A general mechanism for intracellular toxicity of metal-containing nanoparticles† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c4nr01234h Click here for additional data file.

We demonstrate a general mechanism for the toxicity induced by metal-containing NPs, named “lysosome-enhanced Trojan horse effect”, which provides design rules to engineer safer NPs.

. Behavior of CdSe/ZnS QDs in neutral (pH 7, 37 °C) and acidic conditions (pH 4.5, 37 °C) over time. (A) Photoluminescence spectra (the reference PL spectrum at time 0 completely overlaps with the blue curve and is not shown for clarity); (B) Absorption spectra; (C) DLS measurement of QDs after 96 h incubation in acidic conditions. As shown, CdSe/ZnS QDs, when incubated in acidic conditions, gradually but significantly lose their fluorescence property, due to particle core degradation, unlike QDs in neutral environment that, after 96 h, completely maintain their original photo-physical properties (in terms of both PL spectrum and emission intensity). Such behavior is also confirmed by absorption measurements, revealing that incubation in the acidic environment results in the loss of the QDs excitonic peak. Furthermore, the QDs in the acidic conditions degrade and became significantly less stable in solution, showing significant precipitation and presence of aggregates (panel C). It is evident that the lysosomal-like environment strongly promotes NP corrosion and ion release, unlike neutral (cytoplasmic-like) conditions. In particular, TEM imaging clearly reveals that incubation in acidic conditions results in strong NP degradation, with the formation of smaller nanoparticles (see panels C and D). The presence of some aggregates is also visible. AgNP degradation in acidic conditions is revealed also by DLS analyses, which show significant NP aggregation in solution. It should be clarified that the findings of the ion release experiments are that NP degradation and ion release are much faster in acidic conditions than in neutral conditions, namely the acidic (lysosomal-like) conditions strongly promote NP dissolution. This means that it may be possible to detect ion release also in neutral conditions (for instance, by increasing the NP concentration or the detection sensitivity) but, in any case, the acidic environment will promote significantly higher release. For example, in the case of AgNPs, there is indeed some ion release also in neutral conditions (this is the basis of their antibacterial behavior), but in acidic conditions such process is significantly higher 1 .  In the case of gold, the NP corrosion and ion release is lower (see also Fig. 1), so it can be appreciated only as a significant loss of their stability in solution with consequent NP aggregation. Such phenomenon is evident for both striped and unstructured AuNPs in both DLS and absorption experiments. Fig. S5. Time-dependent gold ion release from striped and unstructured AuNPs (50 nM), probed by ICP-AES, at 37 °C, at neutral (pH 7.0, blue symbols) or acidic (pH 4.5, red symbols) conditions. Data represent the average from 3 independent measurements (6 replicates for each experiment) and the error bars indicate the standard deviation. The two AuNPs release comparable amounts of gold ions in acidic conditions, while at neutral pH the gold ion release was not detectable. . ICP-AES analyses of the amount of AuNPs internalized by U937 and HeLa cells (normalized to the control at 37 °C) in the presence of specific inhibitors of endocytic processes, namely 4 °C, sodium azide/2-deoxyglucose (NaN3/DoxG), Cytochalasin D (CytD), confirming that the striped AuNPs are able to penetrate the cells regardless of the activation/inhibition of the active energy-dependent internalization processes, unlike unstructured AuNPs.       Experimental points represent the average from 5 independent experiments and the error bars (reported as diameter of points) indicate the standard deviation. The lifespan curves were validated by the non-parametric log-rank (Mantel-Cox) test followed by Gehan-Breslow-Wilcoxom post-test (CTRL vs unstructured AuNPs: p<0.0001; CTRL vs striped AuNPs: not significant (ns)). A significant lifespan reduction was induced by the treatment with the unstructured particles ( 50 was ca. 47% lower than the control), as opposed to striped particles, despite a similar bioaccumulation of the two AuNPs in the organisms (ca. 0.25 pg of gold/organism were detected by ICP-AES after 30 days of treatment with both AuNPs).  Results are mean ± SD and differences between treated samples and controls (n=8) were considered statistically significant for *P<0.05, non significant for P>0.05. Results are mean ± SD and differences between treated samples and controls (n=8) were considered statistically significant for *P<0.05, non significant for P>0.05.

