Over the past two decades, scientists’ ability to track neural activity with light has changed the face of neuroscience. Using genetically encoded calcium indicators, scientists can simultaneously monitor thousands of neurons, granting a much broader view of brain activity than was previously possible. “Now we can see the brain producing this beautiful symphony,” says Vincent Pieribone, a neuroscientist at Yale University.
But ever since these indicators were first developed, scientists have been striving to make them better. They want sensors that are brighter or glow in different colors, operate faster or slower, or can target different locations. Scientists presented their newest engineering advances at the third annual meeting of BRAIN Initiative investigators, held in Bethesda, Maryland, in December. They described calcium sensors that fluoresce in red, far-red and near infrared, as well as sensors designed to directly monitor changes in voltage. Whereas implanted electrodes can directly measure voltage changes in the brain, genetically engineered sensors have the potential to measure orders of magnitude more cells and to do so less invasively.
“The way it looks right now, voltage sensing, calcium sensing and electrophysiology will all have important roles to play but address different kinds of problems,” says Karel Svoboda, a neuroscientist at the Howard Hughes Medical Institute’s Janelia Research Campus and an investigator with the Simons Collaboration on the Global Brain.
Genetically encoded calcium sensors are designed to detect the rush of calcium ions that flows into neurons when the cells fire, providing an indirect measure of neural activity. The molecules are typically fashioned from a fusion of three naturally occurring proteins — a calcium indicator called calmodulin, which changes shape when it binds calcium, a peptide that binds to the calcium-bound calmodulin, and a fluorescent molecule, typically green fluorescent protein (GFP). When calcium is present, the sensor’s structure changes shape to reveal a fluorescent glow, signaling that the neuron is active.
Calcium has several benefits as a proxy for electrical activity. Its concentration increases several-fold in response to an action potential, triggering an easily detectable change in fluorescence. In addition, calcium’s baseline concentration in neurons is extremely low, meaning that calcium indicators are dark in the absence of electrical activity. “We think it’s by far the most sensitive way to look at neural activity,” Svoboda says.
But calcium indicators also have limitations. Most existing versions employ GFP, which is difficult to image deep in the brain. And calcium signals change far more slowly than the electrical signals they mirror, making it difficult to measure the timing of action potentials.
In the past few years, a handful of labs have begun to use experimental evolution to develop new and improved sensors. Scientists start with an existing sensor molecule and genetically tweak it in myriad ways, creating large libraries of candidate molecules. The candidates are then run through a series of assays to find those with the desired property, be it a faster, slower or brighter response. “Protein design is more art than science, so it’s difficult to predict which modification will ultimately be beneficial,” Svoboda says. “We find you have to test a lot of sensors to make progress.”
Svoboda and his collaborators have developed high-throughput screens to identify new sensor molecules, dubbed the GENIE project (for Genetically Encoded Neural Indicators and Effectors). That effort produced GCaMP6, the most widely used protein calcium sensor today. The GENIE project is producing even better GCaMPs, which researchers aim to distribute by the middle of 2017. GENIE scientists give away the sensors they make, even before they publish the research. They are “basically a biotech company whose business model is to lose $1.5 million per year,” Svoboda said at the BRAIN Initiative meeting in December.
Svoboda is particularly interested in developing calcium sensors capable of capturing large populations of neurons at single-cell resolution, perhaps across multiple brain regions. To do that, scientists need to reduce background fluorescence, making it even easier to detect the change in fluorescence in response to calcium. This type of sensor will help bridge the gap between individual neurons and brain areas, Svoboda says.
The mismatch between rapid action potentials, which last about a millisecond, and the slow flow of calcium, which spans 10 to 100 milliseconds, can be both helpful and problematic. Like a slow-motion sports replay, calcium’s sluggish rise and fall gives scientists more time to measure the signal, which in turn makes it easier to distinguish real activity from background noise. “That’s why I think calcium sensors are so important,” Svoboda says. “You can image them with high signal-to-noise ratio and sensitivity.” For this reason, Svoboda thinks most people don’t mind the slow speed. “If they were much faster, we wouldn’t know how to image them,” he says. “I think fast time resolution is the realm of electrophysiology and will remain so.”
But this sensitivity comes at a cost. While existing calcium indicators can detect spikes that are well spaced out, tracking the timing of action potentials within bursts of spikes is much more difficult. Purkinje cells of the cerebellum, for example, can fire 100 times per second. Svoboda and collaborators on the GENIE project are working on faster calcium sensors that mimic the 10 millisecond rise in calcium concentrations that accompany action potentials. (Existing sensors operate on a 100 millisecond scale.)
Samuel Wang, a neuroscientist at Princeton University, and his collaborators are working on sensors that dim more quickly once calcium levels plummet. GCaMP, for example, contains a central fluorophore split in half like a clamshell. When calcium binds to GCaMP, the two halves of the clamshell are brought together, making the fluorophore glow more brightly. Once closed, it sticks, says Wang. “It gets turned on and stays stuck on,” making even a fast calcium signal look slow. “We’re trying to make it less sticky so that it comes apart better,” he says.
