Optical imaging in an Alzheimer’s mouse model reveals amyloid-β-dependent vascular impairment

Abstract. Alzheimer’s disease (AD) and cerebrovascular disease are often comorbid conditions, but the relationship between amyloid-β and in vivo vascular pathophysiology is poorly understood. We utilized a multimodal, multiscale optical imaging approach, including spatial frequency domain imaging, Doppler optical coherence tomography, and confocal microscopy, to quantify AD-dependent changes in a triple transgenic mouse model (3xTg-AD) and age-matched controls. From three months of age (naïve) to 20 months (severe AD), the brain tissue concentration of total and oxy-hemoglobin (Total Hb, ctO2Hb) decreased 50 and 70%, respectively, in 3xTg-AD mice. Compared to age-matched controls, significant differences in brain hemoglobin concentrations occurred as early as eight months (Total Hb: 126±5  μM versus 108±4  μM; ctO2Hb: 86±5  μM versus 70±3  μM; for control and AD, respectively). These changes were linked to a 29% vascular volume fraction decrease and 35% vessel density reduction in the 20-month-old 3xTg-AD versus age-matched controls. Vascular reduction coincided with increased brain concentration of amyloid-β protein, vascular endothelial growth factor (VEGF), and endothelial nitric oxide synthase (eNOS) at eight and 20 months compared to the three-month baseline. Our results suggest that amyloid-β blocks the normally reparative effects of upregulated VEGF and eNOS, and may accelerate in vivo vascular pathophysiology in AD.


Introduction
Cerebrovascular pathology is often a comorbidity of Alzheimer's disease (AD). 1 In patients, 60 to 90% of AD cases are associated with ischemic vascular disease and 90% of cases show amyloid-β deposition in the vessels. 2 Blood vessels become less pliable early in the process of AD, and this lack of vessel reactivity has been detected with transcranial Doppler, [3][4][5] fMRI, 6,7 PET, 8 and SPECT (Ref.9) techniques.The reduced vascular reactivity is also seen in mouse models genetically modified to overexpress amyloid-β pathology, 10 and these mice are rescued from pathology and memory impairment when treated with anti-hypertensive drugs 11 or Viagra (regulates nitric oxide pathway). 12A proposed hypothesis of AD pathogenesis is that reduced oxygen delivery to the brain leads to hypoxic stress in neurons, causing further amyloid-β production 13,14 and vessel damage. 1 However, it is unknown (1) when the brain becomes hypoxic, (2) how hypoxic the brain becomes, and (3) the role of vessel constriction versus vessel loss in AD-related hypoxia.
Diffuse optical imaging (DOI) 15 quantitatively measures the tissue concentration of oxy-and deoxy-hemoglobin, making it well-suited for studies of brain hemodynamics.We recently described intrinsic optical and hemodynamic contrast between 20-month-old triple-transgenic APP/tau/presenilin 1 (3xTg-AD) mice and age-matched controls in vivo using spatial frequency domain imaging (SFDI), a noncontact, camera-based DOI technique. 16SFDI employs a model of light transport to solve for tissue absorption (μ a ) and reduced scattering (μ 0 s ) coefficients on a pixel-by-pixel basis. 17Oxy-and deoxy-hemoglobin concentrations were derived from the μ a spectra (650 to 970 nm) using the Beer-Lambert law.Compared to age-matched controls, the 3xTg-AD mice had 16 and 21% lower total and oxy-hemoglobin concentrations.A dynamic hyperoxia inhalation challenge also revealed half the functional response found in controls.
In this study, we utilized SFDI to characterize how and when this hypoxic state arises in the 3xTg-AD mouse model compared to age-matched controls.Doppler optical coherence tomography (dOCT) was also used to detect cortical vascular volume in vivo in some 20-month-old animals.After imaging, the mice were sacrificed and vessel density was counted using confocal microscopy, and brain protein levels of amyloid-β, endothelial nitric oxide synthase (eNOS), neuronal nitric oxide synthase (nNOS), and vascular endothelial growth factor (VEGF) were measured.We assumed the toxic effects of amyloid-β on the brain vasculature are global, allowing comparison between the various sampling depths and volumes of the multiple imaging modalities used.This multimodal optical imaging approach allowed us to show that brain hypoxia occurs as early as eight months in the 3xTg-AD brain; vessel constriction and vessel loss both contribute to reduced blood perfusion, and the normal vasodilatory and angiogenic effects of eNOS and VEGF, respectively, may be blocked by amyloid-β.

