Recently, a follow-up on a subsample of participants of the Maastricht Aging Study (MAAS) (25) was conducted, in which 61 individuals who had shown no evidence of cognitive or functional decline were included (Mini-Mental State Examination score ³ 25; Disability Assessment for Dementia score > 90%; no diagnosis of dementia, mild cognitive impairment or other psychiatric or neurological disorders; no structural brain abnormalities; no cognitive impairment due to substance abuse). All participants provided written informed consent before participation. For the current study, we used the imaging data of the 10 youngest (median age = 53.5 years, age range = 47 - 56) and the 10 oldest (median age = 73.5 years, age range: 70 - 91) male participants (Additional file 1, Table 1.1.). Previous studies have demonstrated that the estrous cycle influences CVO activity (26-28), so only male individuals were included in this feasibility study.
2.2. MRI acquisition
Anatomical and DCE MRI data were acquired using a 3 Tesla MRI system (Achieva TX, Philips Healthcare, Best, the Netherlands) with a 32-channel head coil. The imaging protocol included a 3D T1-weighted inversion recovery fast gradient echo (repetition time (TR) of 8 ms; inversion time (TI) of 800 ms; echo time (TE) of 4 ms; flip angle of 8°; 1 mm cubic voxel size) for anatomic reference; a 3D T2-weighted fluid attenuation inversion recovery (FLAIR) (TR/TI/TE of 4800/1650/290 ms; flip angle of 90°; 1 mm cubic voxel size) for localizing the CVOs; and a dual-time resolution dynamic contrast-enhanced (DCE) MRI acquisition. The dual-time DCE MRI protocol consisted of two nested pulse sequences, a slow and a fast sequence with a saturation recovery preparation pulse, as described earlier (29). In short, the fast sequence used a short dynamic scan interval of 3.2 s during the steep signal changes in initial circulations of the contrast agent, while the slow sequence used a longer interval of 30.5 s during the later extravasation phase when the signal changes are much slower. Before contrast administration (pre-contrast), scans of both sequences were acquired. Subsequently, a bolus injection of gadolinium-based contrast agent was performed during the fast sequence (0.1 mmol/kg gadobutrol, Gadavist®, Bayer AG, Leverkusen, Germany), intravenously in the antecubital vein (injection rate 3 mL/s, 20 mL saline flush). The fast sequence consisted of 29 volumes (TR/TE/delay time (TD) 5.3/2.5/120 ms, voxel size 2 x 2 x 5 mm) and the slow sequence consisted of 30 volumes (TR/TE/TD 5.6/2.5/120 ms, voxel size 1 x 1 x 2 mm). To minimize partial volume, fold over, and inflow effects (of the sagittal sinus superior), an odd number of sagittal orientated slices, 11 for the fast sequence and 75 for the slow sequence, was acquired with the frequency-encoding direction in the craniocaudal direction. T1-mapping with variable delay time settings was performed prior to contrast administration and DCE imaging to enable the conversion of the contrast-enhanced tissue signal intensities to contrast agent concentrations (30).
2.3. Brain regions of interest
One researcher (I.C.M.V.) was trained by an experienced neuroradiologist (A.A.P.) in determining the locations of the CVOs on the basis of brain anatomy from the mid-sagittal anatomical FLAIR images. Recognizable anatomical identification marks were used to determine the location of each CVO (e.g. the hypophysis and mammillary body for the NH and ME; the anterior commissure for the OVLT; the hippocampus and fornix for the SFO; the fourth ventricle for the AP), and a limited number of voxels on this exact location was selected, which were saved as region-of-interest (ROI).
Additionally, a ROI was placed in the neck muscle, to observe the temporal enhancement time curves for qualitative comparisons of the contrast agent distribution. As quantitative control, the white and gray matter regions were selected, which were segmented using automated software (FreeSurfer, version 6.0.0 (31)). The FreeSurfer segmentation was visually checked by one researcher (I.C.M.V.) with manual adjustments. From the segmented brain regions, the total gray matter (including cortical gray matter, deep gray matter (thalamus, caudate nucleus, putamen, pallidum, amygdala, and accumbens area), and hippocampus) and total white matter volume were extracted (31).
The slow and fast dynamic series were motion corrected and spatially aligned using a linear registration procedure with six degrees of freedom (FLIRT, FMRIB’s linear registration tool), with the average pre-contrast slow volume as reference. Next, the motion-corrected dynamic series, the FLAIR images and tissue masks, and T10-map were subsequently registered onto the participant’s structural T1-weighted images.
Individual vascular input functions (VIFs) were extracted from manually (I.C.M.V.) selected voxels (≥20) in the superior sagittal sinus (32, 33). Conversion of MRI signal enhancement to contrast agent concentration was performed differently for the VIF and tissue, and has been described earlier in detail (34). In short, the VIF signal-to-concentration conversion was implemented using in vitro data (diluted MnCl2 stock solution with different gadobutrol concentrations (1-40 mM), baseline T1 relaxation time of 1650 ms, comparable to human blood), whereas the conversion to contrast agent concentration in tissue was performed assuming a linear relationship and a tissue relaxation time calculated from the T10-map. Representative contrast agent concentration maps pre- and post-contrast agent injection are shown in Figure 3. A video of the dynamic contrast agent concentration maps during the whole DCE MRI sequence is included as Additional file 2.
