How do capillary occlusions impact brain microcirculation in Alzheimer's disease ? Numerical Simulations and Experimental validation

The most common form of dementia is Alzheimer's disease, making it a major health problem. While studied intensively, no cure or even clear diagnostic procedures have been found yet (Nelson JNEEN 2012). A recent study (Iturria-Medina Nat Com 2016) has shown that a significant decrease in blood flow arises in the onset of Alzheimer's disease. In vivo two photon microscopy of cortical vasculature in APP/PS1 mouse models suggests that the mechanism underlying this blood flow reduction is capillary occlusions due to leucocytes adhering to inflamed blood vessel walls (Cruz-Hernandez Neuroscience meeting 2016). However, these techniques generate massive amounts of anatomical and functional data that are difficult to interpret without proper theoretical and numerical frameworks. Here, we develop a model describing blood flow in large microvascular networks that accounts for complex blood rheology (hematocrit-dependent viscosity and non-linear phase separation at bifurcations, Pries Microvasc Res 1989). We validate it by comparison with in vivo measurements in mice and extrapolate the model to humans. We then quantify the impact of capillary occlusions on regional perfusion for both species. The numerical simulations, based on an iterative algorithm derived from Newton's fixed point method (Lorthois, Neuroimage 2011) are run in 4 postmortem mice anatomical networks with more than 10,000 vessels each acquired by two-photon imaging (Tsai J Neuroscience 2009). For human we use a dataset composed of 300 μm thick slices (over 20,000 vessels per slice) acquired by confocal microscopy (Cassot Microcirculation 2006). We correct post-mortem vessel variations in both dataset using shape-preserving transformations based on available in vivo measurements. We validate our model and corrections by direct comparison with in vivo red blood cell velocity measurements performed in mice (Santisakultarm AJPHCP 2012, Taylor JCBFM 2016). We then investigate, for both humans and mice, the impact of an increasing percentage of capillary occlusions on regional perfusion, defined as the sum of flows in all arterioles supply the studied domain. We observe a clear linear dependence in both species, with similar slopes and no threshold effect. Our results imply that even a small percentage of occluded vessels (2-4%) yield a drastic reduction of regional perfusion (5-12 % in mice and humans), which is of the same order of magnitude as observed experimentally in APP/PS1 mice. Finally, we show that these slopes are controlled by the capillary network connectivity and spatial distribution of occlusions across the network, while only slightly affected by uncertainties on vessel diameters