Near-lifespan longitudinal tracking of brain microvascular morphology, topology, and flow in male mice

Walek, Konrad W.; Stefan, Sabina; Lee, Jang-Hoon; Puttigampala, Pooja; Kim, Anna H.; Park, Seong Wook; Marchand, Paul J.; Lesage, Frederic; Liu, Tao; Huang, Yu-Wen Alvin; Boas, David A.; Moore, Christopher; Lee, Jonghwan

doi:10.1038/s41467-023-38609-z

Near-lifespan longitudinal tracking of brain microvascular morphology, topology, and flow in male mice

Article
Open access
Published: 24 May 2023

Volume 14, article number 2982, (2023)
Cite this article

Download PDF

You have full access to this open access article

From

View current issue

Near-lifespan longitudinal tracking of brain microvascular morphology, topology, and flow in male mice

Download PDF

4452 Accesses
4 Citations
143 Altmetric
19 Mentions
Explore all metrics

Abstract

In age-related neurodegenerative diseases, pathology often develops slowly across the lifespan. As one example, in diseases such as Alzheimer’s, vascular decline is believed to onset decades ahead of symptomology. However, challenges inherent in current microscopic methods make longitudinal tracking of such vascular decline difficult. Here, we describe a suite of methods for measuring brain vascular dynamics and anatomy in mice for over seven months in the same field of view. This approach is enabled by advances in optical coherence tomography (OCT) and image processing algorithms including deep learning. These integrated methods enabled us to simultaneously monitor distinct vascular properties spanning morphology, topology, and function of the microvasculature across all scales: large pial vessels, penetrating cortical vessels, and capillaries. We have demonstrated this technical capability in wild-type and 3xTg male mice. The capability will allow comprehensive and longitudinal study of a broad range of progressive vascular diseases, and normal aging, in key model systems.

Quantitative Assessment of Age-Associated Alterations in Brain Vasculature in Wild-Type Mice and in Bigenic Mice that Model Alzheimer’s Disease

Article 11 July 2019

Intensity distribution segmentation in ultrafast Doppler combined with scanning laser confocal microscopy for assessing vascular changes associated with ageing in murine hippocampi

Article Open access 26 April 2022

Aging-related cerebral microvascular changes visualized using ultrasound localization microscopy in the living mouse

Article Open access 12 January 2022

Introduction

There is an increasing need for methods to longitudinally track microvascular alterations in the aging brain. As one example, in Alzheimer’s disease, vascular decline is thought to emerge up to decades before the onset of cognitive symptoms and is widely viewed as one potential contributor to symptomology¹. Similarly, changes in vascular structure and dynamics are a signature of several other progressive neurodegenerative diseases, including Parkinson disease², Huntington disease³, and multiple sclerosis⁴, suggesting a direct role in disease progression. Deriving effective early treatments for these conditions, and better understanding their etiology, likely requires a better understanding of these evolving vascular changes.

However, studying the role of vascular factors in age-related neurodegenerative diseases has been challenging because of difficulties in longitudinally tracking their development. To date, multi-endpoint terminal approaches have been used for determining such vascular alterations at different ages^5,6,7. While effective in some respects, these approaches are analytically less efficient, vulnerable to subject-specific random effects, and require an increasing number of animals when involving more measurement time points (Supplementary Fig. 1). Fluorescence multi-photon microscopy has revolutionized the in vivo visualization of neocortical vascular structure and function (flow)⁸, but has not met the need for long term, longitudinally repeated imaging. Bleaching and phototoxicity^9,10 and the instability of genetically-encoded indicators limit the use of fluorescent methods for stable repeated measurements over several months^11,12,13,14.

Optical coherence tomography (OCT) is a label-free technology that provides several unique imaging capabilities without the aid of fluorescence^{15,16,Full size image}

Every imaging session included acquisition of Doppler OCT images as well. We used these images to track changes in the diameter and blood flow of individual penetrating vessels (Fig. 2). The diameter of each vessel was measured by fitting its en-face cross-section to a 2D Gaussian function, while the blood flow was measured by the area-integral method¹⁹ after noise reduction²¹. This enabled us to longitudinally track the same set of 151 penetrating vessels of 13 mice for 7 months (gray lines in Fig. 2c, f).

**Fig. 2: Longitudinal monitoring of blood flow and diameter in penetrating arterioles and venules.**

Arteriolar diameter showed sigmoidal decreases in both AD and WT but faster in AD, with the rates of change defined from the sigmoid fits (Methods) being −13% per month in AD (95% CI, −18 to −9) and −7% per month in WT (95% CI, −11 to −3; p = 0.008; Fig. 2c). In turn, their fractional changes became significantly different between AD and WT at 18 weeks of age (WOA) (p < 0.05, Fig. 2d), which is termed as the age of significance (AOS) hereafter. Second, the venular diameter decreased in AD with the rate of change of −13% per month (95% CI, −18 to −8), significantly different from WT (1.3% per month; 95% CI, 0.9–1.7; p < 0.001), leading to the AOS of 12 WOA (Fig. 2e). Third, the arteriolar flow decreased in both AD and WT but faster in AD (−13% versus −10% per month, p < 0.001; AOS, 21 WOA; Fig. 2g). Fourth, the venular flow decreased in AD only (−16% per month; 95% CI, −25 to −8; AOS, 18 WOA; Fig. 2f, h). Finally, the penetrating vessel density in vessel number per unit area decreased in AD (−6% per month; 95% CI, −11 to −1), but the slopes were not significantly different between AD and WT (p = 0.11). It is interesting to see the AOS be earlier for the structural degenerations than their flow counterparts (see Discussion for detailed interpretation).

