Palmitic acid induces neurotoxicity and gliatoxicity in SH-SY5Y human neuroblastoma and T98G human glioblastoma cells

Introduction

Obesity is now a global health issue that presents a major risk for serious diet-related noncommunicable diseases including diabetes mellitus, cardiovascular diseases, hypertension, stroke, and some cancers (World Health Organization, 2018). High-fat diet (HFD) rich in saturated fatty acids (SFAs) has long been recognised to contribute to many adverse metabolic health issues. In the past two decades, it was found that HFD-induced obesity has been associated with neuroinflammation and reactive gliosis (Dorfman & Thaler, 2015), leading to CNS pathologies such as neurodegenerative diseases (Guillemot-Legris & Muccioli, 2017). A case-control study demonstrated a link between the intake of fat from animal sources rich in SFA and PD (Logroscino et al., 1996). Several prospective studies also showed association of body mass index (BMI) with increased risk of Alzheimer’s disease (Gustafson et al., 2003) and PD (Hu et al., 2006); and on the contrary, lower BMI is associated with lower risk of PD (Sääksjärvi et al., 2014). Furthermore, the first study in humans using positron emission tomography with [¹¹C]-palmitate and [¹⁸F]fluoro-6-thia-heptadecanoic acid showed increased fatty acid (FA) uptake and accumulation in the brain of obese subjects with metabolic syndrome (Karmi et al., 2010). This suggests that FAs are able to cross the blood–brain-barrier and are able to be taken up by brain cells. In fact, peripheral FAs were shown to have relationship with central FAs, as reported by a study done on human whole blood and cerebrospinal fluid (Guest et al., 2013).

PA, a long chain 16:0 SFA, is the most common FA found in animals and plants such as palm oil, palm kernel oil, butter, cheese, milk and meat (Gunstone, Harwood & Dijkstra, 2010). Despite its crucial biological functions such as energy storage, acting as a signalling molecule and maintaining integrity of cellular membranes (Gunstone, Harwood & Dijkstra, 2010), PA has been found to be increased in diseased brains. Particularly, PA level appeared to be increased in the frontal cortex lipid rafts in PD (Fabelo et al., 2011); and in parietal cortex in AD (Fraser, Tayler & Love, 2010). Furthermore, PA was found to cause lipotoxicity to several cell lines in vitro. For instance, PA triggered the release of tumour necrosis factor-α and interleukin-6, activating inflammatory signalling in astrocytes (Gupta et al., 2012). PA also induced apoptosis in human hepatoma cell line, HepG2 (Zhang et al., 2004), neural progenitor cells (Park et al., 2011) and neuronal cell line, SH-SY5Y (Hsiao et al., 2014).

Other than neurons, the brain also comprises and depends on surrounding non-neuronal cells such as glial, epithelial cells, pericytes and endothelia, for them to function correctly (Freire-Regatillo et al., 2017). Glial cells were once thought of to serve as a supportive system for neurons. Now they are found to possess modulatory, trophic and immune functions in the brain (Gupta et al., 2012). Astrocyte is a type of glial cell, which is the most plentiful and varied non-neuronal cell in the CNS (Argente-arizón et al., 2015). Of note, marked astrogliosis was observed in the hypothalamus of HFD-induced or genetically obese mice (Buckman et al., 2013), suggesting that astrocyte plays a role in reactive gliosis leading to CNS pathologies.

A family of neuronal proteins that are implicated in both obesity and CN pathologies is the synucleins, i.e.,  α- and γ-synucleins (α-syn, γ-syn). They are small, soluble, highly conserved proteins, predominantly expressed in the neurons. α-syn is a major component of Lewy bodies and Lewy neurites appearing in the postmortem brain of PD and other synucleinopathies (Spillantini & Goedert, 2000). Genetic mutations in α-syn, including point mutations-A53T, A30P and E46K, have been linked to familial PD (Polymeropoulos et al., 1997). Neuronal expression of either human wild-type (wt) or PD-related mutant α-syn induces neurodegeneration associated with pathological accumulations of α-syn. Previous studies also revealed that α-syn-containing inclusion bodies were present in astrocytes of sporadic PD (Wakabayashi et al., 2000; Braak, Sastre & Del Tredici, 2007) and overexpression of wt α-syn in U373 astrocytoma cells induces of astroglial apoptotic cell death (Stefanova et al., 2001). Moreover α-syn and γ-syn were found to affect lipid uptake and metabolism in brain and adipocytes (Golovko et al., 2005; Millership et al., 2013; Hsiao, Halliday & Kim, 2017).

