One of the major differences between our methods and theirs was in the duration of the assays. In Vidal-Gadea et al., 2015 we experimentally determined and reported that 30 min away from food was sufficient to flip the magnetotaxis behavior of the worms from positive to negative. This was initially unexpected, because C. elegans does not flip its orientation preference after 30 min away from food for other orientation behaviors including chemotaxis to benzaldehyde. We therefore described this time as sufficient to induce the ‘starved’ state in animals, and went on to perform several experiments with worms in the ‘fed’ or ‘starved’ states. For ‘starved’ assays, we ensured that worms were away from food for 30 min prior to starting an experiment. From reading Landler et al., 2018 it is now clear that we did not explicitly mention that this definition of starvation implied that for worms to be tested in the ‘fed’ state, animals would need to complete their assay within 30 min.

With practice, we found that we could run our behavioral assays to near completion within this 30 min time window and did not need to immediately tally immobilized animals. This allowed us to run many assays in parallel without having to stop to conduct the time-consuming tallying step for each pipette or plate before starting the next. Tallying animals in the magnetotaxis assay, however, was a much simpler (and faster) procedure, which we could do at the 30 min mark. Therefore, it is important to note that we ran assays so that worms migrated to a particular direction within 30 min of initial removal from seeded growth plate.

It is clear from Landler et al., 2018 that they decided to modify our magnet assay to last 60 min rather than 30 min precisely because they continued to see moving animals all the way until this time point (see their Methods). Unfortunately, this also implies that many worms participating in the assay (which they described was a sufficiently large number to make them deviate from our protocol) would have transitioned to the ‘starved’ state. By our described definition of ‘fed’ and ‘starved’ (also adopted by Landler et al.) they report testing animals under both ‘fed’ (first half of the assay), and ‘starved’ (second half of the assay) conditions. This issue may have been obviated when Landler et al. tested pre-starved worms (their Figure 3B); however, no-magnet controls and horizontal-oriented tube controls for these assays were not reported (see below). We believe this singular, and crucial, difference might explain their different results.

A second major difference between our methods involved the rearing conditions for our animals. Like many C. elegans labs, our worms were grown in laboratories maintained at 20°C. We deliberately kept animals away from artificial magnetic fields (generated by electrical equipment or wiring). Landler et al. grew their animals in incubators. This attempt at controlled culturing could have accidentally grown their animals under extreme magnetic and electric fields conditions generated by their incubators. This is not trivial. Recent studies demonstrated that extreme magnetic inhomogeneities are produced within these type of devices (Makinistian and Belyaev, 2018). For example, animals cultured mere centimeters apart would be exposed to fields differing by up to a factor of 36. This includes hypomagnetic field areas, where the absence of magnetic fields might affect the development of magnetic organs. Therefore, it appears that while rearing animals at 20°C Landler et al. might have unintentionally cultured animals under wildly variable magnetic conditions. This method might also produce preferences for a cultivation temperature in animals that could confound behavior when they are tested in a chamber with a different ambient temperature. We are not certain of what effect these conditions might produce in the magnetic machinery or behavior of worms, however we think it prudent to control magnetic exposure of animals to be used in magnetic studies.

Landler et al., 2018 assayed whether N2 worms injected into agar-filled cylinders burrowed up or down. They report no bias for burrowing up or down, with or without an imposed inverted magnetic field.

The magnetic coil systems used in the two studies differ. In our 2015 study, we used three sets of four-coil systems described by Merritt (Merritt et al., 1983) which was reported to produce more stable test fields for larger samples (e.g. 50 vs 20 cm³) than with the Helmholtz loops (Kirschvink, 1992; Magdaleno-Adame et al., 2010) used by Landler et al., 2018. Before we conducted our burrowing assays inside our magnetic coil system, we measured the field properties throughout the test volume to ensure that our system produced a homogeneous field. Landler et al. (2017) did not report such measurements of homogeneity.