Structural and functional properties of capillary vessel networks

To longitudinally track alterations in smaller vessels like capillaries, every imaging session also included acquisition of another OCT angiography dataset with a higher spatial resolution (3 µm) and an optical focus on the cortical capillary bed (“Methods”). To extract as much information as possible from these microangiography images, we first converted grayscale microangiograms into a graph of the capillary vessel network, using our deep learning-based toolbox for enhancement, segmentation and vectorization of vessels²² (Fig. 3a). From these network graphs and the original grayscale images, our methods enabled measurements of numerous angio-architectural properties, including morphological properties like the length, diameter, and tortuosity of each vessel; and topological properties such as the branching order, betweenness, closeness, and shortest cycle of each vessel. Betweenness, closeness, and the shortest cycle are used in graph theory for elucidating how nodes are connected within a network (see Supplementary Fig. 4 for conceptual illustration and their implication when applied to a vascular network). While these properties were measured for each capillary segment (108,964 segments in total across 13 animals and seven ages of measurement), we calculated the mean and heterogeneity of every property per animal per age and then tracked them as a function of age, where the heterogeneity was quantified by the coefficient of variation (COV; the standard deviation divided by the mean). Other related measures include capillary number density, capillary length density, and fractal dimension (see Table 1 for a full list).

**Fig. 3: Longitudinal monitoring of microvascular morphological and topological properties.**

Table 1 Summary of statistical results of 25 vascular properties

Full size table

Prior to analyzing how the properties vary with age, we compared our measures to those reported in the literature when available. The distributions of capillary lengths, diameters, branching orders, and tortuosity obtained from our WT group (Fig. 3b) looked similar in shape and mean value to those previously reported from WT mice^23,24,25,26. We also observed agreement with the literature in the capillary length density and the shortest cycle (Supplementary Tables 1 and 2).

LME analysis revealed several differences emerging between AD and WT. Mean capillary length steadily decreased in AD mice only (−1.5% per month; 95% CI, −2.6 to −0.3; AOS, 25 WOA; Fig. 3c, d). Mean capillary tortuosity also significantly increased in AD only (2.2% per month; 95% CI, 1.3–3.1), but its fractional changes did not become significantly different between AD and WT until the end of measurement (35 WOA). Regarding network topology, the shortest cycle increased in AD only (1.9% per month; 95% CI, 0.3–3.4; AOS, 20 WOA; Fig. 3e). The betweenness also increased in AD only (3.5% per month; 95% CI, 1.3–5.8; AOS, 21 WOA; Fig. 3g). In contrast, the closeness did not change significantly with aging in both AD and WT (Supplementary Fig. 3), but heterogeneity (COV) decreased in AD (−3.0% per month; 95% CI, −7.9 to −2.0) whereas it increased in WT (3.6% per month; 95% CI, 0.9–6.4), with an AOS of 18 WOA (Fig. 3i).

To assess the pattern of blood flow through a large microvascular network involving hundreds to thousands capillary vessels, we previously presented a method for high-throughput measurements of red blood cell (RBC) flux based on OCT detection of RBCs passing through a voxel²⁰. However, this method tends to underestimate high flux values²⁷ and requires user-selected thresholds. To overcome these weaknesses and improve accuracy, we developed a deep learning-based method by using two sets of RBC-passage data that were simultaneously obtained with OCT and two-photon microscopy (“Methods”). This new method significantly outperformed the previous method (Fig. 4a, b, Supplementary Text 2).

**Fig. 4: Longitudinal tracking of capillary network RBC flux.**

To examine long-term changes in capillary network blood flow, every imaging session included OCT acquisition of RBC-passage data, and the deep learning method was used to produce a 3D map of capillary RBC flux values (Fig. 4c). As a result, WT mice exhibited a gradual increase with aging (2.1% per month; 95% CI, 0.9–3.2; Fig. 4e). This increasing trend agrees with previous findings from WT rats²⁸ although the previous study measured the flux values only at two ages in a non-longitudinal manner. Interestingly, AD mice also showed an increase in flux with aging but at a significantly higher rate (6.0% per month; 95% CI, 3.0–9.0; p < 0.001 against WT), making the AOS to be 22 WOA (Fig. 4f). The capillary flux heterogeneity (COV) decreased in both AD and WT but faster in AD (−6.7% versus −3.0% per month; p < 0.001; AOS, 23 WOA; Fig. 4g).

Temporal relationships between microvascular degenerations and cognitive impairment

To simultaneously obtain the time course of cognitive impairment, the animals also underwent novel object location (NOL) tests every month (Methods; the NOL test assesses spatial cognition and memory²⁹). While WT mice showed no significant changes in the discrimination index of NOL, AD mice showed cognitive decline with the discrimination index becoming significantly lower than WT at 27 WOA (Fig. 5a, b), consistent with previous findings from the identical 3xTg AD model³⁰.

**Fig. 5: Cognitive testing result, and integrative analysis of the measured vascular properties.**

As listed in Table 1, 10 out of 25 vascular properties showed significant differences in fractional changes between AD and WT during the ages of measurement (11–35 WOA). To illustrate the temporal relationships of these vascular degenerations to the cognitive impairment, we plotted a chronological graph as shown in Fig. 5c. This graph revealed three interesting relationships. First, all the observed vascular degenerations preceded the cognitive decline, up to 15 WOA early (about a seventh of the typical lifespan). Second, degenerations in arterioles and venules appeared earlier than those in capillary vessels. Third, structural changes generally preceded corresponding flow changes (arteriolar/venular diameter versus flow, and capillary betweenness/shortest cycle versus RBC flux; see Discussion for interpretation).

To reveal the correlation between the simultaneously observed vascular alterations and cognitive impairment, we calculated the correlation coefficient with age lags of 0, 4, 8 and 12 weeks in the AD group (Methods). Our analysis found that 19 vascular properties were significantly correlated with the NOL discrimination index (Fig. 5d). Some correlations were expected, such as the positive correlations between the decrease in NOL test score and the decreases in arteriolar and venular diameter and flow. Other correlations were unexpected; for example, the betweenness was strongly negatively correlated with the NOL score. This betweenness result was interesting when considering that it exhibited the youngest AOS among capillary vessel properties (Fig. 5c) and formed the greatest number of inter-correlations with other vascular alterations (Supplementary Fig. 8, see “Discussion” for interpretation).

