Proteomic analysis of protein carbonylation: a useful tool to unravel nanoparticle toxicity mechanisms
Driessen et al. Particle and Fibre Toxicology
Proteomic analysis of protein carbonylation: a useful tool to unravel nanoparticle toxicity mechanisms
Marc D. Driessen 0
Sarah Mues 2
Antje Vennemann 1
Bryan Hellack 5
Anne Bannuscher 0
Vishalini Vimalakanthan 0 3
Christian Riebeling 0
Rainer Ossig 2
Martin Wiemann 1
Jürgen Schnekenburger 2
Thomas A. J. Kuhlbusch 4 5
Bernhard Renard 3
Andreas Luch 0
Andrea Haase 0
0 Department of Chemicals and Product Safety, German Federal Institute for Risk Assessment (BfR) , Berlin , Germany
1 IBE R&D gGmbH, Institute for Lung Health , Münster , Germany
2 Biomedical Technology Center, Westfälische Wilhelms-University , Münster , Germany
3 Robert-Koch-Institut (RKI), Junior Research Group Bioinformatics , Berlin , Germany
4 Center for Nanointegration CENIDE, University of Duisburg-Essen , Duisburg , Germany
5 Institute of Energy and Environmental Technology (IUTA) e.V., Air Quality & Sustainable Nanotechnology , Duisburg , Germany
Background: Oxidative stress, a commonly used paradigm to explain nanoparticle (NP)-induced toxicity, results from an imbalance between reactive oxygen species (ROS) generation and detoxification. As one consequence, protein carbonyl levels may become enhanced. Thus, the qualitative and quantitative description of protein carbonylation may be used to characterize how biological systems respond to oxidative stress induced by NPs. Methods: We investigated a representative panel of 24 NPs including functionalized amorphous silica (6), zirconium dioxide (4), silver (4), titanium dioxide (3), zinc oxide (2), multiwalled carbon nanotubes (3), barium sulfate and boehmite. Surface reactivities of all NPs were studied in a cell-free system by electron spin resonance (ESR). NRK-52E cells were treated with all NPs, analyzed for viability (WST-1 assay) and intracellular ROS production (DCFDA assay). Carbonylated proteins were assessed by 1D and/or 2D immunoblotting and identified by matrix assisted laser desorption time-of-flight mass spectrometry (MALDI-TOF/TOF). In parallel, tissue homogenates from rat lungs intratracheally instilled with silver NPs were studied. Results: Eleven NPs induced elevated levels of carbonylated proteins. This was in good agreement with the surface reactivity of the NPs as obtained by ESR and the reduction in cell viability as assessed by WST-1 assay. By contrast, results obtained by DCFDA assay were deviating. Each NP induced an individual pattern of protein carbonyls on 2D immunoblots. Affected proteins comprised cytoskeletal components, proteins being involved in stress response, or cytoplasmic enzymes of central metabolic pathways such as glycolysis and gluconeogenesis. Furthermore, induction of carbonyls upon silver NP treatment was also verified in rat lung tissue homogenates. Conclusions: Analysis of protein carbonylation is a versatile and sensitive method to describe NP-induced oxidative stress and, therefore, can be used to identify NPs of concern. Furthermore, detailed information about compromised proteins may aid in classifying NPs according to their mode of action.
Surface functionalization; Protein carbonylation; Oxidative stress; Silica nanoparticles; Zirconium oxide nanoparticles; Silver nanoparticles; Rat lung; ESR; DCFDA
The range of industrial processes and products taking
advantage of nanotechnology has been growing rapidly
in recent years. Current uses of nanoparticles (NPs)
comprise e.g. the use of ZnO and TiO2 as UV protection
agents in cosmetics and, due to their biocidal activity,
silver NPs in packaging or medical devices, and ZrO2
and SiO2 as binders in ceramics and fillers in modern
polymers. With this ever increasing diverse and common
use of NPs it has become important to address concerns
regarding possible adverse health effects. Currently,
adaptation of test guidelines for NPs and debates on
their validity for NPs are still ongoing . In particular,
fast and reliable toxicity screening methods are urgently
needed as well as knowledge on the underlying toxic
modes of action [2, 3]. Especially the latter aspect will
foster the development of reliable alternative testing methods
and may help to guide production of safe-by-design NPs.
Furthermore, a more detailed knowledge of toxicity
mechanisms is helpful for a regulatory prioritization and also
for any successful grouping approach.
One prevailing paradigm explaining NP-mediated
toxicity is the induction of oxidative stress [4–8]. Oxidative
stress has been linked to various adverse outcomes such
as inflammation, DNA damage, and general cytotoxicity
. It results from an imbalance between reactive
oxygen species (ROS) generation and cellular antioxidants.
Intracellular ROS are generated as regular byproduct of
the respiratory chain and other oxygen consuming
reactions . Furthermore, ROS are induced in to cells by
NPs in different ways [2, 11–17]. Among these, surface
reactions driven by excitation of electrons via UV-light,
Fenton-type reactions, catalytic chemistry at the NP
surface, or via dissolved (metal) ions are implicated most
prominently . The types of ROS generated depend in
particular on the NP and its chemical environment .
Various detection methods are available. Free radicals can
be detected in vitro via electron spin resonance (ESR)
spectroscopy using reagents called spin traps. These
molecules form adducts to stabilize the radicals which then
exhibit a paramagnetic resonance detectable by
spectroscopy . For instance, singlet oxygen and superoxide
radicals can be detected using
1-hydroxy-3-carboxy-pyrrolidine (CPH), whereas 5,5-dimethylpyrroline N-oxide
(DMPO) is sensitive to hydroxyl and superoxide radicals
. Furthermore NPs may induce ROS in cells due to
other mechanisms as described below . Xia and
coworkers suggested that disruption of phagosomes leads to
a first ROS peak, while a second increase stems from
mitochondrial damage being part of apoptosis . Thus
a time and site-specific ROS generation along with specific
byproducts may be expected. Moreover, it has been
demonstrated that NPs may elicit a respiratory burst in
professional phagocytes such as macrophages [14, 15], which
use a NADPH-oxidase dependent defense mechanism
primarily against viruses or bacteria together with expression
of cytokines [12, 13]. In tissues or complex cell models,
prolonged activation of cytokines and ROS may lead to
oxidative damage and stress in surrounding cells, causing
inflammation and potentially promoting DNA damage
and tumorigenesis . Finally, chronic depletion of
cellular antioxidants such as glutathione renders cells
vulnerable for oxidative stress  and stimuli by cytokines ,
possibly amplifying NP effects.
ROS detection in cells often uses fluorescent
dyebased approaches, for example loading with
dichlorofluorescein (DCFDA). However, these dyes when used in
nanotoxicological studies may suffer from interferences
with the NPs  and, albeit widely used, are prone to
some risk of falsification. Thus, an alternative method to
report on cellular ROS generation is highly needed. It is
long been known that ROS cause oxidative
modifications of cellular components, and especially proteins
have often been described as predominant targets for
oxidative modifications as they may scavenge up to 70 %
of ROS [20, 21]. Thus, formation of protein carbonyls
are among the most prevalent oxidative lesions of
proteins . Carbonylation is irreversible and usually
results in an impairment or even loss of protein function,
often associated with protein unfolding and aggregation.
It may also be involved in signal transduction .
Different types of protein carbonyls (aldehydes and ketones)
are formed either by peptide backbone fragmentation,
side chain oxidation, or by a secondary reaction with
oxidized cellular metabolites . Analysis of oxidatively
modified proteins in neurodegenerative diseases
revealed that affected proteins are mainly involved in
glucose metabolism, mitochondrial function, cellular
motility/structural integrity, and protein degradation
. Various metabolic diseases [20, 26, 27],
neurodegenerative diseases  as well as aging and age-related
diseases occur along with elevated protein carbonyl levels
[28, 29]. Furthermore several chemicals can enhance
protein carbonylation. For instance in black tiger shrimp, the
banned insecticide and acaricide endosulfan increased
lipid peroxidation levels and protein carbonylation  as
did the hepatotoxic and hepatotumorigenic fungicide
propiconazole in mice  and ethanol exposure in rats .
Protein carbonyls are increasingly studied, e.g., in
environmental toxicology where they are used as biomarkers of
oxidative stress, as well as indicators of toxicological
modes of actions [33–36] as usually the observed
carbonylation pattern is dependent on the toxicant. Silver NPs
were also shown to induce carbonyls in Daphnia
Protein carbonyls can react with
2,4-dinitrophenylhydrazine (DNPH) and the resulting 2,4-dinitrophenylhydrazones
can be detected with 2,4-dinitrophenyl specific antibodies in
precipitates  or immunoblots [39, 40]. The latter
approach has been used to detect protein carbonyls in obese
mice , and humans . With respect to NPs it has been
shown that silver but not gold NPs induced protein
carbonylation in THP-1 macrophages, primary neuronal cells [42,
43] and in a human colon epithelial cell line in a particle size
dependent manner .
Especially the data on NP-treated cells suggest that
analyzing protein carbonylation may be a useful tool for
studying qualitatively and quantitatively the level of
oxidative stress which had been acting upon NP-exposed cells
or tissues. The derivatization of carbonyls with DNPH
followed by immunoblotting is a specific and powerful
technique, as it separates analytes and NPs and, therefore,
is not expected to suffer from any NP interference.
