Introduction

Radiofrequency electromagnetic fields (RF-EMF) emitted by wireless communication systems are an omnipresent environmental factor for several decades. In Europe, the 700 MHz band, recently allocated for 4G and early 5G services, has gained particular importance due to its favorable propagation characteristics, ensuring wide-area coverage and indoor penetration1,2. The International Agency for Research on Cancer (IARC) classified RF-EMF as “possibly carcinogenic to humans” (Group 2B) based on limited evidence in humans with an observed association between exposure to RF-EMF from wireless phones and glioma and acoustic neurinoma in epidemiological data and limited evidence in animal data3. Hence, despite decades of investigation, whether wireless communication-relevant RF-EMF exposures influence central nervous system (CNS) physiology remains controversial, and international and national agencies1,4,5,6,7 emphasize the need for mechanistic studies under strictly controlled exposure conditions to clarify potential non-thermal deleterious effects.

One of the main hypotheses related to the possible non-thermal deleterious effects of RF-EMF is that they could alter cellular redox balance and induce oxidative stress, that could then lead to deleterious effects in cells and tissues, such as, in the CNS, cancer or neurodegenerative diseases8,9,10,11. Several early studies indeed reported enhanced oxidative stress or DNA strand breaks following exposure to mobile phone-related radiofrequency fields, sparking ongoing debate12,13,14,15,16. However, these findings have often lacked reproducibility, and recent evaluations highlight inconsistencies, methodological heterogeneity, and the importance of excluding thermal confounds6,17,18,19,20.

Astrocytes and neuronal cells represent particularly relevant CNS models as astrocytes play central roles in homeostasis and antioxidant defense in the brain, while SH-SY5Y neuroblastoma cells are widely used as neuronal surrogate. For example, neuroectodermal stem cells (NE-4C) exposed to 1950 MHz W-CDMA at 2 W/kg for 48 h showed no changes in proliferation, apoptosis, or cell cycle, although proteomics revealed some alterations at the protein level21. Similarly, human neuroblastoma SH-SY5Y and BE(2)C cells exposed to 900 MHz GSM-modulated RF-EMF at 1 W/kg displayed cell-type-specific DNA methylation changes without consistent directionality22. In line with these observations, rat neuroblastoma cells exposed to RF-EMF also showed no significant alterations in oxidative stress markers or cell viability under controlled exposure conditions23. In this context, the present study specifically addresses a knowledge gap by investigating whether 5G-modulated 700 MHz RF-EMF exposure-applied at SARs of 0.08 and 4 W/kg under rigorously thermal-controlled and blinded conditions- affects mitochondrial ROS production, viability, early and late apoptosis, and proliferation in primary rat cortical astrocytes and human SH-SY5Y neuroblastoma cells.

Material and methods

Cell lines

Two in vitro models relevant to central nervous system biology were used to investigate the cellular effects of 5G-modulated 700 MHz RF-EMF exposure: primary rat cortical astrocytes and human SH-SY5Y neuroblastoma cells.

Primary rat cortical astrocytes were obtained from Thermo Fisher Scientific (Cat# N7745100) at passage 1 and cultured according to the manufacturer’s protocol. SH-SY5Y cells were acquired from the ATCC (Cat# CRL-2266) and maintained under culture conditions similar to those used for astrocytes, ensuring direct comparability between models. Both cell types were grown in Dulbecco’s Modified Eagle Medium (DMEM; Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS), 1% penicillin-streptomycin, and 1% L-glutamine. Cultures were incubated at 37 °C in a humidified atmosphere containing 5% CO2 and passaged at approximately 80% confluence24,25. Cells were detached using 0.05% trypsin-EDTA (Sigma-Aldrich) and seeded into 35 mm Petri dishes (3 mL medium per dish). After 48 h, cells were either sham-exposed or subjected to RF-EMF exposure.

Exposure system and dosimetry

Exposure of biological samples to 5G-modulated RF-EMF at 700 MHz was performed using transverse electromagnetic (TEM) cells. TEM cells are specifically designed to generate a spatially uniform electromagnetic field within the exposure area, allowing precise control of exposure conditions26,27,28,29. In this study, three TEM cells were employed to simultaneously and blindly expose biological cells to three SAR levels: 0 W/kg (Sham), 0.08 W/kg, and 4 W/kg.

