Four-Dimensional Microvascular Analysis Reveals That Regenerative Angiogenesis in Ischemic Muscle Produces a Flawed MicrocirculationNovelty and Significance
Rationale: Angiogenesis occurs after ischemic injury to skeletal muscle, and enhancing this response has been a therapeutic goal. However, to appropriately deliver oxygen, a precisely organized and exquisitely responsive microcirculation must form. Whether these network attributes exist in a regenerated microcirculation is unknown, and methodologies for answering this have been lacking.
Objective: To develop 4-dimensional methodologies for elucidating microarchitecture and function of the reconstructed microcirculation in skeletal muscle.
Methods and Results: We established a model of complete microcirculatory regeneration after ischemia-induced obliteration in the mouse extensor digitorum longus muscle. Dynamic imaging of red blood cells revealed the regeneration of an extensive network of flowing neo-microvessels, which after 14 days structurally resembled that of uninjured muscle. However, the skeletal muscle remained hypoxic. Red blood cell transit analysis revealed slow and stalled flow in the regenerated capillaries and extensive arteriolar-venular shunting. Furthermore, spatial heterogeneity in capillary red cell transit was highly constrained, and red blood cell oxygen saturation was low and inappropriately variable. These abnormalities persisted to 120 days after injury. To determine whether the regenerated microcirculation could regulate flow, the muscle was subjected to local hypoxia using an oxygen-permeable membrane. Hypoxia promptly increased red cell velocity and flux in control capillaries, but in neocapillaries, the response was blunted. Three-dimensional confocal imaging revealed that neoarterioles were aberrantly covered by smooth muscle cells, with increased interprocess spacing and haphazard actin microfilament bundles.
Conclusions: Despite robust neovascularization, the microcirculation formed by regenerative angiogenesis in skeletal muscle is profoundly flawed in both structure and function, with no evidence for normalizing over time. This network-level dysfunction must be recognized and overcome to advance regenerative approaches for ischemic disease.
It is well established that a robust angiogenesis response can occur after ischemic injury to skeletal muscle. This neovascularization response, together with opening of collateral vessels, can restore blood flow to the otherwise compromised muscle.1,2 The innate capacity to regenerate a vasculature in muscle has also provided an important rationale for efforts to augment angiogenesis for individuals with peripheral vascular disease. However, to date, such strategies have met with little to no success.3,4 There may be several reasons for this lack of success, but a fundamental and unanswered question is the extent to which a regenerated microvasculature in adult muscle can meet the metabolic needs of the tissue.
Editorial, see p 1379
In This Issue, see p 1367
Meet the First Author, see p 1368
The microcirculation in normal skeletal muscle consists of a highly organized network of arterioles, capillaries, and venules, structurally arranged to optimize oxygen transport. This microcirculatory network provides the necessary conduits for red blood cells (RBCs) to deliver their oxygen. In addition, the network dynamically controls the passage of RBCs throughout the tissue, ensuring that local RBC delivery is tightly coupled to the metabolic needs of the tissue. An important driver of this control loop is the local oxygen content, which can inform the feeder arterioles to dilate or constrict as appropriate.5 Thus, integration of precise structural and functional attributes of the microcirculation are critical to oxygenating muscle tissues. However, the attributes of a regenerated microcirculation in skeletal muscle are largely uncharacterized. The network microarchitecture is poorly understood, and whether the network can dynamically regulate RBC transit is unknown.
One of the best-studied models of regenerative angiogenesis after ischemia is ligation or excision of the femoral artery in mice.2 Use of laser Doppler imaging has established near complete return of bulk blood flow to the ischemic mouse hindlimb, particularly in the C57Bl/6 strain.6,7 This return of flow has been shown to be as a result of both collateral vessel opening1,6–8 and generation of new microvessels,2,6 as established from angiographic and histological studies. However, standard histology techniques do not address network-level architecture of the microvasculature. As well, in vivo imaging tools typically lack the spatial resolution to resolve distinct microvessels or to measure blood flow within them. Accordingly, answering whether the regenerated microcirculation is optimized for oxygen delivery requires different methodologies that can provide higher spatial and temporal resolution and are suitable for evaluating skeletal muscle angiogenesis.
Herein, we describe an integrated, high-resolution 4-dimensional imaging approach for studying the regenerated microvasculature in skeletal muscle. We used this approach to determine whether the reconstructed microcirculation recapitulates the anatomy and physiology of a normal microvasculature. We report that despite extensive angiogenesis and remodeling, the regenerated microcirculation is a profoundly flawed network that cannot adequately control the delivery of RBCs.
A detailed description of methods and materials is provided in the Online Data Supplement and includes information on histology, immunostaining, hypoxia detection, analysis of network architecture, quantifying RBC velocity, and statistical analysis.
