Impairment of cerebral vascular reactivity and resting blood flow in early-staged transgenic AD mice: in vivo optical imaging studies

Background Alzheimer’s disease (AD) is a neurodegenerative disorder with progressive cognitive decline in aging individuals that poses a significant challenge to patients due to an incomplete understanding of its etiology and lack of effective interventions. While “the Amyloid Cascade Hypothesis,” the abnormal accumulation of amyloid-β in the brain, has been the most prevalent theory for AD, mounting evidence from clinical and epidemiological studies suggest that defects in cerebral vessels and hypoperfusion appear prior to other pathological manifestations and might contribute to AD, leading to “the Vascular Hypothesis.” However, assessment of structural and functional integrity of the cerebral vasculature in vivo in the brain from AD rodent models has been challenging owing to the limited spatiotemporal resolution of conventional imaging technologies. Methods We employed two in vivo imaging technologies, i.e., Dual-Wavelength Imaging (DWI) and Optical Coherence Tomography (OCT), to evaluate cerebrovascular reactivity (CVR; responsiveness of blood vessels to vasoconstriction as triggered by cocaine) in a relatively large field of view of the cortex in vivo, and 3D quantitative cerebrovascular blood flow (CBF) imaging in living transgenic AD mice at single vessel resolution. Results Our results showed significantly impaired CVR and reduced CBF in basal state in transgenic AD mice compared to non-transgenic littermates in an early stage of AD progression. Changes in total hemoglobin (Δ[HbT]) in response to vasoconstriction were significantly attenuated in AD mice, especially in arteries and tissue, and the recovery time of Δ[HbT] after vasoconstriction was shorter for AD than WT in all types of vessels and cortical tissue, thereby indicating hypoperfusion and reduced vascular flexibility. Additionally, our 3D OCT images revealed that CBF velocities in arteries were slower and that the microvascular network was severely disrupted in the brain of AD mice. Conclusions These results suggest significant vascular impairment in basal CBF and dynamic CVR in the neurovascular network in a rodent model of AD at an early stage of the disease. These cutting-edge in vivo optical imaging tools offer an innovative venue for detecting early neurovascular dysfunction in relation to AD pathology and pave the way for clinical translation of early diagnosis and elucidation of AD pathogenesis in the future.


