Arsenic Stimulates Myoblast Mitochondrial Epidermal Growth Factor Receptor to Impair Myogenesis

Abstract

Arsenic exposure impairs muscle metabolism, maintenance, progenitor cell differentiation, and regeneration following acute injury. Low to moderate arsenic exposures target muscle fiber and progenitor cell mitochondria to epigenetically decrease muscle quality and regeneration. However, the mechanisms for how low levels of arsenic signal for prolonged mitochondrial dysfunction are not known. In this study, arsenic attenuated murine C2C12 myoblasts differentiation and resulted in abnormal undifferentiated myoblast proliferation. Arsenic prolonged ligand-independent phosphorylation of mitochondrially localized epidermal growth factor receptor (EGFR), a major driver of proliferation. Treating cells with a selective EGFR kinase inhibitor, AG-1478, prevented arsenic inhibition of myoblast differentiation. AG-1478 decreased arsenic-induced colocalization of pY⁸⁴⁵EGFR with mitochondrial cytochrome C oxidase subunit II, as well as arsenic-enhanced mitochondrial membrane potential, reactive oxygen species generation, and cell cycling. All of the arsenic effects on mitochondrial signaling and cell fate were mitigated or reversed by addition of mitochondrially targeted agents that restored mitochondrial integrity and function. Thus, arsenic-driven pathogenesis in skeletal muscle requires sustained mitochondrial EGFR activation that promotes progenitor cell cycling and proliferation at the detriment of proper differentiation. Collectively, these findings suggest that the arsenic-activated mitochondrial EGFR pathway drives pathogenic signaling for impaired myoblast metabolism and function.

arsenic, epidermal growth factor receptor, mitochondria, signal transduction, SS-31, myogenesis, environmental toxicology, predictive toxicology

Arsenic exposure poses a significant risk of cardiovascular and metabolic (cardiometabolic) diseases in hundreds of millions of individuals worldwide (Kuo et al., 2017; Moon et al., 2017; Sung et al., 2015). Despite strong epidemiological evidence of these associations, the etiology for arsenic-promoted cardiometabolic diseases is unclear. Over the past decade, skeletal muscle metabolic dysfunction and decline of muscle composition have become a primary focus in establishing the major underlying risks for insulin resistance and type 2 diabetes (Goodpaster et al., 2003; Granados et al., 2019; Miljkovic et al., 2013), as well as cardiovascular disease and all-cause mortality (Miljkovic et al., 2015; Reinders et al., 2016). We previously reported that arsenic exposures increase ectopic skeletal muscle adiposity (Garciafigueroa et al., 2013), an earlier indication of insulin resistance and impaired metabolism than elevated serum free fatty acids or glucose (Goodpaster et al., 2000; Sell et al., 2006; Vigouroux et al., 2011). Arsenic also degrades muscle maintenance and regeneration (Ambrosio et al., 2014; Zhang et al., 2016). The fundamental mechanism for the effects of low to moderate arsenic exposures (10–100 ppb in drinking water) leading to muscle dysfunction and quality decline appears to be disruption of muscle and stem cell mitochondrial function and mitochondrial control of epigenetic regulatory programs (Ambrosio et al., 2014; Cheikhi et al., 2019). Indeed, arsenic reportedly increases DNA methyl transferase 3a (DNMT3a)-dependent DNA methylation of myogenic genes to impair muscle progenitor cell differentiation (Cheikhi et al., 2019; Hong and Bain, 2012; Steffens et al., 2011). However, it remains unclear how low levels of arsenic, which are stoichiometrically unlikely to affect mitochondrial respiration directly, signal for sustained mitochondrial dysfunction.

Low (nM) levels of arsenic activate receptor-initiated signal amplification cascades, which promote pathogenic phenotypic change and transformation in target cells (Andrew et al., 2009; Garciafigueroa et al., 2013; Simeonova et al., 2002; Straub et al., 2009). The epidermal growth factor receptor (EGFR) has long been a focus of investigation of receptor-mediated arsenic actions, especially those contributing to carcinogenesis (Andrew et al., 2009; Germolec et al., 1998; Simeonova and Luster, 2002; Simeonova et al., 2002) and stimulation of progenitor cell proliferation at the expense of differentiation (Patterson and Rice, 2007; Reznikova et al., 2010). Arsenic and other stressors stimulate noncanonical (ligand-independent) EGFR activation where proliferative signals are sustained as the receptor is not degraded upon ligand binding (Andrew et al., 2009; Simeonova et al., 2002; Tan et al., 2016). In cancer cells, activated EGFR scaffolds mitochondrial remodeling enzymes, disrupts respiration, increases reactive oxygen species (ROS) generation, and enhances cell growth and motility (Bollu et al., 2014; Demory et al., 2009; Li et al., 2017; Yao et al., 2010). Although the role of sustained mitochondrial EGFR activity has not been explored in muscle progenitor cells or skeletal muscle regeneration, it is clear that mitochondrial dysfunction impedes myogenic differentiation and muscle maintenance (Sahu et al., 2018; Tsitkanou et al., 2016; Wagatsuma and Sakuma, 2013). Dysfunctional mitochondrial metabolism also underlies development of myosteatosis (intramuscular adipose tissue; Gumucio et al., 2019), a pathogenic mechanism in the etiology of cardiometabolic diseases (Miljkovic et al., 2013, 2015; Reinders et al., 2016).

