sensitive in vitro model system as they have shown stronger responses. Therefore, and in line with the 3R concept, they provide a valuable tool in acute toxicity testing. Overall, it appears that the in vitro models inadequately re ected the in vivo situation. However, it must not be overlooked that a 24 h exposure of selected cell types to NMs cannot re ect what is happening in the complex lung after several days of exposure. Signaling pathways may be switched on or off rapidly, cell types will interact and respond differently. Bearing this in mind, it is important to underline that the classi cation of "active" and "passive" NMs was largely congruent between the tested model systems.

Conclusions
In summary, we show that NM-induced cellular effects can lead to a different assignment to the tiers of oxidative stress, which strongly depends on the used model system. Opposite effects on the GSH/GSSG signaling were observable in type II cells and macrophages, with type II cells leading to an increased GSH production and thus only tier 1 and macrophages displaying a decreased GSH production and signi cant changes over all three tiers. Hence, it was not surprising that the assignment was not as clear for the in vivo results of rat lungs that consist not only of type II cells and macrophages but also of type I cells, endothelial cells and interstitial cells. However, the in vitro and in vivo results were consistent regarding the classi cation of the tested NMs, thus indicating that both tested cell lines are suitable to assess NM toxicity. Since the macrophages exhibited more signi cant changes over all tiers of oxidative stress, they might be a more sensitive model system. Taken together, the presented results illustrate that the MoAs of NMs vary in different model systems and given that, care needs to be taken when using biomarkers for NM risk assessment.

Selected NMs and NM dispersion
SiO2_15_Amino and SiO2_15_Unmod (both precipitated) were provided in suspension by BASF SE. SiO2_7 and SiO2_40 (both pyrogenic) were manufactured by Evonik Industries and provided as powders. TiO2_NM105 was obtained as powder from the JRC repository. The physico-chemical properties for these NMs were described in detail earlier [11,12]. A summary of selected properties can be found in Table 1. Before use, NMs were dispersed by an indirect probe sonication protocol [79].

Cell culture
Cell culture of alveolar macrophages (cell line NR8383, ATCC, CRL-2192, USA) and alveolar type II cells (cell line RLE-6TN, ATCC, CRL-2300, USA) was conducted as described before [11,12]. Macrophages were exposed to 45, 22.5 and 11.25 µg/mL of SiO2_15_Amino, SiO2_15_Unmod, SiO2_40, SiO2_7, and TiO2_NM105 for 24 h under serum free conditions. Thus, nominal doses assuming complete sedimentation were 10, 5, and 2.5 µg/cm 2 . Selected results obtained with this experimental setup were previously published [12]. Type II cells were treated with 1, 10, or 50 µg/cm 2 NMs for 24 h and 48 h, respectively. TiO2_NM105 was investigated at doses of 0.1, 1 and 10 µg/cm 2 due to its high cytotoxicity. The TiO2_NM105 results were brie y described in one of our previous studies [11]. Untargeted proteomics were additionally conducted after 6 h of treatment.
For macrophages, three to four replicates were used for metabolomics and four replicates for proteomics. In case of type II cells, the metabolomics approach was conducted in four replicates, while proteomics was performed in four to ve replicates.

Short-term inhalation studies (STIS)
STIS were conducted with SiO2_7 and TiO2_NM105 under agreement «UN 18 306 DO», approved by the Committee on the Ethics of Animal Experiments of the University of Namur. The procedure was described before [80]. Brie y, female Wistar rats (Charles River, France) were exposed to the nanoaerosols (0.5, 2 or 5 mg/m 3 SiO2_7 and 0.5, 2, 10 or 50 mg/m 3 TiO2_NM105) for 6 h/d over 5 consecutive days or to ltered air (controls) and either sacri ced immediately after exposure (exposure groups, E) or 21 d after exposure (recovery groups, R) by a Nembutal intraperitoneal injection of (60 mg/kg) (Ceva Sante Animale, France).
For each group, ve rats were used. The lungs were directly frozen in liquid nitrogen.

Instillation studies
The experimental design for intratracheal instillation with SiO2_15_Amino, SiO2_15_Unmod, SiO2_7, and SiO2_40 was approved by The Ethics Committee of the "Vasile Goldis" Western University of Arad and authorized by the National Sanitary Veterinary and Food Safety Authority of Romania with registration no. 007/27.11.2017. The applied procedure has been previously described [80]. In brief, male Wistar rats from "Cantacuzino" National Institute of Research (Bucharest, Romania) were exposed to 0.36, 0.18, and 0.09 mg NM in PBS for 3 d (exposure groups, E) and directly sacri ced afterwards or after an additional recovery time of 21 d (recovery group, R). For each group, four rats were used. The lungs were directly frozen in liquid nitrogen.

