If we are ever going to have brain-computer interfaces (BCIs) that do more than tap a soccer ball five feet, we are going to have to build a better mousetrap than the crude exoskeleton recently demoed at the World Cup. If the current consensus is to record cells directly from the motor cortex, then let’s try to figure out how to do that from the ground up. In other words, instead of just jabbing an electrode array into the cortex the same way we might push an eight-pin DIP chip into black conductive foam for safe storage, we need to construct brain and machine together. Researchers at Tufts have now taken the first major steps towards this goal by assembling a piece of cortex-like tissue from scratch in a dish.
This is obviously not the first time that building in vitro brain balls has ever been attempted. We recently described work by Austrian researchers who constructed so-called cerebral organoids from embryonic stem cells. These crude tenderloins were coaxed from an initial sheet of cells into forming a hollow mass with interior ventricles reminiscent of a mini brain. But in order to build something more like an actual cortex ? something with cortex-like cells that project out into a region where their efforts can be directed and efficiently extracted with recording devices ? a little more geometrical finesse is needed.
It is not too bold to say that the most obvious feature of the brain is its partition into grey matter containing mostly cells, and white matter containing their long range axons. The Tufts team was able to recreate this basic layout by combining two different kinds of substrates: a stiff porous scaffold made from silk onto which they seeded rat cortical neurons, and a soft collagen gel through which the axons can grow. The neuron-bearing scaffolds were formed into six colored concentric rings which were intended to mimic the six layers found naturally in cortex. The axons were then directed toward the central interior and would then presumably exit out either the top or bottom of the 3D assembly. We should caution against taking that six-layered standard too strictly though, because in some regions of the cortex assigning cells to layers is more like an astrological art.
The team performed various pharmacologic and electrophysiological manipulations to demonstrate their creation’s brain-like behavior. It also appears they actually dropped weights on this hunk of synthetic meat paste in order to simulate some kind of traumatic brain injury. While curious, it is tough to know some of these details considering we did not get any response from our request to corresponding author David Kaplan for a simple preprint.
A logical next step would be to begin to introduce recording structures like electrode arrays and optical fiber “optodes” into the ring regions so that long term stability and integration can be assessed experimentally. (So far the researchers have kept their tissue alive for over two months which is not too bad for a start.) However, as suggested before, the idea of recording neurons in the grey matter close to the cell bodies may be more a matter of convenience rather good long term sense.
The reason is this: while the cortex may be the first thing you see when approaching the brain from the outside (and so shoving in a pincushion recording array might seem like a no-brainer), these arrays not only fail to retain the ability to record for the long haul, they kill the cells in their wake. Now consider the white matter. When recording at the head of the cell the constant movements found in real brains ? osmotic swelling, growth and decay of synapses, vascular pulsations, and extracellular matrix changes ? tend to shift the cell out of range over time, or else impale it on its own its electrode. Recording the axons in the white matter might be considerably safer. In other words, if you are going to get shot, better to take the bullet in your arm then in your core.
The traditional problem with recording axons is that since they are protected by thick shields of myelin, you have to get lucky and find one of the many tiny nodes between the myelin segments to record them. But that simple fact ? that each neuron can be recorded at many sites on the axon rather than just one body location ? is the key that turns a seeming bug into a feature. The folks who need BCIs most, quadriplegics and paraplegics, can’t wait for their connectome drive to be handed to them. Recording at the cell body tells you nothing about where the signal is going. But if you have more than one node picked off, you not only have a coordinate, but you have a direction. A simple correlation over time will permit nodes on the same axon to be linked as one.
Axons branch in the white matter so the more nodes the better. To date there is no theory of axon branching that allows one to predict the stable forms axon trees eventually settle on in mature brains. Much can be gained in this regard just by looking at the gross structure of the cortex. A gyrus (or bump) in the cortex obviously has a very different white matter structure than that of a sulcus or depression. As far as these new bioengineered cortical models go, it may make sense to first try to extend the structure into a full-sized model gyrus. Then the ideal white matter device, tailored to a particular gyrus geometry might be explored. For example, those who have the canonical “Omega fold” like Einstein, or a “central fissure split” like Gauss, might be able to order up an ideal prosthetic model to fit their needs.
It goes without saying that this technology is in the early stage. But developing a realistic cortex model in a dish is essential to building the BCIs desperately needed by many.
Address: 5636 Lemon Ave.
Dallas TX 75209
Phone: +1 214 5203694