Neurons See What We Tell Them to See

Neurons programmed to fire at specific faces—such as the famously reported “Jennifer Aniston neuron”—may be more in line with the conscious recognition of faces than the actual images seen. Subjects presented with a blended face, such as an amalgamation of Bill Clinton and George W. Bush, had significantly more firing of such face-specific neurons when they recognized the blended or morphed face as one person or the other. Results of the study led by Christof Koch at the Allen Institute for Brain Science, and carried out by neuroscientists Rodrigo Quian Quiroga at the University of Leicester, Alexander Kraskov at University College London and Florian Mormann at the University of Bonn, under the clinical supervision of the neurosurgeon Itzhak Fried at the University of California at Los Angeles Medical School, are published online today in the journal Neuron.

Some neurons in the region of the brain known as the medial temporal lobe are observed to be extremely selective in the stimuli to which they respond. A cell may only fire in response to different pictures of a particular person who is very familiar to the subject (such as loved one or a celebrity), the person’s written or spoken name, or simply recalling the person from memory.

“These highly specific cells are an entry point to investigate how the brain makes meaning out of visual information,” explains Christof Koch, Chief Scientific Officer at the Allen Institute for Brain Science and senior author on the paper. “We wanted to know how these cells responded not just to a simple image of a person’s face, but to a more ambiguous image of that face averaged or morphed with another person’s face.”

For the trials, subjects were shown either the face of individuals such as Bill Clinton or George W. Bush (the “adaptor” image), and then an ambiguous face which was a blend of both faces. Primed with the Clinton image, subjects tended to recognize Bush’s face in the blended image, while subjects who saw Bush’s face first recognized the blended face as Clinton. That is, even though the blended images were identical, subjects tended to consciously perceive the identity of face to which they were not adapted.

Researchers wanted to know whether these selective neurons responded to the actual image on the screen, or whether they responded more to the perception that the image caused in the brain of the subject. When subjects recognized the ambiguous face as belonging to Clinton, their Clinton-specific neurons fired. However, when subjects recognized that same face as Bush, the neurons fired significantly less. These results indicated that conscious recognition of the face played a crucial role in whether the neurons fired, rather than the raw visual stimulus.

“This study provides further evidence that stimulus-specific neurons in the medial temporal lobe follow the subjective perception of the person, as opposed to faithfully reporting the visual stimulus the person sees,” explains Koch. “This distinction may help us glean insight into how the brain takes raw visual information and transforms it into something meaningful, which can be further modulated by other aspects of experience in the brain.”

(Image caption: Archer1 fluorescence in a cultured rat hippocampal neuron. By monitoring changes in this fluorescence at up to a thousand frames per second, researchers can track the electrical activity of the cell. Credit: Nicholas Flytzanis, Claire Bedbrook and Viviana Gradinaru/Caltech)

Sensing Neuronal Activity With Light

For years, neuroscientists have been trying to develop tools that would allow them to clearly view the brain’s circuitry in action—from the first moment a neuron fires to the resulting behavior in a whole organism. To get this complete picture, neuroscientists are working to develop a range of new tools to study the brain. Researchers at Caltech have developed one such tool that provides a new way of mapping neural networks in a living organism.

The work—a collaboration between Viviana Gradinaru (BS ‘05), assistant professor of biology and biological engineering, and Frances Arnold, the Dick and Barbara Dickinson Professor of Chemical Engineering, Bioengineering and Biochemistry—was described in two separate papers published this month.

When a neuron is at rest, channels and pumps in the cell membrane maintain a cell-specific balance of positively and negatively charged ions within and outside of the cell resulting in a steady membrane voltage called the cell’s resting potential. However, if a stimulus is detected—for example, a scent or a sound—ions flood through newly open channels causing a change in membrane voltage. This voltage change is often manifested as an action potential—the neuronal impulse that sets circuit activity into motion.

The tool developed by Gradinaru and Arnold detects and serves as a marker of these voltage changes.

