• Neuronal Connectivity in Brain Function and Disease

    Neuronal Connectivity in Brain Function and Disease

    Organizers: Thomas F. Franke (NYU School of Medicine), Eric Nestler (Icahn School of Medicine at Mount Sinai), Sonya Dougal (The New York Academy of Sciences), and Caitlin McOmish (The New York Academy of Sciences)Presented by the Biochemical Pharmacology Discussion Group
    Reported by Evguenia Alexandrova | Posted June 2, 2016


    The brain is a puzzle of remarkable genetic, structural, and functional complexity. Projects that aim to map the brain and define its circuitry are complemented by advances in our understanding of how development and experience shape the connectivity underlying its functions. On March 22, 2016, the Academy's Biochemical Pharmacology Discussion Group presented Neuronal Connectivity in Brain Function and Disease: Novel Mechanisms and Therapeutic Targets to explore both the importance of structural synaptic plasticity in shaping neuronal connections and the technical advances that enable functional analysis of the brain's ultrastructure and circuitry. Mapping brain circuitry and its relationship to behavior elucidates the formation of thoughts and emotions as functional outputs of brain connectivity. As biomedical and computational technologies expand, so too does research on neuronal connectivity, promising unprecedented insights into the brain and its disorders and new options for diagnosis and therapy.

    Bruce McEwen of the Rockefeller University described how neural circuits are shaped by stress hormones, with moderate stress improving focus and memory but too much stress having the opposite effect. Neural plasticity is the result of neurogenesis combined with growth and shrinkage of dendrites and synapses, modulated by experience. During acute stress, adrenal steroids and excitatory amino acids, such as glutamate, improve synaptic function and memory. But repeated or intense stress suppresses neurogenesis, neuronal excitability, and memory. After serious challenges, such as seizure, stroke, or head trauma, adrenal steroids and glutamate can cause irreversible damage. In the aging brain, glutamate, free radicals, and inflammation are involved in Alzheimer's disease.

    Glutamate mediates stress-related plasticity via dendritic remodeling. Hippocampal CA3 neurons are particularly vulnerable to damage from stress but can sometimes recover. Indeed, reversible shrinkage of apical dendrites of CA3 neurons after repeated stress may be protective. A dramatic example of such CA3 plasticity is found in the hibernating European hamster, which has apical CA3 dendrites that atrophy during hibernation and rapidly grow back afterwards. Glutamate is regulated by stimulatory glucocorticoids and suppressed by metabotropic glutamate receptors (mGlu2) in presynaptic terminals, which prevent an overflow of glutamate that is known to contribute to depression, neurodegeneration, and brain aging. The amygdala and prefrontal cortex turn the hippocampal stress response on and off, respectively. "Whatever is happening in the circuitry ... is going to alter the balance between the areas," McEwen said. Most stress-related plasticity occurs in the hippocampus, and structural changes in this area are not always a sign of damage but sometimes show resilience.

    The human hippocampus under stress: different experiences and activities increase and decrease hippocampus size. (Image courtesy of Bruce McEwen)

    Stress vulnerability differs among individuals. Anxious animals respond to stress with downregulated mGlu2, leading to glutamate overflow and depression, and are vulnerable to social defeat and inflammation. (Epigenetic de-repression of mGlu2 by naturally produced L-acetylcarnitine or by histone deacetylase inhibitors has antidepressant effects.) Chronically stressed animals show permanent changes in hippocampal gene expression, revealed in gene expression signatures after a one-hour swim test. Animals recovering from chronic restraint stress (CRS) are sensitized to acute novel stress and show a distinct gene expression signature after the swim test. This result suggests that a history of stress is forever imprinted in the hippocampus. "As time goes by, we cannot roll back the clock—the brain becomes continually different," McEwen said.

    Depression and posttraumatic stress disorder (PTSD) can cause hippocampal shrinkage in humans. Small hippocampus, a predisposing factor for PSTD, is found in airplane personnel with chronic jet lag and associated with both lack of exercise and chronic inflammation, in diseases such as type 2 diabetes. Exercise, learning, and antidepressant medications can increase hippocampus volume. In one study, elderly sedentary adults showed hippocampus enlargement and had improved memory and prefrontal cortex function after completing one hour of exercise 5 days a week for 6–12 months. "We are never too old to benefit from physical activity," McEwen said.

