Update (October 1): On September 30, Nature Neuroscience published the work by Zhu’s and Celikel’s groups, along with a third article also finding Magneto ineffective and a reply from Güler.


Several recent studies in high-profile journals reported to have genetically engineered neurons to become responsive to magnetic fields. In doing so, the authors could remotely control the activity of particular neurons in the brain, and even animal behavior—promising huge advances in neuroscientific research and speculation for applications even in medicine. “We envision a new age of magnetogenetics is coming,” one 2015 study read.

But now, two independent teams of scientists bring those results into question. In studies recently posted as preprints to bioRxiv, the researchers couldn’t replicate those earlier findings.

“Both studies . . . appear quite meticulously executed from a biological standpoint—multiple tests were performed across multiple biological testbeds,” writes Polina Anikeeva,...

The promise of magnetogenetics

Being able to use small-scale magnetic fields to control cells or entire organisms would have enormous potential for research and medical therapies. It would be a less invasive method than optogenetics, which requires the insertion of optical fibers to transmit light pulses to specific groups of neurons, and would provide a more rapid means of inducing neural activity than chemogenetics, which sparks biochemical reactions that can take several seconds to stimulate neurons.

In a 2016 study in Nature, geneticist Jeffrey Friedman from Rockefeller University and colleagues reported to have stimulated neural activity in glucose-sensing neurons in the mouse hypothalamus. Those neurons fired when the animals were exposed to a magnetic field, causing blood glucose concentrations to rise and insulin levels to fall. Ultimately, the mice ate more.

What I find most impressive about these reports . . . is just the level of care and effort that has gone into this.

—Markus Meister, Caltech

The researchers did so by genetically engineering a construct to be expressed specifically in those neurons. The introduced sequences coded for the iron-based blood cell protein ferritin coupled to the TRPV1 membrane channel, a temperature-sensitive protein that allows positively charged ions such as calcium to enter cells. Stimulation of ferritin’s iron through a magnetic field was thought to prompt TRPV1 to open, although the precise mechanism is unclear.

In a different 2016 study in Nature Neuroscienceneuroscientist Ali Güler of the University of Virginia and colleagues used a similar construct they named Magneto—this time coupling ferritin to the TRPV4 membrane protein, sensitive to mechanical forces as well as temperature changes. Expressing this in dopamine-receptor neurons in the mouse striatum caused the rodents to preferentially spend time in a magnetized area of their cage.

The year prior, researchers of Tsinghua University in Beijing had expressed the gene for a different iron protein–membrane channel construct, dubbed MAR, in specific sensory neurons of the nematode worm Caenorhabditis elegans. Applying a magnetic field resulted in changes to the worms’ movement, they reported in Science Bulletin. All three research groups presented multiple lines of evidence to back up their claims, such as electrophysiologic techniques to monitor the activity of individual neurons in brain slices and in vitro calcium imaging assays, in addition to the behavioral studies.

The studies received a mixed reception from the scientific community. Some, like Boston neuroscientist Steve Ramirez, were enthusiastic, calling the work “badass” on Twitter, while others were skeptical, critiquing the findings in journals and on blogs. That latter includes Markus Meister, a physicist-turned neuroscientist at Caltech, who says he’s aware of several research groups that had difficulties replicating some of the findings—spurring some to conduct lengthy, systematic investigations of the function of these constructs.

Magnetogenetics under the microscope

The new replication studies used a range of methods to investigate whether the constructs work as described in earlier research. In one study, neurophysiologist Tansu Celikel of Radboud University in the Netherlands and his colleagues focused their research on the Magneto construct used in Güler’s study.

Like Güler’s group, they used a virus to deliver DNA encoding Magneto to neurons in the mouse cortex and waited two weeks for the cells there to express the construct. Using permanently implanted microelectrodes, they recorded cortical neural activity as they exposed the animals to a magnetic field. The stimulus didn’t change the rate of action potentials in those neurons, they observed, and the same was true for in vitro experiments. “We argue that the utility of Magneto to control neural activity in vivo is not warranted,” the authors write in the preprint.

In the second study, neuroscientist Julius Zhu of the University of Virginia and his team conducted a systematic investigation of all three constructs that had been used in previous studies: Magneto, the TRPV1-ferritin complex developed by the Rockefeller group, and the MAR construct. (Güler, who is also at Virginia provided some materials for the experiments, but the two labs did not collaborate.)