The signal generation unit (Fig. 1A) located externally to an incubator was consisted of the following components: a vector signal generator (SMBV100A, Rohde & Schwarz, Munich, Germany) delivering a 700 MHz 5G NR signal (release 15, Digital Standards SMBVB‐K444; Rohde & Schwarz) with FDD duplexing, QPSK modulation and 100 MHz channel bandwidth, a high power amplifier (ZHL-16W-43+, Mini-Circuits, NY, USA), a power circulator (PE8419, Pasternack, CA, USA), a bidirectional coupler (ZGBDC30-372HP+, MiniCircuits, NY, USA). To obtain the required exposure levels, two directional couplers with 10 dB and 6 dB coupling values (C1-C74-10N, C1-C74-06N, RDmicrowaves, NJ, USA) and a 1 dB attenuator (BW-N1W5+, Mini-Circuits, NY, USA) were inserted into the setup to attain 10 dB (1:10) and 7 dB (1:5) attenuation. Input power was split into the three TEM cells using three dual RF-EMF SP4T switch matrix (RC2SP4T-26, Mini-Circuits, NY, USA). This setup allowed to select two different SAR levels among three different values, i.e., 4, 0.4, 0.08 W/kg, with 0-, 10- and 17-dB attenuation, respectively. Through mechanical switching, different levels are directed to different TEM cells. In this experiment only 0 (sham), 0.08 and 4 W/kg SAR levels were applied. Incident and reflected power were continuously measured (1 sample/second) using wideband power sensors (N1921A Agilent, CA, USA) connected to a dual channel power meter unit (N1912A, Agilent, CA, USA), where sensors were connected to the bidirectional coupler and power values were recorded on a PC. The TEM cell output powers, attenuated by 30 dB, were also recorded via a data unit acquisition (34970 and 34901A, Agilent, CA, USA) using three detectors (ZX47-40+, Mini-Circuits, NY, USA). A graphic user interface was developed to monitor the experiment allowing blind exposure. The TEM cells were placed within a single temperature-controlled incubator, one per shelf (see Fig. 1A and B). Each TEM cell accommodated six 35-mm Petri dishes containing 3 mL of culture medium with a relative dielectric permittivity of 76 and a conductivity of 1.6 S/m. Dielectric parameters were computed from experimental measurements performed on a medium sample at 37 °C and 700 MHz using a dielectric probe (85070E Dielectric probe kit, Agilent, Santa Clara, CA, USA) connected to a vector network analyser (HP 8753E, Agilent, Santa Clara, CA, USA).

Fig. 1
Fig. 1
Full size image

Exposure system and dosimetry. (A) Schematic representation of the experimental setup. Cells were exposed in three TEM cells placed inside a temperature-controlled incubator, enabling blinded exposure at 0, 0.08, and 4 W/kg. A 700 MHz 5G-NR signal (FDD, QPSK, 100 MHz bandwidth) was generated externally, amplified, and distributed to the TEM cells through an RF switch matrix. (B) Photograph of the interior of the incubator showing the three TEM cells and the water reservoir. (C) Experimental dosimetry. Local SAR was obtained from RF-induced temperature increases measured with a fluoroptic thermometer. Incident and reflected powers were continuously monitored to ensure exposure stability and accuracy. (D) Numerical dosimetry. FDTD simulations confirmed homogeneous SAR distribution across the Petri dishes, with a bottom-layer SAR of 0.35 ± 0.08 W/kg (normalized to 1 W input) and spatial variability below 30%.”

Accurate exposure levels were ensured through both experimental and numerical dosimetry. Local SAR values were experimentally determined by measuring RF-EMF-induced temperature increases in the culture medium using a fluoroptic thermometer (Luxtron m920, LumaSense Technologies, Illinois, USA) (Fig. 1C). Specifically, the local SAR was quantified within an approximated volume of 1 mm3 corresponding to the region at the temperature‑sensing tip of the probe, in accordance with:

$$SAR = C\frac{\partial T}{{\partial t}}$$

where C is the specific heat capacity of the culture medium and \(\frac{\partial T}{{\partial t}}\) is the initial slope of the temperature increase versus time30. The specific heat capacity of the culture medium was assumed equivalent to the water value equal to 4186 J kg−1 K−1. This approximation is justified because the difference in specific heat capacity between standard cell culture media and water is typically within approximately 2%, as reported in31.

In parallel, finite-difference time-domain (FDTD) simulations were performed to get the numerical local SAR values similar to the experimental values, and to get the bottom layer SAR homogeneity as illustrated on Fig. 1D. Numerical dosimetry was performed using a custom FDTD solver that computes Maxwell’s equations in the time domain over a spatially discretized computational domain composed of elementary cells. The simulated geometry is presented in Fig. 1D. A nonuniform spatial discretization was implemented incorporating a fine‑mesh region with a resolution of 100 μm for the biological sample. The Petri dish material was modeled with a dielectric permittivity of 2.1. The 3 mL volume of culture medium was modelled in the numerical simulations using experimentally determined electrical properties at 700 MHz i.e., a relative dielectric permittivity of 76 and a conductivity of 1.6 S/m.

Considering the 6 Petri dishes, the bottom layer SAR value is 0.35 ± 0.08 W/kg (mean value ± standard deviation) normalized to 1 Watt incident power, which confirm the uniformity and accuracy of SAR distribution within the exposed samples. The spatial variability of SAR across the petri dish was below 30 %, in accordance with bioelectromagnetic exposure guidelines1. Following dosimetry, the input power was adjusted to achieve the desired SAR levels, and cells were exposed for either 1 hour or 24 hours. Temperature rise during exposure was measured with a maximum temperature increase of less than 2 °C at the highest SAR level (4 W/kg). This was compensated by adjusting the incubator settings to maintain the temperature between 36 °C (sham and 0.08 W/kg) and 37 °C (4 W/kg) ensuring a iso-thermal exposure of the cells.

