2024 Undergraduate Summer Poster Symposium

Neural circuits within the mammalian neocortex are formed through synaptic connections between cells over a range of spatial scales. Long-range (>1 mm) axons enable communication between cortical areas that subserve distinct sensory, motor, and cognitive modalities. Surprisingly, the structural underpinnings of these long-range projections remain poorly understood. Recent work has begun to shed light on the complex patterns of bouton density by soma location. Still, this analysis does not differentiate between synaptic contacts and non-synaptic boutons, each with distinct functional properties. We aim to apply novel synaptic labeling strategies to classify how synaptic density varies along the length of long-range contralateral projections in the somatosensory cortex of the mouse.

We take advantage of a mouse line with Cre recombinase-dependent conditional expression of the synaptic protein synaptophysin tagged with the red fluorophore tdTomato. We used adenoassociated viral (AAV) vectors to drive the expression of Cre alongside a green fluorescent marker GFP in cortical neurons extending both local and long-range contralateral-projecting axons, which can be imaged in fixed tissue with confocal microscopy. Preliminary analysis indicates that synaptic density varies along a mirrored gradient, with greater synaptic density as axons approach the cortical surface. On the contralateral side, we observe an opposite pattern in the density of axonal boutons without synaptophysin colocalization. This work provides the basis for further understanding of the morphological principles guiding long-range connectivity in the cortex and underscores the need to differentiate axonal boutons with and without synaptic contacts.   
 

Membrane transport proteins move different small molecule metabolites and large macro-molecules like proteins across membranes, allowing a spatial control of countless biological processes between cells or between different subcellular compartments within cells. Despite being such vital components to metabolism and life, in humans, hundreds of these proteins are poorly characterized, and how they cooperate with lipids within the residing membrane needs to be better understood.  

Our research is focused on metabolite transport across mitochondria, a subcellular organelle that is critical for oxidative phosphorylation energy metabolism. Specifically, we are investigating how cardiolipin, the signature lipid for mitochondrial proteins, can support this process. To this end, we are characterizing the role of cardiolipin in SLC25A39 function, a recently identified mitochondrial transporter for the major antioxidant glutathione and critical for cellular physiology. By combining sequence alignment and structure modeling, we have predicted three sites in the A39 protein that likely correspond to its cardiolipin binding sites. Subsequent mutagenesis experiments have revealed that altering the charges of specific amino acids within these sites affects A39 function and cell growth, potentially by impairing mitochondrial function. Our findings strongly suggest that cardiolipin binding is crucial for A39 transporter function, thereby significantly enhancing our understanding of the mitochondrial transport process. These findings could have profound implications for the development of new therapeutic strategies for conditions associated with mitochondrial dysfunction and oxidative stress, providing a wealth of knowledge for the scientific community.

Brain-computer interfaces (BCIs) are embedded computer systems that interface directly with the human brain. BCIs have numerous applications in treating neurological diseases and controlling prosthetics. In order to be effective, implantable BCIs must record, process, and react to neural data in real time. However, BCIs must also adhere to a strict power budget to maintain safe operating temperatures for the user. As a result, there are many useful but power-intensive features that remain unrealized in BCI devices.

Neuroscientists have long recognized spike sorting—the process of discerning which recorded signals originated at which individual neurons—as a critical precursor to precise brain stimulation. However, most spike sorting algorithms prioritize accuracy at the cost of speed and efficiency, making traditional spike sorting impractical for BCI devices.

This work outlines the initial stages in the development of TempoSort, a real-time spike sorting algorithm that employs a temporal neural network (TNN) to accomplish accurate sorting for a low computational cost. Specifically, we explore the possibility of leveraging neurons’ rhythmic syncopation to enhance spike detection, a crucial step in spike sorting. In addition, we show that average-based filtering methods can be employed in low-fidelity spike sorting to save power while maintaining accuracy. In the future, we hope to implement TempoSort in hardware, benchmark its performance in-silicon, and integrate a TempoSort processing unit into our group’s SCALO BCI device.

The vagus nerve, the longest nerve in the human body, innervates every major organ and plays a crucial role in regulating stress and attention. Initially used to treat epilepsy, vagus nerve stimulation (VNS) can now be applied non-invasively through the skin, known as transcutaneous vagus nerve stimulation (tVNS). Despite numerous tVNS devices on the market claiming to reduce stress and enhance attention, existing research reveals only modest and occasionally mixed effects from acute stimulation. This study aims to investigate the cumulative effects of tVNS on stress reactivity and cognitive performance.

In this psychophysiology study, participants undergo three 30-minute tVNS sessions over three consecutive days. We will evaluate cognitive effects using the Stroop Task and assess social stress reactivity and recovery using the Trier Social Stress Test (TSST) on the third day. Additionally, participants will complete at-home surveys to monitor changes in sleep and mood throughout the study.

We will collect a range of physiological data, including electrocardiography (ECG), impedance, respiration, skin conductance, and pulse transit time, to measure changes in stress and attention. Self-reported data and behavioral observations will complement these physiological measurements. We hypothesize that cumulative tVNS will enhance cognitive performance by narrowing attention, as evidenced by improved Stroop Task results. Moreover, we predict that tVNS will increase vagal tone, resulting in either an attenuated stress response, indicative of a "challenge" state, or heightened attunement to social feedback, indicative of a "threat" state during the TSST. Regardless of the specific outcome, we expect significant physiological changes in response to the stimulation condition.