The aim of the present study was to describe and
compare the protein carbonyl pattern in NRK-52E cells
subjected to a representative set of 24 NPs comprising
amorphous silica (6 different types), zirconium dioxide
(4 different types), silver (4 different types), titanium
dioxide (3 different types), zinc oxide (2 different types),
multiwalled carbon nanotubes (MWCNT, 3 different
types), barium sulfate and boehmite (AlOOH). Results
were evaluated and discussed with respect to chemical
composition and different surface functionalizations. To
investigate whether there is a relevance of NP treatment
also in vivo we analyzed lung tissue of silver NP instilled
rats in parallel. In this study we applied and compared
different methods in parallel to the analysis of
All nanoparticles (NPs) were dispersed in H2O and in
complete cell culture medium (CCM; DMEM cell
culture medium supplemented with 10 % fetal calf serum),
the latter of which was the relevant biological test
medium for the in vitro studies. About half of the NPs
were dispersed by a stirring-based dispersion protocol
(see Table 1) to preserve the functionalized surfaces.
However, some of the NPs were hydrophobic and
therefore difficult to disperse by stirring. Thus, we also
included sonication-based dispersion methods for some of
the NPs (see Table 1). We analyzed dispersion quality in
water and CCM and also tested for stability of the
dispersions in CCM over a time course of 24 h. As shown in
Table 1 most NPs were well dispersed in H2O and CCM,
and mostly dispersions were stable for up to 24 h.
Nanoparticle surface reactivity measured by ESR
Next, we measured the surface reactivity of the NPs in a
cell free system in H2O by ESR spectroscopy (Table 2)
using CPH and DMPO as spin reagents. In total, 12 out
of 23 NPs tested by ESR either had a significant surface
reactivity with CPH or indicated fenton-like reactions
occuring with DMPO. Applying the CPH spin probe we
detected positive surface reactivity with SiO2
(unmodified, NM-200, NM-203), Ag 50 PVP, Ag 200 PVP, TiO2
NM-103 and ZnO (NM-110, NM-111). In addition, SiO2
(unmodified, Amino, Phosphate), TiO2 NM-105, MWCT
NM-402 as well as ZnO (NM-110, NM-111) were found
reactive in the presence of DMPO (Table 2).
Cell viability after NP treatment
All NPs were tested for cytotoxic effects on an
immortalized rat kidney epithelial cell line, NRK-52E cells using
NP concentrations of 0–100 μg/mL (0–30 μg/cm2) in a
WST-1 assay (Table 2). This cell model has been chosen
as it appeared highly sensitive with respect to the
development of oxidative stress after NP treatment in pilot
studies and closely resembles differentiated tissue in
contrast to many dedifferentiated tumor cell lines. Using
this dose range four NP treatments reached an IC50,
namely ZnO NM-110 and NM-111, Ag 50 PVP and Ag
NM-300 k (Table 2). Interestingly, the IC50 of NM-110
differed when dispersed by stirring (7.5 μg/mL = 2.2 μg/
cm2) or sonication (12.5 μg/mL = 3.7 μg/cm2). TiO2
NM-105 was cytotoxic as well. SiO2 unmodified and
SiO2 NM-203 were slightly toxic and did not reach an
IC50 in the tested concentration range.
ROS generation by DCFDA assay
Three out of 22 NPs that were tested by the DCFDA
assay induced ROS in NRK-52E cells, namely the three
MWCNTs. NM-400 and NM-402 showed an about 1.4
fold increase in fluorescence at a concentration of
34 μg/mL (10 μg/cm2), while NM-401 showed a similar
increase only at 170 μg/mL (50 μg/cm2). However, it
should be noted that due to interferences in the
readout some NPs (SiO2 unmodified/PEG/Amino/Phosphate,
all ZrO2; Ag 50 PVP/Citrate, Ag 200 PVP, Ag NM-300,
TiO2 NM-105) could be tested only up to
concentrations of 34 μg/mL (10 μg/cm2).
Protein carbonylation by 1D immunoblotting
To analyze protein carbonylation we treated NRK-52E
cells with increasing NP concentrations of up to
100 μg/ml (30 μg/cm2). In total 11 out of 24 NPs
were found to induce protein carbonylation (Table 2,
Additional file 1: Figure S2). Particularly high reactivity
was detected for SiO2 unmodified, Ag 50 PVP, and the
ZnO variants. SiO2 Amino, Ag 50 Citrate, TiO2 NM-105,
SiO2 NM-200, SiO2 NM-203, and Ag NM-300 K were
also positive, albeit with lower reactivity. A weak reactivity
was detected with SiO2 Phosphate. Two NPs, i.e. ZrO2
Amino and SiO2 PEG, showed a weak response, but this
could be attributed to the supernatant controls, rendering
the results unspecific and both NPs were considered
Table 1 Basic NP characterization
SiO2 Unmod. None Stir
PEG Polyethylene-glycol (MW 500 Stir
Amino Aminopropyltrimethoxysilane Stir
Phosphate TPMP Stir
NM-200 None US
NM-203 None US
ZrO2 Acrylate Acrylate Stir
PEG Polyethylene-glycol (MW 500 Stir
Amino Aminopropyltrimethoxysilane Stir
TODS TODS Stir
Ag 50 PVP Polyvinyl-pyrrolidone Stir
200 PVP Polyvinyl-pyrrolidone Stir
50 Citrate Citrate Stir
NM-300k Polyoxyethylene Glycerol Trioleate/ US
Polyoxyethylene (20)/Tween 20
TiO2 NM-103 (rutile) Polydimethyl siloxane US
NM-104 (rutile) Polydimethyl siloxane US
NM-105 (15 % rutile/ None Stir
85 % anatase)
aStir = dispersion by stirring, US = dispersion by sonication
btaken from JRC reports
cnote that DLS is not suitable for measurement of MWNCT
US Stir Stir
negative in Table 2. The NP free supernatant controls
were obtained after prolonged ultracentrifugation and
should mainly contain the dispersant and stabilizers. All
other NPs induced no protein carbonyls and were
considered negative. Additional 1: Figure S2 depicts all
immunoblots for the positive NPs, and Additional file 1: Figure S3
for the negative NP.
Comparison of oxidative stress assessment
Comparing DCFDA and ESR measurements we found
an agreement for only 8 out of 22 NP treatments
(Table 2). In most cases these NPs were completely
negative in both assays. In total we observed an overlap
of 36 % (Table 3). Often, NPs were found positive in the
ESR assay, but DCFDA results remained negative. We
found a similar overlap between DCFDA and the
carbonyl assay (8 of 22 NPs, equals 36 %), here no positive
overlaps occurred. All NPs that gave a positive result in
the carbonylation assay did not induce ROS measured
by DCFDA. However, when comparing the
carbonylation assay and ESR measurements we observed an
excellent overlap. For 19 out of 23 NPs we found an
agreement between the carbonyl and the ESR results,
i.e. an overlap of 83 % (Tables 2 and 3). Interestingly,
Table 2 NP oxidative stress potential
Table 3 Comparison of oxidative stress assessments
Carbonylation (23 NPs)a +
DCFDA (21 NPs)a,b +
aAg 50 Citrate not analyzed by ESR
bBaSO4 and AlOOH not analyzed in DCFDA
in most cases (20 out of 24 NPs) we also found an
excellent correlation between results of the protein
carbonyl assay and cell viability (Table 2), i.e. 83 %
Each NP induces a unique pattern of protein carbonyls in
All NPs that were tested positive in the 1D screening assay
were subjected to a more detailed analysis of the
carbonylation patterns using 2D immunoblots. In total, we tested
11 different kinds of NPs: SiO2 (unmodified, Amino,
Phosphate, NM-200, NM-203), Ag 50 PVP Ag 50 Citrate,
Ag NM-300 K, TiO2 NM-105, ZnO NM-111 and ZnO
NM-110 (note: NM-110 was tested with two different
dispersion protocols). Figure 1 shows the 2D
immunoblotting results for control (untreated) cells and cells treated
with SiO2 unmodified, Ag 50 PVP, and ZnO NM-110. All
other results are given in Additional file 1: Figure S4. In
total we detected 202 carbonylated protein spots, out of
which 55 % could be identified by mass spectrometry. In
untreated control cells we found 52 protein spots
carbonylated, which were also detected in most of the
NPtreated samples. Each NP induced a unique pattern of
carbonylated proteins. For a few NPs the spot pattern was
similar to controls, although not identical. These NPs had
a similar number of carbonylated proteins when
compared to the control but displayed enhanced signal
intensities in several of the carbonylated proteins, thus
indicating a higher level of carbonylation for the respective
proteins. We also analyzed the overlaps in the
carbonylation pattern for all NPs (see Venn diagrams in Fig. 2). For
instance, SiO2 NPs share 79 to 95 % of their respective
carbonylated protein spots. Interestingly we could detect
differences for ZnO NM-110, depending on the dispersion
protocol used. Clearly, the stirring-based protocol resulted
in a lower number of protein carbonyls (68 spots)
compared to the sonication-based protocol (80 spots), 48 from
these were found in common.
Proteins involved in glycolysis, ATP synthesis and cell
integrity/motility are carbonylated after NP treatment
From the total of 202 carbonyl spots 112 proteins
(55 %) were successfully identified by mass
spectrometry (Additional file 1: Table S1 and Additional file 1:
Figure S5). Identification of some spots was impeded by
very low protein amounts despite strong carbonyl
immunoblot signals. In other cases, identification was hampered
by very strong signals in immunoblots covering several
neighboring protein spots in the corresponding duplicate
gel. In these cases we excised all possible protein spots
and these proteins were labeled “multiple” in Table S1.
The majority of carbonylated proteins comprised
cytoskeletal proteins, proteins being involved in the cellular
stress response, or enzymes of central metabolic
pathways such as glycolysis or gluconeogenesis. To analyze
affected signalling pathways in more detail we used
Ingenuity pathway analysis (IPA). IPA confirmed that 8 out
of 10 cytoplasmic enzymes involved in glycolysis and
gluconeogenesis were carbonylated after NP treatment.