So far, Wang’s team has developed sensors that come apart in as little as 10 milliseconds, approaching the speed of synthetic sensors. “Now we’re testing our best variants to see if they work as well in neurons as they do in biochemical conditions,” Wang says. “I think we’re within a few months of creating a sensor that can detect single spikes at high rates.”
Another way to circumvent calcium’s slow dynamics is to focus on specific parts of the neuron. In synaptic boutons — structures that form connections with other neurons — calcium moves in and out of the cell much faster than it does in the cell body, about 1 millisecond and 20 milliseconds, respectively. (These structures have a higher surface-to-volume ratio than the cell body, meaning that calcium is removed much more quickly, via membrane pumps.) Wang and his collaborators are developing methods to monitor calcium flux at boutons. But to be successful, microscopists will have to develop fast scanning strategies that take full advantage of the new sensors, Wang says.
Other researchers are designing sensors that glow in different colors by attaching calmodulin to different indicator molecules. “It’s like a universal adapter,” says Robert Campbell, a biochemist at the University of Alberta in Canada. “We can take advantage of calmodulin and its conformational change and connect it to a fluorescent protein.”
Red, far-red and near-infrared sensors are of particular interest because these longer wavelengths of light can penetrate further into brain tissue. According to Campbell, the latest generation of red indicators perform similarly to the best available green ones. His team is now delving into longer wavelengths, such as near-infrared. “We believe that will get us deeper tissue penetration, which will be a big advantage for many applications, especially BRAIN Initiative type projects where we are trying to image as large a volume of brain as possible,” Campbell says.
Expanding the rainbow of sensors also gives scientists more options for combining different types of tools, such as optogenetics. “Now we can combine red sensors with green optogenetics and stimulate cells in green or blue and record them in red with no overlap in wavelengths of light,” Pieribone says. “Every time you get away from standard colors, it opens up lots of possibilities.”
Searching the oceans for red fluorescent proteins
When Vincent Pieribone and his collaborators first proposed scouring the seas for red fluorescent proteins in the early 2000s, others told them it was a fool’s errand. “People said, ‘Don’t look in the ocean, because it’s blue,’” he says. “But there are animals down there that have red pigments and use cryptic wavelengths to signal each other.” Manta shrimp, for example, send mating signals using red fluorophores that are invisible to other creatures. To find these proteins, the researchers venture into the water at night, armed with special lights and cameras. They grab the rare red specimens and then try to decipher the relevant protein. Sometimes these proteins resemble GFP’s structure and sometimes they are completely different, constructed from a sequence that scientists would not have predicted would glow. The researchers are working with proteins from eels, lizard fish, and sharks.
Of course, calcium indicators give only an indirect measure of electrical activity in neurons. “We’re not really seeing the real activity,” Campbell says. “That’s why there is lots of interest in developing voltage indicators — it would let us see the real signal.”
Electrical activity in neurons isn’t limited to action potentials. Cells also experience subtle changes in membrane potential that influence activity. For example, inhibitory inputs can hyperpolarize neurons, generating a strong negative charge that makes the cell more difficult to fire. “I feel like we’ve been sidetracked by calcium indicators,” Pieribone says. “Of course they are useful — the signals are beautiful and large in response to cell firing. But nerve cell function is about electrical properties. Neurons have subtle changes in membrane potential that define their personalities.”
These subthreshold changes may prove interesting, showing how electrical signals are integrated in different parts of the neuron. Voltage sensors may be the only way to track these signals in large numbers of neurons. “There’s really no other way of doing that,” Svoboda says.
Development of voltage sensors lags behind that of calcium sensors, largely because nothing like calmodulin — a naturally occurring, calcium-sensitive protein — exists for voltage. The closest candidates are voltage-sensing domains, portions of proteins that sit in the cell membrane and change shape when the membrane depolarizes. But Campbell says figuring out how to connect these domains to fluorescent proteins so that they change shape has been a challenge. As a result, “It’s a much more nascent field,” Campbell says. “Most people would agree it hasn’t gotten to the point where it’s a broadly useful tool.”
For voltage sensors to become as user-friendly as their calcium counterparts, they need to become much brighter — exhibiting a larger fluorescence change in response to a change in membrane potential. But most of the currently available voltage sensors glow when the cell is quiet and dim when the cell is active, the reverse of calcium sensors. That pattern is more difficult to monitor, akin to spotting a dim light in a bright building during daylight, versus a bright light in a dark building at night. Pieribone’s team managed to flip the pattern using experimental evolution, creating a voltage sensor that glows when the cell fires. “It seems trivial, but it’s a make or break for a good probe,” he says.
Voltage sensors’ rapid dynamics are both an advantage and a challenge. “One problem with voltage sensors is the signals are too brief,” says Wang. Voltage changes happen so fast that they are difficult to detect with the cameras that scientists use to detect fluorescence changes in genetically engineered sensors. “You have to use a short exposure time, typically one millisecond, to capture changes,” Campbell says. “That means you get few photons and very noisy data, similar to taking photographs in very low light.”