Mouse Hindlimb Ischemia
Experiments were conducted in accordance with the University of Western Ontario’s Animal Care and Use Subcommittee. Male C57BL/6J mice (Jackson Laboratories, Bar Harbor, ME) 12 weeks of age were anesthetized with ketamine (80 mg/kg) and xylazine (10 mg/kg) administered intraperitoneally. Hindlimb ischemia was induced by ligating the right femoral artery above and below the profunda femoris branch using 6-0 silk sutures and excising the intervening 5- to 6-mm portion of artery.2,9
Intravital Video Microscopy
RBC transit in individual microvessels within a 50-μm-deep zone across the entire mouse extensor digitorum longus (EDL) surface was assessed by epifluorescent intravital video microscopy. Briefly, mice were anesthetized with ketamine and xylazine intraperitoneally, and a longitudinal incision was made over the anterior hindlimb. Tibialis anterior and peroneus longus muscles were separated from surrounding fascia and spread apart to reveal the underlying EDL.10 The EDL was covered with an 8×8-mm glass coverslip and positioned face-down on the stage of an inverted microscope (Olympus IX81), maintaining body temperature at 37°C. After a 20-minute stabilization period, the EDL was epi-illuminated with a 120 W Mercury X-Cite high-pressure bulb light source via a 10× objective (Olympus UPlanSApo). RBC transit was visualized by either ultraviolet light epi-illumination (DAPI U-MWU2: 330–385 nm excitation filter, 420 nm emission filter) or blue light epi-illumination (U-MWIBA2: 460–490 nm excitation filter, 510–550 nm emission filter)9,11 after intrapenile injection of fluorescein isothiocyanate–labeled dextran (2×106 MW; 20 mg/mL, 30 μL; Sigma). Video recordings (696×520 pixels, 21 images/s) were captured using a cooled coupled device camera (Rolera-XR; QImaging) and displayed in real-time on a computer monitor for at least 315 images (15 seconds). For all studies, the entire EDL surface was recorded and analyzed (7–10 fields of view). Video sequences were digitized and stored as uncompressed AVI (audio video interleave) files for postprocessing using custom acquisition software (NeoVision) and in-house software written in the MATLAB (Mathworks) programming environment. RBC hemoglobin O2 saturation in capillaries was assessed using transmission intravital video microscopy and dual-camera detection of differential oxy- and deoxyhemoglobin white light absorption, as described previously.12
Assessment of Network Responsivity to Hypoxia
To evaluate the network responsivity to an acute hypoxic challenge, we constructed a localized O2 delivery system. The hindlimb and exposed EDL muscle was positioned in a custom-built frame (Sugru; FormFormForm Ltd.) fashioned to stabilize the EDL muscle for an extended period (≥8 minutes) of imaging, with a 2 cm×2 mm opening. A custom-built gas flow chamber controlled by computer-modulated flow meters was integrated within the microscope stage.13,14 The hindlimb was positioned above the chamber, with the EDL muscle contacting either an optically transparent O2-permeable 100-μm-thick polydimethylsiloxane polymer membrane or a 50-µm-thick fluorosilicone acrylate disk.14 Chamber O2 levels were measured using a fiber-optic Po2 sensor (Ocean Optics) in the gas outlet. Gas consisting of 12% O2, 5% CO2, and 83% N2 at 37°C was passed through the chamber for 5 minutes to generate a hyperoxic surface O2 environment across the entire EDL and standardize the baseline O2 content for both normal and regenerated muscle. This %O2 was selected based on a previously established capillary normoxia tension in the rat EDL of 6.3% O2.15 Real-time video microscopy blood flow recording was initiated, and 30 seconds later the EDL muscle was subjected to the hypoxia challenge by changing the gas content to 2% O2, 5% CO2, and 93% N2, effectively removing O2 from the EDL muscle. RBC transit was continuously recorded before and after the hypoxia challenge for a minimum of 3 minutes, and RBC velocity (VRBC) and RBC supply rate (No of RBCs per second) were quantified.
Mean VRBC and mean RBC supply rates were determined in a total of 255 capillaries from control and regenerated EDL muscles. Hyperoxia RBC measurements were based on 30 seconds of recording prior to induction of hypoxia. Hypoxia RBC measurements were acquired from 120 to 180 seconds of recording after a 20-second transition phase. A stable hyperemic response to hypoxia was considered sustained if the plateau VRBC persisted for at least 120 seconds.
Laser Scanning Confocal Microscopy and 3-Dimensional Reconstruction of Precapillary Arterioles
For thick-section immunostaining, mice were euthanized by isofluorane overdose and perfused sequentially with PBS and 4% paraformaldehyde via left ventricle cannulation at physiological pressure. EDL muscles were dissected, immersed in 4% paraformaldehyde for 2 hours, and cryoprotected with 15% sucrose for 2 hours and 30% sucrose overnight at 4°C. Tissues were then embedded in optimal cutting temperature embedding medium (Tissue-Tek) and stored at −80°C. One hundred-micrometer-thick longitudinal cryosections were permeabilized with 0.5% Triton-X in PBS and double-immunostained using biotinylated rat anti-mouse CD31 (cluster of differentiation 31) antibody (1:50) and mouse anti–smooth muscle (SM) α-actin alkaline phosphatase–conjugated antibody (1:50, Clone 1A4; Sigma A5691). Bound antibodies were visualized using Streptavidin-488 (1:100; Sigma) and Alexa Fluor 546–conjugated goat anti-mouse IgG (1:200; Life Technologies). Nuclei were visualized with TO-PRO-3 iodide (1:500; Life Technologies). Thick sections were transferred to positively charged glass slides, mounted with PermaFluor (Thermo Scientific), and flanked with 100-μm-thick plastic coverslip spacers (Thermo Scientific) prior to coverslipping and sealing.