Proteins involved in Rho signaling, the unfolded protein
response, actin-based signaling, integrin signaling, and
clathrin-mediated endocytosis appeared carbonylated as
well. The complete list of assigned pathways is given in
Fig. 3. Figure 4 shows an IPA of possibly affected cellular
functions. Thus, elevated carbonylation of proteins is
likely to be correlated with increased necrosis and cell
death and impaired glycolysis and ATP synthesis.
Principal component and hierarchical cluster analysis
Based on the 2D immunoblot data, i.e. based on the spot
numbers and median intensities, we performed a
statistical analysis (principal component and hierarchical
cluster analysis) in order to analyze similarities between
different treatments. These powerful tools can represent
large amount of data in a concise way and identify
possible clusters, which importantly could indicate a similar
mode of action for the respective NPs.
Hierarchical cluster analysis (HCA) in this case
compares the differences in normalized spot intensities among
all data, thereby revealing clusters of nearest neighbors.
We used a Euclidian complete linkage algorithm for the
HCA (Fig. 5) which finds compact clusters of
approximately equal diameters. We identified SiO2 NM-200 as
being most similar to control cells. ZnO NM-110 and
ZnO NM-111 were grouped together, indicating a similar
Fig. 1 Protein carbonylation analyzed via 2D immunoblot. NRK-52E cells were treated for 6 h with 10 μg/ml of the indicated NP in at least 3
independent biological repeats. Carbonyls were detected using an anti-DNP antibody after coupling to 2,4-Dinitrophenylhydrazine and
visualized by ECL. Representative blots are shown
Fig. 2 Venn diagrams. Similarities and differences in the spot pattern of protein carbonylation in the different treatments groups are visualized in
Venn diagrams. Depicted are results for ZnO (a), nanosilver (b) and SiO2 variants (c). The sum of all numbers in one oval equals the total number
of carbonylated spots for that respective NP. Numbers given in an overlap of two or more ovals represent the number of spots shared by the
mode of action. ZnO NM-110 (stirring protocol) was
allocated to a different group, more similar to Ag 50 Citrate,
but not together with ZnO (NM-110 and NM-111,
ultrasonication-based protocol), again indicating a
difference in mode of action based on the dispersion protocol.
Interestingly SiO2 unmodified and Ag 50 PVP were
grouped together as well, indicating a similar mode of
action with respect to protein carbonylation.
Principal component analysis (PCA) finds the largest
possible variability in the normalized data. Each
subsequent principal component (PC) describes the largest
variance in a direction orthogonally to the previous PC;
thereby each PC is linearly uncorrelated. Plotting of the
PCs against each other may then reveal groups of lower
variance, i.e. they are similar in the parameters that
contribute to the variance covered by the plotted PCs.
Applying PCA with five PCs we could explain 65 % of the data
variance. Figure 6 depicts the first two PCs covering 34 %
of the data variance. Here we found SiO2 NM-200, SiO2
Amino and SiO2 Phosphate located closest to the
untreated controls. SiO2 NM-203 was an outlier probably
because it exhibits the highest number of spots (c.f. Fig. 2)
albeit all with weak intensity. The ZnO NPs, TiO2, and Ag
50 Citrate grouped together around the axis of PC1 in the
positive half of PC2, all of which display medium to strong
carbonylation. SiO2 unmodified and Ag 50 PVP found in
the negative half of PC1 and PC2, both very strong
inducers of carbonylation and they were clearly separated
from the other NPs. Overall analysis with PCA was in
good agreement to results from HCA.
Carbonylation in rat lung tissues after intratracheal
instillation of silver NPs
To analyze whether carbonylation upon Ag 50 PVP
treatment also occurs in vivo we analyzed lysates from
rat lung tissue blocks isolated 21 days after intratracheal
instillation of Ag 50 PVP (37.5 μg, 75 μg, 150 μg, per
lung). Figure 7 shows that the overall level of protein
carbonylation as detected in 1D immunoblots was
increased in a dose-dependent manner upon treatments
with Ag 50 PVP. Compared to controls robust signals
were especially obtained for high molecular weight
proteins upon instillation of doses of 150 μg/rat lung.
Analysis of protein carbonyls and ESR prove useful to
analyze oxidative stress induced by NPs
In this study we compared three different approaches to
analyze the oxidative stress potential inferred by a set of
24 NPs. Overall, ESR and DCFDA results were in
Fig. 3 IPA Analysis of protein carbonylation in NRK-52E cells. Pathways are listed with decreasing probability of being affected by NP treatment.
Color intensity reflects the number of carbonylated proteins belonging to the respective pathways. Signaling pathways are listed in the upper
panel. Metabolic pathways are listed in the lower panel
agreement for 38 % of all NPs. Comparing the results of
the 1D immunoblot carbonylation assay with ESR
measurements, we found a much better overlap and results
were in good agreement for 83 % of the NPs tested. Nearly
all NPs, except for Ag NM-300 k, which induced protein
carbonyls, were also tested positive by ESR. A few NPs
(Ag 200 PVP, TiO2 NM-103, MWCNT NM-402) were
tested positive in ESR but were negative in the carbonyl
assay. Thus, surface reactivity, as indicated by either one
of the two ESR reagents, appeared to be contributing to
oxidative stress induction in cells, measurable as protein
carbonylation. Since the results were not in perfect
agreement for all cases, other aspects seem important as well
(see below). Nevertheless, we conclude that ESR is well
suited to analyze the biological relevant surface reactivity
of NPs. However, it appears important to combine ESR
Fig. 4 IPA Analysis of affected cellular functions. Cellular functions are listed with decreasing probability of being affected by carbonylation. Color
intensity reflects the number of carbonylated proteins belonging to the respective pathways
results obtained with different reagents to fully describe
biological ROS formation. It should be underlined that
ESR analysis of NPs as produced (i.e. irrespective of
changes that may occur in biological surroundings)
appears to predict at least in part carbonylation of
Fig. 5 Hierarchical clustering analysis (HCA). HCA was performed
using the median intensities of spots determined by image analysis
of 2D immunoblots with Delta 2D software using a Euclidian
complete linkage algorithm in R
cytoplasmic proteins as a biological endpoint. This is an
intriguing finding, as all particles were likely surrounded
by a protein corona whose influence on direct oxidative
processes remains elusive and requires further study.
Investigating oxidative stress of NP-treated cells by
analyzing carbonylated proteins has some general
advances. First, it circumvents all interferences with dyes
or assay systems, as observed here for the DCFDA assay,
in which the range of possible test concentrations was
limited . Secondly, by applying 2D-based proteome
analysis it is possible to obtain detailed information on
affected proteins and signaling pathways and these data
can also be used to gain insight into the subcellular
localization of oxidative processes. Thirdly, compared to
the DCFDA assay, protein carbonyl analysis appears to
be much more sensitive. Finally, analysis of protein
carbonylation is also possible in NP-treated tissues after
in vivo testing. The approach therefore appears useful to
compare effects of NP treatment in vitro and in vivo in
more detail, which is urgently required to further refine
the conditions of in vitro testing.
Analysis of protein carbonylation
All NPs that were tested positive in the 1D carbonyl
screening assay were subsequently tested in a 2D-based
approach in more detail. When comparing the results
from the 2D approach two aspects need to be considered.
One is the absolute number of carbonyl spots. For
several NPs we observed an increased number of
Fig. 6 Principal component analysis (PCA). PCA was performed
based on 5 main principal components using the median intensities
of spots determined by image analysis of 2D immunoblots by Delta
2days software. Separation in the first two principal components
carbonylated protein species upon treatment. The
second aspect to consider is the spot intensity. For a
few NPs we did not detect much increase in the
absolute spot number compared to untreated control
cells but we detected increased spot intensities for
several carbonylated protein species. Finally both
kinds of information (i.e. spot number and spot
intensity) need to be combined to obtain information
on the biological activity. For instance, for SiO2
NM-203 elicited the highest number of carbonylated
protein species (121 spots) but all of them showed
moderate intensities on a quantitative scale such that
overall carbonylation activity of SiO2 NM-203 is only
medium (Table 2).
Carbonylation in untreated samples may be explained
by a low amount of ROS that are produced in cells as a
result of leakage from the mitochondrial electron transport
chain. Moreover, it has repeatedly been reported that
tumors exhibit a higher level of carbonylation compared
to healthy tissue, which also holds true for cell lines
(which are often derived from tumors) compared to e.g.
primary cells . In control cells, we mostly found
carbonylation of proteins that are involved in stress responses
such as chaperones, but also in proteins of the
cytoskeleton or enzymes of the glycolysis. Although each type of
NP induced a characteristic pattern of carbonylated
protein spots, a quite significant overlap occurred which may
reflect physico-chemical similarities of NPs.
Cytoskeletal proteins have often been identified as a
main target of carbonylation  and this effect is
not specific for NP. For instance, carbonylation of
actin could be detected in vitro after treatment of
cells with known toxicants, e.g. acrolein, hypochlorite,
or chloramines. Mussels exposed to environmental
pollutants showed increased actin carbonylation .
Furthermore, carbonylation of actin has been
frequently linked to aging and also to several diseases,
in particular neurodegenerative diseases like
Alzheimer's Disease (AD), but also to heart failure, tumor
growth, or prolonged inflammatory conditions . In
general, carbonylation of actin results in unfolding of
actin monomers, depolymerization of actin strands,
and aggregation of carbonylated actin. Since
carbonylation of actin is also detected in untreated control
cells it may be considered a ROS scavenger molecule
under normal conditions.
