Neuroscientists find the internal workings of next-word prediction models resemble those of language-processing centers in the brain.

In the past few years, artificial intelligence models of language have become very good at certain tasks. Most notably, they excel at predicting the next word in a string of text; this technology helps search engines and texting apps predict the next word you are going to type.

The most recent generation of predictive language models also appears to learn something about the underlying meaning of language. These models can not only predict the word that comes next, but also perform tasks that seem to require some degree of genuine understanding, such as question answering, document summarization, and story completion. 

Such models were designed to optimize performance for the specific function of predicting text, without attempting to mimic anything about how the human brain performs this task or understands language. But a new study from MIT neuroscientists suggests the underlying function of these models resembles the function of language-processing centers in the human brain.

Computer models that perform well on other types of language tasks do not show this similarity to the human brain, offering evidence that the human brain may use next-word prediction to drive language processing.

“The better the model is at predicting the next word, the more closely it fits the human brain,” says Nancy Kanwisher, the Walter A. Rosenblith Professor of Cognitive Neuroscience, a member of MIT’s McGovern Institute for Brain Research and Center for Brains, Minds, and Machines (CBMM), and an author of the new study. “It’s amazing that the models fit so well, and it very indirectly suggests that maybe what the human language system is doing is predicting what’s going to happen next.”

Joshua Tenenbaum, a professor of computational cognitive science at MIT and a member of CBMM and MIT’s Artificial Intelligence Laboratory (CSAIL); and Evelina Fedorenko, the Frederick A. and Carole J. Middleton Career Development Associate Professor of Neuroscience and a member of the McGovern Institute, are the senior authors of the study, which appears this week in the Proceedings of the National Academy of Sciences. Martin Schrimpf, an MIT graduate student who works in CBMM, is the first author of the paper.

Making predictions

The new, high-performing next-word prediction models belong to a class of models called deep neural networks. These networks contain computational “nodes” that form connections of varying strength, and layers that pass information between each other in prescribed ways.

Over the past decade, scientists have used deep neural networks to create models of vision that can recognize objects as well as the primate brain does. Research at MIT has also shown that the underlying function of visual object recognition models matches the organization of the primate visual cortex, even though those computer models were not specifically designed to mimic the brain.

In the new study, the MIT team used a similar approach to compare language-processing centers in the human brain with language-processing models. The researchers analyzed 43 different language models, including several that are optimized for next-word prediction. These include a model called GPT-3 (Generative Pre-trained Transformer 3), which, given a prompt, can generate text similar to what a human would produce. Other models were designed to perform different language tasks, such as filling in a blank in a sentence.

As each model was presented with a string of words, the researchers measured the activity of the nodes that make up the network. They then compared these patterns to activity in the human brain, measured in subjects performing three language tasks: listening to stories, reading sentences one at a time, and reading sentences in which one word is revealed at a time. These human datasets included functional magnetic resonance (fMRI) data and intracranial electrocorticographic measurements taken in people undergoing brain surgery for epilepsy.

They found that the best-performing next-word prediction models had activity patterns that very closely resembled those seen in the human brain. Activity in those same models was also highly correlated with measures of human behavioral measures such as how fast people were able to read the text.

“We found that the models that predict the neural responses well also tend to best predict human behavior responses, in the form of reading times. And then both of these are explained by the model performance on next-word prediction. This triangle really connects everything together,” Schrimpf says.

“A key takeaway from this work is that language processing is a highly constrained problem: The best solutions to it that AI engineers have created end up being similar, as this paper shows, to the solutions found by the evolutionary process that created the human brain. Since the AI network didn’t seek to mimic the brain directly — but does end up looking brain-like — this suggests that, in a sense, a kind of convergent evolution has occurred between AI and nature,” says Daniel Yamins, an assistant professor of psychology and computer science at Stanford University, who was not involved in the study.

Game changer

One of the key computational features of predictive models such as GPT-3 is an element known as a forward one-way predictive transformer. This kind of transformer is able to make predictions of what is going to come next, based on previous sequences. A significant feature of this transformer is that it can make predictions based on a very long prior context (hundreds of words), not just the last few words.

Scientists have not found any brain circuits or learning mechanisms that correspond to this type of processing, Tenenbaum says. However, the new findings are consistent with hypotheses that have been previously proposed that prediction is one of the key functions in language processing, he says.

