A major technical limitation in the study of complex social behavior of freely moving rodents is the manual annotation of behavior because it is subjective, extremely time-intensive, and prone to observer drift. Simple Behavioral Analysis (SimBA) utilizes machine learning (ML) applications to automate behavioral analysis by using pose estimation to create supervised ML predictive classifiers of rodent social behavior. In a single project, thousands of videos need to be preprocessed, which includes locating individual trials, identifying behavioral events within each trial, and choosing trimming points to focus on specific outcomes or the presence of multiple animals. Manual editing amounts to thousands of hours, often yielding clips with inaccuracies in timing or content, and leads to inaccurate ML predictive classifiers. To address this problem, my research is centered around integrating behavioral metadata into a tool that can automate preprocessing steps with high precision to improve the quality of resulting classifiers. I attempt this via two major improvements. First, to eliminate the reliance on manual record-keeping, I will implement a function that utilizes metadata to link relevant time-stamped events to their corresponding behavioral experiment video. Second, I will leverage ML-based object detection, such as YOLO4, to determine time points when two animals are present, which often indicates that a social reward has been obtained. Through this project, when compared to manual scoring, I expect that: (1) there will be a significant cut in the time necessary to preprocess videos, (2) the yield of usable trials for SimBA will increase, and (3) that there will be an improved accuracy of pose estimation and classifier performance. Overall, these additions will greatly enhance the ease and flexibility of data preparation for highly specific behavioral analyses during the task, enhancing the efficiency of ML procedures to yield powerful behavioral classifiers.
 

The Neuropixels (NP) probe is a multielectrode array that can record from large populations of neurons with high temporal and spatial resolution, along a shank spanning multiple brain regions. Identifying specific neural populations recorded along the shank is critical for later determining their structural connectivity, adding further insight into their behavioral function. Due to the shank’s material, fluorescent dyes cannot be used for this purpose as the dye will disperse broadly. To solve this, we will use silk fibroin, a biocompatible molecule derived from the cocoon of Bombyx mori to encapsulate a fluorescent protein-encoding viral vector in a silk film that degrades after a controllable period of time. Viral approaches allow for genetic isolation of specific cell-types and circuits. We will combine herpes simplex virus with the silk film and apply it to discrete sections along the shank before insertion, to induce expression of green fluorescent protein (HSV-GFP) in nearby recorded neurons for visualization. First we will test a range of fibroin/HSV-GFP solutions to optimize targeted expression for acute recording applications. Following optimization, we will test the silk/HSV-GFP solution while recording from the mouse amygdala. Then, we will image the brain to visualize the neural populations that were recorded. We predict that the chosen silk/HSV-GFP solution will yield high expression of HSV-GFP in a localized region of the brain that corresponds to the coated subsection of the probe. In future experiments, we will combine the optimal solution with anterograde and retrograde viral tracers along the probes, allowing us to dissect the connectivity patterns of recorded neuronal populations. These experiments will integrate structure and function to derive greater insight from neurophysiological experiments during behavior in mice.

Current models of pain research involve restrictive forms of resident-intruder pairing where experimental mice are involuntarily placed in social situations. These methods have limited application as the research does not account for individual variability and the dynamic social decision-making characteristic of humans. Our research uses a novel volitional social procedure that more accurately represents human behavior in the context of pain. I conducted social self-administration protocols on C57 strain mus musculus to quantify changes in voluntary social interaction before and after neuropathic pain has been induced via spared nerve injury. In addition, I utilized a Von Frey filament test to measure changes in pain sensitivity over this time period. Two social self-administration (SA) experiments were conducted on separate cohorts of C57 mice. In Experiment I, SA was run intermittently at 3-day intervals following neuropathic injury, providing lapses between voluntary social engagement. In Experiment II, SA was run continuously following neuropathic injury. We found that the continuously run SA group experienced a rebound in social interaction to levels matching their pre-surgery states and sham controls, whereas the intermittent group displayed a stark decline in voluntary social interaction that reached statistical significance from sham controls on day 8. Interestingly, tests of allodynia that were conducted to determine prolonged mechanical sensitivity typical of chronic neuropathic pain showed that both groups were experiencing equal levels of increased pain sensitivity throughout behavioral testing. Our results show promise in revealing the dynamic connection between social interaction and pain perception. Research has already identified key areas of interest such as the medial prefrontal cortex (mPFC) and nucleus accumbens (NAc) as hubs responsible for regulating social behavior. We aim to further examine the physiological changes that occur in these areas as a result of persistent pain using a variety of sophisticated analytical techniques.

