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).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Histology of the extensor digitorum longus (EDL) muscle before and after femoral artery excision. A–D, Hematoxylin and eosin stained sections of EDL muscle before surgery (A), 1 day after surgery showing pale, shrunken myocytes without nuclei (B), after 10 days showing repopulated myocytes with central nuclei (arrows; C), and after 28 days (D). E, Fluorescence micrographs of EDL muscle before and 3 days after surgery immunostained for CD31 (cluster of differentiation 31), endomucin, von Willebrand factor (vWF), and isolectin-B4 showing capillaries (arrow) surrounding myofibers (asterisk) in control EDL and loss of capillaries after ischemic insult. Nuclei are visualized with DAPI (4′,6-diamidino-2-phenylindole). F, Quantitative data for capillary density as identified by each marker. (n=5 native; n=3 each for days 1 and 3, *P<0.0001, †P=0.001, ‡P=0.0002, §P=0.0013, ‖P=0.0004 vs native). G, Photomicrographs of EDL muscle immunostained for CD31 showing capillaries in uninjured EDL muscle (left, arrows), their disappearance 1 day after surgery (middle), and regenerated capillaries 28 days after surgery (right, arrows). H, Graphs showing density of capillaries and capillary-to-muscle fiber ratios in native (n=3) and regenerated (28 days, n=3) EDL muscle (*P=0.006).

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.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Evidence for hypoxia in the regenerated extensor digitorum longus (EDL) muscle. A and B, Photomicrographs of sections of native (A) and regenerated (day 28; B) EDL muscle harvested after infusion of Hypoxyprobe-1 and immunostained for pimonidazole. C, Graph depicting quantified pimonidazole signal in regenerated EDL muscle not infused with Hypoxyprobe-1 (technical control, n=3), in native (contralateral) muscle (n=4), and regenerated muscle (n=4, *P=0.007). D, Graph depicting relative mRNA abundance of hypoxia-dependent genes in uninjured (n=4) and regenerated (n=4) EDL muscles (*P=0.004, 0.043, 0.008, 0.029, and 0.003 vs respective control). ANGPT1 indicates angiopoietin-1; ANGPT2, angiopoietin-2; LOXL2, lysyl oxidase; PLOD2, procollagen lysyl hydroxylase; and VEGF-A, vascular endothelial growth factor-A.

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.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Microvascular network architecture before and after ischemic injury. A, Red blood cell (RBC) transit maps within native extensor digitorum longus (EDL) muscle microvasculature and 1, 10, 28, 56, and 120 days after femoral artery excision, showing no RBC transit 1 day after ischemic insult and subsequent regeneration of a flowing EDL microvasculature. Hatched lines denote EDL muscle borders. Corresponding real-time movies for each panel are available in the Online Data Supplement. B and C, Graphs showing EDL muscle microvascular network length density (B) and branch density (C; n=3–5 for each time point). B: *P=0.037, 0.041, 0.019, <0.0001, and <0.0001, respectively, vs control; C: *P<0.0001 vs control, †P<0.0001 vs day 10. D, Transit maps of branching arterioles of native and regenerated EDL muscle showing symmetrical bifurcation, asymmetrical bifurcation, trifurcation, and quadrification. Arrows depict the direction of parent vessel flow. The curved arrow denotes orthogonal inflow from below the plane.

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 (V_RBC) of 526.8 (337.0–727.0) μm/s (median [IQR]; Figure 4A). In contrast, within the early regenerated vasculature (10 days), median V_RBC 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 V_RBC remained low, at 52% of that of normal capillaries (P<0.0001). Notably, V_RBC was even lower on days 56 and 120 (23% of that of normal capillaries; P<0.0001; Figure 4A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Red blood cell (RBC) transit in capillaries within the extensor digitorum longus (EDL) microvasculature. A, Box and whisker (10–90 percentile) plot of red blood cell velocities (V_RBC) in individual capillaries within native and injured/regenerating EDL muscle evaluated at the designated times. Values were obtained from a total of 860 capillaries. Each value is derived from RBC velocities, averaged over a 15-second imaging time frame, within each capillary. B, Sequential movie still frames depicting dynamic RBC transit in control muscle (boxes, top row) and stalled perfusion in 28-day regenerated muscle (boxes, middle and bottom row). RBCs appear in relief against the fluorescent plasma. Arrows depict stagnant RBCs. Corresponding real-time movies are found in the Online Movie VII. C, Graph depicting density of capillaries with flow halted for at least 15 seconds. n=4 native and n=3 for each of day 28, 56, and 120 EDL muscles.

