Atherosclerosis is associated with a decrease in cerebral microvascular blood flow and tissue oxygenation

Chronic atherosclerosis may cause cerebral hypoperfusion and inadequate brain oxygenation, contributing to the progression of cognitive decline. In this study, we exploited two-photon phosphorescence lifetime microscopy to measure the absolute partial pressure of oxygen (PO2) in cortical tissue in both young and old LDLR-/-, hApoB100+/+ mice, spontaneously developing atherosclerosis with age. Capillary red-blood-cell (RBC) speed, flux, hematocrit and capillary diameter were also measured by two-photon imaging of FITC-labelled blood plasma. Our results show positive correlations between RBC speed, flux, diameter and capillary-adjacent tissue PO2. When compared to the young mice, we observed lower tissue PO2, lower RBC speed and flux, and smaller capillary diameter in the old atherosclerotic mice. The old mice also exhibited a higher spatial heterogeneity of tissue PO2, and RBC speed and flux, suggesting a less efficient oxygen extraction.


Introduction
Chronic atherosclerosis has a lifelong impact on brain health [1][2][3][4][5][6][7][8][9]. It can cause cerebral hypoperfusion and cerebrovascular disorders [10][11][12], which, in turn, impairs normal cognitive function [12][13][14][15]. Reduced supply of oxygen to brain tissue due to the atherosclerosis-related disruption in cerebral blood flow may induce neurodegeneration [16,17], leading to cognitive decline. The cerebral capillary network is the major site of vessel-tissue oxygen exchange [18]. The role of capillaries in oxygen delivery was emphasized in several modeling studies showing that brain tissue oxygenation benefits from not only the increase in absolute capillary blood flow but also from its homogenization [19][20][21]. Therefore, alterations of brain tissue oxygenation, and capillary structural and functional properties can indicate deviations from normal brain condition. PLOS  mice, but declined significantly starting from 6-month old, as compared to their age-matched controls [28]. For experiments, mice were first tracheotomised to facilitate animal respiration. Following tracheotomy, a femoral artery cannulation was performed to measure arterial blood gases.
Then, an open-skull craniotomy (3 mm in diameter) was done over the primary somatosensory area in the left hemisphere. After the craniotomy, a solution of PtP-C343 (0.6-1.2 μL at 150 μM) was microinjected into brain tissue at 200 μm depth below the brain surface. The cranial window was then filled with agarose diluted in artificial CSF, sealed using a 5-mm diameter microscope coverslip, and the animal was transferred to the two-photon microscope for imaging. Dextran-conjugated FITC (100-200 μL at 50 mg/mL, FD2000S Sigma) was then administered intravenously to visualize micro-vessels enabling localization of the capillary bed for PO 2 imaging and to measure capillary RBC flow properties. The animals were anesthetized by intraperitoneal administration of urethane (1-1.5 g/kg at 10% W/V) throughout all surgical and experimental procedures. Anesthesia was confirmed by the loss of withdrawal response to toe pinch, and could last for �4 hours, long enough for performing both surgery and imaging experiment. During all the surgical procedures and imaging, mice were spontaneously breathing an air/oxygen mixture (air mixed with 10-20% oxygen). The mice were euthanized after experiments using CO 2 inhalation. All the surgical and experimental procedures were performed in accordance with the Canadian Council on Animal Care recommendations, and were approved by the animal ethics committee of the Montreal Heart Institute research center (Protocol No: 2013-1619).
Prior to imaging, arterial blood gases were measured by a commercial blood gas analyzer to insure physiological levels. In the young group, arterial partial pressure of oxygen (P a O 2 ) was 102.3±1.4 mmHg; arterial partial pressure of carbon dioxide (P a CO 2 ) was 37.7±3.2 mmHg; and arterial oxygen saturation (SO 2 ) was 97.5±0.3%. In the old group, P a O 2 was 122.0±3.5 A representative CCD image capturing the mouse brain surface vasculature. Scale bar: 500 μm. c. Illustration of the 2D grid of locations (red dots) marked for PO 2 acquisition performed in the cortical tissue within the capillary bed. The red dashed square encloses a capillary, used as an example of blood flow and tissue PO 2 gradient acquisition in d. Scale bar: 50 μm. d. The red dots in the vascular image illustrate the measurement locations for tissue PO 2 gradient, perpendicular to the capillary axis. In the same capillary, perpendicular and longitudinal line-scans were performed, as illustrated by the yellow and red arrows, respectively. The arrows indicate the line-scan directions. e and f. Representative perpendicular (e) and longitudinal (f) line-scan images utilized to extract capillary RBC flow properties and diameter. In e, ø, d and t represent diameter, distance and time, respectively. In f, the RBC speed V was calculated by the ratio of the distance interval (Δd) to the time interval (Δt). mmHg; P a CO 2 was 40.2±3.0 mmHg; and SO 2 was 98.0±0.6%. Body temperature was maintained at *37˚C in all mice throughout all experiments. Data are expressed as mean±SEM.