If non-invasive tVNS proves effective in reducing stress and enhancing cognitive performance, it could have significant implications as a therapeutic or performance-enhancing tool. Such applications could benefit individuals preparing for exams, performances, or managing stressful job environments.

Spinal Cord Injury (SCI) is a catastrophic neurological trauma with a high prevalence rate of 20.6 million cases worldwide. Although there are management strategies for SCI secondary symptoms, effective therapeutic interventions remain elusive. SCI can be attributed to a failure of axon regeneration in the adult CNS upon trauma as a result of both intrinsic cellular factors and extracellular environmental inhibitors. Apolipoprotein E (ApoE), a gene implicated in several neuropathologies such as Alzheimer’s disease, has been shown to exhibit differential effects on axonal regeneration through various alleles. While the ApoE4 variant is linked to reduced regeneration after injury, the rare ApoE2 allele shows increased regeneration after injury. Despite advancements in understanding axonal regeneration in in-vivo and in-vitro mice models, a comprehensive molecular explanation for the ApoE-mediated regeneration in human neurons is lacking. This research aims to fill in this gap by generating and optimizing the conditions of an in-vitro model designed to evaluate axonal regeneration of human neurons derived from Induced Pluripotent Stem Cells (iPSCs). Furthermore, this pioneering model is to be utilized for a thorough evaluation of previously identified prominent gene candidates limiting axonal growth through differential ApoE regulation. Hence, this research will not only provide a novel model of examination but will also identify notable mechanisms of specific reparative axon growth, providing tractable translational leads for SCI and beyond.
 

Stress-related disorders (SRDs) such as anxiety, depression, and posttraumatic stress disorder often emerge during adolescence, a period of extensive structural and functional brain maturation. Adolescents are particularly vulnerable to SRDs, especially when exposed to trauma or stress, which affects up to two-thirds of this age group. The endocannabinoid (eCB) signaling system, crucial for regulating emotional behavior and frontolimbic connectivity, has been identified as a promising target for clinical interventions. However, the eCB system itself is plastic during adolescence and can be disrupted by stress.

In order to develop adolescent-focused behavioral and pharmacological interventions that boost eCB signaling or rescue ELS-induced changes, we first aimed to conduct an extensive literature review. Our review of 61 scientific papers aimed to understand how genotype, sex, and early-life stress (ELS) exposure influence the eCB system’s development and function. The studies show that the eCB system’s mechanisms, expression, and function follow developmentally structured, genotype-specific, and sex-specific trajectories. ELS exposure disrupts these patterns, especially in fear-related brain regions, and has long-lasting effects on eCB signaling. Pharmacological interventions to mitigate ELS effects on the eCB system show promise, but their efficacy depends on the administration route and treatment timing. This literature review and the forthcoming textbook chapter summarize recent advances in eCB system development, aiming to catalyze the development of novel and more effective interventions for adolescents at risk of developing SRDs.

Alcohol use disorder is the continuous drinking of alcohol despite adverse consequences, and it affects over 10% of the world’s population. Recent data show that women tend to drink more to cope with stress and negative affect compared to men. The cyclic adenosine monophosphate (cAMP)-dependent protein kinase A (PKA) intracellular signaling pathway has been implicated in ethanol reinforcement and behavioral responses to stress. Phosphodiesterases (PDE) break down cAMP, and PDE inhibitors have been shown to reduce ethanol drinking and anxiety in male rodents by increasing cAMP. However, it is unclear if these pathways affect stress-induced ethanol drinking in female mice. We hypothesize that apremilast, a selective PDE4 inhibitor, will decrease ethanol consumption in a stress-induced drinking model in female mice.

To test this, mice were randomized into four groups: apremilast-Stress, apremilast-NoStress, vehicle-Stress, and vehicle-NoStress. The “Stress” groups were subjected to repeated foot shock stress and re-exposure to the stress-paired environment before subsequent ethanol drinking sessions, which has been shown to increase ethanol intake. The “NoStress” groups remained undisturbed between drinking sessions. All mice received either vehicle or apremilast injections at least 30 minutes prior to ethanol drinking sessions. We expect increased ethanol drinking in the vehicle-stress group but not in the apremilast-stress group. To confirm apremilast-induced elevation of PKA activity, PKA-mediated phosphorylation of GluR1 in cortical areas will be assessed using Western blotting. Understanding the role of the PDE-PKA signaling pathway in female-specific stress-induced ethanol drinking will help to identify new therapeutic targets, potentially leading to personalized treatments for alcohol use disorders in women.

