Moritz et al. implanted an electrode into a monkey's motor cortex. The electrode was designed to only record from a single neuron at a time. Then the output of that cell -- after a little amplification and transformation in a computer -- was connected to a muscle in the monkey's wrist. Finally, the nerves that innervated that muscle were temporarily anesthetized. The monkey was trained to play a little game that involved moving the wrist muscle to get a reward.
The researchers wanted to know whether the monkey could use the link up from a single neuron to coordinate the motions of their wrist. This is of particular interest to neural prostheticists who would like to invent ways to work around spinal injuries that prevent the brain from talking to the muscles.
Surprisingly, the researchers found that not only was one neuron enough, it didn't matter which one you picked. For nearly all the neurons, the animal figured out how to move their wrist using the machine.
The setup is illustrated below (Figure 1A from the paper).
The results of this paper are surprising and significant for several reasons.
1) Most researchers in this field would have assumed that you would have to select a neuron associated with the wrist in order for this setup to work. If you record from motor cortex neurons while an animal is moving, you find that each neuron is tuned to a particular set of motions in a particular set of muscles. We could associate activity in that neuron with a vector for the motion of the muscle.
The researchers in this case found that if a neuron had a vector associated with wrist motion, the animal could more quickly use that neuron to move its wrist. During a 10 min practice period, for instance, these previously associated wrist neurons were used effectively to get the reward.
But over time and training, the animal could use any neuron -- even those not associated with the wrist -- to move the wrist muscles. This indicates either a level of volitional control or a level of plasticity in the motor cortex that I don't think many people anticipated.
2) I was surprised that one neuron was enough to get accurate control over the wrist muscles. We typically think of motion being controlled by ensembles of neurons. The researchers in this case show that single neurons were enough. (They also found that they could link two neurons up -- one to the flexor muscle and one to the extensor -- to get even better control.)
Both the findings above are very interesting and significant, but I do have some questions before I think this technology could be applicable to those with spinal injuries.
Simple motions with single muscles groups are nothing like the complex motions involved in say walking. We know from experiments with cats that many of these complex motions are coordinated through intrinsic circuitry in the spine rather than in the brain. It is as if the brain sends a signal to the spine to initiate walking, and the spine takes care of the business of coordinating the many muscle groups involved.
The researchers do mention that this technique could be applied to spinal rather than muscle stimulation -- utilizing these motor programs. I think that is a better way to go about it. I doubt that you could get the sort of fine control necessary for navigation by direct brain to muscle linkage.
Also, the researchers show that they can harness neuron pairs to move synergistic pairs of muscles. I think it would be a fair assumption to say that while it is possible to manipulate muscles with single neurons, the more neurons that are present in ensemble would lend greater accuracy to that motion. I don't think I am quite ready to abandon the ensemble concept yet.
In any case, a very cool study. Here is the author's comments on the significance of their work:
These findings have several implications for future approaches to neuroprosthetic control. In contrast to the conventional strategy of deriving control signals from the combined activity of a neural population, it may prove efficacious to maintain separate signal pathways from cells to muscles. Using direct channels from single cells to specific muscles may provide the brain with more distinguishable outcomes of the cell activity and allow innate motor learning mechanisms to help optimize control of the new connections. The ability of the brain to adapt to new but consistent sensorimotor contingencies has been amply documented, and motor cortex can adapt rapidly to learn new motor skills. Motor circuitry can compensate for drastic changes in connectivity, such as surgically cross-connected nerves controlling wrist flexor and extensor muscles, or targeted reinnervation for control of prosthetic limbs.
Our finding that monkeys could learn to use virtually any motor cortex cell to control muscle stimulation -- regardless of the cell's original relation to wrist movement -- suggests another advantage of directly tapping single cell activity. Strategies based on decoding the activity of neural ensembles to obtain movement parameters or muscle activity depend on finding cells that modulate sufficiently with the output variables during actual or imagined movements. Instead, arbitrary cells available on recording arrays could be brought under volitional control using biofeedback, substantially expanding the source of control signals for brain-machine interfaces. This and previous biofeedback studies have shown that even cells with no discernable relation to muscles can be volitionally modulated after brief practice sessions. (Citations removed.)
Chet T. Moritz, Steve I. Perlmutter, Eberhard E. Fetz (2008). Direct control of paralysed muscles by cortical neurons Nature DOI: 10.1038/nature07418
Is there any speculation on what the search procedure is for this? To discover that some neuron over in the ear-wriggling region now also moves a box on the screen would seem to take a lot of experimentation.
Two comments, Murray.
1) We know from tracer studies that the motor homunculus specifying the motor cortex into ear regions, wrist regions, etc. is much less discrete than we used to think. Hence, recording in different parts of the motor cortex would still be likely to find a neuron that was activated in conjunction with wrist motion.
2) The whole idea of the study is that an arbitrarily selected neuron can be recruited to move the wrist via the apparatus.
One of the controls they used is measuring the neurons activity when the muscle was not anesthetized. This allowed them to determine whether that neuron was associated with wrist motion. They found that a minority of neurons were, and that these neurons could be more easily used to move the wrist using the machine.
However, they found that eventually the vast majority of neurons -- regardless of what they were used for normally -- could be used to move the wrist. Hence, their key finding that specification is mostly irrelevant means they don't have to go looking.