Given the involvement of FAs particularly PA in CNS pathologies and the possible role of α-syn and γ-syn in modulating the effects of FAs, therefore, the objective of this study was to first evaluate the effects of PA, oleic acid (OA; long chain FA with lipid number of C18:1 cis-9 and a major constituent in plant oil such as olive oil, almond oil, pecan oil and canola oil) and lauric acid (LA; medium chain 12:0 SFA which comprises about 50% of FA content in coconut oil, coconut milk, laurel oil and palm kernel oil) on the viability of human neuroblastoma SH-SY5Y and human glioblastoma T98G cell lines. SH-SY5Y cells were selected for the experiments as they have been widely used as a cell model of dopaminergic neurons for PD research (Xie, Hu & Li, 2010), while T98G cells were selected due to its biological resemblance with primary astrocytes and its broad use in research as an astrocyte cell model (Avila Rodriguez et al., 2014; Cabezas et al., 2015; Avila-Rodriguez et al., 2016). The effects of stable overexpression of γ-syn in SH-SY5Y and transient overexpression of α-syn (wt and PD mutants A53T, A30P and E46K) in SH-SY5Y and T98G cells were also evaluated. We found that PA is neurotoxic and gliatoxic to SH-SY5Y and T98G cells, respectively. To investigate the potential synergistic effect of environmental factors for dopaminergic neurotoxicity, SH-SY5Y cells were co-treated with PA (to mimic HFD exposure), and increasing concentrations of paraquat (PQ), a herbicide that is implicated in the development of PD (Pezzoli & Cereda, 2013). Since leptin, a hormone that is involved in the brain-adipose axis, has been shown to have neuroprotective effect in SH-SY5Y cells (Russo et al., 2004; Lu et al., 2006; Weng et al., 2007), we also investigated whether leptin pre-treatment could rescue SH-SY5Y cells from PA neurotoxicity. The mode of cell death induction by PA in SH-SY5Y and T98G was investigated using Annexin V/PI staining followed by flow cytometry analysis. Lastly, to attribute whether apoptotic cell death is caused by oxidative stress, intracellular ROS and extent of lipid peroxidation (TBARS level) were assessed.

Materials and Methods

Results

PA, but not OA and LA, is neurotoxic and gliatoxic

Generally, PA induced cytotoxicity towards SH-SY5Y and SH-SY5Y-γ in a time- and dose-dependent manner at concentrations above 300 µM at 24 h, and at concentrations above 200 µM at 48 h. Figures 1A and 1B show the effect of PA treatment on the cell lines for 24 and 48 h. At 24 h, the percentage of viable cells decreased dramatically at PA concentrations ≥300 µM, while at 48 h, the cell viability decreased drastically at PA concentrations ≥200 µM. Except at 500 µM, the viability of SH-SY5Y-γ cells was significantly higher than SH-SY5Y cells at concentrations ≥300 µM (24 h) and at concentrations ≥200 µM (48 h). The concentrations of 50% cytotoxicity (LC₅₀) for SH-SY5Y and SH-SY5Y-γ were 420 and 440 µM, respectively, at 24 h, and 320 and 380 µM, respectively, at 48 h. At lower concentrations, PA did not cause cytotoxicity but promoted the viability of SH-SY5Y-γ instead at 48 h.

The treatment of OA for 24 and 48 h generally increased the cell viability in both SH-SY5Y and SH-SY5Y-γ, as shown in Figs. 1C and 1D. The cell viability increased as a function of OA concentration and reached the peak of ∼140% at 400 µM and 300 µM OA for SH-SY5Y and SH-SY5Y-γ, respectively. For the treatment of LA, the cell viability of SH-SY5Y, but not that of SH-SY5Y-γ, was generally increased, as shown in Figs. 1E and 1F. The results indicate that LA promoted the viability of SH-SY5Y but had no effect on SH-SY5Y-γ at concentrations <300 µM and exerted significant decrease in cell viability (8–18%) at ≥400 µM at 48 h.