Using simple behavioral analysis (SimBA) and Deep Lab Cut (DLC), we can create predictive behavior classifiers using pose estimation (PE) data obtained through DLC. PE is a computerized technique to track and predict the location of mice by training the video dataset with labeled frames using specific regions of interest (ROIs). With this, we can create machine-learning (ML) predictive classifiers of complex social behavior in SimBA. Social behaviors and interactions are difficult to manually track due to their rapid successions. To overcome this, I plan to use ML classification using our SimBA pipeline for behavioral classification allowing us to exceed human performance and increase throughput and consistency. I plan to create accurate classifiers for social behaviors that I will use to analyze the behavioral motifs of mice undergoing an operant social stress procedure. First, we train male and female C57BL/6J mice to self-administer (SA) their same sex cage mate. Experimental mice are then subjected to either physical stress for males or witness stress for females. Following social stress, non-reinforced SA is used to assess social reward seeking. Next, social interaction (SI) tests are performed to document time spent approaching the familiar same-sex conspecific cage mates and the aggressive CD-1 mice. All behavior was recorded, and transferred to DLC, followed by frame extraction. Using these frames, we trained the operant behavioral dataset to track the orientation of the mice. Next, we evaluate the dataset for a low error margin as observed by a continuous plateau of iteration loss. Although not complete, I expect to create behavioral classifiers for mice during social decision making in a social reward context following social stress inclusive of sex differences. Providing descriptive statistics of both movement and probability of successive behaviors as they occur in real-time.

Single cell neural activity mapping is a novel experimental approach used to understand the relationship between neural activity and behavior/thought across the intact brain. The current approach utilizes immunodetection of Fos, a reliable and endogenous protein marker for neuronal activity that has unique induction and decay properties. This method combines whole-mount brain tissue clearing, IHC staining, and high speed volumetric imaging. However, whole-mount IHC is incredibly challenging due to many factors including variable antibody lots, lengthy processing protocols, and inconsistent timing. To overcome these limitations, various genetic methods including direct gene modification and replacement have been produced. However, these methods limit brain-wide expression profiles, display inaccurate signal to noise ratios, and yield low signal expression levels. Here, we have developed a novel activity-dependent tool that allows viral vector-compatible Fos-like reporting of neuronal activity. Our strategy relies upon non-promoter-based regulatory sequences that endows downstream genes with Fos-like induction profiles. Here, we present its effectiveness in ectopically labeling Fos+ cells in the mouse brain in-vivo, and report its comparison to other conventional genetic strategies. Further, we extend its range of use with the creation of multiple versions that enables a range of activity level reporting and through multiple wavelengths of fluorescence. In summary, this novel genetic tool can be used to ectopically map single cell neural activity more effectively in order to better understand the anatomical basis of neural coding driven by specific cell-types distributed across the entire brain.

