1. Introduction

Alzheimer’s disease (AD) is the most well-known form of cognitive impairment (dementia) which the early diagnosis seems crucial for the better treatment of the disease due to its irreversible nature. Among many factors that are known to be linked to AD, vascular dysfunction has recently gained some attention with its high prevalence in AD [1], for example, rapid change in vasculature, blood-brain barrier breakdown, increased tortuosity, reduced capillary density, and regional hypoperfusion [2–6]. In particular, vascular dysfunction was often observed earlier than cognitive decline or even earlier than neuropathology, suggesting the role of vascular dysfunction as a possible etiology. For example, at the preclinical stages of AD, several studies found cerebral blood flow decreases observed in the brain regions such as hippocampus, entorhinal cortex and amygdala, that are closely linked to cognition [7,8]. From population-based studies, it has been reported that this cerebral hypoperfusion is associated with accelerated cognitive decline and an increased risk of dementia [9,10]. Thus, many research groups have been putting their efforts on investigating the correlation between vascular abnormalities and cognitive impairment in AD.

To investigate vascular alterations, there are several imaging modalities such as laser speckle contrast imaging (LSCI), fluorescence microscopy, optical coherence tomography, and so on. LSCI can measure blood flow changes in cortical surface but is limited in depth-resolved volumetric imaging capability, and only provides relative change values [11,12]. For a deeper volumetric and target-specific imaging, fluorescence microscopy such as two-photon microscopy (TPM) can be a great option with the high resolution for visualizing vasculature [13,14]. However, TPM generally requires scanning over all three axes, limiting its speed for blood flow imaging. Also, having to use fluorescence dyes makes TPM not suitable for longitudinal and intact imaging in general. In this regard, spectral-domain optical coherence tomography (SD-OCT) can be a suitable option with its label-free, in-depth volumetric imaging [15], which enables us to visualize the changes in cerebral vasculature in longitudinal manner [16]. Plus, for microcirculation, doppler optical coherence tomography (DOCT) enables the quantification of absolute blood flows in penetrating arterioles and venules based on phase shifts of blood cells flowing through vessels [17]. With these advantages of SD-OCT, it is capable for providing the long-term stability in vivo imaging of vascular alterations in the brain.

In this study, we demonstrated a label-free SD-OCT imaging to thoroughly observe temporal dynamics of vascular alterations in vivo using AD transgenic mouse model (3xTg-AD) [18,19]. We acquired OCT angiogram (OCTA) and DOCT datasets starting at 11 weeks of age (WOA) then repeated every 4 weeks until 57 WOA for each mouse, along with a behavioral test to confirm cognitive impairment. As a result, we found the cognitive decline occurring at 40 WOA whereas the pial vessel diameter change occurred around 20 WOA for AD group. To determine whether vascular alteration affects cognitive function, we used 3 different drugs with vasodilating effects: nicorandil (NC), memantine (MM), donepezil (DN). These drug-treated groups showed no sign of cognitive decline and no significant change in pial vessel diameter, similar to the results of wild-type (WT) group, implying that the vasodilating effect seems to be correlated with cognitive impairment. Furthermore, we quantified the relative change in diameter and blood flows in penetrating arterioles and venules using the DOCT data. Similar to the overall tendency found in OCTA results, the change in diameter for arterioles in the AD group showed early change at around 13 WOA, whereas the change in venules seemed trivial. Interestingly, the change (from 11 to 57 WOA) in mean diameter for arterioles (3.6 µm) was about 10-fold high compared to that of venules (0.4 µm) in AD group. For blood flow change, both arterioles and venules showed the decay tendency with rapid change occurring at early stage (before 20 WOA) in the AD group. Three drug-treated groups showed the similar tendency to that of WT group in both diameter and blood flow change for penetrating arterioles and venules. Our finding supports that understanding the temporal dynamics of vascular alterations in relation to cognitive decline could serve as a basis for translational research toward early diagnosis of AD.

