Abstracts#
Understanding the interactions between different brain regions is vital for unravelling the complexities of brain disorders and brain functions such as cognition, behaviour, and memory. While previous investigations of inter-regional communication were primarily performed via neuroimaging and electroencephalography, these are of limited spatial and temporal resolution. Recent technological advances and improved recording techniques have opened new avenues in cross-region intracranial recordings, allowing us to explore the interactions between brain regions at a finer scale than ever before. This thesis addresses the analysis of inter-regional interactions using a variety of data types obtained from simultaneous recordings of brain activity.
To facilitate this analysis, we introduce Simuran, a software tool that enables the analysis of different data formats and promotes interoperability between neural data and analysis tools. Using Simuran, we investigate the impact of anterior thalamic lesion or inactivation in rats on the hippocampus, subiculum, and retrosplenial cortex. Our results reveal that anterior thalamic lesions disrupt spatial cell coding in the subiculum while leaving CA1 spatial cell coding intact. Interestingly, at the local field potential level, the effect of these lesions is not clear in the subiculum alone; rather, differences emerge when comparing local field potentials across brain regions and examining the relationship between local field potentials and spiking activity.
Next, we explore interactions at the neuron-level and consider the likelihood of recording structurally connected neurons across regions using simulations and statistics. Our findings indicate that modern recording techniques offer a high probability of capturing synaptically connected neurons across well-connected brain areas. Leveraging this connection probability, we turn to analysing large-scale datasets. The results suggest that task performance involves a balance between consistency in the represented information via highly correlated neural activity, while avoiding excessive redundancy. Additionally, patterns in neural ensembles exhibit similarities during both correct and incorrect task performances.
This research builds upon solid foundations to consider cross-region interactions in the brain, shedding light on observed behaviours and highlighting exciting opportunities to leverage modern technology and analytics.
Chapter 1 - introduction#
This thesis chapter addresses the complexity of behaviors requiring the coordinated interaction of multiple brain regions, such as the intricate process of reading a book. While existing knowledge outlines the general functions of individual brain regions, and functional Magnetic Resonance Imaging (fMRI) studies have identified active areas during specific tasks, there remains a critical gap in understanding the fundamental interactions at the level of neurons and neural circuits. The research aims to explore these interactions across brain regions, offering valuable insights into behaviors and brain disorders. The investigation is grounded in the understanding that unraveling the neural dynamics at a finer spatial and temporal scale can provide a more comprehensive view of complex cognitive processes. Through this exploration, the research endeavors to contribute to a deeper understanding of the neural underpinnings of behaviors and their implications for brain-related disorders.
Chapter 2 - background#
This thesis chapter focuses on capturing and understanding the instructional functionality of the brain by integrating information from different regions. The approach involves studying behaving animals to maintain a direct link between observed signals and freely moving animals' behaviors. Utilizing machine learning techniques, the research analyzes sets of neurons and local field potential signals from various brain areas, establishing connections with behavioral outcomes. The subsequent development of software enables the integration of diverse data types, facilitating multi-faceted analyses. Moving into experimental manipulation, the study explores the impact of lesions on the interior thalamic nuclei on specific areas within the limbic system. Building upon preliminary findings, the research delves into investigating interactions among ensembles of neurons across diverse brain regions. Statistical estimation techniques are employed to assess the likelihood of recording synaptically connected neurons from multiple brain areas using recent recording methodologies. In the final stage, the study aims to unveil the relationship between task performance and the intricate interactions among multiple brain regions. This comprehensive approach seeks to advance our understanding of the complex neural dynamics and their role in shaping behavioral outcomes.
Chapter 3 - Simuran#
Recent advances in neural recording technologies, computing power, and experimental methods have made it possible to simultaneously investigate neural recordings from multiple brain regions. However, the availability and continued development of numerous software tools and data formats have presented a challenge for neuroscientists analysing this type of data. To address this challenge, we developed Simuran, a Python library that simplifies the process of analysing large neural data recordings and facilitates interfacing with other analysis software. Simuran supports a variety of data formats and sources, including Axona, NWB, the Allen Institute, and the International Brain Laboratory, in addition to neuroscience analysis libraries such as NeuroChaT, MNE, Elephant, and Neo. Simuran also provides a node-based user interface to streamline the analysis process for inexperienced coders. We have successfully used Simuran to analyse data from multiple brain regions simultaneously, including handling large experiments and analysing open-source datasets. Simuran is used to convert Axona experiments to common format and analyse spiking activity and local field potentials from the subiculum, retrosplenial cortex, and CA1. We also demonstrate using Simuran to analyse data from the Allen Institute and the International Brain Laboratory.
In summary, we have developed a flexible tool that enables the analysis of neural data from multiple brain regions simultaneously. Simuran is modular and designed to support new tools and data formats, making it a valuable resource for the neuroscience community. We close by discussing future directions for Simuran, and the promise such a tool holds for the analysis of large-scale neural data.