Arterioles 7 to 20 μm in diameter were imaged with an LSM 510 Meta Confocal Microscope (Zeiss) using a 40× water-immersion objective and Argon2 (488 nm excitation) and HeNe1 (543 nm excitation) lasers, generating ≤50, 0.5-µm-thick z-slices at 2048×2048 image resolution. Z-slices were reconstructed into 3-dimensional projections with Image Viewer (Leica) software. To quantify the spacing between adjacent SM cell (SMC) processes, the SM α-actin (red) signal was isolated from 3-dimensional projections with Photoshop (Adobe), and the maximum distance from one SM α-actin containing process to the next along the arteriolar center axis was ascertained using ImageJ. Vessel coverage by SMCs was determined after measuring the total interprocess gap area and total arteriole segment area ImageJ.
Femoral Artery Excision Induces Widespread Infarction and Capillary Obliteration in the EDL Muscle Followed by Extensive Regeneration
To study the regeneration of microvessels after ischemic injury, C57BL/6 mice were subjected to right femoral artery excision. We screened all muscles in the hindlimb for the extent of infarction and identified that the anterior muscle bundle distal to the knee was the most extensively damaged (data not shown). Within this bundle is the EDL muscle, which we found to be 2 cm long, 2 mm wide, and with well-defined borders, making it well suited to a comprehensive assessment. Furthermore, the EDL was accessibly located for live imaging. Histological assessment revealed that 1 day after femoral artery excision, the entire EDL muscle was infarcted and necrotic. Myofibers were shrunken, had lost their nuclei, and were stained palely with eosin (Figure 1A and 1B). However, 10 days after surgery, the muscle had repopulated with intensely staining myocytes with large central nuclei, indicative of regenerated skeletal myocytes (Figure 1C).16 By 28 days, the regenerated myocytes had matured, with larger diameters and more condensed nuclei (Figure 1D).
To evaluate the microvasculature, sections were immunostained using a panel of distinct endothelial markers. Immunostaining uninjured EDL muscles for CD31, endomucin, von Willebrand factor, and isolectin-B4 revealed capillaries at the interface between the skeletal myofibers (Figure 1E). One day after femoral artery excision, immunoreactivity relative to control muscle fell to 14%, 11%, 23%, and 44% for each marker, respectively (Figure 1F). Remarkably, 3 days after surgery, immunoreactivity for CD31, endomucin, and von Willebrand factor was entirely absent, with some residual isolectin-B4 staining (34%; Figure 1E and 1F). Notably, the latter stains basement membranes in addition to endothelial cells.17 Furthermore, hematoxylin and eosin staining revealed a complete absence of capillaries, with only the occasional fibrous ghost structure without nuclei (Online Figure I). Capillary obliteration was evident down to ≈150 μm below the muscle surface, with only scattered capillaries in the muscle core. Interestingly, neocapillaries emerged thereafter throughout the EDL muscle, in concert with the skeletal myogenesis (Figure 1G). By 28 days, their density exceeded that of control EDL muscle by 1.65-fold (P=0.006; Figure 1G and 1H), although with a capillary-to-muscle fiber ratio that was similar to that of control muscle (1.09±0.06 versus 1.07±0.06; P=0.841; Figure 1H). Collectively, these findings reveal extensive capillary destruction in EDL muscle after ischemic injury, followed by a vigorous microvascular regeneration program that was associated with skeletal muscle regeneration.
Regenerated EDL Muscle Displays Chronic Hypoxia Despite Robust Angiogenesis
We next asked whether the regenerated EDL muscle returned to a well-oxygenated state. To assess for hypoxia, tissues were immunostained for pimonidazole adducts after in vivo perfusion with hypoxyprobe-1. Control EDL muscle showed no hypoxia signal (Figure 2A). However, regenerated EDL muscle (28 days) displayed widespread hypoxia signals in the muscle fiber cytoplasm and nuclei, with myofiber to myofiber variability (Figure 2B and 2C). We also identified increased expression of the hypoxia-dependent, HIF (hypoxia-inducible factor)-1α-regulated genes adrenomedullin (P=0.004), lysyl oxidase (P=0.043), and procollagen lysyl hydroxylase (P=0.008; Figure 2D). Interestingly, expression of VEGF-A (vascular endothelial growth factor-A), which can stabilize sooner than that of adrenomedullin,18 was not greater than that in the control muscle (P=0.272) and neither was that of angiopoietin-1 (P=0.370). However, angiopoietin-2 was substantially upregulated (P=0.029) as was the ratio of angiopoietin-2/angiopoietin-1 (P=0.003). These findings suggest that despite having a capillary–myofiber ratio similar to baseline, the regenerated EDL muscle was subjected to a chronic, nonangiogenic level of hypoxia.