Discussion

We have demonstrated that the integrated set of label-free in vivo imaging and image-processing methods produces an unprecedented set of data regarding age-related alterations in the morphology, topology, and function of the cortical vasculature across all scales. Such information-rich, multi-feature, longitudinal datasets allow to simultaneously investigate multiple aspects of slowly develo** cerebral microvascular degeneration while affording high statistical power (Supplementary Fig. 1). Among other results presented here, the chronological and correlation graphs between cognitive impairment and vascular alterations (Fig. 5c, d) are a representative example of what type of information can be obtained by the presented methods. Such information may provide insight into how cognitive impairment and different types of vascular alterations develop along with or independently of each other. Moreover, the presented methods align with the “3R’s” principles of animal research, particularly reduction. For example, while a vascular corrosion casting study demonstrated age-related changes in capillary morphology, it required more animals due to the terminal nature of the method⁵. Similarly, another casting study identified that changes in capillary morphology preceded cognitive decline, but it was limited to investigating only one age point out of a similar number of animals³¹. In contrast, the in vivo nature of the OCT method used in this study provides a clear advantage in reducing the number of animals required, while simultaneously providing detailed information about the dynamics of cerebrovascular changes.

In the presented demonstration, we stopped the longitudinal imaging when the AD mice exhibited obvious differences from WT in the cognitive function test for two consecutive sessions (i.e., for eight weeks). The tracking period was approximately a third of the typical mouse lifespan: The potential tracking period could be even longer as the image quality did not significantly degrade over time in many animals (Fig. 1b for example). One limitation in the presented demonstration is that we conducted all imaging experiments under anesthesia, although anesthesia unlikely affects our major outcome, i.e., difference in the rate of change with age (RCA) between AD and WT (see Supplementary Text 9 for related data and discussion). The presented methods are readily applicable to awake imaging as being shown in our follow-up studies. In the context of longitudinal assessments of awake animals, our methods can offer detailed, microvasculature-level assessments under a head-fixed condition. In comparison, diffuse correlation spectroscopy has shown the ability to provide assessments under a freely behaving condition but at a less detailed, macroscopic level³². Combining these two methods could yield a more comprehensive understanding of the changes and mechanisms underlying cerebrovascular dysregulation in Alzheimer’s disease and other progressive vascular diseases. Supplementary Text 3 provides further discussion of other challenges and opportunities related to the presented methods. Supplementary Text 10 discusses the sensitivity of our approach to detecting changes in vascular diameter and blood flow. Supplementary Fig. 7 summarizes how the various types of OCT data went through the described image-processing pipeline, and all processing software codes are publicly opened (see “Code availability”).

In our results with the 3xTg model (see Supplementary Text 4 for use of the specific model within the scope of this study), the earliest cerebral microvascular degeneration (CMD) was detected in the diameter of penetrating vessels (thinner in AD mice), followed by degeneration in their blood flow (lower in AD). It is interesting to see the structural degenerations precede the flow counterparts (Fig. 2). We speculate that once penetrating vessels start becoming pathological in structure (i.e., becoming statistically different from age-matched WT), a compensating adaptation occurs to autoregulate cerebral blood flow. However, this adaptation may not be sustained indefinitely, and eventually blood flow in penetrating vessels decreases below the normal physiological range (95% CI of age-matched WT; see Supplementary Text 5 for further discussion). Although the primary focus of our study is technical demonstration rather than biological discovery (Supplementary Text 4), the observed cortical hypoperfusion in AD is consistent with findings from perfusion magnetic resonance imaging and transcranial Doppler ultrasound studies, which have reported hypoperfusion in various cortical regions in early human AD (see refs. ^33,34 for reviews) as well as findings from AD model mice^35,36.

On a smaller scale, the presented results revealed degenerations in both morphology (the mean capillary length) and topology (the shortest cycle, betweenness, and closeness COV) of capillary vessel networks. The mean capillary length result (becoming shorter with aging in AD only, Fig. 3d) is consistent with the previous finding that capillary length is shorter in AD than WT²³, although the previous study did not longitudinally track the length and used a different AD mouse model (see Supplementary Text 7 for further discussion). The capillary tortuosity did not show significant difference in its fractional changes between AD and WT until the end of measurement (35 WOA), but its rate of change with age was positive in AD only. While vessel diameter and tortuosity can be correlated (e.g., with thicker vessels being less tortuous), it is unlikely that the observed significance in the tortuosity change rate of AD is an influence of diameter changes because the diameter change rate did not show significant difference between AD and WT (Table 1). In contrast, the topological properties have been little studied in the context of aging. Interestingly, the topological changes became pathological earlier than the morphological change (Fig. 3). Also, both the shortest cycle and the mean betweenness increased with aging in AD, likely indicating a higher degree of susceptibility to vascular insult than WT (Supplementary Text 6). This higher susceptibility may be important in AD when considering the increased rate of capillary stalling in AD models³⁷. The presented results also showed degenerations in capillary blood flow (see Supplementary Text 7 for interpretation and potential relation to hypertension).

Finally, the chronological graph showed that most CMDs became apparent between 12 and 25 WOA and preceded the cognitive decline observed from the same animals (Fig. 5c). Some of these CMDs slightly precede even extracellular Aβ deposits in the 3xTg model (see Supplementary Text 11 for details). The correlation graph (Fig. 5d) showed that 19 vascular properties were significantly correlated with the cognitive NOL test score. While some of these confirmed expected correlations, such as those of arteriolar and venular diameter and flow, others revealed new findings. Of particular interest is the betweenness property, which displayed a strong negative correlation with the NOL test score, meaning that higher betweenness was associated with later cognitive decline. It was the earliest capillary vessel property to show significant differences between AD and WT mice (Fig. 5c) and was the most inter-correlated with other vascular alterations in both AD and WT, but with different inter-correlation patterns (Supplementary Fig. 8). These interesting findings suggest that the betweenness of the capillary network is an important component of CMD in neurodegenerative diseases. This spotlight on the relatively novel topological property, now longitudinally measurable in vivo using the presented methods, encourages further studies on how other vascular and non-vascular factors interact with capillary network betweenness and how this interaction contributes to the pathology of vascular and other related systems in aging research.