 Within the field of neuroscience, optogenetics is an established experimental tool which can be used to alter specific cell function or trigger enzymatic reactions under millisecond time-scale precision. The high temporal precision of optogenetic recombinases allows for precise identification of cell populations which causally regulate specific behaviors in health and disease. Furthermore, the inducible optogenetic control specifically of recombinase activity for cell-type targeting eliminates the use of other inducible methods including exogenous chemicals that operate on lower time-scales and notoriously cause non-specific effects. Currently available optogenetic recombinases are driven by low wavelengths of light (e.g., Yao et al, 2020) which limit their in-vivo use to local and/or superficial areas of the brain in an invasive manner. Our recently engineered optogenetics-based protein pair, NOC (Near-IR Optogenetic Cre recombinase) induces Cre recombinase activity with near-infrared (NIR) light through dimerization of split-Cre fragments. We have previously demonstrated NOC’s capability for functional Cre recombinase activity under 650 nm light administration, and now aim to optimize the inducibility profile of NOC through two additional modifications. These include modifications to Cre split-site locations based off of previously designed blue light-inducible recombinases (Yao et. al, 2020), and a novel 660 nm inducible photoreceptor pair (Zhou et. al, 2022). The inducibility of these new configurations will be tested using our in-vitro fiber optic system. The optimal configuration of NOC will lay the groundwork for NOGen (NIR OptoGenetic ensemble capture), an alternative version of NOC which will include a calcium-sensing domain on one of the protein fragments. This will limit NOGen’s activity to active neurons only, offering greater precision in identifying specific cell-populations which drive particular behaviors. These improved molecular tools can be used to further our understanding of brain anatomy and function, which serve as an important catalyst for the development of improved brain disorder treatment.

Maladaptive aggression characterizes - or is comorbid with - many neuropsychiatric illnesses, and can have devastating effects on individuals, their caretakers, and healthcare professionals. Human aggression is typically demarcated as exhibiting either reactive (defensive) or appetitive (rewarding) components. Despite a significant clinical awareness of the differences between these aggression presentations, preclinical characterization of their relative circuitry and associated neuronal mechanisms are absent. Using recently established protocols within our lab, we are able to study and compare these aggression phenotypes in outbred male mice in a high throughput manner. Briefly, for appetitive aggression, we train mice to self-administer a novel subordinate intruder over 7 days using a trial design. In the reactive condition, we non-contingently administered intruders with the same frequency distribution as the appetitive mice. In the current experiment, we used CD1xVgat-Cre or CD1xVglut1-Cre mice injected with pGP-AAV-syn-FLEX-jGCaMP7s in the lateral septum (LS) to examine cell-type specific activity via fiber photometry. GABAergic activity in the lateral septum has historically been implicated in the control of reactive aggression, but little is known about the role of excitatory activity in the LS in reactive or appetitive aggression. My roles in this project have included behavioral testing and filming of the mice, as well as scoring these videos for first attacks following intruder presentation. Using these timestamps, I will next analyze the changes in population level dynamics across different time points of aggression motivation, seeking, and consumption using the open source photometry analysis program guPPY. We expect that the photometry results for mice in reactive and appetitive environments will show different patterns of activity, with more glutamatergic activity in the appetitive group, and more GABAergic activity in the reactive groups. I hope to help understand and prevent unnecessary aggression through this research.

Rigorous ethological observation via machine learning techniques, termed computational neuroethology, is a rapidly expanding field. Our lab has created an open-source pipeline for automated behavioral analysis using supervised machine learning called Simple Behavioral Analysis (SimBA), to aid in the high throughput analysis of social behavior. Using pose estimation data of socially interacting animals obtained through open source pipelines such as SLEAP or DeepLabCut, we are able to create large training sets of video frames that are hand scored as positive or negative for a behavior, which we then feed into supervised random forest algorithms. These algorithms then build classifiers which can detect the behaviors in novel videos. My work has focused on building and titrating classifiers for two important social behaviors: face and body sniffing by a dominant mouse toward a subordinate. So far, I have hand-scored a large dataset of social interaction videos to create a sizable training set. I have begun the initial phases of training my classifiers, which involves finding appropriate hyperparameters for the random forest algorithms so that they can differentiate positive and negative behaviors, and refrain from overfitting to our training datasets. Using both machine learning performance metrics as well as hand versus machine comparisons, I am able to understand the generalizability and accuracy of my classifiers. As I continue with this project, I will selectively add more positive and negative examples to correct false positives and boost the confidence of the classifiers through subsequent iterations. This work allows me to gain an understanding of the principles of machine learning techniques, and create classifiers that we openly provide to behavioral neuroscience labs across the world. We expect that the pooling of these classifiers with outside labs will promote a high level of standardization of behavioral definitions in behavioral neuroscience, ultimately increasing reliability and reproducibility.