However, we found excessive actin carbonylation in
particular for SiO2 unmodified and Ag 50 PVP particles.
Somewhat lower actin carbonylation was detected
after treatment with SiO2 Amino/Phosphate/NM-200. A
lesser degree of actin carbonylation was detected after
treatment of the cells with Ag 50 Citrate and ZnO
NM-110. In a similar manner tubulin carbonylation can
be discussed. Actin and tubulin carbonylation are often
detected together [34, 47].
Glycolysis / gluconeogenesis
Enzymes involved in glycolysis/gluconeogenesis are
also often reported to be carbonylated. We identified
up to 8 out of 10 glycolytic proteins carbonylated.
Carbonylation of glycolytic enzymes can substantially
decrease glycolysis turnover as shown in HL60 cells
after treatment with etoposide (VP16) . It has been
speculated that impairment of glycolysis can put cells in
an inactive state . Depleting cellular ATP levels as a
result of decreased glycolysis could lead to the inhibition of
apoptosis . In our study we found two glycolytic
enzymes carbonylated in control cells, triosephosphate
isomerase and glyceraldehyde-3-phosphate dehydrogenase.
SiO2 NM-203 induced carbonylation of 8 out of 10
glycolytic enzymes. SiO2 phosphate, SiO2 NM-200 and Ag 50
PVP induced carbonylation of glycolytic enzymes as well.
In summary, nearly all tested NPs augmented the
carbonylation of glycolytic enzymes to various extents.
However, there was no obvious link between toxicity and
carbonylation of glycolytic enzymes. England et al. 
suggested that carbonylation of glycolytic enzymes rather
serves as protective mechanism which decreases the level
of apoptosis. This is in line with the results observed for
SiO2 NM-203, which showed no cytotoxicity despite a
high level of carbonylation of glycolytic enzymes.
Fig. 7 Elevated protein carbonylation in tissue lysates of rat lungs. a Tissue fragments from the left lungs of rats harvested 3 weeks after intratracheal
instillation of Ag 50 PVP NP were analyzed for carbonylated proteins by 1D immune blotting. Actin served as a loading control. Control animals received
0.5 ml of vehicle (0.9 % NaCl). b Densitometric analysis of the immunoblot depicted in A). DNPH signals were summarized and adjusted to the actin signal.
Animal #2 was excluded from analysis of the controls as the actin signals revealed that significantly less protein has been loaded compared to the other lanes
Several other pathways were found to contain
carbonylated proteins after NP treatment. In particular, we
found proteins of the unfolded protein response
pathway carbonylated, especially chaperones or
disulfide isomerases. This pathway was strongly affected
by SiO2 unmodified/Phosphate and ZnO NM-110
(stirring). Furthermore, proteins of the Rho signaling
pathway were found carbonylated after NP treatment,
especially after treatment with SiO2 unmodified/Amino/
NM-200 and Ag 50 PVP. The Rho signaling pathway is
particularly important for cell migration, cell adhesion, cell
polarity, changes in the actin cytoskeleton, and
intracellular transport processes by vesicles. Considering that actin
was modified by carbonylation as well, cell motility and
cellular transport processes may be impaired by
carbonylation. Our data suggest a good correlation between
NPinduced carbonylation of proteins of the Rho signaling
pathway (together with proteins of cytoskeleton/cell
motility), and NP toxicity. At present, no information is
available for membrane proteins associated with cellular
organelles as they are hardly detectable in the 2D gels.
Influence of NP’s chemical composition and surface
modification on oxidative stress
Earlier studies have demonstrated that the capacity of a
NP for inducing oxidative stress depends on its size
(mostly due to the higher surface area/mass ratio) [49, 50]
and surface modification [51, 52]. For silica NPs it has
been reported that they can induce oxidative stress
in vitro  and in vivo . Yoshida et al. showed that
surface modification of uncoated 70 nm silica particles by
amination or carboxylation decreased cytotoxicity, DNA
damage, and intracellular ROS generation in HaCaT and
TLR-1 cells . Similarly, Imai et al. reported a decreased
inhibition of CYP3A4 activity by 30 and 70 nm silica NPs
in HepG2 cells after surface carboxylation .
In our study we confirm that unmodified silica NPs
cause stronger effects with respect to carbonylation than
any of the surface modified SiO2 NPs. Yet, differences
between the modified SiO2 Amino or SiO2 Phosphate
and the larger, unmodified SiO2 NM-200 and SiO2
#NM-203 are sparse compared to the strong effects of
Also for silver NPs, oxidative stress has been reported
in vivo  and in vitro . Comparing the results
between the different types of silver that have been
tested in our study we found both a size-dependent
effect (i.e. compare 50 and 200 nm PVP coated silver)
and a coating-dependent effect (i.e. compare 50 nm
silver citrate vs PVP coating). A size-dependent
induction of oxidative stress has previously been
demonstrated for nanosilver. Carlson et al. found an up to
10-fold increase of ROS activity (via DCFDA assay) in a
rat alveolar macrophage cell line (NR8383) for 15 nm
and 55 nm silver particles at the same concentration
. However, Li et al. reported a higher O-2 formation
under UV light for citrate coated particles when
compared to PVP coated NPs of similar size . Our study
confirms the following ranking with respect to carbonyl
induction: Ag 50 PVP > > Ag 50 Citrate = NM-300 k >
Ag 200 PVP. Interestingly, ESR data did not reveal a size
dependency, as 50 and 200 nm PVP coated silver NPs
were rather similar (Ag 200 PVP even had stronger
effects when using the CPH probe). Thus, the size
dependency that we observed in the carbonylation assay
in vitro might be caused by other effects such as
different uptake rates.
For the different TiO2 variants ESR predicted an
activity for NM-103 [rutile] and NM-105 [15 % rutile/85 %
anatase]. However, in our biological in vitro test system
(NRK-52E cells) we detected an effect only for TiO2
NM105. In general, anatase TiO2 NPs are more cytotoxic than
rutile NPs . This effect might be caused by
photocatalytic activity and has also been demonstrated by Yin et al.
who compared anatase particles with rutile and
mixedstructure particles . Under UV irradiation they found
that the NPs with mixed anatase/rutile structure showed
the highest toxicity in HaCaT cells followed by anatase
particles, while rutile TiO2 showed the lowest toxicity .
Both ZnO variants (NM-110, NM-111) were classified
as positive in our protein carbonylation study and were
highly cytotoxic. ZnO NM-111, which is a coated
material, appeared to be slightly less reactive in the 1D
carbonyl screening. ZnO NM-110 was tested with two
different dispersion protocols. ESR measurements
indicated that NM-110 is far less reactive when dispersed by
ultrasonication compared to the stirring procedure. This
may be due to alteration of the NP surface by
ultrasonication, similar to what has been reported for TiO2 .
However, in biological systems we found that ZnO
NM110 is slightly less reactive, when being dispersed by
stirring with respect to carbonylation. Obviously in biological
systems the situation is more complex. At present we can
only speculate that particles obtained after stirring for
24 h or ultrasonication differ with respect to dissolution of
Zn2+ ions, agglomeration stability and/or corona
formation. All of which will influence uptake rate and cell death.
Note for instance that dispersions via ultrasonication are
achieved in a significantly shorter time frame. Therefore,
both protocols may have distinct effects on the
formation of a protein corona. As both techniques have
different energy intakes they may also result in different
ZnO dissolution rates.
For none of the ZrO2 particles, BaSO4 NM-220 or
AlOOH a clear cytotoxic response could be found.
MWCNTs elicited a response only in the DCFDA assay.
Interestingly, cytotoxicity in case of the MWCNTs was
dependent on the kind of cell model chosen. We have also
tested MWCNTs in A549 cells and in THP-1 derived
macrophages (data not shown). A strong cytotoxic effect
and oxidative stress as detected by protein carbonylation
was observed in THP-1 macrophages for NM-401
(MWCNT with the highest aspect ratio). We did not
detect cytotoxicity of MWCNT in NRK-52E or A549 cells.
BaSO4 and AlOOH may be considered as negative
Our study employed three different approaches to
analyze the oxidative stress potential of a panel of 24
different NPs. Cell-free ESR spectroscopy and in vitro
analysis of carbonyl patterns appear to be well suited for
this purpose while the use of the DCFDA assay is
limited due to NP interferences.
The analysis of protein carbonylation appears to be a
very sensitive method to detect possible adverse effects of
NPs. Furthermore, we observed a good correlation
between ESR and carbonyl results, suggesting that most NPs
used in our study induce oxidative stress mostly due to
surface reactivity. In addition, we found a good correlation
between the results obtained in the carbonyl assay and the
overall NP toxicity as tested by WST-1 assay. Thus, we
propose that the analysis of protein carbonylation after 1D
immunoblotting can be used as a predictive screening
method to identify NPs of concern. An advantage of this
approach is that in a second step the identification of
specifically altered proteins is possible via 2D separation. The
more detailed proteomic analysis appears to be a
promising tool to unravel underlying toxicity mechanisms.
Hence, redox profiling might prove useful for NP
classification according to their mode of action.
project to introduce different surface modifications on
three different core materials (i.e. silica, zirconia and silver,
Table 1). The SiO2 variants (SiO2 unmodified, SiO2
Amino, SiO2 Phosphate and SiO2 PEG), all 15 nm in
nominal size, were obtained from BASF. ZrO2 (ZrO2 Acryl,
ZrO2 TODS, ZrO2 Amino and ZrO2 PEG), all 10 nm in
nominal size, were obtained from CeraNovis (Germany)
and 50 nm as well as 200 nm silver NPs (Ag 50 PVP, Ag
50 Citrate, Ag 200 PVP) from Bayer Technology Services
(Germany). These 11 NPs were synthesized by sol–gel or
precipitation and were available in aqueous suspensions.