We previously reported that arsenic, in low nM concentrations, targets muscle, and progenitor cell mitochondria to disrupt muscle metabolism, maintenance, and regeneration (Ambrosio et al., 2014; Cheikhi et al., 2019). However, the mechanisms through which low nM levels of arsenic initiate and sustain abnormal mitochondrial morphology and function, as well as alter muscle progenitor cell fate decisions, are unknown. Thus, we investigated the hypothesis that arsenic requires receptor-mediated signal amplification to disrupt mitochondrial function and retain myoblasts in an activated, proliferative state, which ultimately impairs myogenic differentiation.

MATERIALS AND METHODS

Cell culture

C2C12 myoblasts were plated at 10 000 cells per cm² on glass coverslips and cultured until 80% confluent in growth medium (GM, Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum [Hyclone] and 1% penicillin/streptomycin [Invitrogen]) at 37°C and 5% CO₂. Myogenic differentiation was induced with differentiation medium (DM, Dulbecco’s modified Eagle’s medium supplemented with 2% horse serum and 1% penicillin/streptomycin) in the presence or absence of sodium arsenite (2–500 ηM, Fisher Scientific) and AG-1478 (1 µM, Fisher Scientific). At the end of the differentiation period, the cultures were fixed and imaged to quantify myotube formation relative to undifferentiated reserve cell (RC) proliferation (Nagata et al., 2014; Yoshida et al., 1998). In separate experiments, C2C12 cultured in noncoated 25 cm² flasks were differentiated in the absence or presence of 20 ηM arsenic for 4 days, and then mononucleated, undifferentiated RC were collected by mild trypsinization and separated from myotubes by filtering through a 40-µm strainer. The RC were either rinsed and fixed for flow cytometry or replated in arsenic-free GM on glass coverslips and allowed to proliferate for 3 days in the absence or presence of SS-31 or XJB-5-131 before fixing and immunofluorescent imaging.

Impedance sensing

Time- and frequency-dependent complex impedance, Z(t, f), of an electrode-cell interface were monitored using the Electric Cell-Substrate Impedance Sensing (ECIS) system (ECIS Zθ; Applied Biophysics). Cells were grown on 8-well ECIS arrays (8W10E+polyethylene terephthalate, standard thickness 0.25 mm; Applied Biophysics) containing forty 250-μm gold microelectrodes per well at a density of 15 × 10³ cells/well. Two ECIS arrays were recorded in parallel, for 16 wells per experiment. Real-time measurements were performed directly in the cell culture medium as both the ECIS arrays and the measurement station were kept in a humid incubator at 37°C and 5% CO₂. Multifrequency measurements were taken for each of the 16 wells at fixed time intervals in seconds. Z has a resistance component, R (t, f), and capacitive reactance component, Xc(t, f), that correspond to the real and empirically derived part of Z, respectively. Thus, impedance Z(t, f) = R(t, f) + iXc(t, f) and |Z(t, f)|² = R(t, f)² + Xc(t, f)². For each experimental condition, the mean Z(t, f) value and SEM were calculated at each time and frequency.

Quantitative flow and imaging flow cytometry analyses

To quantify receptor protein expression and phosphorylation, RC were harvested at the fourth day of differentiation, washed with phosphate-buffered saline (PBS) and fixed in 1.5% paraformaldehyde in PBS. The cells were permeabilized with 0.05% Triton X-100 in PBS for 15 min at room temperature, washed with PBS plus 0.5% Bovine serum albumin (PBB), and then incubated with primary antibody in PBB for 2 h at 37°C. For flow cytometric analysis of EGFR phosphorylation, cells were immunostained with rabbit antipY⁸⁴⁵EGFR (1:100, Cell Signaling No. 2231) and mouse anti-EGFR (1:100, containing mouse anti-EGFR (1:100, BD Laboratories No. 610017). After 3 washes with PBB, the samples were incubated for 30 min at room temperature with the secondary conjugated antibodies, Alexa Fluor 568-conjugated goat antirabbit IgG (1:500, ThermoFisher Scientific) or Alexa Flour 488-goat antimouse IgG (1:500, ThermoFischer Scientific). The labeled cells were then washed 3 times with PBB, and fluorescence quantified by flow cytometry using a Biosciences FACSCanto flow cytometer.

For colocalization experiments, cells were washed with PBS, fixed with 1.5% paraformaldehyde, and the cells were then processed as above with 1:100 dilutions of primary antibodies: Alexa Fluor-647 mouse antipY⁸⁴⁵EGFR (BD Biosciences clone 12A3), Alexa Flour 488 rabbit anti-CyclinD1 (Abcam ab190194), or Alexa Fluor 568 mitochondrial cytochrome C oxidase subunit II (MTCO2; 1:100, Cat No. A6404, Invitrogen) and the nuclear stain 4’,6-diamidino-2-phenylindole (DAPI). Cells were then analyzed by flow cytometry as above or using an Amnis ImageStream XMark II (Luminex Corporation). For ImageStream analysis, the samples were processed with filtered sheath buffers to ensure removal of large particulates and debris (≥ 1 µm) and imaged with a 60× objective at a resolution of 0.3 μm²/pixel. Both bright field and fluorescent images of the cells were captured using the INSPIRE software with the highest resolution (sensitivity) and lowest speed. Images and cell fluorescence were analyzed with an integrative technical computing framework with multiple machine-learning modules. Statistical analyses of the fluorescent signals were performed with R/Python and Wolfram programming languages.