Data acquisition
Tissue PO 2 imaging. A CCD image of the mouse brain surface vasculature was recorded to guide the selection of a region of interest (ROI) for subsequent two-photon imaging (Fig 1B). PO 2 measurements were collected typically at two depths, spanning from 150 μm to 250 μm below the brain surface. At each imaging depth, we first performed a raster-scan of FITC intensity, which revealed the location of the capillaries within a 400 × 400 μm 2 field of view (FOV). Two sets of tissue PO 2 measurements were collected. First, PO 2 measurements were performed in a 2D grid geometry (e.g., 10 × 10 measurement locations as denoted by red dots in Fig 1C) in the parenchymal tissue of the cortical capillary bed. Second, PO 2 was measured in tissue as a function of distance from the capillary wall in the direction perpendicular to the longitudinal axis of the capillary, with sub-micron steps, over a radial distance of 30 μm (Fig 1D).
At each measurement location, the phosphorescent nanoprobe (PtP-C343) in the focal volume was excited with a 25-μs-long laser excitation at 820 nm gated by the EOM, followed by a 275-μs-long collection of the emitted phosphorescence. Typically, at each location, 3,000 such 300-μs-long excitation/decay cycles were repeated to obtain an average phosphorescence decay with sufficient signal-to-noise ratio (SNR) for accurate lifetime calculation. Further technical details have been described in [23,29].
Imaging of capillary RBC flow. In the same capillaries for which PO 2 gradients were measured in the adjacent tissue, RBC speed, flux and hematocrit were measured by performing longitudinal and perpendicular line-scans (Fig 1E and 1F) [36,37] on the FITC-labelled capillaries. The line-scan sampling frequency was 800 Hz, fast enough to accurately measure RBC speed and flux in most cortical capillaries in mice without aliasing effect [38]. The linescans and tissue PO 2 gradients were acquired sequentially in time, capillary by capillary. The combined acquisitions for each capillary were completed within �10 seconds.

Data analysis
Calculation of PO 2 . The phosphorescence lifetime was calculated by fitting the average 275-μs phosphorescence decay to a single exponential decay function, using a non-linear least square minimization algorithm [22,23,29]. The lifetime was then converted to absolute PO 2 using a Stern-Volmer type calibration plot obtained in an independent oxygen titration experiment.
Capillary PO2 gradient and wall PO 2 . To observe the capillary PO 2 gradient, the PO 2 values were binned with 5-μm bins (Fig 2A). Capillary wall PO 2 was calculated based on tissue PO 2 measured radially in respect to the capillary axis, by averaging the PO 2 values within the first 10 μm from the capillary wall.
Calculation of capillary RBC flow properties and diameter. RBC flux was calculated by counting the number of the dark shadows in the perpendicular (Fig 1E) or longitudinal ( Fig  1F) line-scan images during the acquisition time (t). The angle of the dark streaks in the longitudinal line-scan image (Fig 1F) was estimated to calculate the RBC speed as Δd/Δt, where Δd and Δt are the distance and time intervals, respectively. RBC hematocrit was computed as the ratio (in %) of the number of pixels in the dark shadows to the total number of pixels in the longitudinal line-scan image. Capillary diameter was calculated as the full width at half maximum, by fitting the intensity profile along the distance axes (d) of the perpendicular line-scan image with a Gaussian model (Fig 1E). More details about the line-scan acquisition parameters can be found in [29].
Calculation of RBC flow and tissue PO 2 heterogeneity. For each of the measured parameters, its heterogeneity was quantified as standard deviation (STD) and coefficient of variation (CV) of its data population. Here, CV was calculated as the ratio of STD to mean [39].