“One of the challenges of language processing is the real-time aspect of it,” he says. “Language comes in, and you have to keep up with it and be able to make sense of it in real time.”

The researchers now plan to build variants of these language processing models to see how small changes in their architecture affect their performance and their ability to fit human neural data.

“For me, this result has been a game changer,” Fedorenko says. “It’s totally transforming my research program, because I would not have predicted that in my lifetime we would get to these computationally explicit models that capture enough about the brain so that we can actually leverage them in understanding how the brain works.”

The researchers also plan to try to combine these high-performing language models with some computer models Tenenbaum’s lab has previously developed that can perform other kinds of tasks such as constructing perceptual representations of the physical world.

“If we’re able to understand what these language models do and how they can connect to models which do things that are more like perceiving and thinking, then that can give us more integrative models of how things work in the brain,” Tenenbaum says. “This could take us toward better artificial intelligence models, as well as giving us better models of how more of the brain works and how general intelligence emerges, than we’ve had in the past.”

The research was funded by a Takeda Fellowship; the MIT Shoemaker Fellowship; the Semiconductor Research Corporation; the MIT Media Lab Consortia; the MIT Singleton Fellowship; the MIT Presidential Graduate Fellowship; the Friends of the McGovern Institute Fellowship; the MIT Center for Brains, Minds, and Machines, through the National Science Foundation; the National Institutes of Health; MIT’s Department of Brain and Cognitive Sciences; and the McGovern Institute.

Other authors of the paper are Idan Blank PhD ’16 and graduate students Greta Tuckute, Carina Kauf, and Eghbal Hosseini.

Featured image: MIT neuroscientists find the internal workings of next-word prediction models resemble those of language-processing centers in the brain.

---

Provided by MIT

Researchers from Tokyo Medical and Dental University (TMDU), Takeda Pharmaceutical Co., Ltd. and Ionis Pharmaceuticals, USA, show that heteroduplex oligonucleotide drugs conjugated with cholesterol cross the blood–brain barrier effectively with intravenous or subcutaneous dosing 

Watch video on YouTube: https://www.youtube.com/watch?v=5us0bzH4GAk 

Antisense oligonucleotide (ASO) therapy has the potential to ameliorate many neurodegenerative diseases at the genetic level to suppress the production of harmful proteins or non-coding RNAs. Previously, achieving delivery of ASO with adequate concentrations in the central nervous system (CNS) with systemic dosing was difficult. Now, researchers from Japan and the USA have developed a drug delivery platform that overcomes this hurdle. 

Evolution has equipped the brain with protection against both mechanical and molecular injury. The blood–brain barrier (BBB) is a selectively semipermeable barricade of endothelial cells lining the capillaries; working with specific transporter proteins, it functions as a fastidious gatekeeper between the circulation and the CNS, barring foreign molecules, including drugs.

ASOs are pharmaceutical molecules that can target disease at the genetic level. They comprise a few dozen base pairs arranged in an ‘antisense’ or reverse order and prevent production of pathogenic proteins through binding to the ‘sense’ strand of mRNA targets. Single-stranded ASOs show great promise against CNS disorders such as spinal muscular atrophy. However, they do not enter the CNS effectively following systemic administration and require direct intrathecal injection. This may be hazardous particularly for patients with lumbar spinal deformity or on blood-thinners.

The research team had recently developed DNA/RNA heteroduplex oligonucleotide (HDO) technology capable of highly efficient RNA degradation in vivo. First author Tetsuya Nagata explains, “We found that cholesterol conjugated HDO (Chol-HDO), unlike cholesterol-ASO, efficiently reached the CNS following subcutaneous or intravenous administration in experimental animals. The Chol-HDO platform showed significant dose-dependent target gene reductions with prolonged action in all CNS regions and cell types.” 

Further, the researchers confirmed that this beneficial outcome was not at the expense of vascular barrier integrity. They also investigated the pharmacokinetics of multiple injections as well as subcutaneous dosing (which may be self-administered). Additionally, the effects were confirmed across species and against other neurogenerative disease gene targets such as myotonic dystrophy type 1, Alexander disease and amyotrophic lateral sclerosis.

“Systemic doses being higher, adverse effects such as mild decrease in platelets were expected,” says Nagata. “However, divided or subcutaneous dosing can rescue these. We may also strategize by initiating treatment with intrathecal dosing to rapidly achieve therapeutic concentrations, followed by intravenous or subcutaneous maintenance as needed.” 