Exposure was blind for the person performing the biological assays who worked on coded samples.

Flow cytometry measurements

Flow cytometry analysis

Flow cytometry analyses were performed using a Becton Dickinson (BD) FACSCanto™ II cytometer with BD FACSDiva software. Approximately 10,000 events were acquired per sample. Cells were kept on ice during staining and acquisition to minimize degradation.

For SH-SY5Y cells, phenotypic characterization was performed by flow cytometry using the neuronal marker CD56. Quantitative analyses and statistical evaluations were restricted to the CD56-positive (neuron-like) cell population.

Mitochondrial oxidative stress was assessed using MitoSOX™ Red (Molecular Probes, Invitrogen) to detect mitochondrial superoxide. Cell viability and apoptosis were evaluated using APC-conjugated Annexin V (BD Pharmingen™) in combination with Sytox Blue Dead Cell Stain (Molecular Probes, Invitrogen), allowing discrimination between viable cells, early apoptotic cells, and late apoptotic or dead cells. All staining procedures were performed according to the manufacturers’ instructions. Representative gating strategies are shown in Fig. 2.

Fig. 2
Fig. 2
Full size image

Flow cytometry gating strategy and validation of apoptosis, mitochondrial ROS, and proliferation assays in astrocytes. Representative flow cytometry dot plots obtained in primary rat cortical astrocytes illustrating assay validation and gating strategies. (A, B) Cell viability and apoptosis assessed using APC-conjugated Annexin V and Sytox Blue Dead Cell Stain. Panel (A) shows sham-exposed cells, while panel (B) corresponds to positive control cells treated with hydrogen peroxide (H2O2). Quadrant analysis allows discrimination between viable cells (Annexin V−/Sytox−), early apoptotic cells (Annexin V+/Sytox−), and late apoptotic or dead cells (Annexin V+/Sytox+). (C, D) Mitochondrial reactive oxygen species (ROS) production assessed using MitoSOX™ Red. Panel (C) shows sham-exposed astrocytes, while panel (D) shows astrocytes treated with H2O2, illustrating increased mitochondrial superoxide production in positive control conditions. (E, F) Cell proliferation assessed by CFSE dilution. Panel E corresponds to CFSE labeling at the time of exposure, while panel F represents CFSE profiles obtained at the 24 h post-exposure readout, allowing evaluation of cell division over the post-exposure period.

Positive controls were included in all experiments using hydrogen peroxide (H2O2) to induce oxidative stress. Astrocytes were treated with 500 µM H2O2 for 1 h, whereas SH-SY5Y cells were treated with 20 mM H2O2 for 5 minutes. These conditions were optimized to induce robust responses without excessive cell death. Optimization was based on stain index calculation for each marker and cell type (data not shown), defined as: (median fluorescence intensity of the positive population–median fluorescence intensity of the negative population)/standard deviation of the negative population.

Cell proliferation assay

Cells were seeded in 35 mm Petri dishes and exposed to RF-EMF at day two post-seeding (D + 2), once cultures reached approximately 80% confluence. Prior to exposure, two dishes per TEM cell were labeled with CellTrace™ CFSE (Invitrogen, C34554) at a final concentration of 5 µM for 20 minutes to assess cell proliferation32,33. Culture medium was then replaced with 3 mL of fresh complete medium.

Cells were exposed to 5G-modulated RF-EMF at 700 MHz for either 1 h or 24 h and analyzed immediately after exposure or following an additional 24-hour recovery period. Sham-exposed controls were included for all exposure durations and experimental endpoints (Fig. 3).

Fig. 3
Fig. 3
Full size image

Experimental workflow for 700 MHz RF-EMF exposure and cytometric analysis. Primary rat cortical astrocytes and human SH-SY5Y neuroblastoma cells were seeded in 35-mm Petri dishes (3 mL medium) and exposed on day 2 post-seeding (D+2) to 700 MHz 5G-modulated RF-EMF at 0, 0.08, or 4 W/kg for either 1 h or 24 h. Samples were analyzed at the immediate readout or at the 24 h post-exposure readout under standard culture conditions. Flow cytometry endpoints included cell viability (Annexin V−/Sytox−), early apoptosis (Annexin V+/Sytox−), late apoptosis/necrosis (Annexin V+/Sytox+), mitochondrial ROS (MitoSOX fluorescence), and cell proliferation (CFSE dilution). Positive controls were performed using H2O2 (500 µM for astrocytes; 20 mM for SH-SY5Y). All experiments were conducted under blinded conditions with n = 5 (astrocytes) and n = 6 (SH-SY5Y).