    Bernardo Sabatini of Harvard Medical School described corticostriatal projections and their role in disease. The striatum is the input area of basal ganglia, evolutionarily conserved subcortical nuclei involved in Parkinson's and Huntington's diseases, drug addiction, and other conditions that feature an altered relationship between action and reward. Basal ganglia are also implicated in autism spectrum disorders (ASD). The striatum has a distinctive structure of recurrent loops and cascading inhibition, receiving excitatory glutamatergic stimuli from the cortex and sending inhibitory outputs to the substantia nigra, thalamus, and globus pallidos. The substantia nigra compacta, which promotes reward-oriented behavior, provides dopaminergic innervation to the striatum.

    The cortex–striatum–cortex circuit: cortex sends excitatory stimuli to striatum, while striatum activates cortex via a direct pathway and inhibits it via an indirect pathway. (Image courtesy of Bernardo Sabatini)

    "Ongoing activity in cortex drives synapse formation in the striatum," Sabatini said, recounting research into the development of striatum topography. For example, mouse striatum shows tremendous growth during the second postnatal week, when repeated glutamate release at postnatal day 10 (P10) triggers new dendritic spine growth in cortical pyramidal neurons. These findings suggest that glutamate from axons of L5 pyramidal neurons drives synapse formation downstream, in the striatum. Indeed, moderate optogenetic activation at P10 enhanced synaptic responses in the striatum, while chemogenetic inhibition of corticostriatal neurons led to a 50% loss of striatal synapses. Stimulation of the substantia nigra compacta also promoted striatum development in young animals—suggesting that behaviors such as successful suckling promote structural changes in the striatum that then reinforce the behaviors. Stratum development thus depends on passive input via the cortex and on context-dependent input via dopamine from the substantia nigra compacta.

    To investigate repetitive behaviors, such as stereotypy, obsessions, and compulsions, Sabatini's lab turned to the Shank3 knockout (KO) mouse model of repetitive behavior. Shank3 is highly expressed in postsynaptic density and a major component of corticostriatal glutamatergic synapses. The Shank3 gene maps to the chromosomal region that is deleted or truncated in Phelan-McDermid syndrome, which features symptoms of ASD. Shank3 KO mice have weakened striatal synapses at P60 but also show a hyper-grooming phenotype that causes them to develop skin lesions. The researchers would have expected this phenotype to result from stronger, not weaker, synapses. "Shank3 KO mice don't quite make sense," Sabatini observed. Indeed, optogenetic activation of L5 cortical neurons at P14 resulted in enhanced synapse formation in the striatum and enhanced responses in Shank3 KO mice compared to normal animals. Surprisingly, the researchers also found enhanced cortical activity in young Shank3 KO mice and confirmed that it directly affects hyperactive striatum. These findings highlight important differences in young and adult Shank3 KO mice, showing that activity of corticostriatal neurons guides early development of striatum, resulting in aberrant adult behavior.

    "In humans, a much wider behavioral repertoire and circuit-based reinforcement can capture aberrant behaviors," Sabatini concluded. Early experience could reinforce a particular cortex–striatum–cortex loop, which could be strong even if the overall striatum is weak.

    Amy Robinson, executive director of EyeWire, presented her company's "citizen neuroscience" project to engage the public in building 3D models of retinal neurons and their connections. The project started after computational neuroscience researchers at Massachusetts Institute of Technology realized they needed help to reconstruct neurons from serial electron microscopic images, despite designing software to facilitate the process. "While AI is dramatically improving the rate at which we are able to analyze the data, form new hypotheses, and test them," Robinson explained, "it's still excruciatingly slow—dozens of hours for a single neuron."

    Drawing inspiration from crowdsourced projects like Foldit, a game in which players assemble 3D protein conformations, the team created EyeWire, "a game to map the brain." The interface features 4.5 cubic micron–sized sections of brain, and players are tasked with mapping a neuron branch from one side of the cube to the other. Some cubes have a piece of a neuron branch assembled by AI, and the challenge is to color the missing pieces to eventually obtain a complete 3D image of a cell. "The most difficult [parts] to properly color and interpret are little knobs, but they are actually important to find synapses," Robinson said.

    The EyeWire game interface. (Image courtesy of Amy Robinson)

    The software features checks and balances to ensure that neurons are mapped correctly: players check one another's work, and advanced players called "scouts" and "scythes" search for imperfect cubes, then flag and fix them. Another level of accuracy comes from "game masters"—expert neuron tracers and community managers. EyeWire's quarter of a million players have mapped more than 700 neurons. Robinson ended her presentation with a spectacular video showing a small area of retina packed with about 120 reconstructed neurons with a tangled web of projections. "It gives you a sense of the beauty and the complexity that we are going to see when eventually we start taking these techniques to the brain, which is where we are going next," she said.