Background
Alzheimer's disease (AD) results in gradual impairment of cognitive abilities and disruptive behavioral changes that contribute significantly to morbidity in aging populations.However, neuropathogenesis is not well understood and there are no effective treatments to prevent or treat AD.This knowledge gap reflects in part the heterogeneity of the disease and the diverse trajectories between patients including the severity of amyloid-β (Aβ) and cerebral amyloid angiopathy (CAA), as well as the difficulty in identifying the very early stage of AD.Many hypotheses have been postulated to explain the underlying cause of AD, the most popular one being "the Amyloid Cascade hypothesis," which posits that abnormal accumulation of Aβ in the brain leads to neuronal dysfunction and eventually neurodegeneration based on numerous histologic and genetic data [1,2].However, the lack of a strong correlation between the severity of cognitive impairment and the level of Aβ deposition reported by some studies [3], has led to the hypothesis that senile Aβ plaques are the consequence of neurodegeneration, rather than the cause.A more recent hypothesis based on preclinical and clinical studies, "the Vascular hypothesis", proposes that neurovascular dysfunction and vascular risk factors underlie AD [4][5][6][7].Indeed, there is increasing evidence of vascular defects in the brain of patients in the early stages of AD [8][9][10] including cerebral hypoperfusion [11], leakage in the blood-brain barrier [12], and cerebral amyloid angiopathy that eventually leads to infarcts, microbleeds, and cerebral hemorrhage [13].However, the majority of studies demonstrating structural alterations in the cerebrovasculature have used immunohistochemistry (IHC) in postmortem brain tissue from AD patients or transgenic AD animal models (summarized in Fisher et al. [14]) and could not capture dynamic changes nor three-dimensional (3D) spatial information of the vasculature.On the other hand, most in vivo studies assessing the effects of vascular risk factors such as chronic cerebral hypoperfusion and transient ischemia on brain changes in transgenic AD animals have only examined changes through inflammation and oxidative stress markers and behavioral tests (summarized in Scheffer et al. [7]), thereby providing limited insight into the actual vasculature changes in the brain of animal models of AD.
To emulate the heterogeneous pathological features that AD patients exhibit, several AD animal models have been developed with different AD transgenes inserted, such as amyloid-β-precursor protein (APP), presenilin (PSEN) 1, PSEN 2, apolipoprotein-E (APOE) 4, etc. leading to phenotypic variations [15,16].Recently, a human amyloid-β knock-in (hAb-KI) animal model was developed to express phenotypes that mimic the late-onset features in sporadic AD patients through a knock-in (KI) of a "humanized" Aβ sequence into the endogenous mouse APP protein [17].Within this AD animal model, the changes in mRNA/protein levels, mitochondrial dynamics, an age-dependent decline in cognitive abilities, as well as changes in brain volume, etc. have been documented [17,18] but vascular pathology has not been investigated yet.
Cerebrovascular reactivity (CVR) refers to the dynamic responsiveness of brain blood vessels to vasoconstrictive or vasodilatory stimuli, which can be used as a measure of the vascular physiological conditions within the brain [19].Technologically, CVR can be examined through dynamic changes in cerebral blood flow (CBF) and/or cerebral blood volume (CBV) changes, which have been known as hemodynamic changes in vasculature.Various means of stimuli such as increased arterial CO2 partial pressure (i.e., hypercapnia) and exogenous chemicals have been used to assess pathologies of vascularrelated diseases as well as AD [20][21][22][23].Several clinical studies demonstrated that the progression of cognitive deficits in AD patients is aligned with the reduction of CVR and vascular contractibility [24][25][26][27].
However, this association is not yet conclusive as some studies reported that global CVR after hypercapnia was retained in AD [28,29].Also, studies to trigger vasoconstriction in AD patients are limited due to the risk of hypotension-induced ischemia [23].Therefore, there is a need for further studies on CVR in AD patients and transgenic animals.
In clinical studies, mapping of CVR of the whole brain of AD patients is commonly achieved through functional magnetic resonance imaging (fMRI) for its advantages over positron emission tomography (PET) and single-photon emission computed tomography (SPECT) with non-radiation exposure and high repeatability [30].Although the blood-oxygen-level-dependent (BOLD) [31] signal of fMRI can retrieve CBF signals with a high signal-to-noise ratio (SNR), it yields an indirect, relative CBF measurement that is potentially affected by other physiological parameters such as cerebral blood volume (CBV) and metabolic parameters such as cerebral metabolic rate of oxygen [32].Also, it has been reported that BOLD signals are primarily from venous blood volume changes, not from arterial blood [33,34], thereby providing an incomplete analysis of CVR across the neurovascular network.Alternatively, arterial spin labeling (ASL) [35] can be performed simultaneously with BOLD with the advantage of its direct, quantitative measurement of CBF; however, its SNR is limited [36].While neural imaging tools such as PET, SPECT, and fMRI can provide insight into regional alterations of CVR in the whole brain of AD patients, their limited spatial resolution limits their ability to resolve individual cerebral vessels and the microvasculature, which are crucial to understanding cerebrovascular function associated with CVR abnormalities and AD progression [29].
To bridge this knowledge gap, we applied two cutting-edge in vivo imaging technologies, Dual-Wavelength Imaging (DWI) and Optical Coherent Tomography (OCT) to examine (i) CVR in response to a vasoconstrictive stimulus (i.e., acute cocaine challenge), (ii) cerebrovascular morphology including microvascular density, and (iii) basal CBF velocity (CBFv) of arteries, veins, and capillaries in the somatosensory cortex (SSC) of transgenic AD hA-beta-loxP-KI mice (hAb-KI) in comparison to non-transgenic littermates (WT).The research group that generated this AD mouse model reported that the hAb-KI mice exhibited an age-dependent increase in the level of insoluble Aβ, with the highest level at 18-22 months old, and displayed impaired performance in a cortical-dependent behavioral test from 14 months age [17].These phenotypic alterations suggest that the hAb-KI mice can serve as a model for characterizing late-onset AD, and thus the middle age, 7-10 months old, was chosen to compare their differences with the age-matched WT before the emergence of cognitive deficits and severe accumulation of Aβ.The use of DWI developed in our lab [37] enabled real-time monitoring of dynamic vascular and tissue hemodynamic changes in response to cocaine challenge (1mg/kg, i.v., a vasoconstrictive stimulant [38]) with its high spatiotemporal resolution and large field of view (FOV).Additionally, threedimensional (3D) mapping of the cerebrovasculature in the SSC was achieved at capillary resolution using our ultrahigh resolution OCT [39,40], which simultaneously recorded the morphology of neurovascular trees by high-resolution optical coherence angiography (µOCA) along with the quantitative CBFv of these vessels by high-resolution optical Doppler tomography (µODT).Given that our OCT system can detect both arterial and venous volume changes with quantitative, not relative, CBFv, it allowed us to overcome the limitation of conventional imaging tools previously used to examine rodent models of AD.
Using these optical imaging techniques, we tested the following hypotheses: (1) CVR in AD mice would be impaired compared to WT mice in vivo, including an attenuation in vascular contractibility especially in arteries and (2) Basal CBFv would be lower in AD mice compared to WT with a lower microvascular density in the cortex of AD than control mice.

