In vivo imaging of cerebral microvascular plasticity from birth to death

Abstract

Cerebral function and viability are critically dependent on efficient delivery of oxygen and glucose through the microvasculature. Here, we studied individual microvessels in the intact brain using high-resolution confocal imaging and long-term time-lapse two-photon microscopy across the lifetime of a mouse. In the first postnatal month, we found large-scale sprouting but to our surprise the majority of sprouts underwent pruning and only a small fraction became perfused capillaries. After the first month, microvessel formation and elimination decreased and the net number of vessels stabilized. Although vascular stability was the hallmark of the adult brain, some vessel formation and elimination continued throughout life. In young adult mice, vessel formation was markedly increased after exposure to hypoxia; however, upon return to normoxia, no vessel elimination was observed, suggesting that new vessels constitute a long-term adaptive response to metabolic challenges. This plasticity was markedly reduced in older adults and aging where hypoxia-induced angiogenesis was absent. Our study describes, for the first time in vivo patterns of cerebral microvascular remodeling throughout life. Disruption of the observed balance between baseline turnover and vascular stability may underlie a variety of developmental and age-related degenerative neurological disorders.

Introduction

The brain is one of the most metabolically active organs in the body. Therefore, an appropriate vascular bed must be developed and maintained to meet its high regional oxygen and glucose demands. The brain vascular plexus forms during embryonic stages by endothelial proliferation and sprouting.^{1, 2} This process abates after birth, corresponding to reductions in the expression of proangiogenic molecules.^{3, 4, 5} Significant brain growth, gliogenesis, and synapse rearrangement, which occur in the first postnatal month, are accompanied by concomitant microvascular remodeling. Whether this remodeling shares mechanistic similarities with embryonic angiogenesis remains unclear. As the brain matures into adulthood, vascular density stabilizes, though it is not known whether this is the result of endothelial quiescence or a balanced turnover of vessels. Despite the vascular stability observed at baseline, the brain retains the ability to increase its microvascular density in response to brain activity and hypoxia.^{6, 7} Finally, it remains unclear whether senescence is associated with microvascular regression and if the aging vasculature retains its ability to remodel in response to proangiogenic signaling.^{8, 9}

Current understanding of vascular remodeling relies on traditional quantitative histology, which provides limited information about dynamic cellular events. Here, for the first time, we performed a systematic characterization in vivo of the dynamic properties of brain microvessels, from the early postnatal period into advanced aging. Using two-photon time-lapse imaging in living mice and high-resolution confocal microscopy, we found that the brain microvascular network has distinct cellular and dynamic properties that are specific to each developmental stage.

During the early postnatal period, the microvascular bed undergoes a process of fine-tuning, during which angiogenesis occurs chiefly through short-distance sprouting and concomitant vascular regression. Vascular dynamism wanes in adulthood, though a modest degree of microvascular formation and elimination persists at this stage. In aging, however, the baseline turnover is severely reduced with no remodeling observed over extended imaging periods. While young adult brains retain the ability to form vessels under hypoxic conditions, this capacity is also lost in mature adult and aged brains. Interestingly, large-scale vessel pruning is rarely observed after the neonatal stage and is not characteristic of either hypoxia-induced remodeling or the aging vasculature. In fact, even vessels formed under hypoxia are retained over extended periods after reestablishment of normoxia. The tendency towards vascular stability may be crucial to maintaining a stable neural environment while the low-level vessel turnover may be important for adapting to fluctuations in brain energetic demand.

Our study elucidates long-standing questions regarding the plasticity of microvessels in vivo and provides fundamental information for future studies of the neurovascular unit in the context of developmental and neurodegenerative disorders.

Materials and methods

Mice

CD1 (Charles River Laboratories, Cambridge, MA, USA) or mice expressing green fluorescent protein (GFP) under the Tie2 vascular promoter (Jackson Laboratories, Bar Harbor, ME, USA) were used for neonatal experiments. For in vivo experiments, p7 to p12 mice were used and for fixed tissue analysis, p1 through p45 mice were used. C57BL6 mice (Charles River Laboratories) ranging from 24 to 150 days old were used for adult experiments. CB6F1 (National Institute of Aging) mice from 22 to 25 month old were used for aging experiments. For hypoxia experiments, mice were randomly assigned to hypoxia or control groups. Mice of either sex were used for all experiments. Animals were given ad libitum access to food and water and maintained on a 12-hour light/12-hour dark cycle. Our manuscript was written according to ARRIVE guidelines and experimental protocols were in accordance with the relevant guidelines and regulations of the Institutional Animal Care and Use Committee at Northwestern University and Yale University.