“Our innovative therapeutic platform for blood-to-brain delivery of ASOs may revolutionize management of neurodegenerative diseases,” senior author Takanori Yokota claims. “Future research will help define the specific molecular pathways thus optimizing delivery of ASO pharmacotherapy to the CNS.”

The article, “Systemically administered DNA/RNA heteroduplex oligonucleotides achieve blood to brain delivery and efficient gene knockdown in the CNS” was published in Nature Biotechnology  at DOI:10.1038/s41587-021-00972-x

Featured image: (A) Drastic target gene (RNA) suppression in the cerebral cortex when Blood-brain-barrier heteroduplex oligonucleotide (BBB-HDO) administered intravenously to mice. (B) RNA images in the cerebral cortex after administration of BBB-HDO to mice. The brown signal indicates RNA, but the signal is almost completely suppressed by BBB-HDO. (Scale: 100 μm). (C) Live image of mouse brain observed by in vivo confocal laser microscopy. The single-stranded antisense oligonucleotide (ASO) remains in the blood vessels of the brain, while the BBB-HDO is directly transferred into the brain. © TMDU

---

Provided by Tokyo Medical and Dental University

The potassium channel KCNQ3 is required for our brain to generate accurate spatial maps. In mice, defects in KCNQ3 function have measurable effects on the internal navigation system. The findings of a research team including researchers from FAU recently published in Nature Communications are also relevant for Alzheimer’s-type dementia research.null

In addition to other physiological processes, potassium is required for muscle and nerve cell excitability. Potassium ions cross the outer cell membrane via a variety of ion channels and thereby generate electrical currents. Twenty years ago, Prof. Thomas Jentsch’s team at the Leibniz Research Institute for Molecular Pharmacology (FMP) in Berlin identified the genes encoding the potassium channel family KCNQ2-5 and demonstrated that mutations in KCNQ2 and KCNQ3 can cause hereditary epilepsy in humans. Pharmaceutical companies were able to develop targeted anti-epileptic drugs as a result of this pioneering research.

Now, teams of molecular biologists led by Thomas Jentsch and neurophysiologists supervised by Alexey Ponomarenko, professor at the Institute of Physiology and Pathophysiology at FAU, together with scientists at the University of Conneticut (USA) and the University of Cologne, have discovered that KCNQ3 may also play a role in Alzheimer’s disease and other cognitive disorders.

Precise navigation maps in the brain
Normally, the transmitter acetylcholine inhibits neuronal potassium flow, which is necessary for cortical excitability and thus for memory and attention. It is well established that Alzheimer’s patients gradually lose this cholinergic neuromodulation.

The current study examined the role of KCNQ3 channels in the neuromodulation of the brain’s navigation system. These place fields, a discovery for which a Nobel Prize was awarded several years ago, serve as an internal map for the brain. “We discovered how various signals generated by place cells under the control of KCNQ3 channels interact with brain rhythms to form precise spatial maps,” says Alexey Ponomarenko.

The knock-out mice with a defective KCNQ3 channel generated by Thomas Jentsch’s group revealed a different picture. Although the activity patterns of place cells in healthy mice followed a specific temporal and spatial pattern, in knock-out mice, the synaptic transmission by single or nearly simultaneous multiple (burst) signals of place cells was disorganized. “When bursts are fired, they typically have a certain rhythm. In the mutants, however, the bursts are not controlled by the rhythm, but are fired at completely random times or phases of the rhythm,” explains Prof. Ponomarenko. “This effectively suppresses single action potentials and creates an imbalance in the activity patterns of place cells.”

Combined with optogenetic experiments, recordings using silicon probes measuring 15 micrometers in thickness implanted in the hippocampus of freely behaving rodents, provided exciting insights into brain function. Additionally, the team members in America demonstrated that the absence of the KCNQ3 channel resulted in a significant decrease in neuronal potassium currents (here M-currents).

“While there is insufficient data to date for clinical applications, our findings suggest that the KCNQ3 channels could be a potential target for future drug research to treat Alzheimer’s and other dementias,” emphasizes Prof. Ponomarenko, “at least in the early stages, when place cells are likely still present but cholinergic neuromodulation has already subsided.” Additional research is required to gain a better understanding of KCNQ3’s role in the brain.

