We have already begun to do this, implanting electrodes that measure the spikes of electrical currents created as neurons pass information along their networks. Yet even our best implants can interrogate no more than 200 nearby neurons. To understand the language of the brain, we will need to monitor thousands and then tens and even hundreds of thousands of neurons networked across the brain. Nanotechnology promises to make this – and more – possible.
The Kavli Foundation brought together four experts to discuss how nanotechnology is changing the way we measure brain activity. Our participants are:
- Paul Alivisatos – Director, Lawrence Berkeley National Laboratory, and Fellow of the National Academy of Sciences, American Physical Society, and American Association for the Advancement of Science.
- Anne Andrews – Richard Metzner Chair in Clinical Neuropharmacology, Professor of Psychiatry and Chemistry & Biochemistry, and member of the California NanoSystems Institute at University of California, Los Angeles. She is a Fellow of the American and International Colleges of Neuropsychopharmacology.
- Arto Nurmikko – L. Herbert Ballou University Professor, Engineering and Physics at Brown University and a Fellow of the American Physical Society, Institute of Electrical and Electronics Engineers, and Optical Society of America.
- Hongkun Park – Professor, Chemistry, Chemical Biology, and Physics, associate member of the Broad Institute, and affiliate at the Harvard Stem Cell Institute and Harvard Center for Brain Science. He is also a Fellow of the American Association for the Advancement of Science.
The following is an edited and revised transcript.
THE KAVLI FOUNDATION: Today, we can measure one or two hundred nearby neurons. How can we measure networks of thousands and tens of thousands of neurons? Will nanotechnology help?
Nanoscale tools could have finer probes, so they could measure many more neurons in the same amount of space. They could have built-in wireless telemetry, so they could communicate information on scales heretofore impossible. They could also be active, and not only sense brain activity but stimulate neurons, so we could control each input and watch how the brain responds.
HONGKUN PARK: Nanotechnology, with its scale and feature sizes, certainly can help to interrogate networks of hundreds of thousands of neurons. But we need to consider the context: Are we talking about preparations of neural tissue on a slide, animal models, or human patients?
I think devices with nanoscale electronics and sensors capable of taking several different types of measurements from tissue samples will come relatively soon. Mouse models with built-in telemetry will take some time, and systems for humans will take longer.
PAUL ALIVISATOS: Right now, several technologies could greatly expand the number of neurons we can measure within a short period of time. These include new optical approaches, less invasive electrical approaches enabled by nanomaterials, or hybrids of each of these. We're still in an early period here, and we should try to move forward with a view of focusing down on one or two of the most promising technologies.
ANNE ANDREWS: I expect nanotechnology will play an important role in making better devices for us to implant into tissue slice preparations or brains. I also believe this idea of multiple approaches is really important, because we often learn different things from different measurement techniques. We want to take advantage of the richness we'll get from using different optical, electrical, and other technologies.
TKF: Today's neural implants are large, invasive, and break down in the body. They are as hard as knives in a brain as soft as Jell-O, so subjects cannot move without causing damage. Will nanotechnology let us create probes that are less invasive and last for months or years?
ARTO NURMIKKO: There are issues of cortical damage, as well as immune system response when we embed anything in the brain. But if we want to access larger neural populations so we can understand the language of the brain, we will need more listening posts. Still, we can develop softer materials or adaptive materials that become less rigid and cause less damage in the brain. We can also make them compatible with the immune system so they last longer.
ANNE ANDREWS: At the nanoscale, even hard materials can be quite flexible. People have been making recordings of dopamine chemistry in the brain for months using carbon fiber microelectrodes, and with really good reproducibility. This may be less of an issue than we think, especially when not trying to measure one particular neuron, but rather large amounts of data from many neurons.
PAUL ALIVISATOS: As we record more and more neurons, some tools might have issues, but I think this is going to be manageable. We can, however, extrapolate from our experiences so far, and they have been surprisingly positive. We should be encouraged by how much it has been possible to do electrically and optically. We're just going to have to go in and find out.
HONGKUN PARK: I'd like to point out that the question appeared to refer only to electrical implants in brain. As technology advances, for example, if we could develop optically based implants, they would come with different types of challenges. The questions will change, depending on the technology we develop.
TKF: Spikes, the electrical charges that convey information along neurons, are easy to measure and correlate with physical actions. What other types of brain activity should we measure?
ANNE ANDREWS: I'm a huge advocate of measuring neurotransmitters, the chemicals neurons used to transmit information among themselves.
If all the information encoded in neuronal signaling was contained in spikes, then the brain would only produce only two neurotransmitters: glutamate and gamma-aminobutyric acid, or GABA. But that is not the case. The brain transforms spikes into many different chemicals used to signal other neurons. I'm intrigued by how we might tap the rich information encoded in chemical signaling.
In our lab, we implant devices into the brains of living animals to try to measure neurotransmission on the length and time scales over which it happens. Of course, we're not there yet. Where nanotechnology is really going to help us is in shrinking our recording elements and devices and allowing us to multiplex them so we can measure several neurotransmitters simultaneously.
