Toxicology in Vitro Comparison of di ﬀ erent in vitro cell models for the assessment of pesticide-induced dopaminergic neurotoxicity

Biomedical and (neuro) toxicity research on (neuro) degenerative diseases still relies strongly on animal models. However, the use of laboratory animals is often undesirable for both ethical and technical reasons. Current in vitro research thus largely relies on tumor derived- or immortalized cell lines. Notably, the suitability of cell lines for studying neurodegeneration is determined by their intrinsic properties. We therefore characterized PC12, SH-SY5Y, MES23.5 and N27 cells with respect to the presence of functional membrane ion channels and receptors as well as for the e ﬀ ects of ﬁ ve known neurotoxic pesticides on cytotoxicity, oxidative stress and parameters of intracellular calcium homeostasis using a combined alamar Blue/CFDA assay, a H 2 DCFDA assay and single cell ﬂ uorescent (Fura-2) calcium imaging, respectively. Although all pesticides demonstrated a certain level of functional neurotoxicity in the di ﬀ erent cell lines, our results also demonstrate considerable di ﬀ erences in in- trinsic properties and pesticide-induced e ﬀ ects between the cell lines. This clearly indicates that care should be taken when interpreting (neuro)toxicity data as the chosen cell model may greatly in ﬂ uence the outcome.