"Our overarching goal for this tool was to achieve sensing of neuronal activity with light rather than traditional electrophysiology, but this goal had a few prerequisites," Gradinaru says. "The sensor had to be fast, since action potentials happen in just milliseconds. Also, the sensor had to be very bright so that the signal could be detected with existing microscopy setups. And you need to be able to simultaneously study the multiple neurons that make up a neural network."

The researchers began by optimizing Archaerhodopsin (Arch), a light-sensitive protein from bacteria. In nature, opsins like Arch detect sunlight and initiate the microbes’ movement toward the light so that they can begin photosynthesis. However, researchers can also exploit the light-responsive qualities of opsins for a neuroscience method called optogenetics—in which an organism’s neurons are genetically modified to express these microbial opsins. Then, by simply shining a light on the modified neurons, the researchers can control the activity of the cells as well as their associated behaviors in the organism.

Gradinaru had previously engineered Arch for better tolerance and performance in mammalian cells as a traditional optogenetic tool used to control an organism’s behavior with light. When the modified neurons are exposed to green light, Arch acts as an inhibitor, controlling neuronal activity—and thus the associated behaviors—by preventing the neurons from firing.

However, Gradinaru and Arnold were most interested in another property of Arch: when exposed to red light, the protein acts as a voltage sensor, responding to changes in membrane voltages by producing a flash of light in the presence of an action potential. Although this property could in principle allow Arch to detect the activity of networks of neurons, the light signal marking this neuronal activity was often too dim to see.

To fix this problem, Arnold and her colleagues made the Arch protein brighter using a method called directed evolution—a technique Arnold originally pioneered in the early 1990s. The researchers introduced mutations into the Arch gene, thus encoding millions of variants of the protein. They transferred the mutated genes into E. coli cells, which produced the mutant proteins encoded by the genes. They then screened thousands of the resulting E. coli colonies for the intensities of their fluorescence. The genes for the brightest versions were isolated and subjected to further rounds of mutagenesis and screening until the bacteria produced proteins that were 20 times brighter than the original Arch protein.

A paper describing the process and the bright new protein variants that were created was published in the September 9 issue of the Proceedings of the National Academy of Science.

"This experiment demonstrates how rapidly these remarkable bacterial proteins can evolve in response to new demands. But even more exciting is what they can do in neurons, as Viviana discovered," says Arnold.

In a separate study led by Gradinaru’s graduate students Nicholas Flytzanis and Claire Bedbrook, who is also advised by Arnold, the researchers genetically incorporated the new, brighter Arch variants into rodent neurons in culture to see which of these versions was most sensitive to voltage changes—and therefore would be the best at detecting action potentials. One variant, Archer1, was not only bright and sensitive enough to mark action potentials in mammalian neurons in real time, it could also be used to identify which neurons were synaptically connected—and communicating with one another—in a circuit.

The work is described in a study published on September 15 in the journal Nature Communications.

"What was interesting is that we would see two cells over here light up, but not this one over there—because the first two are synaptically connected," Gradinaru says. "This tool gave us a way to observe a network where the perturbation of one cell affects another."

However, sensing activity in a living organism and correlating this activity with behavior remained the biggest challenge. To accomplish this goal Gradinaru’s team worked with Paul Sternberg, the Thomas Hunt Morgan Professor of Biology, to test Archer1 as a sensor in a living organism—the tiny nematode worm C. elegans. “There are a few reasons why we used the worms here: they are powerful organisms for quick genetic engineering and their tissues are nearly transparent, making it easy to see the fluorescent protein in a living animal,” she says.

After incorporating Archer1 into neurons that were a part of the worm’s olfactory system—a primary source of sensory information for C. elegans—the researchers exposed the worm to an odorant. When the odorant was present, a baseline fluorescent signal was seen, and when the odorant was removed, the researchers could see the circuit of neurons light up, meaning that these particular neurons are repressed in the presence of the stimulus and active in the absence of the stimulus. The experiment was the first time that an Arch variant had been used to observe an active circuit in a living organism.

Gradinaru next hopes to use tools like Archer1 to better understand the complex neuronal networks of mammals, using microbial opsins as sensing and actuating tools in optogenetically modified rodents.