Aβ clearance plays an important role in AD³⁸. Various overlap** and interacting clearance mechanisms are being studied, including the glymphatic pathway. One of these clearance systems, called perivascular system, involves a number of components such as the cerebrospinal fluid, interstitial fluid, periarterial space, water channel aquaporin-4 in astrocytes³⁹, extracellular space affected by astrocytic endfeet structure⁴⁰, and fiber myelination^41,42. Since many of these components are structurally and functionally interrelated with blood vessels, the methods presented in this paper would be useful in studying how the clearance system components interact with the related vascular properties, how the interaction varies with age, and how the age-dependent interaction differs between AD and normal aging.

Future findings enabled by the presented methods in diverse AD models, when combined with follow-up studies at cellular/molecular level, may determine whether CMD is etiological in AD pathogenesis, one of the long-lasting questions in AD research⁴⁶. The 5-mm cover slip serving as the apex of our window acted as an additional safety measure against compression of the cortex, as it was wider than our 3-mm craniotomy and thus any pressure against the window prior to cementing was applied to the bone surrounding the craniotomy and not the brain tissue itself; it also ensured that the base of the window sat flush just above the cortex without compressing it. Once the window was cemented in place, this minimized any dural scarring or scar tissue ingrowth underneath the window for the duration of our experiment.

The use of an open-skull cranial window has clear strengths and weaknesses against a thinned-skull cranial window. The open-skull method in this study provides higher quality microangiography where individual capillaries are clearly visualized (Fig. 3a) and individual RBC passages through those capillaries are clearly captured in the time courses of the image intensity signal (Supplementary Fig. 5). However, the open-skull method is known to be relatively less robust in a long term; we began with 20 animals, but one animal did not survive until the end of our seven-month longitudinal experiment. In three out of the surviving animals, the imaging quality through the window degraded significantly with time, secondary to scar tissue formation underneath the window such that insufficient number of vessels were visible with imaging—these animals were not included in our analysis. Additionally, three animals were euthanized in the middle of experiment due to headpost damage. We also note that installing a chronic cranial window in mice at younger than 10–12 weeks can be problematic due to incomplete skull growth prior to this age⁴⁷.

OCT imaging

All OCT imaging was performed in vivo. Baseline OCT imaging was obtained seven days following the surgery, and subsequently each animal underwent imaging every four weeks for seven months. After induction of anesthesia and for the duration of imaging, the mouse’s head was stabilized against motion by attaching the affixed metal frame to a proprietary platform with a heating pad and inhaled isoflurane delivery system underneath the OCT system.

All OCT measurements were collected with a commercial SD-OCT system (Thorlabs, Newton, NJ, USA). We used custom LabVIEW software to control the OCT system. The system uses a large-bandwidth near-infrared light source with a center wavelength of 1310 nm, and wavelength bandwidth of 170 nm, which leads to a high axial resolution of 3.5 μm. The system uses a high-speed 2048-pixel line-scan camera to achieve 147,000 A-scans/s. During each imaging session, we acquired angiograms of large pial vessels using a 5× objective lens, with a field of view (FOV) of 1024 × 1024 × 512 (x,y,z) corresponding to an imaging volume of about 3 × 3 × 1.8 mm. Ten angiogram volumes were acquired, motion-corrected⁴⁸, and then volume-averaged prior to diameter measurements, in order to minimize the effect of physiological cardiac fluctuations. Microvessels were imaged using a 10X objective lens with a FOV of 1024 × 1024 × 512 (x,y,z) corresponding to an imaging volume of about 1.5 × 1.5 × 1.8 mm (x,y,z). We measured blood flow in penetrating vessels using Doppler OCT by repeating 8 A-scans at each (x,y) position over a FOV of 512 × 512 × 512 (x,y,z) corresponding to a region of about 1.5 × 1.5 × 1.8 mm. RBC flux data consisted of 512 repeated B-scans with a temporal resolution of 1.4 ms, producing a 4D dataset of size 256 × 256 × 512 × 512 (x,y,z,t), corresponding to an imaging volume of 0.8 × 0.8 × 1.8 mm (x,y,z) over a time period of 0.7 s. Each OCT imaging session took about an hour per mouse; 10 min for angiogram, 5 min for Doppler, 10 min for microangiogram, and 30 min for RBC data. We held imaging sessions routinely during the daytime.

Microangiography analysis

We measured the capillary length and diameter from the segmented image and vectorized graph (Fig. 3a as an example) by measuring the length of the centerline of each capillary segment and the thickness of the segment in the segmented image (see ref. ²⁵ for further details), the histograms of which are shown in Fig. 3b. We also measured the capillary tortuosity from the mean curvature of the vessels, calculated by the mean derivative of the tangent along the vessel⁴⁹. Since tortuosity and diameter can be correlated (e.g., with thick vessels being less tortuous), diameter results should be carefully considered when interpreting tortuosity results. To determine branching order, we leveraged the fact that penetrating vessels appear as dark circles (or shadows) in the structural intensity image. This allowed us to label all penetrating vessels and trace connected vessels as they branched out, thereby finding the closest distance between each capillary and a penetrating vessel along the capillary bed.

Closeness and betweenness aim to elucidate how vessels are connected within the network. Closeness is calculated as the reciprocal sum of the number of vessels along the shortest paths between the vessel and every other vessel in the network. Betweenness counts how many times a single vessel is traversed along the shortest path between any other two vessels within the network (Supplementary Fig. 4). However, these metrics depend on the size of the network, so to account for this we divided the mean values by the mean closeness or betweenness of an idealized honeycomb network with the same number of vessels. Supp. Fig. 4d shows the relationship between mean closeness and betweenness and number of vessels in a honeycomb network, which was used to normalize the values obtained from the experimental data. The COV was normalized in a similar manner.

Lastly, we looked at global properties of the network within the imaging volume: we calculated the capillary number density by dividing the number of capillaries by the volume, and the capillary length density by summing the lengths of all capillaries and dividing by the volume. Fractional dimension was determined using the box-counting method⁵⁰.

Capillary RBC flux measurement

Data preparation

We developed the CNN-based, RBC flux-measuring method by utilizing two sets of RBC-passage data that were simultaneously obtained with OCT and two-photon microscopy (TPM). We trained a 1D CNN using the OCT time traces as the training data, and the corresponding RBC flux determined from the TPM time traces as ground truth (Supplementary Fig. 5 and Supplementary Text 2). We augmented the training dataset by reversing the time of the time traces, and split this augmented dataset into training, validation and testing as 70%, 20 and 10% respectively.