Introduction
Current (neuro) toxicity and biomedical research on human (degenerative) diseases of the nervous system relies strongly on animal models, mainly rodents. According to the 2010 EU report on the annual laboratory animal use (SEC2010/1107) biomedical-and toxicological research involved nearly 2.5 million animals. These animal experiments are not only ethically debated but also expensive, time-consuming and unsuitable for high-throughput testing. In vitro approaches, in contrast, provide a relatively fast and cheap way of testing chemicals for their (neuro) toxic properties without the need for the extensive use of laboratory animals (Westerink, 2013).
Available in vitro techniques to study pathways of neurotoxicity and neurodegeneration rely mainly on chemical-induced effects in a wide array of (often tumor-derived, rodent or human) cell lines or pluripotent (embryonic, rodent or human) stem cells. Although stem cells provide the unique opportunity to study (developmental) neurotoxicity in a complex network, homogeneous cell lines allow for in-detail, single-cell studies dissecting (human relevant) molecular pathways. Therefore, toxicological research on potential neurotoxic properties of compounds could focus on in vitro strategies using one or more cell lines from the wide array of cell lines available.
Various cell models are reported to have dopaminergic properties and should thus be suitable for in vitro studies on dopaminergic neurotoxicity, which could aid in studying for example Parkinson's Disease (PD). However, the suitability of a cell model for answering a particular research question relies largely on its intrinsic (functional) properties. Hence, it is striking that many cell lines are poorly characterized with respect to intrinsic properties, including the presence of important membrane receptors and ion channels.
As neurotoxicity of a compound is often determined by the presence of a particular transporter (e.g. dopaminergic toxicity of MPP + via uptake by dopamine membrane transporters (DAT)) or membrane receptor (e.g. excitotoxicity by activation of glutamatergic neurotransmitter receptors), the use of an inappropriate model may lead to an erroneous estimation of (neuro) toxicity. We therefore chose to characterize four catecholaminergic cell lines (rat PC12, human SH-SY5Y, mouse/rat hybrid MES23.5 and rat N27 cells) with respect to functional properties.
Rat pheochromocytoma cells (PC12; (Greene and Tischler, 1976) provide a widely used model in neurotoxicology that has been extensively characterized for neurosecretion and the presence of ion channels and neurotransmitter receptors (Shafer and Atchison, 1991;Westerink and Ewing, 2008). Although these cells originate from an (non-neuronal) adrenal tumor, they display several characteristics of mature dopaminergic neurons (Table 1). This renders the PC12 cell a suitable model for the study of catecholaminergic neurotoxicity in vitro (Heusinkveld et al., 2016;Meijer et al., 2014).
The SH-SY5Y cell line (Biedler et al., 1973) is a catecholaminergic neuroblastoma cell line from human origin, widely used to study neurotoxicity in vitro (Faria et al., 2016;Sala et al., 2016). SH-SY5Y cells synthesize and release noradrenaline (Table 1) through vesicular release (i.e., exocytosis) (Vaughan et al., 1995;Zhao et al., 2012). The hybrid MES23.5 cell line is a product of somatic fusion of rat embryonic mesencephalon cells and the murine N18TG2 neuroblastoma-glioma cell line (Crawford et al., 1992). The resulting hybrid cell line displays many characteristics of mesencephalic dopaminergic neurons (Table 1). The N27 cell line is an immortalized cell line derived from TH-positive fetal rat mesencephalic neurons (Prasad et al., 1994); Table 1). These cells are increasingly used to study in vitro (parkinsonian) dopaminergic neurodegeneration (see e.g. Song et al., 2011;Xu et al., 2016).
As pesticides are implicated in in vivo and in vitro neurotoxicity, a reference set of five pesticides was composed consisting of rotenone, lindane, dieldrin, imazalil and dinoseb. These pesticides all display, to a certain extent, dopaminergic neurotoxicity albeit with a different mechanism of action. Rotenone is a well-known model compound for inducing PD via mitochondrial uncoupling (complex I) in vivo and is therefore extensively used to study dopaminergic neurodegeneration, both in vivo and in vitro (see e.g. Dawson and Dawson, 2003;Greenamyre et al., 2010). The dinitrophenolic herbicide dinoseb is also a mitochondrial uncoupler (complex III; Palmeira et al., 1994), that has recently also been linked to Ca 2+mediated activation of pathways implicated in dopaminergic neurodegeneration in vitro (Heusinkveld et al., 2016).
The organochlorine insecticides lindane and dieldrin are classical neurotoxicants in vitro as well as in vivo and are implicated in the etiology of PD (Corrigan et al., 2000;Franco et al., 2010). The primary mechanism underlying lindane and dieldrin-induced neurotoxicity is inhibition of GABA receptors (Anand et al., 1998;Vale et al., 2003). In addition, both lindane and dieldrin have been linked to disturbance of intracellular calcium homeostasis (Heusinkveld and Westerink, 2012). Furthermore, the azole fungicide imazalil has been linked to adverse neurobehavioural effects in vivo (Tanaka, 1995) and was recently linked to disturbance of the intracellular calcium homeostasis (Heusinkveld et al., 2013).
Intracellular Ca 2+ homeostasis plays a pivotal role in neuronal function, development and survival of dopaminergic cells (Gleichmann and Mattson, 2011). Therefore, we characterized the four cell lines with regards to functional parameters, such as the presence of functional neurotransmitter receptors and ion channels. Furthermore, effects of the reference pesticides on basal-and depolarization-evoked changes in the intracellular Ca 2+ concentration ([Ca 2+ ] i ) were investigated. As cell death and oxidative stress are clearly implicated in neurotoxicity and degeneration (Goodwin et al., 2013), the effects on cell viability and production of reactive oxygen species (ROS) upon exposure to the reference set of pesticides were also assessed in all cell lines.
To investigate compound-induced effects on basal and depolarization-evoked [Ca 2+ ] i , cells were depolarized by high-K + containing saline (100 mM K + ) for 18 s after a 5 min baseline recording. This provides a robust depolarization to~0 mV depending on the resting potential of the cell. This depolarization is sufficient to open both highand low voltage-activated VGCCs. Following a 10 min recovery period, cells were exposed to DMSO (0.1%), or one of the reference compounds for 20 min (indicated by the dashed line in Fig. 2A) to evaluate the effects on basal [Ca 2+ ] i . Subsequently, cells were depolarized for a second time to evaluate effects of exposure on depolarization-evoked Ca 2+ -influx. This is calculated as the treatment ratio (TR; %) between the first and second depolarization-evoked [Ca 2+ ] i peak values relative to control.

Cell viability assays
To assess the effects of the compounds on cell viability a combined alamar Blue/CFDA-AM (aB/CFDA) assay was used (protocol adapted from Bopp and Lettieri, 2008) to determine respectively mitochondrial activity and membrane integrity. Cells were exposed in serum-and phenol red-free medium to concentrations up to 100 μM for up to 24 h. Mitochondrial activity of the cells was recorded as a measure of cell viability with the aB assay, which is based on the ability of the cells to reduce resazurin to resorufin. In the same experiment, membrane integrity was assessed indirectly using a CFDA-AM assay, which is based on non-specific cytoplasmic-esterase activity. Briefly, cells were incubated for 30 min with 12,5 μM aB and 4 μM CFDA-AM. Resorufin was measured spectrophotometrically at 540/590 nm (Infinite M200 microplate; Tecan Trading AG, Männedorf, Switzerland), whereas hydrolysed CFDA was measured spectrophotometrically at 493/541 nm.