Cortical capillary RBC flow and diameter positively correlate with oxygenation in adjacent tissue
We employed a home-built two-photon laser-scanning microscope (Fig 1A) to measure the resting PO 2 in the parenchymal tissue of the somatosensory cortex of urethane-anesthetized young and old atherosclerotic mice. PO 2 imaging with a 2D grid acquisition geometry ( Fig  1C) was performed within a 400 × 400 μm 2 FOV, typically at two depths per mouse from 150 μm to 250 μm below the brain surface. At each imaging depth, PO 2 gradients (Fig 1D) were also acquired in tissue adjacent to capillaries to estimate the capillary wall PO 2 . In the same capillaries, whose wall PO 2 's were estimated, RBC flux, speed, hematocrit and capillary diameter were measured with the perpendicular and longitudinal line-scan techniques (Fig 1E  and 1F) [29,37]. See the METHODS section for details about the experiments, data acquisition and analysis. We first investigated the PO 2 changes as a function of radial distance from the capillary wall, perpendicular to the capillary axis. The tissue PO 2 gradients were grouped into three groups having an equal number of capillaries, based on the wall PO 2 values of the associated capillaries: high (>46 mmHg), median (35-45 mmHg) and low (<35 mmHg). The average PO 2 gradient in the high wall PO 2 group exhibited greatest PO 2 decrease with distance from the capillary wall (ΔPO 2 = 2.2 mmHg between PO 2 at 0-5 μm and 15-20 μm from the wall), while such decreasing trend in the median and low wall PO 2 groups was much smaller (~1 mmHg; Fig 1A).
We next investigated the relations between capillary RBC flow properties, diameter and wall PO 2 . We observed a positive correlation between RBC speed and wall PO 2 up to speed values of 1.6-2.0 mm/s. The capillary wall PO 2 reached a plateau at higher speeds ( Fig 2B). A similar trend was observed between RBC flux and wall PO 2 (Fig 2C), but no clear correlation between RBC hematocrit and wall PO 2 was observed (Fig 2D). Capillary diameter exhibited a strong positive correlation with wall PO 2 (Fig 2E). Linear regressions were carried out on the data to quantify the correlations. The slopes of the correlations between RBC speed, flux, diameter, and wall PO 2 (corresponding to the panels b, c and e) were calculated to be 8.6 mmHg�s/ mm, 0.2 mmHg�s/RBC and 4.7 mmHg/μm, respectively; and the associated coefficients of determination (R 2 ) were 0.08, 0.02 and 0.11, respectively. Furthermore, RBC flux exhibited strong positive correlation with speed ( Fig 3A). A similar trend was observed in RBC flux vs. hematocrit (Fig 3B). RBC hematocrit was found weakly negatively correlated with speed ( Fig 3C). In addition, both RBC flux and speed increased with capillary diameter (Fig 3D and 3E). No clear trend was observed between RBC hematocrit and capillary diameter (Fig 3F). Linear regressions were carried out on the data populations to quantify the correlations. The slopes of the correlations corresponding to panels a-e are 0.03 mm/RBC, 20.06 RBC/s, -0.21 mm/s, 0.09 mm/s/μm and 0.64 RBC/s/μm; and the R 2 values are 0.26, 0.08, 0.01, 0.04 and 0.01, respectively.

Cortical tissue PO 2 and capillary RBC flow were lower and more heterogeneous in the older atherosclerotic mice
The absolute values, STD and CV of tissue PO 2 , capillary RBC speed, flux, hematocrit and diameter were compared between the young and old atherosclerotic mice. PO 2 gradients adjacent to capillaries and tissue PO 2 measurements were also collected using a 2D grid acquisition geometry in the cortical parenchymal tissue adjacent to the capillary bed. The histograms of tissue PO 2 (combining the PO 2 gradient and 2D-grid PO 2 measurements) acquired in the young and old atherosclerotic mice are presented in S1 Fig. Absolute tissue PO 2 in the old atherosclerotic mice (39.4±7.0 mmHg) was lower (P = 0.22) than in the young atherosclerotic mice (49.9±4.0 mmHg; Fig 4A). PO 2 STD and CV in the old atherosclerotic mice were noticeably higher than in the young atherosclerotic mice (Fig 4B and 4C). Similar changes from old to young were observed in the capillary RBC flow properties. Specifically, absolute RBC speed in the old atherosclerotic mice (0.66±0.08 mm/s) was significantly lower than in the young atherosclerotic mice (0.99±0.10 mm/s; Fig 4D). RBC flux in the old atherosclerotic mice (35.1±1.3 RBC/s) was lower than in the young atherosclerotic mice (39.4±2.1 RBC/s), but did not reach statistical significance (Fig 4G). Both RBC Flux STD and CV were significantly higher in the old atherosclerotic mice than in the young atherosclerotic mice (Fig 4H and 4I). Capillary diameters (Fig 4J) were not significantly different between the two groups despite exhibiting a lower trend in the older mice (Y: 5.3±0.2 μm; O: 4.8±0.2 μm, P = 0.12). No obvious difference was found in RBC hematocrit, and its STD and CV (S2 Fig).

