Interaction of biologically relevant proteins with ZnO nanomaterials: A confounding factor for in vitro toxicity endpoints

16 The results of in vitro toxicological studies for manufactured nanomaterials (MNs) are often contradictory 17 and not reproducible. Interference of the MNs with assays has been suggested. However, understanding for 18 which materials and how these artefacts occur remains a major challenge. This study investigated 19 interactions between two well-characterized ZnO MNs (NM-110 and NM-111) and lactate dehydrogenase 20 (LDH), and two interleukins (IL-6 and IL-8). Particles (10 to 640 µg/mL) and proteins were incubated for 24 21 hours in routine in vitro assays test conditions. LDH activity (OD LDH ), but not interleukins concentrations, 22 decreased sharply in a dose-dependent manner within an hour after exposure (OD LDH < 60% of OD ref for 23 both MNs at 10 µg/mL). A Freundlich adsorption isotherm was successfully applied, indicating multilayer 24 adsorption of LDH. ZnO MNs and LDH had neutral to slightly negative surface charges in dispersion, 25 precluding electrostatic attachment. Particle sedimentation was not a limiting factor. Fast dissolution of ZnO 26 MNs was shown and Zn 2+ could play a role in the OD LDH drop. To summarize, ZnO MNs quickly reduced 27 OD LDH due to concentration-dependent adsorption and LDH inhibition by interaction with dissolved Zn. The 28 control of particle interference in toxicological in vitro assays should become mandatory to avoid misleading 29 interpretation of results. 30 A objective of this study was to investigate the interactions occurring in the typical conditions for in vitro toxicological studies. We followed the guidance for test item preparation defined in the large EU FP7 NANoREG project where a major effort was made to harmonize the test procedures used to increase 86 interlaboratory compatibility of the initial in vitro exposure characteristics 30,31 . The experiments reported 87 here were designed to identify the interaction of LDH, IL-6 and IL-8 with well-characterized uncoated (NM- 88 110) and coated (NM-111) ZnO MNs. The ZnO MNs were selected for study due to initial results from 89 screening studies and literature data on potential interaction between ZnO and proteins 6 . ZnO is also a 90 widely used material that can lead to human pulmonary and cardiovascular health effects 32,33 and a large 91 number of studies report on their toxicity in vivo as well as in vitro and NM-111. These studies were conducted using the SensorDish ® Reader system with incubation in the cell incubator at 37  C, 95% relative humidity, and 5% CO 2 atmosphere. ZnO MNs were dispersed in the test medium and dispensed in 24-well plates on the SensorDish ® Readers in the incubator immediately after sonication. The system was then calibrated at 37 °C using the single measurement command three times. Dissolved oxygen and pH were

mechanisms for interaction include optical interference 6,15 , interaction by adsorption of the target molecule 48 onto the test material 7-9,12-14,16-18 and chemical reactions with the assay components 3,8,19 . These types of 49 interaction could either lower or increase the signal, thereby leading to false interpretation of the results. 50 It is essential to ensure that in vitro toxicity assays produce reliable and reproducible data if results will be 51 used for hazard screening leading to decision making for regulatory purposes. Cytotoxicity and inflammatory 52 response are two of the most commonly investigated endpoints. Cytotoxicity is often assessed and quantified 53 by measuring the concentration of the cytosolic enzyme lactate dehydrogenase (LDH) that is released to the 54 medium from damaged cells. In this assay, LDH first catalyzes the conversion of lactate to pyruvate via 55 reduction of NAD + to NADH. Then, NADH reduces a tetrazolium salt to a red formazan product. This 56 product can be detected by spectrophotometric absorbance. Similarly, pro-inflammatory cytokines IL-6 and