Detection of mitochondrial ROS, apoptosis, and viability

For mitochondrial ROS measurements, cells were stained with MitoSOX™ Red (2.5 µM, 30 minutes, dark), washed with phosphate-buffered saline (PBS), detached using trypsin-EDTA, and washed twice in Hanks’ Balanced Salt Solution (HBSS) by centrifugation (5 min, 1200 rpm, room temperature).

Apoptosis was assessed by incubating cells with APC Annexin V (5 µL, 20 minutes, dark, room temperature), followed by viability assessment using Sytox Blue Dead Cell Stain (10 µM for astrocytes and 6 µM for SH-SY5Y cells; 5 minutes). Viable cells were defined as Annexin V−/Sytox−, early apoptotic cells as Annexin V+/Sytox−, and late apoptotic or dead cells as Annexin V+/Sytox+.

All measurements were performed under blinded conditions using coded samples.

Experimental replication

All experimental conditions were assessed using five independent biological replicates for astrocytes (n = 5) and six independent biological replicates for SH-SY5Y cells (n = 6).

Statistical analysis

Decoding samples occurred once all biological analyses had been completed, just before performing the statistical analyses using GraphPad Prism software (version 9.5 for Windows; GraphPad Software, San Diego, CA, USA). Due to the relatively small sample size, non-parametric tests were preferred.

Mitochondrial ROS levels were expressed as arbitrary units (a.u.) representing fluorescent intensity.

Group comparisons in the validation controls were performed using the Wilcoxon rank-sum test (Mann–Whitney U test) while for the exposed samples, the Kruskall–Wallis test was used. A threshold of p < 0.05 was considered statistically significant.

Data are reported as as boxplots unless otherwise specified. Exact p values and test details are provided in the corresponding figure legends.

Results

Validation assays

Positive controls with H2O2 elicited robust responses in both astrocytes and SH-SY5Y cells (Fig. 4). Astrocytes were treated with 500 µM H2O2 for 1 hour, whereas SH-SY5Y cells were exposed to 20 mM H2O2 for 5 minutes prior to cytometric analysis. In astrocytes, oxidative challenge triggered a statistically significant increase in mitochondrial ROS, in the early and late apoptotic cell populations, and a concomitant significant decrease in the proportion of healthy cells (p = 0.0079 vs. controls for all markers). Similarly, SH-SY5Y cells exposed to H2O2 displayed a statistically significant decrease in cell viability, increase in ROS production and in induced apoptosis relative to untreated controls (p = 0.0061 vs. controls for all markers). These results confirmed the sensitivity and dynamic range of the flow cytometry assays used in subsequent RF-EMF experiments.

Fig. 4
Fig. 4
Full size image

Validation assays in astrocytes and SH-SY5Y cells. Hydrogen peroxide (H2O2) used as a positive control induced robust responses in both cell models, confirming assay sensitivity (**p < 0.05). Left panels (astrocytes): cells treated with 500 µM H2O2 for 1 h. Right panels (SH-SY5Y): cells treated with 20 mM H2O2 for 5 min. (A) Healthy cells (Annexin V−/Sytox−). (B) Mitochondrial ROS (MitoSOX fluorescence). (C) Early apoptosis (Annexin V+/Sytox−). (D) Late apoptosis (Annexin V+/Sytox+). Wilcoxon–Mann–Whitney test.

Effect of 5G RF-EMF on cellular viability

Across all RF-EMF exposure conditions, cell viability remained stable in both models compared to sham-exposed samples (Fig. 5.

Fig. 5
Fig. 5
Full size image

Cell viability after 700 MHz RF-EMF exposure. Cell viability, defined as the proportion of Annexin V−/Sytox− cells, was quantified in primary rat cortical astrocytes (left) and SH-SY5Y neuroblastoma cells (right) after exposure to 700 MHz RF-EMF at 0, 0.08, or 4 W/kg. Cells were exposed for 1 h or 24 h and analyzed at the immediate readout (A, C) or at the 24 h post-exposure readout (B, D). Viability remained uniformly high across all exposure conditions in both models. Data shown as boxplots; n = 5 (astrocytes) and n = 6 (SH-SY5Y). Kruskal–Wallis test.

In astrocytes, cell viability was first assessed immediately after RF-EMF exposure. Following a 1 h exposure, the proportion of healthy cells was 90.6 ± 5.4% in sham samples, 91.3 ± 6.3% in samples exposed at 0.08 W/kg, and 89.9 ± 6.7% in samples exposed at 4 W/kg. Immediately after a 24 h exposure, viability values were 91.7 ± 5.8% (sham), 88.6 ± 7.0% (0.08 W/kg), and 91.5 ± 7.3% (4 W/kg). Analysis showed no delayed effects in astrocytes after either 1 h or 24 h of post-exposure incubation compared with sham-exposed samples, as viability levels were comparable to those measured immediately after exposure and to the corresponding sham-exposed controls.