"For the future work it’s useful that this tool is bifunctional. Although Archer1 acts as a voltage sensor under red light, with green light, it’s an inhibitor," she says. "And so now a long-term goal for our optogenetics experiments is to combine the tools with behavior-controlling properties and the tools with voltage-sensing properties. This would allow us to obtain all-optical access to neuronal circuits. But I think there is still a lot of work ahead."

One goal for the future, Gradinaru says, is to make Archer1 even brighter. Although the protein’s fluorescence can be seen through the nearly transparent tissues of the nematode worm, opaque organs such as the mammalian brain are still a challenge. More work, she says, will need to be done before Archer1 could be used to detect voltage changes in the neurons of living, behaving mammals.

And that will require further collaborations with protein engineers and biochemists like Arnold.

"As neuroscientists we often encounter experimental barriers, which open the potential for new methods. We then collaborate to generate tools through chemistry or instrumentation, then we validate them and suggest optimizations, and it just keeps going," she says. "There are a few things that we’d like to be better, and through these many iterations and hard work it can happen."

Researchers Reveal Pathway that Contributes to Alzheimer’s Disease

Researchers at Jacksonville’s campus of Mayo Clinic have discovered a defect in a key cell-signaling pathway they say contributes to both overproduction of toxic protein in the brains of Alzheimer’s disease patients as well as loss of communication between neurons — both significant contributors to this type of dementia.

Their study, in the online issue of Neuron, offers the potential that targeting this specific defect with drugs “may rejuvenate or rescue this pathway,” says the study’s lead investigator, Guojun Bu, Ph.D., a neuroscientist at Mayo Clinic, Jacksonville, Fla.

“This defect is likely not the sole contributor to development of Alzheimer’s disease, but our findings suggest it is very important, and could be therapeutically targeted to possibly prevent Alzheimer’s or treat early disease,” he says.

The pathway, Wnt signaling, is known to play a critical role in cell survival, embryonic development and synaptic activity — the electrical and chemical signals necessary for learning and memory. Any imbalance in this pathway (too much or too little activity) leads to disease — the overgrowth of cells in cancer is one example of overactivation of this pathway.

While much research on Wnt has focused on diseases involved in overactive Wnt signaling, Dr. Bu’s team is one of the first to demonstrate the link between suppressed Wnt signaling and Alzheimer’s disease.

“Our finding makes sense, because researchers have long known that patients with cancer are at reduced risk of developing Alzheimer’s disease, and vice versa,” Dr. Bu says. “What wasn’t known is that Wnt signaling was involved in that dichotomy.”

Using a new mouse model, the investigators discovered the key defect that leads to suppressed Wnt signaling in Alzheimer’s. They found that the low-density lipoprotein receptor-related protein 6 (LRP6) is deficient, and that LRP6 regulates both production of amyloid beta, the protein that builds up in the brains of AD patients, and communication between neurons. That means lower than normal levels of LRP6 leads to a toxic buildup of amyloid and impairs the ability of neurons to talk to each other.

Mice without LRP6 had impaired Wnt signaling, cognitive impairment, neuroinflammation and excess amyloid.

The researchers validated their findings by examining postmortem brain tissue from Alzheimer’s patients — they found that LRP6 levels were deficient and Wnt signaling was severely compromised in the human brain they examined.

The good news is that specific inhibitors of this pathway are already being tested for cancer treatment. “Of course, we don’t want to inhibit Wnt in people with Alzheimer’s or at risk for the disease, but it may be possible to use the science invested in inhibiting Wnt to figure out how to boost activity in the pathway,” Dr. Bu says.

“Identifying small molecule compounds to restore LRP6 and the Wnt pathway, without inducing side effects, may help prevent or treat Alzheimer’s disease,” he says. “This is a really exciting new strategy — a new and fresh approach.”

Researchers make new discovery about brain’s 3-D shape processing

While previous studies of the brain suggest that processing of objects and places occurs in very different locations, a Johns Hopkins University research team has found that they are closely related.