Calculation of adsorption isotherms for LDH on ZnO MNs
The interaction between LDH and ZnO MNs was quantified in terms of adsorption isotherms. Adsorption 167 isotherm experiments were conducted at 37 °C after 24 h incubation at initial concentration of ZnO MNs 168 ranging from 10 to 640 µg/mL and initial LDH concentration ranging from 20 to 1000 ng/mL as described 169 above. A control study without ZnO MNs was included. The amount of LDH adsorbed, q (mg/m 2 ) was 170 estimated from the mass balance shown in equation (1): 171 In this equation, Co (mg/L) is ODref (initial LDH absorbance), Ce (mg/L) is ODLDH (defined as the optical 173 density measured at 490 nm) at time t, V (mL) is the solution volume, A (m 2 /g) is the particle specific surface 174 area per unit mass and M (g) is the particle mass. 175 Experiments were performed at 37 °C and data were recorded over 6 h. After 5 min thermal equilibration, 198 ten measurements were collected with automatic measurement optimization. After the ten first 199 measurements, the measurement position, the attenuator settings and number of sub-runs were fixed and data 200 recording continued. 201

Measurement of ZnO MNs zeta potential and LDH isoelectric point
The agglomeration and sedimentation behaviors were then assessed from the evolution over time in 202 hydrodynamic size and derived count rate. It is worth noting that the detector was placed 3 mm above the 203 base of the DLS cuvette. Thus, the recorded signal describes the accumulation and particle size changes that The ZnO dissolution experiments were conducted at 320 µg/mL using the 24-well plate SensorDish ® Reader 213 system (SDR, PreSens Precision Sensing GmbH, Germany), which allows simultaneous time-resolved 214 The SensorDish ® with 875 µL cHam's F12 added to each well were placed on the plate readers in the 217 incubation chamber. After thermal equilibration to 37 C, 125 µL ZnO batch dispersion or control dispersion 218 medium was added to the wells. Medium samples were collected either at 0.25, 1, 2, 4 and 24 h into the test 219 or after 24 h with no sampling. After collection of the medium samples by pipette, the samples were 220 immediately added to 3 kDa centrifugal filter tubes and centrifuged at 4000 × RCF for 30 min. In order to test whether decrease in LDH absorbance was due to interaction with dissolved Zn, LDH 227 absorbance was measured following incubation with 1.75 to 640 µg/mL Zn 2+ ions in the form of ZnCl2. 228 ZnCl2 was dissolved in 0.05% BSA water (2.56 mg/mL) and added to the test medium for 24 h at 37 °C, 229 following the same protocol as for NM-110 and NM-111. After centrifugation at 20,000 × RCF for 30 min, 230 LDH absorbance was measured in the supernatant as described above. Three independent experiments with 231 two replicates each were conducted. 232 Investigating the role of pH on LDH 233 We observed that ZnO dissolution resulted in changes in the pH in cHam's F12. Therefore, a detailed study 234 was made to further investigate the temporal evolution of pH for both NM-110 and NM-111. These studies 235 were conducted using the SensorDish ® Reader system with incubation in the cell incubator at 37 C, 95% 236 relative humidity, and 5% CO2 atmosphere. ZnO MNs were dispersed in the test medium and dispensed in 237 24-well plates on the SensorDish ® Readers in the incubator immediately after sonication. The system was 238 then calibrated at 37 °C using the single measurement command three times. Dissolved oxygen and pH were 239 The dissolved Zn concentration, pH evolution, and chemical species distribution during addition of either 244 ZnO or dissolved ZnCl2 to Ham's F12 was assessed using the React program and plotted using the Gtplot 245 Apps in Geochemist Workbench ® v. 11.0 42 . A simplified saline Ham's F12 composition was used in the 246 calculations. This was due to lack of thermodynamic data for especially sugars, organic and amino acids, as 247 well as vitamins in the Geochemist Workbench ® database. Table 2   of 100 ng LDH/mL, 500 µg IL-6/mL and 4000 µg IL-8/mL. The applied cHam's F12 recipe is often used for 266 in vitro toxicological testing 43-46 and the tested levels of LDH and interleukins correspond to moderate levels 267 of particle induced cytotoxicity and inflammation in human lung epithelial cell line A549 47-50 . The 268 interaction of LDH with ZnO MNs was described using two adsorption isotherm models. Additionally, the 269 kinetics of the observed effects, the role of particle sedimentation, pH, and particle dissolution was 270 examined. All tests were conducted in a cell incubator at 37 °C with a 5% CO2 atmosphere and a relative 271 humidity of 95%. 272 and interaction with specific proteins in the assay 9,13,17,52-54 . Even though the FBS concentration in cHam's 285 F12 was considerably higher than the BSA concentration used for making the batch dispersions (10% and 286