In addition, we used boehmite (AlOOH) from Bayer
Materials Science (Germany), BaSO4 NM-220 from Solvay
(Germany), TiO2 NM-105 and ZnO NM-110 from the
JRC repository (Ispra, Italy), all of which were obtained as
powders. A complete characterization data set is available
on the nanoGEM webpage.(http://www.nanogem.de/cms/
nanogem /upload/Veroeffentlichungen /nanoGEM_Del1.3.
Additionally, 10 NPs from the EU FP7 project MARINA
were used in this study, all of which were obtained from
the JRC repository (Ispra, Italy), mostly as powders. These
are ZnO (NM-110, NM-111), TiO2 (NM-103, NM-104),
Ag NM-300 K, and MWCNT (NM-400, NM-401,
NM402). Extensive characterization reports for these NPs are
provided by the JRC ([59–63]).
Similarly, the JRC NM nanomaterials differ in surface
modification and size or aspect ratio (refer to Table 1).
For instance TiO2 NM-103 and NM-104 have rutile
structure and are stabilized with polydimethylsiloxane
while NM-105 is a mixture of rutil and anatase and is
unmodified. For ZnO, NM-110 is uncoated, while
NM111 is triethoxycaprylsilane stabilized. The MWCNTs
differ mainly in their aspect ratio, with NM-401 being
the longest, followed by NM-402.
NPs were dispersed using different protocols for
reasons as follows. For all nanoGEM NPs a
stirringbased protocol was applied. These NPs were
hydrophilic and mostly carried chemical surface modifications,
which might be affected by ultrasonication. A stock
dispersion (2.5 mg/mL) in sterile complete cell culture
medium (CCM) consisting of DMEM medium (w/o
phenol red and L-glutamine, PAN Biotech GmbH)
supplemented with 10 % non-heat inactivated fetal calf serum
(FCS gold, PAA Laboratories, GmbH, Germany), 25 mM
HEPES buffer (PAN), 100 IU penicillin (PAN), 0.1 mg/mL
streptomycin (PAN) and 2 mM L-glutamine (PAN) was
prepared by stirring in sterile vials with a magnetic stir bar
at 700 rpm for 24 h. Final working dilutions (between 5
and 100 μg/mL) were prepared by diluting the stock
dispersion and stirring for another 1 h prior to
treatment of cells. Dispersions were prepared freshly for
each experiment. We used non-heat inactivated serum,
which was considered to be more relevant for
comparison to physiological situations.
MARINA NPs were dispersed using sonication-based
protocols because part of the NPs were hydrophobic and
could not be dispersed well by the stirring-based protocol
as determined in preliminary dispersion tests. Hydrophilic
NPs (SiO2 NM-200/203, Ag NM-300 k, TiO2 NM-104,
ZnO NM-110) were dispersed at 1 mg/ml in ddH2O by
sonication for 6 h in an ultrasonic bath (primary
dispersion) (Bandelin, Germany). Hydrophobic NPs (TiO2
NM103, ZnO NM-111) were pre-wetted with ethanol and
then dispersed in ddH2O containing 0.05 % BSA at
2.56 mg/ml by probe sonication with 15 % amplitude for
15 min (Bandelin, Germany). Note, that for the DCFDA
assay all MARINA NPs were dispersed according to the
protocol for hydrophobic NPs. MWCNTs were dispersed
at 1 mg/mL in 1 % (w/v) pluronic in ddH2O by first
stirring for 30 min and subsequent probe-sonication for
90 min as described elsewhere . Dispersions were
stored, but prior to treatment NPs were re-dispersed by
sonication for 30 min. Only after re-dispersion were NPs
added to CCM.
Characterization of NPs
Dynamic Light Scattering (DLS) and zeta potential
Sizes and size distribution in different dispersion media
were measured using a Zetasizer NanoZS (Malvern,
Germany) equipped with a He-Ne laser (633 nm).
Settings of attenuator and voltage were selected
automatically. For DLS measurements dispersions with a
concentration of 50 – 100 μg/mL were used. Three
independent experiments were performed, comprising of five
Surface reactivity analysis by ESR
ROS activity of NPs was determined by ESR spectroscopy
using two different methods. Employing the method of
Papageorgiou et al., that uses CPH
(1-hydroxy-3-carboxypyrrolidine) as spin probe, possible (surface) reactivity was
investigated . Additionally, employing the method of
Shi et al. , potential hydroxyl radical (OH ) formation
was determined in the presence of hydrogen peroxide
(H2O2) and the spin trap
5,5-dimethyl-1-pyrroline-Noxide (DMPO). The surface reactivity was calculated as
ratio between radical formation in the presence of the
NPs and the response of deionized water (dH2O) as
reference signal (Table 2). The suspensions were used as
delivered by the provider and have been diluted by a factor
of four (for DMPO analysis) or two (for CPH analysis),
respectively. For the solid materials we prepared a
suspension of 1 mg/mL meaning that a final concentration of
0.25 mg/mL (DMPO) and 0.5 mg/mL (CPH), respectively,
was analyzed. Supernatants were also used as delivered by
the provider. For the production of the supernatant
ultracentrifugation has been performed by providers
Based on ESR data for BaSO4 NM-220, which we
considered being a negative reference NP showing no
biological responses [67, 68], we defined ESR ratios of
>2.6 as positive. ESR values for BaSO4 NM-220 were
determined to be 2 and we included a 30 % uncertainty
resulting from the method. For some selected NPs we
also measured NP-depleted supernatants. For those we
compared the NP dispersion to the NP-depleted
controls and a difference of >2.6 was considered positive.
These assessment factors are specific to our system and
serve as a guideline only. There are no absolute
assessment criteria for ESR being published. Initial tests
showed no significant differences in the ROS activity of
selected materials after different dispersion methods,
e.g. rigorous mixing on a Vortex for 1 min, or stirring
for 1 h or 24 h. Consequently, 1 min rigorous mixing on
a Vortex was used as dispersion procedure prior to ESR
Based on previous work using BCR 723 Road Dust
no significant interferences with suspended particles
were expected. Note that ESR method is not an
optical method, which are usually more susceptible to
NRK-52E cells (DSMZ, Braunschweig, Germany) were
grown in complete cell culture medium (CCM) consisting
of DMEM medium (w/o phenol red and L-glutamine,
high glucose, PAN Biotech GmbH) supplemented with
10 % non-heat inactivated fetal calf serum (FCS gold, PAA
Laboratories, GmbH, Germany), 25 mM HEPES buffer
(PAN), 100 IU penicillin (PAN), 0.1 mg/mL streptomycin
(PAN) and 2 mM L-glutamine (PAN). Note that ten
different cell lines were tested initially and among them
NRK52E was the most responsive epithelial cell line to induce
NP-mediated oxidative stress.
Cells were split regularly at ~95 % confluence. For
analyzing cell viability with WST-1 assay cells were
seeded at 1.25 × 104 cells/well in 96-well plates. For the
DCFDA assay cells were seeded in 96-well plates at a
density of 3 × 104 cells/well. For the protein carbonyl
assays, cells were seeded at 1.5 × 106 cells/well in 6-well
plates (1D screen) and for the 2D-based proteomic study
cells were treated identically but lysates of 6 wells were
pooled for every treatment.
Cells were treated in 96-well plates for 24 h with
NPs in concentrations ranging from 0 to 100 μg/ml
(0–30 μg/cm2). WST-1 reagent was obtained from
Roche (Germany) and used according to manufacturer
instructions. Before read-out, supernatants were
centrifuged at 18,500 × g for 25 min to remove interfering NPs
and then measured at 450 nm using a Tecan® (Austria)
plate reader. For MWCNTs the centrifugation time was
extended to 45 min. All NPs were tested for interference
with the assay. To that end we incubated NPs with fully
reacted WST-1 reagent (but without cells) and centrifuged
as described above to assure that there were no NP
NP concentrations for in vitro tests were chosen to
correlate to relevant in vivo test concentrations.
Inhalation overload dose should correspond to 1–10 μg/cm2
in in vitro testing, rendering this concentration range
the preferred test concentration for in vitro testing.
Higher doses were tested in addition for completeness.
The DCFDA assay was performed as described
previously . Briefly, 24 h after plating, cells were exposed
to NPs for 1 h in a concentration range from 3.4 μg/mL
to 34 μg/mL (1–10 μg/cm2). For most NPs higher NP
concentrations could not be tested due to the onset of
NP interferences, which had been evaluated beforehand.
ZnO, TiO2, NM-200 and NM-203 and MWCNTs
interfered less and could be tested at concentrations of up to
170 μg/ml (50 μg/cm2). Subsequently, cells were washed
twice with Krebs-Ringer Buffer (KRB), incubated with
H2DCFDA-DA (5 μM in KRB) for 45 min, and washed
again twice before monitoring the fluorescence signal
(excitation 485 nm, emission 520 nm; FLUOstar, BMG
Labtech GmbH, Offenburg, Germany). Each experiment
was repeated at least 3 times with 6 or 7 replicates each.
To test for interference we incubated NP dispersions
(1–100 μg/cm2) in the presence of oxidized fluorescent
DCF, both with and without cells present.
Concentrations with significant changes from the control were
excluded from testing.