Quantitative immunofluorescence imaging

Control and arsenic-primed RC were plated on coverslips in 24 well-plates or in 8-well chamber slides. After 3 days in culture, the cells were washed 3 times with Hank’s balanced salts solution (HBSS), fixed with 2% paraformaldehyde for 10 min at room temperature, and permeabilized with 0.1% Triton X-100 made in PBS solution for 15 min. The cells were then washed with PBS followed by PBB. Cells were then blocked with 5% donkey serum (Millipore, S30-100KC) diluted in PPB for 60 min, and then washed again 3 times with PBB before adding primary antibodies for 60 min. Primary antibodies recognized: Cyclin D1, (1:100, R&D Systems AF4196), pY⁸⁴⁵EGFR (1:150, Cell Signaling Technology 2231), and MTCO2 (1:200, ThermoFisher Scientific 12C4F12). Cell monolayers were washed with PBB and then species specific, fluorescent conjugated secondary antibodies were added for 60 min. These antibodies included: Alexa Fluor 488 donkey antirabbit IgG (H + L; 1:2000, Invitrogen, A21206), Alexa Fluor 647-conjugated donkey antigoat IgG (H + L;1:700, Jackson ImmunoResearch Laboratories 705-605-003); and Cy3-Conjugated donkey antimouse (1:800, Jackson ImmunoResearch Laboratories 715-165-151). All samples were additionally stained with DAPI (Biolegend) for 1 min. Slides were washed with PBS and coverslips mounted with Fluoromount-G (eBioscience 004958-02).

Z stack images were acquired by confocal microscopy using a Nikon A1 Confocal Laser Microscope with NIS-Elements C Software (Center for Biologic Imaging, Pittsburgh, Pennsylvania). Fluorescent intensities and Colocalization measurements were determined using Nikon Elements AR (Nikon Elements) and Imaris software. Statistical analyses of fluorescent signals were performed using GraphPad Prism v8.2 software.

Mitochondrial membrane potential

At the end of the differentiation protocol, RC were re-seeded in a 96-well black culture plate at 50 000 cells/well and incubated at 37°C and 5% CO₂ in a humidified chamber. After 6 h of equilibration, mitochondrial membrane potential was measured. Alternatively, RC were replated in a 96-well black culture plate at 10 000 cells/well and cultured for 4 days with or without 100 ηM of mitochondrial targeting SS-31 (Campbell et al., 2019; Szeto and Liu, 2018). At the end of the incubation periods, the culture medium was discarded, the cells were rinsed with PBS, and then incubated in assay medium (140 mM NaCl, 2.5 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂, and 20 mM 4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid, N-(2-Hydroxyethyl)piperazine-N′-(2-ethanesulfonic acid), pH = 7.4, mOsm = 300) containing an excess of JC-1 (1 μM). After 90 min of equilibration at 37°C and 5% CO₂, the ratio of the fluorescence of JC-1 aggregates formed as mitochondrial membrane potential (ΔΨm) increases (Ex: 353 nm, Em: 595 nm) to JC-1 monomer (Ex: 485, Em: 535) was then measured every 10 min over 3 h.

Measurement of mitochondrial superoxide

Cells were stained with 250 ηM of MitoSOX red (Ex: 488 nm/Em: 575 nm, Cat No. M36008, ThermoFisher) for 15 min at 37°C in the dark. The cells were then collected by trypsinization and washed with HBSS, fixed in 1.5% paraformaldehyde for 10 min, washed to remove any residual paraformaldehyde, and resuspended in 0.5 ml of PBS. Relative MitoSOX fluorescence intensity per cell was then quantified by flow cytometry.

RESULTS

Arsenic Stimulates EGFR to Attenuate Myogenesis

The most common exposures to arsenic come from drinking contaminated water, and humans drinking water containing 50 ppb arsenic have a blood concentration of approximately 90 ηM inorganic arsenic (Hall et al., 2006). Comparatively lower levels should distribute in skeletal muscle tissue. Thus, we used 2–500 ηM concentrations to examine arenite effects on C2C12 myogenic differentiation. We confirmed the observations of Bain et al. (Hong and Bain, 2012; Steffens et al., 2011) that the point of departure for impairing myoblast fusion into myofibers and myofiber length is between 20 and 50 ηM arsenite (Figs. 1A–C). We previously found that arsenic stimulates proliferation of undifferentiated, mononucleated RC (Cheikhi et al., 2019). As seen in Figures 1D and 1E, this proliferation decreased the relative proportion of multinucleated myotubes. To examine the role of EGFR in arsenite-stimulated RC proliferation, we added the selective EGFR tyrosine kinase inhibitor, AG-1478, with arsenite at the start of the differentiation protocol. The ratio of RC to differentiated myotubes in the cultures receiving AG-1478 with arsenic was the same as in control cultures (Figure 1). ECIS analysis provided label-free, real-time monitoring of changes in cell morphology and myoblast differentiation, as described by Park et al. (2016). Impedance curves generated during 64 h of C2C12 differentiation in the presence and absence of arsenic and/or AG-1478 showed distinct phases including: an acute increase, prolonged primary stabilization, a drop phase, and then secondary stabilization as the differentiating myotubes and RC became confluent (Figure 1F). Arsenic reduced impedance throughout all of the phases (Figure 1F), as quantified by linear regression analysis (Figure 1G). As in the optical assays, adding AG-1478 at the start of differentiation prevented arsenic from reducing impedance throughout differentiation. Taken together, these findings confirm that exposure to arsenic promotes RC proliferation and aberrant retention of “stemness” (Cheikhi et al., 2019), and implicate sustained EGFR activity in arsenic-impaired differentiation.