Mammalian thioredoxin reductase (TrxR): interactions with gold based compounds
The pharmaceutical chemistry of gold ions and of their relative complexes used as therapeutic agents has been extensively studied due to their extensive pharmacological use as anti-rheumatic drugs 2 . Moreover, many gold-based drugs (including solganol, myocrisin, auranofin, etc.) have recently gained great attention as very potent and specific blockers of the enzyme thioredoxin reductase (TrxR) 3 . Interestingly, due to the principal function of TrxRs family to reduce oxidized thioredoxin (Trx) 4,5 , the blocking of this enzyme impacts the regulation of many cell functions under Trxs control, such as regulation of the redox signaling pathways, antioxidant activity (by removing the hydrogen peroxides through peroxiredoxin), control of transcription factors binding, and inhibition of apoptosis 6 . In particular, when mitochondrial TrxR is damaged, the thiol redox balance is altered, thus triggering the apoptosis via swelling or improved permeabilization of the mitochondrial outer membrane 7,8 . Such event is the "point of no return", since it primes the cascade release of many pro-apoptotic signaling molecules, including cytochrome c and caspases 9,10 .
Moreover, the TrxR inactivation does not lead to the normal recycling activity of ROS (e.g., the hydrogen peroxide mainly formed by the mitochondrial respiratory chain 6 thus contributing to an overall alteration of the ROS basal levels in the cells 11 .  It is clear that, while BAL is capable to strongly reduce the toxicity of QDs, the unspecific chelator dfx is completely ineffective, showing ROS levels comparable to those elicited by the QDs alone. ROS quantification was performed via DCFH-DA assay. HeLa cells were pretreated for 30 min with/without 100 mM dfx or 1 mM BAL and then exposed to CdSe/ZnS QDs for 24/48 h. CTRL represents the negative control; values are means+s.d. Differences between treated samples and controls (n=8) were considered statistically significant for *P<0.05, non-significant for P>0.05. The treatment with chelating agents is the same described in Fig. 5 in the main text. In all cases, the pretreatment with chelating agents suppresses almost totally the toxicity of the NPs. CTRL represents the negative control; values are means+s.d. Differences between treated samples and controls (n=8) were considered statistically significant for *P<0.05, non-significant for P>0.05. It is clear that the treatment with the chelating agent significantly reduced the NPs toxicity. HeLa cells were pretreated for 30 min with/without 100 M DMSA and then exposed to the NPs for 24 or 48 h. Ctrl represents the negative control; values are means+s.d. Positive controls (not shown) were treated with 0.01% of TritonX100, displaying a strong viability decrease (ca. 80-90%) with respect to the untreated cells. CTRL represents the negative control; values are means+s.d. Differences between treated samples and controls (n=8) were considered statistically significant for *P<0.05, non significant for P>0.05. Fig. S22. Toxicity assessment of gold ions (AuCl 3 ) in HeLa cells. WST-8 proliferation assay upon treatment with increasing amount of AuCl 3 . Note that the highest concentration of gold ions used (100 M) corresponds to a dose of ca. 58 nM of striped and unstructured AuNPs, namely ca. 3 times higher than the maximum concentration used in all cytotoxicity assays with AuNPs. CTRL represents the negative control; values are means+s.d. and differences between treated samples and controls (n=8) were considered statistically significant for *P<0.05, non significant for P>0.05. Representative confocal microscopy images of fluorescent SiO 2 NPs (NPs were doped with Oregon green 488) (green channel), lysosomes (stained with LysoTracker, red channel), bright field (BF), and relative merged images. Images were taken with a Leica SP8 confocal microscope with a 40X, 1.3 NA oil immersion objective.    Hence, the very same NP is largely less toxic because its suitable surface engineering strongly reduces its intracellular release of toxic ions.

