2.1 Experimental design

Wild-type (WT) and human ApoE-Ɛ4 knock-in (TGRA8960) female Sprague Dawley rats were obtained from Horizon Discovery (St. Louis, MO). Rats were housed in pairs in Plexiglas cages, in ambient temperature (22–24°C), on a 12h:12h light-dark cycle (lights on at 7:00 am). Food and water were provided ad libitum. Rats were imaged during the light phase of the circadian cycle. All rats were acquired and cared for in accordance with the guidelines published in the National Institute of Health (NIH) Guide for the Care and Use of Laboratory Animals. All methods and procedures described below were pre-approved by the Northeastern University Institutional Animal Care and Use Committee. The protocols used in this study adhere to the ARRIVE guidelines for reporting in vivo experiments in animal research [33].

This cross-sectional study was performed on a subset of female rats. Images were obtained at 8 months of age. Two groups (n = 5 WT and n = 6 APOE4) were imaged. Animals were effectively the same size with no significant difference in weight or brain sizes (S1 Fig). After this study, the rats were returned to their housing for use in other studies.

2.2 Magnetic resonance imaging

Images were obtained using a Bruker Biospec 7.0 T/20 cm USR horizontal magnet (Bruker, Billerica, MA, USA) equipped with a 20-G/cm magnetic field gradient insert (ID = 12 cm, Bruker). On the day of imaging, rats were anesthetized with 1–2% isoflurane, fixed with a tail-vein catheter, and secured into a custom-built restraining system with a quadrature transmit/receive volume coil with 3cm diameter (S2 Fig). During the scanning session (S3 Fig), rats were given an i.v. bolus of about 14 mg/kg Fe (roughly 2 times the clinical dose), adjusted for each rat to produce an initial blood concentration of 200 μg/ml Fe, assuming 7% blood by body weight. Three averages pre- and post-contrast were obtained with high signal-to-noise ratios (SNR). As seen in Fig 1, The QUTE-CE method creates snapshots of all cerebral arterial and venous vasculature, independent of flow velocity, vessel orientation, or arterial or venous route. Pre-contrast images differentiate few anatomical features, except for time-of-flight effects limited to the arteries at the periphery of the image, prior to magnetization reaching steady state from the hard pulse.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. QUTE-CE MRI angiograms.

Maximum intensity projection images (MIPs) of the whole rat head at 8 months are displayed (a) pre-contrast and (b) post-contrast 14mg/kg ferumoxytol. Unique contrast-enhanced vascular MIPs are obtained, and segmentation of the brain from the pre- and post-contrast images (c and d, respectively) demonstrates that time-of-flight signal enhancement is limited to the arteries at the periphery of the field of view and so the signal does not contribute to biomarker measurements.

https://doi.org/10.1371/journal.pone.0256749.g001

3D UTE MRI acquisition parameters for all scans were consistent. Parameters time-to-echo (TE)/repetition time (TR)/flip-angle (FA) = 13 μs/4 ms/20°, high RF pulse bandwidth (BW) block pulse at 200 kHz, 3×3×3 cm³ field of view (FOV), 200×200×200 matrix size, and 125,456×3 radial projections to produce 150 μm³ isotropic resolution images, with a total scan time of 25 minutes and 5 seconds for three averages.

2.3 Analysis

As previously described [28], the blood volume under steady-state conditions for each voxel (Eq 3) was segmented into 173 anatomically distinct regions using a 3D MRI Rat Brain Atlas © (Ekam Solutions LLC, Boston, MA). From these measurements, two separate structural biomarkers were obtained: the statistical mean per region, as the quantitative calculation (QC) of cerebral blood volume (QC-CBV); and second, the average blood volume in voxels filled primarily with small vessels (QC-SVD). This measure of the microvascular density was obtained by fitting the QC-CBV histogram distribution using a bin size 0.05 QC-CBV fraction and fitting a gaussian around the mode (max of histogram) using a width of 0.10 using the MATLAB function PeakFit [28,34]. This regional mode is considered to reflect small vessel density, given the resolution and neuroanatomical structure of the global vascular distribution. In this study, we focused on the microvasculature, but an analysis on the QC-CBV of each neuroanatomical region can be found in S1 Table.