Production of ROS
The involvement of oxidative stress in the observed reduction in cell viability was investigated using the fluorescent dye H 2 DCFDA as described previously (Heusinkveld et al., 2010). Briefly, cells were loaded with 1.5 μM H 2 DCFDA for 30 min at 37°C. Subsequently, cells were exposed for up to 24 h to 0.1-100 μM compound and fluorescence was measured spectrophotometrically at 488/520 nm (Infinite M200 microplate; Tecan Trading AG, Männedorf, Switzerland).

Data-analysis and statistics
Data from cell viability and ROS experiments were generated in at least 3 individual experiments (N ≥ 3) consisting of at least 3 replicates per experiment (n ≥ 9). Cell viability data (aB/CFDA) are presented as % viability ± standard error of the mean (SEM; calculated using the N) compared to DMSO controls. Data on production of ROS are presented as % increase in ROS ± SEM (calculated using the N) compared to timematched DMSO controls. A relevant effect size in cell viability and ROS experiments, indicated as lowest observed effect concentration (LOEC), is defined as the concentration that induces a ≥ 20% change in the parameter assessed.
Data from single-cell fluorescence microscopy is presented as F340/ F380 ratio (R), reflecting changes in [Ca 2+ ] i , and analyzed using custom-made MS-Excel macros applying a correction for background fluorescence (Heusinkveld et al., 2013). The data represent average values derived from 15 to 63 individual cells (n) in 3-6 independent experiments (N).
Statistical analyses were performed using GraphPad Prism v6.01 (GraphPad Software, San Diego, California, USA). Concentration-response curves were fitted for experiments assessing cell viability and ROS production using a nonlinear sigmoidal or bell-shaped curve-fit when applicable.

Characterization of neurotransmitter receptors and ion channels
To characterize the different cell lines with respect to the presence of functional Ca 2+ -permeable receptors and ion channels, cell lines were stimulated with different stimuli and [Ca 2+ ] i was monitored using single-cell fluorescence microscopy.
Upon depolarization with 100 mM K + , an increase in [Ca 2+ ] i was observed in PC12, SH-SY5Y and MES23.5 cells, indicative of the presence of VGCCs. The largest increase in [Ca 2+ ] i was observed in PC12 cells, followed by MES23.5 and SH-SY5Y cells. N27 cells displayed no increase in [Ca 2+ ] i upon depolarization. Similarly, upon stimulation with 100 μM ACh, an increase in [Ca 2+ ] i was observed in PC12, MES23.5 and SH-SY5Y cells, suggestive for the presence of ionotropic and/or metabotropic ACh receptors. However, no response was detected in N27 cells. PC12, MES23.5 and N27 responded to stimulation with 100 μM ATP, indicative of the presence of purinergic receptors. However, SH-SY5Y did not respond to ATP. When stimulated with 100 μM serotonin (5-HT) only the MES23.5 cells responded with an increase in [Ca 2+ ] i , whereas the other cell lines showed no change in [Ca 2+ ] i . None of the cell lines responded to stimulation with glutamate (100 μM), neither with nor without prior depolarization with 100 mM K + , suggestive for the absence of functional ionotropic glutamate receptors.

Pesticide-induced changes in [Ca 2+ ] i
The concentrations of the pesticides applied to assess the effects on the [Ca 2+ ] i were chosen based on pilot experiments or previous results in which they were proven effective in PC12 cells without overt acute cytotoxicity.

Basal Ca 2+
Upon exposure to lindane (100 μM), a transient increase in basal [Ca 2+ ] i was observed in PC12 cells ( Fig. 2A) in line with earlier measurements (Heusinkveld et al., 2010), whereas no change in basal [Ca 2+ ] i was observed in SH-SY5Y, MES 23.5 or N27 cells (Fig. 2B). Following exposure to dinoseb (30 μM) an increase in basal [Ca 2+ ] i was observed in all cell lines although the nature of the increase in PC12, MES23.5 and N27 was transient, whereas in SH-SY5Y cells a transient increase was followed by a sustained elevated level of basal [Ca 2+ ] i . No change in basal [Ca 2+ ] i was observed upon exposure to dieldrin (10 μM), rotenone (10 μM) or imazalil (30 μM) in any of the cell lines (Fig. 2).