Landler et al., 2018 suggested that unintended temperature gradients generated by our coil system may have resulted in our reported observations. However, aware of this possibility, we used high sensitivity thermometers (0.01°C) to quantify temperature changes (in both magnet, and magnetic coil system assays). A two-way ANOVA (N = 5, p=0.123) showed no significant difference in temperature between cancelled field and one-earth field (Figure 2—figure supplement 1 in Vidal-Gadea et al., 2015). As an extra precaution, we regularly rotated the orientation of our magnetic coil system to a random position before each assay, and used a small fan to circulate air through the coil to prevent temperature gradients. While clearly understanding the risk of unintended thermal gradients, Landler et al., 2018 do not describe doing any such controls and they do not present associated temperature measurements. However, from their discussion they seem to believe that the use of double wrapped coils prevents the generation of meaningful temperature gradients. This is incorrect; all current passing through a metal conductor generates heat. Therefore, under both their test and control conditions, heat would be generated by current passing through their system, potentially building up within the small enclosed metal room used to shield worms from foreign magnetic fields. We welcome their additional control of a mu-metal shielded room as an improvement to minimize potential magnetic contamination, but caution that environmental parameters should be controlled further by positioning the coil system at random orientations within the room, and by empirically measuring and mitigating the thermal gradients that must necessarily develop as current powers any coil system as we did in our original study.

Our burrowing experiments tested in the natural earth field showed that worms migrated differentially based on their global site of origin and satiation state. This observation undermines the likelihood of temperature gradients, or magnetite contamination, being responsible for our results. Furthermore, our findings that worms lacking the transduction channel encoded by the tax-4 gene, or by worms with genetically ablated AFD neurons, failed to burrow preferentially up or down in earth’s natural field strongly point to the involvement of these neurons and molecules in this behavior (Vidal-Gadea et al., 2015). None of these results are mentioned in Landler et al., 2018.

Landler et al., 2018 tested how worms migrate to the edge of a 10 cm diameter plate in a horizontal magnetic field where they were trapped by azide at the edge. Unlike our study, they found no significant degree of orientation in their migration.

In Landler et al., 2018, they point out that we treated each worm as an individual in our horizontal field assays when performing statistical analysis, which could cause a type one error if worms did not act independently. While we believe that the individual timing and trajectory of each worm makes them independent, which justified our choice of statistical analysis, we nevertheless re-analyzed our data averaging the mean heading of worms in each assay as they did. Similar to our previous report, we found that in all but one out of 28 assays, worms displayed a significant migratory preference (Figure 1). The average heading of the assays changed somewhat from the result reported in Vidal-Gadea et al., 2015, but not significantly so.

Figure 1

Download asset Open asset

Landler et al., 2018 also noted that all worms in their magnetic field conditions are set by an experimenter not involved in the analysis. This blinding protocol was also the case for our original study, but not explicitly mentioned. Anecdotally, we expected British worms to migrate towards magnetic north, just like magnetotactic bacteria, but remained puzzled for months when our results unexpectedly showed them consistently migrating at a 132° angle with respect to magnetic north. This illustrates how our expectations did not affect our analysis or results. Indeed, we still do not know why worms prefer this particular angle, although we presented a parsimonious explanation that worms may choose a migratory direction based on the inclination of their native field.

Landler et al. also performed a modified version of our magnetotaxis assay, reporting that worms exhibited no preference for the magnet when compared with a Wilcoxon signed rank test to a control group (although the control group is not described or plotted in their manuscript). However, Landler et al. chose to extend their assay time. We feel it is unacceptable that they fail to display their control data due to implications described below.

In their manuscript, Landler et al., 2018 offered no actual replication for our magnetotaxis experiments, opting instead for modified protocols. For example, Landler et al. chose to extend their assay time to one hour to include worms that may have switched to a starved state. As described above, it is therefore likely that by testing fed and starved animals alongside one another, they effectively combined and measured positive and negative magnetotaxis.

Landler et al. conducted an additional control in the form of a magnetic assay with worms fed 1% magnetite throughout their cultivation. They reported a barely significant (p<0.04) improvement over chance performance data that they chose not to plot; however, their assay was still not significantly different from animals not fed iron that migrated in the presence of a magnet. Their observation that worms contaminated with magnetite fail to migrate to a magnet as readily as in our assays, coupled with the lack of significant difference between their ‘contaminated’ vs ‘uncontaminated’ samples provides strong evidence that contamination with magnetite is likely not responsible for our observations. However, we take this conclusion with caution since their one-hour assay might have masked potentially significant effects arising from their magnetite enrichment.