We’re anxious to understand what the basis for the differences between his results and ours are.

—Jeffrey Friedman, Rockefeller University

Similar to Celikel’s findings, they observed that magnetic fields did not induce an electrical current in Magneto-expressing mouse hippocampal cells in culture, when the construct was delivered either with a plasmid or a virus. They did note, however, that both Magneto-expressing neurons as well as control cells that lacked the construct frequently displayed spontaneous changes in current that sometimes triggered the cells to fire an action potential. Based on this, they suggest that Güler’s reportedly magnetically triggered action potentials “are likely to represent mismatched spontaneous firings.”

The team appears to have had difficulties getting the construct expressed in cells at all. They used a plasmid encoding the Magneto construct to express it human kidney cells in culture, and made electrophysiological recordings of the cells. Neither a magnetic field nor the addition of a protein that stimulates TRPV4 elicited significant electrical currents in the cells. Interestingly, they did observe a current when they repeated these experiments with kidney cells that expressed the wildtype, unaltered version of the gene for TRPV4 expressed separately with ferritin’s gene. Together with other observations, this suggested that Magneto doesn’t form a functional ion channel or incorporate into the plasma membrane, the authors suggest. The construct lacks a portion of the TRPV4 protein considered necessary for its placement in cellular membranes, the researchers note.

In testing the other constructs, Zhu’s group used viruses to express MAR in neurons from cultured rat hippocampal slices, and the TRPV1-ferritin construct in hypothalamic neurons in intact mouse brains. Again, electrophysiologic recordings did not detect a change in action potentials in any of the genetically modified cells when they were exposed to a magnetic field, although the cells did exhibit frequent spontaneous action potentials. “Together, these results support the theoretical conclusion that Magneto, [MAR] and [the ferritin-TRPV1 construct] are incapable of controlling neuronal activity by producing magnetically-evoked action potentials,” they write in the preprint. The senior authors of both studies both declined to comment out of concern it would interfere with the publication of their research in a peer-reviewed journal.

“What I find most impressive about these reports . . . is just the level of care and effort that has gone into this,” remarks Meister. Neither he nor Anikeeva are surprised by the new findings; both have previously critiqued earlier studies. “By now, if it worked as advertised, you would expect a small industry of people doing this and using it for all kinds of purposes,” Meister says.

An unsolved controversy

Neither have a good alternative explanation for the observations reported in earlier studies. Meister suggests it may boil down to human error, while Anikeeva speculates that tethering ferritin, a relatively bulky protein, to TRPV proteins might possibly make the channels leaky and lower the threshold for action potential firing.

Güler, who developed the Magneto construct, points out several differences between his study and the two preprints that may account for the contradictory results. The groups used different viruses to introduce the constructs to cells, and for the most part, didn’t allow as much time for them to be expressed in neurons as Güler’s group did, which may be why they didn’t achieve full presentation in the cellular membranes. For some experiments, they also didn’t verify that the viruses were actually expressing the constructs before they introduced them into cells, he adds. “Some batches will not work, and you have to systematically make sure that your tools are up to par,” he tells The Scientist.

“We acknowledge that the system we have developed is a little finnicky” in that it requires a lot of optimization to get it to work, he adds. “I think that is where the setback is: everybody wants to have something that works immediately.” Magnetogenetics techniques will take some time to refine until they are reliable, he says.

Friedman, the senior author of the Nature study, is similarly puzzled why Zhu’s team couldn’t replicate his findings. “We take the Zhu paper seriously and . . .  we’re anxious to understand what the basis for the differences between his results and ours are,” he says. Zhu’s team expressed the construct indiscriminately into all neurons in the hypothalamus rather than selectively in a subset of cells, as Friedman did. Some hypothalamic neurons are more easily excitable than others, he explains. “It’s possible that by restricting the cells we were recording from, we may have gotten a cell type that . . . seems to be more rather than less responsive.”

Friedman stresses that his team did multiple experiments as part of their study to ensure that they weren’t mistakenly attributing spontaneous neural activity to a magnetic effect. For instance, in the same Nature study they repeated their experiments with an altered version of the TRPV1 channel that acts as a chloride channel rather than a calcium channel. Whereas calcium influx would excite a neuron, chloride flux would inhibit it, Friedman explains. “We get opposite effects when we use the inhibitory version of the construct instead of the activating one,” he says. “We wouldn’t see that if it was spontaneous activity.”