2.5. Pharmacokinetic modeling
The CVOs are the points of communication between the blood plasma and the brain tissue and their functionality involves extensive exchange with the blood circulation comprising influx and reflux of solutes (4, 6). To take both the influx and reflux into account, we applied the extended Tofts model (ETM) to the CVO data, which is a two-compartmental model with a blood compartment and an interstitial compartment with bidirectional transport between these compartments (35). The CVOs are expected have a high permeability due to the lack of a blood-brain barrier and thus a sufficiently high signal-to-noise ratio. Therefore, the ROI averaged concentration-time data per CVO were fitted using the ETM as implemented in ROCKETSHIP (36) (fitting parameters for Ktrans: starting value of 0.001 min-1; lower and upper bound value of -2 and 2 min-1; maximum number of iterations of 50; and a function tolerance of 10-12) to obtain Ktrans [min-1], the transfer constant from blood plasma to extracellular, extravascular space as measure of permeability, vp [-], the volume fraction of blood plasma within a ROI as measure of perfusion, and ve [-], extravascular space volume fraction as a measure of uptake capacity and retention, for each CVO. The ETM provided sufficiently good fits for both secretory and sensory CVOs (Figure 4).
Next, to check whether the ETM sufficiently fitted the CVO data, each fit was visually checked (I.C.M.V.) and rated as ‘good’, ‘doutbful’ or ‘bad’, based on how well the modeled curve matched the data points and whether the time-course of the contrast-enhancement curve was sensible (Table 2).
The graphical Patlak method was used to calculate the leakage rate (influx) of the contrast agent into the interstitial space of the white and gray matter. This method assumes no reflux from the brain back to the blood and has been demonstrated to be most suitable for assessing pharmacokinetic parameters of normal-appearing brain tissue (37). Therefore, we applied the graphical Patlak method to assess Ktrans [min-1] and vp [-] in each voxel in the white matter and gray matter. As not in all voxels significant transfer from blood to brain could be measured due to low Ktrans values in combination with the strong influence of noise, histograms of the Ktrans values in the white and gray matter were created. These histograms were subsequently corrected for noise (19), after which the mean Ktrans was calculated as a representative measure for the permeability in the whole white and gray matter region. This approach has been applied previously in healthy controls (19). For the CVOs, the graphical Patlak method did not appear to be best suitable, as due to their exchange function, reflux must also be taken into account for these structures.
In addition to pharmacokinetic analyses, one- and ten-minute areas under the curve (AUCs) [mM·min] were calculated as proxies of gadolinium-based contrast agent wash-in during the circulation phase, and retention during the accumulation phase, respectively. Contrary to the pharmacokinetic parameter, the AUC is not dependent on the type of pharmacokinetic analysis applied and can serve as a data-driven, thus model-free, approach for characterization of contrast enhancement.
To ensure that the transfer constant (Ktrans) of the CVOs gave a realistic representation of the data, the analyses were performed excluding the values obtained from fits that were classified as ‘bad’. Post-hoc, the analyses were repeated also excluding any fits for which it could be doubted whether they were sufficiently good, to see if this would change the results.
Since our participant sample was relatively small and the transfer constants were not normally distributed for each ROI, non-parametric tests were conducted (Part 1: Wilcoxon signed-rank test; Part 2: Mann-Whitney test). All statistical analyses used a level of significance of p<0.05, and were performed with commercial software (SPSS , version 24.0, IBM Corp., Armonk, NY, USA).
2.6.1. Part 1:transfer constants of the CVOs and normal-appearing brain matter
We expected the CVOs to have strong contrast enhancement and significantly positive transfer constants, so a transfer constant significantly higher than 0.
The transfer constants and AUCs were calculated for all CVOs separately, but also for all secretory CVOs combined and all sensory CVOs combined. The Wilcoxon signed-rank test was conducted, comparing the Ktrans in the secretory and sensory CVOs to a hypothesized median of 0. If a significant result was obtained, post-hoc analyses were conducted to see which specific CVOs were significantly different from 0.
As additional analyses, Wilcoxon signed-rank tests were used to compare the Ktrans between the secretory and sensory CVOs, and between the CVOs and the white and gray matter. Again, a significant result was followed by post-hoc analyses to determine which specific CVOs had a significantly different permeability from the other CVOs, or from the white or gray matter.
2.6.2. Part 2: Age differences
For the second part, we investigate whether the transfer constant (Ktrans) differed between the older and middle-aged group.
For this between-subjects test, the Mann-Whitney test was used and the difference between the older and middle-aged group was assessed within the combined secretory CVOs, combined sensory CVOs, white matter and gray matter. If a significant result was obtained for either the combined secretory or sensory CVOs, post-hoc analyses were conducted to see for which specific CVOs the age groups differed most strongly.