2026 Theses Doctoral

Modeling Neurovascular Coupling and its Pharmacological Perturbation

Malan, Elizabeth A.

In the brain, increased neuronal activity is almost always accompanied by a local increase in blood flow, a phenomenon termed neurovascular coupling (NVC). NVC is essential to normal brain function and is utilized as a proxy for neural activity in noninvasive functional brain imaging. However, the underlying cellular mechanisms that drive NVC are still poorly understood. Mathematical models ranging from simple linear models to dynamic models and machine learning models have been developed in an attempt to relate neural activity to the resulting hemoglobin response, but these often lack either mechanistic relevancy, knowledge of the underlying signals or reliability.

In previous experiments our lab observed distinct spatiotemporal nonlinearities in the coupling response that suggest a two-part endothelial mechanism driving NVC. We hypothesized that the observed spatiotemporal characteristics of coupling could be explained as the sum of two mechanistically distinct components. In order to test this, we first developed a relatively simple two-component model to relate neuronal and hemoglobin activity when applied to averaged stimulus-evoked data. We then extended this model so that it could be applied in more complex situations, such as on single-trial or non-averaged data.

Wide-field optical mapping (WFOM) was used in order to collect simultaneous neuronal and hemodynamic signals in awake, head-fixed Thy1-GCaMP6f mice. In order to observe coupling in response to a controlled event, we used whisker stimulation of durations from 1-7 s with an acrylic bar. Averaged responses were then used to develop a series of simple models. Each model comprised two separate convolutions of an input with a distinct hemodynamic response function (HRF), summed together. Our results show that a simple two-component model performs better on the data than a single HRF model. This approach also enabled us to separate the vascular response into two spatially distinct components.

The model was then expanded in order to generate a more robust and widely applicable version. Information from the GCaMP signal was incorporated into the inputs for both components. This further improved the function of the model on averaged data and enabled its application to more complex situations such as spontaneous behavior and single-trial data.

We then show, through the use of pharmacological perturbations, that these two spatially and temporally distinct responses may utilize separate mechanistic pathways. Mice were imaged after treatment with pharmacological substances known to affect mechanisms of endothelial dilation and signal propagation. Drug application resulted in significantly altered responses to whisker stimulation. We then applied our model to these perturbation experiments and were able to observe changes coupling after pharmaceutical manipulation. Notably, our results showed that perturbations may affect only one of the two components, which indicates that they may be mediated through separate mechanistic pathways.

Through these findings, we demonstrate that the hemodynamic response to neural activity must be modeled using a nonlinear approach and that a simple model based on the sum of two linear convolutions accounts for these nonlinear features of the response. We develop a simple model that can be used a basis for further study of NVC and demonstrate that it can be used to better understand the effects of perturbations on coupling in vivo.