In Vivo Microvascular Imaging Reveals Robust Network Regeneration but With an Aberrant Branching Architecture
To establish the network architecture of the regenerated microvasculature, we performed real-time microscopy after intravenous injection of fluorescein isothiocyanate–labeled high molecular size Dextran. This enabled visualization of flowing microvessels, with RBCs in relief against the fluorescent plasma, within a 50-µm-thick zone along the entire EDL muscle. As well, we generated RBC transit maps, from 15-second video sequences, which revealed the location and geometry of all flowing microvessels, contrasted against stationary and, thus, nonvisible muscle. Transit maps and live video sequences revealed a highly organized microvasculature in uninjured EDL muscle with hierarchical units of feeder arterioles, capillaries, and postcapillary venules (Figure 3A; Online Movie I). Feeder arterioles penetrated orthogonally through the muscle to the surface zone, where they bifurcated into capillaries. Capillaries could also be seen to bifurcate as they coursed parallel to the EDL surface, ultimately converging and draining into venules.
One day after femoral artery excision, blood flow in the EDL microvasculature had ceased, and there was no RBC transit signal (Figure 3A; Online Movie II). However, 10 days after injury, an extensive but relatively chaotic network of flowing neovessels had regenerated (Figure 3A). Quantitative analysis confirmed a hypervascular network, with a 21% greater length density than in normal EDL muscle (P=0.037; Figure 3B) and a 3-fold increase in the number of branches, normalized to total network length (P<0.0001; Figure 3C). Although microvessels of different lumen caliber were present, there was no evidence for arteriole–capillary–venular hierarchy at this time point (Online Movie III). However, the network subsequently underwent rapid structural maturation. Transit maps indicated that by day 14, there had been substantial branch pruning (Figure 3C), and both network length and branch density stabilized by day 14 to 21. Real-time tracking of RBCs also revealed that the network had acquired arteriole–capillary–venule hierarchy by day 14.
Twenty-eight days after ischemic injury, the network resembled that of the uninjured EDL muscle (Figure 3A; Online Movie IV). Notably, however, there was still a 21% greater vessel length density when compared with the normal network (P=0.019) and a residual 1.6-fold increase in bifurcations (P<0.0001). Remarkably, elevated length and branch densities were still present 56 and 120 days after ischemic injury (Figure 3A–3C; Online Movies V and VI).
We also identified several abnormalities in architecture of the feeder arterioles. First, arterioles were seen to course parallel to the muscle fibers, rather than orthogonal to them, running ≤1 mm before bifurcating (Figure 3D). Second, abnormalities in arterial branch caliber were found. In uninjured muscle, arterioles bifurcated into equal-caliber daughter branches, whereas neoarterioles branched into daughter vessels with unequal lumen diameters (Figure 3D; Online Movie VII). Third, neoarterioles were seen not only to bifurcate, but also to trifurcate and quadrificate (Figure 3D), a distinctly abnormal feature of the microcirculation.19,20 Although only bifurcations were found in control muscle, supernumerary branching was identified in 6.5%, 6.7%, and 7.7% of arterioles within the neovasculature on days 28, 56, and 120, respectively.
We next asked how the altered architecture of the regenerated network might impact overall resistance to flow. To assess this, we undertook computational modeling of flow in the network, based on information derived from the videos and the processed RBC transit maps, for control and regenerated (28-day) networks. This predicted a 66% decline in network resistance, indicating a hemodynamic consequence of the altered network architecture (P=0.032; Online Figure II). Thus, despite a rapid microvascular regeneration cascade, these findings reveal multiple architectural abnormalities, even in the late-stage network, that are predicted to impact hemodynamics within the network.
RBCs Transit Slowly Through the Regenerated Microvasculature
We next sought to determine whether RBC transit through the regenerated network differed from that of normal EDL muscle. Within the uninjured EDL, RBCs traversed through capillaries in single file with a velocity (VRBC) of 526.8 (337.0–727.0) μm/s (median [IQR]; Figure 4A). In contrast, within the early regenerated vasculature (10 days), median VRBC was only 36% of that of native capillaries (P<0.0001). RBC transit was chaotic, with frequent direction changes as they passed through the hyperbranched, nonhierarchical network. Furthermore, the smallest caliber neovessels had 2 to 3 RBCs tumbling side-by-side, rather than transiting in single file (Online Movie III). At 28 days, single file RBC transit in neocapillaries was evident, and arteriolar and venular flow was discernable (Online Movie IV). However, median VRBC remained low, at 52% of that of normal capillaries (P<0.0001). Notably, VRBC was even lower on days 56 and 120 (23% of that of normal capillaries; P<0.0001; Figure 4A).
We also identified an abundance of neocapillaries in which blood flow was halted altogether. This was evident either as a static column of plasma devoid of RBCs or as channels with stationary RBCs (Figure 4B; Online Movie VIII). On average, 2.0±1.6 capillaries/mm2 in the entire control EDL microvasculature displayed stopped flow for a duration of at least 15 seconds, and this increased by over 9-, 19-, and 19-fold in the regenerated EDL on days 28, 56, and 120, respectively (Figure 4C).