Analysis of protein carbonyls via 1D immunoblot
In vitro samples: NRK-52E cells were treated with NP
concentrations between 5 and 100 μg/ml (1.7-34 μg/cm2)
for 6 h (using four different concentrations per NP, mostly
below cytotoxic level). Optimal treatment times were
determined at the beginning of our study by testing different
time points (Additional file 1: Figure S1). Where
applicable, NP-depleted supernatants served as a control in
addition to untreated cells. Cells were lysed in a modified
RIPA-Buffer (50 mM Tris/HCl pH 7.4, 150 mM NaCl,
1 mM EDTA, 1 % Igepal, 0.25 % Na-deoxycholate)
containing protease inhibitors (Protease Inhibitor Cocktail Set
III, Millipore, Germany) and 2-mercaptoethanol (1 % v/v)
at 4 °C for 45 min, followed by centrifugation (4 °C,
17,000 × g, 15 min). Each analysis was performed using at
least three independent biological repeats.
In vivo samples: Rat lung tissue was kept at −80 °C
after explantation. The lung tissue was pulverized to a
fine powder at ~ −180 °C (under Liquid Nitrogen) using
a mortar and a pestle and subsequently extracted using
RIPA buffer as described above. Samples of four animals
per treatment were analyzed.
In all samples protein concentrations were determined
using BioRad® Bradford assay (BioRad, Germany)
according to manufacturer instructions. Protein carbonyls were
detected using the Millipore Oxiblot® kit (Millipore,
Germany) according to manufacturer instructions using
10–25 μg protein lysates. Briefly, proteins were separated
by 10 % SDS-PAGE and transferred to nitrocellulose
membranes using a semi-dry transfer (200 mW/Gel, 1 h).
Membranes were blocked with Rotiblock® (Carl-Roth,
Germany), and incubated with primary antibodies
(1:300 in Rotiblock, from the Oxiblot® kit) overnight at 4 °
C. Membranes were washed three times using TBST,
incubated with secondary antibodies (1:150 in TBST, from
the Oxiblot® kit) for 1 h, washed again 3 times with TBST
and visualized using ECL (Thermo Scientific Pierce,
SuperSignal West Pico ECL). Chemiluminescence was
detected using a GelDoc (BioRad, Germany). Analysis was
performed with ImageLab software (BioRad, Germany).
For normalization the tubulin signal of the same
membrane was used. The tubulin antibody was obtained from
Abcam (via NEB, Germany), and used in a 1:5000 dilution
Analysis of protein carbonyls via 2D immunoblot
Cells were treated with 10 μg/ml (3.4 μg/cm ) NP for 6 h
and lysed in 600 μl 2D lysis buffer (7 M urea, 2 M
thiourea, 4 % Chaps, 2 % IPG buffer pH 3 –10, 1 % DTT) at
4 °C for 45 min. 350 μg of the protein lysate were used for
every analysis. For mass spectrometric identification we
used a separate duplicate 2D gel. For isoelectric focusing
(IEF), 24 cm IPG strips (GE Healthcare, Germany) with a
nonlinear pH gradient of pH 3 –10 were used. Active
rehydration and focusing was performed in 6 steps (15 h at
30 V, 1.5 h at 200 V, 1 h at 500 V, a 13.5 h gradient from
500 –1000 V, a 3 h gradient from 1000 –8000 V, and 6 h
at 8000 V). After IEF, protein carbonyls were
derivatized in the IEF strip by incubation with 10 mM
2,4dinitrophenylhydrazine (DNPH) in 2 M HCl at RT for
15 min. Excessive reagent was removed by three washes
of 2 M tris (hydroxymethyl) aminomethane solution in
30 % glycerol/ddH2O for 10 min each. Subsequently,
strips were washed twice with electrophoresis buffer for
10 min. Samples were reduced with DTT and alkylated
using iodoacetamide according to standard protocols prior
to separation in the second dimension. For second
dimension separation 12.5 % SDS-PAGE gels were used
with 3 W per gel for 1.5 h followed by 15 W per gel.
Proteins were transferred onto a nitrocellulose membrane
using wet-transfer (50 mA, 16 h). Membranes were
blocked with Rotiblock® (Carl-Roth, Germany). Carbonyls
were detected using DNBP antibody (Sigma, Germany) at
1:500 in Rotiblock® at 4 °C over night. After three washes,
a goat anti-rabbit antibody (Dianova, Germany) was used
at 1:10,000 in TBST for 2 h at room temperature. For
visualization by ECL a GelDoc imager (BioRad) was used.
In parallel, each sample was separated on another
duplicate 2D gel, which was stained with ruthenium
II-bathophenanthroline disulfonate chelate as described
elsewhere . Gels were scanned using a FLA9500 laser
scanner (GE Healthcare, Germany), using an excitation at
λ = 473 nm and detection at λ = 610 nm. Duplicate gels
were used for spot picking and mass spectrometric
Analysis of 2D gels and 2D immunoblots was
performed using Delta 2D software (Decodon, Germany).
A carbonyl spot was considered relevant if it was
detected in 2 out of 3 biological repeats (for NP
samples) or in at least 4 out of 6 repeats for control gels.
All NP samples were analyzed in at least three
independent biological replicates, controls in 6 independent
Identification of protein by MALDI-TOF/TOF
Spots were excised from the fluorescent stained gels
using a spot picker (Proteome Factory, Germany) and
digested in-gel with trypsin using a standard protocol. In
short, gel spots were incubated with 0.03 μg trypsin in
50 mM ammonium bicarbonate in 95/5 H2O/acetonitrile
over night at 37 °C for digestion. Peptides were extracted
from the gel matrix by subsequent extraction with 60 %
acetonitrile (ACN)/0.1 % trifluoroacetic acid and 100 %
ACN. Digested samples were dried in a centrifugal
evaporator, redispersed in 1 % TFA and purified using a C18
ZipTip (Millipore, Germany). Samples were spotted with
α-cyano-4-hydroxy-cinnamic acid (HCCA) matrix on
AnchorChip targets (384/800) (Bruker, Germany) and
measured using an UltrafleXtreme MALDI-TOF/TOF
(Bruker, Germany) with FlexControl and FlexAnalysis
software. Data were evaluated using Proteinscape (via
MASCOT using Swissprot database). The following
search parameters were used: 1 partial cleavage site,
carbamidomethyl (Cys) as fixed modification and
oxidation (Met) as variable modifications. Taxonomy was
rattus, MS tolerance was 50 ppm and MS/MS tolerance
was 0.7 Da. Proteins were considered reliably
identified if the MASCOT score was above threshold and
in addition at least 2 independent unique MS/MS
spectra were identified. Proteins identified by a
MASCOT score with only 1 unique MS/MS spectrum were
considered as not reliably identified and those were
marked as ‘tbc’ (Additional file 1: Table S1).
For pathway analyses we used Ingenuity Pathway Analysis
software (IPATM, Qiagen, Germany). Data sets of
identified proteins were subjected to core analyses against the
IPA knowledge base (genes only data base version) for
protein enrichment in canonical pathways.
For each spot in each treatment median values were
derived from the spot intensities of the 2D carbonyl
immunoblots as obtained by Delta 2D analysis. Median values
of all carbonyl spots (202 spots) were analyzed using the
statistical computing environment R [R Development
Core Team, 2012]. Some spots appeared only in the
treated samples and were absent in control samples. If
spots appeared both in treated samples as well as in
controls, spots were considered for analysis that were
significantly induced compared to controls (intensities
at least 2-fold over controls, p < 0.05, students t-test, in
total 175 spots). We performed a hierarchical cluster
analysis (HCA) using Euclidian complete linkage. The
basic idea of HCA is to group a given data set step by
step into nested clusters by using a distance measure.
Here we used the spot numbers of carbonylated proteins
together with median intensities resulting from
minimum three biological repeats to analyze distances
between the different NP treatment groups. We employed
Euclidian complete linkage as one of the many possible
methods, which uses a Euclidian distance measure and
always uses the largest possible distance between two
clusters of data points for analysis.
In addition we performed principal component
analysis (PCA) using SIMCA software (Umetrics, Sweden)
using 5 main principal components, which could explain
more than 65 % of data variance. The objective of PCA
is to project data points from a higher into a lower
dimensional space with minimal loss of information. It
visualizes variance in a data set by introducing new axes
(called principle components) such that the first axis lies
in the direction of greatest variation. Each subsequent
principal component (PC) describes the largest variance
in a direction orthogonally to the previous PC, thus
resulting in linearly uncorrelated PCs.
Pathogen-free female rats (Wistar strain WU)
weighing 200–250 g were purchased from Charles River
Laboratories (Sulzfeld, Germany) and maintained with
a 12 h lights-on lights-off cycle. Food and water were
provided ad libitum. For intratracheal instillation
animals were briefly anesthetized using isoflurane. Animal
experiments were approved by the local regulatory
agencies. All study protocols complied with the federal
guidelines. An Ag 50 PVP stock dispersion was diluted in
0.9 % NaCl (which also served as control treatment
without NPs) and indicated doses were intratracheally
instilled using a Penn Century Microsprayer that had been
inserted into the trachea under visual control. On the day
of sacrifice, rats were deeply anesthetized with a mixture
of ketamine and xylazine. The trachea was cannulated and
the left lung filled with Cryomatrix (Thermo Shandon
Ltd., Runcorn, UK): A tissue block (0.5 – 1 cm3) from the
medial hilar region was cut with razor blade, snap frozen
in liquid nitrogen, and stored at −80 °C.
Additional file 1: Figure S1-S5 and Table S1. The file contains various
additional figures and tables, collected in a single portable documend file
(PDF) (DOC 9935 kb).
AB and MDD performed analysis of NPs by DLS, contributed to cytotoxicity
testing of NPs and analyzed protein carbonylation via 1D and 2D gels.