Using software developed at Northeastern University Center for Translational Neuroimaging, in collaboration with Ekam Imaging, the 173-region atlas was custom-fit to each rat dataset using an affine transformation of the atlas to image space. Once the atlas was co-registered to the images, custom MATLAB code was used to analyze individual brain regions.

We conducted a mixed-design ANOVA to analyze the state of the vasculature between genotypes. Next, we performed our planned pairwise comparisons on each brain region in SPSS version 27. We made corrections to control false positives by applying the two-stage step-up method of Benjamini, Krieger and Yekutieli to each family of hypotheses with the false discovery rate set to 0.05 [35] %. This correction was selected over more conservative methods, such as Bonferroni, in order to mitigate significant increases in type-2 error, while still giving control on the amount of type-1 error present [36]. All multiple comparisons correction was performed in GraphPad prism version 8. Power analysis a priori demonstrated that the appropriate data size that can be used to draw our conclusions is 4 animals per group with an alpha 0.05 and power of 85%.

Each intensity image was corrected for B1^- field inhomogeneity along the z-axis with data from a homogenous copper sulfate liquid phantom in a 10ml tube (S4 Fig) and motion aligned spatially using the MATLAB SPM12 toolbox developed at UCL (http://www.fil.ion.ucl.ac.uk/spm/) using nearest neighbor interpolation to preserve original intensity values. The intensity will vary due to B1 inhomogeneity, and therefore both the coil sensitivity (B1^-) and the flip-angle distribution (B1⁺) should be measured throughout the image volume. Therefore, some limitations exist with the current study that are discussed in detail in the Discussion.

QUTE-CE MRI is the combination of acquisition with an optimized 3D UTE pulse sequence to produce a quantitative signal and intra-vascular contrast agent (CA) to selectively enhance the vascular compartment. B₁ field correction is implemented to render a highly quantitative vascular-specific signal throughout the imaging volume [27,28]. The QC-CBV is calculated using a two-volume blood-tissue model. The two-volume model assumes that the measured voxel intensity, I_M, is a linear combination of contributions from the blood and tissue compartments. Thus, the measured signal intensity in each voxel is given by, (1) where I_B is the blood intensity, I_T is the tissue intensity, f_B is the voxel fraction of blood, and f_T is the voxel fraction of tissue. The measured change in intensity between any two scans is given by, (2)

By selecting a CA confined to the blood plasma, it is possible to vary I_B with intravascular contrast without changing I_T, such that but , where prime denotes the measured intensity after CA injection. The blood fraction after CA injection can be measured as, (3) Where I_M is the pre-contrast image intensity and is the post-contrast image before neurofunctional modulation. I_B and are the B₁-corrected blood intensity values pre- and post-contrast, respectively. QC − CBV₀ is therefore a measure of the steady state partial blood volume used for structural biomarkers.

To determine if there are changes in QC-CBV across different neurological states (e.g., 5% CO₂ administration), then the need for multiple pre-contrast images in the corresponding state is avoided by using the calculated , such that the time-varying QC-CBV thereafter can be written as, (4)

Here, the subscript “₀” represents values during the first post-contrast image. The second term in Eq 4 is ≈ 0 because , since the tissue compartment is not enhanced by CA unless the blood brain barrier (BBB) is compromised, in which case an increased intensity would reflect CA deposition [31]. Eq 3 is applied to calculate .

Blood intensity values were measured along a large vessel, the superior sagittal sinus (SSS). The region-of-interest (ROI) identified post-contrast using a rough ROI followed by selection of maximum intensity voxels along the SSS (1 per slice) and the average value per image was used (S5 Fig). All pre-contrast I_B values obtained with this method can be found in S6 Fig and post-contrast blood values for 8 months S7 Fig.

4.1 Vascular abnormality

At 8 months, approximately 22 human years [32], we detected increased signal in 48.3% of the brain associated with the APOE4 phenotype (q<0.05, multiple comparisons). Our supplemental exploration including large vessels demonstrated a similar, but far less pronounced increases (18.8% of the brain for p<0.05 with no correction). Researchers increasingly recognize small vessels as important pathophysiological indicators in ADRD development [8,40]. Old APP23 mice present with disrupted blood flow and abnormalities in microcirculatory architecture, including angiogenesis with disease progression [41]. APP23 mice also have increased vascular density associated with amyloid plaques [42]. Even studies in cerebral ischemia show a pattern of newly formed vessels at the border of infarcts within 2 to 3 days [43]. Tau models of AD also present with small vessel abnormalities and increased small vessel density in the cortex and elsewhere. Interestingly, microvessel density has been found to increase from twelve to fifteen months but decline thereafter in one rodent model [44].