B. The surgical procedure of cranial window implantation prior to in vivo imaging
All surgical apparatus and operation table were sterilized with 70% alcohol before surgery.The animal was anesthetized using a mixture of 1.5-2.5% isoflurane and oxygen, and its physiological changes including respiration rate and body temperature were monitored constantly during the surgical procedure to ensure proper anesthesia.The fur on the dorsal part of its head was shaved using a trimmer, and the animal was secured in a customized stereotaxic frame with a face mask for anesthesia.Lubricating eye drops were applied to protect its eyes, and 70% alcohol and povidone-iodine were applied topically to sterilize the head area.Its cortical skin was incised minimally along the sagittal suture and removed in a round shape to expose parietal bones and bregma.Then, the connective tissue on the skull was cleared by applying a hydrogen dioxide solution.The skull on the SSC was thinned in a rectangular shape (~2×2mm 2 ) using a dental drill, and saline was constantly applied during the surgery to remove skull fragments.Once the skull was removed, a drop of saline was applied and a sterilized glass coverslip window (100µm thick) was fixed on top of the exposed brain surface with glue and dental cement [41].

C. In vivo imaging of cerebrovascular reactivity via DWI and post-image processing
In vivo imaging of hemodynamic changes, i.e., CVR, in the brain of WT and AD mice was carried out using the DWI developed in our laboratory [37,39,42].The DWI consisted of a zoom fluorescence microscope (AZ100, Nikon) and a 2× Plan APO objective with sequential illumination of dual-wavelength LEDs (Spectra Light Engine, Lumencor) in a 10ms exposure for each light source was synchronized with an sCMOS camera (Zyla 5.5, Andor) to retrieve spectral images of the SSC at these two excitation wavelengths, λ1=568nm and λ2=630nm, simultaneously (Fig. 1b).Light exposure time and camera control were managed by a custom LabVIEW software with the high-speed digital triggers.
To compare the CVR of WT (n=8) and AD (n=12) mice in vivo, cocaine was selected as a pharmacological challenge for its known vasoconstricting effects [38,39].A catheter was placed into the tail vein and cocaine (1mg/kg) was acutely administered intravenously (i.v.) after 10 minutes of baseline imaging.After cocaine, the hemodynamic changes in the cortex were recorded continuously for 30 minutes for comparison.
Based on the molar extinction coefficient spectrum illustrated in Fig. 1b, the sequential image stacks collected at λ1=568nm (Fig. 1bi) are sensitive to show hemodynamic changes in both oxygenated hemoglobin (Δ[HbO2]) and deoxygenated hemoglobin (Δ[HbR]) in the cortex including in both veins and arteries, whereas the images obtained at λ2=630nm (Fig. 1bii) show a higher absorbance for [HbR] than that for [HbO2].Accordingly, arteries and veins could be distinguished for quantification by comparing the two sets of image stacks that were simultaneously recorded via DWI, as demonstrated as red and blue traces in Fig. 1bi and bii, respectively.
Quantification of the dynamics of Δ[HbO2] and Δ[HbR] from the spectral images was achieved using the following formula [43], which is based on the assumption that Δ[HbO2] and Δ[HbR] are the major sources for changes in light attenuation at the measured diffuse reflectance: , and   2  refer to the molar extinction coefficients of HbO2 and HbR at their respective wavelengths, λ1 and λ2.  1 and   2 indicate the measured diffuse reflectance matrices at time t, and   1 and   2 are pathlengths of the light propagation.From the sum of Δ[HbO2] and Δ [HbR] acquired at the two wavelengths, changes in the total hemoglobin Δ[HbT], i.e. total blood volume in a given area of the cortex, can be quantified as a function of time.Three regions of interest (ROIs) each for arteries, veins, and tissue were manually selected from the spectral images, and their Δ[HbT] were quantified accordingly.

D. In vivo imaging of basal cerebrovascular morphology and blood flow via OCT
