Introduction

With the widespread use of mobile phones, the possible neurobiological effects of radiofrequency electromagnetic fields (RF EMF) have become a crucial question.

Glial reactivity was reported after an acute RF EMF exposure to the rat brain at an averaged specific absorption rate (SAR) of 6 Watts/kg (W/kg). Mausset-Bonnefont et al. (2004) showed an increase in glial fibrillary acidic protein (GFAP) staining in the hippocampus, the striatum, and the cortex 3 days after exposure. This data was reproduced by Brillaud et al. (2007) who showed GFAP staining increase in the striatum and the frontal cortex 2 to 3 days after the same exposure. Whether RF EMF leads to gliosis is an important matter of debate as data from the literature are contradictory. Mausset et al. (2001), Mausset-Bonnefont et al. (2004), Brillaud et al. (2007), Ammari et al. (2008a, b), Maskey et al. (2010, 2012), and Lu et al. (2014) looked at the link between SAR and GFAP increase using acute or repeated RF EMF exposures and reported positive effects. Fritze et al. (1997), Thorlin et al. (2006), Grafstrom et al. (2008), Watilliaux et al. (2011), Bouji et al. (2012), Court-Kowalski et al. (2015), Bouji et al. (2016), and Petitdant et al. (2016) did not show any positive effect. Astrocytes insure a variety of functions in the brain. Neurologic function could be altered in some important respects if RF EMF mobile phone exposures did cause astrocytic reaction. This question needs new insights.

Astrogliosis is a nonspecific consequence of many insults. It aims at protecting the brain against deleterious processes as excitotoxic glutamate neurotransmitter excess, neuron lesions, oxidative stress, neuroinflammation, or toxicity of the beta amyloid peptide (Aβ). Increased staining of GFAP, the main intermediate filament of astrocytes, is the hallmark of reactive astrogliosis (Eng 1985; Schiffer et al. 1986). Mechanisms of this response are rearrangements of the intermediate filament network, including breakdown of F-actin stress fibers, polymerization of upregulated GFAP, and vimentin and nestin overexpression. Increased GFAP staining is not always linked to significant rise in protein content, but may be correlated to cytoskeleton organization changes, usually detectable by variations in GFAP detergent solubility (Boran and Garcia 2007). The consequent cell morphology modifications, as processes hypertrophy, reflect metabolic activation, as shown by increased organelles number or size, and release of chemical factors (cytokines) in the case of mild to moderate astrogliosis (Schmidt-Kastner and Szymas 1990; Sofroniew 2009).

Astrogliosis can be secondary to demyelinating lesions. Progenitor cell differentiates into myelinating oligodendrocytes to reestablish myelination in the demyelinated zones (Nishiyama et al. 2009; Zhao et al. 2009). Oligodendrocytes selectively express myelin basic protein (MBP). The transitory decrease in brain MBP expression was shown after exposure to several environmental contaminants (chemicals, plants, medicines, or metals) (Coria et al. 1984; Schang et al. 2013; Bhasker et al., 2014; Ma et al. 2015). MBP increase was also shown within 2 to 3 days post-injury in toxicological conditions.

In healthy organisms, astrocytes are involved in neuronal electrical activity regulation through reciprocal communications in the tripartite synapses (Wenzel et al. 1991; Rapanelli et al. 2011). Astrocytes are in a symbiotic relationship with neurons to optimize neural information processing, to modulate synaptic transmission and neuronal activity (Hewett 2009). Neuronal electrical activity in the subcortical structures (as the striatum and the hippocampus) and in cortical regions participates to generate mental functions: memory, emotionality, and novelty perception. Astrocyte-released pro-inflammatory molecules may affect these functions and adaptative behaviors (Eddleston and Mucke 1993; Rothstein et al. 1996). Severe astrogliosis may induce collateral neurotoxicity on biomarkers and cognitive functions by exacerbating the deleterious processes (Sofroniew and Vinters 2010; Middeldorp and Hol 2011). Memory performances have been shown to be impaired by lipopolysaccharide-induced inflammatory (Lee et al. 2009; Spitzer et al. 2010). Some authors indicated that long-term RF EMF exposures impaired memory performances and modified the levels of Aβ (Arendash et al. 2010; Dragicevic et al. 2011). Amyloidogenesis has been the endpoint of recent studies in health assessments of environmental contaminants. Aβ synthesis was shown to be stimulated by several neurotoxic agents including medications, metal mixtures, hormones, or ozone (Blasko et al. 2008; Ma and Liu 2015; Ashok et al. 2015; Deng et al. 2015; Hernandez-Zimbron and Rivas-Arancibia 2015).