SH-SY5Y cells showed similar stability, with viability consistently above 85% across all conditions. At delayed timepoints, cell viability remained unchanged compared to sham-exposed controls. For instance, 24 h after a 1 h exposure, viability was 98.4 ± 1.1% in cells exposed at 4 W/kg versus 98.2 ± 0.9% in sham samples, while 24 h after a 24 h exposure, values were 96.7 ± 2.3% and 96.2 ± 3.1%, respectively.

Overall, no statistically significant differences were detected between RF-EMF-exposed and sham-exposed groups in either cell type (p > 0.05, Kruskal–Wallis test).

Effect of 5G RF-EMF on early apoptosis

The proportion of early apoptotic cells (Annexin V+/Sytox−) remained unchanged following exposure to the 5G-modulated 700 MHz signal in both cellular models.

In astrocytes (Fig. 6, early apoptosis was first assessed immediately after RF-EMF exposure. Following a 1 h exposure, early apoptotic cells represented 5.2 ± 1.8% in sham-exposed samples, 5.6 ± 2.0% at 0.08 W/kg, and 5.8 ± 1.9% at 4 W/kg. Immediately after a 24 h exposure, values were 5.5 ± 2.1% (sham), 6.1 ± 2.3% (0.08 W/kg), and 6.0 ± 2.0% (4 W/kg).

Fig. 6
Fig. 6
Full size image

Early apoptosis after 700 MHz RF-EMF exposure. Early apoptosis (Annexin V+/Sytox−) was measured in astrocytes (left) and SH-SY5Y cells (right) under the same exposure and readout conditions as in Fig. 5. Early apoptotic fractions were low and showed no detectable modulation by RF-EMF at any SAR level, exposure duration, or readout time. Data shown as boxplots; n = 5 (astrocytes) and n = 6 (SH-SY5Y). Kruskal–Wallis test.

In SH-SY5Y cells (Fig. 5B), early apoptosis remained below 2% across all exposure conditions. Immediately after a 1 h exposure, values were 0.7 ± 0.6% (sham), 0.8 ± 0.7% (0.08 W/kg), and 0.8 ± 0.6% (4 W/kg).

After recovery following RF-EMF exposure, delayed early apoptotic cell proportions remained unchanged in both models. In SH-SY5Y cells, analysis performed 24 h after a 1 h exposure yielded values of 0.6 ± 0.5% (sham), 0.6 ± 0.7% (0.08 W/kg), and 0.4 ± 0.4% (4 W/kg).

None of these differences reached statistical significance (p > 0.05, Kruskal-Wallis test).

Effect of 5G RF-EMF on late apoptosis

Late apoptotic cells (Annexin V+/Sytox+) consistently represented less than 3.5% of the astrocytes’ total population, and this percentage was unaffected by 700 MHz RF-EMF exposure (Fig. 7. Immediately after RF-EMF exposure, late apoptosis following a 1 h exposure was 3.1 ± 1.5% in sham-exposed samples, 3.4 ± 1.6% at 0.08 W/kg, and 3.5 ± 1.4% at 4 W/kg. Immediately after a 24 h exposure, values were 3.6 ± 1.6% (sham), 3.8 ± 1.7% (0.08 W/kg), and 3.7 ± 1.5% (4 W/kg).

Fig. 7
Fig. 7
Full size image

Late apoptosis/necrosis after 700 MHz RF-EMF exposure. Late apoptotic cells (Annexin V+/Sytox+) were quantified in astrocytes (left) and SH-SY5Y cells (right). The proportion of late apoptotic cells remained low (0.1–3.8% of the cell population) in both models and did not differ significantly across SAR levels, exposure durations, or readout times. Data shown as boxplots; n = 5 (astrocytes) and n = 6 (SH-SY5Y). Kruskal–Wallis test.

In SH-SY5Y cells (Fig. 7), late apoptotic populations remained below 1% across all conditions, both immediately and 24 h after exposure, with no detectable differences between exposed and sham groups.

Overall, no statistically significant differences were observed (p > 0.05, Kruskal–Wallis test).

Effect of 5G RF-EMF on mitochondrial ROS production

MitoSOX-based analysis revealed no significant changes in mitochondrial ROS production following exposure to 5G-modulated RF-EMF in either cell type (Fig 8).

Fig. 8
Fig. 8
Full size image

Mitochondrial ROS production after 700 MHz RF-EMF exposure. Mitochondrial superoxide levels were quantified using MitoSOX fluorescence in astrocytes (left) and SH-SY5Y cells (right) exposed to 700 MHz RF-EMF at 0, 0.08, or 4 W/kg for 1 h or 24 h, with analysis performed at the immediate readout (A, C) or at the 24 h post-exposure readout (B, D). ROS levels remained stable across all exposure conditions in both cell types, with no significant effects detected regardless of SAR level or exposure duration. Data shown as boxplots; n = 5 (astrocytes) and n = 6 (SH-SY5Y). Kruskal–Wallis test.