Chapter 4 - anterior thalamic lesions#
In this study, we investigated a set of interconnected brain regions within the rodent limbic system, consisting of the subiculum, CA1, retrosplenial cortex, and anterior thalamic nuclei. We focused on the role of the anterior thalamic nuclei for spatial coding in the subiculum and CA1. By understanding the role of the ATN in this circuit, we aimed to shed light on the mechanisms underlying spatial coding and implications for memory and amnesia. Electrophysiological recordings were performed in freely moving male rats with permanent ATN lesions or temporary inhibition of the ATN to obtain single-unit measurements. Recordings of units were obtained from the dorsal subiculum and CA1 of lesioned rats, and the dorsal subiculum of control rats. Control and lesioned rats were tested in behavioural tasks, including spatial alternation tasks and open-field exploration. Our findings revealed that both permanent and reversible lesions of the anterior thalamic nuclei led to a complete cessation of subicular spatial signalling. As a result, spatial memory performance was reduced to chance levels. Interestingly, place cells in the hippocampal CA1 region remained largely unaffected by the lesions. From our results, we can conclude that the outputs from CA1 alone are insufficient for subicular spatial coding. Instead, the anterior thalamic nuclei play a critical role in providing key inputs or modulating the subiculum to support spatial coding. These findings deepen our understanding of the intricate mechanisms underlying spatial coding and memory formation in the limbic system.
Chapter 5 - LFP with anterior thalamic lesions#
The ATN are vital for spatial memory in rodents and appear necessary for episodic memory in humans. We previously showed how ATN lesions abrogate spatial coding in subicular cells. Here, we further elucidate these findings by investigating the relationship between the LFP and single-unit activity in the dorsal subiculum and the LFP in ipsilateral RSC post-lesion of the ATN. Electrophysiological recordings were performed in rats during free-exploration and spatial alternation tasks with permanent ATN lesions via NMDA and temporary ATN inhibition via muscimol to obtain single-unit measurements and LFPs. We primarily focused on analysing theta rhythm activity (related to spatial coding in single units), beta rhythm activity (linked to learning and novelty detection), and sharp-wave ripples during sleep (linked to memory and replay) from LFPs. Lesion effects on subicular LFPs were unexpectedly subtle, given the gross behavioural impairments present post-lesion. Theta and beta rhythms, sharp-wave ripples during sleep, and the relationship between motor activity and theta rhythm activity were all maintained in the subiculum post-lesion. However, power in RSC LFPs was reduced outside of the theta band post-lesion, the coherence between LFPs from RSC and subiculum was reduced in T-maze trials post-lesion (but less so in open-field exploration), and the coupling between the subicular LFP theta phase and single unit-spiking activity was reduced post-lesion. We suggest that a key contributor to the behavioural impairments post-lesion may stem from a change in the transfer or representation of information between regions in the hippocampal circuit following anterior thalamic impairment.
Chapter 6 - neuron connectivity#
Large-scale simultaneous in vivo recordings of neurons in multiple brain regions prompt the question of whether we can capture direct interactions between these neurons. This, in turn, opens the possibility of identifying inter-regional communication rules during behavioural tasks, assuming we can detect coordinated activity among neurons in connected brain regions. Here, we take a first step towards addressing this issue by estimating the number of connected neurons that can be recorded through simultaneous recordings in anatomically connected brain regions. To investigate this, we construct a directed acyclic graph that represents the chemical synapses between neurons in the specific brain regions under study. Through Monte Carlo simulations and statistical analyses, we develop a method for calculating the probability distribution of recordable neurons that are connected. We apply this method to neurons in the neocortex, following the Blue Brain model of the mouse neocortex connectome, as well as to neurons in the hippocampus and subiculum, based on statistical data from anatomical studies. Our findings reveal that the probability of recording a connected pair of neurons is notably high in well-connected brain regions like the hippocampus and subiculum. In the neocortex, the probability is relatively lower but still significant. Furthermore, along chains of synapses, such as disynaptic or trisynaptic connections, the probability of recording connected neurons increases dramatically. By employing high-throughput neural recording technologies in well-connected brain areas, such as specific regions within the auditory cortex, we can enhance the likelihood of capturing a substantial number of neuron pairs that are directly connected across brain regions. Consequently, these techniques hold promise for investigating and validating hypotheses related to interregional communication and source transformation rules.
Chapter 7 - task perfomance and correlations#
The relationship between the timing of action potentials from multiple neurons with stimuli and observed behaviours has been a long-standing topic of interest. From a theoretical view, increased neural correlation implies reduced information transmission. Despite this, evidence suggests that increased neural correlation could enhance task performance. This suggests a delicate balance between neural information efficiency and stability, with task performance potentially influencing this balance. In this study, we investigate this relationship and its dependence on task performance. We compare coherence from local field potentials in the claustrum, anterior cingulate cortex, and retrosplenial cortex across fixed interval and fixed ratio tasks. We then compare correlations between a reduced dimension representation of neural firing in catch and miss trials during a visual change detection task. To identify these low-dimensional representations of neural firing, we employ canonical correlation analysis. Additionally, we use Gaussian process factor analysis to generate latent neural trajectories in visual and limbic system brain areas, allowing us to assess the similarity of computations between catch and miss trials in the same visual change detection task. Coherence differed in a region-specific manner between fixed interval and fixed ratio trials. Latent neural trajectories were distinctly separable between catch and miss trials. However, the shape of the neural trajectories often appeared similar for catch and miss trials, suggesting that related computations were being performed. The correlation between the reduced dimensional activity differed between catch and miss trials, with higher correlation in catch trials for visual areas and lower correlation in catch trials for limbic system areas. Our findings demonstrate that trial-to-trial correlations vary across similar tasks and task performance, but in a region-specific manner. Thus, the correlation-task relationship appears to be highly task- and brain region-dependent, lacking a general rule. These results offer insights that can help explain the conflicting reports on the correlation-task relationship.