Neuropsychiatric disorders pose a difficult challenge for healthcare providers. Treatments for such disorders vary in efficacy and come with detrimental costs for patients and their communities. Historically, preclinical animal models have failed to incorporate the nuances of volitional human social behavior. This project used chronic social defeat stress to induce depression-like behaviors in male and female mice, this was followed by self-administered social interactions within an operant chamber in which lever presses were reinforced by social contact. The goal is to develop preclinical animal models that can be assessed to identify mechanisms responsible for stress-induced social motivation. The mice will be injected with a nuclear localized tag (oNLS) and viral retrograde tracer rAAV2-retro-GFP. Male and female mice will train to self-administer social interaction with a sex and age-matched housing partner over the course of ten 12-trial sessions. Next, experimental male and female mice will be subjected to physical and witness defeats followed by operant social self-administration. Before and after the 10-day operant social stress sessions, we will test social reward seeking via non-reinforced self-administration of social reward followed by a progressive ratio test. Brain tissue will be collected and prepared for immunohistochemistry and iDISCO+ whole-brain clearing for cfos labelling. We predict results will show differential cfos activity in sexually dimorphic brain regions such as the hippocampus, prefrontal cortex, amygdala and the bed nucleus of stria terminalis. We determine that operant social stress can be used to discern differences in social motivation in male and female mice as a result of stress-induced factors. There is great potential in using whole-brain activity mapping to identify brain structures activated during social reward following social stress, as this can also serve as a technical resource for the field by identifying relevant non-canonical brain regions and circuits that govern such behaviors.

Alzheimer’s disease (AD) is characterized by aberrant cleavage of Amyloid Precursor Protein (APP) leading to toxic β-Amyloid (Aβ) peptide aggregates. Presenilin 1 (PSEN1), and presenilin 2 (PSEN2) are intimately involved in this process. We found that alternative splicing of PSEN2 leads to the increased inclusion of a cryptic exon (part of the RNA that codes for a protein), in the typically non-coding intron 9 of PSEN2, titled exon 9B (PS2x9B), which has elevated abundance in individuals with sporadic AD. My aim is to uncover the link between PS2x9B and sporadic AD. I designed a PCR approach to measure the ratio of PS2x9B inclusion compared to non-mutated Wild Type (WT) in cDNA generated from parietal and temporal lobe tissue samples. I distinguished PS2x9B inclusion by amplicon size on a gel and measured the ratio to WT via band intensity using ImageJ. I found that the ratio of PS2x9B inclusion was significantly elevated in sporadic AD cases in the parietal lobe (p = 0.02) and the temporal gyrus (p = <0.001). Next I transfected SH-Sy5y cells with plasmids containing either WT or PS2x9B PSEN2 joined to a Flag-tag. I extracted proteins from these cells and control empty plasmids, and used Western blot to examine size differences between each plasmid’s protein products to research the impact of PS2x9B inclusion on PSEN2 protein levels. I hypothesize that inclusion of PS2x9B will lead to truncated proteins due to an early stop codon. Additionally, I am working to establish the consequences of PS2x9B on APP processing and Aβ cleavage. I expect that PS2x9B could cause either an increase in Aβ production or modify its length. These findings will allow us to expand our understanding of PSEN2 and alternative splicing in sporadic AD which could guide development of novel gene therapy treatments for AD.