CNN architecture and training

We implemented the state-of-the-art architecture in time-series classification, InceptionTime⁵¹, as shown in Supplementary Fig. 6. We designed our CNN to provide an estimate of uncertainty in its prediction by adapting the loss function as shown in Eq. 1 to maximize the probability of a given prediction $\hat{y,}$ assuming this prediction is derived from a normal probability distribution with mean equal to G, and standard deviation, σ. G is the ground truth, and the standard deviation thus serves as an indicator of uncertainty in the predicted value. We trained our network using the Adam optimizer, with a learning rate of 0.0008 for 50 epochs, and a gradient threshold of 500 which was necessary given our adapted loss function.

$${{{{{\rm{loss}}}}}}=-\log \left(\frac{1}{\sqrt{2\pi }\sigma }\exp \left(-\frac{{(G-\hat{y})}^{2}}{2{\sigma }^{2}}\right)\right)$$

(1)

RBC flux estimation

To track long-term changes in the capillary network blood flow pattern, we acquired 4D data (x,y,z,t) consisting of 512 time points over a period of 0.72 s. We first constructed an angiogram from this data (Fig. 4c) and obtained a vascular network graph from the image as described in Fig. 3a. We extracted the time course from the 4D data for each centerline voxel of the graphed vessels, where RBC passage is known to cause a transient increase in intensity. We then applied the CNN to each time course along the centerline and averaged the predicted flux values along the centerline voxels for each vessel by weighting the predictions according to their associated uncertainties, thus determining the average RBC flux for each vessel (Fig. 4c, bottom). Finally, we determined the mean flux for each dataset (per animal per time point) by averaging the flux values over all vessels, and the COV by dividing the standard deviation by the mean.

Novel object location test

Every 4 weeks from 11 weeks of age, each mouse was cognitively evaluated using a novel object location test that is widely used to assess spatial memory⁵². Our test used an open-field apparatus consisting of a square arena (40 × 40 × 49 cm³). The mouse was habituated to an empty arena with four visually distinct quadrants for 5 min, and then removed. The arena was cleaned with 70% ethanol and dried to eliminate any potential odor cues left beforehand. Then, two identical objects were placed into the northwest (NW) and northeast (NE) quadrants, while the other quadrants remained empty. The mouse explored the arena for 5 min while the video tracked its movement. The mouse was removed, the arena was again cleaned, and both objects were placed again in the NW and southeast (SE) quadrants such that one of the objects was placed in a different location (SE). After 20 min, the mouse was again placed into the arena and allowed to explore the arena for 5 min while its movement was tracked. Differences in movement between the first and second 5-min tracking were used to assess recognition of the novel object’s location.

To determine the sign of cognitive decline, we calculated the discrimination index:

$$Z=\frac{{T}_{{{{{{\rm{SE}}}}}}}}{{T}_{{{{{{\rm{NW}}}}}}}+{T}_{{{{{{\rm{SE}}}}}}}}$$

(2)

where T_SE is the amount of time spent exploring the object in the novel position, and T_NW is the amount of time spent exploring the object in the same position. A z-score of 0.5 or below was considered indicating that the mice did not remember one object being moved to a novel location, and thus was considered as the sign for cognitive decline⁵².

Statistical analysis

We used the linear mixed-effects (LME) method to fit each set of measured time courses to either a linear model or a nonlinear model following a sigmoidal curve ($y=c+\frac{L}{1+\exp (k(b-{{{{{\rm{age}}}}}}))}$), and then selected the model with the lowest Akaike Information Criterion (AIC) following the widely used method^53,54. The AIC also informed our choice of random effects based on the intercept, slope and/or time-group interaction. We normalized each metric for AD and WT groups by the intercept as determined by the fitted model. Outliers were excluded based on Cook’s distance: any value with a Cook’s distance greater than 3 times the mean was considered an outlier, as practiced in many studies^55,56.

The rate of change was simply defined by the slope for the linear model. For the sigmoid model, we defined the rate of change by measuring 90% of the total changes observed during the ages of measurement (11–35 WOA; from the bottom 5% to the top 5%) and then by dividing the 90% change by the age duration corresponding to the 90% change. The same was done on the 95% confidence intervals of the nonlinear fitting to obtain the confidence intervals for the nonlinear slope. When comparing the rate of change between AD and WT, for linear models, we recorded the p value of the interaction term to determine significant differences between slopes. For nonlinear models, we determined the p value by conducting a t test between the estimated parameters $L$ for each nonlinear model. We selected vascular properties for further analysis which were associated with significant p values at an overall level of $\alpha$ = 0.05 after Benjamini-Hochberg correction to control false discoveries. For these metrics, we determined the age at which AD and WT groups differed by assessing the degree of overlap between two confidence intervals (CIs). To account for multiple hypothesis testing using Bonferroni correction, we aimed to find the age at which the groups were significantly different at a rate of α = 0.05/7, which in turn leads to the age at which the 93% CIs just touch. In detail, we computed the distance between the lower CI bound of the group with higher values and the upper CI bound of the other group. The age from which this distance becomes larger than zero was selected as the age at which AD and WT groups started to differ.

Correlation analysis

To study the correlation between cognitive decline and the vascular features studied, we fitted LME models to determine the relationship between the fractional change of each vascular feature and NOL discrimination index. Two such examples are demonstrated on the right of Fig. 5d. We fitted these models while also introducing time-lags of 0, 4, 8 and 12 months. The features which exhibited a significant slope (p < 0.05) were considered to be correlated with cognitive function, with the sign of the slope indicating positive or negative correlation.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The vascular property measurement and novel object location test data generated in this study have been deposited in the Figshare database: https://doi.org/10.6084/m9.figshare.19178885. The raw image data is not available due to its size (larger than 50 terabytes). Source data are provided with this paper.