Sample Preparation for Omics Studies
For the in vitro samples, the preparation was conducted as described before [11,12]. Metabolites and proteins were extracted from individual samples.
After STIS and instillations, the frozen lungs were homogenized as described before [80], and proteins as well as metabolites were extracted according to the in vitro studies.

Targeted Metabolomics and untargeted Proteomics
The here presented metabolomics data were published before [80] but re-evaluated in analogy to the proteomics data. In brief, the AbsoluteIDQ p180 Kit (Biocrates, Austria) was used as described previously [11,12,80,81], and samples were analyzed with an API 5500 triple quadrupole mass spectrometer (ABSciex, Germany), coupled to an Agilent 1260 In nity HPLC system (Agilent, USA). Analyst® software and MetIDQ were used for data analysis. Values below the limit of detection were excluded. Fold changes (FCs, treatment vs. control) of the average values were used for further analyses.
For untargeted proteomics, tandem mass tag (TMT)-labeling (Thermo Scienti c, USA) was applied. Therefore, 25 µg protein per sample were used and labeling was conducted with 0.1 mg label.
For type II cells, the work ow was conducted according to the manufacturer's instructions and as described before [11,82,83]. For these samples TMT-6-plex (Thermo Scienti c, USA) was used, while for all other sample sets TMT-10-plex (Thermo Scienti c, USA) was applied.
Effects on macrophages were investigated as described previously [12]. In brief, paramagnetic beads were used, which lead to an improved sample quality and allow for fractionation [84,85]. The paramagnetic bead approach was also applied to the samples from the in vivo studies.
As described before [11,12], labeled samples were analyzed on a nano-UPLC system (Ultimate 3000, Dionex, USA) with trapping column (Acclaim PepMap source (Nanomate, Advion, USA), coupled to the mass spectrometer (QExactive HF, Thermo Scienti c, USA). MS raw data were processed using ProteomeDiscoverer 2.2. The database search was performed against the UniprotKB reference proteome of Rattus norvegicus (27 February 2019), resulting in replicate FCs (treatment versus control), which were log2-transformed and median normalized before further analyses.

Statistical Analysis
Statistical analysis of log2-transformed FCs was performed in R-3.5.0 with the use of several packages [86,87,88,89,90,91,92,93,94,95]. To unravel signi cant changes compared to controls, the Student's t-test with Benjamini & Hochberg adjustment was performed for proteins and metabolites that were quanti ed in at least three biological replicates. FCs and adjusted p-values (p.adj) for all data sets can be found in the Additional le 1 (Table E3 -E10) together with replicate values and calculation results (Table E11 -E24).
Enrichment analyses were performed using Ingenuity Pathway Analysis (IPA, Qiagen, Germany) [96,97]. For this purpose, the data were ltered for signi cantly (p.adj ≤ 0.05) altered analytes, lung was selected as tissue and rat as organism. The used mapping of metabolites to identi ers from the Human Metabolome Database (HMDB) that was used for this analysis can be found in the Additional le 1 (Table E2), where also the Uniprot Accession mapping to genes is stored (Table E1). Furthermore, the obtained IPA results are summarized in Table E25

Authors' contributions
The study was planned by KS, AH, and MvB with contributions from AL. MvB and AH secured funding. AB conducted experiments with alveolar type II cells,  with different NMs at varying doses for analytes that are connected to the tiers of oxidative stress. Signi cant changes (p.adj ≤ 0.05) are labeled with * (a). Furthermore, the effects observed for the three tiers after treatment with 10 µg/cm2 SiO2_7 are presented and signi cant changes (p.adj ≤ 0.05) are indicated with bold letters (b).

Figure 3
Summary of in vivo results with focus on the tiers of oxidative stress. Shown are the percentages of signi cantly altered metabolites and proteins after STIS (left) and instillation (right) for ve different NMs that were applied at varying doses with distinguishing exposure groups (E) and recovery groups (R) (a).
Furthermore, the observed changes are summarized for proteins and metabolites that are connected to the GSH/GSSG signaling pathway as well as the tiers of oxidative stress and signi cant changes (p.adj ≤ 0.05) are labeled with * (b).