Both Güler and Friedman note there are three additional studies that report having successfully used similar genetic techniques to excite neurons under magnetic fields. In 2017, a team of researchers engineered a construct made of the genes for ferritin and the heat-sensitive channels—either TRPV1 or TRPV4—into neural crest cells of chick embryos, claiming to have stimulated the neurons with electromagnetic fields. In 2018, another group combined the TRPV1-ferritin construct with a protein involved in cell migration, and showed that human kidney cells expressing the introduced genes had an unusual migration pattern when under a magnetic field. And earlier this year, a third set of researchers replicated Güler’s findings by expressing a TRPV4-ferritin construct in a human kidney cell line to better understand its function, also observing a response to magnetic stimulation.

© istock.com, marcin klapzcynski

Mysterious mechanism

It’s not quite clear how these constructs might work. One possibility is that magnetic fields cause the iron atoms in the ferritin to flip periodically, generating heat that causes the temperature-sensitive TRPV1 channel to open. Another option is that the stimulated ferritin would tug open the central pore of the membrane channels. The group that was able to replicate Güler’s results in kidney cells suggested that the magnetic sensitivity of the TRPV4 channel has more to do with thermal energy than with mechanical force.

Meister has argued that these proposed mechanisms “conflict with basic laws of physics,” on the grounds that ferritin doesn’t have the characteristics necessary to prompt a mechanical stimulus under a magnetic field. In several back-of-the-envelope calculations outlined in his 2016 eLife paper, Meister shows that magnetic interactions between ferritin and a magnetic field would be “between five and ten orders of magnitude” too weak to generate the mechanical force to cause a membrane channel to open.

The core of ferritin consists not of a truly magnetic substance, but ferrihydrite, which is only weakly paramagnetic at room temperature. This means that the molecule requires a more powerful magnetic field to induce a magnetic moment—that is, to align all iron atoms with the magnetic field—than those used in previous studies. Even if the iron ferritin was truly magnetic, the forces would still be too small to account for the proposed mechanisms, notes Anikeeva, who made similar arguments in a separate eLife paper.

Those biophysical arguments could be overcome if physicist Mladen Barbic of the Howard Hughes Medical Institute’s Janelia research campus is right. Earlier this year in eLife he proposed several new alternative mechanisms whereby magnetic stimulation of ferritin could open an ion channel. One, for instance, is based on the Einstein-de-Haas effect, by which iron oxide particles would rotate under a magnetic field, producing energy which could perhaps cause the ion channel to open. Other groups are exploring the possibility of a chemical mechanism through the release of free iron, Friedman says. “I think all these are on the table,” he says.

It’s not the end of magnetogenetics

The lure of a non-invasive method to control neural activity has kept scholars in pursuit of a reliable method of magnetogenetics, including those that aren’t based on ferritin. For instance, Anikeeva’s group has shown that it’s possible to open TRPV1 and stimulate neuronal activity with synthetic nanoparticles made of the iron oxide magnetite. The particles are known to dissipate heat, and that opens the channels, she explains. However, these particles can’t be genetically expressed because they are synthetic. Rather, they have to be injected into the brain.

Another route is to look at organisms in nature that have already evolved systems that respond to magnetic fields. Magnetotactic bacteria, for instance, produce particles similar to the ones Annikeeva synthesized in her lab, she writes. Scientists could also examine the mechanisms that migratory organisms such as pigeons, butterflies, and fish use to sense magnetic fields to navigate, she suggests.

What may help speed these efforts along, and help untangle the controversies around magnetogenetics, is better communication between physics and neuroscience, Anikeeva notes. “There should be more interaction between physical and biological sciences, especially in the context of training of both biologists and engineers in each other’s disciplines and vocabularies.”

Katarina Zimmer is a New York–based freelance journalist. Find her on Twitter @katarinazimmer. 

Interested in reading more?

magnetogenetics neuroscience techniques

The Scientist ARCHIVES

Become a Member of

Receive full access to more than 35 years of archives, as well as TS Digest, digital editions of The Scientist, feature stories, and much more!
Already a member?