Similar with SH-SY5Y cells, PA at lower concentrations (<200 µM), increased the viability of astrocytic T98G cells, and while at higher concentrations (≥200 µM), reduced the cell viability significantly, in a dose-dependent manner (Fig. 1G). The LC₅₀ of PA in T98G cells was 320 µM. Like in SH-SY5Y cells, OA also increased the cell viability of T98G at concentrations ≥200 µM, while LA did not affect the cell viability significantly (Fig. 1G).

Transient transfection of α-syn and its PD mutants led to modest rescue from the neurotoxicity of PA

To investigate the gene-environment interaction in affecting neurotoxicity and gliatoxicity, SH-SY5Y and T98G cells were transiently transfected with wt, A30P, E46K or A30P α-syn for 48 h before further treated with LC₅₀ of PA (320 µM for both SH-SY5Y and T98G cells). Instead of exacerbating the cytotoxic effect of PA, transient overexpression of wt α-syn or its PD mutants (A30P and E46K, but not A53T) significantly rescued SH-SY5Y cells from the toxicity of PA, by increasing their viability to up to 10% (Fig. 2A). The similar phenomenon could also be observed in T98G cells, where transient overexpression of A30P and E46K significantly increased the cell viability to around 10%, compared with the untransfected cells treated with PA (Fig. 2B). Furthermore, transient overexpression of wt, A30P or E46K α-syn also increased the viability of SH-SY5Y cells (but not T98G) to up to 20% when subjected to OA treatment (Fig. 2A).

Since the discrepancy in the cytotoxic effect of α-syn overexpression might be due to the transfection method used (transient in this study vs. stable in Stefanova et al., 2001), we investigated whether the neurotoxicity results of PA in stable SH-SY5Y-γ cells could be replicated by transient transfection of γ-syn in SH-SY5Y and T98G cells. Indeed, transient transfection of γ-syn did not significantly increase the cell viability of SH-SY5Y cells when subjected to PA treatment (compared with results in Fig. 1B). In short, transient transfection of α-syn or γ-syn will have different effects in SH-SY5Y or T98 cells compared with stable transfection.

PA induces neurotoxicity and gliatoxicity via apoptotic cell death

To investigate the mode of cell death induced by PA in SH-SH5Y and T98G cells, dual staining with Annexin V and PI was performed and cells were analysed by flow cytometry. The results showed that PA at lower concentrations did not induce apoptosis in SH-SY5Y cells (less than 20% cells undergoing apoptosis) and T98G cells (less than 5% cells undergoing apoptosis), but at 300 µM, PA significantly increased the percentage of apoptotic cells to about 2 fold in both cell lines (Figs. 3A, 3B and 3E). Single staining with PI for cell cycle analysis by flow cytometry also revealed that DNA fragmentation was statistically significant in 200 µM and 300 µM PA-treated SH-SY5Y and T98G cells (Figs. 3C, 3D and 3F). For SH-SY5Y cells, the increment in the percentage of DNA fragmentation as indicated by the R1 gate was 1.3-fold (22.56 ± 0.53% to 28.23 ± 0.03%) and 2.6-fold (from 22.56 ± 0.53% to 57.75 ± 0.19%) for 200 µM and 300 µM PA treatments, respectively (Fig. 3F). While for T98G, the increment was 1.5-fold (from 6.74 ± 0.17% to 10.26 ± 0.22%) and 4.1-fold (from 6.74 ± 0.17% to 27.66 ± 0.35%) for 200 µM and 300 µM PA treatments, respectively (Fig. 3F). These were associated with the decrease in cell percentages in the other cell cycle phases.

