Magnetic resonance imaging (MRI) is a non-invasive and non-ionizing imaging modality used to study cellular and molecular events as well as track disease progression in tissues. Due to limited sensitivity, genes, proteins and other molecules can be utilized to increase contrast in specifically labeled tissues and cells for MRI images. MRI reporters are able to provide information on gene expression, cell tracking and migration, and cellular energy metabolism. Additionally, many genes and molecules used to produce contrast may yield therapeutic benefit. Here, we take a closer look at two established MRI reporter genes: creatine and mms6, and determine their therapeutic potential.We tested the efficacy of creatine (Cr) supplementation on the brain. Recent studies suggest that Cr supplementation may improve cognitive function and memory, but the major hurdle is bypassing the blood brain barrier. Initially, we developed a creatine nasal spray and hypothesized that we could increase creatine and phosphocreatine (pCr) concentration in mouse brain. Despite multiple rounds of experiments with varying creatine concentration and duration of administrations, we did not confirm an increase in Cr or pCr in experimental groups. Due to issues resolving Cr and pCr on small tissue samples, we developed a novel pre-incubation technique to convert Cr and pCr to creatinine (Crn) and calculate the concentration of total creatine using commercially available creatine and creatinine assay kits. This technique was demonstrated on Cr and pCr standards as well as mouse muscle and brain tissue. We tested the ability of MRI reporter mms6 to enhance gemcitabine cytotoxicity on pancreatic tumor cell line, PANC-1. Mms6 is an iron binding protein found in magnetotactic bacteria. When the mms6 gene is expressed in mammalian cells, they increase iron uptake and storage producing measurable contrast in T2-weighted MRI images. We created a PANC-mms6 cell line capable of producing MRI contrast, and treated them with a common pancreatic cancer drug, gemcitabine. Given its resistant nature in PANC-1 cells, we hypothesized that we could increase cytotoxicity and cell death in PANC-mms6 cells over non-expressing PANC-WT cells. The results revealed a significant reduction in viability of PANC-mms6 cells after treatment with gemcitabine while PANC-WT cells remained viable under the same conditions.

The human brain is a complex neurobiological system characterized by nonlinear spatio-temporal dynamics. While functional magnetic resonance imaging (fMRI) has widely been used in the field of brain dynamics, the whole-brain nonlinear maps remain to be characterized.

We firstly applied phase space embedding on blood oxygen level dependent signals from resting state fMRI data. We mapped their phase space dynamics with SE, which is the sum of the length of the portrait edges. We studied the effects of repetition time (TR) and bandpass filter on the optimal model parameters, including the embedding time delay and the embedding dimension. The present method was applied to two psychiatric data sets, i.e., schizophrenia (SZ) and autism spectrum disorder (ASD), to demonstrate its capability in the characterization of abnormal brain dynamics. The statistically parametric maps were compared between patients and controls. A significant increase of SE was found in the default mode network, salience network, and auditory network regions of SZ patients, indicating increased dynamic change along with SZ. A significant increase of SE was also found in DMN and SN of ASD patients, besides the visual network. These results suggest that the present method is robust and promising in characterizing the dynamic changes of the BOLD signal among patients with mental disorders.

In addition, most of the fMRI studies which related to SZ and ASD had concentrated on the neuronal activity of patients in 0.01–0.1 Hz frequency band. However, it remains unknown whether underline dynamics in different frequency bands of BOLD signals in the two types of mental disorders. In this study, we firstly filtered BOLD signals into four sub-bands: slow-5, slow-4, slow-3, and slow-2. And then we used phase embedding to make more use of nonlinear dynamics on these multi-bands signals and calculated six geometric features in the embedding space. We used a SVM classifier to characterize SZ and ASD patients and we did a 10-fold cross validation with 100 initialization to compute mean accuracy. We achieved a highest accuracy of 85% and 73% for SZ and ASD, respectively. Out result proved that the present method is robust and promising in characterizing the BOLD signal among patients with two types of mental disorders.

While creatine levels in skeletal muscles can be enhanced by exogenous creatine supplementation, the elevation of creatine levels in the brain with oral creatine administration remains a challenge due to a lack of effective creatine transportation through the blood-brain barrier (BBB). As consequence, intranasal administration has been established as a method to bypass the blood-brain barrier and deliver drugs directly to the brain. In our studies, we focused on using intranasal delivery to bypass the BBB and deliver creatine to the brain. Our first objective was to use magnetic resonance spectroscopy to measure creatine accumulation in the brain after a 14-day period of intranasal administration. The spectra showed that the brain creatine/choline ratio was significantly increased after creatine administration. This proved the accumulation of exogenous creatine in the brain. In the second objective, we studied the effects of intranasal creatine administration on wild-type rats and evaluated the cognitive performance after administration via the Barnes maze experiments. This study demonstrated that intranasal creatine administration can improve the performance of rats in the Barnes maze. The results of biochemical assays showed that administration of creatine can increase the creatine content in the rat brain. Despite the success of intranasal administrations, there are still drawbacks, especially when it comes to pH and solubility. To address the solubility limitation of creatine concentration per unit volume and ensure a pH value close to neutral, we assembled creatine microparticles. We tested the characteristics of these creatine microparticles, including size distribution, functional group composition, assembly rate, stability, and cytotoxicity. In vitro permeability measurements indicate creatine microparticles are permeable to human respiratory epithelial cells. Overall, we developed methods to evaluate efficacy of intranasal creatine administration in vivo and methods to evaluate characteristics and permeability of creatine microparticles in in vitro drug delivery models. Our experiments indicate intranasal administration of creatine can improve wild type rats’ cognitive performance in Barnes maze and increases creatine content in the brain. Moreover, creatine microparticle suspensions were proven to exceed the limits of creatine solubility. This approach enables the delivery of creatine intranasally beyond their solubility limits without significant cytotoxicity and alteration of cell layer integrity.

