The potential effects of radiofrequency (RF) electromagnetic fields used for wireless telephony on public health or welfare are still an open question. Solving this issue requires determining biological effects of these RF, including pulse-modulated signals such as global system for mobile communications (GSM) operating in the 900 and 1800 MHz bands. Preclinical animal models and cell culture studies have been widely used to search for effects of RF at behavioral, cellular, or molecular levels. In particular, the effects of GSM-associated RF on the healthy CNS have been investigated with regard to memory functions (Dubreuil et al. 2003; Klose et al. 2014; Kumlin et al. 2007; Nittby et al. 2008; Schneider and Stangassinger 2014), blood-brain barrier integrity (Stam 2010), oxidative stress (Poulletier de Gannes et al. 2011; Xu et al. 2010), neural cell survival (Marino et al. 2011), neuronal firing rate (Beason and Semm 2002; Moretti et al. 2013), neurotransmission (Mausset et al. 2001; Mausset-Bonnefont et al. 2004), and astroglial reactivity (Ammari et al. 2010; Bouji et al. 2012; Court-Kowalski et al. 2015; Mausset-Bonnefont et al. 2004; Petitdant et al. 2016; Watilliaux et al. 2011). However, the occurrence and the amplitude of RF effects are still a matter of debate and appear to strongly vary according to experimental settings. In vivo studies have used whole-body or head-only exposures to RF, the latter being closer to conditions in which human are exposed to RF emitted by a cellular phone. In both cases, key parameters include the number and duration of exposures to RF, as well as the specific absorption rate (SAR), which measures the rate at which RF energy is absorbed by a unit mass of tissue (W/kg).

Only few studies have addressed the effect of RF on the brain affected by pathological processes. It was reported that in transgenic mice model of Alzheimer’s disease, chronic exposures to 900 or 1950 MHz RF improve cognition and brain mitochondrial function while reducing ß amyloid deposits and reactive gliosis (Arendash et al. 2010; Dragicevic et al. 2011; Jeong et al. 2015). The effect of RF was also documented in the context of epilepsy. In rats injected with subconvulsive doses of picrotoxin, a single whole-body exposure to GSM-900 MHz can potentiate neuronal and astroglial responses to the proconvulsive agent and promote seizures (Carballo-Quintas et al. 2011; Lopez-Martin et al. 2006).

It is now well established that many neuropathological conditions involve an inflammatory reaction in which CNS resident phagocytes, such as microglia, play key roles owing to their capacities to produce a variety of components, among which are cytokines and reactive oxygen species (Heneka et al. 2014; Shemer et al. 2015). Previous studies of microglia exposed to modulated RF ranging from 900 to 1950 MHz led to conflicting results, showing either a lack of proinflammatory cell response (Hirose et al. 2010; Thorlin et al. 2006) or a stimulation in cell expression of proinflammatory genes (Lu et al. 2014). However, microglia are highly plastic cells. In pathological states, they undergo an activation process marked by morphological, transcriptomic, and proteomic changes that can dramatically modify their capacities to interact with neighboring neural cells and to sense external chemical or physical cues (Srinivasan et al. 2016; Wes et al. 2016). It is therefore conceivable that activation of microglia may affect their sensitivity or responses to RF.

Neuroinflammation triggered by an intraperitoneal injection of lipopolysaccharide (LPS) has been widely used to investigate microglial responses in a variety of pathological contexts (Cunningham et al. 2009; Hagberg et al. 2012). This peripheral immunological challenge causes an activation of microglia associated with increased brain expression of proinflammatory markers, including interleukin (IL)1ß, tumor necrosis factor-alpha (TNF-α), IL6, and enzymes catalyzing the formation of reactive oxygen species, such as inducible NO synthase (NOS2) and NOX2-dependent phagocyte NADPH oxidase (NOX2) (Hoogland et al. 2015; Li et al. 2014; Qin et al. 2013). The LPS-triggered neuroinflammation is associated with sickness and depressive-like behaviors. These behaviors implicate changes in the excitatory neurotransmission and are thought to involve proinflammatory cytokines and glutamatergic agonists released by activated microglia (Dantzer et al. 2008; Dantzer and Walker 2014).

In this study, we assessed the effect of GSM-1800 MHz on the expression of proinflammatory genes and on the morphology of microglia in the cerebral cortex of rats injected with LPS. We analyzed gene responses to RF both in adult and in developing rats as neuroinflammatory profiles and the brain sensitivity to RF may vary according to developmental stages (Christensen et al. 2014; Feychting 2011; Kaplan et al. 2016; Odaci et al. 2008). We also looked for the effect of RF on α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptors (AMPAR) that mediate fast excitatory neurotransmissions. These receptors are homomeric or heteromeric channels assembled from differing combinations of GluA1–4 subunits (formerly termed GluR1–4), with a predominance of GluA1/GluA2 (GluR1/GluR2) heteromers in forebrain areas (Lu et al. 2009; Shepherd and Huganir 2007). We focused on the phosphorylation of serine (ser) residues in the intracytosplamic terminal domains of GluA1 and GluA2. These post-translational modifications impact AMPAR activity and subcellular trafficking (Lu and Roche 2012; Shepherd and Huganir 2007) and are modulated by acute stress (Caudal et al. 2010, 2016). Head-only exposures to GSM-1800 MHz were performed using an experimental setting that did not trigger cell stress responses or glial reactions in healthy rats aged from 5 to 35 days (Watilliaux et al. 2011).

Here, we show that under neuroinflammatory conditions, a single exposure to GSM-1800 MHz results in transient changes in the expression of proinflammatory genes, microglial cell morphology, and in the level of phosphorylation of the GluA1 AMPAR subunits in the cerebral cortex.