A custom ultrahigh-resolution optical coherence tomography (µOCT) system developed in our laboratory [39,40] consists of two functional extensions that enable simultaneous imaging of cerebrovascular morphology by ultra-high resolution optical coherence angiography (µOCA; Fig. 1ci) and a 3D cerebral vascular network with cerebral blood flow velocity (CBFv) quantification in the cortical brain by ultra-high resolution optical Doppler tomography (µODT; Fig. 1cii).As illustrated in Fig. 1c, an ultra-broadband light source (λ=1310nm and λFWHM=220nm) illuminates a spectral-domain OCT engine with a 2×2 broadband fiber-optic Michelson interferometer, and it yields a relatively high axial resolution (~2.5µm) due to a short coherence length (  = 2(2) 2 /Δ  ).In the sample arm of the system, a microscopic objective (f16mm/NA0.25)collimates the light over the cranial window transversely by a servo mirror, and a high-speed line scan InGaAs camera (2048-pixels, 145k-lines/s; GL2048, Sensors Unlimited) collects the transverse scans of backscattered light from the cortical vessels and tissue, resulting in the acquisition of 2D/3D µOCT in a transverse resolution of ~3µm.The camera was set to operate at 6k A-lines/sec for 3D µODT and 20k A-lines/sec for 3D µOCA.
CBFv was quantified by applying a custom Doppler flow reconstruction algorithm on µODT images [39,40].This reconstruction is based on the phase subtraction method (PSM) and phase intensity method (PIM).Simply, a phase shift induced by the Doppler flow of red blood cells in the vessels was detected between two consecutive A-scans, and it yielded the flow velocity (v) of the red blood cells by the following equation [40,44] : , where λ0 refers to the central wavelength (0.00131mm), n is the refractive index (1.38),T is the time interval between two adjacent A-scans (0.000167sec), and ϴ is the angle between blood flow and light incidence (assumed ϴ=0).The DWI images acquired prior to OCT imaging were used to identify 1 st order arteries (the primary branch of arteries), and 2 nd order arteries, which are bifurcated from the 1 st order vessels in each animal.ROI selection was initially done based on the OCA images using ImageJ and then applied to the ODT images because vessels were not entirely visible in the ODT images if their flow was too low.Multiple ROIs were selected sequentially in a closely spaced arrangement along each artery to account for the variability of CBFv within each vessel and the measurements on the same artery were averaged.7-10 per order of arteries in n=6 per group were analyzed and presented.
The diameters of 1 st and 2 nd order arteries were quantified using a customized MATLAB program that implemented the Frangi-Hessian filter [45].An OCT image was first normalized and denoised of speckles by applying a median filter with a 3×1 window, and a "vesselness" score based on eigenvalues of the Hessian matrix was quantified at each pixel to generate a binarized segmentation [46].Here, large and small vessels were segmented separately via the threshold segmentation method to avoid artifacts that commonly occur when the same Frangi-Hessian scale factor is applied to an image with a highly variable vessel profile [47].The resulting vasculature masks of large and small vessels were combined, and values for diameter were assigned based on the Euclidean Distance Transformation algorithms [48].Similar to the CBFv quantification, the 1 st and 2 nd order arteries were identified using the DWI images, and multiple ROIs were selected continuously along each artery from the processed OCA images.The multiple measurements on the same artery were averaged and the total mean of 6-9 arteries per order of arteries in n=6 per group is presented.
To compare the microvascular density in WT (n=8) and AD (n=6) mice, our lab developed a custom MATLAB program that processed an OCA image into a density map on a 0.00-0.15scale.This program binarized and skeletonized the OCA image sequentially by employing the Frangi vesselness filter [49] and thinning-based skeletonization algorithm [50].The resulting skeleton map S was further modified through morphological dilation and erosion operations to refine breakages in the skeletonized vessels caused by the inherent noise of OCA imaging.Then, the vascular fill factor (FF) at each pixel (x, y) (the measure of vascular density), was quantified over the whole image using a sliding window technique to generate a density map as shown in the following [51]: Here, S(x, y, w) refers to the image patch of size w = 60 at a given pixel on the skeleton map S. As a final step, a Gaussian filter (σ=3) was applied to the generated density map to achieve numerical stability and remove boundary artifacts.Then, three ROIs per post-processed image were manually selected in an area without large-or medium-sized vessels (>ϕ = 50µm), and their maximum and mean density were quantified.The resulting densities from the three ROIs per animal were averaged and the mean values across each group (n=9 for WT, n=6 for AD) were presented.