---

Reference: Xiaojie Gao et al, Place fields of single spikes in hippocampus involve Kcnq3 channel-dependent entrainment of complex spike bursts, Nature Communications (2021). DOI: 10.1038/s41467-021-24805-2

---

Provided by Friedrich-Alexander-Universität Erlangen-Nürnberg

OHSU study reveals that synthetic compound regulates gene implicated in neurodegenerative diseases

Researchers from Oregon Health & Science University have for the first time demonstrated it’s possible to use a synthetic thyroid hormone to regulate a gene implicated in neurodegenerative diseases like Alzheimer’s, Parkinson’s and multiple sclerosis.

The findings from tests in cells and mice, published today in the journal Cell Chemical Biology, raise the possibility of development of new medication to treat debilitating diseases.

“This is the first example reported that shows it’s possible to increase the expression of the TREM2 gene in a way that will lead to healing in certain diseases,” said senior author Tom Scanlan, Ph.D., professor of physiology and pharmacology in the OHSU School of Medicine. “This will generate a lot of excitement.”

The paper’s first author is Skylar J. Ferrara, Ph.D., a postdoctoral fellow in the OHSU School of Medicine’s chemical physiology and biochemistry department. 

The discovery builds on a 2013 publication linking genetic variants of TREM2 to risk of Alzheimer’s disease.

The new research from OHSU builds on that work by showing that it’s possible to turn on TREM2 expression and the TREM2 pathway using a compound originally developed more than two decades ago to lower cholesterol.

Researchers administered an analog of the compound that penetrates into the central nervous system of mice. They discovered they were able to increase the expression of TREM2 and reduce damage to myelin. Myelin is the insulation-like protective sheath covering nerve fibers that’s damaged in disorders like multiple sclerosis.

The pathway activated by the TREM2 gene is also implicated in neurodegenerative diseases, including Alzheimer’s and Parkinson’s.

“TREM2 is a receptor,” Scanlan said. “It senses damaged cellular debris from disease and responds in a healing, productive way. The thought is, if you can simply turn up its expression, then that’s going to lead to a therapeutic effect in most neurodegenerative diseases.”

Joseph Quinn, M.D., professor of neurology in the OHSU School of Medicine, who treats patients with Parkinson’s and Alzheimer’s, said the findings are promising. Quinn wasn’t involved in the research.

“TREM2 is a viable ‘target’ for treatment in Alzheimer’s disease, based on genetics and other studies,” Quinn said. “This new report has important implications for testing a new therapeutic approach for Alzheimer’s, including raising the potential for developing a new medication to regulate TREM2.”

The synthetic thyroid hormone compound, known as sobetirome and similar analogs, is already licensed by an OHSU spinoff company to conduct clinical trials for central nervous system diseases, including multiple sclerosis. In contrast to other basic science discoveries in mice, Scanlan said this latest discovery connects this class of compounds to Alzheimer’s, Parkinson’s and other neurodegenerative diseases, advancing the science that much closer to clinical trials in people with debilitating disease.  

“The possibility of doing clinical trials is not millions of miles away,” Scanlan said. “It would be an achievable thing.”

REFERENCE: Skylar J. Ferrara, Priya Chaudhary, Margaret J. DeBell, Gail Marracci, Hannah Miller, Evan Calkins, Edvinas Pocius, Brooke A. Napier, Ben Emergy, Dennis Bourdette, Thomas S. Scanlan, TREM2 is thyroid hormone regulated making the TREM2 pathway druggable with ligands for thyroid hormone receptor, Cell Chemical Biology, Aug. 9, 2021, DOI: https://doi.org/10.1016/j.chembiol.2021.07.014

This research was supported by the National Institutes of Health, awards DK52798 and GM133804; the National Multiple Sclerosis Society, awards RG5199A4 and RG 1607-25053, RG5106A1/1 and RG 2001-35775; the Race to Erase MS, and the OHSU Laura Fund for Innovation in Multiple Sclerosis.

Featured image: Researchers from Oregon Health & Science University have for the first time demonstrated it’s possible to use a synthetic thyroid hormone to regulate a gene implicated in neurodegenerative diseases like Alzheimer’s, Parkinson’s and multiple sclerosis. (Getty Images)

---

Provided by OHSU

Brain Signals That Help Memories Form May Influence Blood Sugar

A set of brain signals known to help memories form may also influence blood sugar levels, finds a new study in rats.