Protein-particle interactions 289
To test the protein-particle interactions, 10 to 640 µg/mL (NM-110 and NM-111) were incubated in cHam's 290 F12 with LDH, IL-6, and IL-8 for 24 h. We observed that only the LDH levels were significantly reduced in 291 the test medium ( Figure 1 and Table 2   ZnO MNs, is in agreement with results from Kroll et al. 6 who detected a specific interference of ZnO MNs 304 with the LDH ELISA assay among a set of 24 MNs of diverse natures and physicochemical properties. 305 Furthermore, evidence was provided that IL-8 measurements were strongly affected by TiO2 MNs but not by 306 ZnO MNs. This is in agreement with other studies demonstrating a decrease in IL-8 levels after incubation 307 with TiO2 MNs 13,55 . As the LDH assay is often used to screen and control for cytotoxicity, it is important to 308 further our understanding of the kinetics of the LDH interaction and investigate whether determinant 309 physicochemical characteristics can be identified. The kinetics of the interaction can play a role in for 310 example in vitro genotoxicity studies, where early time points are used to understand initial effects on DNA 311 and signaling 29 . 312

Kinetics of LDH interaction 313
Repeated measurements of LDH absorbance levels were made during 120 min incubation with 640 μg/mL 314 NM-110, dosing 50, 100, 200, and 1000 ng LDH/mL cHam's F12 (Figure 2 and Table 3). The ODLDH 315 reduction rate (initial slope) increased with increasing initial LDH concentration. Statistically significant 316 differences were observed (p-value < 0.05) between 50 and 200 ng/mL, 50 and 1000 ng/mL and 100 and 317 1000 ng/mL. In all tests, a plateau was reached after approximately 60 min (borderline for 1000 ng/mL) 318 where the LDH levels were inversely proportional to the initial LDH concentration (statistically significant 319 difference, except between 200 and 1000 ng/mL, Tukey pairwise comparison, p-value < 0.001), i.e., the 320 more LDH that was initially present in the test medium, the more had interacted with the NM-110. 321    These results show that the interaction between LDH and NM-110 was completed within the first hour of 331 incubation. The fast onset of LDH disappearance from the test medium was also observed after exposure to 332 soot and oxidized soot particles 7 . The occurrence of a plateau after 60 min incubation regardless of the 333 initial LDH concentrations suggests that a material-related phenomenon occurs in the test medium, which 334 prevents further interaction between ZnO MNs and LDH after 60 min. In addition to saturation of ZnO MNs 335 adsorption sites, the physicochemical conditions in the dispersions, including particle sedimentation and/or 336 dissolution may play a role. The timing of measurement in interaction tests appears to be highly important. between adsorbed molecules. The results for the calculations are plotted in Figure 3 while the constants 356 obtained from applying the models to each set of independent experiment (each conducted in duplicate) and 357 the correlation coefficients R 2 are listed in Table 4.

364
The correlation coefficients (Table 4) and the graph (Figure 3) show that the adsorption data fit better to the 365 While LDH adsorption can be governed by electrostatic mechanisms or physical properties, the negative 369 correlation between MNs concentration and adsorbed LDH suggests that other phenomena in the test 370 medium play a role. It is likely that the increase in MNs concentration leads to an increase in particle 371 sedimentation rate and particle-particle interaction which could result in increased MNs agglomeration 372 altogether reducing the surface area available for adsorption. Besides, ZnO MNs are known to be partially 373 soluble in test medium 57 . On one hand, particle dissolution will release Zn ions, which have previously been 374 proposed to interact with LDH 6 . On the other hand, dissolution will further reduce the surface area available 375 for interaction. Finally, the test medium pH could change during dissolution of ZnO MNs and affect LDH in 376 the medium. The role of these phenomena in the interaction between ZnO MNs and LDH is assessed below. 377