Animal Model
All mice used were bred and obtained from the LaFerla laboratory at University of California, Irvine.3xTg-AD mice were studied at three months (n ¼ 9), eight months (n ¼ 6), and 20 months of age (n ¼ 5) to coincide with naïve, moderate, and advanced plaque and tangle pathology, respectively.Agematched C57 BL/6-129svj mice (three months: n ¼ 6; eight months: n ¼ 6; 20 months: n ¼ 9) were used to control for changes due to normal aging.All procedures were performed in accordance with the regulations of the Institutional Animal Care and Use Committee of the University of California, Irvine (protocol no.2010-2934).
Imaging windows were created under gas mask isoflurane anesthesia (2% maintenance in 21% oxygen balanced by nitrogen) by removing the skin from bregma anteriorly to lambda posteriorly and bilaterally to the temporalis muscle attachments.The mouse skull, 200-to 500-μm thick, was left intact to reduce trauma artifact from removing or thinning it.This approach has been used widely in intrinsic signal imaging of mouse brain. 18uring the initial surgery and for the imaging session, the mouse was kept at 37°C by a thermister-controlled heating pad (CWE Inc., Ardmore, Pennsylvania).The mouse's head was secured in a stereotactic frame (Stoelting Co, Wood Dale, Illinois) to prevent motion and a thin layer of mineral oil was applied onto the cranium to reduce skull drying. 19,20Following surgery, isoflurane was turned down to 1% maintenance.Excess gases were scavenged via a f/air adsorber unit (A.M.Bickford Inc., Wales Center, New York) and gas flow was maintained at 1 L∕ min.

Steady-State Measurement and Hypercapnia Challenge
An initial 17-wavelength (650 to 970 nm, every 20 nm) baseline measurement was acquired in triplicate, lasting for ∼2 min with the mouse breathing 21% O 2 (normoxia).Immediately following the 17-wavelength scan, in 20-month-old mice, baseline and dynamic measurements were recorded during an eight-min vasodilation challenge using only 670-and 850-nm light to achieve a time resolution of ∼5 s∕measurement while still containing hemoglobin chromophore information.Normoxia was used to measure the baseline optical properties again for three min followed by hypercapnia (5% CO 2 , 95% normoxia air) as an experimental perturbation for five min.

SFDI Instrument and Analysis
A schematic of the experimental arrangement is illustrated in Fig. 1.A complete description of SFDI instrumentation and data analysis has been previously presented in detail. 16,17,21riefly, broadband near-infrared light was projected onto the mouse cranium in sinusoidal patterns at two spatial frequencies (0 and 0.125 mm −1 ) and phase-shifted 120 deg apart (six projections total), and the remitted reflectance was captured at 17 wavelengths from 650 to 970 nm by a Nuance Multispectral Imaging System (CRI Inc., Woburn, Massachusetts) with a liquid-crystal tunable filter.The phase-shifted images (e.g., I 1 , I 2 , I 3 ) were demodulated [Eq.( 1)] and calibrated to a mouse-shaped silicone phantom of known optical properties to correct for any system response, such as light inhomogeneity or lens aberrations.A known Monte Carlo forward model 21 was then used to generate a look-up table to solve for μ a and μ 0 s pixelby-pixel.
Quantitative tissue concentration values of oxy-hemoglobin (ctO 2 Hb), deoxy-hemoglobin (ctHHb), total hemoglobin (THb ¼ ctO 2 Hb þ ctHHb), and tissue oxygen saturation (S t O 2 ¼ 100 Ã ctO 2 Hb∕Total Hb) were calculated from the absorption spectrum according to the Beer-Lambert law using a least squares linear fit.For further analysis, a region of interest (ROI) was selected between the suture junctions, bregma and lambda, and bilaterally to the temporalis muscle attachments (Fig. 1).The average of pixel intensities in the ROI for each mouse was calculated.The within-group standard deviation (i.e., 3xTg-AD versus control) was calculated using the average ROI pixel values for each mouse in the group.All averages, standard error bars, and p values shown were calculated from mean ROI and SD values between animals in each group.A two-tailed student's t-test analysis was used to determine significance between Alzheimer's and control mice.All image processing and analyses were done with MATLAB® software (Mathworks, Natick, Massachusetts).