E. Statistics
All data are presented as mean ± standard error of the mean (SEM).The differences between WT and AD mouse groups were analyzed using a two-tail Student's t-test on SigmaStat software (Systat Software Inc.).To assess whether there was an interaction between sex and AD, two-way analysis of

Female AD mice exhibited more severe impairment in vascular reactivity than males.
While AD exhibits high prevalence in females [52], reports on its sex-biased differences in pathologies and pathogenesis have been equivocal, largely because there has been an insufficient number of sex-specific analyses in preclinical studies and sex has not been sufficiently recognized as a critical factor that may influence response to treatments in clinical trials of AD (summarized in Guo et al. [53]).
In addition, the RTs of female AD (Fig. 3cii) were shorter in all types of vessels and tissue than those of female WT: 12.14±1.75minvs 19.46±1.77minfor veins (p=0.024),7.95±1.42minvs 17.10±1.77minfor arteries (p=0.003), and 7.46±1.47minvs 18.80±1.41minfor tissue (p<0.001),respectively for AD and WT.When males of WT and AD mice were compared, the RTs of cortical tissue areas of AD mice (7.47±1.23min)were shorter than those of WT mice (13.05±1.76min;p=0.027;Fig. 3ci) but not in veins (10.58±2.41min(AD) vs 13.33±1.58min(WT) on average; p=0.422) nor arteries (7.26±1.61(AD) vs 12.83±1.73min(WT); p=0.051).Though the female AD mice appeared more impaired than male AD as compared with their wildtype counterpart the interaction between sex and AD was only significant for arterial and tissue IRs, which might reflect the small sample sizes.