In Vivo Time-Lapse Imaging

Microvessels were imaged through a thinned-skull preparation as previously described.¹⁰ Briefly, mice were anesthetized with isoflurane (neonates) or Ketamine/Xylazine (adults) and the skull was exposed with a midline scalp incision. A skull region of 1-mm diameter over the somatosensory or motor cortex was thinned with a high-speed drill and a microsurgical blade to a final thickness of ∼30 μm. The neonatal skull is very thin and only requires thinning using a blade. Skull thinning is a minimally invasive procedure that lacks many of the side effects associated with craniotomy. The skull was attached to a custom-made steel plate to stabilize the head while imaging. Pups received an intravascular injection of clear blue dye (75 μL/mL in phosphate-buffered saline (PBS), RR co., Santa Ana, CA, USA) and the adults were injected with 100 μL of 1% Thioflavin-S (Sigma, St Louis, MO, USA) to label blood plasma. A mode-locked Ti-sapphire laser (Coherent, Santa Clara, CA, USA) was used with a two-photon microscope (Prairie Technologies, Middleton, WI, USA) and tuned to 720 nm for imaging of Thioflavin-S and 890 for imaging of GFP and clear blue dye using a 490-nm dichroic filter. Images were taken using a water-immersion objective ( × 40, 0.8 NA; zoom × 1, Olympus, Center Valley, PA, USA) at depths up to 300 μm below the pial surface in adults and 120 μm in pups (likely corresponding to layers 2/3) (see Supplementary Figure 2). In pups, the pial vasculature is dramatically changed as the perineural plexus remodels into mature pial vessels. As our studies were focused on the brain, we excluded this area from our quantifications and only used intraparenchymal vessels. Multiple stacks per animal at different X–Y coordinates were taken using a precision motorized stage. Following imaging, the skull plate was removed, the scalp sutured, and the mouse was treated with analgesics and returned to its home cage. For subsequent time points, the scalp was reopened and the same regions were relocated using the pattern of pial blood vessels. Skull windows required cleaning and debridement if time between imaging was short (1 to 10 days), but necessitated additional drilling over longer time points (months).

Labeling of Vessels In Vivo and Immunohistochemistry in Fixed Tissue

For fixed tissue experiments, mice were trans-cardially perfused; the brains were removed, postfixed in 4% paraformaldehyde (PFA) and immersed in sucrose before cryosectioning. For Collagen IV/bromodeoxyuridine (BrdU) dual labeling experiments, pups were injected with 5 mg/kg BrdU (Sigma) 2 hours before perfusion. Collagen IV antigen retrieval was performed by heating sections in sodium citrate buffer. Sections were incubated with rabbit anti-Collagen IV antibody (1:250, Abcam, Cambridge, MA, USA) and goat anti-rabbit Alexa488 secondary antibody (1:500, Invitrogen, Carlsbad, CA, USA). For BrdU labeling, sections were fixed, washed and denatured with 5 M HCl for 15 minutes. Sections were incubated with rat anti-BrdU antibody (1:400, AbD Serotec, Raleigh, NC, USA) followed by goat anti-rat Alexa 555 secondary antibody (1:500, Invitrogen). Tomato-lectin labeling was obtained by transcardiac perfusion of tomato-lectin (50 μL, Vector Lab, Burlingame, CA, USA) before kill. Isolectin B4 (1:50, Vector Lab) stain was performed by incubating sections in IsoB4/NG2 in PBlec buffer followed by 488-conjugated streptavidin (1:100, Invitrogen). Rabbit anti- NG2 antibody (1:50, Millipore, Billerica, MA, USA) and Alexa647 secondary antibody (1:100, Invitrogen) was used to label pericytes. Rabbit ant-Ki67 (1:100, Abcam) incubations were followed by 647-Alexa secondary antibody (1:100, Invitrogen). Sections were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) (1:1,000, Sigma) where indicated. For colabeling experiments of endothelial cells and blood flow in adults, we used the protocol described by Li et al.¹¹ with modifications. Mice were sequentially perfused with PBS, DiI (125 μL/mL), 4% PFA and finally FITC-Dextran in 4% PFA (3 mg/mL).