Researchers at NYU Grossman School of Medicine discovered that a peculiar signaling pattern in the brain region called the hippocampus, linked by past studies to memory formation, also influences metabolism, the process by which dietary nutrients are converted into blood sugar (glucose) and supplied to cells as an energy source.

The study revolves around brain cells called neurons that “fire” (generate electrical pulses) to pass on messages. Researchers in recent years discovered that populations of hippocampal neurons fire within milliseconds of each other in cycles. The firing pattern is called a “sharp wave ripple” for the shape it takes when captured graphically by EEG, a technology that records brain activity with electrodes.

Published online in Nature on August 11, the new study found that clusters of hippocampal sharp wave ripples were reliably followed within minutes by decreases in blood sugar levels in the bodies of rats. While the details need to be confirmed, the findings suggest that the ripples may regulate the timing of the release of hormones, possibly including insulin, by the pancreas and liver, as well of other hormones by the pituitary gland.

“Our study is the first to show how clusters of brain cell firing in the hippocampus may directly regulate metabolism,” says senior study author György Buzsáki, MD, PhD, the Biggs Professor of Neuroscience in the Department of Neuroscience and Physiology at NYU Langone Health.

“We are not saying that the hippocampus is the only player in this process, but that the brain may have a say in it through sharp wave ripples,” says Dr. Buzsáki, also a faculty member in the Neuroscience Institute at NYU Langone.

Known to keep blood sugar at normal levels, insulin is released by pancreatic cells, not continually, but periodically in bursts. As sharp wave ripples mostly occur during non-rapid eye movement (NREM) sleep, the impact of sleep disturbance on sharp wave ripples may provide a mechanistic link between poor sleep and high blood sugar levels seen in type 2 diabetes, say the study authors.

Previous work by Dr. Buzsáki’s team had suggested that the sharp wave ripples are involved in permanently storing each day’s memories the same night during NREM sleep, and his 2019 study found that rats learned faster to navigate a maze when ripples were experimentally prolonged.

“Evidence suggests that the brain evolved, for reasons of efficiency, to use the same signals to achieve two very different functions in terms of memory and hormonal regulation,” says corresponding study author David Tingley, PhD, a postdoctoral scholar in Dr. Buzsáki’s lab.

Dual Role

The hippocampus is a good candidate brain region for multiple roles, say the researchers, because of its wiring to other brain regions, and because hippocampal neurons have many surface proteins (receptors) sensitive to hormone levels, so they can adjust their activity as part of feedback loops. The new findings suggest that hippocampal ripples reduce blood glucose levels as part of such a loop.

“Animals could have first developed a system to control hormone release in rhythmic cycles, but then applied the same mechanism to memory when they later developed a more complex brain,” adds Dr. Tingley.

The study data also suggest that hippocampal sharp wave ripple signals are conveyed to hypothalamus, which is known to innervate and influence the pancreas and liver, but through an intermediate brain structure called the lateral septum. Researchers found that ripples may influence the lateral septum just by amplitude (the degree to which hippocampal neurons fire at once), not by the order in which the ripples are combined, which may encode memories as their signals reach the cortex.

In line with this theory, short-duration ripples that occurred in clusters of more than 30 per minute, as seen during NREM sleep, induced a decrease in peripheral glucose levels several times larger than isolated ripples. Importantly, silencing the lateral septum eliminated the impact of hippocampal sharp wave ripples on peripheral glucose.

To confirm that hippocampal firing patterns caused the glucose level decrease, the team used a technology called optogenetics to artificially induce ripples by re-engineering hippocampal cells to include light-sensitive channels. Shining light on such cells through glass fibers induces ripples independent of the rat’s behavior or brain state (e.g., resting or waking). Similar to their natural counterparts, the synthetic ripples reduced sugar levels.

Moving forward, the research team will seek to extend its theory that several hormones could be affected by nightly sharp wave ripples, including through work in human patients. Future research may also reveal devices or therapies that can adjust ripples to lower blood sugar and improve memory, says Dr. Buzsáki.

Along with Dr. Tingley and Dr. Buzsáki, study authors were Ekin Kaya, Kathryn McClain, and Jordan Carpenter at NYU Langone Health. The work was funded by National Institutes of Health grants MH122391, U19 NS104590, and U19 NS107616.