Discussion

MTT assay results revealed that only PA, but not OA and LA, is cytotoxic to all of the three cell lines; namely, SH-SY5Y, SH-SY5Y- γ and T98G. The range of FA concentrations used was within the physiological range of circulating free FAs in human plasma, which is ∼0.1–1.0 mM (Yuan et al., 2013). The rate of PA β-oxidation was also found to increase as a function of PA concentration, from 15 µM to 2 mM, and peaked at 200 µM, in homogenates of astrocytes cultured from neonatal mouse brain (Murphy et al., 1992). The similar range of FA concentrations had been used in many studies, and the cytotoxic effect of PA observed was also similar to that reported earlier (Gupta et al., 2012; Yuan et al., 2013; Hsiao et al., 2014). Dose- and time-dependent apoptotic effects were observed in SH-SY5Y after PA treatment (Kim et al., 2011). Further elucidation of pathways leading to PA-induced apoptosis revealed that endoplasmic reticulum (ER) stress, as indicated by the expression of spliced X-box binding protein 1 mRNA and immunoglobin heavy chain-binding protein, was involved (Kim et al., 2011). In addition, PA inactivated 5′ adenosine monophosphate-activated protein kinase (AMPK), and re-activation of AMPK by N1-(β-D-Ribofuranosyl)-5-aminoimidazole-4-carboxamide, ameliorated PA-induced cytotoxicity with diminished ER stress-mediated apoptosis (Kim et al., 2011). Correspondingly, Hsiao et al. (2014) also showed that PA triggered SH-SY5Y apoptotic cell death via protein palmitoylation, which induced G₂/M cell cycle arrest and ER stress.

On the other hand, OA was shown to promote the cell growth in both neuronal SH-SY5Y and astrocytic T98G cell lines, parallel to previous studies in which OA was found to have neuroprotective and neurotrophic effects. For example, OA was found to be superior to docosahexaenoic acid or linoleic acid in the protection of mouse neuroblastoma Neuro-2a and primary rat cortical neurons against PA- or ceramide-induced cytotoxicity (Kwon, Lee & Querfurth, 2014). OA pre-treatment attenuated PA-induced mitochondrial dysfunction and insulin resistance by inhibiting the phosphorylation of mitogen-activated protein kinase and nuclear translocation of NF-κB p65 induced by PA (Kwon, Lee & Querfurth, 2014). It also stimulated adenosine triphosphate (ATP) production, decreased mitochondrial superoxide generation, increased triglyceride and blocked PA-induced diacylglycerol accumulation, with the involvement of protein kinase A and peroxisome proliferator-activated receptor gamma coactivator-1α (Kwon, Lee & Querfurth, 2014). In addition to its neuroprotective effect, OA was found to be synthesised and released by astrocytes with the stimulation of albumin (Tabernero et al., 2001). Released OA was then incorporated into neurons for the synthesis of phospholipids, and stimulated neuronal differentiation by promoting axonal growth, neuronal clustering and expression of axonal growth-associated protein-43 (Tabernero et al., 2001). The neurotrophic effect of OA involved the binding to its receptor peroxisome proliferator-activated receptor-α (Bento-Abreu, Tabernero & Medina, 2007).

The same cytotoxic effect was also observed in the astrocytic cell line, T98G. This PA-induced cytotoxicity was also reported previously in primary cultured mouse astrocytes (Wang et al., 2012) and rat cortical astrocytes (Wong et al., 2014). Wang et al. (2012) showed that PA-induced apoptosis in primary mouse astrocytes involved the rise in Bax/Bcl-2 ratio and caspase-3 activation. It involved ROS generation and subsequent mitochondrial membrane potential collapse (Wang et al., 2012; Wong et al., 2014). PA was also found to induce inflammatory response in astrocytes via Toll-like receptor 4, by releasing pro-inflammatory cytokines like tumour necrosis factor-α and interleukin-6 (Gupta et al., 2012). The inflammatory response was also observed with the treatment of other SFAs, namely, LA and stearic acid, but not unsaturated FAs (like OA) (Gupta et al., 2012).

PA induced cytotoxicity in both SH-SY5Y and SH-SY5Y-γ, albeit to a lesser extent in the latter, as indicated by the higher LC₅₀. As a marker of cancer progression, overexpression of γ-syn in ovarian cancer cell lines, A2780 and OVCAR5, was found to enhance tumorigenicity by constitutively activating Extracellular Signal-regulated Kinase-1/2 and down-regulating c-Jun N-terminal kinase 1 in response to a host of environmental stress signals (Pan et al., 2002). In human breast cancer cell lines T47D and SKBR3, knockdown of γ-syn sensitized the cells to ER stress-induced apoptosis dependent on c-Jun N-terminal kinases or caspase-3 and caspase-7 activation (Hua et al., 2009). Consistent with these, our study suggests an anti-apoptotic role of γ-syn in response to PA-induced cytotoxicity, since the two cells lines used are cancer cell lines.

