DISCUSSION

Taken as a whole, our results show that the Antarctic amphipod G. antarctica uses a magnetic compass for orientation. If all the samples from the four different rotations of magnetic axis (Fig. 1A–D) are plotted with respect to magnetic north, a clear escape in the seaward direction appears (Fig. 1E). This escape bearing corresponded with the magnetic position of the beach from which the animals were collected (Fig. S1). Correct orientation in the sea–land axis (y-axis) has been reported as being crucial for small animals living in the littoral zone. Supratidal crustaceans dwelling in the seagrass wrack accumulated on beaches above the water line prefer to escape in the shoreward direction (Rothsey, 2006; terHorst, 2012; Ugolini and Pezzani, 1995), but this motivation may be reversed by environmental factors; for example, beach hoppers (Orchestia sp.) move towards the shore in wet environments and towards the water in dry environments (terHorst, 2012; Rothsey, 2006).

Our animals, which occupy open seawater, were expected to prefer an escape direction towards the sea, and this hypothesis of seaward movement motivation was confirmed. As to the biological meaning of seaward escape, we had assumed that animals trapped at low tide in shallow pools on rocky beaches without being able to find their way back to the open ocean were exposed to a high risk of predation, overheating or desiccation. Despite a rather limited swimming efficiency in water currents, movement toward the sea could be a universal escape reaction of the animal in danger. As to accessible orientation cues, amphipods occupying sandy beaches are known to orient themselves along the y-axis with the aid of visual markers, as well as a complex of environmental stimuli dependent on time, slope, magnetism (Arendse, 1978), humidity, etc. (reviewed in terHorst, 2012; Scapini, 2006). However, orientation according to visual or other clues seems to be rather difficult in Antarctic waters because of strong winds and rough seas. Magnetosensitivity could make orientation in this turbid environment much more reliable, and MFE is one of the very few stable guideposts for small aquatic animals.

The phenomenon of RF interfering with the magnetic compass orientation of animals is remarkable not only for its possible impact on organisms living and migrating near systems emitting RF, but also for its diagnostic evidence in the search for the biophysical basis of the magnetoreception mechanism (Henbest et al., 2004). Indeed, this was the primary goal of the first experiments with RF and magnetic orientation (Ritz et al., 2004). Despite 60 years of research, the principles of magnetoreception – the unique ability to use omnipresent MFE clues for orientation and/or navigation – are still not fully understood. There are two plausible mechanisms of terrestrial magnetoreception. In salmon and other vertebrates such as subterranean mole-rats (Thalau et al., 2006) and bats (Holland et al., 2008), magnetoreception is probably based on the presence of magnetic particles of iron oxides in their tissues (Eder et al., 2012, but see also Edelman et al., 2015). Other animals, including invertebrates, seem to possess a different system for magnetic field detection: according to radical-pair (RP) theory, MFE affects the dynamic of biochemical reactions of molecules forming transient RPs. Cryptochromes found in the retina of birds (reviewed in Liedvogel and Mouritsen, 2010) and Drosophila (Mazzotta et al., 2013) have been suggested to create RP with the cofactor Flavin and are the most widely accepted candidates for the molecules sensitive to MFE direction (e.g. reviewed in Dodson et al., 2013). A combination of both mechanisms is expected in birds (e.g. Wiltschko and Wiltschko, 2007). The principle of the disruptive effect of RF on magnetosensitivity was originally suggested as a selective diagnostic tool separating RF-sensitive RP photochemical processes from RF-insensitive iron particle mechanotransduction processes (Ritz et al., 2004). However, the current situation is complicated by the fact that the disrupting effect of extremely weak RF on the RP reaction to a static MFE some 3000 times stronger seems to be problematic even within the RP paradigm, and is not generally accepted by biophysicists (Hogben et al., 2009). Similarly, the thermal or other reactions of iron particles are not fully clear, especially in broadband RFs (Engels et al., 2014).

There are, however, specific resonance effects bound to the interaction between RF and unpaired electrons of an RP. If RF shows a peak of its effect at the LF, such a finding would be much more compatible with an RP mechanism than reception mediated by iron oxide particles (Ritz et al., 2004). In theory, spinning electron(s) of the RP interact with an external magnetic field in a frequency-dependent manner that corresponds to the LF (Ritz et al., 2009). Provided that one or both radicals are free of hyperfine magnetic interactions with nuclei and the earth's field vector is the only relevant local field, the LF falls into frequencies of radio waves close to 1 MHz (Ritz, 2011). If the RF of this frequency is applied in a non-parallel way to the earth's field, it causes resonance effects on spinning electrons interfering with their singlet–triplet interconversion, and possibly hindering sensitivity to the direction of the MFE (Ritz et al., 2009). The condition of no hyperfine interaction points to radical species involving oxygen as likely magnetoreceptors sensitive to LF, though some other key prerequisites are considered to be questionable (Hogben et al., 2009; Mouritsen and Hore, 2012; Lee et al., 2014).