Figure 1.

Arsenic impairs myoblasts differentiation efficiency through an EGFR-dependent mechanism: (A–C) C2C12 cultures were exposed to the indicated range of arsenite concentrations during the myogenic differentiation protocol. The cultures were fixed and myotubes were immunostained antibody to MHC (green) and nuclei were stained DAPI (A). A fusion index was calculated (B) and tube length measured (C). Significant differences from control (0 ηM As) were determined using 1-way ANOVA with Dunnett’s test and difference at p < .01 is designated by ** (n = 6). D, Cultures were differentiated in the absence or presence of 20 ηM arsenite (As) and/or 1 µM AG-1478 (AG) for 4 days, and then fixed and brightfield images were captured (scale bar = 50 µm). The multinucleated myotubes and mononuclear RC were enumerated and the ratio of RC to myotubes calculated. The graph in (E) presents the mean ± SEM of the RC/myotube ratio and significant differences between groups was calculated using ANOVA followed by Tukey’s multiple comparison (**p < .01 from control, ^p < .05 from As). F, Real-time, label-free monitoring of C2C12 cells cultured with and without arsenic and/or AG-1478 throughout differentiation by measuring ECIS (Z) at 500 Hz. Impedance at 4 phases of cell behavior (increase [I], 1° stabilization [1°S], drop phase [DP], and 2° stabilization [2°S]) were measured. G, Linear spline regression (n = 3/group) and 1-way ANOVA with Dunnett’s test were used for statistical comparisons. All data are presented as mean ± SE (***p < .001 relative to control).

Arsenic Stimulates Prolonged EGFR Activation and Colocalization with Mitochondrial Regulatory Proteins

Given the protective effects of AG-1478, we next sought to investigate whether arsenic-stimulated noncanonical EGFR activation by quantifying EGFR phosphorylated on tyrosine 845 (pY⁸⁴⁵EGFR) in RC that were replated after the differentiation protocol and cultured in arsenic-free GM for 6 h. Src family kinase phosphorylation of EGFR tyrosine 845 is a hallmark of oxidant-mediated, noncanonical EGFR activation, as well as activated mitochondrial EGFR (Boerner et al., 2004; Bollu et al., 2014; Demory et al., 2009; Tan et al., 2016). As seen in Figure 2A, RC from C2C12 cultures exposed to arsenic during the differentiation protocol and replated in the absence of arsenic retained elevated pY⁸⁴⁵EGFR, relative to RC from the arsenic-free differentiation protocol. Flow cytometric analyses of the RC confirmed that arsenic exposure during the 4 days of differentiation increased pY⁸⁴⁵EGFR without affecting EGFR expression (Figure 2A). However, the flow cytometric analysis indicated that only a portion of the RC from the arsenic-exposed cell population retained activated EGFR.

Figure 2.

Sustained EFGR activation and mitochondrial localization in arsenic-primed RC. A, Protein levels of total EGFR and pY⁸⁴⁵EGFR were compared by flow cytometry in control RC (green) and arsenic-primed RC (red) replated for 6 h in the absence of arsenic. The bar charts report the mean + SEM of fluorescent intensity in cells from a combined 3 separate experiments. ***p < .001 by 2-tailed Student’s t test. B, Replated RC were immunostained with antibodies to pY⁸⁴⁵EGFR (green) and MTCO2 (red) and imaged with confocal microscopy (scale bar = 50 microns). Arsenic (As, 20ηM) and AG-1478 (AG, 1 µM) were added only in the differentiation phase of the experiment. C, Pearson’s colocalization coefficients for pY⁸⁴⁵EGFR and MTCO2 were calculated and the graph shows mean ± SEM for 6 separate cell images from each group. D and E, Pearson’s colocalization coefficients for pY⁸⁴⁵EGFR and MTCO2 were calculated 6 separate cell images of control or arsenic-primed RC, as well as control and arsenic-primed RC that received either XJB-5-131 (XJB, 100 ηM) or SS-31 (100 ηM) after being replated and cultured for 4 days. Statistical differences were determined by 1-way ANOVA, followed by Tukey’s post hoc test (**p < .1 different from control and ^^p < .01 different from arsenic).