Because of the important neurological consequences if RF EMF mobile phone exposures did cause memory impairments, this question needs new insights. Up to date, there are no consensual conclusions on the possible effect of RF EMF exposure on memory. Both positive (Preece et al. 1999; Koivisto et al. 2000; Krause et al. 2000; Maier et al. 2004; Keetley et al. 2006; Nittby et al. 2008; Arendash et al. 2010) and negative reports (Sienkiewicz et al. 2000; Dubreuil et al. 2002; Haarala et al. 2007; Ammari et al. 2008a, b; Petitdant et al. 2016; Bouji et al. 2016) were published.

Here, we aimed to reproduce the neurotoxicity of an acute high SAR RF EMF exposure (Mausset-Bonnefont et al. 2004; Brillaud et al. 2007). Rats were exposed at 0 and 1.5 W/kg for 15 min, 6 W/kg for 15 min, or 6 W/kg for 45 min. Responses to a single exposure aimed to assess mechanisms for RF EMF and nervous tissue interaction, which are not necessarily detectable using intermittent exposures due to habituation or sensitization processes of the organism. The exposure durations of 15 and 45 min are compatible with the duration of daily calls in moderate and intensive mobile phone users (INTERPHONE group 2010; Hardell and Carlberg 2013). In order to test if biological responses would be in the direct ratio of the total absorbed energy, four rising levels were tested by increasing the brain averaged SAR (BASAR) and the exposure duration of the highest BASAR (to 45 min). We evaluated emotional memory, anxiety-like behavior, and locomotor activity. Tests were performed in the fear conditioning paradigm, the elevated plus maze (EPM), and the open field. Three days after RF EMF exposure, markers were assessed using enzyme-linked immunosorbent assay (ELISA). Previous studies indicated that the increase of GFAP and Aβ peptide as well as the decrease of MBP was rapidly observed, 2 to 3 days after a toxic stimulation (Morin-Richaud et al. 1998; Ma and Liu 2015). Two phenotypes of astrogliosis were quantified: the total GFAP expression and the (soluble monomeric GFAP)/(total GFAP content) ratio (Eng et al. 2000; Tseng et al. 2006; Maskey et al. 2010; Watilliaux et al. 2011; Lu et al. 2014), as well as MBP and Aβ 1–40. Six brain structures involved in mental functions were exanimate: the cortex (anterior and posterior), the striatum, the hippocampus, the olfactory bulb, and the cerebellum).

Material and methods

Animals

Sixty-eight 6-week-old Sprague Dawley male rats (OFA Iffa Credo, France) were daily weighed and handled for 1 week before the start of experiments. We used male rats to be consistent with previous studies from our laboratory on dosimetry, behavioral paradigms, and restraint effect. The rats were housed two per cage in controlled environmental conditions (room temperature 22 °C; 12-h light/dark cycle) with food and tap water ad libitum. The cage environment was enriched with a plastic tube identical in size, shape, and appearance to RF EMF exposure rockets (Brillaud et al. 2005). The rats were free to explore, and move in and out of the tube from their arrival in the vivarium until sacrifice. INERIS Use Committee and French State Council guidelines were followed for the care and use of laboratory animals (Decree no. 87–849, October 19, 1987).

Experimental groups

The rats were randomly assigned to four independent groups (n = 14–20 rats/group) exposed to SAR doses: 0 W/kg for 45 min, 1.5 W/kg for 15 min, 6 W/kg for 15 min, and 6 W/kg for 45 min. Restraining in the exposure rocket lasted 45 min for all the groups. On the first day, the rats were subjected to fear conditioning training, immediately followed by an acute RF EMF exposure. The contextual memory test was performed 3 h after the end of the training session (and 3 h after the beginning of RF EMF exposure). On the second day, cue memory test was performed in the morning (21 h after the beginning of RF EMF exposure) and the 24 h contextual memory test in the afternoon (25 h after the beginning of RF EMF exposure). The 24-h contextual memory test was performed 24 h following the end of the 3-h contextual memory test during which there was a recall of the context/tone/electrical shock presentation. On the third day, the rats were tested in the open field in the morning (45 h after the beginning of RF EMF exposure) and in the elevated plus maze (EPM) in the afternoon (50 h after the beginning of RF EMF exposure). The open field and the EPM tests were applied at the same day as in previous study from Petitdant et al. (2016). The rats were sacrificed on the fourth day (72 h after the beginning of RF EMF exposure).