 

The Developmental Origins of Health and Disease (DOHaD) hypothesis evaluates how the prenatal environment affects health after birth. The placenta is a multi-faceted organ that sustains life during human development and is key to evaluating the DOHaD hypothesis. Glial Cells Missing Transcription Factor 1 (GCM1) is a transcription factor that plays a critical role in placental development. Our goal is to understand the downstream effects of GCM1 on various genes necessary for placental development by evaluating gene expression after GCM1 knockdown. The BeWo choriocarcinoma cell line is a model of placental syncytiotrophoblasts cells which undergo a cell fusion process called syncytialization in the placenta to form multinucleated cells that help exchange nutrients. Previously, we knocked down GCM1 in full-term primary placental cells that spontaneously syncytialize and assessed gene expression using RNA sequencing. We identified 10 differentially expressed genes. Based on those findings, we hypothesized that GCM1 plays a greater role during early pregnancy leading us to repeat the GCM1 knockdown in BeWo cells. BeWo cells were treated with 20µM, 50µM and 100µM forskolin (FSK) for 48hr to induce syncytialization which was confirmed via qPCR of syncytialization markers GCM1 and Syncytin-2 and through fluorescence microscopy. GCM1 expression increased 3.15, 1.3, and 1.2 fold respectively after treatment with 20µM, 50µM, and 100µM FSK, whereas Syncytin-2 increased 78.1 fold after 50µM FSK treatment. We then performed an siRNA knockdown of GCM1 in unsyncytialized BeWo cells with two concentrations of siRNA (25nM and 50nM) for 24hrs and observed a 70% and 80% reduction in GCM1 expression, respectively. Next steps include optimizing the siRNA procedure for syncytialized BeWo cells and comparing these results to our previously conducted experiment. Overall, this will improve understanding of how GCM1 coordinates gene expression in the placenta during pregnancy.

Pangolins are the most widely trafficked mammal in the world. Pangolin scales are smuggled for use in traditional medicine and this is a concern due to all eight species of pangolins; four in Africa and Asia, respectively, currently being in the vulnerable to critically endangered category. My research provides insight into the practices and routes of smugglers using stable isotope analysis to obtain information on locality of origin from scale samples of Smutsia gigantea, Smutsia temminckii, Phataginus tetradactyla, and Phataginus tricuspis, the four species of pangolins native to the African continent. I collected powdered scale samples from the four pangolin species and analyzed the samples using an Elemental Analyzer attached to an Isotope Ratio Mass Spectrometer (EA-IRMS). Scale samples were sourced from recent seizures in multiple ports. Data from the EA-IRMS provide Sulfur, Carbon, Nitrogen and Oxygen stable isotope composition estimates. This information was then used to place individuals into groups and then predict where those groups lived prior to being poached. Better understanding of smuggling routes and poaching habits is an aim of this project. Stable isotope analysis may also help our collaborators identify the pangolin species when scale samples contain insufficient or otherwise poor quality DNA. This project has valuable implications in future forensic studies of organic materials and will lead to better policing and policy making with regards to pangolin conservation.

The glymphatic system, which is primarily active during sleep, is a network of astroglial perivascular channels within the brain that allows for cerebrospinal fluid (CSF) influx and exchange. Glymphatic exchange plays a crucial role in the clearance of amyloid, a hallmark in the development of Alzheimer’s. Recently, a bidirectional relationship between Alzheimer's disease and sleep has also been suggested with amyloid deposition associated with mid-life sleep disruption. However, the mechanistic link between sleep disruption, particularly over chronic time scales, and the development of Alzheimer’s pathology remains unclear. This study investigated whether chronic sleep disruption, similar to that experienced in aging population, impacts downstream Alzheimer’s-related neuropathology. We hypothesized chronic sleep disruption will result in decreased glymphatic function and increased amyloid plaque burden. This experiment utilized a chronic sleep disruption model using Lafayette Sleep Fragmentation chambers, where mice underwent either chronic sleep disruption every two minutes during normal sleeping periods (daylight hours) or normal sleeping conditions (sham) from 10 weeks to 18 weeks of age (n=120). After eight weeks of sleep disruption or sham exposure, glymphatic function was assessed by dynamic in vivo near infrared imaging following stereotactic CSF tracer injection. Animals were perfusion fixed, cryosectioned, and glymphatic function was further assessed by measurement of fluorescent cerebrospinal fluid tracers in brain tissue. Aquaporin-4 localization, amyloid plaque deposition, and markers of astroglial and microglial activation were assessed by immunofluorescence. The collected data demonstrated that sleep disruption significantly increased neuropathological outcomes. The measured impact of glymphatic function was also correlated with these downstream pathological effects. These findings could be an indicator of interactions between neurological disease progression and an inflammatory expression after sleep disruption. They can also shed more light on the complex relationship between Alzheimer’s disease progression, the glymphatic system, and chronic sleep disruption.