Activated mitochondrial pY⁸⁴⁵EGFR binds to and phosphorylates the MTCO2 subunit of complex IV in the respiratory chain (Boerner et al., 2004; Demory et al., 2009). Quantitative immunofluorescent imaging demonstrated that arsenic-increased colocalization of pY⁸⁴⁵EGFR with MTCO2 (Figs. 2B and 2C) in the replated RC progeny, and adding AG-1478 with arsenic during differentiation prevented this sustained colocalization (Figs. 2B and 2C).

At the level of mitochondrial function, low-level arsenic exposure led to a sustained excess of mitochondrial superoxide production in RC progeny, as well as morphological change, over multiple divisional generations in arsenic-free GM; an indication of altered epigenetic memory (Cheikhi et al., 2019). Since elevated ROS is a noncanonical activator of EGFR, we confirmed a role of mitochondrial ROS and/or disrupted respiration in maintaining the pY⁸⁴⁵EGFR/MTCO2 colocalization by reversing arsenic effects with 2 mitochondrially targeted compounds that have separate mechanisms of reducing ROS. Control and arsenic-primed RC progeny (replated in arsenic-free medium) were cultured for 4 days with either XJB-5-131, a mitochondrial targeting catalytic electron scavenger (Javadov et al., 2015), or SS-31, a mitochondrial targeting peptide that preserves the integrity of the mitochondrial membrane phospholipid cardiolipin (Campbell et al., 2019; Szeto and Liu, 2018). Addition of either compound to arsenic-primed RC progeny reversed the arsenic-promoted pY⁸⁴⁵EGFR colocalization with MTCO2 to a level comparable to the control RC progeny (Figs. 2D and 2E and Supplementary Figure 1). Note that despite removing arsenite at the time of replating, pY⁸⁴⁵EGFR remained elevated in that arsenic-primed RC.

Cell Population Analysis of Arsenic-Activated pY845EGFR Colocalization with MTCO2

Consistent with our previous demonstration of arsenic effects on cellular memory (Cheikhi et al., 2019), the data in Figures 1 and 2 suggest that arsenic has greater effects on a subpopulation of myoblasts that resist differentiation and retain a proliferative phenotype. To interrogate the impact of arsenic on this subpopulation, we conducted a cell population-level quantitative multispectral flow cytometry image analysis using the Amnis ImageStream XMark II. This combined traditional quantitative flow cytometry gating strategies (ie, bright field and fluorescence intensity-based clustering) with image segmentation and subsequent morphological- and texture-based object characterizations at single cell-level resolution. Focusing on fluorescence intensity of direct fluorophore-conjugated antibodies against pY⁸⁴⁵EGFR and MTCO2, the data in Figure 3A demonstrated that the probability of a cell to contain colocalized pY⁸⁴⁵EGFR and MTCO2 (colocalization coefficient >2.5) was greatly increased in the RC progeny of arsenic-exposed cells. Again, treatment with AG-1478 reduced the probability of colocalization in the arsenic-primed RC progeny. SS-31 added in the differentiation protocol also attenuated the arsenic-stimulated increase in the portion of cells with colocalized pY⁸⁴⁵EGFR and MTCO2 (Figure 3A). Principle component analysis showed clear separation of the population of arsenic-primed RC progeny with increased colocalized pY⁸⁴⁵EGFR and MTCO2 from the control or intervention groups (Figure 3B).

Figure 3.

Single-cell analysis of pY⁸⁴⁵EGFR/MTCO2 colocalization. A, Representative bright field and fluorescence images of single cells captured by ImageStream XMark II flow cytometry and corresponding graphs of the probability of cell having colocalized pY⁸⁴⁵EGFR/MTCO2 in the different treatment groups. B, Principle component analysis of colocalization showing separation of arsenic-primed RC progeny from control, as opposed to the tendency of the intervention groups to cluster with controls.

Mitochondrial Dysfunction from Arsenic-Stimulated Mitochondrial pY845EGFR

Phosphorylation of MTCO2 by activated pY⁸⁴⁵EGFR causes mitochondrial morphological change and increased ROS generation as cytochrome C oxidase becomes dysfunctional (Demory et al., 2009). We previously observed that arsenic exposure during the differentiation protocol promotes mitochondrial remodeling, increased mitochondrial mass, increased ΔΨm, and increased mitochondrial ROS in the replated RC progeny (Cheikhi et al., 2019). Flow cytometric analysis with JC-1 dye demonstrated that treating cells with EGF caused a similar ΔΨm increase and that adding EGF with arsenic resulted in no more of a ΔΨm increase than adding either agent alone (Figure 4A). Treating the arsenic-exposed cells with AG-1478 or SS-31 during differentiation prevented the increased ΔΨm in the arsenic-primed replated RC progeny (Figs. 4B and 4C). Dysfunctional cytochrome C oxidase promotes cardiolipin oxidation and remodeling of the inner mitochondrial membranes (Szeto and Liu, 2018). Addition of either EGF or arsenic decreased the amount of reduced cardiolipin (10-nonyl acridine orange [NAO] staining) in the RC progeny (Figure 4D). Addition of EGF and arsenic together was again less than additive in oxidizing cardiolipin, suggesting that they act through the same mechanism. This mechanism is likely driven by activating mitochondrial EGFR, since AG-1478 attenuated mitochondrial ROS generation in the arsenic-primed progeny (Figure 4E).

Figure 4.