Exposure system

Exposure setup was previously described (Mausset et al. 2001; Dubreuil et al. 2002). A radio frequency power source (900–64 type, radio frequency power amplifier, France) emitting a 900 MHz RF EMF (1/8 duty factor) pulse modulated at 217 Hz was connected to a four-output divider. Each output was connected to a loop antenna allowing local exposure of four animal’s head simultaneously in an anechoic chamber. During exposure, each animal was placed in a Plexiglas rocket capped with a truncated cone containing an individual loop antenna. The rocket body was lined with air holes to facilitate breathing and minimize the rise in body temperature. SAR calculations were made with homogeneous phantoms (Leveque et al. 2004).

Behavior

Fear conditioning

Fear conditioning is a Pavlovian conditioning, in which a neutral conditioned stimulus (a sound) is paired with an aversive unconditioned stimulus (a footshock). It occurred in two standard fear conditioning apparatus purchased from Bioseb (France). Two freezing boxes (25.5 × 27 × 26 cm) were each located inside a larger, insulated plastic chamber that provided protection from outside light and noise. The boxes were made with black methacrylate walls and a transparent front door. The floor consisted of 20 stainless steel rods through which a shock could be delivered. The rods were connected to a high-sensitivity weight transducer system that measured animal movements. The conditioning apparatus was controlled by the experimenter with specific software (Panlab, Barcelona). Freezing response was defined as complete immobilization of the rat, except for respiration (Blanchard and Blanchard 1972). It was measured in response to stimuli presentation (context and tone) by amplitude of movement of the animal and was scored as percentage of time spent freezing. At the end of each session, the animals were returned to their home cages, and the boxes were cleaned with non-alcohol disinfectant.

The training phase began with a 2-min habituation period (no auditory or aversive stimulus), followed by five training cycles (28 s of sound (2 kHz, 100 dB); 2 s of sound + electrical shock (0.25 mA); a 30-s exploration period (no auditory or aversive stimulus)). To test contextual memory, the animals were introduced for 3 min in the boxes used for the training. To avoid extinction, the 3-h contextual memory test ended with two training cycles. To perform the cue memory test, the animals were introduced in boxes with modified environment (different box size, wall color, and floor texture) for 2 min and were exposed to the sound for 3 min.

Open field

The open field test was performed according to previously described methods (Prut and Belzung 2003). Movements were monitored for 30 min using an automated apparatus consisting of four Plexiglas boxes (60 × 60 × 40 cm) equipped with 16-bottom and 8-top (for vertical activity evaluation) infrared beams located along two adjacent sides of one chamber (Imetronic, France). Activity was measured by the number of counts, i.e., the number of infrared beam interruptions. Anxiety-related behavior was assessed by the time spent in the central area of the apparatus.

Elevated plus maze

The EPM test was performed according to previously described methods (Daniels et al. 2004). The apparatus was composed of a plus shaped acryl maze with two opposite open arms and two opposite closed arms (50 cm in length, 10 cm in width, and 31 cm in height), extending out from a central platform (10 × 10 cm) and elevated 50 cm above the floor. The rats were placed in the center of the maze, the head facing an open arm, and were allowed to explore for 5 min. The maze was cleaned after each test to prevent influences of previously tested rat. Two paws had to be inside the entrance line to each arm, which signaled the start of the time spent in the specific arm, and then the end time was recorded when all four paws were outside the line again. Tests were recorded and the time spent in the different arms was analyzed using Smart software (Panlab). Data were presented as percentage of time spent in the open arm compared to the time spent in the four arms during the first 30 s of the test.

Brain sample preparation

Animals were anesthetized with isoflurane. Intra-cardiac perfusion was performed with 0.9 % NaCl solution. Following decapitation, the brains were immediately removed and placed on a cold plate. The brains were divided in two hemispheres. The striatum, the hippocampus, the cortex, the olfactory bulb, and the cerebellum were dissected from one hemisphere with curved forceps. The third ventricle was the landmark to dissect the anterior and the posterior cortex. Samples of brain tissue were snap-frozen on a bed of dry ice and then grounded into a powder. Tissue was stored at −80 °C.

Enzyme-linked immunosorbent assay

Total protein assay

Total protein was assayed by the method of Bradford. Bovine serum albumin (BSA, Sigma-Aldrich) was used as standard. Absorbance was read at 595 nm using a Molecular Devices UV Max microplate reader (Infinite M200 TECAN, Menlo Park, CA, USA) coupled to a Macintosh computer running a Soft Max (Molecular Devices, Menlo Park, CA, USA) program.