All experimental procedures were carried out in compliance with the INSERM and CNRS ethical rules and the European Community Council Directive for the Care and Use of laboratory animals (2010/63/EU). All protocols were approved by the institutional ethics committee (CEEA 59, approval number 03729.02). Adult Wistar females with their adopted male pups and adult Wistar males were obtained from Janvier Labs (Le Genest Saint Isle, France). Male pups originating from a same litter were assigned to the same adoptive mother. Each adoptive mother nursed male pups stemming from at least two different litters. The animals were housed in a humidity (50–55%)- and temperature (22–24 °C)-controlled facility on a 12 h:12 h light/dark cycle (light on at 7:30 a.m.) with free access to food and water. The experiments were carried out with males. Adults and developing rats were accustomed to the colony rooms for at least 1 week before any treatment (LPS injection and/or exposure to RF). Two groups of developing rats and two groups of adults (n = 6 animals per group) were used to determine the effect of LPS per se on the expression of proinflammatory genes. Six groups of developing rats (n = 5 to 7 animals per group) and ten groups of adults (n = 6) were used to analyze the effects of RF exposures combined or not with LPS treatment. Half of these groups were submitted to sham exposures. Each group of developing rats comprised pups nursed by two to three different adoptive mothers. For each adopted litter (nursed by a same adoptive mother), some animals were assigned to RF exposure and others to sham exposure.

LPS Injection

Adult (2-month-old) and 14-day-old male rats were injected intraperitoneally (i.p.) with Ecoli LPS (250 μg/kg, serotype 0127:B8, SIGMA) diluted in sterile endotoxin-free isotonic saline. Two groups of developing or adult rats were sacrificed 24 h after injection of LPS or vehicle (isotonic saline) for assessment of the cortical neuroinflammation.

Exposure to GSM-1800 MHz

GSM exposure in rats was carried out 24 h after LPS injection or without LPS pretreatment, according to a previously described procedure (Watilliaux et al. 2011). The animals were lightly anesthetized with a mixture of ketamine/xylasine before exposure in order to prevent movement and to ensure reproducible positioning of the animals’ heads close to the loop antenna emitting the GSM signal. A small tissue support was placed under the head of the anesthetized P15 rats to keep the head position as close as possible to the loop antenna. Half of the adult rats and half of the developing rats from each litter were placed in the device without exposure to RF and served as control (sham exposure). The anesthetized animals were placed on a metal-free heating pad, to maintain body temperature around 37 °C throughout the experiment. The exposure duration was set at 2 h. After exposure, the animals were placed on another heating pad until they woke up (usually in less than 45 min) and then returned to their home cage. Rat pups exposed or sham-exposed at postnatal day 15 (P15) (e.g., 1 day after LPS injection) were separated from their adoptive mother for less than 3 h. The animals were sacrificed 24 or 72 h after exposure for biochemical or immunohistological analyses of the cerebral cortex. Figure 1 shows the different steps of the protocol.

Fig. 1
figure 1

Diagram of the experimental protocol designed to test the effects of GSM-1800 MHz on adult or developing rats injected (a) or not (b) with LPS

RF Exposure System

The exposure system has been described in detail elsewhere (Leveque et al. 2004). Briefly, a radiofrequency generator emitting a GSM electromagnetic field (RFPA, 1/8 duty factor) pulsed at 217 Hz (pulse emission every 4.6 ms for 546 μs) was connected to a four-output divider, allowing simultaneous exposure of the head of four anesthetized animals. Each output was connected to a loop antenna (Sama-Sistemi srl; Roma), enabling local exposure of the head of an anesthetized animal lying on a non-electric heating pad (3M coldhot pack, made of water/urea 1/1 v/v), as illustrated in Fig. 2a. The loop antenna consisted of a printed circuit, with two metallic lines engraved in a dielectric epoxy resin substrate (dielectric constant ε r = 4.6). At one end, this device consisted of a 1-mm-wide line forming a loop that was placed close to the animals’ head. Specific absorption rates (SARs) were determined numerically with the finite-difference time-domain (FDTD) method, using numerical models of adult and P15 rats. Anatomical magnetic resonance imaging of adult (250–300 g) and P15 rats maintained in the exposure set up were carried out and the tissues (skin, skull, brain, marrow, fat, muscles) were segmented and assigned to different dielectric properties, as described in Leveque et al. (2004). The dielectric properties of rat tissue change as a function of age. In our study, the dielectric properties of the P15 rat model were extrapolated from Peyman et al. (2001), where the dielectric properties changes were measured in the frequency range of 130 MHz to 10 GHz. At 1800 MHz, the decrease in dielectric properties of several tissues in the P15 to adult rats was evaluated to around 10%. For the dosimetric characterization, a detailed uncertainty analysis was performed. The procedure to provide the uncertainty analysis followed the guidance for the uncertainty assessment of in vivo exposure systems (Kuster et al. 2006). The combined uncertainty of the numerical dosimetry including the discretization, the mechanical and electrical variations, the loop antenna support, and heating pad properties gave a 0.7 dB standard deviation, a value close to the one determined in a recent detailed analysis of a rat exposure system (Collin et al. 2016). SARs were also determined experimentally in homogeneous rat phantoms using a Vitek or Luxtron probe for measurement of temperature rises. In this case, SARs, expressed in watts per kilogram, were calculated using the following equation: SAR = CΔTt with C being the calorific capacity in J/(kg K); ΔT, the temperature change in °K; and Δt, the time in seconds. Numerically determined SAR values were compared with experimental SAR values obtained using homogenous phantoms, especially in the equivalent rat brain area. The agreement between numerical and experimental SAR values was satisfactory.