Magnetite is pervasive in soils adjacent to the decaying fruit and plant matter where worms live (Stanjek et al., 1994; Moskowitz, 1995; Schulenburg and Félix, 2017; Schulenburg and Félix, 2017). It is likely that worms in their natural habitat may have access to much greater amounts of iron than those provided for them in the lab via a diet of E. coli on an agar surface. If C. elegans does use magnetite to build its own magnetic field detector, it is possible that enrichment of the lab culture conditions may result in increased magnetic indices for adults, or even succeed in enabling larval stage worms to perform this behavior better than what we previously reported (Bainbridge et al., 2016). While we consider this experiment informative, we do not agree with Landler et al., 2018 in their suggestion that ingestion of magnetic particles might confer the ability to migrate within magnetic fields to C. elegans. This would not explain how worms migrate at different angles in a satiety-dependent manner, or that this behavior can be systematically abolished and rescued by cellular and molecular tinkering. To our knowledge, ingestion of magnetite or iron is yet to be demonstrated in any animal to be sufficient to confer the ability to migrate towards and/or away from magnets using their own power. Thus, although exogenous iron has been suggested to contaminate cells proposed to be magnetoreceptors (e.g. Edelman et al., 2015), in contrast to Landler et al., 2018 claim, there is no example of false positives associated with actual magnetoreceptive behavior or physiology displayed by any wild-type animal. Additionally, in our initial study, we tested more than 30 mutants with impairments in a broad range known sensory modalities that C. elegans has been shown to detect (including temperature, electric fields, soluble and volatile chemicals, touch, light, osmolality, acidity, water, CO₂, and O₂). Only mutants related to AFD function were unable to orient to magnetic fields (Supplementary file 1).

One important control provided in our manuscript but left out in Landler et al., 2018 was the response of worms in the absence of a magnet. In their Figure 3F the authors display results for two assays in which worms are in the presence of a magnet, with the only difference being whether or not their diet was enriched with magnetite. No control assays where worms tested in the absence of a magnet are provided. Additionally, worms tested with a magnet are confusingly labelled ‘baseline’. We find the omission of this no-magnet control, and the nomenclature used, concerning. Based on our experience, fed worms display magnetotaxis indices with high positive values, and starved worms display indices with high negative values. If Landler et al. inadvertently combined in effect starved and fed worms in their assays, then we expect them to observe a broad range of magnetotaxis indices with an average centered at zero. These results would starkly differ from control assays with no magnet, where indices would more tightly cluster around zero (Vidal-Gadea et al., 2015). If their no-magnet control data are significantly more narrowly distributed around zero than their test data, this would suggest that the broad distribution of indices obtained in their magnet assays is likely the result of combining fed and starved animals that are in fact orienting to the magnetic field, but in opposite directions thus resulting in an average calculated index of zero.

To investigate the possibility that Landler et al. had obtained positive results that were masked by testing worms in both fed and starved states, we reanalyzed their data. First, we plotted their original results from Figure 3F in bar format alongside results from our original study including our no-magnet control (Figure 2A). Next, to control for satiation dependent reversals in magnetotaxis preferences, we plotted the absolute value of magnetotaxis indices (Figure 2B). With this alternative analysis, we found that worms migrating in the presence of a magnet displayed a significant migratory bias when compared to worms migrating in the absence of a magnet (Figure 2B). Landler et al., 2018 reported that their magnetite-enriched worms did not migrate significantly better than worms that were not fed magnetite (p=0.652). Our reanalysis confirms their finding, but also shows that their worms in both test conditions displayed a significant difference from our no-magnet control. We also reanalyzed burrowing data plotted in Figure 3B and Figure 3—figure supplement 1 from Landler et al., 2018. Once again, we provided missing controls in the form of worms burrowing in horizontally oriented pipettes and compared these to data from both groups (Figure 2C&D). We found an identical pattern of results where the absolute value of the burrowing index values for all six test group worms were higher than our horizontal control worms with three of the four groups showing significant differences (Figure 2D, Mann-Whitney Rank Sum Test). Importantly, our reanalysis suggests that had no-magnet, or horizontal controls been included and analyzed by Landler et al., they would have obtained positive results for magnetic orientation in C. elegans.

Figure 2

Download asset Open asset

Figure 3

Download asset Open asset

Since our initial description of this behavior in C. elegans (Vidal-Gadea and Pierce-Shimomura, 2012a), we are aware of several groups joining the study of magnetic field detection using nematodes. While conducting our original study, Ilan et al. (2013) reported that parasitic nematodes migrated preferentially south over north when placed in a magnetic field. We recently became aware of a group at the University of Quilmes, Argentina who independently reproduced our findings with minor modifications (Vidal-Gadea et al., 2018).

