Date of Award

Spring 5-15-2023

Author's School

Graduate School of Arts and Sciences

Author's Department

Biology & Biomedical Sciences (Neurosciences)

Degree Name

Doctor of Philosophy (PhD)

Degree Type



Neuroimaging has revolutionized the way in which we understand the hierarchical organization of the amazingly complex, interconnected human brain. Neuroimaging techniques, like functional magnetic resonance imaging (fMRI), have provided high quality structural and functional data, providing multiple in-depth analyses and biomarkers of disease processes. In animal models, mechanistic studies can uncover root pathologies that aren’t explorable in humans. In mice, brain functional connectivity (FC) can be measured via Optical Intrinsic Signal (OIS) imaging – a modality that measures vascular reactivity as a surrogate for neural activity via quantification of fluctuations in oxygenated-hemoglobin (similar to the blood oxygen level dependent (BOLD) signal used in fMRI).

Another advantage of optical neuroimaging in mice is the expression of genetically encoded calcium indicators (GECIs), which provide cell-specific and network-level functional imaging of brain activity at speeds up to at least 4Hz. Imaging in higher frequency bands (compared to <0.2Hz in fMRI or other hemoglobin-based imaging modalities) allows for resolution of neural specific phenomena on the order of milliseconds, such as the global ∼1Hz slow oscillation that is characteristic of anesthesia and non-rapid eye movement (NREM) sleep. We imaged mice expressing the GECI GCaMP6 in excitatory neurons while awake, in NREM (verified by EEG), or under ketamine/xylazine (K/X) or Dexmedetomidine (Dex) anesthesia and reconcile discrepancies between activity dynamics observed with hemoglobin vs. calcium (GCaMP6) imaging. Alterations in correlation structure were most obvious in delta band calcium NREM and anesthesia data, resulting in maps with large regions of polarized positive and negative correlations covering the field-of-view (FOV). We use principal component analysis (PCA) to provide evidence that the slow oscillation superimposes on FC rather than replaces FC patterns typical of the alert state.

While consciousness state can oscillate on the order of seconds, many studies of disease processes are most informative across a longer period of time. Surgical preparations coupled with optical imaging allow for longitudinal experiments on varying timescales. For example, sequalae of subarachnoid hemorrhage (SAH) include vasospasm, microvessel thrombi, and other delayed cerebral ischemic (DCI) events around 3 days post SAH. These DCI events have been shown to coincide with up-regulation of the neuroprotective peptide Sirtuin1 (SIRT1), using an endovascular perforation mouse model. Here, we display global FC disruption caused by SAH and DCI events in parallel with behavioral deterioration. Normal brain connectivity and behavior was maintained during SAH and DCI via two different treatments targeting SIRT1 activation. SIRT1-specific (resveratrol) and non-specific (hypoxic conditioning) treatments both protected against the FC deficits induced by SAH and DCI, with the latter providing the largest protective effect. This indicates that conditioning-based strategies targeting SIRT1-directed mechanisms provide multifaceted neurovascular protection in experimental SAH – data that further supports the overarching hypothesis that conditioning- based therapy is a powerful approach with great potential for improving patient outcome after aneurysmal SAH.

Studies involving focal injury (e.g., stroke, SAH) usually exhibit functional deficits surrounding the injured tissue, however, it is less clear how diffuse processes, such as novel models of acute septic encephalopathy (i.e., Delirium), and encephalitis caused by Zika virus infection, alter brain dynamics. Septic encephalopathy leads to major and costly burdens for a large percentage of admitted hospital patients. Elderly patients are at an increased risk, especially those with dementia. Current treatments are aimed at sedation to combat mental status changes and are not aimed at the underlying cause of encephalopathy. Indeed, the underlying pathology linking together peripheral infection and altered neural function has not been established, largely because good, acutely accessible readouts of encephalopathy in animal models do not exist. In-depth behavioral testing in animals lasts multiple days, outlasting the time frame of acute encephalopathy. Here, we propose optical fluorescent imaging of neural FC as a readout of encephalopathy in a mouse model of acute sepsis. Imaging and basic behavioral assessment was performed at baseline, Hr8, Hr24, and Hr72 following injection of either lipopolysaccharide (LPS) or phosphate buffered saline (PBS). Neural FC strength decreased at Hr8 and returned to baseline by Hr72 in somatosensory and parietal cortical regions. Additionally, neural fluctuations transiently declined at Hr8 and returned to baseline by Hr72. Both FC strength and neural fluctuation tone correlated with behavioral neuroscore indicating this imaging methodology is a sensitive and acute readout of encephalopathy.