Also, these neurochemicals are the major targets of most drugs used to treat brain diseases. If we extend brain activity mapping to include brain chemistry, it has the potential to help us understand disease processes and design new treatments.
ARTO NURMIKKO: I totally subscribe to that. These are complementary methodologies. We should complement spike measurements by tracking neurochemicals. Having both types of information will enable us to get at the actual function of neural circuits in the brain.
PAUL ALIVISATOS: While I agree about wanting to do multidimensional measurements, I believe that if we could measure just spikes for 100,000 or 1 million neurons, it would allow scientists to answer questions they cannot even ask today. Just operating at the level of spikes, we would have an enormously powerful tool.
Clearly, these other measurements would generate useful information. But it is not easy to measure everything, everywhere, all at once. There's going to be an evolution on that. I believe that if we do spikes alone, it would be a huge deal.
ANNE ANDREWS: I certainly don't mean to advocate not measuring spikes. But going back again to what we can do today, there are nice examples of people using carbon fiber microelectrodes to make high-speed electrophysiological and electrochemical recordings with the same device. I think it will be possible to make these multiplexed measurements, and I would like to see this incorporated in our thinking regarding the BRAIN Initiative from beginning rather than as an add-on later, because the brain does not communicate by electrical discharges alone.
HONGKUN PARK: Yes, multiplexing different types of measurements should be included in brain activity mapping. But I also believe that measuring the electrical spikes should be our primary goal.
ARTO NURMIKKO: Clearly, we need spikes to model the brain's networks and create computational models of brain functions. But I have to agree with Professor Andrews entirely. Spikes are only part of the story. Neurochemical activity complements the information we can observe with electrical and optical recording. It's really important.
I would even add another piece of the puzzle, metabolics. We will want local probes that measure brain metabolism near a single neuron or clusters of neurons.
TKF: So far, we've talked about networks of neurons. Do we need to understand what happens inside neurons?
HONGKUN PARK: There are many different things going on inside the brain, such as chemical signaling and metabolic changes. What's happening inside neurons might also be important. In fact, we do not even know how many different types of neurons there are in the brain or even in a single sensory organ, or how they affect brain function.
Yet I do not believe that these issues should not be the primary goal of brain activity mapping, at least initially. We understand that they play an important role. Down the road, when we have begun to assemble the basic circuit diagrams of the brain, we will want to study what happens inside neurons, how neurotransmitters work, and brain metabolism.
" While we need to study what goes on inside neurons, we should address how many neurons come together to execute functions. Trying to get that by mapping activity rather than connectivity feels to me the essence of what we’d like to achieve." — Paul Alivisatos
ANNE ANDREWS: I want to understand how the brain codes, stores, recalls, and uses information. So we need to think about spikes, but also how those spikes are translated into chemicals that move between neurons and then are transduced back into spikes.
PAUL ALIVISATOS: While we need to study what goes on inside neurons, we should address how many neurons come together to execute functions. Trying to get that by mapping activity rather than connectivity feels to me the essence of what we’d like to achieve.
These higher-level interconnections are the part of the brain about which we have the least information. It's hard to get even a clear hypothesis about how our neural circuits work. Over the near term, we should focus on how large numbers of neurons act together. We should carry out other work in parallel.
HONGKUN PARK: For our research to succeed, we should have clearly outlined goals. I believe how the neurons talk to each other is most important. Other research, such as what goes on inside neurons, will serve to supplement what we learn from spike activity.
ARTO NURMIKKO: If we look at the brain, it is an electrochemical machine. You cannot divorce its electrical aspects from the underlying chemistry. You have to have both out of the gate.
ANNE ANDREWS: I couldn't agree more.
TKF: No one wants to keep reinventing the wheel. Can we create nanotechnology systems that are adaptable enough to measure many different types of data?
This is not a simple problem, but I think we've been reasonably successful. At every step of the way, we've worked to make the chemistry and the technology generalizable, so that materials could present a variety of small-molecule neurotransmitters.
We plan to do the reverse as well, using biomolecules to recognize neurotransmitters. Artificial biomolecules could be added to various nanodevices, possibly made from carbon nanotubes or nanowires, to signal the presence of brain chemicals. They would be truly nanoscale in terms of the spatial and temporal resolution we will get from them.
PAUL ALIVISATOS: The goal here is to make truly integrated, flexible measurement systems. Having said that, let's think for a moment about telescopes. If you look at a telescope today, it doesn't look very much like the one Galileo used. There have been many, many generations of improvements, from better optics to new designs and instrumentation.
With brain activity mapping, we're talking about integrated measurement platforms that will evolve over many generations, and each one will add more functions. In my mind, this initiative will not have succeeded if it does not develop tools that very large numbers of researchers can use to ask interesting questions.
"[T]his idea of multiple approaches is really important, because we often learn different things from different measurement techniques. We want to take advantage of the richness we'll get from using different optical, electrical, and other technologies." — Anne Andrews
ARTO NURMIKKO: I would add that any toolkit should include capacity for bidirectionality. It should be able to stimulate and excite neurons in a controlled fashion, so we can elicit specific responses and conduct repeatable experiments. That type of bidirectionality seems really important. In contrast with astronomy, where we only observe, we want to talk to neurons, as well as listen.