Fig. 2
figure 2

Dosimetric analysis of specific absorption rates (SARs) in the rat brain during exposure to GSM-1800 MHz. The heterogeneous model of phantom rat and loop antenna described by Leveque et al. (2004) was used to evaluate the local SAR in the brain with a 1-mm3 (adult brain) or 0.5-mm3 (P15 brain) cubic mesh. a Global view of the rat phantom (adult rat) in the exposure setup with the loop antenna above its head and the heat pad (yellow) under the body, and a heat map of head and body SAR values. b Enlarged view of the area boxed in (a) and showing the interface between the loop antenna and the rat’s head. Note that highest SAR values correspond to the first millimeters of tissues below the loop antenna (i.e., the skin, the skull, the dura mater, and the cortex). c Distribution of SAR values in the adult brain at 1-mm3 spatial resolution. d SAR values in the P15 brain at 0.5-mm3 spatial resolution, the areas delimited by black contours in sagittal sections (c and d) correspond to the motor and somatosensory cortices that were used for biochemical and histological analyses. The color-coded scale of SAR values shown in (d) applies to all numerical simulations shown in the figure

Figure 2 shows the SAR distribution in rat models corresponding to 2-month-old adults (Fig. 2a–c) or developing rats at postnatal day 15 (P15) (Fig. 2d). These models match the actual rats used in our study in terms of weight and size.

Figure 2a, b provides global views of the rat brain model (with the loop antenna above) and show that the rat’s head absorbed most of the total power absorbed by the animal. SAR values for the whole body, the whole brain, and the cerebral cortex are reported in Table 1. The mean values of whole body SAR for the adults (0.13 W/Kg) or the P15 rats (0.63 W/kg) were above the public exposure limit of 0.08 W/kg or the occupational exposure limit of 0.4 W/kg for human whole-body exposure, which are set by the international commission on non-ionizing radiation protection (ICNIRP) and European guidelines (2013/35/EU). In adult rats, high values of SAR were limited to the animal’s head (Fig. 2a, b) confirming previous dosimetric studies with 900 MHz (Leveque et al. 2004) and with 1800 MHz RF (Watilliaux et al. 2011). Figure 2c shows the distribution of the SAR values within the adult brain calculated with a 1-mm3 spatial resolution. The mean SAR value was 2.94 W/kg in the dorso-medial part of the cerebral cortex that was selected for biological analyses (Fig. 2c and Table 1). This SAR value was above the public exposure limit of 2 W/kg set for human head exposure.

Table 1 Results of the dosimetric study performed by FDTD with phantom rats matching the size of adult and P15 rats. The values of specific absorption rate (SAR) were computed as specified in the Methods section

In the P15 rat model (Fig. 2d, Table 1), the consequences of the small size of the animals were (i) a higher Brain Average SAR (BASAR) and (ii) a higher whole body SAR, but the SAR values in the dorso-medial cortical region were similar to those determined in the adults (mean SAR, 2.90 W/kg).

Tissue Preparation for Biochemical Analyses

Rats were killed by decapitation under isoflurane anesthesia, and the brains were quickly removed from the skulls. The dorso-medial part of the cortical mantle corresponding to motor and somatosensory areas was dissected out on ice. Tissues were stored frozen at −80 °C until use. RNA and proteins were extracted from cortical tissues collected from the right and left hemispheres, respectively.

RNA Extraction and Quantitative RT-PCR

RNA extraction from cortical tissues was performed using RNeasy Lipid Tissue mini kit (Qiagen, Les Ulis, France). Complementary DNA (cDNA) was synthesized from 400 ng of total RNA using Maxima First strand cDNA synthesis kit (Thermo Scientific, Waltham, USA). Quantitative PCRs were performed with LightCycler®480 SYBR green I Master mix (Roche Diagnostics, France, Maylan). cDNA was amplified with RT2-qPCR primer sets from Qiagen (Les Ulis, France), for rat TNF-α (ref, PPR42247F, accession number NM_012675.3) and HPRT1 (ref: PPR42247F; NM_0125583.2), and with primers for IL1ß, IL6, CCL2, NOS2, NOX2, Iba1 (AIF1), and CD11b (ITGAM) genes, which were designed to span introns and which are listed in Table 2. Specificity of designed primers was checked by Ensembl BLAST analyses. The samples were amplified by 45 cycles: (15 s at 95 °C, 1 min at 60 °C). Each reaction was performed in triplicate. Single peak dissociation curves with expected Tm were verified for each sample. Gene expression was normalized to HPRT1 and analyzed by calculating 2−ΔCt values.

Table 2 Primers used for real-time qPCR analysis

Western Blotting

Tissue samples were homogenized by sonication in cold 1% SDS mixed with Halt™ protease and phosphatase inhibitor cocktail (Thermo Scientific, Waltham, USA, ref: 784442). The total protein concentration was determined with a BCA assay using a Multiskan Ex spectrophotometer (Thermo Electron Corporation). Proteins (50 μg per sample) were separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and electrophoretically transferred to nitrocellulose membranes. The membranes were then incubated for 30 min at room temperature in blocking buffer (0.1 M Tris, 150 mM NaCl 0.1% Tween, 5% skimmed milk) and probed over night at 4 °C with mouse or rabbit specific primary antibodies. Primary antibodies were mouse anti-GluA1 diluted, 1:500; rabbit anti-Phospho-Ser831GluA1, 1:500; rabbit anti-Phospho-Ser845GluA1, 1:500; rabbit anti-GluA2, 1:500; rabbit anti-Phospho-Ser880GluA2, 1:500 (all from PhosphoSolution, Aurora, CO); and rabbit anti-ß tubulin, 1:500,000 (Abcam, Cambridge, UK). Then, membranes were washed and incubated for 2 h at room temperature with goat anti-rabbit IRDye 680 or goat anti-mouse IRDye 800W secondary antibodies diluted 1:10,000 (LI-COR Biotechnology, GmbH, Homburg, Germany). Bands were visualized and their intensities were determined and normalized to that of ß tubulin in each sample using an Odyssey® CLx imager (LI-COR).


