Decreased inspired oxygen stimulates de novo formation of coronary collaterals in adult heart

Abstract

Rationale-

Collateral vessels lessen myocardial ischemia when acute or chronic coronary obstruction occurs. It has long been assumed that although native (pre-existing) collaterals enlarge in obstructive disease, new collaterals do not form in the adult. However, the latter was recently shown to occur after coronary artery ligation. Understanding the signals that drive this process is challenged by the difficulty in studying collateral vessels directly and the complex milieu of signaling pathways, including cell death, induced by ligation. Herein we show that hypoxemia alone is capable of inducing collateral vessels to form and that the novel gene Rabep2 is required.

Objective-

Hypoxia stimulates angiogenesis during embryonic development and in pathological states. We hypothesized that hypoxia also stimulates collateral formation in adult heart by a process that involves RABEP2, a recently identified protein required for formation of collateral vessels during development.

Methods and Results-

Exposure of mice to reduced FiO2 induced collateral formation that resulted in smaller infarctions following LAD ligation and that reversed on return to normoxia. Deletion of Rabep2 or knockdown of Vegfa inhibited formation. Hypoxia upregulated Rabep2, Vegfa and Vegfr2 in heart and brain microvascular endothelial cells (HBMVECs). Knockdown of Rabep2 impaired migration of HBMVECs. In contrast to systemic hypoxia, deletion of Rabep2 did not affect collateral formation induced by ischemic injury caused by LAD ligation.

Conclusions-

Hypoxia induced formation of coronary collaterals by a process that required VEGFA and RABEP2, proteins also required for collateral formation during development. Knockdown of Rabep2 impaired cell migration, providing one potential mechanism for RABEP2’s role in collateral formation. This appears specific to hypoxia, since formation after acute ischemic injury was unaffected in Rabep2^−/− mice. These findings provide a novel model for studying coronary collateral formation, and demonstrate that hypoxia alone can induce new collaterals to form in adult heart.

1. INTRODUCTION

Collaterals are arteriole-to-arteriole anastomoses present in most tissues within the watershed region between adjacent arterial trees where they cross-connect a small fraction of the outermost branches. Under normal conditions, blood flow in collaterals slowly ebbs to-and-fro and averages at or near net-zero, owing to the absence of a significant pressure drop across them [1,2]. However, when the trunk or a branch of one of the trees becomes narrowed or occluded, significant collateral flow is recruited to the territory downstream of the obstruction (ie, the area at risk), allowing collaterals to serve as critical bypass vessels. The amount of protection against ischemia that they provide is proportional to their number and diameter present (ie, collateral extent) [3–5]. Previous studies have shown that the extent of native collaterals in tissues of healthy adult mice varies widely due to genetic-dependent differences in their formation, a process termed collaterogenesis that occurs late during gestation [4–7]. Collateral extent is also affected by aging and presence of risk factors such as hypertension and diabetes which cause collateral rarefaction, ie, a decline or pruning away of their number and a smaller diameter in those that remain [8]. Although technical limitations have precluded precise measurement of collateral extent in humans, collateral conductance varies widely among individuals in heart, brain and lower extremities, presumably due to genetic differences and the above-mentioned “environmental” factors, with only a fraction of patients having a well-developed collateral circulation in these tissues [9–12]. Therefore, there has been growing interest in understanding the vascular biology of collaterals.

It has long been assumed that although arterial obstruction in coronary artery disease (CAD), acute myocardial infarction (AMI) and other ischemic conditions induces collaterals to undergo shear-stress induced anatomic outward remodeling, these conditions do not induce additional new collateral vessels to form [9,13]. However, addressing this question has not been possible due to absence of a method to distinguish newly formed collaterals from native ones that have undergone remodeling. Recently, arterial obstruction has been shown to induce new collaterals to form in mouse models of stroke and peripheral artery disease [3,14,15]. This was also observed in a mouse model of AMI, a finding aided by the absence of native collaterals in heart (but not other tissues) in this species [4]. Identifying the pathway that drives this process is made difficult by the complex milieu of signaling factors induced by the hypoxia, ischemia, inflammation, cell injury, and cell death that accompany coronary artery occlusion. A possible way forward was recently uncovered by the finding that additional collaterals form in the brain of mice when subjected to sustained reduction in atmospheric oxygen [16], demonstrating that systemic hypoxia, alone in the absence of arterial obstruction, is capable of stimulating de novo formation of collaterals in this tissue.