GFAP

The method was slightly modified from O’Callaghan (1991). Brain samples were weighed and homogenized using an homogenizing vessel in hot (90–95 °C) 1 % SDS in PBS supplemented with 1× protease inhibitor cocktail (1/25, Roche) to obtain the total cell fraction. Part of the total cell fraction was centrifuged (15 min 15,000g) to obtain the cytosolic fraction. Both total and cytosolic fractions were analyzed for their GFAP content. Standard curve samples were prepared with human GFAP (Merck) in sample buffer at dilutions 0.0075, 0.015, 0.03, 0.06, 0.1, 0.25, 0.5, and 1.0 μg/100 μl. Immulon-2 flat-bottom microtiter plates (Greiner Bio-One) were coated with the rabbit polyclonal anti-GFAP (Dako Cytomation) (1.0 μg/100 μl in PBS) for 1 h at 37 °C and then overnight at 4 °C. Blocking was performed with BSA 2 %. Standards and samples were incubated for 1 h in duplicates (100 μl/well). Monoclonal mouse anti GFAP (Abcam) (1:500 in BSA containing Triton ×100) was then added (for 1 h). Finally, rabbit polyclonal to mouse IgG horseradish peroxidase (HRP)-conjugated (Abcam) (1:3000 in BSA containing 0.5 % Triton ×100) was incubated for 30 min. Revelation was performed by adding 3,3′,5,5′-tetramethylbenzidine (TMB) (Interchim) and was stopped 10 min later with H3PO4 1 M (Uptima). Optical densities were read at 450 nm.

MBP

ELISA assay was performed using commercialized kits for rat MBP (USCN Life Science Inc.). Tissue samples were diluted in PBS 1× supplemented with protease inhibitor cocktail. MBP standard curve was prepared in standard diluent at 0.156, 0.312, 0.625, 1.25, 2.5, and 5 ng/ml. Blank, standard dilutions and samples were incubated for 2 h at 37 °C in duplicate in the pre-coated plate. Detection reagent containing the primary antibody anti-MBP was added (for 1 h at 37 °C). After washing, detection was performed using the reagent containing the HRP-conjugated secondary antibody (30 min at 37 °C). Finally, the substrate solution was added to each well (15–25 min at 37 °C) before stopping the reaction. Optical densities were read at 450 nm.

Amyloid β 1–40

Aβ 1–40 concentration was measured using commercialized kits allowing Aβ 1–40 detection with a 0.049 pmol/l sensitivity (Wako Chemicals, GmbH, Germany) and based on the sandwich ELISA method (Silverberg et al. 2010). Tissue samples were prepared with tissue powder extracted by sonication with 70 % formic acid and centrifugation (100,000g for 1 h). The supernatant was removed and neutralized with 20 volumes of Tris Base (1 mol/l). Samples and standards (0, 1, 2.5, 5, 10, 25, 50, and 100 pmol/l) were diluted with the kit diluent and added in duplicate to the monoclonal antibody anti-Aβ11–28 (BNT77)-pre-coated 96-well microtiter plate (for 2 h at room temperature). Then, HRP-conjugated antibody solution was incubated for 2 h at 4 °C. Finally, TMB was added to each well, and reaction was terminated 30 min later using the kit stop solution. Total Aβ 1–40 concentrations (pg/mg total protein) were expressed as percentages of sham group. Optical densities were read at 450 nm.

Statistical analysis

Statistical analyses were performed using SPSS 17 software (Inc., Chicago, IL, USA). Values are given as mean ± standard error of mean per group. Control for normal distribution was performed using Levene’s test. Analyses were performed using two-way analysis of variance (ANOVA) (for time × SAR level, time × exposure duration) and one-way ANOVA (for SAR level or exposure duration effect). SAR level effect was assessed by comparing the 0, 1.5, and 6 W/kg groups exposed for 15 min, while exposure duration effect was assessed by comparing the 0 and 6 W/kg 15 min, and 6 W/kg 45 min groups. Main effects were analyzed using Bonferroni’s post hoc corrected t test. Time was analyzed as within-subject factors while exposure doses were analyzed as between-subject factors. Significant effects were found when p < 0.05. The p value was one-tailed for total GFAP increase, MBP decrease, and Aβ increase and two-tailed for behavioral tests and cytosolic GFAP. Linear models and correlations (Pearson) were performed using R statistical environment. The linear model aimed to explain long-term memory by Aβ in the hippocampus and in the striatum and by their interactions with RF EMF exposure.