α-syn and γ-syn, except A53T α-syn, generally increased the cell viability of both SH-SY5Y and T98G cells after PA treatment. Previously, Da Costa, Ancolio & Checler (2000) reported that wt α-syn, but not A53T, had anti-apoptotic effect against staurosporine, etoposide and ceramide C in murine neuronal cell line TSM1. Similarly, our group reported that only wt α-syn, but not the mutant variants, protected SH-SY5Y from the toxicity of rotenone and maneb (Parkinsonian pesticides), by attenuating mitochondrial membrane potential changes and ROS level (Choong & Say, 2011). On the other hand, overexpression of α-syn led to oxidative stress-mediated apoptotic death in U373 astrocytoma cells (Stefanova et al., 2001). In this study, we found that transient overexpression of not only wt, but also A30P, E46K α-syn and γ-syn, were able to ameliorate (albeit modestly) PA-induced cytotoxicity in SH-SY5Y and T98G cells. The discrepancy in the cytotoxic effect of α-syn overexpression might be due to the transfection method used (transient in this study vs. stable in Stefanova et al., 2001). A further confirmatory study involving stable overexpression of α-syn and its PD mutants in SH-SY5Y and T98G cells is needed to further elucidate whether α-syn protects against PA-induced cytotoxicity, since α-syn has been shown to bind to the surface of triglyceride-rich lipid droplets and protects them from hydrolysis in cultured cells as well as in primary neurons (Cole et al., 2002).

The co-treatment of PA and PQ brought about intensified cytotoxicity in SH-SY5Y cells as compared to PA or PQ treatment alone. There is no previous study reporting on the interaction between PA and PQ in dopaminergic neurotoxicity, but rotenone, another Parkinsonian environmental toxin which inhibits mitochondrial complex I, was found to increase the incorporation of radioactively-labeled PA into acetyl coA by β-oxidation in SH-SY5Y cells (Worth et al., 2014). In bovine cerebral mitochondria, PQ was reduced to the PQ radical via complex I in mitochondria, leading to accelerated lipid peroxidation, an effect similar to that triggered by rotenone (Fukushima et al., 1995). It is predicted that the PQ-induced lipid peroxidation is further enhanced by PA, which serves as a substrate for carnitine palmitoyltransferase-1-dependent mitochondrial β-oxidation, a process that leads to enhanced ROS production (Magtanong, Ko & Dixon, 2016). In addition, PA treatment also triggers mitogen-activated protein kinase- and caspase-dependent signaling pathways leading to apoptosis in neuronal N2a cells (Kwon, Lee & Querfurth, 2014). PQ was found to trigger SH-SY5Y neuronal cell apoptosis via the intrinsic mitochondrial pathway by the upregulation of p53 protein and subsequently, its target pro-apoptotic Bax protein (Yang & Tiffany-Castiglioni, 2008). The impairment of mitochondria complex I activity was followed by the release of cytochrome c, increased caspases 9 and 3 activities, nuclear condensation and DNA fragmentation (Yang & Tiffany-Castiglioni, 2008). Thus, it is suggested that PQ and PA synergistically enhance SH-SY5Y cytotoxicity.

Leptin, secreted primarily by adipocytes, is transported across the blood–brain barrier and acts on leptin receptors in CNS to regulate food intake by modulating activity of appetite control neurons in the brain (Zhang et al., 1994). Obesity is associated with leptin resistance, where high plasma leptin concentration was observed in most obese humans and rodents but this hyperleptinemia may not reduce appetite or increase energy expenditure (Frederich et al., 1995; Considine et al., 1996). It was reported that leptin receptors, the long and the short isoforms, are expressed in SH-SY5Y (Russo et al., 2004). However, in this study, pre-treatment of leptin was found to exert no neuroprotective effect against PA-induced cell death in SH-SY5Y cells. In a previous study, leptin was found to stimulate cell proliferation in a dose- and time-dependent manner involving the mechanism of apoptosis suppression in SH-SY5Y cells (Russo et al., 2004). SFAs (including PA) were shown to induce ER stress and inflammatory response via toll-like receptor 4 signalling, leading to leptin resistance in rat hypothalamus, whereas rats fed with monounsaturated FAs (including OA) did not develop hypothalamic leptin resistance (Milanski et al., 2009). Therefore, the result suggests a PA-induced leptin resistance in neurons, diminishing its neuroprotective effect.