The aim of this study was to determine if hypoxia also induces collaterals to form in the heart, characterize the process in detail, and begin to examine underlying mechanisms. There is indirect evidence that supports this hypothesis. Individuals with sleep apnea and chronic obstructive pulmonary disease have exceptionally abundant angiographically identified coronary collaterals [17,18], as do humans residing at high altitude [19] and species indigenous to high altitude environments [20]. Increased coronary collateral conductance was also observed in dogs made chronically anemic and in piglets exposed to several weeks of reduced inspired oxygen [21,22]. The finding in the present study that hypoxia induces formation of coronary collaterals in adult mice provides a model for future studies aimed at determining how collaterals form, whether a marker can be identified that distinguishes newly formed from native collaterals, and whether a means to augment this process can be developed. In addition, the fact that hypoxia has now been shown to increase collateral abundance in both heart and brain—tissues with high oxygen requirements—provides additional evidence for the concept [16,20] that collaterals not only provide an alternate source of perfusion in occlusive disease but also serve a physiological purpose to optimize oxygen delivery to meet oxygen demand, particularly when oxygen availability is limiting.

2. METHODS

2.1. Animals

Data presented in the figures and supplemental figures were obtained from 185 mice. The Institutional Animal Care and Use Committee approved all studies. Three-to-four months-old C57BL/6J (B6; Jackson Labs, Bar Harbor, ME) and the following mutant mice were used: B6.Rabep2^em1Jef/J (JAX #29463) mice were described previously [5]; heterozygote B6.Cg-Tg(CAG-cre/Esr1*)5Amc/J (JAX #4682) (“CAG-CreERT”) [23] were crossed with B6;129Sv-Vegfa^tm2Gne/J mice homozygous for the floxed Vegfa allele (“Vegfa^Fl/Fl) [24] to generate Vegfa^Fl/Fl;CAG-CreERT^+/− and Vegfa^Fl/Fl;CAG-CreERT^−/− animals. Efficacy of knockdown was assessed. Controls were age- and sex-matched wildtype (WT) littermates. Although approximately equal number of both sexes were studied, the experiments were not powered with n-sizes necessary to test for sexual dimorphism. A previous study found no sex-dependent differences in native collateral number or diameter [25].

2.2. Hypoxia

An oxygen controller (Biospherix, Parish, NY) maintained normobaric hypoxia in a custom-made plexiglass chamber using automated nitrogen injection. “Normoxia” (control) mice were maintained in an identical chamber. A fan provided mixing of oxygen, nitrogen, CO₂ and water vapor. Mice were maintained 4 per cage. Drierite and soda lime controlled moisture and CO₂ respectively and were changed weekly. CO₂ was measured using a Cozir Wide-Range sensor and GasLab software, (#CM-0123 CO2meter.com, Ormond Beach, Fl). Mice were acclimated to 7% fractional inspired oxygen content (FiO₂) over a 14-day period followed by a 14-day exposure to 7% FiO₂ (“hypoxia” mice). In selected experiments, mice were instead exposed to 8.5% or 10% FiO₂. Retro-orbital blood was collected for determination of hematocrit. In some experiments, exposure to hypoxia was followed by a 7-day recovery period during which FiO₂ was gradually returned to 21% (“recovery” mice). Tamoxifen (Sigma-Aldrich T5648) was dissolved in 100% ethanol, diluted to 80 mg/ml in sterile filtered corn oil and administered 2 days before exposure to hypoxia and once per week thereafter until conclusion of the experiment.