When individuals are faced with adversity from various life experiences and social interactions, some are able to positively adapt to these challenges and emerge resilient. In contrast, others struggle to overcome these setbacks. Although the direct relationship between resilience and stress remains complex, the global prevalence of stress has markedly increased. The study aims to explore the long-term impacts of early life stress (ELS) on social decision-making and its corresponding neuronal properties. Neonatal rats were subjected to chronic stress using limited bedding material to simulate ELS conditions. Following stress induction, pairs of conspecific rats will be challenged to toggle between strategies that are more prosocial versus antisocial in an open-field environment. Future research will focus on interventions to promote prosocial behavior and enhance resilience in response to adverse events. By furthering our understanding of the impacts of ELS on social decision-making, we aim to demonstrate the translational potential of our research and close the gap between basic neuroscience studies in rodents and clinical applications in humans.

Human brains are often more efficient than classical computers but operate via a vastly different architecture. In an effort to understand neural algorithms, Cognitive Scientists program computational models hypothesized to recapitulate aspects of mental life. Such models are fundamentally limited, however, in that they can only be compared to the brain at the output level. Recurrent Neural Networks (RNNs) provide a partial solution to this problem, expressing algorithms in a distributed network structure whose processing dynamics can be compared directly to neural data. RNNs cannot run traditionally written code, however, and current training paradigms obscure the algorithms they implement.

Here, we investigate and develop upon a new method for directly programming algorithms into RNNs to bridge the gap between network-structured and algorithmically-transparent models. Expressed in the language of dynamical systems, the method maps the evolution of network states to a set of outputs defined as symbolic equations of inputs. Despite the programmable RNN method’s promise, the programming procedure remains time-consuming and inflexible. This summer, we began the development of a Domain Specific Language (DSL) capable of compiling traditional code to the language of RNNs. The compilation procedure for such a DSL will represent user code as a circuit of connected ‘assembly instruction’ networks. Additionally, we aim to establish a platform for users to define their own more complex assembly networks for research applications. By providing a paradigm for the network-level comparison of neural data and programmed algorithms, we hope to enable new insight into the mechanisms of neural computation.

Research has documented that socioeconomic conditions can shape individual decision-making. When the availability of resources is scarce, a person may value decisions differently depending on the probability and amount of reward. The present study aims to understand the relationship between socioeconomic conditions and decision-making under uncertainty (including both risk [when outcomes are known] and ambiguity [when outcomes are unknown]). First, we compared four different computational models of risk propensity and ambiguity aversion during a financial decision-making task. Next, we selected the model that best explained participant behavior. Using the best-fitting model within a Bayesian framework, we constructed a socioeconomically-informed decision-making model that embedded neighborhood disadvantage (Area Deprivation Index [ADI]) and household-level income as predictors of risk propensity and ambiguity aversion. We found that higher neighborhood disadvantage, but not household income, was associated with lower risk propensity (ADI = -2.35, 90% CI [-4.49, -0.17]). There was no significant relationship between ADI or household income and ambiguity aversion. Thus, neighborhood disadvantage can impact decision-making processes, such as risk propensity. These findings suggest that, on average, individuals from more disadvantaged neighborhoods – who likely have a scarcity of resources – are less likely to sacrifice those resources in a risky situation because they cannot afford to lose them.

Mitochondria are integral as an energy source for synapses and are thought to be potentially important in both short-term and long-term plasticity. However, much about their role in various aspects of neural processing remains unknown. Further study may improve our knowledge of fundamental synapse function as well as our understanding of degenerative disease. We demonstrate that mitochondria in the nematode C. elegans display highly significant cell-type specific patterns in their positioning that are correlated with the connectivity of local synapses. We find that mitochondria associate with presynapses that connect to specific downstream neurons, implying that mitochondrial placement is coordinated across the Connectome. These findings lead to a more granular investigation of mitochondria-related mechanisms. Interestingly, we also find that these relationships hold if synapses are conditioned on neuron morphologies typical of the presence of mitochondria, even when mitochondria are absent. This suggests that external factors mediating both neuron morphology and mitochondrial location may be responsible for aspects of synapse function rather than mitochondria themselves, as currently hypothesized.

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by deficits in social communication and interactions, restricted and repetitive behaviors, and often comorbid with epilepsy and hyperactivity. Advances in genomics have identified de novo mutations contributing to ASD, including the Cyclin-dependent kinase-like 5 (CDKL5) gene. CDKL5 deficiency disorder (CDD) features early-onset seizures, central visual impairment, severe neurodevelopmental impairment, and circadian dysregulation.
CDKL5 knockout mice model many human CDD phenotypes, such as impaired social behaviors, altered arousal, impaired learning, and sleep dysregulation. Understanding the neural activity underlying atypical social interaction in these mice can shed light on similar phenotypes in humans, aiding in devising strategies to mitigate social difficulties in people with ASD.

Mesoscopic imaging, a widefield, single-photon, fluorescence-based technique, can analyze neural dynamics in these mice. To overcome the mobility limitations of head-fixed imaging, a new technique allows mesoscopic imaging of multiple mice in an open field, enabling social interactions. Here, using DeepLabCut, we analyzed the behavior of CDKL5 knockout mice in an open field to investigate the association between hierarchy and behavior by comparing our data to the rank established by the tube test. Through this work, we will better understand how to analyze the social dynamics of CDKL5 mice from open field data, enabling us to relate neural activity to social behavior in an assay useful for studying ASD.