Featured image: ROST-9D/GETTY

---

Reference: Tingley, D., McClain, K., Kaya, E. et al. A metabolic function of the hippocampal sharp wave-ripple. Nature (2021). https://doi.org/10.1038/s41586-021-03811-w

---

Provided by NYU Langone

A new study on bats reveals an unexpected representation of three-dimensional space in the brain

In a new study published in Nature today, Weizmann Institute of Science researchers, in collaboration with colleagues from the Hebrew University of Jerusalem, unveiled for the first time how three-dimensional space is represented in the mammalian cortex by the brain’s “GPS” system. The team of researchers, led by Prof. Nachum Ulanovsky of Weizmann’s Neurobiology Department, were surprised to find that this representation is very different from the way in which two-dimensional space is represented, turning several long-standing hypotheses on their heads.

Mammals, including humans, know their position in space, owing to several types of specialized neurons in the hippocampus and its next-door neighbor the entorhinal cortex – regions located deep inside the brain. Head-direction cells, the internal compasses of the brain, indicate to the animal the direction in which its head is turned. Place cells, thought to construct a mental map of the environment, are activated when an animal crosses a specific location. Grid cells, by contrast, respond not to one, but to multiple such locations, and they are thought to provide the brain with a GPS system of sorts.

The study of grid cells and the brain’s GPS was awarded the Nobel Prize in 2014. However, these and other studies focused solely on how two dimensions are represented and said very little about the representation of three-dimensional space. To bridge this gap, Ulanovsky and colleagues set out to elucidate how grid cells act in three dimensions in freely behaving bats.

In the past, when grid cells were studied in rodents running on two-dimensional surfaces, they were found to be activated in multiple circular areas, known as firing fields, which are arranged in a symmetrical hexagonal pattern – resembling millimeter graph paper – that tiles the surface. This unparalleled symmetry and periodicity suggest that these cells may be involved in geometric spatial computations that form the core of the cerebral GPS. The entorhinal cortex, where grid cells are located, is the brain area that is first affected in Alzheimer’s disease, and it is possible that spatial disorientation, one of the early manifestations of Alzheimer’s, is due to the grid cells’ dysfunction – and the loss of the hexagonal “millimeter paper” of grid cells.

“The well-ordered global grid that is the hallmark of their two-dimensional activity was altogether gone”

Mathematically, the optimal way to pack circles in two dimensions is in a hexagonal pattern, like a honeycomb: This is possibly the reason why the circular firing fields of grid cells are represented in the brain in a hexagonal lattice when animals walk over two-dimensional surfaces. Therefore, the researchers expected the activity pattern in three dimensions to be similarly symmetrical and hexagonal. “We and many other researchers hypothesized that we’d see hexagonally stacked balls, like oranges in a grocery store neatly stacked in a pyramid, or any other extremely ordered three-dimensional arrangement,” Ulanovsky says.

To test this hypothesis, the researchers, led by doctoral student Gily Ginosar, together with Staff Scientist Dr. Liora Las, recorded the activity of grid cells in bats that had small mobile devices mounted on their heads, as the bats were flying around a room the size of a large living room. Feeding stations at different heights ensured that each bat covered most of the room’s volume in every run. Once the data started coming in, the researchers saw that grid cells did not behave as expected when responding to three-dimensional coordinates. “The well-ordered global grid that is the hallmark of their two-dimensional activity was altogether gone,” explains Ulanovsky.

Instead, the three-dimensional firing fields of the grid cells, shaped in this case as spheres rather than circles, were packed like a box full of marbles. They were not completely disordered, but were certainly less organized than the three-dimensional equivalent of a hexagonal lattice – as the new arrangement allowed the “marbles” some extra degrees of freedom. Whereas any noticeable global order was lacking, the spheres did commit to a local order wherein the distance between one sphere and its nearest neighbors remained constant.

Compared to other longstanding theories, the new theoreitcal model was the most loyal to the experimental data

To offer a mechanistic explanation of this phenomenon of local rather than global order, the experimental team – Ginosar, Las and Ulanovsky – collaborated with theoreticians Dr. Johnatan Aljadeff, a former postdoctoral fellow at Weizmann and now a professor at the University of California in San Diego, and Prof. Haim Sompolinsky and Prof. Yoram Burak from the Hebrew University of Jerusalem. Together they constructed a model that uses principles, borrowed from statistical physics, that describe the interaction between particles. The model revealed that the spherical firing fields of grid cells seem to interact in almost the same way as particles do – they are “attracted” to one another when at a distance and are “repelled” once they get too close. In particular, the balance of forces acting on particles could explain the local order that kept the spheres at constant local distances from one another, while avoiding any global lattice. Compared to other models that were used in the past to predict the three-dimensional organization of grid cells’ firing fields, the new model was the most loyal to the experimental data.