Doppler Optical Coherence Tomography
After SFDI, some 20-month-old mice (three Ctrl versus three 3xTg-AD) were imaged with dOCT in vivo while breathing 21% O 2 .A fiber-based swept source OCT system with a central wavelength of 1310 nm, an A-line rate of 50,000 Hz, and sensitivity of 108 dB was used in the experiment.The system has a depth resolution of 9.3 μm in air and a lateral resolution of ∼9 μm.For in vivo mouse brain vasculature mapping, a three-dimensional (3-D) OCT data volume containing 4096 frames with 1024 A-lines per frame was obtained for each sample and the processed 3-D OCT volume covered an area of ∼2 mm × 2 mm × 3 mm.A recently developed intensitybased Doppler variance method was used to reconstruct the cortical vessel network. 22In each A-line, the OCT system provided 512 pixels over a depth of 4.3 mm in air (∼3 mm in tissue).However, because of scattering from the overlying skull and the shadowing artifact from the high density of blood vessels in mouse brain, dOCT analysis was limited to the superficial vessels of the brain.The en-face dOCT images obtained were analyzed to calculate vessel volume fraction.Vessel volume fraction was calculated from the most superficial (color-labeled white) layer (180 μm thickness).For each mouse, 30 en-face images imported into MATLAB® were intensity-thresholded using the MATLAB® function greythresh, which applies Otsu's method, and a 7 × 7 pixel median filter was applied.Vessel volume fraction was calculated as the number of pixels with vessels throughout the tissue volume divided by the total number of pixels in the tissue volume.Mean apparent hematocrit was calculated from Eq. ( 2).
where THb is the mean ROI total hemoglobin value determined from SFDI measurements and hemoglobin weight per 1 ml packed red blood cells is derived from young C57 BL/6J mice. 23or 3-, 8-, and 20-month-old mice, we saved the overlying skull and a single OCT B-scan frame was obtained for each sample ex vivo.Using ImageJ software, 24 the distance (pixel number) between the upper and lower skull boundary was found manually at three points and multiplied by the pixel resolution (6 μm∕px) to determine the average thickness of each mouse skull.

Histology
In 3-and 20-month animals, 150 μl of 10 mg∕ml Texas Red lysine-fixable dextran dye (Invitrogen, Grand Island, New York) was administered via cardiac injection prior to sacrificing the animals.The right hemispheres of the brains were preserved in 10% formalin and subsequently flash frozen and sliced in horizontal sections to correlate to the top-down imaging of SFDI.Citrate-treated 40-μm slices were incubated overnight at 4°C with eNOS primary antibodies (Abcam, Cambridge, Massachusetts) diluted 1:400 and/or amyloid-β 6e10 primary antibodies (Millipore, Bellerica, Massachusetts) diluted 1:500 and the appropriate secondary fluorescent antibodies.Cortical slices (40 μm) ∼300 μm down from the dorsal surface of the brain were selected and imaged with confocal microscopy for the vessel-labeling dextran and eNOS.Vessel density analysis was done on either the dextran (3-and 20-month animals) or eNOS images (8-month animals) with a semiautomated code written in MATLAB®, 25 with user-defined thresholds for intensity and object size of the vessels.Mean vessel density was calculated as the total length of the vessels in the image divided by the area of the image.Three to six animals were used in each group for statistical analyses.

Western Blots
The left hemisphere of each brain was flash frozen and soluble homogenates of the forebrain were run on a polyacrylamide gel and probed with antibodies for amyloid-β 6e10 (Signet, Dedham, Massachusetts), eNOS (Abcam), nNOS (Abcam), VEGF (Millipore, Temecula, California), and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (Santa Cruz Biotechnologies, Santa Cruz, California).Protein band intensities were analyzed with ImageJ software and normalized to the GAPDH band intensity.Relative comparisons were done between groups from a 26-well gel with four to five mice in each group.

Baseline SFDI Brain Measurements
Optical property map (μ a and μ 0 s ) images of the brain were determined at 21% O 2 from 650 to 970 nm at 17 wavelengths spaced every 20 nm.Absorption contrast from 730 to 970 nm decreased significantly (p < 0.05) from 3 to 20 months of age in both Alzheimer's and control mice [Fig.2(b)].Significant absorption differences were also found between 8-month-old 3xTg-AD and control mice from 790 to 930 nm and between 20-month-old cohorts at all 17 wavelengths measured from 650 to 970 nm.
We found that normal aging from 3 to 20 months of age caused a 26, 46, and 26% decrease in Total Hb, ctO 2 Hb, and tissue hemoglobin oxygen saturation (S t O 2 ), respectively.
Similarly, aging in the 3xTg-AD model caused a 50, 70, and 42% decrease in Total Hb, ctO 2 Hb, and S t O 2 , respectively.Total hemoglobin maps [Fig.2(a)] illustrate the progression of perfusion decrease in normal and 3xTg-AD aging, and ROI averages for total hemoglobin [Fig.2(c)] and ctO 2 Hb [Fig.2(d)] show that significant differences occur at 8 and 20 months of age.Table 1 has the calculated baseline ROI averages of hemoglobin concentrations at each age and condition group.
Reduced scattering values from 650 to 830 nm rose significantly (p < 0.05) in normal aging from 3 months to 20 months of age [Fig.3(a)].The same aging in 3xTg-AD mice caused significant scattering rises in all 17 wavelengths from 650 to 970 nm.Scattering was also significantly higher in the 20month 3xTg-AD mice compared to age-matched controls from 650 to 890 nm.Skull thicknesses as measured by OCT ranged from 175 to 474 μm and increased significantly from the three-and eight-month time points to the 20-month time point [Fig.3(b)].