AD decreased cerebrovascular blood flow velocities with no vessel diameter differences in arteries
of AD with WT mice.
To examine whether the blood flow in cortical arteries of AD mice is changed in the basal state, we took images of 3D vasculature of WT and AD mice (n=6 per group; 6-9 arteries per order of arteries via µOCA; Fig. 4ai) simultaneously with µODT to obtain quantitative CBFv measures (as demonstrated in Fig. 4aii).We observed that CBFv was decreased, and some arteries seemed occluded in AD mice (e.g., dashed trace of vessel visualized by µOCA (AD-ai), but for which no flow was detected shown in (ADaii)) as compared to WT mice, despite having similar diameters, as illustrated in Fig. 4a.To quantitatively compare the arterial diameters of the two groups, the arteries identified using the DWI images were first categorized into 1 st order, primary branch, and 2 nd order, secondary branch bifurcated from the 1 st order for more comprehensive analysis.The vessel diameters (ϕ; µm) quantified from the µOCA images showed no significant differences (p = 0.868 and p = 0.085) between WT (79.28 ± 2.54 and 72.20 ± 2.94 µm) and AD mice (78.61 ± 3.09 and 63.77 ± 3.12 µm) for 1 st and 2 nd order arteries, respectively (Fig. 4b).
Interestedly, the 2 nd order arteries of AD mice were significantly narrower than those of 1 st order arteries (p = 0.004), whereas there was no statistically significant difference for WT mice (p = 0.095).
We examined whether AD affected vascular blood flow by quantifying CBFv in the arteries for which diameters were quantified.As summarized in Fig. 4c, AD mice showed significant attenuation in CBFv compared to WT mice for both 1 st and 2 nd orders, suggesting a reduced basal circulation possibly due to partial occlusion of the vessels and/or narrowed vessels in high-order arteries.The mean CBFv for AD mice were 0.624 ± 0.075 and 0.537 ± 0.043 mm/s, and for WT, they were 0.787 ± 0.014 and 0.746 ± 0.025 mm/s for 1 st and 2 nd order arteries, respectively.Overall, simultaneous imaging with µOCA and µODT revealed that CBFv of arteries in the cortex were reduced significantly in AD mice, despite having negligible differences in arterial diameters compared to WT mice.

Microvascular density was decreased in AD mice.
Neuronal activations are known to be coupled with increased CBF, especially in capillaries to meet the metabolic needs of cells, and thus impairment of capillary flow or density may lead to neuronal dysfunction [54,55].Hence, we assessed microvascular density in the cortex of WT (n=9, ROIs=3/animal) and AD mice (n=6, ROIs=3/animal) by post-processing the µOCA images into microvascular density maps (Fig. 5a).Maximum capillary densities in the selected areas without large-and medium-sized vessels (ϕ 50µm~) were significantly lower for AD (0.0839 ± 0.0049) than WT mice (0.1045 ± 0.0034, p=0.003) (Fig. 5b).Consistently, Figure 5c further demonstrates a decline in mean microvascular densities of AD (0.0600 ± 0.0054) compared to WT mice (0.0801 ± 0.0034, p=0.005), indicating a global disruption in the microvascular recruitment in the brain of AD compared to WT (Fig. 5c).