There has been recent growth in using hidden Markov models (HMMs) to chronicle the latent spatiotemporal dynamics of brain activity acquired from functional magnetic resonance imaging (fMRI). This technique can be used to model resting state data or dynamic processing, such as attention. The locus coeruleus (LC), a small subcortical structure, is the main source of norepinephrine throughout the brain and therefore is involved in modulating attention and arousal. An HMM can be used to identify latent fMRI-based brain states comprised of a combination of networks, and to determine how the behavior of these brain states change as a function of attentional perturbations.The purpose of this investigation is two pronged. We aimed to create a comprehensive theoretical and empirical overview of various HMM subtypes in effort to make informed decisions about which should be used under different circumstances for future investigations. We then fit this probabilistic model to an fMRI dataset optimized to image the effects of the LC to gain insight into its dynamic relationship with attention and arousal. To accomplish this, we first theoretically contrasted three HMM subtypes, then applied them to an fMRI resting state dataset to obtain an empirical understanding of the strengths and weaknesses of each one. One model type, an activation-based HMM, was applied to a pseudo-resting state dataset where LC activity was noninvasively up-regulated via a handgrip task. This aimed to analyze how HMM-related measure focusing on attentional networks changed as a function of actively squeezing a squeeze-ball. We found that an activation-based HMM is ideal for capturing temporal dynamics and is preferred if activation and connectivity state patterns are to be analyzed in conjunction. An HMM where functional connectivity values were summed at each time point is advantageous if a study’s goal is to examine the general connectedness between nodes. We also noted that the HMM subtype fitted to all unique correlation values from a sliding window analysis is preferred if an investigation wishes to explore specific nodal connectivity. Our findings from HMM-based measures extricated from the LC-focused dataset show potential evidence of norepinephrine depletion and attentional allocation.

Autism Spectrum Disorder (ASD) is a highly heterogeneous developmental disorder with diverse clinical manifestations. Neuroimaging studies have explored functional connectivity (FC) of ASD through resting-state functional MRI (fMRI) studies, however findings have remained inconsistent, thus reflecting the possibility of multiple subtypes. The Autism Brain Imaging Data Exchange (ABIDE) contains neuroimaging data from more than 17 international scanning sites and has become a useful tool in studying the brain-behavior relationships of ASD. However multi-site databases impose site effects that confound FC due to the use of different scanning hardware, models, and parameters. Although there are established methods to mitigate site effects, these strategies often result in reduced effect size in FC features known to be affected in diseased populations. In this work, we propose a site-wise de-meaning (SWD) strategy in multi-site FC analysis of fMRI and evaluate the performance against two common site effect mitigation methods (Generalized Linear Model and ComBat Harmonization). These methods were tested on two multi-site psychiatric consortium: ABIDE and Bipolar and Schizophrenia Network on Intermediate Phenotypes. Preservation of consistent FC alterations in patients were evaluated for each method through the calculation of effect size (Hedge’s g) between patients and controls. The SWD method demonstrated superior performance in preserving the effect size in FC features associated with neurodevelopmental and psychiatric disorders compared to the original data and commonly used methods. We then aim to identify the relationships between clinical symptoms and FC measures to help clarify the inconsistencies in earlier findings and advance our understanding of ASD subtypes. Canonical correlation analysis was performed on two-hundred and ten ASD subjects from ABIDE to identify significant linear combinations of resting-state connectomic and clinical profiles of ASD. Then, hierarchical clustering defined three ASD subtypes based on distinct brain-behavior relationships. Overall, we reduce heterogeneity in multi-site fMRI databases and elucidate the heterogeneity of ASD clinical manifestations and connectomic profiles. The reduction of site effects and preservation of FC associated with disorders can lead to a better understanding of brain connectivity in diseased populations, and identification of distinct ASD subtypes may lead to better targeted therapies for individuals.

Functional magnetic resonance imaging (fMRI) measures brain activity through the blood-oxygen-level-dependent (BOLD) signal, which has been widely used to study the human brain in neuroimaging studies. Resting-state fMRI has been acquired to study functional connectivity between brain regions without any specific tasks being involved during the scan. Raw fMRI data, a time series data with three-dimensional images in stereotaxic space, provide rich information related to neural activities. Based on recurrent neural networks (RNNs), the temporal dynamics behind the brain network has been characterized. To address the functional connectivity between spatially distant regions within/across functional networks relying on regions-of-interest (ROIs) based fMRI data, we propose two deep learning architectures for fMRI analysis. First, Convolutional RNN (ConvRNN) has been introduced with convolutions for spatial feature extraction on ROI-based fMRI data. Local interactions among ROIs from the same functional network have been extracted, and temporal features have been processed by the RNN-based architecture. Better classification performance has been achieved over previous studies. The in-place visualization based on ConvRNN reveals the informative regions related to individual identification, leading to the same conclusions from previous studies. Second, we introduce a connectivity-based graph convolution network (cGCN) architecture for fMRI analysis. fMRI data are represented as the k-nearest neighbors graph based on the group functional connectivity, and spatial features are extracted from connectomic neighborhoods through Graph Convolution Networks (GCNs). We have demonstrated our cGCN architecture on two scenarios with improved classification accuracy. cGCN on the graph-represented data can be extended to fMRI data in other data representations, which provides a promising deep learning architecture for fMRI analysis.