Neurological trauma, from a spinal cord injury (SCI), can have devastating effects on the quality of life of individuals, often resulting in a significant loss of motor function and accompanying bowel and bladder complications. Recent research has combined use-dependent physical rehabilitation, which is the current gold standard of treatment, with the delivery of neuromodulators or electrical stimulation to improve motor-recovery outcomes after SCI. However, little work has been done to understand the scope of psychological changes that occur after SCI, or how these changes may, in turn, affect an individual's ability to recover from the trauma. To fill this gap, our project examined changes in motivation levels after a chronic cervical contusion of the spinal cord that produces impairments in forelimb-motor function. We also studied how subsequent motor recovery altered motivation. Our experiments were conducted in an adult rodent contusion model of chronic cervical SCI. We used a skilled forelimb reach-and-grasp behavioral task to assess the motor performance and injury severity of rats. To assess motivation, we recorded the duration that rats attempted to complete the reach-and-grasp task over a thirty-minute time window. Motivation levels and motor performance were assessed pre-and post-SCI and before, during, and after therapy. Our results show that motivation levels were significantly impacted by the SCI, with motivation loss positively correlated to injury severity. Surprisingly, despite significant motor recovery, motivation levels continued to remain low months after the injury. These results provide new insights into the effect of SCI on psychological factors, which will inform future investigations and the design of therapies targeting neurological trauma.

The ocean and atmosphere interact through air-sea exchanges of heat and energy across the air-sea interface. These air-sea fluxes have important implications on global weather and climate patterns. Because estimation of covarying turbulent variations is not feasible in Numerical Weather Prediction (NWP) models, the turbulent air-sea exchanges are typically estimated using bulk air-sea flux algorithms based on state variables. However there are large differences in the values estimated by different NWP and even when they agree, without a reference data set, it may be that all NWP are equally biased. For this project, I used in situ observations collected by Saildrone Uncrewed Surface Vehicles (USV) in the central tropical Pacific to assess bulk flux estimates from multiple atmospheric reanalyses including NCEP Climate Forecast System Reanalysis (CFSR), ECMWF Reanalysis v5 (ERA5), NCEP/NCAR Reanalysis 1 (NCEP1), and NCEP/DOE Reanalysis 2 (NCEP2). Preliminary results, based upon hourly, spatially-interpolated, co-located values that are then made into 24-hour “daily” averages, indicate that all of the reanalyses had a strong correlation with USV observations for net heat flux and net SWR, but the correlation was much weaker (0.5 to 0.7) for other flux components and very weak (~0.25) for the net longwave radiation for NCEP1 and NCEP2. The root mean square errors for the 24-hour-averaged differences were 55 to 66 W/m^2 for solar radiation and 20 to 30 W/m^2 for latent heat flux. In my analysis of my results, I looked at the differences region by region for each of the flux components and state variables as well as for each of the products. As Saildrone technology becomes more widely used and more intercomparison studies such as this are conducted, observations from Saildrones could eventually be integrated into NWP models, possibly improving forecast accuracy.