Regenerated Microvasculature Contains Arteriolar–Venular Malformations
The movement of oxygenated RBCs from arterioles to capillaries is fundamental to efficient oxygen delivery. However, we found that despite the return of single file capillary transit, not all arterioles diverged into a capillary network. Instead, we observed neoarterioles from which one branch diverged into a capillary mesh, while the other branch flowed directly into a venule (Figure 5A and 5B; Online Movies IX–XI). RBC transit within the direct arteriovenous connection was not single file, and flux was demonstrably greater than in the adjacent capillary system, revealing conditions for diverting oxygenated RBCs directly into the venule. Quantitative analysis revealed that 37%, 32%, and 33% of all neoarterioles on days 28, 56, and 120, respectively, branched into an arteriovenous connection, whereas arteriovenous shunts were rarely observed (2%) in control muscle (P<0.0001; Figure 5C).
RBC Velocity Dispersion Is Severely Blunted in the Regenerated Microvasculature With an Abnormal Oxygen Delivery Profile
We next assessed for heterogeneity in RBC transit among capillaries within a network.21,22 This heterogeneity is fundamental to the skeletal muscle circulation and critical for precisely matching the metabolic demands of each myofiber in the muscle with RBC delivery.14,22 To assess network heterogeneity, we compared the frequency distributions of mean VRBC among capillaries over the entire EDL zone. The intercapillary VRBC histogram derived from the 28-day regenerated network was significantly different than that of the native network, with both a leftward shift and narrower range (K-S stat =0.54; P=1.12×10-13; Figure 6A). In addition, the normal microcirculation was found to contain a hyperemic subpopulation of capillaries with a particularly high VRBC (>1000 μm/s). In contrast, there was no hyperemic subpopulation of capillaries in the regenerated network on day 28. These abnormalities did not normalize over time and in fact were worse on days 56 and 120. Thus, the regenerated microcirculation displayed profoundly reduced dispersion in VRBC, suggesting an inability of the network to tune the delivery of O2 to meet the needs of individual myofibers.
To determine whether the homogeneity in VRBC was associated with abnormal O2 delivery, we directly measured hemoglobin O2 saturation in RBCs transiting through the capillaries. Using dual wavelength absorption microscopy to distinguish oxy- and deoxyhemoglobin, we found that median RBC O2 saturation in regenerated capillaries (120 days) was 75% of that in normal capillaries (P=0.0003). This reduction was evident not only for RBCs at the downstream end of the capillary (69%; P=0.002) but also at the upstream end (77%; P=0.005; Online Figure III). Furthermore, there was a strikingly wide spread in RBC O2 saturation in the regenerated capillaries, with O2 saturations as low as 2% and as high as 87%. These findings indicate the variable presence of regions with profoundly unmet O2 demands, further supporting a failure of local tuning of RBC delivery.
Regenerated Microvasculature Has Impaired Flow Responsiveness to Hypoxia
To directly determine whether the regenerated microcirculation could regulate blood flow, we generated a methodology to locally control the O2 content in the mouse EDL muscle. A defined O2-containing gas mixture was delivered to the EDL muscle through a gas-permeable membrane. After generating a stable Po2 of 12% at the muscle site, we abruptly decreased the ambient oxygen content to 2%, establishing conditions for microremoval of muscle O2. The ability of the microvasculature to respond to this hypoxic challenge was then evaluated by tracking both RBC velocity and RBC supply rate through individual capillaries.
Figure 6B depicts RBC velocity response profiles to local hypoxia, illustrating a brisk and sustained VRBC increase in a native capillary but a blunted and transient response in a regenerated capillary. Although there was a range of responses among capillaries (Figure 6C and 6D), overall, there was a significant reduction in both VRBC and RBC supply rate responsivity within the regenerated microvascular network. Among 163 capillaries in normal EDL muscle, hypoxia induced a 1.53-fold increase in median VRBC and a 1.7-fold increase in median RBC supply rate (Figure 6C and 6D; Online Movie XII). Relative to these responses, the VRBC and RBC supply rates were found to be significantly lower on each of days 28, 56, and 120 after ischemic injury (Figure 6C and 6D). These findings identify that the regenerated microvasculature is impaired in its ability to effectively augment RBC delivery when faced with local hypoxia.
Smooth Muscle Cell Wrapping Around Regenerated Arterioles Is Aberrant
The combination of impaired responsivity to hypoxia and the homogeneity of RBC velocities throughout the network implicated failure of a vasomotor tuning mechanism. We considered that a unifying basis for this failure might reside in the cellular architecture of the feeder arterioles. To elucidate the microstructure of normal and regenerated arterioles, we generated 3-dimensional confocal reconstructions from 100-µm-thick sections of EDL muscle double-immunolabeled for CD31 and SM α-actin. In normal EDL, arterioles were found to be intimately invested by SMCs. The precise architecture varied with arteriolar caliber and location within the tree. SMCs on arterioles of diameter 12 to 20 μm enveloped the artery with circumferential processes containing SM α-actin, which we could resolve as individual microfilament bundles (Figure 7A and 7B). SMCs on arterioles of 7 to 12 μm diameter also circumferentially wrapped the vessel with SM α-actin cytoplasmic protrusions, but these processes were somewhat thinner and more variably oriented (Figure 7A and 7C). SMC investment and protrusion-based wrapping terminated once the vessel diameter fell to 5 to 8 μm, consistent with transition to a capillary (Figure 7A).