Zeta potential of LDH and ZnO MNs in cHam's F12 379
The interaction between LDH and ZnO MNs can be governed by several mechanisms, including electrostatic 380 interactions. The particle surface charge in a given medium varies with pH. Therefore, to elucidate whether 381 electrostatic forces were responsible for the observed interaction between LDH and ZnO MNs, we 382

Temporal sedimentation 396
Particle agglomeration and sedimentation in a given dispersion medium are governed by surface charge or 397 steric stabilization but also by the particle concentration 58 . Due to the neutral to slightly negative zeta 398 potential of the ZnO MNs (reported above), we expected that the ZnO MNs could have relatively fast 399 sedimentation rates in cHam's F12 in our experiments. This was confirmed by testing the stability of the two 400 ZnO MNs dispersions at 320 µg/mL in cHam's F12 over 6 h at 37 C. Sedimentation plots were obtained 401 using the variation in the scattered light intensity (It , detected as mean count rate) relative to the initial value 402

pH and dissolution of ZnO MNs and interaction with LDH
ZnO is known to be partially soluble in water as well as in different media used for in vitro toxicological 417 testing and Zn has been proposed to interfere with the LDH due to binding of zinc ions to histidine tails of 418 The full composition of the Ham's F12 nutrient mixture is, however, chemically more complex than the one 426 that could be used for modeling based on available chemicals in the thermodynamic database. The full 427 Ham's F12 nutrient mixture also contains various additional organic compounds, amino acids, nutrients and 428 vitamins, as well as the 1% v/v penicillin/streptomycin and 10% v/v FBS. In addition, the test medium with 429 the ZnO MNs was incubated in a 5% CO2 enriched atmosphere, which can also influence the system. Finally, 430 due to sedimentation, there may also be temporal differences between the chemical microenvironment at the 431 bottom of the wells, where the ZnO MNs accumulate, compared to the volume above, which is progressively 432 being depleted of MNs over time. This could be important when cells are attached at the bottom of the wells, 433 which is the case for e.g., lung epithelial cells. 434 Therefore, we investigated the pH reactivity and release of Zn during dissolution of ZnO MNs using the 435 SensorDish™ Reader (SDR) method. The results showed that the pH values rapidly exceeded 9 (the upper 436 limit of detection of the SDR sensors) (Figure 4). The strongest and most prolonged effect was observed for 437 NM-110 where the pH was above the detection limit between 170 and 520 min.