Adult rats were deeply anesthetized with pentobarbital and perfused with 0.1 M PBS followed by PFA 4% in 0.1 M PBS, pH 7.4. Brains were post-fixated overnight in PFA 4%, then cryoprotected by immersion in PBS containing 25% (w/v) sucrose and frozen in melting isopentane. Immunostaining was performed on coronal sections (20-μm thick) cut on a cryostat (Microm, Heidelberg, Germany) and mounted on gelatin-coated Superfrost glass slides (Menzel-glazer, Freiburg, Germany). Sections were blocked in PBS containing 0.1% Triton-X100 and 10% goat serum (30 min at room temperature) and then incubated overnight at 4 °C with mouse monoclonal anti-CD11b (1:100 OX42, AbD Serotec, Kidlington, UK) or rabbit polyclonal anti-Iba1 (1:400, Wako Chemicals). Bound antibodies were detected by applying biotinylated goat anti-mouse or anti-rabbit IgG (GE Healthcare, Velizy Villacoublay, France) for 2 h at room temperature, followed by Alexa Fluor 488-conjugated streptavidin (Life Technology, Saint Aubin, France) for 30 min at room temperature. Sections were counterstained with Hoechst 3342 dye (2 μg/ml) and mounted in fluoromount G (CliniSciences). Control staining performed by omitting primary antibodies was negative.

Image Analysis

Quantitative image analyses were performed as described previously (Cheret et al. 2008). Images were captured using a Zeiss AxioImager Z1 microscope equipped with the apotome system. A 10× NA 0.50 Fluar objective (Zeiss microscope) was used to acquire images with equal exposure times. For each animal, immunostained microglial cells were analyzed in microscopic fields (6 × 104 μm2 area) acquired in 12 sections of the dorso-medial cortex and distributed over a cortical region starting 2 mm rostral to the bregma and extending over 2400 μm in the rostro-caudal axis. Image analysis was performed blind relative to the experimental groups. The microglial cell bodies, confirmed by nuclei counterstaining, with Hoechst dye were counted manually. The area stained by Iba1 or CD11b antibodies and the area of Iba1-stained cell bodies were determined following assignment of a threshold to eliminate background immunofluorescence, using Image J software.

Statistical Analysis

Differences between groups of animals (n = 6 adults or 5 to 7 developing rats) were analyzed using non-parametric Mann–Whitney U tests with Benjamini–Hochberg multiple testing corrections. The false detection rate set at 0.05 was applied to determine significant differences.


Effect of LPS on the Expression of Proinflammatory Genes in Adult and in Developing Rats

Acute neuroinflammation was achieved by injecting adult (2-month-old) or developing (2-week-old) rats with LPS, which triggers an early fever response. In 2-week-old rats, the fever response can be biphasic with an initial period of hypothermia followed by an increase in body temperature within less than 6 h following LPS injection (Heida et al. 2004). The delay between LPS injection and exposure to RF was always set at 24 h to avoid exposure at the fever peak. Our investigation focused on dorso-medial regions of the cerebral cortex localized close to the loop antenna delivering the GSM signal (Fig. 2). We first assessed the level of expression of genes associated with proinflammatory profiles at the time of RF exposure (i.e., 24 h post LPS injection) using RT-qPCR analyses. In adult rats (Fig. 3a), LPS treatment led to significant increases in the levels of transcripts encoding proinflammatory cytokines or enzymes catalyzing the formation of reactive oxygen or nitrogen species, such as TNF-α, IL1ß, CCL2, NOX2, or NOS2. The most prominent differences appeared in NOX2 and IL1ß transcripts, the mean levels of which were increased 4- and 12-fold, respectively. At this time point, the level of IL6 transcripts was not significantly modified by the LPS treatment (Fig. 3a).

Fig. 3
figure 3

Neuroinflammation 24 h after peripheral immune challenge in adult (a) and in developing P15 (b) rats. Adult and 14-day-old rats were injected i.p. with LPS or vehicle (control) and cortical tissues were collected 24 h later. Levels of TNF-α, IL6, CCL2, NOS2, NOX2, and IL1ß mRNA levels were determined by RT-qPCR. Data expressed in arbitrary units are means ± SEM from six animals (n = 6). *p < 0.05 in Mann–Whitney test

Similarly, in LPS-injected developing rats (Fig. 3b), optimal up-regulations corresponded to IL1ß (sixfold increase in the mean level of transcripts) and NOX2 genes (fourfold increase). In contrast with adults, LPS-injected developing rats showed down-regulation of NOS2 and IL6 genes, whereas the level of TNF-α transcript did not significantly differed from those measured in control non-injected animals (Fig. 3b).

Effect of RF on the Expression of Proinflammatory Genes

We then examined whether the expression of LPS-modulated genes could be affected by a single exposure to RF. LPS-injected animals were exposed “head only” for 2 h to GSM-1800 MHz. Exposures were performed under light anesthesia to prevent uncontrolled head movements that could strongly modify the local intensity of the RF (SAR level) delivered to the investigated dorso-medial cerebral cortex (Fig. 2). Anesthesia also avoided acute stress that would have been generated by a 2-h restrained condition necessary to expose awake animals, which could interfere with CNS cell responses to RF. The levels of LPS-modulated transcripts were assessed 24 h after exposure. For developing rats, comparison of sham vs. RF-exposed animals revealed that GSM signals induced significant 60 and 56% reductions in the mean levels of IL1ß and NOX2 transcripts, respectively. The expressions of the two other LPS-modulated genes (NOS2 and IL6) were not significantly affected by the RF (Fig. 4a). The effects of GSM signals on the levels of IL1ß and NOX2 transcripts were transient and did not persist 72 h following RF exposure (Fig. 4b).