Quantification of Vessel Changes In Vivo and in Fixed Tissue

For neonatal branching and endothelial proliferation experiments, images were acquired using Zeiss LSM 510 Meta confocal microscope (Thornwood, NY, USA, × 40, NA 1.35, zoom × 1). Equivalent cortical areas were located and z-projections were made from a fixed number of 1 μm optical sections. Vessel branching points were manually counted using the Cell Counter function on NIH ImageJ software (Bethesda, MD, USA) and normalized to imaging area. For BrdU quantification, proliferating endothelial cells were manually counted from each image section and normalized to surface area. For length quantification, vessel images were projected from 15 μm stacks and thresholded on ImageJ. They were converted to a mask, despeckled, and skeletonized. The area of the skeletonized stack was calculated to obtain vessel length. Vessel length calculated in adult mice was comparable to previously published values.¹² High-resolution imaging of Collagen IV and BrdU allowed us to distinguish proliferating perivascular from endothelial cells (see Figure 1B). For in vivo neonatal experiments, NIH ImageJ was used to analyze and quantify formation and elimination of sprouts and mature vessels as a percentage of total branches at time point 1. Sprouts were defined as vascular structures that are connected to another vessel only at one branch point. Mature vessels were defined as structures connected on both ends with plasma flow. Vessels or sprouts with plasma entry were defined as structures that express Tie2–GFP and contain intravascular dye. The degree of perfusion in sprouts and vessels was determined by identifying at time point 1 unperfused sprouts or vessels and checking whether these same structures were perfused at time point 2. Sprouts at time point 1 that turned into perfused vessels by time point 2 were not included in this quantification because it could not be ascertained whether perfusion had occurred during sprouting or after anastomosis. For in vivo experiments in adults and aging mice, branch points throughout the entire image stack were manually quantified at both time points 1 and 2. New branch points or loss of branch points were counted at time point 2 representing vascular formation and elimination, respectively. Vessel formation and elimination were calculated as a percentage of total branch points counted at time point 1. The recorded vascular events were evenly distributed within the quantified volume ranging from 10 to 300 μm below the pial surface. Only when branches could be clearly visualized, were eliminations and formations recorded. We excluded tissue in which the signal-to-noise ratio became too low to recognize existing branches. For in vivo length quantification, we used the Fiji Simple Neurite tracer to trace individual vessels in a fixed tissue volume. Total vessel length was determined for first and final time points and percent difference was calculated. For dual labeling experiments with DiI and FITC-dextran, brain slices were imaged and a standardized number of stacks were compared for colocalization.

Figure 1.

The brain microvasculature undergoes extensive endothelial proliferation and branching in the first postnatal month. (A) Confocal images from Collagen IV (green)- and BrdU (red)-labeled cortical sections showing increasing branching in the postnatal period. The majority of BrdU-labeled cells are located outside vessels and thus are not endothelial. (B) An example of higher magnification images used to distinguish endothelial BrdU nuclei from other perivascular nuclei. (C) Quantification of changes in branching and endothelial cell proliferation in the postnatal period shows that branching plateaus between P15 and P25 while endothelial cell proliferation peaks around P10. One-way analysis of variance (ANOVA) (branching): F_(5,19)=40.60, P<0.0001. One-way ANOVA (endothelial proliferation): F_(5,19)=22.81, P<0.0001. Values are presented as mean+s.e.m. (D) Quantification of vessel length (mm/mm³) in a fixed tissue volume reveals an increase over time. Analysis of endothelial BrdU and total BrdU (cells/mm²) reveals that while endothelial proliferation peaks postnatally, nonendothelial proliferation declines steadily after birth, reaching near-zero levels by p10. (E) Collagen IV labeling shows that vascular branching and vessel length increase over the first postnatal month are highly correlated (r=0.76). Each purple circle represents a single animal (animals used were ages p0, p4, p15, and p25). Black lines denote 95% confidence intervals. Scale bars: (A) 200 μm; (B) 5 μm.

Hypoxic Exposure

Mice were imaged on the two-photon microscope, returned to their cages and placed in a normobaric hypoxic chamber (Coy Laboratory, Grass Lake, MI, USA, O₂ Control Glove Boxes for In Vivo Studies). Gradual oxygen deprivation was attained over a 2-hour period after which mice were housed in the chamber for varying time periods according to the experimental design. At the end of each time period, mice were removed from the chamber and immediately subjected to live imaging as described above. For hypoxia reexposure experiments, mice were initially imaged at baseline (time point 1), introduced into the hypoxic chamber for 28 days and immediately imaged afterwards (time point 2), and finally allowed to return to normoxia and imaged after 56 days (time point 3).

Statistical Analysis

All statistical parameters were calculated using STATISTICA and Prism (GraphPad, La Jolla, CA, USA) softwares. Group differences were determined using one-way or factorial analysis of variance (ANOVA). Post hoc comparisons were made using Tukey's HSD (Honest Significant Difference) test. Paired t-test was used for repeated measurements. Proportion differences were determined using χ² test. Correlations were determined using linear regression with confidence interval of 95%. P<0.05 was considered statistically significant.

Results

During the First Postnatal Month There Is Extensive Expansion of the Brain Microvascular Network