2. Materials and methods

2.1 Optical coherence tomography angiography (OCTA)

Spectral domain OCT system (Telesto III, Thorlabs, USA) was used to collect all experimental data. It uses a near-infrared (NIR) laser source with a center wavelength of 1310 nm and wavelength bandwidth of 170 nm which leads to an axial resolution of 3.5 µm in air (calculated). The system uses a high-speed 2048-pixel line-scan camera to achieve 147,000 A-scan/s providing a relatively large imaging depth (2.5 mm maximum). The 5× NIR objective lens (NA = 0.14, Mitutoyo, Japan) was used and produced the lateral resolution of 7 µm in air (measured with a USAF target). The field of view (FOV) had 1024 × 1024 × 1024 voxels corresponding to 3.08 × 3.08 × 3.58 mm³, and transverse and axial sampling intervals of 3 µm and 3.5 µm, respectively. For OCTA and DOCT, to minimize surface reflections, the focal plane was located at approximately 100 µm below the cortical surface. It should be noted that this focal depth may cause slight defocus on the imaged pial vessels that are located at the cortical surface, but its effect on the major outcome of our study would be minimal as the study focuses on differences in the diameter between ages and groups rather than the absolute value. The volumetric scan was repeated 5 times for reducing speckle noise. Acquired data was reconstructed to 3D images using the widely used method which involves numerical dispersion compensation, k-space interpolation, and inverse Fourier transformation [20]. After volume averaging, median and gaussian filtering, we obtained maximum intensity projection (MIP) images and calculated full width half maximum (FWHM) values for vessel diameter measurement. In detail, when the user selected a point near a vessel, our software user interface identified the centerline of the vessel, extracted 10 cross-sectional intensity profiles from 10 adjacent pixels along the centerline, fitted the average profile to a Gaussian fitting, and measured the full-width half-maximum as the diameter of the selected vessel. This averaging along the vessel centerline and Gaussian fitting-based sub-pixel measurement allows for more robust measurements than a simple counting of pixel numbers along the vessel width.

2.2 Doppler-OCT

DOCT images were acquired by repeating 8 A-scans at each (x,y) position with the temporal sampling interval of 6.8 µs, and the FOV was 512 × 512 × 1024 voxels corresponding to 1.54 × 1.54 × 3.58 mm³. After 3D reconstruction, we obtained blood flow maps using improved clutter rejection algorithm with Kasai autocorrelation method previously published [21,22]. Median and low-pass filtering were applied. Absolute blood flow of selected vessel area was calculated using the following equation:

To extract the information, we drew a circle around the selected vessel for finding the center of mass, then set the region of interest (ROI) with around 60 × 60 µm². A line profile was drawn at each of four angles, passing the center of ROI, where the four angles represent the horizontal, vertical, 45 deg, and 135 deg. We then applied a gaussian fit to the average of the four profiles for obtaining FWHM as diameter for penetrating arterioles and venules. When we selected penetrating arterioles and venules from an en face Doppler OCT image, we distinguished between the two types by considering that the anatomy of cerebral vasculature in the cortex has arterioles and venules result in positive and negative velocity values in Doppler data, respectively. To select the same vessels across different time points within each animal, we registered the en face images to a reference time point image and considered spatial distribution of the penetrating vessels and relative positions between each other as appeared in the registered images. We also considered the corresponding OCTA image of the same animal and age point when helpful. This selection approach was consistently applied across data of all animals and ages.

2.3 Novel object location test (NOL)

Mice were placed into an open-field apparatus consisted of a square arena (40 × 40 × 49 cm³). The arena is digitally divided into 4 quadrants, with the center of the northeast (NE), northwest (NW), southeast (SE), and southwest (SW) corners which serve as placement locations for two identical objects (rods, radius = 1 cm, height = 5 cm). For 5 min, each mouse’s behavior was recorded by the automated tracking software (EthoVision XT, Noldus) to calculate the exploration time determined by a circular zone (< 5 cm) from the objects’ center. Two different sessions were carried out with the delay of 20 min in between: a habituation stage for 5 min (objects placed on NE and NW) and a testing stage for 5 min (objects placed on SE and NW). The arena is cleaned with a 70% ethanol and dried in between sessions to eliminate any potential odor cues left beforehand. To determine the sign of cognitive decline, we calculated Z-score using the following equation:

