Repetitive low intensity magnetic field stimulation in a neuronal cell line: a metabolomics study

Cell culture

Rat neuroblastoma cells from the B50 cell line were seeded directly onto 6-well plates and grown for 24 h in media containing DMEM with 5% (v/v) heat-inactivated foetal calf serum, 2 mM L-glutamine, 100 U/ml streptomycin and 100 U/ml penicillin. Cells were grown at 37 °C within a CO₂ incubator (5% CO₂ + 95% air). Cells from each 6-well plate were later pooled during extraction to make one replicate. Each stimulus condition or control had six replicates in total.

LI-rMS stimulation

We used LI-rMS parameters that have previously been shown to increase intracellular calcium in primary cultured neurons in vitro (Grehl et al., 2015). Stimulation was delivered to cells in the incubator using custom built round coils (34 mm diameter, 17.1 mm height, 0.812 mm thickness, 138 turns). In order to deliver reproducible stimulation to each well, coils were designed to fit within a single well of a 6-well plate so that a plate containing cells could be placed on top of a plate containing five coils, resulting in reliable and reproducible placement at a distance of 2.8 mm from the base of each well (Figs. 1A and 1B). As the stimulator could only accommodate five coils, only five wells were stimulated on each plate. The coils were driven by a 12 V magnetic pulse generator under control of a programmable micro-controller card (CardLogix, Irvine, CA, USA), which delivered monophasic pulses (rise time of 0.725 ms). To approximate the stimulation dose in the 6-well plate, we measured the magnitude of the magnetic field in the x and y axes with a Hall Effect probe (SS94A2D; Honeywell, Morris Plains, NJ, USA). Magnetic field measurements were made at a distance of 0.5 cm from the base of the coil surface. Magnetic field strengths were converted to dB/dT, as a measure of the change in magnetic field over time of each delivered pulse. A map of dB/dT across the 6-well plate was generated with Matlab (Fig. 1C). Peak magnetic field and dB/dT inside each well ranged between ∼2.2 and ∼2.4 mT and ∼3 to ∼3.2 T/s respectively. Coil temperature did not rise above 37 °C, ruling out confounding effects of temperature change. Vibration of the coil was assumed to be within vibration amplitude of background (bench surface), as shown previously in Grehl et al. (2015). For the experiments, 6-well plates were assigned to one of three conditions: control (pulse generator switched off, no LI-rMS) (n = 6), 1 Hz (continuous stimulation for 10 min at 1 Hz) (n = 4) and 10 Hz (continuous stimulation for 10 min at 10 Hz) (n = 6). Stimulation and sham (control) stimulations were performed at 37 °C within the CO₂ incubator.

Metabolite extraction and derivatisation

Cells were immediately quenched and washed twice post-stimulation with 2 ml of ice-cold phosphate-buffered saline. Cells were then scraped, aggregated and frozen in liquid nitrogen before being freeze-dried. 500 µl of extraction solution consisting of 2.6 µg/ml of ¹³C₆-sorbitol, as an internal standard, was dissolved in methanol and added to the dried cells. Cells were lysed with a Precellys 24 tissue lyser (Bertin Technologies, Saint-Quentin en Yveline, France) at 6,500 rpm for 40 s. The supernatant was evaporated before adding 500 µl of water and freeze-drying again. Samples were derivatised according to the protocol described in Abbiss et al. (2012). 20 µl of methoxyamine HCl solution (20 mg/ml in pyridine) was added to each extract and agitated at 1,200 rpm and 30 °C for 90 min. 40 µl of N-methyl-N-(trimethylsilyl) trifluoroacetamide (MSTFA) was subsequently added to each extract and agitated at 300 rpm and 37 °C for 30 min. The derivatised samples were transferred to GC vials and 5 µl of alkanes(C₁₀, C₁₂, C₁₅, C₁₉, C₂₂, C₂₈; 0.156 mg mL⁻¹; C₃₂, C₃₆; 0.313 mg/ml) in hexane were added to each sample for calculation of a Kovat’s retention index, which aids comparison of data between samples.