    Retinal neurons fully reconstructed and assembled by EyeWire players. (Image courtesy of Amy Robinson)

    Wenbiao Gan of NYU Langone Medical Center explained the importance of sleep for maintaining new neuronal connections. His lab explores the synaptic basis of memory and the mechanisms by which synapses, highly dynamic structures, hold information for a lifetime. In one experiment, mice on an accelerating rotating rod must learn to keep up with its changing speed, and researchers use two-photon microscopy to count the number of dendritic spines on L5 pyramidal neurons before and after training. After 2 days of training, young and adult mice had double the number of new dendritic spines compared to control animals. New spines formation increased again when the direction of rotation reversed. "Any time animals learn something new, they form new connections," Gan said. Of the new spines, 15%–20% persisted for 5 days, and a small percentage persisted for 16 months. The researchers predicted that 0.8%, about 2 million new synapses, would remain for life, maintaining the new skill. Further testing supported this idea: animals trained at 2 months old showed better performance at 4 months old compared to untrained animals.

    After noticing that animals slept more after training, the researchers tested the effects of sleep deprivation on spine formation and learning. When animals were prevented from sleeping for 7 hours after training, new synapse formation was severely reduced, and even a second training session did not restore it to normal levels. "This suggests that it's much better to go to sleep after studying than to keep studying without sleep," Gan said. Intriguingly, NREM (deep) sleep is important for new synapse formation, but REM (shallow) sleep is not. Moreover, the same motor neurons active during training remain active ("awake") during subsequent NREM sleep, which may be a sign of new synapse formation. Gan's work shows that experience promotes the formation of synapses that then undergo selective stabilization. "Since only very small percentage [of new synapses are] integrated, they probably do not disrupt existing circuits too much, and therefore do not disrupt existing memory storage, but are sufficient to generate new memory," he said.

    David Sulzer of Columbia University Medical Center proposed a new cellular mechanism of ASD: deficient autophagy, resulting in excessive neuronal connections. No single cause of ASD has yet been found, and a unifying mechanism could change the diagnosis and management of these disorders. Autophagy ("self-eating"), an important means of synapse elimination, is a process of cellular degradation by lysosomes. Synapse remodeling involves macroautophagy, the main autophagy subtype, which proceeds via formation of membranous vacuoles, as well as autophagy mediated by chaperones, endosomes, and multilamellar bodies.

    Molecular regulation of synapse growth (new protein) and shrinkage (autophagy) by mTOR. (Image courtesy of David Sulzer)

    Synapse formation is regulated by growth factors such as BDNF, which activate the master regulator of cell growth and shrinkage, mTORC1. Much work in ASD focuses on growth regulation by mTORC1, but mTORC1 is also an important suppressor of autophagy. Sulzer's lab found that rapamycin, an mTOR inhibitor and potent autophagy activator, promotes autophagy in dopaminergic neurons of striatum, resulting in significant reduction of synaptic vesicle formation and lower dopamine release. In contrast, deficient autophagy results in larger presynaptic terminals and, in the developing brain, faster presynaptic recovery.

    Among disorders with autophagy defects, those with defects in the mTOR pathway tend to be neurodevelopmental disorders. For example, mutations in TSC1/2, PTEN, and NF1 genes, which promote benign tumors and epilepsy, are associated with autism predisposition. To test the role of autophagy in ASD, Sulzer's lab analyzed postmortem brains of preadolescent (2–9 year old) and adolescent (12–19 year old) children, with and without ASD. Focusing on L5 pyramidal neurons and the area of the brain involved in speech, the researchers found dramatic—about 50%—pruning of dendritic spines in the normal adolescent brain compared to the normal preadolescent brain, and significantly reduced pruning associated with ASD.

    Decreased dendritic spine pruning in adolescents with ASD. C, control; A, ASD. (Image courtesy of David Sulzer)

    Markers such as PSD95 and synapsin I decrease simultaneously from pre- to postadolescence in normal children but decrease much less in children with ASD. The change in these markers correlates with increased mTOR pathway activation and decreased autophagy in adolescents with ASD. Interestingly, these differences are eliminated by age 45.