2.4 Animal preparation and pharmacological treatment

All animal-based experimental procedures were reviewed and approved by the Institutional Animal Care and Use Committee of Brown University (IACUC-1612000239), according to the guidelines and policies of office of laboratory animal welfare and public health service, National Institutes of Health. We conducted a power calculation ($\alpha$ of 0.05 and 80% power), with the intra-group mean difference in vessel diameter of about 6 µm between young and old AD mice and the inter-group standard deviation of about 2 µm from the literature [24], which suggested 2 mice per group. We used 4 mice per group by considering unexpected risks associated with the longitudinal study. Our design has five groups: wild-type (WT), sham-treated AD (AD), nicorandil-treated AD (NC), memantine-treated AD (MM), and donepezil-treated AD (NC) groups. Following this design, male WT mice (n = 4, 9 WOA, 20–25 g in-weight, strain C57BL/6, Jackson Lab) and AD mice (n = 16, 9 WOA, 20 - 25 g in-weight, strain 3xTg-AD, B6;129-Tg(APPSwe,tauP301L)1Lfa Psen1TM1Mpm/Mmjax, Jackson Lab) [18,19] underwent craniotomy surgeries following the widely used protocol [25]. The animal was kept under isoflurane anesthesia (1.5%, delivered with 1 L/min O₂) during imaging. Oxygen saturation, pulse rate, and temperature were continuously monitored with pulse oximetry and a rectal probe throughout the surgical procedure and imaging experiments. The body temperature was maintained at 37 °C and the pulse rate remained within the normal range of 250 - 350 pulse/min. For pharmacological treatment, we chose three drugs: nicorandil, memantine, and donepezil (N3539, PHR1886, PHR1584, Sigma-Aldrich). The drugs were dissolved by aqueous buffer, phosphate-buffered saline directly, then administered via drinking water starting after 15 WOA, with the dose of 15, 30, 4 mg/kg/day for NC, MM, and DN, respectively. The mouse weight and administered drug amount were checked every week. We performed the longitudinal experiment involving the described OCT imaging, NOL test, and continuous drug treatment from 11 to 57 WOA, but we had to pause data acquisition for 3 months in the middle of this 1-year long experiment for a logistic reason.

2.5 Bielschowsky silver staining (BSS)

The coronal sections were deparaffinized and rehydrated, then incubated in preheated (37 °C) 20% silver nitrate solution for 25 min. While the slides (N = 4 sections per animal) were hold in warm distilled water, ammonium hydroxide 28 - 30% was added drop by drop in the silver nitrate solution until precipitate disappeared. The slides were placed back into the silver nitrate solution with added developer for 5 - 15 min until the sections became dark brown. They were then washed in distilled water for 5 min, placed in 5% sodium thiosulfate solution for 5 min, dehydrated, and mounted [26]. For quantitative analysis, we used ImageJ to binarize two coronal sections for each animal based on a threshold value by which the amyloid plaques were well segmented. We applied this threshold (80 when the max pixel intensity was 255 [8-bit image]) consistently to all the slides, then quantified area fraction of amyloid plaque burden in the hippocampal region.

2.6 Statistical analysis

The statistical results are denoted as mean value ± standard error (SE) measured after each experiment. For the statistical comparison between young and old ages (Fig. 7), the Kolmogorov-Smirnov test was used for testing normality, and then Student t-test was performed since data were normally distributed. The p-value lower than 0.05 was considered statistically significant.

3. Results

3.1 Evaluation of cognitive impairment using NOL and BSS

To confirm cognitive impairment in 3xTg-AD mice known to occur around 36 - 44 WOA [18], we conducted the NOL test every 4 weeks for all mice groups (see Materials and Methods). We estimated all of the NOL Z-scores in which the score of 0.5 (along with linear regressions) were used to determine the sign of cognitive decline in our mice groups. As shown in Fig. 1(a), blue dots represent the mean of Z-scores with standard errors (SE) for WT group, displaying a slight decrease with the slope of -0.00123 estimated by linear fit (black dotted line). The average of Z-scores changed from 0.62 to 0.56 for 11 and 57 WOA, and there was no crossing below the 0.5 threshold, implying no sign of cognitive decline. As a confirmation of the major pathology of AD, we performed BSS as shown in Fig. 1(a) inset and the mean (and SE) of regional plaque burden (fractional area) was 0.15 ± 0.02% for WT group.