Synthesis of NPs
Striped and unstructured gold nanoparticles were prepared as previously reported 12 .
Silver nanoparticles were synthesized by photochemical method by using tyrosine as a photoreducing agent. The synthesis was carried out in a laboratory reactor system fitted with UV lamp and surrounded by quartz tubing for cooling with water. The pre-cooled aqueous solution of potassium hydroxide irradiated with the UV lamp was added with tyrosine and silver ions (Ag 2 SO 4 ) under vigorous stirring conditions. After 30 min, the AgNP suspension was warmed up to room temperature. The concentrated NP suspension was purified and separated by using sephadex G-75.
CdSe/ZnS core/shell NPs were prepared following standard colloidal synthesis procedures 13,14 and transferred from the organic phase to aqueous phase by adopting a polymer coating SiO 2 NPs were synthesized as previously reported 18 .
Nanodiamonds (NDs) were purchased from Sigma. They were functionalized with fluorescein isothiocyanate (FITC) as previously reported 19 , to assess cellular internalization. Before cellular tests, NDs were purified by filtration.
ZnO, AlO, and Ni NPs were purchased from Sigma.
Pt NPs were purchased from HiQ-Nano.
InP/ZnS QDs, donated by Hicham Chibli, were synthetized according to Brunetti et al. 20 Ceria NPs were donated by Fanny Caputo (University of Rome Tor Vergata).

Characterization of NPs
All NPs used in this study were characterized by TEM, DLS and ζ-potential.
For TEM analyses, samples were prepared by dropping a dilute solution of nanoparticles in water on carbon-coated copper grids (Formvar/Carbon 300 Mesh Cu). TEM images were recorded on a JEOL Jem1011 microscope operating at an accelerating voltage of 100 kV. The size of nanoparticles was obtained after measuring the size of more than 100 particles by TEM.
For DLS and ζ-potential measurements, a Zetasizer Nano ZS90 (Malvern, USA) equipped with a 4.0 mW He-Ne laser operating at 633 nm and an avalanche photodiode detector was used.
Measurements were made at 25 °C in water (pH 7.0) or in cell culture medium, DMEM, 10% FBS.
The size distributions were performed by volume (%), while the ζ-potential measurements by intensity. Each sample was measured 5 times and the results analyzed by Malvern Instruments Ltd software.
The characterization data are reported in Tab

Measurements of NP ion release
The evaluation of NP ion release was performed at 37 °C both in acidic conditions (sodium citrate buffer, pH 4.5, an acidic medium mimicking the lysosomal environment 22,23

Measurements of NP cellular internalization
To . Then, samples were digested and analyzed by ICP as described above.

Measurements of intracellular NP ions
To estimate the amount of intracellular NP ions, HeLa cells were seeded in 150 cm 2 flasks (Sarsted) at a density of 10 6 cells/ml and exposed to 20 nM of striped and unstructured AuNPs for

WST-8 Cytotoxicity Assay
The metabolic activity of all the aforementioned cells was measured after 24 and 48 h of exposure to striped and unstructured AuNPs, using a standard WST-8 assay (Cell Counting Kit-8, code: 96992, Sigma) and following the experimental procedures described by Sabella et al. 25 .
Briefly, cells were treated with increasing concentration of AuNPs ranging from 0.03, 0.1, 0.3, 1, 3,  Table S33: TEM, DLS and ζ-potential data of Tstriped and Astriped AuNPs.   Cells were exposed to unstructured, Tstriped or Astriped AuNPs at a concentration of 65 nM for 12 h in serum free conditions, afterwards the medium was replaced with fresh, fully supplemented medium without NPs, and WST-8 viability measurements were done at 24 and 48 h as described above.