To determine the degree of changes in the postnatal microvasculature, we analyzed confocal images of the mouse cerebral cortex at ages ranging from P0 to P45. Vessels were immunostained for the extracellular matrix protein Collagen IV or isolectinB4, which gives a strong and reliable labeling of microvascular structures (Supplementary Figures 1A and 1B). We found a sharp increase in vessel branching at P10 that stabilized to adult levels between P15 and P25 (Figures 1A and 1C; Supplementary Figure 1B) and which was accompanied by a steady increase in vessel length (Figure 1D), which strongly correlated with branch increases (Figure 1E, P<0.0001). To determine the rate of endothelial cell proliferation, we double-labeled brains with an antibody for the proliferation marker BrdU and the vascular markers collagen IV and isolectinB4. At various postnatal ages, pups were given an injection of BrdU 2 hours before brain harvesting. We then obtained high-resolution confocal images to distinguish BrdU-positive endothelial nuclei (embedded within the collagen IV basal lamina) from those of pericytes (sandwiched between two collagen IV or isolectinB4 layers) and other vessel-associated cells (lying outside the collagen IV layer) (Figures 1A and 1B). Despite careful analysis and high-resolution imaging, it is possible that some of the BrdU-positive cells identified as endothelial were, in fact, pericytes, leading to a slight overestimation of endothelial proliferation. Therefore, tissues were also stained with an NG2 antibody to label pericytes (Supplementary Figure 1C), which more clearly differentiated the two cell types. We found a substantial number of BrdU-positive cells in all areas of the brain, but the majority of them were not associated with blood vessels and, as they were often NG2-positive, they likely represented oligodendrocyte precursor cells (Figures 1A and 1B; Supplementary Figure 1D). The number of these BrdU-positive cells declined sharply after birth, reaching near-zero levels after p10 (Figure 1D). However, the smaller population of proliferating endothelial cells demonstrated a peak around P10, which gradually dropped to negligible levels by P25 (Figures 1C and 1D; Supplementary Figure 1D). Pericyte proliferation, by contrast, steadily declined after birth (Supplementary Figure 1E). Thus, during the first postnatal month, a period characterized by significant gliogenesis and synaptic rearrangement, there is ongoing endothelial proliferation and vessel branching that significantly expands the vascular network.

Postnatal Vascular Branching Occurs by Sprouting Over Short Distances and Does Not Require Local Endothelial Proliferation

It has been shown in the embryonic brain, that vascular expansion occurs predominantly through a process of widespread sprouting.¹³ In contrast, in the postnatal period, the degree and mechanisms of brain vascular network expansion have not been fully characterized. To visualize sprouts, we intravascularly injected the carbocyanine dye, DiI, which labels vessel membranes, including those of unperfused sprouts.¹¹ High-resolution confocal imaging in P5 to P10 mice revealed the unambiguous presence of endothelial tip cells (Figures 2A, 2B, 2D, 2E, and 2H). These specialized cells, which have been mostly studied during early development, are localized at the leading fronts of sprouts where they respond to chemotactic guidance cues and are trailed by actively proliferating stalk cells.^{13, 14} In the postnatal brain; however, we found that sprouts often consisted only of one tip cell (Figure 2A), sometimes still located within the parent vessel (Figures 2B and 2C). These unicellular sprouts spanned short distances between nearby microvessels (Figure 2C). Immunolabeling with Collagen IV demonstrates the presence of thin segments of basal lamina, which are pervasive during the first postnatal month (Figures 2F–2I). These segments may represent empty basal membrane sleeves of retracted vessels, or, alternately, newly connected sprouts that are beginning to form and whose nuclei have not migrated from the parent vessels. Interestingly, proliferating endothelial cells were rarely observed adjacent to these thin segments (Figure 2F; BrdU-positive endothelial cells within 50 μm of thin segments: 5/181 vascular segments, n=7 mice p15 to 25 injected with BrdU every other day for 10 days). If basement membrane segments represent growing sprouts, this would suggest that sprouting over short distances leads to increased branching, independently of local endothelial cell proliferation. Indeed, sprouts were also found to contain no BrdU or Ki67 staining (Figures 2D and 2E). This is consistent with reports suggesting that endothelial migration and proliferation are functionally distinct and separable processes.^{14, 15}

Figure 2.

Short-distance sprouting occurs without local endothelial proliferation. (A–C), Confocal images of DiI-labeled vessels in P4 mice showing protruding vascular sprouts with numerous filopodia. At this early age, 4′,6-diamidino-2-phenylindole (DAPI) staining reveals that sprouts span short distances and are usually unicellular (Ai, asterix); in some cases tip cell nuclei may remain within the parent vessel (Bi, Ci) entailing that sprouts may reach their nearby targets through the extension of cellular processes (C). (D, E), IsolectinB4-labeled sprout (green) exists in the absence of local endothelial cell proliferation using BrdU (D) or Ki67 (E) proliferation markers (red). (F, H), Collagen IV-labeled structures (green, arrowheads) that may be sprouts, anastomosed segments, or retracting vessels are not adjacent to proliferating endothelial cells (BrdU, red). (H, I) Collagen IV/DAPI staining shows that thin Collagen IV structures may contain a nucleus (H) or be acellular (I). Scale bars: (A–F) 10 μm; (G–I) 5 μm.

In Vivo Two-Photon Imaging Reveals that Microvascular Patterning Occurs Through Concurrent Vascular Formation and Elimination

To better understand the precise dynamic processes through which the final microvascular pattern is achieved in the postnatal brain, we developed methods to image vessels using in vivo time-lapse two-photon microscopy in pups aged P7 to P12. We utilized a Tie2–GFP transgenic reporter mouse in which endothelial cells are fluorescently labeled, in combination with a fluorescent dye injected intravascularly, to simultaneously detect vascular structures and plasma flow (Figure 3A). This technique allowed us to repeatedly image the superficial cortex over intervals of up to 2 days, providing information about the fates of individual sprouts and vessels and the timing of lumen formation.