Taken together, the surprising experimental data and theoretical model offer a new way of looking at the neural basis of three-dimensional navigation and the role that grid cells play in this cognitive process. While previous models extrapolated a similar three-dimensional arrangement from the two-dimensional grid, the work of Ulanovsky and colleagues and their “box of marbles” model show that things are much more complex. Since no periodic lattice is formed in three-dimensional space, the classical theories for understanding the intriguing behavior of grid cells will need to be revised.

Featured image: The Egyptian fruit bat: The representation of three dimensional space in the mammalian cortex resembles a box of marbles. Photo: Steve Gettle; Design: Maayan Visuals

---

Reference: Ginosar, G., Aljadeff, J., Burak, Y. et al. Locally ordered representation of 3D space in the entorhinal cortex. Nature (2021). https://doi.org/10.1038/s41586-021-03783-x

---

Provided by Weizmann Institute of Science

Hormones released after a brain injury contribute to movement problems on the left and right sides of the body, scientists from Uppsala University and elsewhere can now show in a new study in rats. The results also suggest that hormone-blocking treatments may help counteract these effects, a finding that has implications for treating people with traumatic brain injuries or stroke. The study has been published in eLife.

A stroke or injury to one side of the brain causes movement difficulties on the opposite side of the body. Scientists have previously thought that this is because nerves from one side of the brain control activity on the opposite side. But recent studies have shown that giving rats without a brain injury certain hormones can cause movement responses similar to human motor deficits on one side of the body. 

“This led us to ask whether pituitary hormones might mediate in part the side-specific movement problems humans can experience after brain injury,” explains Georgy Bakalkin, Professor at the Department of Pharmaceutical Biosciences, Uppsala University, Sweden, and a co-senior author of the study.

To investigate further, Bakalkin and the team examined the effects of a one-sided brain injury in rats that lacked the connection between the brain and nerves that regulate the hindlimbs. They found that, even without this connection, the hindlimb on the opposite side to the injury had impaired reflexes.

However, animals that lacked the pituitary gland, a hormone-producing gland connected to the brain, did not experience these problems. Two pituitary hormones ß-endorphin and Arg-vasopressin appeared to play a role. When the team gave rats without a brain injury these two hormones, the rats also developed hindlimb contraction on the right side.

Next, they tested what would happen if they gave the rats with a left-sided brain injury drugs that block the effects of these two hormones. They found that the animals did not develop right-sided movement problems. This suggests that the hormones convey side-specific signals after a brain injury and treating patients who have a similar injury with drugs that block the effects of these hormones might be beneficial.

“These observations suggest that the endocrine system through its hormones in the blood may selectively target the left and right sides of the animals’ bodies,” Bakalkin concludes. “This is an unusual phenomenon that requires further studies and verification in other animal models. We must be cautious in the interpretation of these findings and their biological implications before further research is carried out. But if future studies confirm the benefits of treatments that block these hormones, they may offer a new approach to treating movement problems following stroke or injury. Now having this published we could proceed with analysis of underlying mechanisms and a role of this phenomenon in control of our body plan and in neurological disorders”.

Georgy Bakalkin and Jens Schouenborg (Lund University, Sweden) are co-senior authors of the study. The team also includes co-first authors Nikolay Lukoyanov (University of Porto, Portugal), Hiroyuki Watanabe (Uppsala University, Sweden), Liliana Carvalho (University of Porto), and Olga Nosova and Daniil Sarkisyan (both Uppsala University), as well as Mengliang Zhang and Marlene Storm Andersen (University of Southern Denmark), Elena A. Lukoyanova (University of Porto), Vladimir Galatenko (Lomonosov Moscow State University, Russia), Alex Tonevitsky (National Research University Higher School of Economics, Russia), and Igor Bazov and Tatiana Iakovleva (both Uppsala University).

---

Reference: Lukoyanov, et al., Left-right side-specific endocrine signaling complements neural pathways to mediate acute asymmetric effects of brain injury, eLife 2021;10:e65247. DOI: 10.7554/eLife.65247

---

Provided by Uppsala University

Watson College assistant professor of mechanical engineering to lead $587,853 NSF project

Many mysteries continue to surround the human brain, but among the most important are how it forms and how those early weeks affect the rest of a person’s life.