Landler et al., 2018 notes that a study by Njus et al., 2015 reported that worms failed to respond to magnetic fields. Njus et al. restricted their study to crawling velocity and omega bends and not orientation. Nevertheless, in Figure 6 of Njus et al., 2015, they show a 10 ± 7% to 90 ± 20% change omega bends when a 5-mT magnetic field was introduced or removed respectively. Such difference in turns would have a significant effect on course trajectory and orientation. Rather than comparing these paired measurements to each other, Njus et al. compared them to the number of worms turning in the absence of a magnetic stimuli for which they report an average of 70 ± 50%. A 50% variability in omega bends is surprising and not consistent with previous reports in the literature (e.g. Vidal-Gadea et al., 2012), or even with the variability they report for the rest of their data (22.5 ± 8%, obtained by measuring and averaging standard deviations from test conditions reported by Njus et al., 2015: Figure 6). Not surprisingly, no test condition was significantly different from such a variable control. Therefore, the Njus et al., 2015 study appears to offer little evidence to counter the idea that C. elegans orients to magnetic fields.

How do worms move in our magnetotaxis assay? As described above, worms are placed in the center of an agar-filled Petri plate with a 0.29 T strength, 1.5-inch diameter, neodymium magnet placed north-side facing up 1 cm beneath the agar surface on one side of the plate (Figure 3A–C). Azide is pipetted on magnet and control sides to immobilize worms that reach either location. Landler et al., (2017) correctly point out that the intensity of the magnetic field generated by the magnet is many times stronger than the earth’s field. They ask how could worms orient to this magnetic gradient if they never encountered magnetic fields this size during their course of evolution.

We agree with Landler that the worms are unlikely to distinguish the difference between magnetic field intensities higher than earth strength. This is consistent with our finding that the AFD magnetosensory neurons failed to generate larger responses when presented with larger than earth-strength fields (Vidal-Gadea et al., 2015). Instead, we believe that C. elegans pays attention to the directional information contained in the magnetic field gradient. To help visualize the magnetic field in our magnetotaxis assay, we scattered iron filings across the agar surface of the assay plate (Figure 3A). The filings stand straight up directly above the center of the magnet, indicating that the magnetic field is perpendicular to the agar surface here. Filings tilt at increasingly shallow angles as the distance from the magnet center increases until they become parallel with the plate surface. This simple experiment shows that magnetic field lines pierce the surface of the agar at a variety of angles in a radial pattern which could provide abundant directional information pointing away from the center of the magnet in our, and Landler et al.’s, experiments.

We next calculated the direction and magnitude of the magnetic field across the assay plate. This was done using the online calculator by K and J Magnetics, and validated experimentally with the aid of a magnetometer. We found that the magnetic field was strongest at the center of the magnet and began to strongly dissipate near the inner edge of magnet (Figure 3D). The horizontal component of the field that is parallel to the plate surface was zero at the magnet center and increased away from this point up until near the inner edge of the magnet (red line, Figure 3D). The sign of the horizontal component switched at the center of the magnet reflecting how the field lines point radially away from the magnet center. The angle of field penetration varied across the assay plate as expected from the iron filings (Figure 3A,B&E). The field pierces out of the agar surface at 90° only at the center of the magnet, and starts tilting until it reaches 180° about 21.5 mm from the center of the magnet (Figure 3B&E). Beyond this point, the magnetic field starts to tilt further, piercing into the surface of the agar. Therefore, the magnetic field in the magnetotaxis assay varies in polarity and inclination which we hypothesize worms may use as a cue to orient (Figure 3B&C).

Armed with an empirically validated model of the magnetic field in our plates, we can predict how worms move in the magnetotaxis assay. We previously found that in a uniform horizontal field, N2 worms preferred to migrate approximately 132° away from magnetic north when well fed (Vidal-Gadea et al., 2015). Given this preference, we expect that worms would make a leftward arc when viewing the assay plate from above (Figure 3C). This is because worms started at the center of the assay plate would consistently bear 132° away from the horizontal field lines (purple arrows, Figure 3C). To test this prediction, we retrieved photographs of assay plates from our original 2015 study. We found that a significant portion of well-fed worms migrate towards the magnet along the left side of the plates (Figure 3F). This unusually asymmetric arced trajectory contrasts greatly from the typical symmetric trajectory that worms make when migrating to the peak of an attractant chemical gradient during chemotaxis (e.g. Pierce-Shimomura et al., 1999).

Taken together, this reanalysis of data from our 2015 study unifies the migratory patterns observed in all three behavioral assays and yields new predictions strengthening our original findings.

• Andres Vidal-Gadea

• Jonathan Pierce-Shimomura

• Jonathan Pierce-Shimomura