The Neurochip


Motivation and Aims:  The goal of the neurochip research project is to design and fabricate a silicon-micromachined device that continuously records from and selectively stimulates individual neurons that are part of a small network of cultured neurons.   (The cultured neurons are embryonic rat hippocampal cells, because the hippocampus is known to play a central role in learning in memory.)

With this device we (and others) will be able to map connections between the individual neurons that are part of the network and study plasticity--changes in the network connectivity—over times of several weeks.  As the culture grows and develops on the neurochip, we will have, for the first time, an opportunity to study interactions of the networked neurons in detail.  Furthermore, the neurochip will allow us to investigate the effect that selective stimulation has upon the network.  Thus, we will be able to study “learning”—modifiable connectivity in response to stimulation—in living neural networks.

Structure and Fabrication:   The micro-machining wizards from Prof Y.C. Tai’s Lab—Dr. Ellis Meng, Qing He, and Angela Tooker--are in charge of fabrication of  the neurochips.

The chip consists of a 4x4 array of “neurowells”, or “neurocages”.  The neurocage consists of a top loading access hole, the cage body, and several (usually ½-1 dozen) thin channels that protrude from the bottom of the cage.  The cages are spaced 100 mm apart.  The diameter and height of a single well are 30 and 15 mm, respectively.  These dimensions accommodate loading a freshly dissociated cell (~15 mm in diameter), as well as matured cell (~20 mm in diameter, 10 mm thickeness).  In addition, the channels leading out of the main body of the cage—the space through which neurites may grow, while keeping the cell body trapped—are about 2-3 mm high.   A gold electrode will be pattered at the bottom of each neurowell for stimulation and recording of the neuron.

Making these structures is a multi-step process that involves patterning with multiple layers and removal of sacrificial layers.  This process is summarized in the figure below; for more detailed info on fabrication, see the publications section.

Above:  Generalized neurowell fabrication process (left); SEM of finished neurowell (right)

Challenges Ahead:  Two immediate challenges lie ahead--to evolve the neurowell structure so that supports confined growth and development of the neural network, and to design a method of rapidly loading individual neurons into wells.

It is well known that neurons are very mobile during early stages of development, as individual neurons grow and as groups of neurons begin to network.  (A movie showing this mobility is available here.).  This mobility enables a neuron to climb up and out of the neurowell or to squeeze through one of the channels (intended only for neurite growth), thus exiting the neurocage—sometimes at the cost of cell damage or death.  So the challenge is to design a neurocage that prevents the cell from escaping, keeping it caged in close proximity to the recording/stimulating electrode at the base of the well, while allowing for neurite growth through the channels which is necessary for normal neuronal growth and network connectivity.  Of course, we can not simply make the top loading access hole arbitrarily small because we have to load the neuron into the neurocage in the first place.

The task of how to load individual neurons into wells quickly--ideally 16 wells in (less than) 16 minutes--presents its own challenge.  The current method is completely manual and proceeds as follows:   A glass micro-pipette attached to the end of a manually adjusted “vacuum” is used to suck a neuron off the surface of a “non-stick” chip, capturing it in the micro-pipette.  (Finding a non-toxic substance to which neurons do not adhere is non-trivial!)  The micropipette is attached to a micromanipulator which is used to position the pipette over one of the neurowells.  The pressure in the “vacuum” is adjusted so as to slowly release the neuron out of the pipette tip and into the pH buffered medium in which the neurochip and neurons bath during loading.  The position and pressure of the micro-pipette are adjusted to skillfully guide the neuron into the neurowell as it sinks through the medium.  This process is repeated, loading as many wells as possible, for about 30 minutes.  Any longer, and the neurons become stressed or damaged.  This manual method proceeds slowly and is not fast enough to load all wells in the 4x4 array--at best, only about 10 neurowells can be loaded in 30 minutes.  So the challenge is to develop a system that can load the wells much more efficiently. 

The new, faster method will utilize a laser tweezers in combination with a computer-controlled mechanical stage.  The neurochip (eventually made on a glass substrate) will sit atop a mechanical stage that is mounted on top of an optical microscope stage.  The neuorchip will be loaded with about 20,000 in the periphery (away from the central 4x4 array).  With the aid of the microscope, the user will search the periphery and select a neuron to be loaded.  A laser tweezers then turns on, “grabbing” that neuron—the neuron is actually levitated in the laser beams.  The software controls the mechanical stage  positioning it so that the desired neurowell sits directly below the levitated neuron.  The laser tweezers will turn off, allowing the neuron to drop into the well.  Once fine tuned, this system will be fast enough to load 64 wells in 30 minutes.

In addition to solving the problems discussed above, a suite of software and electronics is being developed that will fully automate the simulation and recording of the neurochip.