In astrocytes, mitochondrial ROS levels were first assessed immediately after RF-EMF exposure. After a 1 h exposure, mean fluorescence intensities were 2352 ± 472 a.u. in sham-exposed samples, 2402 ± 456 a.u. at 0.08 W/kg, and 2394 ± 405 a.u. at 4 W/kg. Immediately after a 24 h exposure, values remained comparable across all groups.

In SH-SY5Y cells, mitochondrial ROS levels were similarly stable. Immediately after a 1 h exposure, fluorescence intensities were 388 ± 51 a.u. in sham samples, 409 ± 74 a.u. at 0.08 W/kg, and 399 ± 60 a.u. at 4 W/kg.

24 h after RF-EMF exposure, mitochondrial ROS levels remained unchanged in both cell models, with no significant differences between exposed and sham groups (p > 0.05, Kruskal–Wallis test).

Effect of 5G RF-EMF on cell proliferation

CFSE-based proliferation assays showed no effect of exposure to the 5G-modulated RF-EMF signal on astrocytes or SH-SY5Y cells (Fig. 9). In both models, proliferation indices were comparable between sham and exposed groups at both SARs and all time points. No significant differences were detected (Kruskal–Wallis test).

Fig. 9
Fig. 9
Full size image

Cell proliferation after 700 MHz RF-EMF exposure. Cell proliferation was measured using CFSE dilution in astrocytes (left) and SH-SY5Y cells (right) following exposure to 700 MHz RF-EMF at 0, 0.08, or 4 W/kg. Cells were exposed for 1 h or 24 h and analyzed at the immediate readout (A, C) or at the 24 h post-exposure readout (B, D). CFSE profiles and proliferation indices showed no differences across SAR levels in either model. Data shown as boxplots; n = 5 (astrocytes) and n = 6 (SH-SY5Y). Kruskal–Wallis test.

Discussion

The objective of the present work was to evaluate whether exposure to 5G-modulated 700 MHz RF-EMF at two SAR levels, 0.08 and 4 W/kg, could induce measurable biological effects in two distinct cell models relevant to the nervous system, i.e. primary rat astrocytes and human SH-SY5Y neuroblastoma cells. These models were chosen to represent both glial and neuronal phenotypes. Astrocytes are key regulators of oxidative balance and inflammatory processes in the brain34,35,36. SH-SY5Y cells, while immortalized, represent a neuronal model widely used in neurobiology and toxicology studies37 and have been extensively employed in radiofrequency exposure research38. By including both cell types, we addressed the possibility of differential sensitivity between glial and neuronal lineages. Both astrocytes and SH-SY5Y cells exhibited comparable responses following RF-EMF exposure with no significant alterations detected, indicating that the responses observed is consistent across distinct CNS cell types under the conditions tested.

The endpoints considered: viability, apoptosis, mitochondrial ROS, and proliferation were selected because they reflect fundamental aspects of the cellular health and are frequently investigated in the RF-EMF literature. Our results show that across these endpoints and conditions, no significant effects could be observed when compared to sham exposures, either immediately or after 24 h of recovery.

These findings are highly relevant, as they directly address one of the underexplored frequency bands now implemented in mobile communication (700 MHz with 5G modulation) and provide further evidence to support the absence of adverse effects under non-thermal RF-EMF exposure. The importance of reporting such data cannot be overstated, since publication bias in favor of positive results has been a long-standing concern in this field1. By adding to the body of well-controlled studies reporting no effects, we contribute to a more balanced understanding of RF-EMF bioeffects.

Within this context, our experimental findings can be positioned relative to the existing body of in vitro RF-EMF studies.

Our observations are thus in agreement with numerous prior investigations that found no bioeffects of RF-EMF in brain cells under carefully controlled conditions. Sekijima et al.39 showed that 2.1 GHz RF-EMF signals, either continuous wave (CW) or W-CDMA modulated, had no impact on cell proliferation or gene expression in human cell lines including neuroblastoma and neuroglioma cells, using exposure duration and SAR values comparable to those used here. Xu et al.13 reported similar conclusions in mouse primary cultured neurons, where exposure to 1800 MHz RF-EMF failed to induce consistent oxidative damage. Both studies underscore the importance of rigorous dosimetry and replication in determining outcomes. Poulletier de Gannes et al.40 found no increase in ROS production in human brain cell lines, including SH-SY5Y, after a 24 h- exposure or 24 h after a 1-h exposure to Enhanced Data rate for GSM Evolution (EDGE)– modulated signal at 1800 MHz at SAR levels of 2 and 10 W/kg. More recently, Su et al41 also found no change in cell viability and proliferation after an intermittent (5 min on/10 min off) exposure to an 1800 MHz RF-EMF signal at 4.0 W/kg for up to 24 h in three glioblastoma cell lines, including SH-SY5Y cells. Joushomme et al.42 observed no alteration in global cell behavior (as determined by cellular impedance measurement) in SH-SY5Y neuroblastoma cells and primary rat astrocytes continuously and isothermally exposed to CW or GSM-, UMTS-, LTE-, and Wi-Fi-modulated signals at 1800 MHz for 3–5 days, and at SAR of 5–24 W/kg. In a study similar to the present work, Höytö et al.43 could not detect changes in human SH-SY5Y cells’ viability, proliferation, apoptosis, and oxidative stress as evaluated through lipid peroxidation and glutathione concentration when exposed to 872 MHz, CW or GSM-modulated signal (5 W/Kg for 1 h or 24 h).