Code availability

All codes for image processing and analysis were developed in MATLAB. The MATLAB codes are available for non-commercial use in GitHub (https://github.com/optobrain/adp-1-3xtg-cortex)⁵⁷.

References

Zlokovic, B. V. Neurovascular pathways to neurodegeneration in Alzheimer’s disease and other disorders. Nat. Rev. Neurosci. 12, 723–738 (2011).
Article CAS PubMed PubMed Central Google Scholar
Melzer, T. R. et al. Arterial spin labelling reveals an abnormal cerebral perfusion pattern in Parkinson’s disease. Brain J. Neurol. 134, 845–855 (2011).
Article Google Scholar
Chen, J. J., Salat, D. H. & Rosas, H. D. Complex relationships between cerebral blood flow and brain atrophy in early Huntington’s disease. NeuroImage 59, 1043–1051 (2012).
Article PubMed Google Scholar
Ingrisch, M. et al. Quantification of perfusion and permeability in multiple sclerosis: dynamic contrast-enhanced MRI in 3D at 3T. Invest. Radiol. 47, 252–258 (2012).
Article PubMed Google Scholar
Meyer, E. P., Ulmann-Schuler, A., Staufenbiel, M. & Krucker, T. Altered morphology and 3D architecture of brain vasculature in a mouse model for Alzheimer’s disease. Proc. Natl Acad. Sci. USA 105, 3587–3592 (2008).
Article ADS CAS PubMed PubMed Central Google Scholar
Bennett, R. E. et al. Tau induces blood vessel abnormalities and angiogenesis-related gene expression in P301L transgenic mice and human Alzheimer’s disease. Proc. Natl Acad. Sci. USA 115, E1289–E1298 (2018).
Article CAS PubMed PubMed Central Google Scholar
Tarantini, S. et al. Nicotinamide mononucleotide (NMN) supplementation rescues cerebromicrovascular endothelial function and neurovascular coupling responses and improves cognitive function in aged mice. Redox Biol. 24, 101192 (2019).
Article CAS PubMed PubMed Central Google Scholar
Shih, A. Y. et al. Two-photon microscopy as a tool to study blood flow and neurovascular coupling in the rodent brain. J. Cereb. Blood Flow. Metab. 32, 1277–1309 (2012).
Article CAS PubMed PubMed Central Google Scholar
Patterson, G. H. & Piston, D. W. Photobleaching in two-photon excitation microscopy. Biophys. J. 78, 2159–2162 (2000).
Article ADS CAS PubMed PubMed Central Google Scholar
Hopt, A. & Neher, E. Highly nonlinear photodamage in two-photon fluorescence microscopy. Biophys. J. 80, 2029–2036 (2001).
Article ADS CAS PubMed PubMed Central Google Scholar
Masamoto, K. et al. Repeated longitudinal in vivo imaging of neuro-glio-vascular unit at the peripheral boundary of ischemia in mouse cerebral cortex. Neuroscience 212, 190–200 (2012).
Article CAS PubMed Google Scholar
Yoshihara, K. et al. 3D analysis of intracortical microvasculature during chronic hypoxia in mouse brains. Adv. Exp. Med. Biol. 765, 357–363 (2013).
Article CAS PubMed Google Scholar
Yang, G., Pan, F., Chang, P. C., Gooden, F. & Gan, W.-B. Transcranial two-photon imaging of synaptic structures in the cortex of awake head-restrained mice. Methods Mol. Biol. 1010, 35–43 (2013).
Article CAS PubMed PubMed Central Google Scholar
Nishiyama, N., Colonna, J., Shen, E., Carrillo, J. & Nishiyama, H. Long-term in vivo time-lapse imaging of synapse development and plasticity in the cerebellum. J. Neurophysiol. 111, 208–216 (2014).
Article PubMed Google Scholar
Huang, D. et al. Optical coherence tomography. Science 254, 1178–1181 (1991).
Article ADS CAS PubMed PubMed Central Google Scholar
Choi, W. J. Optical coherence tomography angiography in preclinical neuroimaging. Biomed. Eng. Lett. 9, 311–325 (2019).
Article ADS PubMed PubMed Central Google Scholar
**ng, F., Lee, J.-H., Polucha, C. & Lee, J. Three-dimensional imaging of spatio-temporal dynamics of small blood capillary network in the cortex based on optical coherence tomography: a review. J. Innov. Opt. Health Sci. 13, 2030002 (2020).
Article Google Scholar
Srinivasan, V. J. et al. Rapid volumetric angiography of cortical microvasculature with optical coherence tomography. Opt. Lett. 35, 43–45 (2010).
Article ADS PubMed PubMed Central Google Scholar
Srinivasan, V. J. et al. Optical coherence tomography for the quantitative study of cerebrovascular physiology. J. Cereb. Blood Flow. Metab. 31, 1339–1345 (2011).
Article PubMed PubMed Central Google Scholar
Lee, J., Wu, W., Lesage, F. & Boas, D. A. Multiple-capillary measurement of RBC speed, flux, and density with optical coherence tomography. J. Cereb. Blood Flow. Metab. 33, 1707–1710 (2013).
Article PubMed PubMed Central Google Scholar
Akif, A., Walek, K., Polucha, C. & Lee, J. Doppler OCT clutter rejection using variance minimization and offset extrapolation. Biomed. Opt. Express 9, 5340–5352 (2018).
Article CAS PubMed PubMed Central Google Scholar
Stefan, S. & Lee, J. Deep learning toolbox for automated enhancement, segmentation, and graphing of cortical optical coherence tomography microangiograms. Biomed. Opt. Express 11, 7325–7342 (2020).
Article PubMed PubMed Central Google Scholar
Haft-Javaherian, M. et al. Deep convolutional neural networks for segmenting 3D in vivo multiphoton images of vasculature in Alzheimer disease mouse models. PLoS ONE 14, e0213539 (2019).
Article CAS PubMed PubMed Central Google Scholar
Kirst, C. et al. Map** the fine-scale organization and plasticity of the brain vasculature. Cell 180, 780–795.e25 (2020).
Article CAS PubMed Google Scholar
Todorov, M. I. et al. Machine learning analysis of whole mouse brain vasculature. Nat. Methods 17, 442–449 (2020).
Article CAS PubMed PubMed Central Google Scholar
Ji, X. et al. Brain microvasculature has a common topology with local differences in geometry that match metabolic load. Neuron 109, 1168–1187.e13 (2021).
Article CAS PubMed PubMed Central Google Scholar
Marchand, P. J., Lu, X., Zhang, C. & Lesage, F. Validation of red blood cell flux and velocity estimations based on optical coherence tomography intensity fluctuations. Sci. Rep. 10, 19584 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Desjardins, M., Berti, R., Lefebvre, J., Dubeau, S. & Lesage, F. Aging-related differences in cerebral capillary blood flow in anesthetized rats. Neurobiol. Aging 35, 1947–1955 (2014).