 

Fig. 1. Cognitive impairment determined by NOL Z-scores and the quantification of amyloid deposits by BSS images. (a) NOL Z-score result with the fitted line (black dotted line) for (a) WT group, (b) AD group, and three drug-treated groups of (c) NC group, (d) MM group, (e) DN group. (f) Relative change of fitted slopes with WT group as a baseline. (g) BSS images with amyloid plaque visualized. Inset represents BSS image with amyloid plaque in AD group (red arrow heads). Scale bars: 500 µm and inset 50 µm.

Download Full Size | PDF

For AD group, the fitted line had the slope of -0.00476 (about 4-fold decrease compared to WT) and did cross the 0.5 threshold at around 40 WOA (denoted as red shaded region with the 5 weeks width), and the mean of Z-scores for 11 and 49 WOA decreased from 0.64 to 0.46, as shown in Fig. 1(b). The mean (and SE) of regional plaque burden for the AD group was significantly higher (1.52 ± 0.30%, p = 0.0013) than that of the WT group. A typical amyloid plaque was indicated by red arrow heads, and also represented as enlarged view (red box) as shown in Fig. 1(g) inset. In Figs. 1(c) – 1(e), the three drug-treated groups did not show neither a decreasing trend nor a crossing below 0.5 threshold, with the slopes of 0.00229, 0.00253, 0.0004 for NC, MM, and DN, respectively. From BSS analysis, the means of regional plaque burden were significantly less in three drug-treated groups than in the AD group: 0.48 ± 0.09% (p = 0.013), 0.89 ± 0.03% (p = 0.039), 0.78 ± 0.06% (p = 0.024), for NC, MM, and DN groups, respectively. In addition, we compared the relative change of the fitted slopes in NOL result between these mice groups and found AD group was the only group showing a negative value of -2.85 when compared to WT group. The three drug-treated groups had the positive values of 2.86, 3.06, and 1.35 for NC, MM, and DN group, respectively, as shown in Fig. 1(f).

3.2 Temporal dynamics of early vascular alterations in AD

We investigated WT group for quantifying pial vessel diameter change using 4 mice from 11 to 57 WOA with the total of 16 vessels (see color lines in Fig. 2(a) for an example of selected vessels, with one of them shown in detail in Fig. 2(b)). As exemplified in Fig. 2(b), we measured the diameter of each vessel at each age point by fitting the averaged cross-sectional profile of the vessel to a Gaussian function and taking the full width at half maximum (FWHM) as the diameter. The WT group (N = 16 vessels) showed a slight increase in pial vessel diameter (from 30.8 ± 2.3 µm to 37.4 ± 3.6 µm between 11 and 57 WOA), with a fitted slope of 0.078 as shown in Fig. 2(c).

 

Fig. 2. Quantification of pial vessel diameter using OCTA images. (a) Example MIP images of a selected set of age points (15, 23, 41, and 45 WOA) for one of the WT mice. Color lines indicate the selected set of same vessels over which the diameter was measured and tracked. The red boxes indicate the representative vessel with which the vessel diameter measurement process is shown in detail in (b). (b) Left: enlarged views of the representative vessel at each age point. The blue lines represent the line along which the cross-sectional profile was extracted. We extracted ten adjacent cross-sectional profiles and used their average (see Materials and Methods for details). Right: the intensity profile (circles) with its Gaussian fit (line). Red line illustrates an example of the measured FWHM as diameter. (c) Pial vessel diameter change of WT group (N = 16 vessels) with mean (blue dots) and standard error from 11 to 57 WOA. (d) MIP images for AD group. (e) Left: diameter tracking of the same vessel (numbered red boxes) for AD group. Right: intensity profile of the vessel with gaussian fit and estimated FWHM values as diameter. (f) Pial vessel diameter change of AD group (N = 11 vessels) with mean (red dots) and standard error. Red shaded region with yellow dotted vertical line represents the decay time constant τ (around 20 WOA). Scale bars: 500 µm in (a) and (d), 50 µm in (b) and (e).

