I am a physicist-become-neuroscientist. I have just recently (Sept. 1993) begun working in neuroscience, after receiving my PhD in condensed matter physics from Cornell. There, I worked with Rob Thorne on the electronic properties of new materials; specifically, the collective motion of large numbers of electrons in a Charge Density Wave.
I've left that all behind me. Now, I'm helping to develop the technology to chronically interface neurons with electronics, to help determine how biological neural networks operate. The way we're choosing to do this is to put immature neurons into small wells etched into a silicon substrate. Each well has an electrode attached to external electronics, to record the activity of the neuron and to provide a means of stimulation.
Lately, we've been experimenting with growth of dissociated neurons from the embryonic rat hippocampus. It turns out that they are extremely mobile and flexible. Click here to see what I know about culturing fetal rat hippocampal neurons.
I'm currently interested in exploring the following questions.
(1) How does the brain store information?
(2) How does the brain process information?
(3) Can we use living neurons as components of an integrated circuit?
(4) Can we directly connect a computer with a living neuronal network?
Actually, these are the questions that we hope to be able to tackle with the neurochip technology. I came at these questions from a more practical standpoint - - namely, how can I use my training in physical sciences to advantage in the biological sciences?
In order to really get at the dynamics of a functioning neuronal network, you need to know what all the different components are doing at the same time. To study plasticity, where interconnections between the neurons change based on the activity patterns, you need to be able to influence the cells without damaging them. To study development, where the various pieces of the network are changing, you also need to be able to measure the activities of the same neurons over time. Otherwise you're stuck making statistical analyses of the activities, which is far less helpful.
Conventional electrophysiological measurement techniques fall short in one or more of these key points. Once we get the neurochip working, we should have all these abilities at once.
The concept of the neurochip was developed through research in Jerry Pine's lab over the past ten years or so. The idea is to trap a neuron in close proximity to a non-invasive, extracellular electrode. In a close collaboration with Prof. Yu-Chong Tai in Electrical Engineering, this is accomplished by etching a small well into a piece of silicon. The top of the well is made to be confining enough to prevent the cell body from escaping, but not so constrictive that it prevents normal growth and development.
The picture below illustrates a neurochip well. There is a diamond-shaped hole in the top of the grillwork, so that we can insert a small, young, spherical neuron. The neuron will then fall to the bottom of the well, on top of the electrode (the roundish area near the center of the picture). This electrode is a thin film of gold attached to the external electronics.
Within a few hours of being placed in the well, a neuron will begin to grow long, thin extensions (called 'processes' or 'neurites') which adhere strongly to the surface. They will climb the walls of the well, and hunt around until they find the smaller, triangular holes in the grillwork. Then they will continue growing, seeking out their buddies and making functional connections called synapses.
Our first designs had a small (1 micron) overhang at the edge of the wells. This turned out to completely inhibit their growth. Below is a scanning electron microscope (SEM) photomicrograph of such a well. The scale bar kinda got trashed by the scanner; it's 10 microns long. This particular well has grillwork made of silicon nitride. You can see many of the important features of the neurochip in this picture. The bright pattern in the middle is the grillwork itself. It looks bright where it is suspended in space because it is only half a micron thick, and the electron beam can penetrate all the way through it. That means secondary electrons from the material come from both sides of the material, making it brighter than nitride which is backed by silicon (secondary electrons only come from the front side).
You can also see the inverted pyramid shape of the well through the corner holes. Notice that the electron-bright (i.e. thin nitride layer) extends past the corner holes for about 2 microns. This was our first hint that we had an overhang problem. When a neuron in the well starts to grow, it sends out objects called 'growth cones' to guide it. The neuron's process then trails it, like the slime trail left by a snail. This growth cone responds to many different cues, including chemical and touch cues. In a live animal, these cues tell the process where to find its targets. We now know that hippocampal neuron growth cones can't cross a micron-wide 90 degree angle, so the processes couldn't get out of the well and without any contacts to neighbors, the neurons would die.
This problem is summarized in the figure below. This is a cross-sectional view of a cell in a well. The grillwork on the left side has no overhang and imposes no barrier to the outgrowth of the cell. The grillwork on the right overhangs the well a little, and so the growth cone can't get past it.
We then tried a design which has no overhang. An example of the initial outgrowth of a rat SCG (a neuron from a ganglion of the peripheral nervous system) growing from such a well is shown below. To get an idea of the sizes involved, the black square is the edge of the inverted pyramid of this well, and it is 30 um wide. This picture was taken one day after loading a cell into the well, and extensive processes are visible. (For a time-lapse 'movie' of this process, click on the picture).
Some advice on viewing the movie: If you get an error saying that the helper application is not available, you'll need to set the preferences to get a movie player. On the Mac, I use Fast Player. The contrast on the neuron's processes is low, so for the best effect, play the movie fast - - about 3 frames per second (you may have to do this manually). If you have any problems, e-mail me so I can fix it up (maher@cco.caltech.edu).
The problem with this design is that the corner holes are too big. After they grow, the neurons' processes shorten a little, pulling on the cell body. Uneven tesion will pull the neuron towards the 'strongest' process. The neurons are 15-20 um in diameter, just smaller than the triangular holes in the grillwork at the corners of the wells. With just a little squeezing, the neurons pop out of the wells and escape the electrode! !
So, the holes were too big. We designed chips with a variety of hole sizes, from 10 x 3 microns down to 3 x 1 microns. And guess what: The cell bodies STILL escaped!! They take longer to get out of the smaller holes, but they still always got away. This is really amazing, because the cell bodies are about 20-25 microns in diameter.
Below is a picture (courtesy of Hannah Dvorak) of a well with 3 x 1 micron holes, with a hippocampal neuron inside the well. This picture was taken when the cell had grown for 3 days. Notice the long, healthy process growing from the top left side. The dark grey square with the black diamond in the middle is the well; it is 30 microns on a side.
Now look at the same well the next day. The oval shape 25 microns from the well is the cell body, which squeezed its way out of the hole. Also note the new process growing out of the lower right corner - - these cells can grow and develop very quickly.
So we had to come up with a way to keep cells in the wells. The cells were capable of squeezing through the smallest holes we cared to make, and many cells would kill themselves getting out through these tiny apertures. So, rather than using a small hole to contain the cell body, we came up with a design in which the processes have to grow through a long, narrow tunnel. This way, there would be so much pressure and friction on the cell body if it did squeeze partway into the tunnel that it couldn't get all the way through. And - - it worked ! ! ! !
The following image is a picture of fetal rat hippocampal neurons growing from our latest, fourth-generation neurochip at one week in culture. The long, squiggly lines are the neurites. Notice that several wells have multiple processes emerging, and that they are long and branching. This is indicative of healthy cells.
The black squares are the wells; they are 30 microns across. The grey-and-white pattern surrounding the wells are the tunnels. The white lines radiating out from the center are the tunnels, while the grey areas between them are the supports for the tunnels. This grillwork extends out over the wells, so that the cell body gets stuck in the well after the processes grow through the tunnels.
This particular chip has no electrodes. The grillwork design allows the neurons to grow, and contains them indefinitely. We are currently building full chips with this design, and with electrodes.
Keep an eye out for this page. Once we get fully functional chips, it shouldn't be long before I can show some real experiments and data.
Back to Pine Lab home page.
September 1995 Michael Maher (maher@cco.caltech.edu)/ADDRESS>