MALDI-MS/MS identification as well as the pathway analysis was done by AB,
MDD and AH. CR, BR and VV performed the statistical data analysis. MW and
AV conducted the animal experiments. SM, RO and JS performed DCFDA
assays. TAJK and BH analyzed NPs by ESR. The study was planned by AH
with contributions from AL, TAJK, MW and JS. The manuscript was written
by MDD, CR and AH with contributions from all co-authors.
The authors also thank all institutions for supporting this project. In particular
we would also like to acknowledge support from Dr. Ralf Nehring and the
State Ministry for Environment, Agriculture, Food, Viniculture and Forests,
Rhineland-Palatinate, Germany for a part of this study. We thank D. Wittke
(BfR), N. Dommershausen (BfR) and J. Tharmann (BfR) for excellent technical
support. The authors also thank Wendel Wohlleben for supporting this analysis
and for critical comments on the manuscript. In addition, the authors would like
to acknowledge support from the Joint Research Centre (JRC) by providing NPs
used in this study.
We thank the German BMBF for funding the project nanoGEM (03X0105)
and the EC for funding from EU FP7 project MARINA (263215).
1. Hankin S , Boraschi D , Duschl A , Lehr C-M , Lichtenbeld H. Towards nanotechnology regulation - Publish the unpublishable . Nano Today . 2011 ; 6 : 228 - 31 .
2. Nel A , Xia T , Madler L , Li N. Toxic potential of materials at the nanolevel . Science . 2006 ; 311 : 622 - 7 .
3. Oberdorster G , Maynard A , Donaldson K , Castranova V , Fitzpatrick J , Ausman K , et al. Principles for characterizing the potential human health effects from exposure to nanomaterials: elements of a screening strategy . Part Fibre Toxicol . 2005 ; 2 : 8 .
4. McShan D , Ray PC , Yu H. Molecular toxicity mechanism of nanosilver . J Food Drug Anal . 2014 ; 22 : 116 - 27 .
5. Meng H , Xia T , George S , Nel AE. A predictive toxicological paradigm for the safety assessment of nanomaterials . ACS Nano . 2009 ; 3 : 1620 - 7 .
6. Sayes CM , Reed KL , Warheit DB . Assessing toxicity of fine and nanoparticles: comparing in vitro measurements to in vivo pulmonary toxicity profiles . Toxicol Sci . 2007 ; 97 : 163 - 80 .
7. Sayes CM , Wahi R , Kurian PA , Liu Y , West JL , Ausman KD , et al. Correlating nanoscale titania structure with toxicity: a cytotoxicity and inflammatory response study with human dermal fibroblasts and human lung epithelial cells . Toxicol Sci . 2006 ; 92 : 174 - 85 .
8. Xu A , Chai Y , Nohmi T , Hei TK . Genotoxic responses to titanium dioxide nanoparticles and fullerene in gpt delta transgenic MEF cells . Part Fibre Toxicol . 2009 ; 6 : 3 .
9. Halliwell B , Whiteman M. Measuring reactive species and oxidative damage in vivo and in cell culture: how should you do it and what do the results mean? Br J Pharmacol . 2004 ; 142 : 231 - 55 .
10. Reczek CR , Chandel NS . ROS-dependent signal transduction . Curr Opin Cell Biol . 2014 ; 33 : 8 - 13 .
11. Xia T , Kovochich M , Brant J , Hotze M , Sempf J , Oberley T , et al. Comparison of the abilities of ambient and manufactured nanoparticles to induce cellular toxicity according to an oxidative stress paradigm . Nano Lett . 2006 ; 6 : 1794 - 807 .
12. Forman HJ , Torres M. Reactive oxygen species and cell signaling: respiratory burst in macrophage signaling . Am J Respir Crit Care Med . 2002 ; 166 : S4 - 8 .
13. Babior BM . Phagocytes and oxidative stress . Am J Med . 2000 ; 109 : 33 - 44 .
14. Prietl B , Meindl C , Roblegg E , Pieber TR , Lanzer G , Frohlich E. Nano-sized and micro-sized polystyrene particles affect phagocyte function . Cell Biol Toxicol . 2014 ; 30 : 1 - 16 .
15. Cruz T , Gaspar R , Donato A , Lopes C. Interaction between polyalkylcyanoacrylate nanoparticles and peritoneal macrophages: MTT metabolism, NBT reduction, and NO production. Pharm Res . 1997 ; 14 : 73 - 9 .
16. Mari M , Morales A , Colell A , Garcia-Ruiz C , Kaplowitz N , Fernandez-Checa JC . Mitochondrial glutathione: features, regulation and role in disease . Biochim Biophys Acta . 1830 ; 2013 : 3317 - 28 .
17. Hussain SP , Hofseth LJ , Harris CC . Radical causes of cancer . Nat Rev Cancer . 2003 ; 3 : 276 - 85 .
18. He W , Liu Y , Wamer WG , Yin JJ . Electron spin resonance spectroscopy for the study of nanomaterial-mediated generation of reactive oxygen species . J Food Drug Anal . 2014 ; 22 : 49 - 63 .
19. Aggarwal BB , Shishodia S , Sandur SK , Pandey MK , Sethi G. Inflammation and cancer: how hot is the link ? Biochem Pharmacol . 2006 ; 72 : 1605 - 21 .
20. Dalle-Donne I , Aldini G , Carini M , Colombo R , Rossi R , Milzani A. Protein carbonylation, cellular dysfunction, and disease progression . J Cell Mol Med . 2006 ; 10 : 389 - 406 .
21. Davies MJ . The oxidative environment and protein damage . Biochimica et Biophysica Acta (BBA) - Proteins and Proteomics . 2005 ; 1703 : 93 - 109 .
22. Stadtman ER . Protein oxidation and aging . Free Radic Res . 2006 ; 40 : 1250 - 8 .
23. Wong CM , Marcocci L , Liu L , Suzuki YJ . Cell signaling by protein carbonylation and decarbonylation . Antioxid Redox Signal . 2010 ; 12 : 393 - 404 .
24. Madian AG , Regnier FE . Proteomic identification of carbonylated proteins and their oxidation sites . J Proteome Res . 2010 ; 9 : 3766 - 80 .
25. Butterfield DA , Perluigi M , Reed T , Muharib T , Hughes CP , Robinson RA , et al. Redox proteomics in selected neurodegenerative disorders: from its infancy to future applications . Antioxid Redox Signal . 2012 ; 17 : 1610 - 55 .
26. Dalle-Donne I , Giustarini D , Colombo R , Rossi R , Milzani A. Protein carbonylation in human diseases . Trends Mol Med . 2003 ; 9 : 169 - 76 .
27. Frohnert BI , Sinaiko AR , Serrot FJ , Foncea RE , Moran A , Ikramuddin S , et al. Increased adipose protein carbonylation in human obesity . Obesity (Silver Spring) . 2011 ; 19 : 1735 - 41 .
28. Ergin V , Hariry RE , Karasu C. Carbonyl stress in aging process: role of vitamins and phytochemicals as redox regulators . Aging Dis . 2013 ; 4 : 276 - 94 .
29. Cabiscol E , Tamarit J , Ros J. Protein carbonylation: proteomics, specificity and relevance to aging . Mass Spectrom Rev . 2014 ; 33 : 21 - 48 .
30. Dorts J , Silvestre F , Tu HT , Tyberghein AE , Phuong NT , Kestemont P. Oxidative stress, protein carbonylation and heat shock proteins in the black tiger shrimp, Penaeus monodon, following exposure to endosulfan and deltamethrin . Environ Toxicol Pharmacol . 2009 ; 28 : 302 - 10 .
31. Bruno M , Moore T , Nesnow S , Ge Y. Protein carbonyl formation in response to propiconazole-induced oxidative stress . J Proteome Res . 2009 ; 8 : 2070 - 8 .
32. Newton BW , Russell WK , Russell DH , Ramaiah SK , Jayaraman A. Liver proteome analysis in a rodent model of alcoholic steatosis . J Proteome Res . 2009 ; 8 : 1663 - 71 .
33. Dowling V , Hoarau PC , Romeo M , O'Halloran J , van Pelt F , O'Brien N , et al. Protein carbonylation and heat shock response in Ruditapes decussatus following p, p'-dichlorodiphenyldichloroethylene (DDE) exposure: a proteomic approach reveals that DDE causes oxidative stress . Aquat Toxicol . 2006 ; 77 : 11 - 8 .
34. Hu W , Culloty S , Darmody G , Lynch S , Davenport J , Ramirez-Garcia S , et al. Toxicity of copper oxide nanoparticles in the blue mussel, Mytilus edulis: a redox proteomic investigation . Chemosphere . 2014 ; 108 : 289 - 99 .
35. Ching B , Chew SF , Wong WP , Ip YK . Environmental ammonia exposure induces oxidative stress in gills and brain of Boleophthalmus boddarti (mudskipper) . Aquat Toxicol . 2009 ; 95 : 203 - 12 .
36. Almroth BC , Sturve J , Stephensen E , Holth TF , Forlin L. Protein carbonyls and antioxidant defenses in corkwing wrasse (Symphodus melops) from a heavy metal polluted and a PAH polluted site . Mar Environ Res . 2008 ; 66 : 271 - 7 .
37. Rainville LC , Carolan D , Varela AC , Doyle H , Sheehan D. Proteomic evaluation of citrate-coated silver nanoparticles toxicity in Daphnia magna . Analyst . 2014 ; 139 : 1678 - 86 .
38. Levine RL , Williams JA , Stadtman ER , Shacter E. Carbonyl assays for determination of oxidatively modified proteins . Methods Enzymol . 1994 ; 233 : 346 - 57 .