Dysfunctional mitochondrial effects of arsenic and EGF. A, Flow cytometric analysis of mitochondrial membrane potential using JC-1 ratiometric dye in arsenic (20 nM) or EGF (100 ηM) exposed RC, relative to control. Percentages show the relative abundance of cells in quadrant 2 (JC-1 aggregates) and quadrant 4 (JC-1 monomers). B, Quantification of JC-1 aggregates that indicate more positive membrane potential. AG-1478 (1 µM) was added to cells simultaneously with arsenic in the differentiation protocol and JC-1 fluorescence was measured in undifferentiated RC. C, SS-31 (100 ηM) was added to RC replated in arsenic-free medium and JC-1 fluorescence was measure after 3 days in culture. D, Flow cytometric analysis of cardiolipin oxidation (loss of NAO fluorescence) in RC treated with EGF or arsenic in the differentiation protocol. The percentage of the cell population in quadrant 3 (decreased NAO staining) is given. E, Mitochondrial superoxide was measured in undifferentiated control RC and RC exposed to arsenic and AG-1478 during the differentiation protocol. All experiments were repeated 3 times and group comparisons were made using ANOVA followed by Tukey’s post hoc test for significance (**p < .01, ***p < .001 relative to control, ^p < .05, ^^^p < .001 relative to arsenic).

Arsenic-enhanced RC Stemness and Proliferation Requires Activated EGFR and Mitochondrial Dysfunction

Arsenic exposure increases the “stem-like character” (ie, increased CD34 and CD133 expression) and proliferation of RC progeny, and this proliferative response was associated with RC populations with sustained cyclin D1 levels (Cheikhi et al., 2019). AG-1478 has been shown to inhibit EGFR-induced cyclin D1 upregulation and cell cycle progression (Liu et al., 2013; Zhu et al., 2001). Adding AG-1478 with arsenic during differentiation also decreased arsenic-increased nuclear cyclin D1 levels in replated RC progeny (Figs. 5A and 5B). Nuclear cyclin D1 levels in the arsenic-primed RC progeny were also returned to control levels by adding either XJB-5-131 or SS-31 to the replated RC progeny (Figs. 5C and 5D and Supplementary Figure 2). Taken together, these data suggest that the adverse effects of arsenic on myogenesis may be sustained through noncanonical mitochondrial EGFR activation, which is known to dysregulate cyclin D1-dependent myogenic cell cycle progression (Owusu-Ansah et al., 2008).

Figure 5.

Arsenic effects on EGFR and mitochondrial regulation of nuclear cyclinD1. A, Representative images of cyclin D1 immunofluorescence in undifferentiated RC replated in arsenic-medium. AG-1478 was added simultaneously with arsenic in the differentiation protocol. B–D, Quantitative comparison of nuclear cyclin D1 levels in replated RC. Groups of cells in (C) and (D) were treated with SS-31 or XJB-5-131 after replating and all groups in (B–D) were fixed and analyzed after 3 days in arsenic-free culture. Representative images of cells from (C) and (D) are provided in Supplementary Figure 2. All experiments were repeated 3 times and the data are presented as mean ± SEM of relative fluorescence per cell captured in 5 separate images. Group comparisons were made using ANOVA followed by Tukey’s post hoc test for significance (*p < .05, **p < .01 relative to control, ^p < .05, ^^p < .01 relative to arsenic).

Classifying Cellular Arsenic-Induced Mitochondrial Signaling as an Adverse Outcome Pathway

Mitochondrial adverse outcome pathways have been proposed as means to link mitochondrial dysfunction to organismal and population health in the context of adverse environmental exposures (Dreier et al., 2019). As the myoblasts are inherently heterogeneous and we observed subpopulation specific responses to arsenic (Figs. 2 and 3), we used machine-learning algorithms to assess the robustness of dysfunctional arsenic-induced mitochondrial EGFR signaling, as well as efficacy of the drug interventions in preventing or reversing arsenic effects. First, the analytical workflow was extended with several supervised machine-learning classifiers, to analyze myoblast population heterogeneity based on response to arsenic and the interventions using the relative expression levels of pY⁸⁴⁵EGFR, MTCO2, and cyclin D1, as well as the quantitative colocalization of pY⁸⁴⁵EGFR/MTCO2. Of all supervised classifiers tested, gradient boosted trees and random forest classifiers had the highest predictive accuracy (≥70%; Figure 6, Table I) . These data suggest that arsenic exposure is predicted a by significant mitochondrial translocation of pY⁸⁴⁵EGFR and/or increase tyrosine phosphorylation of mitochondrial membrane proteins. To capture the maximum heterogeneous spectrum of arsenic effects on mitochondrial dysfunction, we improved the accuracy of the predictive model using a bootstrapping method. The empirical dataset was iteratively resampled to randomly increase the number of training observations. This approach yielded a classification accuracy of over 99% in the entire empirical data (n = 40 507). The larger numbers of training observations (n = 648 100; Figure 6, Table II) best captured the population-level influence and predictive power of the pY⁸⁴⁵EGFR-MTC02 signaling pathway in mediating the arsenic response. Confusion matrices are presented that summarize and validate the machine-learning predictions (Figs. 6A and 6B). Next, we used information- and statistically based scoring methods to determine which marker ranked highest in the ability to distinguish between arsenic exposed and nonexposed cells, with or without drug interventions. Information gain versus Gini coefficient and ANOVA versus χ². As revealed mitochondrial pY⁸⁴⁵EGFR as the most discriminative adverse outcome pathway biomarker of arsenic exposure in our model (Figure 6, Table III).