Results

Effect of RF EMF on total GFAP

Figure 1 shows total GFAP in the six cerebral structures 3 days following RF EMF exposure. ANOVA performed on each structure showed significant SAR level effect in the striatum (p = 0.03), but no exposure duration effect. There was no exposure duration or SAR level effect in the other structures (p > 0.05). Bonferroni’s post hoc corrected t test showed significant GFAP increase in the striatum in the 1.5 W/kg-exposed rats when compared to the sham-exposed group (p < 0.05).

Fig. 1
figure 1

Total GFAP in the striatum (a), the hippocampus (b), the olfactory bulb (c), the posterior cortex (d), the anterior cortex (e), and the cerebellum (f) in the sham- and RF EMF-exposed rats. Total striatal GFAP in the 1.5 W/kg-exposed rats was significantly increased compared to that in the sham-exposed group (*p < 0.05). N = 8/group

Full size image

Effect of RF EMF on cytosolic GFAP

Figure 2 shows cytosolic GFAP. ANOVA performed on the each structure showed a significant SAR level effect in the striatum (p = 0.01) and the hippocampus (p = 0.02). Both SAR level and exposure duration effects were obtained in the olfactory bulb (p = 0.04 and p = 0.05), while no significant SAR level or exposure duration effect was obtained in the other structures (p > 0.05). Post hoc Dunnett’s-corrected t test showed that soluble GFAP was significantly decreased in the 1.5 W/kg-exposed group when compared to the sham group in the striatum (p = 0.05) and significantly increased in the 6 W/kg–15 min-exposed group when compared to both the sham and the 1.5 W/kg–15 min-exposed groups in the hippocampus (p = 0.05) and in the olfactory bulb when compared to the sham group (p < 0.05).

Fig. 2
figure 2

Cytosolic GFAP in the striatum (a), the hippocampus (b), the olfactory bulb (c), the posterior cortex (d), the anterior cortex (e), and the cerebellum (f) in the sham- and RF EMF-exposed rats. Cytosolic striatal GFAP in the 1.5 W/kg-exposed rats was significantly decreased compared to that in the sham-exposed group (*p < 0.05). Cytosolic GFAP in the hippocampus and the olfactory bulb in the 6 W/kg (15 min)-exposed rats was significantly increased compared to that in the sham-exposed group (*p < 0.05). N = 8/group

Full size image

Effect of RF EMF on MBP

Figure 3 shows MBP in the six cerebral structures 3 days following RF EMF exposure. ANOVA performed on MBP for each structure did not show any SAR level effect (p > 0.07) or exposure duration effect (p > 0.09).

Fig. 3
figure 3

MBP in the striatum (a), the hippocampus (b), the olfactory bulb (c), the posterior cortex (d), the anterior cortex (e), and the cerebellum (f) in the sham- and RF EMF-exposed rats. No RF EMF effect was reported. N = 14/group

Full size image

Effect of a post-conditioning RF EMF exposure on emotional memory

Figure 4a–c shows time spent freezing in response to the context 3 and 24 h after the conditioning, and to the tone, respectively. Figure 5d shows the comparison between the 3- and 24-h tests. ANOVA performed separately on the three tests (Fig. 4a–c, respectively) showed significant time effect (F (35,2240) = 4.7, p < 0.001; F (35,2240) = 10, p < 0.001; and F (35,2240) = 15.9, p < 0.001), but no SAR level effect (F (2,45) = 0.6, p = 0.6; F (2,45) = 0.9, p = 0.4; and F (2,45) = 1, p = 0.4), time × SAR level interaction (F (70,1575) = 1.1, p = 0.4; F (70,1575) = 1.1, p = 0.3; and F (70,1575) = 0.8, p = 0.9), exposure duration effect (F (2,51) = 0.7, p = 0.5; F (2,51) = 0.4, p = 0.7; and F (2,51) = 0.4, p = 0.6), and exposure duration × time interaction (F (70,1785) = 0.8, p = 0.9; F (70,1785) = 0.9, p = 0.7; and F (70,1785) = 0.8, p = 0.8). ANOVA performed to compare contextual memory at 3 and 24 h showed no significant test delay × SAR level interaction (F (3,64) = 0.4, p = 0.7) or SAR level effect (F (3,64) = 0.9, p = 0.5) but a significant test delay effect (F (1,64) = 4.1, p = 0.04). Bonferroni’s post hoc corrected t test showed that contextual memory was significantly decreased at 24 h when compared to the test at 3 h in the 6 W/kg (15 min)-exposed group (p = 0.03) (Fig. 4d). Exploration of the data using a linear model indicated that long-term memory was explained by the level of GFAP in the striatum and by its interaction with RF EMF exposure (1.5 W/kg). The calculation of the Pearson coefficient indicated that long-term memory scores correlated with total striatal GFAP (r = −0.38, p = 0.047) and with short-term memory (r = 0.38, p = 0.0013).