The apoptotic cell death was confirmed with flow cytometric analysis using annexin V/PI staining and cell cycle analysis. The results are consistent with previous studies (Wang et al., 2012; Hsiao et al., 2014). In SH-SY5Y, PA induced neuron cell cycle G₂/M arrest at 24 h and increased in sub-G₀ phase at 48 h (Hsiao et al., 2014). However, the percentage of apoptotic cells in T98G as quantified by flow cytometry was much lower than that in SH-SY5Y, despite the similar LC₅₀ obtained from the MTT assay. This indicates that the use of MTT assay has a limitation, as its rate of tetrazolium reduction may reflect the general metabolic activity or the rate of glycolytic NADH production, and not the cell number (Berridge, Herst & Tan, 2005). The lesser PA-induced apoptotic cells observed in astrocytic T98G as compared to neuronal SH-SY5Y cells was not in the odd, owing to the high FA β-oxidation rate in astrocytes as compared to the poor utilisation of FA as fuel in neurons, as discussed earlier (Schönfeld & Reiser, 2013).

In both cell lines, PA and OA treatments were shown to induce ROS generation in a dose-dependent manner. However, the increase in ROS level was found to be higher in T98G than in SH-SY5Y. Also, the degree of lipid peroxidation was higher in T98G cells as compared to that in SH-SY5Y cells. However, there is a similar pattern of increment at 100 µM PA, decrement at 200 µM and rise again at 300 µM in both cell lines, except for the 48 h treatment in SH-SY5Y. At 48 h, the fold change in lipid peroxidation degree gradually decreased with increasing PA concentrations in SH-SY5Y. These observations may be attributed by the high FA β-oxidation rate in astrocytes, a prominent source of ROS generation (Seifert et al., 2010; Rodrigues & Gomes, 2012). β-oxidation of PA yields 15 molecules of flavin adenine dinucleotide (FADH₂) and 31 molecules of nicotinamide adenine dinucleotide (NADH) (FADH₂/NADH ratio ≈ 0.5) as compared to that of glucose degradation in which 2 molecules of FADH₂ and 10 molecules of NADH are generated (FADH₂/NADH ratio = 0.2) (Schönfeld & Reiser, 2013). Enhanced ROS generation is observed during the oxidation of FADH₂ for ATP generation by the electron transfer flavoprotein-ubiquinone oxidoreductase. Thus, the higher FADH₂/NADH ratio of PA β-oxidation in astrocytes would result in elevated ROS level. On the other hand, OA treatment in both cell lines was also shown to increase the ROS level, despite no cytotoxicity effect was detected by MTT assay. This is not unexpected, as OA was found to induce the production of ROS in rat aortic smooth muscle cells (Lu et al., 1998), human neutrophil (Carrillo et al., 2011) and cultured endothelial cells (bEnd.3) (Gremmels et al., 2015).

Conclusions

In summary, PA, but not OA and LA, induced cytotoxicity in SH-SY5Y, SH-SY5Y-γ and T98G cell lines in a time- and dose-dependent manner. The PA-induced cytotoxicity was found to be lower in SH-SY5Y-γ, suggesting its possible role in neuroprotection. Co-treatment of PA and PQ revealed that the PA-induced cytotoxicity was exacerbated by PQ. Leptin did not protect SH-SY5Y cell line from PA-induced neurotoxicity, suggesting a PA-induced leptin resistance. Annexin V/PI and sub-G₀ cell cycle analysis by flow cytometry revealed that PA-induced apoptotic cell death in both SH-SY5Y and T98G cell lines, but the percentage of apoptosis was much lower in T98G with similar concentrations of PA treatment. This indicates that neurons are more susceptible to PA-induced cytotoxicity. The PA-induced apoptotic cell death was found to be associated with increased lipid peroxidation and ROS generation. Taken together, the results suggest that HFD may cause neuronal and astrocytic damage, which indirectly proposes that CNS pathologies involving neuroinflammation and reactive gliosis could be prevented via the diet regimen. Apart from that, the signalling pathways involved in PA-induced apoptotic cell death and the neuroprotection of γ-syn warrant further investigation.