The present results extend these findings to the 700 MHz band and 5G modulation, strengthening the argument that isothermal RF-EMF exposures do not systematically affect basic cellular functions in neuronal or glial cells.

It is worth noting that while many studies are in agreement with our results, others have described positive findings, often pointing toward oxidative stress or apoptosis. For instance, in primary rat astrocytes, a significant increase in ROS production and DNA fragmentation was reported after 20 min of exposure to CW or 50 Hz amplitude-modulated 900 MHz RF-EMF44. In this study however, no SAR level was calculated (10 V/m electric field) while a 0.03 °C temperature increase was measured. Wang et al.45 observed an increased ROS production and oxidative DNA base damage in Neuro-2a mouse neuroblastoma cells exposed to a 900 MHz GSM-talk signal (24 h, 2 W/Kg, isothermal exposure). Interestingly, silencing 8-oxoG DNA glycosylase-1 (OGG1), known as the major protective enzyme against the mutagenic effects of 8-oxoG resulting from oxidative stress, led to an increase in oxidative DNA base damage in Neuro-2a cells when exposed at 1 and 2 W/Kg, but not at 0.5 W/Kg. The authors found however no change in the cellular viability.

In a recent review dealing with the effects of RF-EMF on apoptosis in mammalian cells, Romeo et al.46 concluded that 84.7% of the 42 papers selected based on three quality criteria (presence of sham-exposed samples, dosimetry analysis, and temperature control) did not observe RF-EMF induced changes in apoptosis. Among these, Zielinski et al.47 showed that an intermittent exposure (2 min on/2 min off) of murine N9 microglial cells or SH-SY5Y human neuroblastoma cells to a GSM- modulated 935 MHz signal at 4 W/kg for 2 and 24 h did not impact the expression of apoptosis markers such as Apoptosis-Inducing Factor-AIF, Bcl-2, and Bax. Other studies using human neuroblastoma cells (SH-SY5Y, SK-N-SH or LAN-5 cell lines) showed no cell death through apoptosis after exposure to 872 or 900 MHz signals, CW or GSM-modulated, at SAR levels ranging from 0.2 to 5 W/Kg and exposure duration ranging from 2 to 72 h43,48,49,50. Those data are in agreement with the results obtained in this work. However, using SH-SY5Y cells, Buttiglione et al.51 observed an increase in apoptotic markers after 6 and 24 h of exposure to a GSM-900 RF-EMF and an SAR of 1 W/Kg.

These results illustrate that biological responses to RF-EMF exposure are highly context dependent and inconsistently reported across studies. Recent systematic and critical evaluations of the literature, including the review by Meyer et al.20, emphasize that much of the reported variability arises from substantial heterogeneity in experimental designs, exposure conditions, and endpoint selection. In line with this observation, Miyakoshi52 previously concluded that many apparent inconsistencies in RF-EMF bioeffects could be attributed to methodological factors, such as differences in exposure systems, dosimetry accuracy, or assay sensitivity, rather than to genuine biological variability.

Indeed, a number of earlier studies lacked robust dosimetric characterization or adequate control of temperature, thereby leaving room for thermal confounding. By contrast, the present study relied on an exposure system as described by Christ et al.26, ensuring stable electromagnetic fields and accurate SAR determination. The use of three TEM cells enabled parallel sham and exposed conditions, reducing experimental variability and strengthening internal validity. In addition, all biological analyses were performed under blinded conditions. Under such rigorously controlled experimental settings, the interpretation of null effects is more robust than in studies conducted under less stringent exposure conditions.

Considering the broader literature, the divergences observed also likely stem from experimental design differences. Lu et al.15, for instance, observed oxidative stress and apoptosis in immune cells, which may have intrinsic susceptibility to ROS-related pathways compared to neuronal or glial cells. Diem et al.12 reported DNA strand breaks in human fibroblasts (ES1 cells) and in human lymphoblastoid HL-60 cells after 1800 MHz exposure, although a true replication study in another laboratory, using the same cell lines, failed to confirm these findings53,54. Such variability suggests that sporadic findings of positive effects may also reflect experimental artifacts, specific susceptibilities of certain models, or publication bias.