Membrane integrity, mitochondrial activity and ROS production
To assess differences between cellular responses to pesticide exposure (0.1-100 μM; 24 h), mitochondrial activity and membrane integrity were assessed using a combined alamar Blue (aB) and CFDA assay. To assess the involvement of ROS production in the observed changes in cell viability, cumulative ROS production was measured using the fluorescent dye H 2 DCFDA.

Discussion
In this study we compared four catecholaminergic neuronal cell lines with respect to functional characteristics and toxicity of five known neurotoxic pesticides. The results presented in this paper clearly indicate that the observed effects of a compound depend on the cell line used and are thus likely related to intrinsic properties of the cell line, such as the expression of ion channels and receptors.
Calcium plays a pivotal role in many inter-and intraneuronal processes, including gene transcription (Carrasco and Hidalgo, 2006), neurotransmission (Westerink, 2006), neurodegeneration (Mattson, 2012) and neurodevelopment (Pravettoni et al., 2000). As such, changes in calcium homeostasis can be considered an important key event linking a molecular initiating event to a downstream adverse effect in so-called adverse outcome pathways (AOPs). Considering the different responses in [Ca 2+ ] i of the cell lines to the applied stimuli ( Fig. 1), considerable differences in channel/receptor expression exist between the cell lines with MES23.5 cells being the most-and N27 cells the least versatile. The strong response of SH-SY5Y cells to ACh (100% response rate) indicates that SH-SY5Y cells can be used as a model for effects on ACh receptors. Although the approach used cannot be used to discriminate between ionotropic and metabotropic ACh receptors, data from literature suggests that both receptor types are present in undifferentiated SH-SY5Y cells (Vetter and Lewis, 2010). Therefore, the observed increase in [Ca 2+ ] i possibly consists of a combined influx from extracellular Ca 2+ and store-mediated release of Ca 2+ . In contrast, the SH-SY5Y cells did not respond to ATP indicating that these cells do not express P2X or P2Y purinergic receptors. This is in line with findings of Vetter and Lewis (Vetter and Lewis, 2010), but in contrast to earlier findings of Larsson and co-workers (Larsson et al., 2002) who reported the presence of P2X7 receptors in SH-SY5Y cells. This effectively illustrates that differences also may occur between clones of the same cell line, stressing the importance of knowing the properties of the cell line used when addressing compound-induced effects. As opposed to the non-responsive SH-SY5Y cell line, 32% of the N27 and 100% of the MES23.5 and PC12 cells in this study responded to ATP indicating the expression of functional purinergic receptors.
Furthermore, 74% of the MES23.5 cells responded to serotonin exposure indicative of the presence of 5-HT receptors, whereas the other cell lines did not show signs of functional 5-HT receptor.
Importantly, no response to glutamate was observed in either of the cell lines. While studies have been published in which these cell lines were exposed to high millimolar levels of glutamate to evoke cell death, the absence of glutamate-evoked calcium influx suggests it can be debated if this truly represents excitotoxicity. Consequently, the study of this important pathway of neurotoxicity requires other (primary) cell models. The observation that PC12, MES23.5 and SH-SY5Y cells respond to high-K + -induced depolarization and thus express VGCCs is in line with other research (Reuveny and Narahashi, 1991;Schneider et al., 1995;Shafer and Atchison, 1991). However, the types of VGCCs expressed differ among the cell lines as PC12 cells express L-, N-and P/ Q-type VGCCs (Dingemans et al., 2009), whereas SH-SY5Y cells are reported to express L-and N-type VGCCs (Reuveny and Narahashi, 1991) and MES23.5 cells are reported to express predominantly N-type VGCCs (Schneider et al., 1995). As N27 cells do not respond to high-K +evoked depolarization, we conclude that N27 do not express functional VGCCs.
The differences in Ca 2+ machinery between cell lines can lead to false-negative results when testing for neurotoxicity as illustrated by the observed absence of a lindane-induced effect on basal [Ca 2+ ] i in SHSY-5Y and MES23.5 cells. In PC12 cells, the increase in basal [Ca 2+ ] i upon exposure to lindane has been linked to a lindane-induced depolarization of the membrane of 32 mV (± 7 mV) causing opening of ωconotoxin sensitive high-voltage activated calcium channels (N-and P/ Q-type; Heusinkveld et al., 2010). This depolarization is apparently enough to reach the activation potential (V a ; Catterall et al., 2005) of the N-and P/Q-type VGCCs in PC12 cells, and to cause a Ca 2+ -influx. As MES23.5 and SH-SY5Y are documented to express N-type VGCCs, but in particular the MES23.5 has a more negative membrane potential (see: Åkerman et al., 1984;Colom et al., 1998), the V a is possibly not reached in these cells which may explain the absence of lindane-induced Ca 2+ influx. Hence, the less negative membrane potential in PC12 cells renders these cells a good sentinel for neurotoxicity in excitable cells.