Figure 3.

Microvascular patterning occurs through concurrent vascular formation and elimination. (A) Two-photon imaging through thinned skull of neonatal Tie2–GFP pups for time-lapse imaging at various ages between P7 and P12. (B–G) Using Tie2–GFP to label endothelial cells and an intravascular fluorescent dye (clear blue), the same vascular structures can be observed over a time interval of 30 hours. Dynamic changes include sprout formation (B, Bi, yellow arrowheads) and elimination (D, Di, blue arrowheads) as well as vessel formation (C, Ci, yellow arrowheads) and elimination (E, Ei, blue arrowheads). Some sprouts contained a trickle of plasma (F, red arrowheads) whereas others were unperfused (F, white arrowheads). Unperfused vessels (G, white arrowheads) were partially (Gi, green arrowhead) or fully (Gi, red arrowheads) perfused at a second time point. (H) Sprout outcome quantified as a percentage of total branch points at TP1. The majority of sprouts were eliminated by the second time point of imaging. One-way analysis of variance (ANOVA): F_{(2, 15)}=26.81, P<0.0001; post-hoc Tukey's tests: P<0.001 (n=98 sprouts; six animals). (I) Newly perfused vessels or sprouts quantified as percentages of total lumenization events at TP2. Establishment of flow was significantly more likely to occur following complete anastomosis than concurrent with the sprout growth. χ² test: P<0.0001 (n=45 new perfusions, six animals). (J) Eliminated vessels or sprouts quantified as percentages of total eliminations at TP2. Sprouts were significantly more likely to get eliminated than vessels. χ² test: P<0.0001 (n=83 eliminations, six animals). Values are presented as mean+s.e.m; *P<0.05 throughout. Scale bar: 20 μm.

Consistent with our observations in fixed tissues, in vivo imaging revealed the presence of a large number of vascular sprouts (defined as GFP-labeled processes connected to only one parent vessel) at the initial time point (Figures 3C, 3D, and 3F). Surprisingly, a large majority of sprouts were eliminated by the second imaging time point (Figures 3D and 3H) and only a small percentage had matured into a fully connected vessel (Figures 3C and 3H). To determine the relative contribution of sprouting to the overall increase in vascular length, we manually traced the vessels using 3D tracing software, in a subset of neonatal mice (p6 to 9) over 1-day in vivo imaging intervals. This revealed that blood vessel length, within a standardized tissue volume, grew by 23%±8% over 1 day. New vessels only contributed to 30%±5% of this increase. The remaining length is likely a result of the elongation of existing vessels. The contribution of sprouts to the overall length is likely to be small given that they are often quite short (10 to 15 μm) and very transient.

The majority of sprouts observed were either unperfused or contained only a trickle of plasma and remained so even when sprouts were imaged over hours (Figures 3B, 3C, 3D, and 3F). Examining the dynamics of perfusion in the growing sprouts revealed that lumen maturation and plasma entry were more likely to occur following completion of anastomosis rather than concurrent with sprout growth (Figure 3I). This is consistent with developmental studies that show that vessel maturation and stability closely follow anastomosis.^{16, 17} Indeed, once connected, vessels do appear to become very stable, compared with sprouts, as very few vessels were eliminated (10.4%±2.6% P<0.0001) (Figure 3J).

The initial formation of excess sprouts, of which only a small number go on to form mature vessels, is reminiscent of the process of transient filopodia formation during synaptic postnatal maturation.¹⁰ This pruning strategy may allow for a more optimized matching between regional metabolic needs and final microvascular density.

The Adult and Aging Microvasculature Display Long-Term Stability but Low-Level Vessel Turnover Is Lost in Aging

Following the first postnatal month, synaptic remodeling and glial proliferation are significantly reduced.¹⁸ However, it remains unclear to what extent microvascular remodeling declines once the brain has matured into adulthood and how it is affected by aging. To address these questions, we used long-term in vivo imaging in mice 24 to 900 days old. Vessels were visualized using an intravascular dye because GFP expression in adult Tie2–GFP mice is down-regulated and insufficiently bright. Imaging at 1-month interval revealed a steady decline in both vessel formation and elimination following the first postnatal month such that in 74- to 94-day-old mice there was only a small turnover over 1 month (Figures 4A–4G). Though modest, this rate of change could have a substantial cumulative impact over the lifetime of a mouse.

Figure 4.