Sample analysis

An Agilent 6890 series gas chromatograph with an Agilent 7863 autosampler coupled to an Agilent 5973N single quadrupole gas chromatography-mass spectrometer(GC-MS) (Agilent Technologies, Australia, Mulgrave, Australia) was used along with a Varian Factor Four-fused silica capillary column VF-5ms (30 m × 0.25 mm × 0.25 µm + 10 m EZ-Guard). The analytical method was also as described in Abbiss et al. (2012). 1 µl of the sample was injected splitless into the inlet that was set to 230 °C. The oven temperature was initially set to 70 °C with an initial temperature ramp of 1 °C/min for 5 min and was subsequently set to 5.63 °C/min, to a final temperature of 330 °C and held for 10 min. The ion source was set to 70 eV and 230 °C while the transfer line to the mass spectrometer was set to 330 °C. The detector, set to full scan, monitored a mass range of m/z 45–600 at 1 scan per second. The carrier gas, helium, was set at a flow rate of 1 ml/min.

Data analysis

Due to co-elution, the peaks of the ¹³C₆-sorbitol internal standard and glutamate could not be deconvoluted. Results were therefore normalized using the total ion chromatogram (TIC). The presence of aberrant peaks suggested that two of the 6-well plates (1 Hz stimulation condition) were not adequately washed during the extraction process and these were therefore excluded from the final analysis.

GC-MS data were viewed with AnalyzerPro v2.70 (SpectralWorks, Runcorn, UK). The mass spectra of peaks from the chromatogram were matched against the NIST (National Institute of Standards and Technology) mass spectral library. Metabolites with a similarity index of more than 60% were tentatively identified as metabolites. Metabolites with multiple derivative peaks were summed and treated as a single metabolite. For the peak area matrix, features that occurred in less than 80% of all samples were removed, unless they were unique to one treatment group in >75% of replicates. Peak areas were normalised to the total ion chromatogram (TIC) and imported into Unscrambler X version 10.1 (CAMO Software, Oslo, Norway). The data was log transformed [X = log(x + 1)] and a principal component analysis (PCA) was performed using a non-iterative partial least squares algorithm, cross validation and no rotation. Statistical comparisons between treatment samples were conducted with SPSS v21 (IBM, Corporation, Armonk, NY, USA) using a one-way ANOVA with Tukey’s post-hoc test.

LI-rTMS induces specific metabolic changes associated with GABA synthesis

Following LI-rTMS, we did not detect any significant change in the levels of glucose, fructose or galactose, which are key energy supplies to the cell. This is in contrast with outcomes of high intensity rTMS in human and in cat studies, where changes in glucose levels were detected during and immediately after stimulation (Valero-Cabré, Payne & Pascual-Leone, 2007). In the latter studies, glucose uptake was significantly reduced during stimulation, suggesting a local suppression of neuronal firing (Valero-Cabré, Payne & Pascual-Leone, 2007). However, immediately after high frequency, high intensity rTMS, the same regions had increased metabolism (Valero-Cabré et al., 2005), presumably due to long lasting plastic changes in intracortical circuitry (Walsh & Pascual-Leone, 2003). Together, these findings suggest that the intensity and frequency of stimulation contribute to changes in metabolic response.

However, in contrast to the stability of major carbohydrates, the metabolic profiles of metabolites implicated in GABA synthesis were reduced following LI-rMS. GABA is synthesised via glutamate, from α-ketoglutarate, a principal component of the TCA cycle (Fig. 3, box a). We observed depletion of three amino acids involved in de novo synthesis of TCA cycle intermediates, aspartate, phenylalanine and isoleucine (see Fig. 3, boxes b, c and d, respectively). A possible explanation is that LI-rMS caused increased spontaneous neurotransmitter release in B50 cells, requiring the use of α-ketoglutarate to replenish GABA pools and depleting the downstream TCA cycle intermediates and their substrates (Table 1 and yellow boxes, Fig. 3). Similarly, increased GABA synthesis could explain the LI-rTMS dependent reduction in pyroglutamate and alanine (Kumar & Bachhawat, 2012; Westergaard et al., 1993), both of which can be converted to glutamate and subsequently to GABA. In support of this, spontaneous transmitter release is increased following LI-rMS in excitatory neurons (Ahmed & Wieraszko, 2008), and our results suggest a similar increase in spontaneous GABA release from B50 cells.