SMCs were also found to layer around arterioles of the regenerated network and send out processes. However, in contrast to native arterioles, the coverage was discontinuous and disordered (Figure 7A–C). In some neoarterioles, there were zones up to 13 μm in length that were largely devoid of SM α-actin bundles. Moreover, even in regions where SMC processes wrapped the vessel, the spacing between adjacent processes was increased compared with that of normal arterioles of similar caliber (Figure 7B and 7C arrows). Quantitative analysis revealed that the median space from one SM α-actin containing process to the next in normal terminal arterioles was 0.74 μm (Figure 7D). These gaps constituted 7% of the arteriolar surface area (Figure 7E). In contrast, 28 days after injury, the median interprocess space was 1.40 μm (P<0.0001), which corresponded to 22% of the neoarteriolar surface area (P<0.0001; Figure 7D and 7E). Increased process spacing was also observed at days 56 and 120, although this improved somewhat. However, prominent gaps in SMC process coverage were still evident, with widely variable process orientations, rendering a haphazard and disordered coverage profile (Figure 7C arrows). These findings strongly implicate arteriolar wall microdisorder as a basis for impaired control over RBC delivery through the regenerated skeletal muscle.
Through high-resolution mapping of RBC transit, dynamic tracking of RBC velocities, and optical reconstruction of terminal arterioles, we have discovered that the reconstructed microvasculature in EDL muscle is a highly flawed network. The array of abnormalities we uncovered in this regenerated tissue included (1) aberrant arteriolar branching; (2) arteriolar–venular shunting; (3) slow RBC transit through capillaries; (4) a monotony of RBC transit velocities throughout the capillary system; and (5) an impaired vasomotor control system. These abnormalities were still evident 4 months after injury, indicating a failure to normalize. The functional importance of these abnormalities was supported by the presence of hypoxia signals and profoundly abnormal capillary O2 saturations in the EDL muscle, despite a capillary-to-skeletal myocyte ratio that was similar to that of normal EDL muscle. Thus, notwithstanding an extensive regenerative response, the microvascular neonetwork was ill-suited for effective oxygen delivery.
This is the first study to evaluate regenerative angiogenesis in skeletal muscle at a network level and at the site where gas and nutrient exchange occur. The mouse EDL subjected to femoral artery excision proved to be a valuable model for this because the entire EDL muscle was infarcted and the native microcirculation was destroyed, as evidenced by staining with hematoxylin and eosin and a panel of endothelial markers. All subsequently identified microvessels could, thus, reliably be attributed to a regeneration program. This contrasts with other sites in the mouse hindlimb, where infarction is patchy and a mix of preexisting and new microvessels would coexist.1,7 High-resolution mapping of RBC transit in the regenerating EDL revealed that the vascularization response was in many respects impressive. It entailed an early hypervascular phase that covered the entire EDL muscle surface zone. Branching was extensive at this early stage, and the network was not yet hierarchical. However, rapid branch pruning and realignment of microvessels quickly ensued, yielding a network that coursed primarily parallel to long axis of the maturing skeletal myofibers. Moreover, within 14 days of the ischemic insult, these vessels had differentiated into arteriole–capillary–venular circulation units, further indicating the rapid angiogenesis and network remodeling program.
Nevertheless, discordance between the microvascular restoration process and functionality of the RBC delivery system was remarkable. Whether the identified array of microcirculatory defects exists in all models of ischemia, as well as different mouse strains and forms of injury, is unknown. However, femoral artery excision in C57BL/6 mice constitutes a substantial insult and is also a setting where the return of bulk flow is robust, reaching 85% to 100% that of the contralateral limb.1,6,7 The current findings, therefore, highlight the hidden pathophysiology that can exist in the microcirculation. Some differences in capillary network architecture might be expected, given the different size and arrangement of regenerated myofibers. However, the abnormal RBC transit, impaired vasoreactive control, and violations of arteriole–capillary–venule geometry and hierarchy constitute previously unrecognized pathologies in this regenerative setting. Moreover, these defects would not be detectable with currently used techniques for assessing blood flow, such as laser-Doppler, Doppler-ultrasound, and magnetic resonance imaging strategies.