Article PubMed Google Scholar
Vogel-Ciernia, A. & Wood, M. A. Examining object location and object recognition memory in mice. Curr. Protoc. Neurosci. 69, 8.31.1–17 (2014).
Article PubMed Google Scholar
Stover, K. R., Campbell, M. A., Van Winssen, C. M. & Brown, R. E. Early detection of cognitive deficits in the 3xTg-AD mouse model of Alzheimer’s disease. Behav. Brain Res. 289, 29–38 (2015).
Article CAS PubMed Google Scholar
Kelly, P. et al. Microvascular ultrastructural changes precede cognitive impairment in the murine APPswe/PS1dE9 model of Alzheimer’s disease. Angiogenesis 20, 567–580 (2017).
Article CAS PubMed PubMed Central Google Scholar
Brothers, R. O., Atlas, N., Cowdrick, K. R. & Buckley, E. M. Cerebrovascular reactivity measured in awake mice using diffuse correlation spectroscopy. Neurophotonics 8, 015007 (2021).
Article PubMed PubMed Central Google Scholar
Wolk, D. A. & Detre, J. A. Arterial spin labeling MRI: an emerging biomarker for Alzheimer’s disease and other neurodegenerative conditions. Curr. Opin. Neurol. 25, 421–428 (2012).
Article CAS PubMed PubMed Central Google Scholar
Sabayan, B. et al. Cerebrovascular hemodynamics in Alzheimer’s disease and vascular dementia: A meta-analysis of transcranial Doppler studies. Ageing Res. Rev. 11, 271–277 (2012).
Article PubMed Google Scholar
Niwa, K., Carlson, G. A. & Iadecola, C. Exogenous A beta1-40 reproduces cerebrovascular alterations resulting from amyloid precursor protein overexpression in mice. J. Cereb. Blood Flow. Metab. 20, 1659–1668 (2000).
Article CAS PubMed Google Scholar
Niwa, K., Kazama, K., Younkin, S. G., Carlson, G. A. & Iadecola, C. Alterations in cerebral blood flow and glucose utilization in mice overexpressing the amyloid precursor protein. Neurobiol. Dis. 9, 61–68 (2002).
Article CAS PubMed Google Scholar
Cruz Hernández, J. C. et al. Neutrophil adhesion in brain capillaries reduces cortical blood flow and impairs memory function in Alzheimer’s disease mouse models. Nat. Neurosci. 22, 413–420 (2019).
Article PubMed PubMed Central Google Scholar
Tarasoff-Conway, J. M. et al. Clearance systems in the brain-implications for Alzheimer disease. Nat. Rev. Neurol. 11, 457–470 (2015).
Article CAS PubMed PubMed Central Google Scholar
Iliff, J. J. et al. A paravascular pathway facilitates CSF flow through the brain parenchyma and the clearance of interstitial solutes, including amyloid β. Sci. Transl. Med. 4, 147ra111 (2012).
Article PubMed PubMed Central Google Scholar
Li, Y. et al. The mechanism of downregulated interstitial fluid drainage following neuronal excitation. Aging Dis. 11, 1407–1422 (2020).
Article PubMed PubMed Central Google Scholar
Wang, A. et al. The drainage of interstitial fluid in the deep brain is controlled by the integrity of myelination. Aging Dis. 10, 937–948 (2019).
Article PubMed PubMed Central Google Scholar
Wang, R. et al. The alteration of brain interstitial fluid drainage with myelination development. Aging Dis. 12, 1729–1740 (2021).
Article CAS PubMed PubMed Central Google Scholar
Liu, P.-P., **e, Y., Meng, X.-Y. & Kang, J.-S. History and progress of hypotheses and clinical trials for Alzheimer’s disease. Signal Transduct. Target. Ther. 4, 29 (2019).
Article PubMed PubMed Central Google Scholar
Jack, C. R. et al. NIA-AA research framework: toward a biological definition of Alzheimer’s disease. Alzheimers Dement. 14, 535–562 (2018).
Article PubMed PubMed Central Google Scholar
Mostany, R. & Portera-Cailliau, C. A craniotomy surgery procedure for chronic brain imaging. J. Vis. Exp. JoVE 680. https://doi.org/10.3791/680 (2008).
O’Reilly, M. A., Muller, A. & Hynynen, K. Ultrasound insertion loss of rat parietal bone appears to be proportional to animal mass at submegahertz frequencies. Ultrasound Med. Biol. 37, 1930–1937 (2011).
Article PubMed PubMed Central Google Scholar
Goldey, G. J. et al. Removable cranial windows for long-term imaging in awake mice. Nat. Protoc. 9, 2515–2538 (2014).
Article CAS PubMed PubMed Central Google Scholar
Lee, J., Srinivasan, V., Radhakrishnan, H. & Boas, D. A. Motion correction for phase-resolved dynamic optical coherence tomography imaging of rodent cerebral cortex. Opt. Express 19, 21258–21270 (2011).
Article ADS PubMed PubMed Central Google Scholar
Bullitt, E., Gerig, G., Pizer, S. M., Lin, W. & Aylward, S. R. Measuring Tortuosity of the Intracerebral Vasculature from MRA Images. IEEE Trans. Med. Imaging 22, 1163–1171 (2003).
Article PubMed PubMed Central Google Scholar
Gould, D. J., Vadakkan, T. J., Poché, R. A. & Dickinson, M. E. Multifractal and lacunarity analysis of microvascular morphology and remodeling. Microcircirculation 18, 136–151 (2011).
Article Google Scholar
Ismail Fawaz, H. et al. Inceptiontime: finding AlexNet for time series classification. Data Min. Knowl. Discov. 34, 1936–1962 (2020).
Article MathSciNet Google Scholar
Antunes, M. & Biala, G. The novel object recognition memory: neurobiology, test procedure, and its modifications. Cogn. Process. 13, 93–110 (2012).
Article CAS PubMed Google Scholar
Bozdogan, H. Model selection and Akaike’s Information Criterion (AIC): the general theory and its analytical extensions. Psychometrika 52, 345–370 (1987).
Article MathSciNet MATH Google Scholar
Müller, S., Scealy, J. L. & Welsh, A. H. Model selection in linear mixed models. Stat. Sci. 28, 135–167 (2013).
Article MathSciNet MATH Google Scholar
Brinkman, L. et al. Independent causal contributions of alpha- and beta-band oscillations during movement selection. J. Neurosci. 36, 8726–8733 (2016).
Article CAS PubMed PubMed Central Google Scholar
Ramirez, D. G. et al. Detecting insulitis in type 1 diabetes with ultrasound phase-change contrast agents. Proc. Natl Acad. Sci. USA 118, e2022523118 (2021).
Article CAS PubMed PubMed Central Google Scholar
Stefan, S. & Lee, J. Near-lifespan longitudinal tracking of brain microvascular morphology, topology, and flow in male mice. GitHub repository release. https://doi.org/10.5281/zenodo.7834685 (2023).