Figure 6.

Machine-learning modeling of arsenic adverse outcomes. Machine-learning models were built using different classifiers of arsenic signaling data in order to predict whether an individual cell belonged to the control conditions, an arsenic group, and/or an intervention group. The input data consisted of the pY⁸⁴⁵EGFR, MTCO2, or cyclin D1 protein expression levels and pY⁸⁴⁵EGFR/MTOC colocalization. The capacity of these single-cell markers to discriminate the experimental conditions was comparatively tested by several classifier models using cross-validation of the training set according to the various metrics. Table I: The training set (approximately two-third of the experimental data) was used to construct features attuned to the molecular fingerprint of the noncanonical EGFR pathway and combine them into a generalizable classifier model learned on the empirical dataset, and applied to the test dataset (ie, the remaining one-third of the experimental data) to evaluate the model’s predictive accuracy. Table II: To account for the potential statistical limitations caused by the relatively small training dataset and possible misrepresentation of the overall cell population heterogeneity, a bootstrapping approach of the original empirical data was utilized to generate a larger sample size. The increased accuracy of estimating errors in the classified data is described by the confusion matrices (A and B). Table III: The condition-labeled datasets and scores of the classification attributes according to their correlation with the class using Gini coefficient, information gain, ANOVA, and χ².

DISCUSSION

Skeletal muscle homeostasis and regeneration is critically dependent on the balance between muscle stem cell selfrenewal and commitment to differentiation. Here, we demonstrate that arsenic exposure disrupts this critical balance to impair myogenesis, and identify mitochondrial EGFR activation as an upstream mechanism for arsenic dysregulation of RC fate. Indeed, low-level arsenic stimulation of mitochondrial pY⁸⁴⁵EGFR was associated with remodeling of the mitochondrial membranes, increased ΔΨm, increased mtROS, and proliferative signaling at the expense of myogenic differentiation. This mechanism for low-level arsenical effects on mitochondrial function and cell fate varies greatly from mechanisms through which much higher levels of arsenic inhibit mitochondrial ATP generation and promote cell death.

Skeletal muscle maintenance and regeneration depend on resident stem cells (satellite cells, MuSC), that are normally found in a quiescent state (Mauro, 1961; Schultz and McCormick, 1994). In response to injury, MuSCs activate, enter a state of transient amplification, and begin to express muscle regulatory transcription factors (MRFs; (Charge and Rudnicki, 2004; Hawke and Garry, 2001)). The commitment to terminal differentiation occurs with expression of the MRF, myogenin, and degradation of the transcriptional repressor of differentiation, Pax7 (Ciciliot and Schiaffino, 2010; Olguin et al., 2007). In vivo, EGFR is activated in adult MuSCs and RCs as they exit quiescence in the early stage of myogenic differentiation (Golding et al., 2007; Nagata et al., 2014) and this response is recapitulated in differentiating myoblasts in vitro (Andrechek et al., 2002; Ford et al., 2003; Harper and Buttery, 2001; Kim et al., 1999). However, prolonged EGFR activity inhibits myogenic differentiation as the activated MuSCs are locked in a proliferative state (Golding et al., 2007; Leroy et al., 2013). Although most attention has focused on ligand-dependent canonical action of plasma membrane EGFR, noncanonical, ligand-independent EGFR signaling in different cellular organelles is now recognized to be a primary driver of pathogenic cell proliferation and proliferative diseases (Demory et al., 2009; Tan et al., 2016).

Murine C2C12 myoblasts provide a useful model for studying the consequence of adverse EGFR activation in quiescent MuSCs (Charville et al., 2015; Golding et al., 2007; Leroy et al., 2013; Nagata et al., 2014; Yoshida et al., 1998). C2C12 RC display key characteristics that are shared by MuSC in vivo, including quiescence, activation, self-renewal and generation of myotubes (Yoshida et al., 1998). The C2C12 differentiation model was shown to be extremely sensitive to arsenic and arsenic-induced epigenetic repression of myotube differentiation (Hong and Bain, 2012; Steffens et al., 2011). These previous studies found that ηM arsenic suppressed MRF activation with increased DNMT3a-dependent methylation of the Myod promoter. We found that ηM arsenic induces C2C12 Dnmt3a through mitochondrially regulated epigenetic changes, and that this epigenetic regulation enhances C2C12 proliferation over myogenic differentiation (Cheikhi et al., 2019). These findings suggest that low, ηM levels of arsenic promote signal amplification through mitochondrial EGFR and its downstream pathways that disrupt respiratory chain activity, promote mitochondrial remodeling, and alter mitochondrial to nuclear communication.

