Filter exchange imaging with crusher gradient modelling detects increased blood-brain barrier water permeability in response to mild lung infection.

Blood-brain barrier (BBB) dysfunction occurs in many brain diseases, and there is increasing evidence to suggest that it is an early process in dementia which may be exacerbated by peripheral infection. Filter-exchange imaging (FEXI) is an MRI technique for measuring trans-membrane water exchange. FEXI data is typically analysed using the apparent exchange rate (AXR) model, yielding estimates of the AXR. Crusher gradients are commonly used to remove unwanted coherence pathways arising from longitudinal storage pulses during the mixing period. We first demonstrate that when using thin slices, as is needed for imaging the rodent brain, crusher gradients result in underestimation of the AXR. To address this, we propose an extended crusher-compensated exchange rate (CCXR) model to account for diffusion-weighting introduced by the crusher gradients, which is able to recover ground truth values of BBB water exchange (k(in)) in simulated data. When applied to the rat brain, k(in) estimates obtained using the CCXR model were 3.10 s(-1) and 3.49 s(-1) compared to AXR estimates of 1.24 s(-1) and 0.49 s(-1) for slice thicknesses of 4.0 mm and 2.5 mm respectively. We then validated our approach using a clinically relevant Streptococcus pneumoniae lung infection. We observed a significant 70 ± 10% increase in BBB water exchange in rats during active infection (k(in) = 3.78 ± 0.42 s(-1)) compared to before infection (k(in) = 2.72 ± 0.30 s(-1); p = 0.02). The BBB water exchange rate during infection was associated with higher levels of plasma von Willebrand factor (VWF), a marker of acute vascular inflammation. We also observed 42% higher expression of perivascular aquaporin-4 (AQP4) in infected animals compared to non-infected controls, while levels of tight junction proteins remain consistent between groups. In summary, we propose a modelling approach for FEXI data which removes the bias in estimated water-exchange rates associated with the use of crusher gradients. Using this approach, we demonstrate the impact of peripheral infection on BBB water exchange, which appears to be mediated by endothelial dysfunction and associated with an increase in perivascular AQP4.