|Welcome to The Neuromorphic Engineer|
Applications » Neural Interface
The neurophotonic interface: stimulating neurons with light
PDF version | Permalink
At the end of the 18th century, Luigi Galvani demonstrated that nerves could be excited with electrical stimuli. Since then, scientists and engineers have been working on the development of neuroelectronic interfaces such as those popularised in the fictional works of Mary Shelley (Frankenstein) and William Gibson (Neuromancer). Despite advances in the miniaturization of electronics, materials science, and stimulation biophysics, neuroelectronic interfaces suffer from many fundamental drawbacks. These include the following: poor spatial resolution, since the extracellular microelectrodes typically used today simultaneously interface with all neurons within approximately 100μm; poor selectivity, as it is not possible to preferentially stimulate specific neurons; and non-flexible electrode-neuron contact. In addition, this kind of stimulation is invasive.
In 1971 Richard Fork showed that a high power laser can stimulate neurons by physically punching temporarily holes in their membranes. Equally, ultraviolet- (UV-) activated release of caged neurotransmitters to stimulate neurons was developed in the 1970s.1 However, the real excitement began in the end of 2003 when Peter Hegemann's group from the Max-Planck Institute for Biophysics discovered a light-activated ion channel in a swamp algae. This ion channel, called the ChannelRhodopsin-2 (ChR2), is the first light-activated ion channel that can transport the sodium and calcium ions necessary for neuron stimulation.2 Since then there has been a race to genetically engineer this ion channel in animal cells, resulting in around 40 papers in just the last 12 months.2 The field is now moving from genetics to biophysics and bioengineering.
The neurophotonics interface
Our group is mainly interested in using this ion channel as a novel type of neurointerface based on light instead of electricity. We use special custom-made light-emitting-diode (LED) matrix to stimulate multiple neurons in parallel. The micro-LED array we currently use is based on Gallium Nitride (GaN) technology. It emits 470nm blue light, which matches the absorption peak of ChR2.
In our first set of experiments we used the LEDs to stimulate action potentials in rat hippocampal neurons photosensitized with ChR2.3 We recorded the responses from single cells with a standard patch-clamping technique and, using the unique spatiotemporal resolution of the micro-LED array, succeeded in stimulating an arbitrary combination of neuron cells and a single cell with sub-cellular resolution. We believe this is the first such demonstration. Figure 1 shows a single blue micro-LED stripe illuminating the body (soma) of a ChR2 transfected neuron (the ChR2 is coupled to yellow fluorescence protein and therefore has a green fluorescenting appearance). The inset on the bottom-left shows the neuron response (white) to a train of four light pulses (10ms pulse duration) at a frequency of 10Hz. We showed that single action potentials could be accurately and reliably triggered up to a frequency of 40Hz.
The initial array that we developed consisted of 120 micro-stripes,3 but we are now moving to matrix arrays of 64×64 pixels (20μm radius; 50μm pitch) that can be flip-chip bonded to complimentary-metal-oxide-semiconductor (CMOS) controllers (see Figure 2). The CMOS control is particularly exciting as it allows us to implement independent oscillator control of the individual pixels rather than raster scanning.
There are many advantages in using light to interface with neurons. The technique has micron spatial resolution and millisecond temporal resolution. It has flexible 'electrode'-neuron connections, high specificity (by targeting the expression of ChR2) and offers a remote (non-invasive) control of neural activity. Moreover, another recent exciting discovery has been HaloRhodopsin, which can be used to optically inhibit action potentials. Thus, while we are still largely at the biophysics stage, the field will soon be transferring out to neuroscientists, bioengineers, and neuromorphic engineers. It will then become a very important tool for probing brain function, and for developing novel prostheses such as those for the retina.4 It may even be possible to have an optical link between biological and silicon components in a hybrid neurocomputer: this would present many interesting possiblities.
The objective of our own work is to develop a neurophotonic visual prosthesis. Electronic retinal implants have not been able to follow in the footsteps of cochlear implants, largely because of the high power consumption required for stimulation (10μW is acceptable if you need 16 electrodes for a cochlear implant but becomes problematic if you require 500-10,000 which is the minimum requirement to return the most rudimentary vision). Photonic stimulation would allow the system to be non-invasive, and keep all the power requirements external, thus massively increasing efficiency.
There are, however, several issues that need to be addressed before the neurophotonic interface can be fully functional. The kinetics of the triggering process and its side effects on the neuronal response must be better understood. Additionally, for our own purposes, we want to bring the stimulation requirement for the ChannelRhodopsin down so that we can reduce the power consumption of the final prosthesis. Nevertheless, the future looks bright!
Tell us what to cover!
If you'd like to write an article or know of someone else who is doing relevant and interesting stuff, let us know. E-mail the editor and suggest the subject for the article and, if you're suggesting someone else's work, tell us their name, affiliation, and e-mail.