Download Full Size | PDF

In contrast, the AD group (N = 11 vessels) showed a 21% decrease in pial vessel diameter from 31.6 ± 1.7 µm to 25.0 ± 4.5 µm between 11 and 57 WOA shown in Fig. 2(f). Interestingly, the tendency of decreasing was noticeably different between early and late stages with the turning point around 20 WOA, consistent with the early alterations of brain vascular volume and morphology observed in the 3xTg-AD mice [27,28]. The slopes were -0.25 and -0.08, for the early and late stage, respectively (green and magenta lines in Fig. 2(f)). The absolute value of these two slopes differs by about 3-fold. Considering this tendency, we applied the exponentially decaying function ($y = A + B{e^{( - x/\mathrm{\tau })}}$) resulting in the decay rate of 0.052 (R² = 0.74). The decay time constant (τ: corresponding to 1/e) was 19.4 (denoted as a red shaded region with yellow dotted line in Fig. 2(f)) implying that the 36.7% of the total diameter changes already occurred at 19.4 WOA. As a result, the change in pial vessel diameter occurred around 20 WOA, much earlier than that of cognitive decline observed at 40 WOA.

To further investigate vascular alterations with respect to dynamics, we acquired the 3D blood flow maps using DOCT which can distinguish between penetrating arterioles and venules. We focused the depth between 250 µm and 350 µm (to which the visualization of arterioles and venules were optimized) and extracted the flow and diameter information of same individual vessels (arterioles and venules) over time. Figure 3(a) upper represented the merged en face image of DOCT and corresponding OCTA images from two different timepoints (15 and 45 WOA). A typical arteriole and venule (denoted as cyan and a magenta boxes) for WT group are displayed in Fig. 3(a) lower. For the selected arteriole, the values for diameter and blood flow were 40 µm and 1.3 µL/min for both 15 and 45 WOA. The selected venule had the values of 25 µm and 24 µm for diameter and blood flow of 0.3 µL/min at 15 and 45 WOA in WT group. Further, we calculated the relative change in diameter and confirmed no significant change over time with the slopes of 0.0010 and 0.0009 for arterioles and venules, approximately, as shown in Figs. 3(b) and 3(c). For blood flow, we also confirmed no significant change with the slopes of 0.0004 and -0.0010 in arterioles and venules, approximately, as shown in Figs. 3(d) and 3(e).

 

Fig. 3. Measurement of vessel diameter and blood flow using DOCT. (a) Upper: merged images of DOCT and OCTA en face image for 15 and 45 WOA of WT group at the depth of 320 µm beneath the cortical surface. Lower: enlarged views of numbered boxes, arteriole (cyan), venule (magenta). Intensity profile (dotted curve) with gaussian fit (solid curve) and estimated FWHM values as vessel diameter (dotted red lines). Relative change for vessel diameter of WT group (b) for arterioles and (c) for venules. Also, for blood flow (d) in arterioles and (e) in venules. (f) Upper: merged images of DOCT and OCTA en face image for 15 and 45 WOA of AD group at the depth of 290 µm beneath the cortical surface. Lower: enlarged views of numbered boxes, arteriole (cyan), venule (magenta). Intensity profile with gaussian fit and estimated FWHM values as vessel diameter. Relative change for vessel diameter of AD group (g) for arterioles and (h) for venules. Also, for blood flow (i) in arterioles and (j) in venules. Red shaded region with yellow dotted line represents the decay time constant. Scale bar: 500 µm.