Fig. 4
figure 4

Effect of GSM-1800 MHz on LPS-regulated mRNAs in developing rats. Fourteen-day-old rats were injected with LPS 24 h before exposure (LPS + GSM) or sham exposure (LPS + sham) to GSM signals. Cerebral cortex was collected 24 h (a) or 72 h (b) following exposure. mRNA levels of LPS-responsive genes were determined by RT-qPCR and normalized to mean values from sham-exposed animals. Data are means ± SEM from five animals. *p < 0.05, Mann–Whitney test

In adult rats, exposure to RF resulted in a significant 50% reduction in the mean level of IL1ß transcripts assessed 24 h after exposure. The RF had no clear effect on the abundance of transcripts encoded by other LPS-responsive genes such as TNF-α, CCL2, or NOS2 (Fig. 5a). The mean level of NOX2 transcripts showed a trend toward reduction that did not reach significance (p = 0.3) (Fig. 5a). This trend mostly reflected unexpected heterogeneity in the levels of NOX2 transcripts assessed in the group of sham-exposed adult rats, due to a contrastingly high level of NOX2 transcripts (2.69 time higher than the mean value of the group) in one of the six sham-exposed animal. As in developing rats (Fig. 4b), the effect of GSM signals on cortical IL1ß transcripts could not be detected 72 h following RF exposure (Fig. 5b).

Fig. 5
figure 5

Effect of GSM-1800 MHz on LPS-regulated mRNAs in adult rats. Animals were injected with LPS 24 h before exposure to GSM signals (LPS + GSM) or sham exposure (LPS + sham). Cerebral cortex was collected 24 h (a) or 72 h (b) following exposure and assessed for mRNA levels of LPS-responsive genes. Data are means ± SEM from six animals (n = 6). *p < 0.05, Mann–Whitney test

To verify that the proinflammatory gene responses to RF required LPS stimulation, we assessed the level of transcripts in the cerebral cortex of adult rats 24 h after sham or GSM exposure, in the absence of LPS treatment. Although there was a tendency for an increase in several transcripts, there was no significant difference between GSM-exposed and sham-exposed rats (p > 0.38 for any of the assessed genes) (Fig. 6).

Fig. 6
figure 6

RT-qPCR analyses of the cerebral cortex collected from adult rats 24 h after GSM or sham exposure in the absence of LPS injection (GSM alone, sham alone). Data are means ± SEM from six animals. The differences between GSM- and sham-exposed groups are not significant (p > 0.05 in Mann–Whitney test)

Effect of RF on Microglial Morphology in LPS-Injected Rats

In young and adult rats, cortical microglia display a typical ramified cell morphology marked by the occurrence of highly motile and branched cell processes. These processes are continuously remodeled through cycles of extensions and retractions, which are responsive to microglial activators such as LPS treatment (Orr et al. 2009; Walker et al. 2014). To determine whether RF impacted the morphology or density of LPS-activated microglia in the cerebral cortex, we performed immunofluorescent detections of Iba1, a cytosolic protein constitutively and selectively expressed by microglia, which is commonly used to reveal microglial cell morphologies in mammals (Kettenmann et al. 2011; Walker et al. 2014). In LPS-injected developing rats, the morphologies of Iba1-stained cortical microglia were very similar in brains that were fixed 24 h after sham or RF exposure (Fig. 7a). Counts of microglial cell bodies and quantitative assessment of the area covered by Iba1 immunoreactivity showed that RF had no effect on the density of microglial cell bodies and did not change the extent of the cortical domain occupied by microglial cell bodies and processes (Fig. 7b). In contrast, examination of microglia in cortical section from LPS-injected adult rats revealed significant cell responses to RF. Twenty-four hours after GSM exposure, microglial cell processes were more developed in GSM-exposed than in sham-exposed adult animals (Fig. 8a). Quantifications showed that the mean area of Iba1 immunoreactivity was increased by 38% in GSM-exposed rats compared with that in sham-exposed animals (Fig. 8b), whereas counts of Iba1-positive cell bodies showed no change in the number of microglial cells (Fig. 8c). Further image analyses showed that the area of Iba1-stained cell bodies was not modified by the RF (Fig. 8c); thus, the change in the bulk Iba1-positive area reflected increases in the abundance or the length of microglial processes. The enrichment in microglial cell processes was confirmed by a significant increase (24%) in the mean area of CD11b immunoreactivity (Fig. 8b), which labels an integrin subunit selectively localized to microglial membranes in the brain parenchyma (Kettenmann et al. 2011). RT-qPCR analyses showed that GSM exposure had no significant effect on the cortical level of Iba1 or CD11b messenger RNA (mRNA), indicating that the enhanced spatial coverage of microglial cell processes was not associated to increases in the expression of Iba1 or CD11b genes (Fig. 8d). As for the effect on IL1ß gene (Fig. 5), the GSM-induced modifications in microglial cell processes were transient. Quantifications on brains fixed 72 h after GSM or sham exposure showed no difference in the cortical area of Iba1 immunoreactivity or in the microglial cell body density (Fig. 8e).