Download references

Acknowledgements

This work was supported by the National Institute on Aging award R01AG067228.

Author information

These authors contributed equally: Konrad W. Walek, Sabina Stefan.

Authors and Affiliations

Warren Alpert Medical School, Brown University, Providence, RI, 02912, USA
Konrad W. Walek & Yu-Wen Alvin Huang
School of Engineering, Brown University, Providence, RI, 02912, USA
Sabina Stefan, Jang-Hoon Lee & Jonghwan Lee
College of Medicine, Drexel University, Philadelphia, PA, 19104, USA
Pooja Puttigampala
Department of Neuroscience, Brown University, Providence, RI, 02912, USA
Anna H. Kim, Seong Wook Park & Christopher Moore
Department of Electrical Engineering, École Polytechnique de Montréal, Montréal, QC, H3T 1J4, Canada
Paul J. Marchand & Frederic Lesage
Department of Biostatistics, Brown University School of Public Health, Providence, RI, 02912, USA
Tao Liu
Department of Molecular Biology, Cell Biology and Biochemistry, Brown University, Providence, RI, 02912, USA
Yu-Wen Alvin Huang
Carney Institute for Brain Science, Brown University, Providence, RI, 02912, USA
Yu-Wen Alvin Huang, Christopher Moore & Jonghwan Lee
Department of Biomedical Engineering, Boston University, Boston, MA, 02215, USA
David A. Boas

Authors

Konrad W. Walek
View author publications
You can also search for this author in PubMed Google Scholar
Sabina Stefan
View author publications
You can also search for this author in PubMed Google Scholar
Jang-Hoon Lee
View author publications
You can also search for this author in PubMed Google Scholar
Pooja Puttigampala
View author publications
You can also search for this author in PubMed Google Scholar
Anna H. Kim
View author publications
You can also search for this author in PubMed Google Scholar
Seong Wook Park
View author publications
You can also search for this author in PubMed Google Scholar
Paul J. Marchand
View author publications
You can also search for this author in PubMed Google Scholar
Frederic Lesage
View author publications
You can also search for this author in PubMed Google Scholar
Tao Liu
View author publications
You can also search for this author in PubMed Google Scholar
Yu-Wen Alvin Huang
View author publications
You can also search for this author in PubMed Google Scholar
David A. Boas
View author publications
You can also search for this author in PubMed Google Scholar
Christopher Moore
View author publications
You can also search for this author in PubMed Google Scholar
Jonghwan Lee
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

K.W. and J.L. conceived of the study. J.L. supervised the study. K.W. acquired main OCT and NOL data. P.J.M. and F.L. acquired the TPM dataset used for training the RBC flux predictor. J.-H.L. acquired supplementary immunofluorescence and OCT data. S.S. and J.L. developed codes for data analysis, and S.S., P.P., J.-H.L., A.H.K., and S.W.P. performed the analysis. S.S. and T.L. performed the statistical analysis. S.S., Y.-W.A.H., D.B., C.M., and J.L. interpreted the results. S.S. and J.L. wrote the manuscript, and all authors revised the manuscript.

Corresponding author

Correspondence to Jonghwan Lee.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Nicolau Beckmann and Hong-Bin Han for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Peer Review File

Reporting Summary

Source data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Walek, K.W., Stefan, S., Lee, JH. et al. Near-lifespan longitudinal tracking of brain microvascular morphology, topology, and flow in male mice. Nat Commun 14, 2982 (2023). https://doi.org/10.1038/s41467-023-38609-z

Download citation

Received: 08 March 2022
Accepted: 09 May 2023
Published: 24 May 2023
DOI: https://doi.org/10.1038/s41467-023-38609-z
Springer Nature Limited

This article is cited by

Neuroinflammation increases oxygen extraction in a mouse model of Alzheimer’s disease
- Chang Liu
- Alfredo Cárdenas-Rivera
- Mohammad A. Yaseen
Alzheimer's Research & Therapy (2024)

Near-lifespan longitudinal tracking of brain microvascular morphology, topology, and flow in male mice

Abstract

Similar content being viewed by others

Introduction

Structural and functional properties of capillary vessel networks

Temporal relationships between microvascular degenerations and cognitive impairment

Discussion

OCT imaging

Microangiography analysis

Capillary RBC flux measurement

Data preparation

CNN architecture and training

RBC flux estimation

Novel object location test

Statistical analysis

Correlation analysis

Reporting summary

Data availability

Code availability

References

Acknowledgements

Author information

Authors and Affiliations

Contributions

Corresponding author

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

This article is cited by

Search

Navigation