Download Full Size | PDF

On the other hand, the diameter in AD group for the selected arteriole (cyan boxes) decreased from 36 µm to 32 µm as shown in Fig. 3(f) lower. Based on the overall tendency found in pial vessel diameter change from OCTA analysis, we applied the exponential fit. For the relative change in diameter for all arterioles, we obtained the decay rate of 0.07 and τ of 12.9 WOA (denoted as a red shaded region with a yellow dotted line) as shown in Fig. 3(g). In addition, the blood flow values for the selected arteriole were 1.3 and 0.9 µL/min shown in Fig. 3(f) lower. The overall tendency showed the decay rate of 0.06 and τ of 17.9 WOA (denoted as a red shaded region with a yellow dotted line) shown in Fig. 3(i). For venules, the values for diameter and blood flow of the selected venule (magenta boxes) were 29 µm and 25 µm, 0.4 and 0.2 µL/min for 15 and 45 WOA, as shown in Fig. 3(f) lower. The relative change of the diameter in venules did not show exponential decay tendency, however, did show a slight decrease with the slope of -0.0007 using a linear fit as shown in Fig. 3(h). Despite this linearly decreasing tendency, the relative change of blood flow showed exponentially decaying tendency with the obtained decay rate of 0.08 and τ of 11.8 WOA (denoted as yellow dotted line in a red shaded region) in Fig. 3(j). Considering the overall tendency in pial vessel diameter change found in OCTA analysis, interestingly, there seemed to be more influence of penetrating arterioles over penetrating venules in diameter change. On the other hand, the relative change of blood flow in both arterioles and venules showed early decrease with exponentially decaying tendency. A reason for such decrease in cerebral blood flow is possibly due to several factors such as tortuous arteries and excessive collagen deposition in veins [29], and this could also be reflected as a different tendency found between diameter and flow changes in arterioles and venules.

3.3 Correlation between the vasodilating effect and vascular integrity in AD

To investigate the correlation between vasodilating effect and vasculature/vasodynamics over time, we administered NC, MM, and DN, known to have some vasodilating effect [30–32]. We found that these drug-treated groups showed the different tendency in the NOL Z-score compared to AD group, as shown in Fig. 1(f). Using OCTA images, the total number of selected pial vessels (color lines) were 19, 19, and 38 for NC, MM, and DN group using 4 mice, respectively, as shown in Fig. 4. Unlike AD group, the overall tendency of pial vessel diameter change in NC group showed a slightly increasing tendency with the slope of 0.065, shown in Fig. 4(d), implying vasodilating effect may have contributed to prevent the decrease in pial vessel diameter found in AD group. The similar tendency was found for MM and DN groups, with the slopes of 0.042 and 0.105, respectively, seen in Figs. 4(e) and 4(f). This result from MM and DN group seemed to similarly represent the vasodilating effects found in NC group.

 

Fig. 4. Pial vessel diameter quantification using OCTA for 3 drug-treated groups. (a) - (c) Left: a MIP image of 15 WOA. Scale bar: 500 µm. Right: a diameter tracking (intensity profile and estimated FWHM) of the same individual vessel (numbered red boxes) of 15 and 49 WOA for NC, MM, and DN groups. Scale bar: 50 µm (e) - (f) pial vessel diameter change from 11 to 57 WOA for NC (N = 19 vessels), MM (N = 19 vessels), and DN (N = 38 vessels) groups.

Download Full Size | PDF

We further investigated on vascular alterations using DOCT to compare three drug-treated groups in relation to AD group. NC group showed no significant change in diameter for both arterioles (N = 12 vessels) and venules (N = 11 vessels) with the slopes of 0.0007 and 0.0009, as shown in Figs. 5(a) and 5(d). For MM and DN group, the tendency for vessel diameter change seemed analogous to NC group as shown in Fig. 5 and Table 1. In particular, the slope for arterioles in DN group was 0.0034, about 6-fold high compared to NC and MM group. This could be possibly due to DN specifically dilating parenchymal arterioles [32], thus led to the higher slope than others. This finding supports the notion that the vasodilatory intervention may play a crucial role in maintaining the integrity of vasculature over time.

 

Fig. 5. Relative change of vessel diameter from DOCT analysis for 3 drug-treated groups using linear fit. (a) – (c) for arterioles. (d) – (f) for venules of NC, MM, and DN group.