In research funded by the National Institutes of Health and published today in the journal Neuron, a team led by Johns Hopkins researcher Charles E. Connor reports that a major pathway long associated with object shape also carries information about landscapes and other environments.

Siavash Vaziri, then a biomedical engineering graduate student and now a post-doctoral fellow in the Connor lab, studied how neurons in the ventral visual pathway of the monkey brain respond to 3-D images. In one channel of the ventral pathway, neurons responded to small, discrete objects as expected. But in a neighboring, parallel channel, the researchers were surprised by the overwhelming responsiveness of many neurons to large-scale environments that surround the viewer, extending beyond the field of view.

"We were entirely surprised ourselves," said Connor, senior author of the paper. "Based on decades of research, we expected that all neurons in the ventral pathway would be primarily concerned with objects."

The ventral pathway is one of the two major branches of high-level visual processing in humans and other primates. It is sometimes called the “what” pathway, based on its role in identifying objects based on their shapes and colors.

"Dr. Vaziri’s finding is exciting because it puts environmental shape information together with object shape information in two densely connected neighboring channels. This could be a site for integrating object information into environmental contexts in order to understand scenes," Connor said.

Vaziri used microelectrodes to study how individual neurons responded to a large variety of 3-D shapes projected onto a large screen. Depth structure was conveyed by shading, texture gradients, and stereopsis, the effect used in 3-D movies. The shape stimuli evolved during the experiment based on the neuron’s responses, sometimes in the direction of small objects near the viewer, sometimes in the direction of environments filling the screen and surrounding the viewer.

Connor, a professor of neuroscience and the director of the Zanvyl Krieger Mind/Brain Institute at Johns Hopkins, is a noted expert on the neural mechanisms of object vision. His research focuses on deciphering the algorithms that make object vision possible and explain the nature of visual experience.

"Many people would say that vision is our richest and most vivid experience," said Connor. "We want to understand the brain events that create that experience."

Connor said that the next step will be to understand how object and environment information are integrated between the two channels.

"We don’t typically experience objects in isolation," Connor said. "We experience scenes, that is, environments containing multiple objects. We now think that the ventral pathway may be where all that information gets put together to create scene understanding."

Neurons express ‘gloss’ using three perceptual parameters

Japanese researchers showed monkeys a number of images representing various glosses and then they measured the responses of 39 neurons by using microelectrodes. They found that a specific population of neurons changed the intensities of the responses linearly according to either the contrast-of-highlight, sharpness-of-highlight, or brightness of the object. This shows that these 3 perceptual parameters are used as parameters when the brain recognizes a variety of glosses. They also found that different parameters are represented by different populations of neurons. This was published in the Journal of Neuroscience.

The gloss of an object surface provides information about the condition of that object. For instance, whether it is wet or dry, whether food is fresh or old. Several gloss-related physical parameters such as specular reflectance and diffuse reflectance have been described and used in computer graphics so far. However, the parameters used when neurons respond to gloss have not yet been found.

A Japanese research group led by Hidehiko Komatsu, professor of the National Institute for Physiological Sciences (NIPS), National Institutes of Natural Sciences (NINS), in collaboration with the Advanced Telecommunications Research Institute International (ATR) prepared 16 images representing various glosses and showed them to monkeys. In a circumscribed area in the inferior temporal cortex of the brain, neurons strengthened their responses proportionately as the contrast-of-highlight and/or sharpness-of-highlight got higher. Neural responses also vary greatly depending on the brightness, for instance, whether the object is black, gray, or white. Furthermore, the perceptual gloss parameters of the presented image could be fairly precisely predicted from the strengths of the population neural responses.

By the application of these findings in an artificial image recognition system, the researchers are expecting that it would be able to develop robots that recognize gloss like humans.

Tipping the Balance of Behavior

Humans with autism often show a reduced frequency of social interactions and an increased tendency to engage in repetitive solitary behaviors. Autism has also been linked to dysfunction of the amygdala, a brain structure involved in processing emotions. Now Caltech researchers have discovered antagonistic neuron populations in the mouse amygdala that control whether the animal engages in social behaviors or asocial repetitive self-grooming. This discovery may have implications for understanding neural circuit dysfunctions that underlie autism in humans.