2.3. Collateral vascular casting and planimetry

Methods were described previously [4]. Briefly, mice were heparinized (1000 units, ip) and 5 min later deeply sedated with ketamine and xylazine (100 and 10mg/kg, ip, respectively) followed by thoracotomy. The distal thoracic aorta was cannulated (pulled-out PE-50) and PBS with 0.1 mM nitroprusside was perfused at 100 cm H₂O. The left anterior descending artery (LAD) was ligated (7–0 polyprolene) immediately below the left auricle (approximately 3–4 mm from the origin of the LAD). Sixty percent yellow Microfil (Flowtech Inc, Carver, MA) was injected retrograde via the aorta. Care was taken to avoid introduction of bubbles. Injection was stopped after Microfil first appeared in the coronary venous circulation. Microfil was allowed to cure for at least 30 min, and the heart was fixed in 2% PFA, dehydrated through an ethanol gradient, cleared with methyl salicylate and imaged at 25X with a stereomicroscope. ImageJ was used to measure the area-at-risk (AAR) from anterior and posterior views of each heart, clearly demarcated by the area lacking Microfil. Thresholding was applied with identical parameters for each image to mask the Microfil cast of the LAD distal to the ligation within the AAR. The total sum of the pixel area of the LAD distal to the ligation was expressed as a percentage of the AAR. Measurements were normalized to vascular density as defined by the vascular area within a 1 mm² region of the left ventricle remote zone imaged at 50X.

2.4. Retrograde fill-time of the ligated LAD

Methods for determining retrograde fill-time (RFT), which is inversely proportional to collateral conductance, were described previously [4]. Mice were heparinized, anesthetized, euthanized, distal thoracic aorta cannulated, and PBS with 0.1 mM nitroprusside was perfused at 100cm H₂O as above. The heart was exposed in situ and the left anterior descending coronary artery (LAD) was ligated as above. Perfusion of 40% yellow Microfil without curing agent at 100 cm H₂O was video recorded under a surgical microscope at 6X magnification. RFT was defined as the time in seconds between Microfil reaching the LAD ligation point and the start of retrograde filling of branches of the distal LAD tree mediated by collaterals connecting to the surrounding coronary artery trees. The above is in part an anatomical method and thus provides an indirect index of relative collateral conductance.

2.5. Myocardial infarction

Mice were anesthetized as above, intubated and ventilated (Harvard Apparatus) using 100% O₂ at 0.2 L/min and 1% isoflurane. A left lateral thoracotomy was performed between the 3^rd and 4^th rib. Following pericardiectomy, the LAD was ligated at its midpoint with 7–0 suture. After confirmation of ligation by tissue blanching and myocardial hypokinesia, the chest wall, muscle and skin were closed with 5–0 suture, intramuscular cefazolin (50 mg/kg) and buprenorphine (0.1 mg/kg) were administered, and recovery was under a heat lamp. Mortality was < 5%. 24 h after ligation, mice were heparinized (1000 units, ip) and after 5 min deeply anesthetized followed by thoracotomy, as above. The heart was removed and frozen at −20°C for 1–3 h wrapped in foil to allow cutting of 1mm thick slabs which were incubated in fresh 1% triphenyl tetrazolium chloride in PBS at 37°C for 15 min. The reaction was stopped with addition of 4% paraformaldehyde (PFA) to a final concentration of 1% PFA. Slabs were imaged at 15X and infarct area (white tissue) was expressed as percentage of total left ventricular+septum area. AAR after LAD ligation was measured by retrograde perfusion of 5% phthalocyanine blue (Phthalocyanine 15:3, Heucotech, Fairless Hills, PA) in 0.9% NaCl. The heart was removed, frozen, cut into 1mm thick slabs, and imaged as above. AAR was expressed as a percentage of total left ventricular+septum area.

2.6. Blood collection

In selected experiments, hypoxia mice were anesthetized and subject to retro-orbital collection of up to 0.6ml blood followed by retro-orbital injection of an equivalent volume of heparinized sterile mouse plasma (BioreclamationIVT, Westbury, NY) performed on the 3^rd and 5^th days of a 7-day recovery period.

2.7. Cell culture and siRNA transfection

Primary human brain microvascular endothelial cells (HBMVECs, Cell Systems, ACBRI 376) were cultured on collagen-coated dishes or glass coverslips in Endothelial Growth Media-MV2 (EGM-MV2, PromoCell, C-22022). For knockdown experiments, the following siRNA oligos were purchased from Qiagen: human Rabep2_12 Flexitube siRNA (I04776520), target sequence TTCAATAAATAAGCAGCTCAG; AllStars Negative Control siRNA (SI03650318, proprietary). HBMECs were cultured in 6-well dishes until ~50% confluent. For each well, 5 ul of a 20 uM siRNA oligos stock solution was mixed with 200 ul serum-free Optimem (Gibco)—Tube A; and 4 ul Oligofectamine plus 22 ul Optimem—Tube B, and incubated separately at RT for 10 min. Tubes A and B were then combined and incubated for 15 min at RT. HBMVEC monolayers were rinsed with serum-free Optimem, and siRNA complexes were added dropwise to each well containing 1 ml of Optimem. After incubation for 4h at 37°C, 1 ml antibiotic-free EGM2-MV media supplemented with 10% FBS was added to existing media. 48h post-transfection, cells were trypsinized and replated (1 × 10⁵ cells/24-well coverslip) onto collagen-coated 24-well dishes for 24h (72h total post-transfection) to allow monolayer formation. Knockdown was confirmed by immunoblot of whole cell lysates 72h post-transfection.