39. Nakamura A , Goto S. Analysis of protein carbonyls with 2,4-dinitrophenyl hydrazine and its antibodies by immunoblot in two-dimensional gel electrophoresis . J Biochem . 1996 ; 119 : 768 - 74 .
40. Robinson CE , Keshavarzian A , Pasco DS , Frommel TO , Winship DH , Holmes EW . Determination of protein carbonyl groups by immunoblotting . Anal Biochem . 1999 ; 266 : 48 - 57 .
41. Grimsrud PA , Picklo Sr MJ , Griffin TJ , Bernlohr DA . Carbonylation of adipose proteins in obesity and insulin resistance: identification of adipocyte fatty acid-binding protein as a cellular target of 4-hydroxynonenal . Mol Cell Proteomics . 2007 ; 6 : 624 - 37 .
42. Haase A , Rott S , Mantion A , Graf P , Plendl J , Thunemann AF , et al. Effects of silver nanoparticles on primary mixed neural cell cultures: uptake, oxidative stress and acute calcium responses . Toxicol Sci . 2012 ; 126 : 457 - 68 .
43. Haase A , Arlinghaus HF , Tentschert J , Jungnickel H , Graf P , Mantion A , et al. Application of laser postionization secondary neutral mass spectrometry/ time-of-flight secondary ion mass spectrometry in nanotoxicology: visualization of nanosilver in human macrophages and cellular responses . ACS Nano . 2011 ; 5 : 3059 - 68 .
44. Verano-Braga T , Miethling-Graff R , Wojdyla K , Rogowska-Wrzesinska A , Brewer JR , Erdmann H , et al. Insights into the cellular response triggered by silver nanoparticles using quantitative proteomics . ACS Nano . 2014 ; 8 : 2161 - 75 .
45. Castro JP , Jung T , Grune T , Almeida H. Actin carbonylation: from cell dysfunction to organism disorder . J Proteomics . 2013 ; 92 : 171 - 80 .
46. McDonagh B , Tyther R , Sheehan D. Carbonylation and glutathionylation of proteins in the blue mussel Mytilus edulis detected by proteomic analysis and Western blotting: Actin as a target for oxidative stress . Aquat Toxicol . 2005 ; 73 : 315 - 26 .
47. Keshavarzian A , Banan A , Farhadi A , Komanduri S , Mutlu E , Zhang Y , et al. Increases in free radicals and cytoskeletal protein oxidation and nitration in the colon of patients with inflammatory bowel disease . Gut . 2003 ; 52 : 720 - 8 .
48. England K , O'Driscoll C , Cotter TG . Carbonylation of glycolytic proteins is a key response to drug-induced oxidative stress and apoptosis . Cell Death Differ . 2004 ; 11 : 252 - 60 .
49. Carlson C , Hussain SM , Schrand AM , Braydich-Stolle LK , Hess KL , Jones RL , et al. Unique cellular interaction of silver nanoparticles: size-dependent generation of reactive oxygen species . J Phys Chem B . 2008 ; 112 : 13608 - 19 .
50. Prasad RY , McGee JK , Killius MG , Suarez DA , Blackman CF , DeMarini DM , et al. Investigating oxidative stress and inflammatory responses elicited by silver nanoparticles using high-throughput reporter genes in HepG2 cells: effect of size, surface coating, and intracellular uptake . Toxicol In Vitro. 2013 ; 27 : 2013 - 21 .
51. Li Y , Zhang W , Niu J , Chen Y. Surface-coating-dependent dissolution, aggregation, and reactive oxygen species (ROS) generation of silver nanoparticles under different irradiation conditions . Environ Sci Technol . 2013 ; 47 : 10293 - 301 .
52. Yoshida T , Yoshioka Y , Matsuyama K , Nakazato Y , Tochigi S , Hirai T , et al. Surface modification of amorphous nanosilica particles suppresses nanosilica-induced cytotoxicity, ROS generation, and DNA damage in various mammalian cells . Biochem Biophys Res Commun . 2012 ; 427 : 748 - 52 .
53. Chen Q , Xue Y , Sun J. Kupffer cell-mediated hepatic injury induced by silica nanoparticles in vitro and in vivo. Int J Nanomedicine . 2013 ; 8 : 1129 - 40 .
54. Imai S , Yoshioka Y , Morishita Y , Yoshida T , Uji M , Nagano K , et al. Size and surface modification of amorphous silica particles determine their effects on the activity of human CYP3A4 in vitro . Nanoscale Res Lett . 2014 ; 9 : 651 .
55. Shrivastava R , Kushwaha P , Bhutia YC , Flora S. Oxidative stress induced following exposure to silver and gold nanoparticles in mice . Toxicol Ind Health . 2014 .
56. Landsiedel R , Sauer UG , Ma-Hock L , Schnekenburger J , Wiemann M. Pulmonary toxicity of nanomaterials: a critical comparison of published in vitro assays and in vivo inhalation or instillation studies . Nanomedicine (Lond) . 2014 ; 9 : 2557 - 85 .
57. Yin J-J , Liu J , Ehrenshaft M , Roberts JE , Fu PP , Mason RP , et al. Phototoxicity of nano titanium dioxides in HaCaT keratinocytes-Generation of reactive oxygen species and cell damage . Toxicol Appl Pharmacol . 2012 ; 263 : 81 - 8 .
58. Auffan M , Pedeutour M , Rose J , Masion A , Ziarelli F , Borschneck D , et al. Structural Degradation at the Surface of a TiO2-Based Nanomaterial Used in Cosmetics . Environ Sci Technol . 2010 ; 44 : 2689 - 94 .
59. Cotogno G , Gibson P , Per Axel Clausen , Pieter-Jan de Temmerman, Signe Hjortkjaer Nielsen , Ceccone G , et al. Synthetic Amorphous Silicon Dioxide (NM-200, NM-201, NM-202, NM-203, NM-204): Characterisation and PhysicoChemical Properties In: NM-Series of Representative Manufactured Nanomaterials. Ispra, Italy: Joint Research Centre of the European Commission ; 2013 .
60. Singh C , Friedrichs S , Levin M , Birkedal R , Jensen KA , Pojana G , et al. Zinc Oxide NM-110, NM-111, NM-112, NM-113: Characterisation and Test Item Preparation In: NM-Series of Representative Manufactured Nanomaterials. Ispra, Italy: Joint Research Centre of the European Commission ; 2011 .
61. Comero S , Klein C , Stahlmecke B , Romazanov J , Kuhlbusch T , Doren EV , et al. NM-300 Silver Characterisation , Stability, Homogeneity In: NM-Series of Representative Manufactured Nanomaterials. Joint Research Centre of the European Commission ; 2011 .
62. Rasmussen K , Mast J , Temmerman P- Jd , Verleysen E , Waegeneers N , Steen Fv , et al. Multi-walled Carbon Nanotubes, NM-400, NM-401, NM-402, NM-403: Characterisation and Physico-Chemical Properties In: NM-Series of Representative Manufactured Nanomaterials. Ispra, Italy: Joint Research Centre of the European Commission ; 2014 .
63. Cotogno G , Ceccone G , Gilliand D , Nihtianova D , Nikolova R , Pianella F , et al. Titanium Dioxide, NM-100, NM-101, NM-102, NM-103, NM-104, NM-105: Characterisation and Physico-Chemical Properties In: NM-Series of Representative Manufactured Nanomaterials. Ispra, Italy: Joint Research Centre of the European Commission ; 2014 .
64. Piret JP , Detriche S , Vigneron R , Vankoningsloo S , Rolin S , Mejia Mendoza JH , et al. Dispersion of multi-walled carbon nanotubes in biocompatible dispersants . J Nanoparticle Res . 2010 ; 12 : 75 - 82 .
65. Papageorgiou I , Brown C , Schins R , Singh S , Newson R , Davis S , et al. The effect of nano- and micron-sized particles of cobalt-chromium alloy on human fibroblasts in vitro . Biomaterials . 2007 ; 28 : 2946 - 58 .
66. Shi T , Schins RPF , Knaapen AM , Kuhlbusch T , Pitz M , Heinrich J , et al. Hydroxyl radical generation by electron paramagnetic resonance as a new method to monitor ambient particulate matter composition . J Environ Monit . 2003 ; 5 : 550 - 6 .
67. Buesen R , Landsiedel R , Sauer UG , Wohlleben W , Groeters S , Strauss V , et al. Effects of SiO2 , ZrO 2 , and BaSO 4 nanomaterials with or without surface functionalization upon 28-day oral exposure to rats . Arch Toxicol . 2014 ; 88 : 1881 - 906 .
68. Landsiedel R , Ma-Hock L , Hofmann T , Wiemann M , Strauss V , Treumann S , et al. Application of short-term inhalation studies to assess the inhalation toxicity of nanomaterials . Part Fibre Toxicol . 2014 ; 11 : 16 .
69. Kroll A , Dierker C , Rommel C , Hahn D , Wohlleben W , Schulze-Isfort C , et al. Cytotoxicity screening of 23 engineered nanomaterials using a test matrix of ten cell lines and three different assays . Part Fibre Toxicol . 2011 ; 8 : 9 .
70. Rabilloud T. Detecting proteins separated by 2-D gel electrophoresis . Anal Chem . 2000 ; 72 : 48A - 55A .
71. Farcal L , Torres Andon F , Di Cristo L , Rotoli BM , Bussolati O , Bergamaschi E , et al. Comprehensive in vitro toxicity testing of a panel of representative oxide nanomaterials: first steps towards an intelligent testing strategy . PLoS One . 2015 ; 10 , e0127174.