Fig. 4
figure 4

Emotional memory in the sham- and RF EMF-exposed rats. There was no RF EMF effect on the time spent freezing to the context (after 3 h (a) and after 24 h (b)) and to the tone (c). There was a significant RF EMF effect when comparing the short- and long-term memories to the context (d). The 1.5 W/kg-exposed rats decreased significantly there contextual memory over time (*p < 0.05). N = 17/group

Full size image
Fig. 5
figure 5

Anxiety-like behavior (a) and locomotor activity (b, c, d) in the sham- and RF EMF-exposed groups. No RF EMF effect was reported in the percentage of time spent in the open arms of the EPM (a). There was no RF EMF effect on the vertical (b) and horizontal (c) activities measured in the open field (N = 14/group) and on the time spent in the central area (d) (N = 8/group)

Full size image

Effect of RF EMF exposure on anxiety-like behavior and locomotor activity

Figure 5a shows the percentage of time spent in the open arm of the EPM. No effect was detected with ANOVA comparing SAR level (F (2,43) = 0.8, p = 0.4) or exposure duration (F (2,49) = 1.6, p = 0.2). Figure 5b, c shows vertical and horizontal locomotor activities measured in the open field. ANOVA performed on vertical activity showed significant time effect (F (5,105) = 26.1, p < 0.001) but no SAR level effect (F (2,21) = 0.2, p = 0.8), exposure duration effect (F (2,21) = 0.01, p = 0.9), time × SAR level interaction (F (10,105) = 0.5, p = 0.9), or time × exposure duration interaction (F (10,105) = 0.4, p = 0.9) (Fig. 5b).

ANOVA performed on horizontal activity, comparing SAR levels or exposure durations, showed significant time effect (F (5105) = 85.7, p < 0.001 and F (5105) = 79.4, p < 0.001, respectively) but no SAR level effect (F (2,21) = 0.2, p = 0.8), exposure duration effect (F (2,21) = 0.04, p = 0.9), time × SAR level interaction (F (10,105) = 0.9, p = 0.5), or time × exposure duration interaction (F (10,105) = 1.0, p = 0.4) (Fig. 5c). Similarly, there was a significant time effect (p < 0.05) but no SAR level effect, exposure duration effect, time × SAR level interaction, or time × exposure duration interaction (p > 0.05) for the time spent in the central area of the open field (Fig. 5d).

Effect of RF EMF on Aβ 1–40

Figure 6 shows Aβ 1–40 quantities in the six cerebral structures 3 days following RF EMF exposure. ANOVA performed on Aβ 1–40 for each structure did not show any SAR level or exposure duration effect (p > 0.2). Exploration of the data using a linear model indicated that long-term memory was explained by the level of Aβ 1–40 in the striatum and by its interaction with RF EMF (1.5 W/kg). Long-term memory was explained by the interactions of RF EMF exposure with Aβ (in the striatum, p = 0.022 and in the hippocampus, p = 0.047). Long-term memory was explained by Aβ in the striatum (estimate −0.74, standard error 0.31, p = 0.02), by its interactions with RF EMF exposure (at 1.5 W/kg, estimate 1.38, standard error 0.64, p = 0.04 and at 6 W/kg 15 min, estimate 1.38, standard error 0.64, p = 0.04), and by the interaction Aβ in the hippocampus × RF EMF exposure (at 6 W/kg 15 min, estimate −1.66, standard error 0.64, p = 0.02). Multiple R-squared was 0.38 and adjusted R-squared was 0.18 (p = 0.10).