Long-term microvessel stability in adulthood by low-level turnover is lost in aging. (A–F, H), Two-photon time-lapse imaging of microvessels labeled with intravascular thioflavin-S (TS). (A, B), Ongoing vascular remodeling towards the end of the first postnatal month (24 to 31 day age group). (A) Representative images showing sprout growth into a fully connected vessel (blue arrowheads) over a 1-month imaging interval. Sprouting was not observed in older age groups. (B) Representative images showing vessel elimination (yellow arrowhead) and vessel formations (red arrowheads) in close proximity over a 1-month imaging interval. (C, D), Microvascular stability in the maturing brain is concomitant with low-level turnover (32 to 50 day age group). Representative images showing largely unchanged vascular structures over a 1-month imaging interval with one vessel elimination (C, yellow arrowhead) and one vessel formation (D, red arrowhead) observed in two separate areas. (E, F), Microvascular stability and low-level turnover is persistently seen in older adults (74 to 94 day age groups). Representative images showing unchanged vascular structures over a 1-month imaging interval (F) and a single vessel elimination (E, yellow arrowhead) observed in two separate areas. (G) Quantification of branch point formation and elimination during a 1-month imaging interval (percentage of total branch points at TP1). Microvascular formation and elimination are significantly higher in the 24 to 31 day age group compared with older mice; however, only the aging group displays complete absence of vascular remodeling. One-way analysis of variance (ANOVA): F_{(6, 16)}=6.8437, P<0.001; post-hoc Tukey's tests: P<0.05 (n=4,209 branch points; 13 animals). (H) Representative images from a 22-month-old mouse showing remarkable microvascular stability over a 7-month interval. No elimination or formation events were noted for this group. Scale bar: 10 μm; values are presented as mean+s.e.m; *P<0.05 throughout.

We next applied our imaging strategy to mice 22- to 25-month-old over 1- to 7-month imaging intervals to determine the effect of aging on microvascular stability. It is now mostly accepted that, in the absence of neurodegenerative conditions such as Alzheimer's disease, there is no overt neuronal death in aging^{19, 20} though there is likely a gradual decline in the number of synapses.²¹ It has been shown that concomitant functional changes occur in the cerebral microcirculation;^{22, 23} however, the structural effects of aging on microvessels remain controversial. Histologic studies characterizing cerebral microvascular density in aging have shown changes ranging from microvascular increases to vessels loss or stability.^{24, 25, 26}

Our ability to track individual vessels longitudinally allowed us, for the first time, to precisely measure the rates of vessel formation and elimination in vivo over long intervals. Surprisingly, we found that in aging there was no acceleration of vessel loss; in fact, we observed a trend toward even greater vascular stability than seen in adults (Figures 4G and 4H). Therefore, although existing vessels appear to be very stable into advanced senescence, they have lost the limited turnover observed in adults.

Hypoxia Induces Robust Angiogenesis in the Adult Brain and Newly Formed Vessels Remain Stable in the Long Term

The brain vasculature must provide a steady supply of oxygen and glucose to fulfill the high metabolic needs of neural cells. Even a small mismatch between metabolic demand and blood flow could have a deleterious impact on normal brain function and structure.²⁷ Previous studies have demonstrated that chronic hypoxia can cause an expansion of the capillary network;^{28, 29, 30} however, the dynamic vascular events contributing to this expansion are poorly understood. Therefore, it is unclear to what extent this process resembles neonatal angiogenesis in terms of sprouting, vessel formation, regression, and the fate of newly formed vessels over time. To address these questions, we used time-lapse imaging in young adult mice exposed to chronic normobaric hypoxia and this allowed us to track even rare or transient changes to the microvascular tree over time. Exposure to 15% O₂ for 28 days led to no detectable change in vessel formation or elimination rates (Figure 5D). However, when oxygen was reduced to 10%, vessels underwent robust angiogenesis, which reached a plateau following 2 weeks of exposure (Figures 5A, 5D, and 5E). Interestingly, the effect of hypoxia was restricted to vessel formation, as no significant change was observed in the rates of vessel elimination (Figure 5D). Sprouting vessels were observed using both confocal imaging in fixed tissues and in vivo two-photon microscopy (Figures 5B and 5C), albeit much less frequently than during early postnatal development, when the overall magnitude of vascular remodeling is much greater than with hypoxia-induced angiogenesis in adults. Furthermore, quantification of equivalent tissue volumes revealed that vessel formations only contributed to a 4.4%±1% increase in vascular length. Unlike in neonates, the increase in vascular length could be fully accounted for by new vessels. Thus, though robust at this age, hypoxia-induced angiogenesis cause proportionally minor changes to the dense vascular beds in adult brain tissue. Combined with the lack of appreciable vessel elimination and sprout pruning, this suggests that microvascular remodeling is constrained in the adult brain compared to neonates.

Figure 5.