Discussion
It has been reported that chronic atherosclerosis could cause cerebral hypoperfusion and cerebrovascular disorders [10][11][12], leading to cognitive impairment [12][13][14][15]. In this study we measured PO 2 in the cortical tissue of a mouse model of atherosclerosis. Mice lacking the LDL receptor develop atherosclerosis spontaneously, after the age of 3 months, without the need of high cholesterol diet [25,26]. The PO 2 measurements were performed both densely in cortical tissue adjacent to capillaries and over wider areas containing the capillary bed. Capillary RBC speed, flux, hematocrit and diameter were also measured by two-photon imaging of FITClabelled blood plasma. We analyzed the pairwise relations between tissue PO 2 , and capillary RBC flow properties and diameter. We also performed group comparisons to explore the impact of chronic atherosclerosis on brain oxygenation and capillary RBC flow.
We first characterized the tissue PO 2 gradient driven by the oxygen diffusion from capillaries (Fig 2A). Tissue PO 2 gradients around capillaries have been theoretically predicted [40], but not experimental confirmed. We found that around capillaries associated with high wall PO 2 , the gradient was higher when compared to those associated with medium and low wall PO 2 . There is possibility that the capillaries with high wall PO 2 were topologically and/or physically closer to pre-capillary arterioles in the network, and had higher oxygenation than the capillaries having medium and low wall PO 2 , thus contributing more to brain tissue oxygenation [41]. Next, we focused on the relations between capillary wall PO 2 and RBC flow and diameter. It has been reported that tissue PO 2 in proximity to the capillary wall could represent, to some extent, the intra-capillary PO 2 [42]. Our results in Fig 2 show that higher capillary RBC flux results in a greater oxygenation in the adjacent tissue. This result could be inferred from a previous study, in which a positive relation between capillary RBC flux and intra-capillary PO 2 was reported [43]. Furthermore, we analyzed the relations between capillary RBC flow parameters and diameter (Fig 3). Our results showed that both RBC speed and hematocrit were positively related with flux, but hematocrit was slightly negatively related with RBC speed, in agreement with the previous studies [37,43]. In addition, RBC speed and flux increased with capillary diameter, but hematocrit had no clear relation to diameter. Similar results have been reported in previous studies in anesthetized rats [37,44] and mice [36,45].
Absolute tissue PO 2 measured in the capillary bed tended to be lower in the old atherosclerotic mice (Fig 3), although a significant difference was not reached. However, reduced capillary RBC speed, flux and diameter were observed in the old atherosclerotic mice, which differs from previous studies reporting increased RBC speed and flux, and larger capillary diameter in aged mice [29] and rats [44]. Atherosclerosis is associated with vascular endothelial damage [46,47] and cerebral hypoperfusion [11,12] which could decrease capillary diameter thereby limiting RBC flow.
The reduced capillary RBC flow and diameter in the old atherosclerotic mice was accompanied by an increase in the spatial heterogeneity of the same variables (quantified by STD and CV, Fig 4), implying a compromised oxygen delivery to tissue as predicted by the biophysical models [19,20]. Capillary flow pattern disturbances are not unique to this mouse model of atherosclerosis, but were also found to be involved in the progression of neurodegeneration [21]. It was reported that capillary flow heterogeneity could play a crucial role in focal ischemic stroke, as it could induce local hypoxia and neuronal death, without a detectable decrease in the absolute global cerebral blood flow [48].
The origin of capillary RBC flow heterogeneity remains difficult to identify. We have found smaller average capillary diameter but larger diameter heterogeneity in the old atherosclerotic mice. As capillary RBC flow was reported to be correlated positively with diameter and intracapillary PO 2 [43,44], our findings about diameters are in good agreement with the findings about capillary RBC flow and tissue PO 2 (Figs 2 and 3). Therefore, these results imply that a disturbed flow pattern might essentially originate from microvascular structural changes [49].
Limitations exist in this study. Although the animal physiology was carefully controlled in experiments, anesthesia inevitably introduce hemodynamic changes relative to the awake state [50], thus adding variability to the measurements. Furthermore, we cannot exclude the possibility of an anesthetic effect with age which remains a limitation of this study. Nonetheless, in our previous studies in either anesthetized [44] or awake wild-type animals [29], comparisons of the absolute capillary blood flow and flow heterogeneity between the young and old mice followed similar trends, implying that anesthesia may not differentially affect the brain physiology among different age groups. Measurements in age-matched wild-type mice were not available in this study. In the previous work with the same atherosclerotic model, difference in cerebral blood flow between the 3-month and 6-12 month mice was found significantly different [28], but the aging effect on cerebral blood flow, oxygenation and capillary blood flow properties in the healthy wild-type mice occurred much later (e.g., 2-year old) [29].
To conclude, our study for the first time demonstrated that tissue PO 2 was positively correlated with capillary RBC flow and capillary diameter. In addition, we found lower tissue PO 2 , lower capillary RBC flow and smaller capillary diameter in the old atherosclerotic mice when compared to younger mice. We also found more spatially heterogeneous distributions of tissue PO 2 , RBC flow properties and capillary diameter in the old atherosclerotic mice when compared to the younger mice. Overall, our results confirm a detrimental impact of atherosclerosis on brain oxygenation and microvascular function.