Zika virus (ZIKV) emerged as a prominent global health concern due to the severe neurologic injury in infants born to adults who had ZIKV infection during pregnancy. However, neurologic manifestations in healthy adults were subsequently reported during Zika pandemics in South America and Southeast Asia. In this population, infection can result in severe cases of encephalitis and have lasting impacts on cognition, and learning and memory, even after recovery from acute infection. Recent studies have uncovered extensive ZIKV- related neural apoptosis within the trisynaptic circuit involving the entorhinal cortex, the cornu ammonis, and the dentate gyrus of the hippocampus in adult mice. However, there are many contributing regions and circuits involved in cognition and learning and memory outside of this trisynaptic circuit. Communication within the cortex and between the cortex and hippocampus is necessary for a variety of neurological processes, such as performing cognitive tasks or for memory consolidation during sleep. Here, we investigate cortical networks and connectivity utilizing wide-field optical fluorescence imaging. We demonstrate that functional deficits congregate in regions of cortex that are highly communicative with hippocampus, such as somatosensory and retrosplenial cortices. Further, we prove that these functional imaging deficits are correlated with other metrics of disease severity, such as encephalitis score and increased delta power, providing a potentially useful clinical biomarker of disease. Finally, these imaging deficits resolve after recovery from acute infection.

While optical methods have obvious advantages when used to study animal models, the technique is relatively novel (compared to fMRI) therefore, there are many avenues for data processing algorithms to improve. Similar to fMRI, historically, optical methods use a remarkably simple bivariate Pearson-based approach to mapping FC, leading to quick and easy-to-interpret models of brain networks but also susceptibility to global sources of variance (e.g., motion, Mayer waves). Previously, we demonstrated the binarizing effect of the slow oscillation on FC during NREM and K/X anesthesia. While PCA effectively removed the slow oscillation, it is reasonable to assume that a biological process cannot be completely explained in algebraically orthogonal components. Therefore, we pioneer a multivariate approach to imputing individual neural networks from spontaneous neuroimaging data in mice in an effort to map connectivity with less susceptibility to confounding variance. Calcium dynamics in all brain pixels are holistically weighted via support vector regression to predict activity in a region of interest (ROI). This approach yielded remarkably high prediction accuracy, suggesting the optimized pixel weights represent multivariate functional connectivity (MFC) strength with the ROI. Additionally, MFC maps were largely impervious to the slow oscillation. Moreover, MFC maps more closely aligned with anatomical connectivity as modeled through axonal projection images, than FC maps. Lastly, MFC analysis provided a more powerful connectivity deficit detection following stroke compared to standard FC. These results show that MFC has several performance and conceptual advantages over standard FC and should be considered more broadly within the FC analysis community.

Further, with study of diffuse processes (e.g., LPS and ZIKV infection), statistical developments are crucial to solve the multiple comparisons problem when examining all cortical regions within the FOV. Therefore, part of this thesis focuses on the development of a streamlined, open source, user friendly data processing toolbox that contains multiple statistical approaches to make the aforementioned studies possible. Together, the following presents the multiple ways wide-field optical imaging can be used to learn more about the brain’s functional architecture in health and disease.


English (en)

Chair and Committee

Joseph P. Culver

Committee Members

Mark Anastasio, Timothy Holy, Erik Herzog, Robyn Klein,