Reduced oxygen triggers sprouting angiogenesis in the young adult brain but newly formed vessels remain stable in the long term. (A, B) Two-photon images before and after hypoxia showing multiple vessel formations (red arrowheads) and sprout formation (blue arrowhead) over a 1-month imaging interval. (C) Representative confocal images of dual perfusion with FITC-dextran and DiI after 15 days of 10% hypoxia exposure showing the presence of a partially perfused sprout (blue arrowhead). (D, E), Branch point formation and elimination quantified as a percentage of total branch points at TP1 after a 1-month imaging interval. (D) Exposure to 10% hypoxia over a 1-month interval causes a robust increase in vessel formation as compared with normoxic controls and mice exposed to 15% hypoxia but it has no effect on vessel elimination. One-way analysis of variance (ANOVA): F_(4,16)=6.0121, P<0.01; post-hoc Tukey's tests: P<0.001 (n=2,542 branch points; 12 animals). (E) Vessel formation is significantly increased and reaches a plateau after 14 days of exposure to 10% hypoxia. Factorial ANOVA (oxygen level × duration of hypoxia interaction): F_{(4, 34)}=3.6853, P=0.01; post-hoc Tukey's tests: P<0.05 (n=3,665 branch points; 24 animals). (F) Branch point formation and elimination quantified as a percentage of total branch points counted at each preceding imaging point. Reestablishment of normoxia following a hypoxic exposure does not lead to increased rates of vessel elimination. Paired t-tests: P>0.05 (n=252 branch points; four animals). (G) Long-term imaging across three time points from a prehypoxic baseline to reestablishment of normoxia. This panel shows the formation of a new a blood vessel after the first hypoxic exposure (day 30, red arrowhead) and its persistence through reestablishment of normoxia (day 86, white arrowhead). Scale bar: 10 μm; values are presented as mean+s.e.m; *P<0.05 throughout.

Given the overall vascular stability observed in the unperturbed adult brain, we asked whether new vessels formed under hypoxic conditions would also remain stable over the long term once normal oxygen concentrations were reestablished. Using in vivo imaging in mice previously exposed to 10% O₂ for 28 days, we observed that both newly formed and preexisting vessels remained remarkably stable over observation periods of up to 2 months (Figures 5F and 5G), suggesting that these vessels are retained as an adaptive response for future hypoxia exposures.

The Mature Brain Loses its Ability to Undergo Angiogenesis in Response to Hypoxia

We have shown that in the young adult brain, the microvasculature displays only modest turnover in normoxia, but retains significant plasticity when exposed to hypoxia. Baseline turnover is markedly reduced as the brain becomes more mature (Figure 4E), suggesting that the mature vasculature is inherently less plastic and may be less responsive to metabolic perturbation. Previous studies in aging mice have reported hypoxia-induced angiogenesis comparable to that seen in adults,^{9, 29} while others have observed an attenuated angiogenic response to exogenous VEGF in aging rodents.⁸ We used time-lapse imaging in aging mice subjected to 10% oxygen over a period of 1 month to definitively answer this question. Remarkably, hypoxic exposure did not lead to a significant increase in vessel formation in 4- to 5-month-old mice at and failed to induce any angiogenic response in aging mice (Figures 6A and 6B). This shows that the cerebral vasculature begins to lose its capacity for remodeling during adulthood, such that it cannot respond to even severe metabolic challenges. Although the mechanism is unclear, a variety of age-related cellular changes or disruption in angiogenesis-⁸ or hypoxia-related signaling pathways³¹ may be involved. In the long term, this deficient vascular response may cause a significant mismatch between metabolic needs and oxygen supply leading to synaptic and neurovascular injury,³² which may constitute an underlying mechanism of age-related neuropathology.

Figure 6.

Hypoxia-induced vessel formation is dramatically reduced in mature brains. (A) Representative image from a 25-month-old mouse showing microvascular stability and lack of angiogenesis despite exposure to 10% hypoxia over a 1-month interval. (B) Branch point formation and elimination quantified as a percentage of total branch points at TP1 after a 1-month imaging interval. There are no significant differences in vessel formation or elimination between hypoxic brains and their normoxic controls in both adult and aging mice; however, in younger adults hypoxic exposure significantly and preferentially increases vessel formation. Factorial analysis of variance (ANOVA) (age × oxygen level interaction): F_(2,18)=20.08, P<0.0001; post-hoc Tukey's tests: P>0.05 for aging vessel formation and P<0.001 for adult vessel formation under hypoxia (n=5,557 branch points; 24 animals). Scale bar: 10 μm; values are presented as mean+s.e.m; *P<0.05 throughout.

Discussion

A better understanding of the mechanisms of microvascular remodeling and maintenance in the brain is critical for studying the neurovascular unit in health and disease. It has been proposed that defects in the ability of the brain to sustain microvascular plasticity may underlie a variety of neuropathological conditions.^{33, 34} However, it is still unclear to what extent cerebral microvascular remodeling occurs after birth and whether features of this remodeling change throughout life. To bridge this gap in knowledge, we performed, for the first time, a comprehensive in vivo imaging study of long-term cerebral microvascular plasticity across all stages of life, ranging from the early postnatal period into late senescence.