Reduction in other amino acids and metabolites

The amino acids serine and glycine were also significantly decreased following 1 Hz and 10 Hz LI-rMS. The technique used in our study could not differentiate between L- and D-serine. Although both stereoisomers are found in the brain, only L-serine is incorporated into proteins, while D-serine acts as a neuromodulator by co-activating NMDA receptors (Wolosker, 2006). The reduction in serine could be due to increased protein synthesis, which has been shown for specific proteins such as BDNF, c-fos and various neurotransmitter receptors following rTMS (Chervyakov et al., 2015; Rodger et al., 2012). Glycine is mostly synthesized de novo in the brain from serine, rather than taken up via the blood–brain barrier (Shank & Aprison, 1970), and accumulation of glycine in neurons requires the activity of the glycine transporter GlyT2 (Gomeza et al., 2003). Thus, disruption of L-serine transport and/or diminished conversion of L-serine into glycine may contribute to the observed reduction in glycine. A key function for serine and glycine is to act as co-agonists for the NMDA receptor, which plays a central role in the long term plastic changes induced by high intensity rTMS(Vlachos et al., 2012), and may also contribute to LI-rTMS effects (Makowiecki et al., 2014; Rodger et al., 2012).

LI-rMS downregulates inositol and cholesterol—implications for calcium signalling and exocytosis

Inositol plays a key role as a precursor for inositol lipid and inositol phosphate syntheses, which are vital for signal transduction and intracellular calcium homeostasis (Wen, Osborne & Meunier, 2011). A previous study showed that LI-rMS increases the levels of intracellular calcium in cortical neurons by release from intracellular stores, possibly via inositol-1,4,5-triphosphate (IP3) signalling (Grehl et al., 2015). Inositol depletion as detected here may thus reflect the changes in intracellular signalling events induced by LI-rMS. Cholesterol was also decreased in our study. This lipid is not only important as a structural constituent of lipid rafts and cell membranes, but also modulates vesicle trafficking and exocytosis (Churchward et al., 2005; Lang et al., 2001; Zhang et al., 2009). A recent study in ageing mice found that rTMS reduced cholesterol that accumulated with age (Wang et al., 2013). This is important because cholesterol is associated with oxidative stress during normal aging (Cutler et al., 2004), and decreasing cholesterol levels can improve cognition in rats (Wang et al., 2013). A decrease of cholesterol following rTMS and LI-rTMS, as suggested by our data, may thus contribute to the cognitive improvements observed following rTMS treatment of Alzheimer’s patients (Hsu et al., 2015), and also in healthy volunteers (Miniussi & Ruzzoli, 2013).

Frequency-dependent metabolic changes

High and low frequency rTMS have different effects on neuronal circuits, with functional and genetic studies identifying distinct mechanisms in different neuronal cell types (Grehl et al., 2015; Pell, Roth & Zangen, 2011). Previous work using high intensity rTMS in anaesthetised cats demonstrated that metabolic changes were also frequency specific with increased glucose metabolism following high, but not low frequency stimulation (Valero-Cabré, Payne & Pascual-Leone, 2007). Our study provides a different approach, because we examined metabolic changes in a uniform population of neuronal cells with an inhibitory phenotype. Thus our finding that the effects of LI-rMS are stronger with 1 Hz than with 10 Hz matches the inhibitory effect of 1 Hz that is consistently reported in animal and human literature (Chen et al., 1997; Maeda et al., 2000; Pell, Roth & Zangen, 2011; Trippe et al., 2009). However, because LI-rTMS does not induce action potentials, it remains unclear whether the metabolic changes we report are a direct consequence of the electromagnetic field, or whether they are secondary to the changes in excitability induced by rTMS.

B50 neuroblastoma cells are derived from the rat central nervous system (Schubert et al., 1974) and have been used extensively to study the pathways involved in cell death, proliferation and migration (e.g., Honma et al., 1996). Although cells grown in culture remain a highly simplified model of the CNS (discussed in Grehl et al., 2015), B50 cells have been reported to fire regenerative action potentials, and contain high levels of the GABA bioythetic enzyme, glutamate decarboxylase (GAD), as well as GABA itself, suggesting that they retain some key functional features of inhibitory neurons (Schubert et al., 1974). Future studies in large, uniform populations of neuronal-like cells will contribute to our understanding not only of the mechanism whereby cells detect electromagnetic fields, but also of the signalling events such as kinase and phosphatase cascades, that underpin the complex cellular responses such as apoptosis, differentiation, migration and proliferation.