In Vivo Doppler OCT
To better determine vascular tone in vivo, we looked at dOCT images of 20-month 3xTg-AD and control mice.As can be seen in Fig. 4, superficial cortical vessels (white) are more dilated in the controls and thinner in the 3xTg-AD mice.Analysis of the superficial vessel volumes reveal a 29% decrease (p ¼ 0.03) in vessel volume in 3xTg-AD mice compared to controls [Fig.4(b)].Using Eq. ( 2) and the dOCT and SFDI data, the calculated apparent hematocrit was 0.16 AE 0.02 for 20-month controls and 0.17 AE 0.03 for 20-month 3xTg-AD mice (p ¼ 0.8).Error reported is standard error.

Vessel Density Analysis
Immunohistochemistry revealed capillary staining in both control and 3xTg-AD mice at all three ages in cerebral blood vessels (Fig. 5).While eNOS capillary staining could also be seen in 3and 20-month mice, the signal-to-noise was much better with the injected dextran dye.Amyloid-β plaques could be seen in the cortex of 20-month-old 3xTg-AD mice.Images of the capillary staining were filtered into binary images [Fig.4(c)] and skeletonized.When vessel density was analyzed, no significant

Western Blots
We confirmed with protein analysis that amyloid-β is elevated in the 8-and 20-month 3xTg-AD mice [Fig.6(a)].Both eNOS and VEGF were upregulated significantly by eight months, and eNOS continued to be significantly raised at 20 months for both 3xTg-AD and age-matched controls [Fig.6(b)].Protein expression of nNOS remained constant at all ages for both 3xTg-AD and controls.

Hypercapnia Vasodilation Challenge
To visualize the brain reactivity to hypercapnia, continuous measurements of 670/850 nm μ a values were fit for concentrations of ctO 2 Hb and ctHHb and three-min baselines at normoxia and hypercapnia were averaged.In this vasodilation challenge, a rise in ctO 2 Hb, Total Hb, and S t O 2 along with a dip in ctHHb were expected and seen in the control mice [Fig.6(c)].Interestingly,