Our study reveals distinct patterns of microvascular remodeling across the different stages. In the early postnatal brain, we found extensive expansion of the microvascular network by both sprouting, endothelial proliferation, and vessel elongation. We observed that many sprouts consisted of a single tip cell, in contrast to the multicellular sprouting observed in earlier developmental stages and other vascular beds.¹⁴ This suggests that at this stage, most sprouting is likely aimed at forming capillaries connecting terminal arterioles and venules. These short distances (20 to 40 μm) could be spanned by unicellular sprouts without the need for actively proliferating stalk cells in trail. Because most oxygen diffusion in the brain occurs through these smallest blood vessels,³⁵ this final process of vascular refinement must be critical for establishing a microvascular network well-adapted to regional metabolic needs. Given the global organizational changes still underway in the developing postnatal brain, a lack of synchrony in the development of neural and microvascular structures could potentially lead to long-lasting neurologic consequences.

In vivo time-lapse imaging revealed that a surprisingly large number of postnatal sprouts are eliminated. This suggests that postnatal refinement of the capillary network occurs by a strategy of redundant vascular sprouting followed by pruning, similar to what is thought to occur during maturation of neural circuits.³⁶ As in neural development, microenvironmental factors, such as neural activity, may play a role in modulating vascular sprouting and pruning. To this end, astrocytes, which closely associate with synapses and vascular sprouts, may modulate the effects of neural activity on vessel growth.

It is noteworthy that most sprouts lacked a visible lumen or had very restricted plasma entry. Shear stress has been shown to prevent endothelial cell sprouting³⁷ and proliferation.³⁸ Therefore, the lack of blood flow in developing sprouts may maintain their proangiogenic profile. Sprouts appear to be more vulnerable to pruning than connected vessels, which are quite stable. This stability may be triggered, in part, by the establishment of laminar blood flow. Shear stress forces activate a host of endothelial signaling cascades, including VEGF receptor 2 phosphorylation³⁹ and has been implicated in vessel stabilization.⁴⁰

Extending our in vivo time-lapse imaging to older age groups showed that in the adult brain, microvascular remodeling declines and the vascular density stabilizes. Despite this, a small number of vascular formations and eliminations continue to take place. This low-level turnover likely leads to a substantial restructuring of the brain vasculature over the lifetime of the mouse. It is possible that this limited plasticity represents the response of the adult vasculature to minor metabolic changes in the brain, while maintaining a neural environment that is fundamentally stable. However, this vascular plasticity appears to be lost over time as aging vessels displayed complete absence of remodeling even over extended imaging intervals. Given the multitude of changes, which occur in the aging brain, it is surprising that under normal conditions, blood vessels are not eliminated. A consensus on the effect of aging on capillary density has not been definitive;²⁶ however, our sensitive methods of tracking individual vessels longitudinally provide reliable evidence for sustained stability during aging. Although active vessel elimination in the aging brain is absent under normal conditions, it is likely that vessel loss occurs during a variety of pathological conditions. Given the lack of plasticity in the aging vasculature, such losses are unlikely to be compensated with the formation of new vessels. The loss of turnover in aging brains might eventually result in a mismatch between energetic supply and demand and may underlie a variety of age-related neural defects.

The persistent plasticity of the young adult (up to 3 months) microvasculature becomes manifest when the brain is exposed to metabolic challenges.^{29, 30} Using time-lapse imaging, we observed robust vessel formation under low oxygen conditions as compared with normoxic controls in this age group. However, even under these conditions, the response was reduced compared with the remodeling observed in the neonate. Angiogenesis was only initiated after oxygen levels were dropped to a moderately low level (10%) and it ceased by the second week of exposure. Moreover, sprouts observed in both fixed tissue and live imaging were much rarer than in the neonate and changes in total vessel length were minor, suggesting that angiogenesis in adults is relatively limited even after strong stimuli such as chronic hypoxia. Furthermore, few vessels were eliminated during hypoxic exposure, underscoring the tendency of the adult vasculature to stabilize, even under conditions favorable to remodeling. Additional evidence for this fundamental feature comes from the finding that vessels formed under hypoxia were retained even after a prolonged return to normoxia. Though previous studies have suggested that these vessels are eliminated,⁷ our method unambiguously shows that both new and preexisting vessels are retained following reestablishment of normoxia. Although the newly formed vessels are unlikely to be critical for the proper function of the brain under normoxic conditions, they may prepare the brain for additional bouts of hypoxia.

While some studies report that metabolic challenges cause vessel formation in the adult brain,⁴¹ our results using time-lapse imaging suggest that the response to hypoxia ceases by 4 months of age. Discrepancies may arise from differences in technical and quantitative approaches. The failure of mature brains to remodel in response to reduced oxygen parallels the declining plasticity observed at baseline normoxic conditions. The diminishing ability of vessels to compensate for changes in metabolic demands could lead to cellular and synaptic dysfunction and may explain the particular vulnerability of aging brains to conditions of reduced oxygen or blood flow.³²

The authors declare no conflict of interest.