Blood oxygen level dependent (BOLD) signal in functional magnetic resonance imaging (fMRI) is frequently used as a proxy for underlying neural activity. Although this is a plausible assumption for experiments where a task is performed, it may not hold to the same degree for conditions of fMRI recording in a task-free, "resting" state where neural synaptic events are weak and, hence, neurovascular coupling and endothelial vascular factors become more prominent (Hillman Annu Rev Neurosci 37:161-181, 2014, 10.1146/annurev-neuro-071013-014111). Here we investigated the magnitude of change of BOLD in consecutive samples over the acquisition time period (turnover of BOLD, "TBOLD") by first-order differencing of single-voxel BOLD time series acquired in 70 areas of the cerebral cortex of 57 cognitively healthy women in a task-free resting state. More specifically, we evaluated (a) the variation of TBOLD among different cortical areas, (b) its dependence on age, and (c) its dependence on the presence (or absence) of the neuroprotective Human Leukocyte Antigen (HLA) gene DRB113 (DRB113:02 and DRB113:01). We found that TBOLD (a) varied substantially by 2.2 × among cortical areas, being highest in parahippocampal and entorhinal areas and lowest in parietal-occipital areas, (b) was significantly reduced in DRB113 carriers across cortical areas (from ~ 15% reduction in orbitofrontal cortex to 2% reduction in cuneus), and (c) increased with age in noncarriers of DRB113 but decreased with age in DRB113 carriers. These findings document significant dependencies of TBOLD on cortical area location, HLA DRB1*13 and age.