Controlling the Brain With Holograms
To truly understand how populations of neurons encode information, scientists need to be able to speak the language of the brain — in other words, they need to precisely target scattered groups of individual cells at superfast speeds. Researchers have made great strides toward this goal over the last several years, developing new tools that use light to target brain cells with better and better precision.
New research, published in Nature Neuroscience in April, pushes the technology further, offering unprecedented control over neural firing. “This paper is exciting because it shows we can simultaneously read and write patterns of activity with millisecond precision in up to 50 neurons in the intact brain,” says Michael Häusser, a neuroscientist at University College London and an investigator with the Simons Collaboration on the Global Brain who was not involved in the study. “By manipulating activity of ensembles in a precise way, we can get new insights into the neural codes used by the cortex for processing and storing information.”
The new research is part of a growing effort to develop ‘all-optical’ tools for controlling and recording brain activity. Scientists have been developing different components of such technology over the last few years, but until now, they haven’t been able to target complex three-dimensional subgroups of neurons with fine temporal precision. “The interventions we have are just too coarse,” says Alan Mardinly, a postdoctoral researcher in Hillel Adesnik’s lab at the University of California, Berkeley, and co-first author on the paper. “Neuroscientists would love to be able to re-create specific activity in the brain.”
In traditional optogenetics, scientists flood brain tissue with light, activating or silencing all neurons tagged with an optogenetic sensor, or opsin. But natural brain activity is much subtler, with subsets of neurons in different regions firing in a precisely orchestrated sequence. Targeting cells with a broad brush prevents scientists from uncovering the specific pattern of activity that underlies a particular behavior, sensation or thought.
A technique called computer-generated holography can better target specific cells by using spatial light modulators to create three-dimensional patterns of light. Neuroscientists first combined holography and optogenetics in 2010. Computer-generated holography is precise, generating many small points of light. But the light can leak above and below the cells of interest, triggering off-target activity and limiting the technique’s utility for controlling larger, distributed groups of cells.
In the new study, researchers added a technique called temporal focusing to limit light not just in the lateral plane but at specific depths. The technology splits light pulses into slightly different bands that take different paths through the tissue, reuniting in the target focal plane. “Say there are 30 neurons we want to target; we can instruct our system to activate those neurons and only those neurons, which is a big step forward,” says Nicolas Pégard, a postdoctoral researcher in Adesnik’s and Laura Waller’s labs at Berkeley and co-first author. The group recently published a technique incorporating temporal focusing, dubbed 3D-SHOT (three-dimensional scanless holographic optogenetics with temporal focusing), which they use in the new study.
“How they create light patterns and rapidly switch them is a significant step forward in the methodology of the holography,” says Ehud Isacoff, a neuroscientist at Berkeley who collaborates with the researchers but was not directly involved in the study.
High-precision delivery of light is only half of the equation. To take full advantage of that precision, the researchers also needed to improve their light-activated switches. Existing opsins are too weak or too slow to activate and silence cells with high spatial and temporal precision, so the researchers engineered a faster and more potent version called ChroME. This version “grants temporal control for writing a precise pattern of activity — you can only be precise if you can control neurons quickly,” says Ian Oldenburg, an SCGB fellow in the Adesnik lab and co-first author on the paper. The activation of the new opsin can mimic peak neuronal firing rates, about 30 times or more per second. The researchers also enhanced an inhibitory opsin to more effectively silence neurons.
The team then added a molecular tag to both opsins limiting their expression to the cell body, rather than the axons or dendrites, to reduce the chances of accidentally triggering neighboring cells. That’s important for re-creating the brain’s sparse code, where few neurons are active at the same time, Isacoff says.
They also carefully tested and benchmarked the new opsins, measuring the spatial and temporal resolution for activating and silencing neurons in slices and in vivo. “It sets a new technical standard in the field with the rigor in which it benchmarks these opsins,” Häusser says.
Other groups have been working on each of these components individually. But the Berkeley team wanted to make sure all the parts functioned together. “We are trying to build an integrated system that anyone can use rather than a niche system that accomplishes speed or potency,” Oldenburg says.
“We’ve done all the necessary tests to make sure it works. That in itself is the main novelty,” Pégard says. “It’s meant to be used by neuroscientists; we went the extra mile to make it practical.” The sequence for the opsins is on Addgene, the code for the holograph technology is available on GitHub, and instructions for how to build the device are in the two papers.
The system integrates both optogenetics and calcium imaging, meaning that scientists can simultaneously monitor and control neural activity. They can then choose which neurons to target by identifying those that are active during a specific task. For example, to understand the neural circuits involved in visual discrimination, researchers can pinpoint the neurons that are most tightly linked to the animal’s ability to discriminate two different stimuli and activate or silence those cells to monitor the influence on both behavior and brain activity more broadly.
“It’s an impressive technical achievement,” Häusser says. “We know that neural codes in the cortex involve millisecond patterns of activity in hundreds to thousands of neurons in specific circuits. Until now, we lacked the ability to read out, reproduce and manipulate patterns with the necessary temporal and spatial precision. The kind of tools this paper benchmarks put us in this ballpark, giving us the ability to interrogate the brain with the necessary temporal and spatial resolution.”
Scientists are now beginning to use the technology to answer specific questions. For example, a temporally precise method for controlling the brain will allow scientists to test theories predicting whether it’s the timing of spikes or the rate at which they occur that matters for encoding information. “We’re getting to the point in the field where we have a suite of tools that enable us to interact with the brain on a crucial scale at which we think it works,” says Adam Packer, a neuroscientist at the University of Oxford. “We can interact with single spikes and single cells and examine how spikes and neurons encode information about the external world and internal state of the animal.”
A number of basic questions remain. It’s unclear whether being able to target 50 neurons is adequate to address fundamental questions in neural coding. “There is still the question about how many neurons you need to activate over what size of brain territory to produce a synthetic percept that you can detect and is meaningful,” Isacoff says. And though the new method targets many more neurons than previously possible, it may be too few, he says. Oldeburg notes that because of the technology’s temporal precision — they can target a new set of neurons roughly every 300 Hz — they can theoretically control many more neurons sequentially.
“No one agrees on the [ideal number of neurons to target] because, until recently, no one has been able to do the experiment where you can control a specific group of neurons in this manner at a precise time,” Packer says.