Discussion
In vivo brain imaging of transgenic AD mice via the DWI and OCT strongly implicated that distinctive abnormalities in the cerebral vasculature emerge at an early stage of AD.Specifically, our data provided the following insights to deepen our understanding of vascular pathology in AD: (i) CVR was impaired in response to the vasoconstrictive stimulus, i.e., cocaine, in cortical arteries and tissue; (ii) CVR defects appeared stronger in females; (iii) arterial CBFv was significantly decreased in the resting state despite similar vessel diameters; and (iv) microvascular network density in the cortex was reduced.Taken together, these in vivo results are aligned with "the Vascular Hypothesis" and uniquely show vascular defects in the brain of AD mice in the early stage of AD progression.
Cerebral hypoperfusion and reduced CVR to vasodilators such as increased arterial CO2 partial pressure have been commonly used to evaluate patients with AD using MRI and PET [20,21]; however, a close examination of individual vessels and their changes in blood volume in vivo has not been feasible due to the limited spatiotemporal resolution of these conventional imaging technologies.The DWI and OCT used in this study allowed us to overcome such shortcomings and identify morphological and functional alterations in the vasculature of a mouse AD model.Additionally, the effect of vasoconstrictors on the blood vessels of AD has been left largely underexplored.Given that the prevalence of stimulant drugs, which are vasoconstrictive [56,57], is increasing, our data may also provide insights into the vascular effect from stimulant drug misuse that could increase the risk of AD [58].
The impaired CVR and decreased CBFv at the resting state observed in AD mice may be attributed to dysfunction of the neurovascular unit (NVU) prior to neurodegeneration.The NVU consists of cerebrovascular cells, including endothelial cells, smooth muscle cells, and pericytes, and non-vascular cells, including neurons and glia, and its proper functioning is crucial for the functional integrity of the with the suggested timeline of the symptom manifestation in AD development: early symptoms of vascular impairment and delayed emergence of Aβ plaques in the later stage of AD progression.
Aβ generated as a consequence of dysregulated NVU can trigger a vicious cycle of vascular and neuronal dysfunction along with further Aβ deposits because Aβ exerts vasoconstrictive effects through endothelial damage and induction of oxidative stress [73][74][75].Aβ has been reported to selectively accumulate on the arterioles and cortical capillaries over venules [76][77][78], leading to the development of cerebral amyloid angiopathy (CAA) that can further weaken vasculature function [79]; this specificity on arteries and capillaries is consistent with the reduced CVR and CBFv especially in arteries and reduced microvascular density demonstrated in this study.Also, diameter constriction by Aβ has only been observed in capillaries, not arterioles or venules in transgenic AD mice in vivo via two-photon microscopy with intraluminal dyes [78].Similarly, this study showed that there were no vascular diameter changes in the arteries of AD compared to WT animals, despite having a significant decrease in CBF in vascular networks.However, the underlying molecular mechanisms of decreased CBFv without changes in vessel diameters specifically in arteries remain unknown.Regardless, the vascular impairment imaged through our DWI and OCT holds important evidence on how vascular defects emerge at the early stage of AD and may undergo progressive decline as the NVU is further impaired and Aβ accumulates.
Furthermore, we have demonstrated that female AD mice exhibited more detectable CVR impairment than males when compared with age-matched WT animals, and sex-biased effect on AD was specifically observed in IRs of artery and cortical tissue.Likewise, female-specific changes during the progression of AD have been observed clinically, and a greater vulnerability of the female brains to AD development has been suggested by several studies (summarized by Guo et al. [80]).Female C57BL/6 mouse brains were shown to exhibit age-related alterations of gene expression earlier than males, and consequent early onset of hypometabolism and susceptibility to AD were reported [81].Additionally, a single-nucleus transcriptomic analysis of AD patients revealed that a substantial number of female cells is represented in the cell subpopulations associated with AD, whereas male cells are mostly associated with non-pathological cell subpopulations.Further, gene activities of excitatory and inhibitory neurons appear to be downregulated in females [82], suggesting that transcriptional dysregulation in females may account for the sex-biased differences in CVR presented in this paper.
Most studies on changes in the vascular density of AD patients and transgenic AD mice demonstrated a decline in vascular density in alignment with disease progression (summarized in Fisher et al. [14]).The severe reduction in microvascular density in the cortex of AD mice observed in our study is consistent with these findings.However, OCT is unique in that it provides images of the vascular network in areas devoid of large-or medium-sized vessels (>ϕ = 50µm) within a large FOV.Also, prior methodology for quantifying vessel length density and vessel counts was largely limited to postmortem tissues and therefore highly dependent on staining protocols and confounding variables such as age imaging with fluorescein-dextran [83].The discrepancy from our study could be attributed to different post-processing plugins to quantify density and the animal models used.We used the APP KI model (hAbeta-loxP-KI), which was recently developed to avoid phenotypic artifacts associated with APP overexpression [84] commonly observed in conventional transgenic mouse models such as the APP/PS1 model used by Bennet et al.The imperative difference in phenotypes between the two mouse models is that the dominant form of Aβ in the APP/PS1 model is Aβ42 [85], while that in hAb-KI is Aβ40, which has a vasoconstrictive effect in cerebral arterioles in vitro whereas Aβ42 does not [86].Hence, this new AD mouse model with the KI approach may be more suitable for defining disease mechanisms concerning vascular defects [87].
One of the limitations of this study is that only one-time point (7-10 months old) was evaluated; therefore, longitudinal studies to examine the potential correlation between cognitive decline and CBF in AD mice and to pinpoint the age when the differences between AD and WT become distinctive will be helpful for understanding the progression of AD.Also in this study, we anesthetized mice with isoflurane to reduce motion artifacts, and future measurements in awake animals would reduce confounds from the vasodilatory effects of isoflurane [88].Nevertheless, the data presented in this study using cutting-edge imaging techniques provides relevant information on the involvement of vascular alterations in AD.