Upcoming research from Binghamton University and Harvard Medical School will use computer modeling and advanced imaging of developing fetal brains to try to answer some of those longstanding questions.

The National Science Foundation’s Biomechanics and Mechanobiology Program recently approved a $587,853 grant to better understand the growth and folding that make each human brain unique.

Leading that study will be Assistant Professor Mir Jalil Razavi from the Thomas J. Watson College of Engineering and Applied Science’s Department of Mechanical Engineering. Co-principal investigator will be Assistant Professor Weiying Dai from Watson’s Department of Computer Science, with Harvard Associate Professor Ali Gholipour as a research partner.

Razavi first became interested in brain research in 2014, when earning his PhD at the University of Georgia. His advisor was working on molecular dynamics modeling to predict the physical movements of atoms and molecules, but wanted to branch out by studying the mechanical modeling of soft tissue.

“One of the important topics with soft biological tissue is the brain,” Razavi said. “It’s a mystery for us how our brains start from the smoothest state at 22 to 25 weeks after gestation, but within a few weeks there is expansion in the surface area and volume as well as brain folding.”

When he arrived at Binghamton in 2018, Razavi focused on the mechanics of human skin, doing research with Associate Professor Guy German from Watson College’s Department of Biomedical Engineering.

With this NSF project, he hopes to chart the formation of brain folds as faster-growing grey matter (the outer layer where higher-level thinking is done) grows on top of white matter (the inner layer that communicates between different gray matter areas and between the gray matter and the rest of the body).

There are about 100 billion neurons in a human brain, and information is transmitted through a complex network of axonal fibers that would stretch the entire 239,000 miles from the Earth to the Moon if connected end to end. Hardwiring for most of this network happens before birth, as growing brains form folds that connect the neurons in random yet significant ways.

“We don’t understand the underlying mechanism from the biological view, but we can say from the mechanical view that we have folds because we have mismatch in the growth rate of the layers,” Razavi said. “This is not just about the brain. If we have a multilayer system and the outer layer grows faster than the inner layers, then we will have instability and folding.”

Gholipour — Harvard Medical School’s director of translational radiology research — took scans of 50 fetuses at 25 weeks and 36 weeks using standard magnetic resonance imaging (MRI) as well as brain-specific diffusion tensor imaging (DTI), which maps the diffusion process of molecules within cells.

“Professor Gholipour and his team have the most precise MRI and DTI for the fetal brain,” Razavi said. “It is really difficult to get those images because the nature of pregnancy means fetuses are in motion. After birth, people can be put into a static condition.”

Dai, who conducts her own brain-related research, will offer her expertise to parse through those results, he added: “We should find a mutual language between the imaging of the fetal brain and the mechanical model. She will help us to process those data to create our mechanical models.”

How those 50 fetal brains grew and folded then will be compared to Razavi’s computer model to see whether the expected patterns match the actual ones.

“If we are precise, the results should be close to each other,” he said. “Otherwise, we will say: ‘What’s the problem? What kind of other factors haven’t we included in our models?’ We have included a lot of data, but we don’t know what will happen.”

Razavi sees this study as a beginning step to understanding some brain disorders, such as autism, schizophrenia and polymicrogyria (in which the surface of the brain has many ridges or folds). From there, he potentially sees a lifetime of possible brain-related research avenues that could branch out.

“I think we will need a very long time to decipher the mystery of the brain, because it’s not comparable with the other organs at all. It’s completely different and very complex,” he said.

Related brain research

In addition to the NSF grant, Razavi also received two recent seed grants through Binghamton University’s Transdisciplinary Areas of Excellence, which encourages researchers to collaborate outside of their usual field of study.

A $15,000 grant with Assistant Professor Guifang Fu from the Mathematical Sciences Department at Binghamton’s Harpur College of Arts and Sciences will offer more expertise when examining the morphology of the brain during early stages of growth.

A second $15,000 grant with German seeks to better understand the mechanical properties of the axons in the inner white matter of the brain. The results could aid treatment for neurodevelopmental disorders and traumatic brain injuries.

Featured image: New research at Binghamton University looks at how brain folds form. © Binghamton University

---

Provided by Binghamton University