Fig. 6
figure 6

Aβ 1–40 in the striatum (a), the hippocampus (b), the olfactory bulb (c), the posterior cortex (d), the anterior cortex (e), and the cerebellum (f) in the sham- and RF EMF-exposed rats. No RF EMF effect was reported. N = 8–17/group

Full size image

Discussion

This study aimed to investigate the neurobiological effect of an acute RF EMF exposure at high SARs. Results showed a +114 % GFAP increase in the striatum following RF EMF exposure at 1.5 W/kg. At 6 W/kg (15 min), GFAP cytosolic fraction was increased in the hippocampus (+119 %) and the olfactory bulb (+46 %), and long-term contextual memory was decreased (−21 %). No MBP or Aβ 1–40 expression modification was shown.

Astrocytes insure a variety of functions in the brain. Thus, neurologic function could be altered in some important respects if RF EMF mobile phone exposures did cause astrocytic reaction. Here, we afforded new insights to this question, which is still an important matter of debate. Several previous investigations using acute and repeated RF EMF exposures looked at the link between SAR and GFAP increase. Mausset et al. (2001), Mausset-Bonnefont et al. (2004), Brillaud et al. (2007), Ammari et al. (2008a, b), Maskey et al. (2010, 2012), and Lu et al. (2014) showed positive effects. Fritze et al. (1997), Thorlin et al. (2006), Grafstrom et al. (2008), Watilliaux et al. (2011), Bouji et al. (2012), Court-Kowalski et al. (2015), Bouji et al. (2016), and Petitdant et al. (2016) did not show any positive effect. There is no clear explanation for these discrepancies. Contradictory data were shown with similar experimental designs: local (to the brain) acute exposure (Brillaud et al. 2007; Bouji et al. 2012) or intermittent (Ammari et al. 2008a, b; Petitdant et al. 2016), whole body continuous exposure (Grafstrom et al. 2008; Fragopoulou et al. 2012), high SAR (Brillaud et al. 2007; Bouji et al. 2012), environmental SAR (Watilliaux et al. 2011; Fragopoulou et al. 2012), mice (Maskey et al. 2010; Kim et al. 2008), or rats (Brillaud et al. 2007; Bouji et al. 2012). Similar data were obtained with Sprague Dawley and Wistar male rats (Bouji et al. 2012, 2016). The method of immunohistochemistry or ELISA may detect differently cytoskeletal protein redistribution and GFAP overexpression (Boran and Garcia 2007). Variability may also be caused by contention used to perform local brain exposures. However, there was no behavioral effect of restraint stress (Petitdant et al. 2016).

Here, total GFAP quantity was increased in the striatum of the 1.5 W/kg-exposed group, and cytosolic fraction of GFAP was increased in the hippocampus and the olfactory bulb in the 6 W/kg (15 min)-exposed group. Total GFAP levels did not increase in the direct ratio of the total absorbed energy. We may hypothesize that GFAP response was modified by activation of thermoregulatory processes or occurred with a different kinetic in the 6 W/kg-exposed group compared to the 1.5 W/kg-exposed group. In the 1.5 W/kg-exposed group, the decrease of the ratio cytosolic/total fraction of GFAP in the striatum is explained by the increase of the total GFAP, while cytosolic fraction remained unchanged. In the hippocampus and the olfactory bulb, the increased ratio cytosolic/total fraction of GFAP suggested either increased synthesis of soluble monomeric GFAP or reduction of its conversion to insoluble GFAP filament. There was no effect of exposure duration or proximity to the RF EMF source. Effect may have occurred in a SAR window in the more vulnerable cerebral structures.

The present study is the first to assess the effect of RF EMF exposure on MBP. There was no impact of RF EMF on MBP quantities in the six brain regions. In addition, there was no correlation between MBP and total or cytosolic GFAP levels. This data suggests that RF EMF-induced GFAP variation in the striatum, the hippocampus, and the olfactory bulb may not have been secondary to oligodendrocytes reactivity and/or myelin injury.

Aβ was largely studied because of its central role in Alzheimer disease neurobiochemical impairment (Silverberg et al. 2010). However, its physiological role state is not clear. In healthy young animals, Aβ peptide seems implicated in activity-dependent regulation of synaptic vesicle release, and in learning and memory (Abramov et al. 2009; Morley et al. 2010). Here, our exploratory linear model suggests that the level of Aβ 1–40 in the striatum explains long-term memory. This exploratory data corroborates a possible role of Aβ 1–40 in memory processes.

The present study is the first quantification of endogenous Aβ 1–40 following acute RF EMF exposure. Our data did not indicate any RF EMF effect on cerebral Aβ 1–40 3 days following an acute exposure. In addition to the absence of RF EMF biological effect on MBP and Aβ 1–40 in the present experimental design, further time points and neurotoxicity markers should be assessed to prove the absence of RF EMF-induced injury. In another study, neuronal damage and cell loss in the hippocampus (CA area) were shown after long-term and repeated RF EMF exposures at 900 MHz (Bas et al. 2009).