Furthermore, since the underlying mechanism of the dinoseb-induced increase in basal [Ca 2+ ] i is Ca 2+ release from the endoplasmic reticulum (ER; Heusinkveld et al., 2016) the detection of dinoseb-induced increase in basal [Ca 2+ ] i in all cell lines is not surprising as all cells contain ER. However, the presence or absence of several channels and pumps as well as the coupling between ER and mitochondria may determine whether an increase in basal [Ca 2+ ] i is transient or sustained. Therefore, the observation that upon exposure to dinoseb a transient increase in [Ca 2+ ] i is observed in PC12, MES23.5 and N27 cells, whereas this transient increase is followed by a sustained increase in [Ca 2+ ] i in SH-SY5Y cells, is indicative for fundamental differences between the cell lines in Ca 2+ -handling machinery. Also, the differences observed in the depolarization-induced increase in [Ca 2+ ] i upon exposure to dinoseb (MES23.5: increase; PC12: no effect) are likely explained by differences in intracellular Ca 2+ handling and Ca 2+ -related feedback loops. The observed dinoseb-induced decrease in depolarization-evoked [Ca 2+ ] i in human SH-SY5Y cells is most likely related to inhibition of VGCCs due to the sustained increase in [Ca 2+ ] i . Altogether, this clearly demonstrates the importance of characterization of cell lines as the presence or absence of a toxic response can depend on rather subtle differences between cell lines.
Also the results from the experiments assessing the more general measures of toxicity (membrane integrity, mitochondrial activity and ROS production) demonstrate differences in sensitivity between cell lines. In general, the results clearly indicate that a change in mitochondrial activity is not necessarily linked to an increase in ROS production. Also, an increase in ROS production is not necessarily linked to more cell death. This is illustrated by the results from the human SH-SY5Y cell line that displays the highest relative ROS production in response to exposure to the mitochondrial toxicants dinoseb Fig. 3. Differential effects of selected pesticides on membrane integrity (CFDA-AM, left column) and mitochondrial activity (alamar Blue, middle column) as measures of cell viability, as well as on ROS production (H 2 DCFDA, right column) in PC12 (o), SH-SY5Y (◊), MES23.5 (□) and N27 (Δ) cells following 24 h of exposure. The effects induced by particular pesticides differ between the various cell lines, resulting in different concentration-dependent changes in mitochondrial activity and membrane integrity. Data points in the concentration-response curves of the cell viability assays represent the average percentage membrane integrity (3A) or mitochondrial activity (3B) (±SEM; ≥ 9 wells from ≥3 independent experiments) compared to control. The data points in the ROS production curves (3C) represent average percentage ROS production (± SEM; ≥ 12 wells from ≥ 3 individual experiments) compared to time-matched control. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) and rotenone, whereas this cell line appears relatively insensitive in the membrane integrity and mitochondrial activity assays. This can likely be explained by differences in antioxidant levels and other coping strategies for oxidative stress between the cell lines. This is illustrated by the difference in results between imazalil and e.g. dinoseb as the imazalil-induced increase in mitochondrial activity only leads to a fairly mild ROS production in selected cell lines, whereas the changes in mitochondrial activity induced by dinoseb are paralleled by an increase in ROS production. Moreover, the hybrid cell line MES23.5 displays a particular, yet unexplained, sensitivity towards organochlorine insecticides as both exposure to lindane and dieldrin induced a change in membrane integrity solely in MES23.5 cells.
Thus, it appears that these cell lines differ considerably in their sensitivity towards toxicity for the different pesticides. It seems therefore justified to conclude that the research question should determine which model is the most appropriate since the intrinsic properties of a particular cell may strongly influence the outcome.
As exposure to all of these compounds is related to neurotoxicity related to changes in cellular function (intracellular Ca 2+ ), but not all are detected in general toxicity assays, measuring of cellular function appears pivotal to avoid false-negative results. Therefore, great care should be taken when interpreting toxicity data and the presented differences clearly highlight the need for thorough characterization of in vitro models. However, since the differential responses can be traced back to differences in membrane channel/receptor composition, using multiple cell lines may also provide mechanistic insight into the underlying mechanisms of action.