Fig. 7
figure 7

Histoquantitative analyses of cortical microglia in LPS-injected developing rats. a Representative views of microglia stained with anti-Iba1 antibody in coronal sections of the dorso-medial cortex from LPS-injected rats that were fixed 24 h after sham (LPS + sham) or GSM (LPS + GSM) exposure. Scale bar: 15 μm. b Spatial coverage of anti-Iba1-stained microglia and microglial cell body densities 24 h after sham or GSM exposure. Data represent the area of anti-Iba1 staining normalized to mean values from sham exposed animals and counts of anti-Iba1-stained microglial cell bodies (means ± SEM from seven animals). The differences between GSM- and sham-exposed groups are not significant (p > 0.05 in Mann–Whitney test)

Fig. 8
figure 8

Effect of GSM-1800 MHz on cortical microglia in LPS-injected adult rats. a Representative stacked views of microglia stained with anti-Iba1 antibody in coronal sections of the dorso-medial cortex from LPS-injected rats that were fixed 24 h after sham (LPS + sham) or GSM (LPS + GSM) exposure. Scale bar: 20 μm. b, c Spatial coverage of microglial markers and microglial cell body area or densities 24 h after sham or GSM exposure. Data represent the area of anti-Iba1 or anti-CD11b staining normalized to mean values from sham exposed animals (b), counts of anti-Iba1-stained microglial cell bodies and the area of microglial cell bodies (c). d RT-qPCR analyses of Iba1 and CD11b mRNA in cortical extracts collected from LPS-injected rat, 24 h after sham or GSM exposure. The mRNA levels were normalized to mean values in sham-exposed animals. e Area of anti-Iba1 staining and microglial cell body densities in the cerebral cortex of LPS-injected rats 72 h after sham or GSM exposure. All the data are means ± SEM from six animals, *p < 0.05, Mann–Whitney test

Effects of RF on the Expression and the Phosphorylation of AMPAR Subunits

Emerging evidence indicates that activated microglial cells and proinflammatory mediators, such as IL1ß, affect glutamate neurotransmission by acting at pre- or postsynaptic levels. We therefore looked for a possible effect of GSM signals on glutamate neurotransmission, focusing on the expression and phosphorylation of AMPAR subunits GluA1 and GluA2. Western blot comparisons of cortical extracts prepared from LPS-injected adult rats sacrificed 24 or 72 h after GSM or sham exposure showed no global effect of RF on the level of expression of the GluA1 subunit (Fig. 9a, b, left panel in b). However, 24 h following GSM exposure, the proportions of phosphorylated Ser 831- and Ser 845-GluA1 were reduced by 50 and 88%, respectively (Fig. 9a, b, middle and right panels in b). The difference in the rate of phosphorylated GluA1 isoforms did not persist over time and was not observed at the 72-h time point (Fig. 9b). In contrast, GSM exposure had no significant effect on the levels of total GluA2 or on the phosphorylation of GluA2-Ser-880 residues, which were assessed 24 and 72 h after GSM or sham exposure in LPS-injected rats (Fig. 10).

Fig. 9
figure 9

Effect of GSM-1800 MHz on GluA1 expression and phosphorylation in LPS-injected rats. a Western blot detection of total GluA1, phosphorylated Ser831- or Ser845-GluaA1 isoforms (pS831-GluA1, pS845-GluA1), and ß tubulin in cortical extracts prepared from LPS-injected adult rats 24 h after sham (six animals S1 to S6) or GSM exposure (E1 to E6). The bands were visualized using the Odyssey device. MW molecular weights markers. b Densitometric analyses of the bands revealed by Western blot detection of total or phosphorylated GluA1 in cortical extracts prepared 24 or 72 h after sham (LPS + sham) or GSM (LPS + GSM) exposure. Tubulin served as an internal standard. Data represents the levels of total GluA1 or the ratio of phosphorylated Ser831- or Ser845-GluA1 isoforms (pSer-GluA1 to total GluA1 ratio) normalized to mean values from sham animals (mean ± SEM form six animals; p value in Mann–Whitney test; *p < 0.05)

Fig. 10
figure 10

Western blot analyses of total GluA2 and phosphorylated Ser880-GluA2 isoform (pS880-GluA2) in the cerebral cortex of LPS-injected rats 24 or 72 h after sham or GSM exposure. a Immunodetection of GluA2 isoforms and ß tubulin in extracts prepared 24 h after sham or GMS exposure. b Densitometric analyses of the bands revealed in extracts prepared 24 or 72 h after sham (LPS + sham) or GSM (LPS + GSM) exposure. Levels of total GluA2 and pS880-GluA2 to total GluA2 ratio were normalized to mean values from sham animals (mean ± SEM from six animals). The differences between GSM- and sham-exposed groups are not significant (p > 0.05 in Mann–Whitney test)

To determine the role of LPS in the effect of RF on GluA1 phosphorylation, we analyzed AMPAR subunits in adults that were exposed, or not, to GSM signals without any pretreatment with LPS. Comparison of cortical extracts from untreated rats sacrificed 24 h after GSM or sham exposure showed no significant effect of the RF on the level of expression or on the phosphorylation of GluA1 or GluA2 subunits (Fig. 11). Thus, LPS treatment is necessary to promote the effect of GSM signals on AMPAR phosphorylation.

Fig. 11
figure 11

Western blot analyses of total and phosphorylated GluA1 and GluA2 isoforms in cortical extracts collected from adult rats 24 h after GSM or sham exposure in the absence of LPS injection (sham alone, GSM alone). Data are means ± SEM from six animals. The differences between GSM- and sham-exposed groups are not significant (p > 0.05 in Mann–Whitney test)


Our study takes advantage of a classical rodent model of acute neuroinflammation triggered by LPS injection to uncover possible effects of RF in the cerebral cortex of adult or developing rats. Proinflammatory gene responses to LPS showed differences between adult and developing rats. A single 2-h head-only exposure to GSM-1800 MHz at a local cortical SAR of 2.9 W/kg resulted in a down-regulation of LPS-stimulated expression of IL1ß mRNA in both adult and developing rats, and led to a significant reduction in the level of NOX2 mRNA in developing but not in adult rats. These effects could be detected 24 h after exposure to RF, but they were no longer present 72 h after the exposure. In the adult cerebral cortex, the reduced expression of IL1ß gene corresponded to an increase in the spatial domain covered by microglial processes and it was accompanied by post-translational modification of AMPAR marked by reduced levels of phosphorylated Ser 845- or Ser 831-Glu A1 isoforms.