Conclusion
In summary, we show cerebrovascular dysfunction in early-stage AD mice using advanced optical imaging.The high spatial resolution allowed us to image the vascular network in a relatively large FOV and with high temporal resolution (at seconds).Our results add new evidence to "the Vascular Hypothesis" that cerebrovascular alterations observed at the early stage of AD progression may precede neurodegeneration in AD.However, further investigation of cellular/molecular mechanisms behind the alterations together with hemodynamic response and high-resolution angiography is needed to assess the biological causality between the health of the underlying NVU dysfunction observed in the AD mice.(b) Vessel diameter ϕ (µm, means ± SEM) of 1st and 2nd order arteries of WT and AD mice (n=6 per group; 6-9 arteries per order of arteries; multiple measurements on the same vessel were averaged).There were no statistically significant differences between the WT and AD mice for both 1st and 2nd arteries but the 2nd order arteries of AD were significantly narrower than those of 1st order (p=0.004).

List of
(c) Quantitative CBFv (mm/s, means ± SEM) of 1st and 2nd order arteries of WT and AD mice (n=6 per group; 7-10 arteries per order of arteries; multiple measurements on the same vessel were averaged).
Statistically significant CBFv decreases were observed in both 1st and 2nd order arteries in the cortex of AD animals (p=0.038 and p=0.002, respectively).

Results 1 .
variance (ANOVA) with Bonferroni post-hoc tests was performed with "sex" and "groups (WT versus AD)" as sources of variation.The following criteria for p values were used: *p<0.05,**p<0.01,and ***p<0.001.Impaired cerebrovascular reactivity (CVR) in response to cocaine in AD mice in vivo.

Figure 2
Figure 2 summarizes a comparative analysis of WT and AD mice in terms of CVR against the

Figure 2b summarizes the
Figure 2bsummarizes the mean time courses of quantitative Δ[HbT] (%) in veins (i), arteries (ii), variation from individual samples and possible tissue distortion in preparation.The feature of µOCA with label-free and high spatiotemporal resolution and automated segmentation technique presented in this study allowed us to detect the microvasculature in vivo and to quantitatively detect the decrease of microvascular density in the cortex of living hAb-KI mice in a large, 3D FOV without inherent limitations of contrast agents and fluorescent labeling.Previously, a study by Bennett et al. demonstrated that there was no distinctive change in blood vessel volume in the cortex of 15-month-old APP/PS1 mice when assessed with in vivo two-photon

Figure 2 .
Figure 2. Comparison of cerebrovascular reactivity in response to cocaine between WT and AD mice

Figure 4 .
Figure 4. Vessel diameters and cerebrovascular blood flow velocity (CBFv) in arteries of WT and

Figure 5 .
Figure 5. Evaluation of microvascular density between WT and AD mice.