Our findings of asymmetrical arteriolar bifurcations, trifurcations, and quadrifications in the regenerated microvasculature identify a distinctly abnormal arteriolar morphogenesis program. Bifurcation of arterioles into symmetrical caliber branches is fundamental to the microcirculation and ensures ordered dispersion of red cells through the capillary network.23–26 Although the molecular underpinnings of this arteriolar dysgenesis response are unknown, it is noteworthy that both asymmetrical arteriolar branching and supernumerary branching are features of the highly disordered microvasculature of tumors.19,20 The similarities between the tumor and regenerated muscle milieu would only be partial, but it is noteworthy that aberrant arteriole branching in tumors is considered to be one of the causes of tumor hypoxia.11,19,20,27 The potential for arteriolar asymmetry to similarly contribute to hypoxia in skeletal muscle is also supported by mathematical modeling, which has indicated skimming of plasma, disparate local hematocrits, and regional heterogeneity in oxygen tensions unrelated to tissue needs.20,25,28
Our discovery of arteriole–venule shunts in the regenerated microcirculation was striking and reveals another basis for impaired oxygen delivery to the muscle. These aberrant conduits were rarely found in the uninjured muscles, yet constituted the output pattern of 37% of terminal arterioles. Their presence in muscle can be expected to compromise gas exchange by diverting red cells away from nearby single file transit capillaries into the high-flow microshunts. Interestingly, these microshunts were scattered throughout the vascular tree, which differs from the focal arteriovenous malformations seen in conditions, such as hereditary hemorrhagic telangiectasia.29,30 The shunts emerged at the same time as capillaries were forming elsewhere in the network, suggesting a primary, albeit localized, failure in capillary differentiation. This differs from recent studies in the brain where arterial–venous malformations arise via pathological transformation of preexisting capillaries.31
The velocity of RBCs within capillaries provides important clues about oxygen delivery in tissues. The sluggish transit of RBCs through regenerated capillaries was particularly noteworthy given that the metabolic demands of regenerating skeletal myofibers would be expected to increase, not decline.32 Our transit maps and video analyses point to several causes of the slow capillary flow, including a greater total path length of the regenerated microvascular network and the steal phenomenon from the abundant arteriole–venule shunts.33,34 The persistent increase in the prevalence of capillaries with entirely halted RBC transit also suggests luminal plugging of microvessels or competing pressure heads within the network. The absence of capillaries in which RBC velocities rose >1000 μm/s is also noteworthy because RBCs transiting through hyperdynamic capillaries have the potential to deliver oxygen not only to the adjacent tissue but also to RBCs in nearby capillaries, effectively reloading them as they approach the distal capillary segment.14,35 Collectively, these abnormalities in RBC transit velocities strongly suggest that slow RBC transit is a pathological entity of the regenerated microvascular network.
The normal microvasculature does not act as an inert set of microtubes but, instead, as an exquisitely controlled vasomotor system that distributes RBCs in an orchestrated manner to meet the needs of the tissue.5 Hypoxia is a powerful stimulus for enhancing delivery of RBCs in skeletal muscle.36,37 The oxygen exchange strategy we used, coupled with direct tracking of RBC transit in individual capillaries, provided a rapid means of generating a hypoxic muscle milieu and directly linking this to network responsivity. RBC velocity and supply rates in control EDL muscle rapidly increased in response to hypoxic challenge. However, the responses in the regenerated vasculature were blunted. Thus, the regenerated microvasculature was defective in a fundamental control process for modulating RBC delivery. This muted responsivity could also explain the striking monotony in RBC velocities we observed throughout the network. Although it is possible that the demands of the regenerated EDL were more uniform than in the normal EDL, this seems unlikely given the striking variations in capillary RBC O2 saturation that we identified. Together, the findings indicate that the regenerated network does not have the ability to tune the delivery RBCs to match the regional needs.
What might be the mechanism for this relative inertness in the regenerated microcirculation? Distal arterioles constitute the critical effector limb for RBC delivery and the site where hormonal, metabolic, or neural regulatory factors impinge. Three-dimensional reconstruction of confocal optical images revealed how SMCs and their contractile elements are organized around these arterioles. Our finding of circumferential wrapping of SMC processes is consistent with studies using scanning electron microscopy38 and has recently been delineated in brain arterioles using genetically encoded mural cells.39 Remarkably, in addition to the banded pattern of the SMC processes, we were also able to resolve individual actin microfilament bundles, which revealed there to be 3 to 5 discrete actin microfilament bundles per process, in 10 to 20 μm diameter arterioles. The finding that SMC processes also wrapped arterioles in the regenerated network is consistent with our intravital microscopy findings of arteriole–capillary–venule units. However, the increased spacing between the SM α-actin–containing processes identifies an architectural defect in the contractile machinery. In fact, in the regenerated vessels, we found that there was more gap between the SM α-actin–containing processes than area covered by the processes. This defect was even more striking in arteriolar segments immediately upstream of the capillary, where the actin processes were neither circumferential nor spiral, but oriented haphazardly. Given that arterioles of the caliber studied (7–20 μm) were invested by only a single layer of SMC processes, the architectural details of these processes stand to be vital determinants of the extent to which RBC entry into the capillaries is controlled. We do not currently know if additional defects in the vasomotor control loop exist, such as in oxygen sensing by the endothelium, gap-junction–mediated signal conduction, or communication between the endothelium and SMCs. Nonetheless, the findings implicate a new reason for microvascular dysfunction, based on disorder of the otherwise exquisitely patterned actin-containing protrusions.
In summary, using 4-dimensional microvascular imaging and an obliteration–restoration model of angiogenesis, we have uncovered profound and persistent network-level dysfunction in the regenerated microcirculation in mouse skeletal muscle. These findings suggest that in the setting of vascular regeneration after skeletal muscle infarction, impaired network functionality rather than limited angiogenesis may be a key determinant of ongoing ischemia. Strategies to impart advanced physiology to the postinfarction microcirculation, including recapitulating normal arteriolar morphogenesis, are warranted.