Discussion
This multimodal imaging study is the first to describe the amyloid-β-dependent time course and extent of hemodynamic and vascular impairment in AD compared to normal aging in a 3xTg-AD mouse model.SFDI results showed reduced brain perfusion and oxygenation in the 3xTg-AD mice starting at eight months.Both controls and 3xTg-AD mice showed significant drops in Total Hb and ctO 2 Hb at 20 months, but 3xTg-AD mice had 30% lower Total Hb and 8% lower tissue oxygen saturation than controls.This mirrors findings of decreased cerebral blood flow in normal human aging, 26,27 the severity of which increases in mild cognitive impairment and AD. 28xTg-AD mice exhibited a 29% decrease in vessel volume fraction compared to controls, despite upregulation of eNOS in both 3xTg-AD and control mice.Thus, eNOS inefficiency at producing the vasodilatory molecule NO in aged endothelial cells 29 may be further exacerbated by amyloid-β 30,31 in the 3xTg-AD mice.Similarly, amyloid-β binds VEGF and limits its efficacy in inducing angiogenesis, 32 showcased in the 3xTg-AD mice by the 35% decrease in vessel density compared to controls.These data suggest that both reductions in vascular tone and density contribute to the brain hypoxia seen in AD.
Hypercapnia challenges also revealed a dynamic contrast between the 3xTg-AD and control mice at 20 months of age.In controls, we saw an expected robust increase in ctO 2 Hb and a symmetric drop in ctHHb.However, in 3xTg-AD mice, the same challenge produced a significantly lower change for both ctO 2 Hb and ctHHb.This could reflect an oxygen-starved state in the 20-month 3xTg-AD mice.Interestingly, nNOS protein, which is largely responsible for vasodilation in response to hypercapnia, 33 did not change with age or between AD and normal mice.Hypercapnia is known to increase heart rate and ventilation in addition to vessel dilation. 34Thus, without control of the former two variables, interpretation of these results is limited.
Multimodality information allowed us to calculate the mean apparent hematocrit for the 20-month control and 3xTg-AD mice to be 0.16 and 0.17, respectively.Normal hematocrit for young C57 BL6 mice is 0.44 AE 0.004. 23While it is possible that the true hematocrit is as low as the calculated apparent hematocrit, the different tissue depths interrogated by SFDI (∼2 to 4 mm) and dOCT (∼200 μm) have different vessel volume fractions.For example, SFDI probes relatively deep cortical and subcortical structures.If the observed vessel fraction is less than that of the regions measured by dOCT, the apparent hematocrit would increase.Nevertheless, dOCT images suggest that SFDI-measured alterations in bulk tissue hemodynamic properties are due to underlying changes in microvascular structure rather than differences in erythrocyte volume fraction between AD and control mice.
Reduced scattering coefficients were significantly elevated in the 20-month 3xTg-AD mice compared to controls, and μ 0 s increased with age in both 3xTg-AD and control mice.Tissue scattering parameters are measured in SFDI (instead of assumed in continuous wave spectroscopy), yielding more accurate absorption values in addition to general structural contrast.For example, cell or organelle swelling from cerebral edema, 35 vasodilation from cortical activation, 36 ex vivo human AD brain tissue, 37 and cell death in vitro 38 and ex vivo 39 all alter light scattering.We found the measured skull thickness correlates to the increase in scattering with age.Better modeling of light through layered tissue and further study of a mouse model that mimics the cellular pathology seen in human AD 40 will need to be done to make the scattering signal an informative biomarker of AD.In summary, we have utilized complementary in vivo optical imaging methods, SFDI and dOCT, combined with conventional histological and protein assays, in a longitudinal mouse model study examining the appearance and progression of AD.Our results suggest that amyloid-β may contribute to ischemia/hypoxia in the brain of 3xTg-AD mice by reducing both vessel density and volume in vivo, despite upregulated VEGF and eNOS protein.This work links in vivo hemodynamic measurements to molecular biomarkers in a preclinical model and suggests potential approaches for utilizing clinical nearinfrared spectroscopy/diffuse optical imaging for characterizing AD vascular pathology and response to therapy in patients.

Fig. 2
Fig. 2 (a) ROI images of total hemoglobin concentration (Total Hb) for control (Ctrl) and AD (3xTg) mice determined from in vivo SFDI measurements.(b) Average ROI absorption spectra for 3, 8, and 20 monthold 3xTg-AD (red) and age-matched controls (black) show decreased absorption with age and disease.(c) Average ROI total hemoglobin concentration versus age shows significant decrease in 3xTg-AD mice by eight months of age.(d) Average ROI oxy-hemoglobin (ctO 2 Hb) concentration versus age shows significant decrease in 3xTg-AD mice by eight months of age.All error bars are standard error between animals in each group.

Fig. 3
Fig. 3 (a) Spectral dependence of reduced scattering parameter, shown for 3, 8, and 20 month-old 3xTg-AD (red) and age-matched controls (black).(b) Optical coherence tomography (OCT) analysis of skull thickness shows significant increase of skull thickness from 3 and 8 months of age to 20 months of age (p < 0.05).All error bars are standard error between animals in each group.

Fig. 4
Fig. 4 (a) Doppler OCT images of cortical vessels in 20-month control and 3xTg-AD mice.Superficial vessels (∼200 μm) are in white and deeper vessels are in pink.Scale bar is 500 μm.(b) Quantification of the superficial vessel (white) area in each group.(c) Computer-processed images of vessel density differences between 20-month control and 3xTg-AD mice derived from immunohistochemistry analysis.Scale bar is 100 μm.(d) Quantification of the cortical vessel density in each group.

Fig. 5
Fig.5Images of capillary staining at 3, 8, and 20 months in controls and 3xTg-AD mice.Texas red dextran perfused into the vessels was used to label the 3-and 20-month mice and endothelial nitricoxide synthase labeling was used to label the 8-month-old mice.Amyloid-β plaques (6e10 labeling) can be seen in the 20-month 3xTg-AD mice.

Fig. 6
Fig. 6 (a) Representative Western blot protein bands for each group of mice.(b) Western blot protein band intensities were quantified and normalized to GAPDH band intensities.(c) Magnitude change of oxy-, deoxy-, total hemoglobin, and percent oxygen saturation with an inhaled carbogen (95% O 2 þ5% CO 2 ) challenge.