Evaluation of interaction between released Zn and LDH 463
Based on the dissolution results, a specific study to investigate the potential role of dissolved Zn on the 464 interaction with LDH was conducted. A marked decrease in LDH absorbance was observed with increasing 465 Zn concentration starting at 10 µg Zn/mL ( Figure 5). The effect of dissolved ZnCl2 on the LDH was 466 comparable to what was observed after incubation with ZnO MNs. However, the effect was observed at 467 much higher Zn concentrations than was reached during dissolution testing of the two ZnO MNs (7 to 8 468 µg/mL after 24 h, Table 5). The decrease in LDH absorbance following dosing and incubation with 1.75 to 469 10 µg dissolved Zn/mL was not significant. Consequently, the previously reported inhibition of LDH due to  repulsive forces between LDH and ZnO MNs. Chemical reaction modeling suggests that an increase in 492 dissolved Zn in cHAM's F12 can lead to a slight drop in pH. Lowering of the pH would reduce the negative 493 surface charge towards neutral as the isoelectrical point of LDH was found to be at pH 4.67. 494 None of the investigated physicochemical parameters could directly justify the observed interaction between 495 LDH and ZnO MNs. Even though dosing with dissolved Zn did not immediately explain the LDH 496 interaction, this may still be an explanation. Indeed, one can expect the dissolved Zn concentration to be 497 higher at the interface and in the surroundings of dissolved ZnO MNs than the average concentrations in the 498 mediums, which is what we measured. Our chemical reaction modeling suggested that the dissolved Zn 499 concentration can be higher than the 24-hour solubility limit measured in our test. However, high dissolved 500 Zn concentration at the ZnO interface does not explain that more LDH interacts when more LDH is added to 501 the cHam's F12 test medium. In this regard, the reasonable fit of the data with the Freundlich isotherm 502 model suggests that the observed interaction between LDH and ZnO MNs could be driven by multilayer 503 adsorption of LDH without saturation of adsorption sites. In summary, we suggest that the observed 504 interferences are mainly due to concentration-dependent physisorption on to ZnO MNs potentially assisted 505 always be considered in order to limit artefacts and erroneous conclusions regarding the toxicity of a test 524 material. However, if other methods are not available, it is recommended to use correction factors to remove, 525 or at least limit, the effect of an interaction. Recommendations for controlling particle interference in the 526 optical analysis in colorimetric assays are insufficient. While standardized methods are needed for 527 dispersing, testing and quantifying toxicological effects, a case-by-case evaluation of the interaction of MNs 528 with the assay is strongly recommended to obtain reliable results. 529 cHam's F12 (Ham's F12 nutrient mixture with 1% v/v penicillin/streptomycin and 10% v/v FBS) under 532 typical in vitro test conditions. The MNs were added to cHam's F12 after dispersion in BSA water following 533 the NANOGENOTOX batch dispersion protocol 25 . 534 When dosed in cHam's F12, the ZnO MNs were, at least partially, coated with BSA and had neutral to 535 slightly negative zeta potential. LDH interacted with the ZnO MNs and the LDH levels in cHam's F12 536 decreased linearly to a minimum plateau within one hour. Moreover, LDH reduction could be augmented 537 with increasing initial LDH concentrations. No interaction was observed with IL-6, IL-8, and FBS. Full 538 sedimentation of ZnO MNs occurred after about 3 h and is hence not a limiting factor for LDH interaction. 539 Addition of 10 to 20 mg Zn/mL (ZnCl2 dissolved in Nanopure™ water) caused a steep reduction in the 540 measured LDH level. However, Zn concentrations only reached 5-6 µg Zn/mL after 1 h dissolution of NM-541 110 and NM-111 and 7-8 µg Zn/mL after 24 h. 542 In spite of partial dissolution, the LDH loss could be explained using a Freundlich adsorption isotherm 543 model. Electrostatic interaction was ruled out since the zeta potential of ZnO MNs, LDH and FBS in cHam's 544 F12 were all neutral to slightly negative. We conclude that the observed LDH interaction with ZnO MNs in 545 cHam's F12 is mainly governed by physisorption. It is possible that high concentrations of Zn ions in 546 microenvironments around dissolving ZnO also play a role in lowering the level of measured LDH due to 547 due to binding of Zn ions to histidine tails of LDH 6,59 . 548 We demonstrate that LDH levels measured in in vitro toxicological tests of ZnO MNs cannot be used 549 directly for interpretation. If not controlled for, artefacts in toxicological assays can lead to erroneous 550 estimation of particles toxicity. We highlight the need for careful considerations and thorough 551 characterization of the test system when conducting in vitro toxicological assays with particulate materials. 552 The current study only covers the interaction between ZnO nanoparticles and LDH and it remains to be 553 tested if other enzyme based assays are prone to similar confounding factors.  (Table S1), protein-563 particle interaction (Table S2), LDH zeta potential ( Figure S1), chemical reaction modeling ( Figure S2). This 564 material is available free of charge via the Internet at http://pubs.acs.org. 565 against oleic acid hydroperoxide-induced oxidative damage in IPEC-J2 cells. International Journal 705 of Molecular Medicine. 706 (47) Corsini, E., Budello, S., Marabini, L., Galbiati, V., Piazzalunga, A., Barbieri, P., Cozzutto, S.,