Download Full Size | PDF

Table 1. Slopes of relative change in diameter and blood flow for arterioles and venules. The numbers (N) represent vessel numbers.

View Table

In addition, we quantified blood flow change for both arterioles and venules in relation to the vasodilating effect. For NC group, the relative change of blood flows in arterioles (N = 12 vessels) and venules (N = 11 vessels) seemed trivial with the fitted slopes of -0.0011 and -0.0004, as shown in Figs. 6(a) and 6(d). For MM and DN groups, the relative change for both arterioles and venules did not show meaningful change in Fig. 6, with the fitted slopes listed in Table 1. Taken all together, although we observed vasodilating effects shown from the results of vessel diameter change in Fig. 5, understanding how much this vasodilating effect can be correlated with blood flow change results is beyond our scope.

 

Fig. 6. Relative change of blood flow from DOCT analysis for 3 drug-treated groups using linear fit. (a) – (c) for arterioles. (d) – (f) for venules of NC, MM, and DN group.

Download Full Size | PDF

4. Discussion

We have investigated the temporal dynamics of vascular alterations such as vessel diameter and blood flow change in AD with respect to cognitive impairment over a 1-year timespan using OCTA and DOCT techniques. AD group showed the sign of cognitive decline at around 40 WOA, consistent with previous studies using the same mouse (3xtg-AD) [18]. Interestingly, the pial vessel diameter change from OCTA analysis had two different tendencies where the slope for early stage (before 20 WOA) was -0.25 (estimated by linear fit) and that for late stage (after 30 WOA) was -0.08, about 3-fold high in magnitude for the early stage. To better interpret, this trend was well fitted by an exponential function with a decay time constant τ (19.4), implying that about ∼37% of total change in pial vessel diameter occurred at 20 WOA in AD group, approximately. This result shows not only early pial vessel diameter change preceding cognitive decline (40 WOA) but also the majority of total diameter change occurring during the early stage. Regarding early vascular alterations in AD, Meyer et al. reported structural alterations in microvasculature such as vessel deformation and elimination are found at young ages (∼3 months) of APP23tg model using fluorescence TPM imaging, even when amyloid beta plaques are not yet present [28]. Thus, our finding shows its consistency with previous studies but also provides a better understanding of temporal dynamics in vasculature change.

DOCT shows its capacity of distinguishing penetrating arterioles and venules based on phase-shifts of red blood cells scattering within the vessels. It provides the absolute blood flow and diameter information from arterioles and venules. AD group showed the exponentially decaying tendency (τ of 12.9 WOA) in the relative change of diameter over time for arterioles, shown in Fig. 3(g). The amount of decrease in mean diameter for arterioles from early (before 20 WOA) to late (after 40 WOA) stage was 18.3% (p < 0.05) as shown in Fig. 7(a) left. On the other hand, for venules, only a slight linear decrease was observed over time, seen in Fig. 3(h). The decreasing amount in mean diameter from early to late stage was 1.7% (with no statistical significance) for venules, shown in Fig. 7(a) right. Interestingly, the change in mean diameter for arterioles was about 10-fold high to that of venules. Such tendencies could be possibly due to tortuous arterioles and excessive collagen depositions in venules [29,33], Thus, more influence of the diameter change in arterioles is believed to be present in our result. With respect to amyloid deposition in AD pathology, this result could be explained by majority of amyloid accumulation occurring in arterioles whereas less involvement in venules [34].

 

Fig. 7. Comparison between early and late values of vessel diameter and blood flow for AD and WT group. (a) and (c) distributions of diameters for penetrating arterioles and venules in AD and WT. (b) and (d) distributions of blood flow for penetrating arterioles and venules in AD and WT. The chart shows the mean value (square inside the box), median value (line inside the box), the interquartile range (box), and the minimum and maximum values within 1.5 times the interquartile range (whiskers).