Meanwhile, an exploratory linear model indicated that the interaction between the levels of Aβ 1–40 in the striatum and RF EMF (1.5 W/kg) explained long-term memory. This exploratory data should be tested in a separated experiment. One can expect Aβ 1–40 changes to be more likely to occur following long-term exposures.

Up to date, there is no consensus regarding the possible effect of RF EMF on memory. Both positive (Preece et al. 1999; Koivisto et al. 2000; Krause et al. 2000; Maier et al. 2004; Keetley et al. 2006; Nittby et al. 2008; Arendash et al. 2010) and negative (Sienkiewicz et al. 2000; Dubreuil et al. 2002; Haarala et al. 2003; Haarala et al. 2007; Ammari et al. 2008a, b) reports were published. RF EMF exposures in freely moving rodents also gave contradictory behavioral effects. Nittby et al. (2008) indicated deleterious effect on episodic like memory with no effect on the open field. Lai (2004) showed alteration of spatial learning and memory. Cassel et al. (2004) and Cosquer et al. (2005) showed no alteration of spatial working memory in free-moving rats. Here, we brought new insights to this question by assessing the effect of RF EMF on retrograde memory using post-training RF EMF exposure in the fear conditioning paradigm. We studied only one type of memory, which was not shown from the literature to be basically affected by RF EMF comparing to other types of memory (spatial, recognition). The asset of this paradigm is to produce rapid, robust, and enduring (for months) learning to assess hippocampal and amygdala functions (Quinn et al. 2008). The 6 W/kg (15 min)-exposed group showed decreased long-term memory compared to short-term contextual memory. In a previous study, 6 W/kg (15 min)-exposed adult rats displayed lower and slower freezing response to context compared to middle age rats (Bouji et al. 2012). RF EMF may be perceived as an external cue (a distracter), which may disturb memory. Perception of RF EMF as a cue was also suggested in previous studies showing that EMF improved cognition and modified glutamatergic and electrophysiological responses (Lai et al. 1990; Eulitz et al. 1998; Lee et al. 2003; Krause et al. 2004). In addition, the correlation between striatal GFAP and long-term memory may suggest that impaired long-term emotional memory was secondary to astrogliosis in the striatum. Moreover, striatal GFAP and its interaction with RF EMF exposure were shown to explain long-term memory in a linear model. Conversely, memory decline was not correlated with the increase of cytosolic fraction of GFAP in the hippocampus or in the olfactory bulb.

Our experimental data indicates that striatal markers (GFAP and Aβ 1–40) as well as the interactions RF EMF (1.5 W/kg) × striatal GFAP and RF EMF (1.5 W/kg) × striatal Aβ 1–40 explain long-term environmental emotional memory. This data support the involvement of striatum in neuroanatomy of emotional memory and the particular sensitivity of the striatum to RF EMF exposure at high SARs. Previous studies showed that contextual fear conditioning was impaired after the selective ablation of striatal neurons and in striatum-specific dopaminergic receptor D1R knockout mice (Ikegami et al. 2014). Moreover, a previous study reported strong RF EMF effects in the striatum on the expression of N-mehyl d-aspartate receptors subunits at the synaptic plasma membrane and on GFAP levels (Mausset-Bonnefont et al. 2004). This structure may be of particular sensitivity due to RF EMF-induced cellular stress through the synthesis of heat shock proteins or molecular pathways of excitotoxic events.

Our data suggests no effect of RF EMF exposure on anxiety and locomotor activity. It confirms Mausset-Bonnefont’s study showing no impairment immediately and 24 h after a single RF EMF exposure. Overall, there was no correlation between GFAP levels in the six structures and locomotion or anxiety-related behavior.

Conclusion

The transposition of our results to human is not straightforward as the SARs (1.5 and 6 W/kg averaged on the whole brain) are higher than in the brain of mobile phone users (0.9 W/kg averaged on 10 g of tissues) and as controlled exposures with loop antennas in rodents differ from cell phone use in human. Our experimental design aimed at explaining biological effects of high SARs. Overall, our data corroborates previous studies indicating RF EMF-induced astrogliosis. This study suggests that RF EMF-induced astrogliosis had functional consequences on memory but did not demonstrate that it was secondary to neuronal damage.