In addition to increasing DNMT3a repression of MRF expression, arsenic may suppress differentiation through EGFR and mitochondrial-dependent temporal dysregulation of cyclin D1 (Figure 5). We previously found that arsenic-induced temporal dysregulation of cyclin D1 oscillatory behavior impedes adipogenic stem cell differentiation (Beezhold et al., 2017). In these studies, arsenic-increased nuclear cyclin D1 levels were attenuated by AG-1478 or reversed by adding XJB-5-131 or SS-31 to arsenic-primed RC. Sustained nuclear cyclin D1 affects self-renewal and differentiation in MPC by preventing MyoD binding to DNA (Wei and Paterson, 2001). Growth factor-stimulated nuclear cyclin D1 bound to cyclin-dependent kinase 4 is essential for allowing MPCs to choose to remain in cell cycle or exit to terminal differentiation (Wei and Paterson, 2001). Thus, by sustaining mitochondrial EGFR activity and its consequent effects on mitochondrial function, arsenic likely prevents the fall in cyclin D1 in G0 of cell cycle that allows MyoD and other MRF complexes to transactivate differentiation.

Predicting and understanding adverse low-dose arsenic-induced cytotoxicity and its role in pathogenic tissue remodeling is a global public health challenge. Moving toward machine learning to parse arsenic adverse outcome pathways is a step toward meeting that challenge. Conventional approaches to establish pathogenic mechanisms of arsenic in different tissues and modes of action for promoting different disease endpoints are constrained by time, costs, and species differences. In contrast, fast computational approaches can be employed to create predictive models and adverse outcome pathways by maximally exploiting prior experimental data. This reduces the number of extraneous experimental tests by better predicting which hypothesis-driven experiments are needed, as well as potential intervention strategies. Mitochondrial signaling and function has been proposed as a fundamental foundation for adverse outcome pathways modeling to link metabolic dysfunction to organismal and population health outcomes in the context of adverse environmental exposures (Dreier et al., 2019). By applying machine-learning tools to data collected from the well-established, arsenic-responsive C2C12 model, we quantified 4 signaling parameters describing typical phenotypic changes driven by arsenic exposure and sustained for multiple cell generations after removal of the arsenic. As primary myogenic cells and in vivo responses are inherently more sensitive than cell lines, the population effects of arsenic could potentially be even more heterogeneous. Although many Gaussian, normal distribution statistics are commonly used to average response intensities, they may be replaced by more statistically robust and predictive analyses to provide deeper insight into the mechanistic basis of arsenic pathogenesis and mode of action. Similar machine-learning approaches are leading toward more efficient study of health risks from a range of environmental exposures (Dreier et al., 2019; Luechtefeld et al., 2018).

The full health impacts of arsenic effects on muscle mitochondrial function and impaired regeneration are unknown. There have been only a few epidemiological studies focused specifically on arsenic-impaired muscle function. Parvez et al. found that even low to moderate level arsenic exposures (<100 ppb) cause strength and motor deficits in children (Parvez et al., 2011). Others found rates of muscle morbidity, atrophy, and weakness in the range of 35%–85% in populations exposed to very high arsenic levels (Chakraborti et al., 2003; Mukherjee et al., 2003). The effects of arsenic on sensory motor coupling and neural maintenance of muscle are a large component of these morbidities. We found direct effects of arsenic on muscle strength, muscle regeneration and muscle compositional change, as well as on isolated muscle cells that are likely independent of neural input (Ambrosio et al., 2014; Cheikhi et al., 2019; Garciafigueroa et al., 2013; Zhang et al., 2016). Our findings of low-level arsenic-induced compromise of muscle mitochondrial function, bioenergetics, and compositional change toward ectopic adiposity (myosteatosis) may explain an underlying mechanism for arsenic-promoted diseases beyond skeletal muscle defects.

Loss of muscle quality is critically important to overall well-being as highlighted by the fact that skeletal muscle comprises 40%–50% of total body metabolism, and that myosteatosis is associated with increased risk of all-cause and cardiovascular disease mortality (Carobbio et al., 2011; Goodpaster et al., 2000; Miljkovic and Zmuda, 2010; Prado et al., 2018), as well as diabetes (Miljkovic et al., 2013). Thus, continued findings of direct effects on low-level arsenic on muscle metabolism, function, and composition warrant further population-level investigation of the role of these environmentally derived pathologies in the etiology of arsenic-promoted diseases. The finding that mitochondrial protectants can mitigate and reverse arsenic-promoted pathogenic determination of progenitor cell fate has broad implications for developing mitochondrially directed strategies to treat disease in arsenic-exposed populations.

SUPPLEMENTARY DATA

Supplementary data are available at Toxicological Sciences online.

FUNDING

The National Institutes of Health (NIH) National Institute on Aging grant (AG052978-01 to F.A.), the Pittsburgh Claude D. Pepper Older Americans Independence Center (P30 AG024827 to A.C.), National Institute of Environmental Health Sciences grant (R01ES023696 to F.A. and A.B. and R01ES025529 to F.A. and A.B.) supported this work. We acknowledge the NIH-supported microscopy resources in the Center for Biologic Imaging (NIH grant number 1S10OD019973-01). Finally, flow cytometry was performed in the University of Pittsburgh Unified Flow Core using an ImageStreamX MARKII (NIH grant 1S10OD019942-01).

DECLARATION OF CONFLICTING INTERESTS

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Author notes

Amin Cheikhi and Teresa Anguiano contributed equally to this study.