Download Full Size | PDF

Furthermore, we measured absolute blood flow for both penetrating arterioles and venules in AD group. For the relative change in arterioles, τ was 17.9 WOA, implying early alterations are highly correlated with both diameter and blood flow changes. For venules, the relative change of blood flow showed τ of 11.8 WOA whereas the diameter change shows a relatively slow decreasing tendency. The amount of decrease in mean blood flow from early to late stage for arterioles was 61.3% (p < 0.001) as shown in Fig. 7(b) left. For venules, the amount of change was 33.3% (p < 0.05) as shown in Fig. 7(b) right. Blood flow and diameter change for venules show a low correlation in overall tendency of alteration. However, it still presents that the blood flow change in venules show an exponentially decaying trend similar to arterioles in AD group. Regarding early blood flow change in AD mice, Guo et al. recently reported that age- and brain region-associated alterations of cerebral blood flow in early stages (at around 2 and 3.5 months) of AβPPSWE/PS1ΔE9 transgenic mice [35]. In contrast, WT group showed no significant change in diameter and blood flow for arterioles and venules over time shown in Figs. 7(c) and 7(d), and consistent with OCTA analysis shown in Fig. 2(c). On the other hand, we had to pause data acquisition for 3 months in the middle of the 1-year long experiment, leading to missing datapoints as can be seen in Figs. 1–6. But this missing data at the middle would not significantly weaken our major conclusion of Fig. 7 which compares the vascular properties between young and old ages. Also, the missing data may not make the fitting of the whole-period data to a linear line untrustworthy, unless a vascular property should exhibit a severely non-monotonic trace like a Gaussian function.

To further understand vascular alterations in relation to cognitive decline (shown at around 40 WOA), we used three drugs that are known to have vasodilating effects and administered (via drinking water) to the same 3xTg-AD mice. NC is a typical balanced vasodilator which acts as both arterial K + ATP channel opener and NO donor. It improves spatial learning and memory in chronic cerebral hypoperfusion-induced vascular dementia [30]. For two FDA-approved drugs, MM improved spatial learning and memory in a rodent model of chronic cerebral hypoperfusion [31]. DN dilates cerebral parenchymal arterioles via the selective activation of neuronal nitric oxide synthase [32]. These vasodilating drugs are believed to be mitigating cognitive impairment as shown in Fig. 1(f). As expected, for both vasculature and vasodynamics, there was no decreasing tendency found in the drug-treated groups although these groups showed high standard errors. This is because of our inevitable choice of drug administration route (via drinking water) due to headpost installation necessary for stable OCT imaging, possibly leading to some variations in determining the exact amount of drug administered for each animal. Nevertheless, within the drug-treated groups, the slope for arteries diameter change in DN group was about 6-fold high compared to NC and MM group as shown in Fig. 5. This could be possibly due to DN specifically dilating parenchymal arterioles [32], resulting in the higher slope than others. These findings support the notion that pharmacological intervention with vasodilating effects, when preceded the events of vasoconstriction and hypoperfusion, may play a crucial role in preventing cognitive decline in AD. Regarding the specific model used in this paper, studies reported that 3xTg-AD mice have sex difference in terms of their pathology [36,37] and treatment response [38,39]. Yet, the present study only used male mice to focus more on the vasodilation effect and cognitive decline. Extending the described approach to both sexes and considering the sex as a key biological variable will be important in future work.

5. Conclusion

In summary, we have demonstrated the capability of SD-OCT imaging for label-free in vivo studies to investigate vascular alterations in AD. With a 3D volumetric imaging, OCTA can measure the diameter change of individual pial vessels over time, and the DOCT technique can distinguish these vessels into arterioles and venules and provide diameter and absolute blood flow measurements. More importantly, we found that vascular alterations such as vessel diameter and absolute blood flow change precede cognitive decline in AD group. Interestingly, the amount of mean diameter change over time for penetrating arterioles was about 10-fold high to that of penetrating venules in AD group, suggesting the importance of monitoring the arteriolar alterations. In addition, by administering drugs with vasodilating effect, we confirmed the correlation between vascular alterations and cognitive decline in AD. We believe that monitoring vascular alterations could provide better insights toward translational research in achieving early diagnosis but also in developing therapeutics strategy for AD.