WST-8 Cytotoxicity Assay in the presence of lysosomotropic and chelating agents
To prove the direct role of intracellular NP ions (whose release is enhanced by acidic lysosomal conditions) in inducing toxicity, cells viability was monitored by WST-8 proliferation assay in the presence of lysosomotropic agents enabling the increase of the acidic lysosomal pH 27 . 24 h after seeding, HeLa cells were pretreated for 30 min with chloroquine (5 M) (Sigma) or ammonium chloride (5 mM) (Sigma) and then exposed to 2.5 nM of Fe 3 O 4 NPs or 20 nM of unstructured AuNPs for 48h. The WST-8 proliferation assay was carried out using 4 independent experiments (8 replicates for each experiment) as described above.
As a further demonstration of the direct relationship between the intracellular released ions by NPs and viability reduction, cells viability was monitored by WST-8 proliferation assay (as well as by ROS and LDH assay) in the presence of metals chelating agents, such as 2,3 dithiopropanol (BAL) 28 and the iron chelator, desferrioxamine (dfx) 29 . 24 h after seeding, HeLa cells were pretreated for 30 min with BAL (1 M) and then exposed to NPs (20 nM of striped or unstructured AuNPs, 2 nM of AgNPs, 5 nM of CdSe/ZnS QDs) for 24-48 h. In the case of Fe 3 O 4 NPs, HeLa cells were pretreated for 30 min with dfx (100 M) and exposed to 2.5 nM of the iron NPs for 24/48 h.
The WST-8 proliferation assay was carried out using 4 independent experiments (8 replicates) as described above. Untreated cells were used as control, while cells pretreated with the lysosomotropic agents or chelating agents without further NP exposure were tested for the toxicity of the applied chemicals (finding no toxicity for both the chemicals used, data not shown).
The same procedure was applied for other NPs (Fig. S21) at the working concentration ranges but with the use of DMSA (100 M).

WST-8 Cytotoxicity Assay in the presence of AuCl 3
Cellular toxicity was assessed in human HeLa upon treatment with increasing concentrations (1, 10 and 100 µM) of a gold salt, AuCl 3. Freshly prepared stock solution of gold salt in water was added to cell culture medium and to cells at the desired concentrations for 24/48 hours. The WST-8 proliferation assay was carried out using 4 independent experiments (8 replicates) as described above.

ROS quantification by DCFH-DA assay
Oxidative stress of all the aforementioned cells was determined after 24 and 48 h of exposure to striped and unstructured AuNPs following the DCFH-DA assay procedures as previously reported 25 . Cells were treated at the same final concentrations of striped and unstructured AuNPs employed for the aforementioned viability assay. In this case, as positive control, a free radical generator (H 2 O 2 at a concentration of 100 µM) was used, finding that it caused an increase of ROS around 190-220% with respect to the untreated cells (data not shown in the graphs). The fluorescence generated by cells from each well was measured immediately, and after 30 min of reaction with DCFH-DA by using Fluo Star Optima. Fluorescence quantification and data analyses were carried out as reported elsewhere 25 . Data were reported as mean ± SD for 4 independent experiments (8 replicates).
The same procedure was applied for all other NPs at their working concentration ranges.

Membrane damage by LDH assay
In these experiments, cells were seeded in a 96-well plate and treated with NPs, following the procedures previously reported. 20

In vivo experiments
The flies of wild-type Drosophila melanogaster (Oregon R+) were cultured at 24±1 °C on standard Drosophila food, containing agar, corn meal, sugar, yeast and nepagin (methyl-phydroxybenzoate). Striped and unstructured AuNPs were formulated in the Drosophila diet (dose: 0.36 µg/g per day), by dispersing the particles in the food. In particular, the NP solution was added to the food before solidification, mixed strongly and finally poured into vials.
For lifespan studies, newly eclosed flies were collected and housed at a density of 10 males and   Fig. 4 The intracellular toxicity of metal-containing NPs is mainly due to their intracellular (lysosomal) release of ions o Fig. S18 o Fig. S19  Fig. S22  Fig. 6 Demonstration that toxicity of AuNPs follows the same molecular mechanisms activated by gold ions, namely inhibition of the TrxR enzyme, leading to mitochondrial membrane depolarization and/or inactivation of mitochondrial enzymes Description of a general mechanism for intracellular toxicity of metal-containing NPs (LETH mechanism)  Schematic of LETH mechanisms and colocalization studies  Fig. 7