Proinflammatory Gene Responses to LPS Are Different in Adult and Developing Rats

In both adult and developing rats, peripheral injection of LPS triggers a sickness syndrome involving changes in body core temperature and behavior, associated with neuroendocrine and neuroinflammatory responses (Dantzer et al. 2008; Eklind et al. 2006; Hoogland et al. 2015). Febrile and neuroendocrine responses were shown to differ according to developmental stages (Boisse et al. 2004; Dent et al. 1999), but only a few studies have addressed the variations in intracerebral production of proinflammatory mediators. For example, Iwasa et al. (2011) reported that LPS-triggered up-regulation of IL1ß gene is significantly reduced in the hypothalamus of 2-week-old rats compared with that in young adults. Assessing transcript levels in the cerebral cortex 24 h after LPS injection, we confirmed that the up-regulation of IL1ß transcript level is more prominent in adults (12-fold increase) than in 2-week-old rats (4-fold increase). On the other hand, our data also revealed further differences in the response of proinflammatory genes (see Fig. 3). In particular, LPS treatment had opposite effects on the level of NOS2 transcripts in developing vs. adult rats, but more detailed analyses on the time course of these changes are required to look for possible biphasic modulation of NOS2 gene either in adult or in developing rats. The differences in neuroinflammatory profiles could possibly arise from developmental changes in intracerebral immune cells, as emphasized by age-dependent differences in the transcriptome of microglia (Bennett et al. 2016). It could also involve changes in the systemic immune signaling that is directly triggered by peripherally injected LPS and which, in turn, causes CNS cell responses (Dantzer et al. 2008).

Exposure to GSM-1800 MHz Reduces LPS-Triggered Expression of Proinflammatory Genes in Developing and Adult Rats

In spite of the differences observed in the neuroinflammatory profiles induced by LPS, the exposure to GSM signals triggered a transient reduction in the level of IL1ß transcript in both adult and developing rats, which was detectable 24 h after a single exposure to the RF. The levels of NOX2 transcript were also significantly reduced in GSM-exposed developing rats but not in exposed adult rats. This difference could stem from the age-dependent change in the absorption of RF energy, which is emphasized by the slight enhancement of the brain-averaged SAR in developing rats compared to adults, even though the mean cortical SAR values were not statistically different (Table 1). However, it is possible that the immature stage of development also promoted the NOX2 gene response to RF.

Interestingly, among the assessed proinflammatory genes, IL1ß and NOX2 were distinguished by an optimal up-regulation at the time of GSM exposure (4- to 12-fold increases in transcript levels). This suggests that the gene responses to RF depend on their level of modulation by the inflammation process. Accordingly, we observed no significant alterations in the levels of cytokine or enzyme transcripts that were assessed in animals exposed to RF without pretreatment with LPS. Consistent with this observation, Hirose et al. (2010) found that in primary cultures of rat microglia, the basal production of IL1ß, TNF-α, or IL6 was not affected by a 2-h cell exposure to modulated 1950 MHz RF applied at a SAR of 2 W/kg. Moreover, Thorlin et al. (2006) could not detect any changes in IL6 or TNF-α production assessed in rat microglial cultures exposed to GSM-900 MHz at a SAR of 3 W/kg for up to 8 h. In contrast, Lu et al. (2014) reported that a 6-h exposure to GSM-1800 MHz at an estimated SAR of 2 W/kg is sufficient to significantly increase the levels of IL1ß, TNF-α, and IL6 transcripts in the mouse N9 microglial cell line. The discrepancies in the reported effects could arise from differences in the exposure systems, or in microglial capacities to sense or to respond to RF according to experimental conditions. Notably, transcriptome profiling has shown important differences between N9 microglial cell line, cultured primary microglia, and in vivo microglia (Butovsky et al. 2014; Gosselin et al. 2014). In our study, reduced IL1ß gene expressions were observed in LPS-treated rats exposed to RF at estimated brain average SAR of 1.7 to 2.2 W/kg (Table 1). In contrast, Bouji et al. (2012) reported that the levels of intracerebral IL1ß were increased in 12-month-old rats following a single head exposure to GSM-900 MHz, at a brain average SAR of 6 W/kg. Beyond parameters such as LPS treatment, age of the animals, or RF spectrum, the marked difference in the SAR values can potentially explain the opposite responses reported for the IL1ß gene.

It is established that the rise in brain IL1ß plays important role in the early symptoms triggered by LPS injection, which comprise fever and a sickness behavior characterized by depressed locomotor activity, social withdrawal, reduced food and water intake, increased slow wave sleep, and impaired cognition (Konsman et al. 2002; Maier 2003). These symptoms are commonly observed in infectious states and are thought to favor appropriate metabolic responses and immune adaptive reactions required to neutralize pathogens (Dantzer 2001). In our study, we chose to avoid exposure manipulation during the febrile period. This led us to assess the effect of RF on IL1ß gene expression 48 h after LPS injection at the earliest (Fig. 1), a time when sickness behavior had already vanished or was attenuated enough to escape detection in the absence of specific behavioral testing. Although the reduction in IL1ß expression raises the possibility that RF exposure could alleviate the sickness syndrome, we found no significant effect of RF on the LPS-triggered stimulation of the TNF-α gene, which is also known to promote the sickness behavior (McCusker and Kelley 2013). Further investigation will be needed to determine whether RF exposure can modify the LPS-triggered sickness behavior.