    Autism is a disease of many genes, but it is remarkable, Sulzer said, that the majority of ASD brains had defects in autophagy and mTOR pathways and reduced dendritic pruning. In mice, pruning of dendritic spines at 21–28 days old is also regulated by autophagy—and suppression of autophagy inhibits pruning. Moreover, in a mouse model of ASD, autistic-like behaviors such as decreased sociability can be reversed by rapamycin. Defective autophagy "certainly is not going to show up in every case of autism and is not a perfect biomarker, but it might be pretty nice for understanding particular—perhaps a very broad—group of people with autism," Sulzer said. "Maybe this is the way we can understand [autism] biology and develop new treatments."

    Eric Nestler from the Icahn School of Medicine at Mount Sinai described how neuronal connectivity is changed by cocaine addiction. His lab studies medium spiny neurons (MSNs) in the nucleus accumbens (NAc), a reward region in the brain. MSNs lack spontaneous activity and require glutamatergic inputs from prefrontal cortex, hippocampus, amygdala, and other regions. In cocaine-addicted animals, withdrawal increases the density of thin dendritic spines after 24 hours and the density of thicker mushroom spines after 2–4 weeks in these neurons. These morphologies generally correlate with immature and mature synapses, respectively, but whether this is true in MSNs is unknown.

    Transcription factors, chromatin remodeling proteins, and actin regulators have been implicated in synaptic plasticity during cocaine withdrawal. The researchers investigated how and whether these molecules are interconnected and whether they regulate thin and mushroom spines. They identified a pathway that controls both types of spines bidirectionally, via PDZ-RhoGEF, using an open-ended in silico screen for targets of miR-132/212, which is known to reduce cocaine-rewarding effects in dorsal striatum. The microRNA does not alter the levels of PDZ-RhoGEF mRNA but causes its translocation from the cytoplasm to the nucleus, where it activates the small GTPase RhoA. In turn, RhoA promotes actin polymerization, which activates transcription factors MAL and SRF.

    Another unbiased screen for genes simultaneously controlled by the MAL/SRF complex (from a published database on non-neural cells) and by cocaine in the NAc revealed a small GTPase, Rap1b, as a top hit. "It was an interesting hit, because studies of Rap1b in hippocampal and cortical neurons showed its direct role in dendritic spines," Nestler said, adding that "one interesting theme is that nuclear actin controls synaptic actin [in this system]." Analysis confirmed that cocaine withdrawal induces Rap1b via PDZ-RhoGEF, RhoA, and MAL/SRF during thin spine formation 24 hours post-cocaine.

    Regulation of thin and mushroom dendritic spines by Rap1b at different times after cocaine withdrawal. (Image courtesy of Eric Nestler. Source: Cahill, et al. Neuron. 2016.)

    Surprisingly, Rap1b knockout also increased formation of thicker mushroom spines in both cocaine-addicted and non-cocaine-addicted mice. Indeed, Rap1b reduction at 3–4 weeks post-cocaine in addicted mice potently activated the formation of mushroom spines. Downstream, Rap1b activates PI3K, leading to Akt–TSC1/2–mTOR pathway activation, known to control the translation of several synaptic proteins in dendritic spines. The Akt–mTOR pathway is broadly activated at 24 hours and suppressed at 3 weeks post-cocaine.

    How do these findings translate into behavioral responses to cocaine? While Rap1b overexpression increases animals' behavioral responses to sub-threshold doses of cocaine, its genetic and pharmacological inhibition decreases those behaviors. Nestler's lab also found, using optogenetic activation, that the critical region for MSN remodeling during cocaine withdrawal is the prefrontal cortex, not the hippocampus or the amygdala. A better understanding of the molecular pathways in cocaine-dependent dendrite remodeling could provide insights into the treatment of cocaine addiction.

    Use the tabs above to find multimedia from this event.


    Presentations available from:
    Bruce McEwen, PhD (The Rockefeller University)
    Eric Nestler, MD, PhD (Icahn School of Medicine at Mount Sinai)
    Amy Robinson (Massachusetts Institute of Technology)
    Bernardo Sabatini, MD, PhD (Harvard Medical School)
    David Sulzer, PhD (Columbia University Medical Center)

    The Biochemical Pharmacology Discussion Group is proudly supported by

    • Boehringer Ingelheim
    • Pfizer


    American Chemical Society

    How to cite this eBriefing


    The New York Academy of Sciences. Neuronal Connectivity in Brain Function and Disease. Academy eBriefings. 2016. Available at:

    Log in or Join Now to continue