Modeling the impact of neurovascular coupling impairments on BOLD-based functional connectivity at rest

Functional magnetic resonance imaging (fMRI) of blood oxygenation level dependent (BOLD) signals during the resting-state is widely used to study functional connectivity (FC) of slowly fluctuating ongoing brain activity (BOLD-FC) in humans with and without brain diseases. While physiological impairments, e.g. aberrant perfusion or vascular reactivity, are common in neurological and psychiatric disorders, their impact on BOLD-FC is widely unknown and ignored. The aim of our simulation study, therefore, was to investigate the influence of impaired neurovascular coupling on resting-state BOLD-FC. Simulated BOLD signals comprising intra- and extravascular contributions were derived from an adjusted balloon model, which allows for independent definitions of cerebral blood flow (CBF) and cerebral metabolic rate of oxygen (CMRO₂) responses, being elicited by a synthetic oscillatory input signal with low frequency (0.05 ​Hz) amplitude modulations. BOLD-FC was then defined by correlations between physiological reference BOLD time curves (seeds of seed-based BOLD-FC) and the test BOLD time curves (targets of BOLD-FC) featuring altered physiological variables (CMRO₂, CBF, cerebral blood volume (CBV)). Impact of impaired neurovascular coupling on BOLD-FC was investigated for three different scenarios with independent changes in (1) CBF and CMRO₂amplitudes, (2) CBF and CMRO₂delays, and (3) coupling between CBF and CBV. For scenario 1, we found 'linear' influences of CMRO₂ and CBF amplitudes on BOLD-FC: for a given CMRO₂ amplitude, BOLD-FC changes from negative to positive FC with increasing CBF amplitude, and increasing CMRO₂ amplitude simply shifts this dependence linearly. For scenario 2, CMRO₂ and CBF delays had a complex 'non-linear' effect on BOLD-FC: for small CMRO₂ delays, we found that BOLD-FC changes from positive to negative BOLD-FC with increasing CBF delays, but for large CMRO₂ delays positive BOLD-FC simply diminishes with increasing CBF delay. For scenario 3, changes in CBF-CBV coupling have almost no effect on BOLD-FC. All these changes were not critically influenced by both signal-to-noise-ratio and temporal resolution modulations. Our results demonstrate the importance of alterations in neurovascular coupling for aberrant resting-state BOLD-FC. Based on our data, we suggest to complement BOLD-FC studies, at least of at-risk patient populations, with perfusion and oxygenation sensitive MRI. In cases where this is not available, we recommend careful interpretation of BOLD-FC results considering previous findings about hemodynamic-metabolic changes. In the future, accurate modeling of the hemodynamic-metabolic context might improve both our understanding of the crucial interplay between vascular-hemodynamic-neuronal components of intrinsic BOLD-FC and the evaluation of aberrant BOLD-FC in brain diseases with vascular-hemodynamic impairments.