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).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Red blood cell (RBC) transit maps of arteriole–capillary–venule units. A, Native extensor digitorum longus (EDL) muscle microvasculature with a parent arteriole that bifurcates into daughter vessels, both of which diverge into capillary meshes that drain into a venule system. A simplified, not-to-scale, schematic of the unit is shown on the right. B, Regenerated EDL muscle (day 28) with a parent arteriole that trifurcates into daughter vessels of unequal caliber. Daughter vessel 1 diverges into a capillary mesh that drains into a venule, whereas daughter vessel 2 drains directly into a venule, also depicted in the adjacent schematic. Corresponding real-time movies are found in the Online Movies III and IV and reveal high RBC flux in the arteriovenous (AV) connection. C, Graph depicting the percentage of arterioles that branch into an AV connection. The entire network from 5 native EDL muscles and from 3, 4, and 4 regenerated muscles on days 28, 56, and 120, respectively, was evaluated as indicated below the graph (*P<0.0001 vs native).

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 V_RBC among capillaries over the entire EDL zone. The intercapillary V_RBC 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 V_RBC (>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 V_RBC, suggesting an inability of the network to tune the delivery of O₂ to meet the needs of individual myofibers.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Red blood cell (RBC) velocity dispersion within the microvasculature and network responsivity to hypoxia. A, Frequency histograms of the average capillary RBC velocity (V_RBC) within native and regenerated extensor digitorum longus (EDL) muscles 28, 56, and 120 days after injury. The distribution is significantly narrowed on day 28 (K-S stat=0.54; P=1.12×10⁻¹³) and remains narrow and left-shifted. B, RBC velocity response profiles to local hypoxia. V_RBC were acquired every second within a single capillary for each condition before and after hypoxia challenge. C and D, Graphs depicting the V_RBC and RBC supply rate responses to local hypoxia. Left, The fold-change response in individual capillaries is shown for day 28. The line depicts the median. Right, The relative median responses (and 45–55 percentile) at the time points indicated are shown. Data are from 164, 65, 27, and 60 capillaries from control (n=6), day 28 (n=3), day 56 (n=3), and day 120 (n=3) EDL muscles, respectively. For V_RBC: *P=0.004, †P=0.001, ‡P=0.021 vs control. For RBC supply rate: *P=0.004, †P=0.0004, ‡P=0.005 vs control.

To determine whether the homogeneity in V_RBC 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 V_RBC 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 V_RBC 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 V_RBC and a 1.7-fold increase in median RBC supply rate (Figure 6C and 6D; Online Movie XII). Relative to these responses, the V_RBC 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).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Aberrant wrapping of smooth muscle cells (SMCs) around arterioles in the regenerated microvasculature. A, Confocal micrographs of distal arteriolar trees from normal and regenerated extensor digitorum longus (EDL) muscle, immunostained for CD31 (cluster of differentiation 31; green) and smooth muscle (SM) α-actin (red). Images are projections of 30 optical sections. Arrows indicate sites of discontinuous SMC coverage. B, Confocal micrographs of projected optical sections of arteriolar segments ≈14 to 15 μm in diameter of control and regenerated (day 28) muscle, immunolabeled for SM α-actin (red). C, Projected optical sections of arteriolar segments ≈9 μm in diameter of control and regenerated (day 56 and 120) muscle, immunolabeled for SM α-actin (red) and TO-PRO-3 (blue). Arrows indicate sites of discontinuous SMC coverage, and arrowhead indicates site of aberrant process orientation. D, Graph depicting the SMC interprocess gap widths. Data are from 25 arterioles from native EDL muscles and 23, 19, and 22 neoarterioles from EDL muscles 28, 56, and 120 days, respectively, after ischemic injury. *P<0.0001 vs native. E, Graph depicting surface area coverage of feeder arterioles by SM α-actin-containing processes. *P<0.0001 vs native.

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.