GSM Signals Trigger Changes in the Morphology of LPS-Activated Microglia and Post-Translational Modifications of AMPAR in Adult Rats

We found that in adult rats, the GSM-induced reduction in IL1ß mRNA levels was associated with an increase in length, or the abundance, of microglial processes in the adult cerebral cortex. This provides direct evidence that RF can affect pre-activated microglial phenotypes. The mechanisms of this effect may be considered in light of the remarkable motility of these cells. In physiological conditions, microglia undergo permanent and fast remodeling of their cell processes, which are thought to play key roles in the microglial control of neuronal wiring, neuronal firing, and synaptic plasticity (Tay et al. 2016). The velocity of cell process extension in cortical microglia was reported to be unchanged or increased in response to systemic treatments with LPS, although retractions of cell processes are prominent during early phases of the microglia response to LPS (Gyoneva et al. 2014; Kondo et al. 2011). Thus, GSM exposures are likely to promote microglial coverage of intercellular spaces through stimulation of cell process extensions and/or limitation of retractions during the first 24 h following GSM exposure. While observed in the adults brains, the change in microglial cell morphology could not be detected in the brain of developing rats examined 24 h after GSM exposure. This difference further emphasizes that the brain responsiveness to RF varies according to the age under neuroinflammatory conditions. However, we cannot rule out that in the developing rats, transient morphological cell responses occurred earlier or later than 24 h following GSM exposure. It is also possible that in the developing brain, the RF-induced down-regulation of NOX2 gene impacted morphological cell response to RF, considering studies showing that NOX2 controls reorganization of the actin-cytoskeleton and cell motility in different cell types including microglia (Lelli et al. 2013; Valdivia et al. 2015).

Current evidence indicates that the activation state of microglia and the remodeling of their cell processes are tightly controlled by components synthesized or released by neighboring neural cells, including purines and neurotransmitters (Hristovska and Pascual 2016; Kettenmann et al. 2011). In particular, the balance between microglial cell process extension and retraction in LPS-treated rodents depends on the availability of extracellular ATP and adenosine, and on the type of purine receptors expressed by activated microglial cells (Orr et al. 2009). Interestingly, it was shown that excitatory neurotransmission mediated by AMPAR regulates microglial cell process motility and branching processes through the release of ATP (Fontainhas et al. 2011). In our study, we probed functional states or recruitment of AMPAR through the detection of phosphorylated AMPAR isoforms. Our data suggest that in GSM-exposed animals, the enlarged domain occupied by microglial processes occurred together with reduced AMPAR-mediated neurotransmission. We found that GSM exposure resulted in a profound decrease in the proportion of phosphorylated Ser 831 and Ser 845 residues in the C terminal region of GluA1, while phosphorylation of GluA2-Ser 880 was not modified. It has been shown that phosphorylation of GluA1-Ser831 increases conductance of AMPAR channel (Derkach et al. 1999), whereas phosphorylation of GluA1-Ser845 enhances mean open probability of homomeric GluA1 receptors (Banke et al. 2000). Furthermore, phosphorylation of GluA1-Ser845 promotes AMPAR trafficking to extrasynaptic membrane and primes extrasynaptic receptors for synaptic insertion (He et al. 2009; Incontro et al. 2013; Oh et al. 2006). Conversely, dephosphorylation of GluA1-Ser 845 favors AMPAR endocytosis (Ehlers 2000; He et al. 2009; Oh et al. 2006). Therefore, by reducing phosphorylation of GluA1-Ser 845 and Ser 831, GSM exposure might possibly alter incorporation, or stabilization, of AMPAR subunits in cortical synapses, and reduce ionic currents through AMPAR channels. Consistent with this assumption, Xu et al. (2006) reported that the amplitude of AMPA miniature excitatory postsynaptic currents is decreased in cultured hippocampal neurons, when the cells are repeatedly exposed to GSM-1800 at a SAR of 2.4 W/kg. However, our data provide evidence that the GSM-induced reduction in the phosphorylation of GluA1 is linked to a neuroinflammatory state, as it was not observed in GSM-exposed rats that were not injected with LPS. Previous studies have shown that double injections of LPS can trigger reductions in the levels of GluA1 subunits incorporated in the plasma membrane of the prefrontal cortex or the ventral tegmental area (Sekio and Seki 2015). This effect is thought to contribute to the development of a depressive-like behavior induced by LPS (Sekio and Seki 2015), which follows the early LPS-triggered sickness behavior (Dantzer et al. 2008; Frenois et al. 2007). Epidemiological investigations have addressed the possibility that depressive symptoms could be favored in inhabitants living nearby mobile phone base stations (Pall 2016), but to our knowledge, the impact of RF on depressive-like behaviors has not been assessed in preclinical rodent models. It remains to determine whether GSM exposure has any detectable effect on the LPS-induced depressive behavior.

Similar to microglia and IL1ß gene responses, GSM-induced changes in GluA1 phosphorylation were transient: they were observed 24 h but not 72 h following GSM exposure. This temporal relationship is remarkable in light of previous evidence showing that AMPA neurotransmission controls morphological remodeling of microglia (Fontainhas et al. 2011) and that conversely, activated microglia can impact excitatory neurotransmission at pre- or postsynaptic levels, including through the reduction of AMPAR phosphorylation (Liu et al. 2015; Zhang et al. 2014). We therefore propose that under a LPS-triggered neuroinflammatory state, GSM exposure may affect crosstalk between cortical neurons and activated microglia.


The effect of GSM-associated RF on the CNS affected by an inflammatory process is an important issue considering the broad incidence of human pathologies involving neuroinflammation. Using a rat preclinical model, our study provides evidence that a single head exposure to GSM-1800 MHz at a local cortical SAR value of 2.9 W/kg is sufficient to modulate proinflammatory genes and microglial cell responses to a peripheral immune challenge, and triggers changes in functional markers of excitatory neurotransmission. Our data also indicate that an acute neuroinflammation promotes cortical cell responses to RF, because these responses were not detected in the healthy rat CNS exposed to RF. So far, however, these cell responses were observed under RF exposure at a power above the limits set for the safety of the general public (2 W/kg). Clearly, the impact of neuroinflammation on the brain responses to RF deserves further investigation, which will help to determine potentially beneficial or harmful effect of RF, according to neuropathological states.