This discovery, which is like a “seesaw circuit,” was led by postdoctoral scholar Weizhe Hong in the laboratory of David J. Anderson, the Seymour Benzer Professor of Biology at Caltech and an investigator with the Howard Hughes Medical Institute. The work was published online on September 11 in the journal Cell. 

"We know that there is some hierarchy of behaviors, and they interact with each other because the animal can’t exhibit both social and asocial behaviors at the same time. In this study, we wanted to figure out how the brain does that," Anderson says.

Anderson and his colleagues discovered two intermingled but distinct populations of neurons in the amygdala, a part of the brain that is involved in innate social behaviors. One population promotes social behaviors, such as mating, fighting, or social grooming, while the other population controls repetitive self-grooming—an asocial behavior.

Interestingly, these two populations are distinguished according to the most fundamental subdivision of neuron subtypes in the brain: the “social neurons” are inhibitory neurons (which release the neurotransmitter GABA, or gamma-aminobutyric acid), while the “self-grooming neurons” are excitatory neurons (which release the neurotransmitter glutamate, an amino acid).

To study the relationship between these two cell types and their associated behaviors, the researchers used a technique called optogenetics. In optogenetics, neurons are genetically altered so that they express light-sensitive proteins from microbial organisms. Then, by shining a light on these modified neurons via a tiny fiber optic cable inserted into the brain, researchers can control the activity of the cells as well as their associated behaviors.

Using this optogenetic approach, Anderson’s team was able to selectively switch on the neurons associated with social behaviors and those linked with asocial behaviors.

With the social neurons, the behavior that was elicited depended upon the intensity of the light signal. That is, when high-intensity light was used, the mice became aggressive in the presence of an intruder mouse. When lower-intensity light was used, the mice no longer attacked, although they were still socially engaged with the intruder—either initiating mating behavior or attempting to engage in social grooming.

When the neurons associated with asocial behavior were turned on, the mouse began self-grooming behaviors such as paw licking and face grooming while completely ignoring all intruders. The self-grooming behavior was repetitive and lasted for minutes even after the light was turned off.

The researchers could also use the light-activated neurons to stop the mice from engaging in particular behaviors. For example, if a lone mouse began spontaneously self-grooming, the researchers could halt this behavior through the optogenetic activation of the social neurons. Once the light was turned off and the activation stopped, the mouse would return to its self-grooming behavior.

Surprisingly, these two groups of neurons appear to interfere with each other’s function: the activation of social neurons inhibits self-grooming behavior, while the activation of self-grooming neurons inhibits social behavior. Thus these two groups of neurons seem to function like a seesaw, one that controls whether mice interact with others or instead focus on themselves. It was completely unexpected that the two groups of neurons could be distinguished by whether they were excitatory or inhibitory. “If there was ever an experiment that ‘carves nature at its joints,’” says Anderson, “this is it.”

This seesaw circuit, Anderson and his colleagues say, may have some relevance to human behavioral disorders such as autism.

"In autism," Anderson says, "there is a decrease in social interactions, and there is often an increase in repetitive, sometimes asocial or self-oriented, behaviors"—a phenomenon known as perseveration. "Here, by stimulating a particular set of neurons, we are both inhibiting social interactions and promoting these perseverative, persistent behaviors."

Studies from other laboratories have shown that disruptions in genes implicated in autism show a similar decrease in social interaction and increase in repetitive self-grooming behavior in mice, Anderson says. However, the current study helps to provide a needed link between gene activity, brain activity, and social behaviors, “and if you don’t understand the circuitry, you are never going to understand how the gene mutation affects the behavior.” Going forward, he says, such a complete understanding will be necessary for the development of future therapies.

But could this concept ever actually be used to modify a human behavior?

"All of this is very far away, but if you found the right population of neurons, it might be possible to override the genetic component of a behavioral disorder like autism, by just changing the activity of the circuits—tipping the balance of the see-saw in the other direction," he says.