HONGKUN PARK: Nanotechnology is a platform technology. Nanoscientists are trying to generate lots of different toolkit pieces, like nanomaterials with optical characteristics, nanowires that couple with neurons, nanosensors that can sense neurotransmitters, and devices that communicate wirelessly. We already have many of these pieces. What we should be doing now is to learn how to put them together in flexible ways to create tools suited for specific research challenges.
TKF: Clearly, nanotechnology will extend existing technologies. Will it also let us do something entirely new? Dr. Alivisatos, you've done work in self-assembling nanostructures. Could we create materials that form nanostructures that access the brain in entirely new ways? What new directions seem most exciting?
PAUL ALIVISATOS: Our colleague, George Church, has made some interesting proposals. He imagines synthetic molecules that could assemble nucleic acids into DNA at rates that vary with voltage. This way, each neuron firing would produce a bit of DNA. That would let us record a history of neuronal firing from DNA formed within artificial cells inside the brain. I find that to be breathtakingly exciting and I wouldn't underestimate it.
People in the nanoscience community are learning how to build these types of artificial systems. They are not as sophisticated as nature's, but we are getting better every day. That's a trend, and it's going to impact our ability to do things.
"Perhaps we can restore functions to the neurologically impaired, enabling them to move their arms or legs or to control electronic devices. If we understand how neurocircuits fail, we could not only diagnose mental illnesses, but perhaps communicate back some means of reestablishing more useful functions. We might be able to repair the errant brain." — Arto Nurmikko
I would say there are two tracks we should follow. On one hand, we want to create technology that is not so dissimilar from what we have now, and use it to extend our reach to thousands and tens of thousands of neurons. That will be very, very powerful.
On the other hand, we should also support research that does this type of exotic, state-of-the-art work, such as sophisticated self-assembling nanomaterials, and see how far we can take that. But this type of extreme research is a parallel process, and not necessarily the heart of the initiative as it is today.
TKF: The Brain Research through Advancing Innovative Neurotechnologies (BRAIN) initiative promises enough funding to change significantly what researchers might accomplish. Where do you see the work going in three to five years?
ARTO NURMIKKO: It might be possible to use nanotechnology to enable existing technologies to retrieve spike and perhaps neurochemistry data from one to ten thousand neurons. This would help us understand some pieces of our brain's motor or sensory circuitry, but only if we develop computational models that act like a Rosetta Stone to translate the language of the brain into something we understand. Then we can learn what each piece of circuitry is actually doing.
ANNE ANDREWS: I'm really excited about optical technologies. We now do optical measurements at really short distances, but we've seen really nice examples of imaging activity deeper and deeper in the brain. This will allow researchers to create larger optical maps of activity across neural circuits.
"Nanoscientists are trying to generate lots of different toolkit pieces, like nanomaterials with optical characteristics, nanowires that couple with neurons, nanosensors that can sense neurotransmitters, and devices that communicate wirelessly. ... What we should be doing now is to learn how to put them together in flexible ways to create tools suited for specific research challenges." — Hongkun Park
PAUL ALIVISATOS: My vote is that in three years, we will have just inaugurated six large-scale centers that bring together a few hundred scientists and engineers to develop brain activity mapping tools, each one designed to answer specific types of questions or make measurements in specific animals, such as zebra fish or rats. We could then exploit the technical approach best suited for tackling each of our big questions. That would be a really exciting development.
HONGKUN PARK: I think a lot of people in the neuroscience community will be using existing tools to study hundreds, thousands, or even tens of thousands of neurons in the cortical regions. A lot of nanoscientists, working with neuroscientists, will be busy building next-generation tools to fill in the brain activity map. It will be an exciting time.
TKF: If we really understood how the brain worked, what could we do? Could we create true artificial intelligence that thinks like us?
PAUL ALIVISATOS: I want to be cautious about the impact of this research, because we don't really know. From the scientific perspective, if we had tools to map brain activity, they would accelerate the pace of discovery enormously. That would have really important consequences for people studying neurological and neurodegenerative diseases. But I want to be cautious about predicting a definite outcome with respect to curing diseases or creating artificial systems.
ANNE ANDREWS: I'm also cautious about artificial intelligence, but I am enthusiastic about the potential for human health. In my own field, psychiatry, all of the drugs we use to treat psychiatric illnesses have descended from four accidentally discovered chemical compounds. We have yet to design a drug rationally.
This is because we have so little understanding of what's wrong with the brain in these disorders. I'm excited about the potential for advances in our understanding of the brain that could lead to fundamentally new treatments. Human health may not be the goal of brain activity mapping, but I think information from the project will be directly applicable to human health.
ARTO NURMIKKO: I think bi-directional probes will let us use our knowledge to help people. Perhaps we can restore functions to the neurologically impaired, enabling them to move their arms or legs or to control electronic devices. If we understand how neurocircuits fail, we could not only diagnose mental illnesses, but perhaps communicate back some means of reestablishing more useful functions. We might be able to repair the errant brain.