We thank Jason Baek for his contributions to the network modeling used in the flow simulations.
Sources of Funding
This work is supported by the Canadian Institutes of Health Research (FRN-11715, FRN-126148, and FDN-143326), the Heart and Stroke Foundation of Canada (T7081), and the University of Western Ontario (POEM). H. Yin was supported by a CIHR Fellowship. J.G. Pickering holds the Heart and Stroke Foundation of Ontario/Barnett-Ivey Chair.
The online-only Data Supplement is available with this article at http://circres.ahajournals.org/lookup/suppl/doi:10.1161/CIRCRESAHA.116.310535/-/DC1.
- Nonstandard Abbreviations and Acronyms
- extensor digitorum longus
- smooth muscle
- smooth muscle cell
- red blood cell velocity
- Received December 23, 2016.
- Revision received January 20, 2017.
- Accepted February 7, 2017.
- © 2017 American Heart Association, Inc.
- Helisch A,
- Wagner S,
- Khan N,
- Drinane M,
- Wolfram S,
- Heil M,
- Ziegelhoeffer T,
- Brandt U,
- Pearlman JD,
- Swartz HM,
- Schaper W
- Duvall CL,
- Taylor WR,
- Weiss D,
- Guldberg RE
- Frontini MJ,
- Nong Z,
- Gros R,
- Drangova M,
- O’Neil C,
- Rahman MN,
- Akawi O,
- Yin H,
- Ellis CG,
- Pickering JG
- Yin H,
- Frontini MJ,
- Arpino JM,
- Nong Z,
- O’Neil C,
- Xu Y,
- Balint B,
- Ward AD,
- Chakrabarti S,
- Ellis CG,
- Gros R,
- Pickering JG
- Ellis CG,
- Goldman D,
- Hanson M,
- Stephenson AH,
- Milkovich S,
- Benlamri A,
- Ellsworth ML,
- Sprague RS
- Paoni NF,
- Peale F,
- Wang F,
- Errett-Baroncini C,
- Steinmetz H,
- Toy K,
- Bai W,
- Williams PM,
- Bunting S,
- Gerritsen ME,
- Powell-Braxton L
- Less JR,
- Skalak TC,
- Sevick EM,
- Jain RK
- Duling BR,
- Damon DH
- Ellis CG,
- Wrigley SM,
- Groom AC
- Koller A,
- Dawant B,
- Liu A,
- Popel AS,
- Johnson PC
- Baish JW,
- Jain RK
- Pries AR,
- Secomb TW,
- Gaehtgens P
- Murphy PA,
- Kim TN,
- Huang L,
- Nielsen CM,
- Lawton MT,
- Adams RH,
- Schaffer CB,
- Wang RA
- Koopman R,
- Ly CH,
- Ryall JG
- Ellis CG,
- Bateman RM,
- Sharpe MD,
- Sibbald WJ,
- Gill R
- Koning NJ,
- Simon LE,
- Asfar P,
- Baufreton C,
- Boer C
- Parthasarathi K,
- Lipowsky HH
Novelty and Significance
What Is Known?
Subjecting mouse hindlimb skeletal muscle to ischemic injury leads to regenerative angiogenesis and near-complete restoration of blood flow.
Strategies to augment skeletal muscle angiogenesis in individuals with peripheral vascular disease have met with little to no clinical success.
Current approaches to evaluating regenerative angiogenesis in muscle lack the combined spatial and temporal resolution required for delineating microcirculatory architecture and function.
What New Information Does This Article Contribute?
The structure and function of the neomicrovasculature that forms after ischemic injury can be quantified, at a network level, by real-time microvascular imaging in the mouse.
Despite robust angiogenesis, the regenerated microvasculature is profoundly flawed.
The microcirculation that forms in muscle after ischemic injury fails to adequately control the delivery and distribution of red blood cells and is ineffective at delivering oxygen in accordance with local needs.
Regenerative angiogenesis is critical for restoring perfusion after ischemic muscle injury. However, strategies that augment angiogenesis have not benefited patients with peripheral vascular disease. Importantly, it is not known whether a regenerated microcirculation, even when extensive, can meet the metabolic needs of muscle. We tested this by developing a real-time imaging strategy with spatial and temporal resolution suitable for probing angiogenesis in the ischemic mouse hindlimb. We found that the formation of a new microvascular network was rapid and robust, but the end result was flawed. Branching architecture was aberrant, red blood cell transit was slow and monotonous, and there were arteriovenous shunts. Moreover, the network did not appropriately respond to hypoxia, and smooth muscle cell processes haphazardly wrapped the terminal arterioles. These findings provide the first network-level indication that the regenerated microcirculation in ischemic muscle does not have the attributes required for effective, local delivery of oxygen. They further imply that (1) capillary density and bulk flow are inadequate indicators of functional angiogenesis after ischemic injury; and (2) therapeutic strategies beyond stimulating angiogenesis in ischemic muscle are required, including normalizing microcirculatory physiology.