Another essential point is the validity of the assays we used. Apoptosis detection was performed using Annexin V and Sytox staining, a gold standard for distinguishing early apoptotic from late apoptotic or necrotic cells. ROS measurements relied on MitoSOX, a mitochondrial superoxide probe widely used in the literature. Zielonka et al.55 have emphasized the limitations of hydroethidine-derived probes, particularly the potential for unspecific oxidation and the importance of distinguishing true superoxide-derived products from artifacts. By combining positive controls and applying these probes consistently across two cell models, we ensured that our assays were capable of detecting oxidative stress when present. The amplitude of responses to hydrogen peroxide validated the sensitivity of the system. Therefore, finding no major biological changes under RF-EMF exposure reflects the absence of detectable immediate and delayed effects rather than methodological insensitivity.

Publication bias is a recognized issue in RF-EMF research, as studies reporting no biological effects are less likely to be published, potentially skewing the literature toward positive findings. As highlighted by Foster and Moulder56, the field is characterized by inconsistent and poorly reproducible results, and the apparent predominance of positive outcomes may partly reflect selective reporting rather than true biological effects. Negative studies such as the present one therefore play an essential role by contributing to a more balanced body of evidence, which is critical for objective risk assessment and informed public health decisions.

Despite the strengths of our work, it is also important to recognize its limitations. First, we investigated exposures of 1 h and 24 h but not intermittent exposures or exposures extending over days or weeks which could have effects not captured here. Xu et al.13, for instance, highlighted the complexity of neuronal responses and the possibility that subtle molecular perturbations might not translate immediately into functional deficits. Second, our endpoints were restricted to apoptosis, viability, ROS production, and proliferation. While these are fundamental parameters of cell health, they do not encompass the full range of possible cellular responses. In particular, subtle phenotypic alterations such as astrocytic activation were not specifically assessed. Markers including glial fibrillary acidic protein (GFAP) or inflammatory mediators could provide additional insight into potential reactive states not captured by viability or apoptosis measurements. Similarly, microglial activation markers such as Iba1 may be relevant in more complex cellular systems. Other endpoints such as transcriptomic or proteomic changes, epigenetic modifications, or ion channel activity could be affected under specific conditions. Third, our study was limited to two cell models. Although astrocytes and SH-SY5Y cells are relevant, they do not represent the entire diversity of cell types in the nervous system, and SH-SY5Y cells, being an immortalized neuroblastoma-derived line, only partially recapitulate the molecular and functional properties of primary neurons. Moreover, the use of monoculture systems does not fully recapitulate the complexity of neuronal–glial interactions occurring in vivo. Crosstalk between neurons and astrocytes plays a critical role in redox homeostasis and inflammatory signaling, and co-culture models could reveal effects not detectable in isolated cell populations. Finally, some studies point to the possibility that RF-EMF could impact the effects of known chemical or physical stressors, although results are also controversial. We previously showed that 3.5 GHz 5G-modulated RF-EMF exposure at 1 W/kg did not alter mitochondrial membrane potential or apoptosis in human fibroblasts or keratinocytes, while a weak, cell-type-dependent increase in mitochondrial ROS appeared after a co-exposure to RF-EMF and UV57. Other co-exposure paradigms also suggested potential interactions such as combining 1800 MHz RF-EMF at 2 W/kg with pharmacological inhibition of mitochondrial calcium uptake, which enhanced DNA damage and apoptosis in mouse embryonic fibroblasts relative to the inhibitor alone21. Moreover, an RF-EMF induced adaptive response was evidenced through the reduction of chemically-induced micronuclei, DNA damage, and ROS production in cells co-exposed to RF-EMF (CW, GSM-, UMTS- or LTE-modulated fields at 900 or 1950 MHz and 0.3–1 W/kg) and pro-oxidant or genotoxic agents such as H2O2 or mitomycin C in different cell types including SH-SY5Y neuroblastoma cells58,59,60. Nevertheless, a label-free quantitative analysis in SH-SY5Y human neuroblastoma cells, HCT116 human colon tumoral cells, primary rat astrocytes, and primary neuronal cell cultures exposed to GSM-, UMTS-, LTE-, Wi-Fi-modulated, or CW signals at 1800 MHz and 1–2 W/kg revealed no alteration in global cell behavior, including under pro-apoptotic or pro-autophagic challenges42. Kang et al.61 also found that the combination of two RF-EMF signals (CDMA plus WCDMA signals, 4 W/kg, 2 h) did not interact with the ROS production induced either by H2O2 or menadione treatments in neuronal cells such as U87 human glioma cells and SH-SY5Y neuroblastoma cells.

Conclusion

In summary, our results show that exposure to a 5G-modulated 700 MHz RF-EMF at SAR levels up to 4 W/kg does not significantly alter cell viability, apoptosis, intracellular ROS, or proliferation in rat primary astrocytes and human neuroblastoma SH-SY5Y cells. These findings align with other studies reporting the absence of immediate or delayed cellular effects under controlled RF-EMF conditions.

We also recognize the need for future work to expand upon our findings for example by using omics-based and epigenetic approaches which could provide a broader view of potential molecular changes.

Overall, this work provides experimental evidence showing that, under our study conditions, 5G-modulated 700 MHz RF-EMF exposure does not induce acute or delayed adverse effects in two complementary brain cell models.