What is the behavioural evidence for LF-mediated effects? While experiments on birds (Ritz et al., 2009) and insects (Vacha et al., 2009) have shown the dominant impact of LF in line with RP-specific resonance bias, no exceptional effects of LF on magnetoreception were reported by Engels et al. (2014) on the European robin under broadband RFs or by Malkemper et al. (2015) on wood mice. Such heterogeneity may still be understandable because of the limited number of studies and the variety of protocols from multidisciplinary fields combining demanding biological and physical approaches. Deciphering the problem of interactions of weak RF fields and living organisms is, therefore, an important task of frontier research on the boundary between biology and physics.

The results of our experiments suggest the remarkable sensitivity of the magnetic orientation of the Antarctic amphipod G. antarctica to RF. Spontaneous escape bearings became random not only after exposure to an RF of 20 nT, 969 kHz (LF) (Fig. 2A), but also under RF 2 nT both at 969 kHz (Fig. 2B) and at a frequency 10 times higher (10 MHz; Fig. 2C,D). Even 2 nT RF, regardless of frequency, was able to disrupt orientation (Fig. 2E). Orientation was lost whenever the set consisting of a signal generator connected to an antenna loop generated an RF. However, biasing factors other than RF potentially caused by feeding the set such as vibrations or ultrasounds are not likely. Measurement of ultrasounds produced by running the generator–antenna system (bat detector Pettersson D980) at a distance of 2 m from the generator gave no signal. To check the possibility that loss of orientation was due to vibrations spread from running the RF generator, we performed additional vibration measurements under conditions resembling as much as possible those at Johann Gregor Mendel station. Measurements by a specialised detector (Seismic Accelerometer, Wilcoxon Research, Model 731A, Wilcoxon, Power unit Amplifier Model P-31, Audacity 1.3 software) showed that differences between generator on/off states were very apparent when the probe was on the table in the direct proximity of the generator (Fig. S3A), as well as under the table on the chipboard floor (Fig. S3B). However, the record curves did not differ when the probe and generator were separated by a distance of 1 m or more (Fig. S3C–E), making the likelihood of vibrations from the RF generator reaching the 3 m distant testing device highly unlikely.

In addition, the animals showed no orientation towards or away from any cue in the laboratory with the exception of the magnetic field (Fig. 2F), which provides evidence of no directional bias. RFs remain, therefore, the most likely factor explaining the loss of magnetic orientation.

Our results confirm the extraordinary sensitivity of animal magnetoreception to electromagnetic fields from a radio wave spectrum of tens of nanoTesla flux densities previously reported in research into the animal magnetic compass (Landler et al., 2015; Engels et al., 2014; Ritz et al., 2009). Our work extends recent findings on vertebrate compass orientation for marine invertebrates living abundantly in polar seas. Should sensitivity to man-made RF be widespread across the animal kingdom, it may point to a new phenomenon with possible important ecological consequences. Animal directional magnetoreception of MFE seems to be the only known biological phenomenon responsive to electromagnetic fields of nanoTesla flux densities and a megahertz spectrum. A magnetic compass need not, however, be the only biological process receptive to extremely weak oscillating fields, as the report of nociception in mice by Prato et al. (2013) shows. Such findings affirm, inter alia, the importance of studying animal magnetoreception and perhaps other selected physiological events in environments isolated from man-made electromagnetic noise, whether in shielded laboratories or far from sources of anthropogenic noise.

There is currently no generally accepted biophysical model covering biologically relevant processes underlying such extreme sensitivity to RF, and the nature of the interference remains unresolved (Lee et al., 2014; Engels et al., 2014; Wiltschko et al., 2015). Our present work cannot contribute to the discussion on the validity of the LF resonance effect on RP magnetoreception because we did not reach a flux density threshold for any frequency. Magnetoreception was lost under all intensities and frequencies of RF used. Similarly, to discuss the possible impacts of the electrical and magnetic component of RF separately, we would have had to screen out or to produce only one or the other component. Unfortunately, such experiments could not be carried out within the limited time frame of the Antarctic season and the limited space of the polar base facility. We believe, however, that our report of extraordinary sensitivity to RF may inspire practical replications as well as new theoretical biophysical and biological insights.