Our results at 8 months of age support the findings of early hyper-vascularization. It is possible that this early increased small vessel density may serve as a compensatory mechanism for vascular dysfunction within the neurovascular unit, such as reduced vasodilatory reactivity [45] or impaired molecular signaling to pericytes [6]. Further studies should be performed to examine these possibilities. Recently, Ferris et al., found that six-month old APOE4 rats of the model used here exhibited hyper-connectivity, which is consistent with our findings. An increased microvascular density could explain the increased BOLD signal found in young human carriers both at rest and during memory tasks, despite having lower CO₂-induced vascular reactivity [46]. Ostergaard suggested that brain capillary dysfunction underlies the development of a neuronal energy crisis which triggers AD due to reduced capillary transit time [47,48]. While the molecular and cellular mechanisms for neurovascular dysfunction and related compensation remain mostly unknown at present for APOE4 carriers [49] a denser microvascular network—one with more red blood cells and surface area for oxygen exchange—could explain the dramatic BOLD signal response found in the fMRI signal of young APOE4 carriers. Finally, it is important to acknowledge the possibility that this trend eventually reverses towards hypo-vascularization, which is more commonly associated with neurodegeneration [50].

Additionally, the role of blood -brain -barrier (BBB) leakage as a possible contributor to these findings is one of great interest. Literature suggests that BBB leakage is positively correlated to aging, cognitive decline and AD in APOE4 humans [51] BBB leakage was reported in 6-month-old ApoE4 mice of unknown gender [52], but it is hard to determine the implications for our study on female APOE4 rats. Female rodents are known to progress differentially than males in APOE4-associated pathology, and a study by Ferris et al, it was found that these APOE4 females were not cognitively dysfunctional at 6 months [39]. Retrospectively considering the possibility of BBB leakage modulation on the QUTE-CE signal, we performed a small study on available rats (n = 5; APOE4 16-month-old (~40 human years) females; see S11 Fig). These animals were on a high fat diet, and older, both of which would potentially exacerbate any leakage. Though, even at this later stage there was still no indication of BBB leakage in this animal model with QUTE-CE MRI using previously described methods [31]. These findings suggest that it is less likely that BBB leakage had a significant contribution to our measurements in 8-month-old APOE4 animals. Even with this in consideration, further exploration with histological analysis is warranted to adequately determine the biological cause of the detected vascular abnormality.

4.2 Limitations of this study

This is a small cross-sectional study; however, we were still able to identify key differences in the APOE4 model due to robust modulations in a quantitative signal. Another limitation here is that it is possible that additional abnormality was present but not detected with p-value tests. On Bruker scanners, we expect that signal-to-noise ratio (SNR) may be augmented by 4 or 8-fold with current available commercial hardware (phased array surface head coil and cryoprobe, respectively). The best combination of hardware for this experiment would be a surface coil, for high sensitivity, combined with a large volume excitation coil, for robust mapping of B1^-/+ fields. In this study, coil inhomogeneity in the x-y plane was ignored in the volume coil because it is rather homogenous. The flip-angle was not measured, which is important for quantitative measurements, but B1⁺ inaccuracy resulting from this comparison will be minimal due to geometric consistency (S1, S2 and S4 Figs). Sex is a limiting factor in this study; this study focused only on female rats, however, more than 60% of humans afflicted with AD and other dementias over the age of 65 are women. Likewise, the Ɛ4 allele is, first, found more often in women with AD than men [53]; second, associated with an earlier age of familial AD onset [54]; and, third, presents an overall greater risk for developing AD [55].

4.3 Implications towards clinical translation

We have investigated the use of QUTE-CE MRI biomarkers for ApoE-Ɛ4 (APOE4) induced vascular abnormality in a knock-in (TGRA8960) female rat model. Our results suggest that quantitative assessment of vasculature is important to evaluate brain health. As research interest has been growing towards vascular causes of ADRD, QUTE-CE MRI is readily translatable to the clinic and could prove to be a uniquely important tool in non-invasive studies of vascular pathogenesis.