2.8. Quantitative RT-PCR

Total RNA was isolated from mouse left ventricle or HBMVECs grown on 6 or 12-well plates using the Qiagen RNeasy Kit per manufacturer’s protocol. 1 ug of RNA was processed with iScript Reverse Transcription Supermix (Bio-rad) following the manufacturer’s instructions. Amplification of targets was by Taqman assays (Supplemental table I; Applied Biosystems) on a StepOnePlus RT-PCR system (Applied Biosystems). TBP and HPRT1 were used as housekeeping genes in assays of mouse heart, RPLP0 and HPRT1 in assays of HBMVECs. These genes were chosen based on an analysis of the literature for studies examining the effect of hypoxia on gene expression and our pilot studies on the effect of hypoxia on expression of these and four other genes often used as housekeeping genes; the latter showed they were not reduced by hypoxia in the experimental designs that we employed. 4–5 samples per group were analyzed in triplicate using the delta-delta CT method.

2.9. Wound healing assay

Prior to imaging, a linear wound was scratched across the cell monolayer in each well with a p200 pipet tip. Monolayers were rinsed once with EGM2-MV medium, then imaged in 1 ml fresh medium. The 24-well dish was placed in a Tokai stage-top incubator set at 37°C, 5% CO₂ and room air with humidity. Wound closure was imaged using a Nikon Ti2 Eclipse microscope, motorized stage, and 10X phase contrast objective. Multi-acquisition captured 2–3 fields per well/wound every 15 min for 24h. Wound area quantification, as an indicator of cell migration, was by Nikon NIS-Elements software to automatically detect cell-free wound area.

2.10. Immunoblotting

Cells were washed in Hepes buffered saline, lysed in 2X Laemmli sample buffer, and boiled for 5 min. For immunoblots of tissue, 30mg of snap frozen tissue was lysed in 2X Laemmli buffer by mechanical homogenization. Samples were resolved on polyacrylamide gels in the presence of SDS (Bio-rad, Hercules, CA). Gels were transferred onto nitrocellulose membranes, blocked with 5% non-fat milk powder in Tris-buffered saline (25 mM Tris, pH 7.6, 150 mM NaCl) plus 0.1% Tween-20 (TBST) and incubated with primary antibodies at room temperature for 2–4h with gentle rocking. Blots were washed extensively in TBST before incubation with species-appropriate HRP-conjugated secondary antibody (Jackson ImmunoResearch, West Grove, PA) for 1h at room temperature. Blots were again washed in TBST, and fluorescence was detected using enhanced chemiluminescent reagent (ThermoFisher Scientific, Waltham, MA) and imaged (Fluorchem, San Jose, CA). Primary antibodies used were rabbit anti-RABEP1 (Proteintech #14350–1-AP), rabbit anti-RABEP2 (Proteintech #14625–1-AP), and mouse anti-B-ACTIN (Abcam, AB6276).

2.11. Histology and immunofluorescence

Tissues were perfused with freshly prepared 2% PFA followed by washing in PBS, cryoprotected with 30% sucrose overnight at 4°C, embedded in O.C.T. compound, and sectioned on the transverse plane at 8 um. Sections were washed in PBS, permeabilized with 0.02% Triton X-100 in PBS, blocked with 5% donkey serum, 1% BSA, 0.02% Trixon X-100 in PBS, and sections were incubated with primary antibodies overnight at 4°C and secondary antibodies for 2h at room temperature. Antibodies were diluted in blocking solution. Primary antibodies used for immunofluorescence were rat anti-CD31 (BD Biosciences #550274) and rabbit anti-LAMININA (Sigma L9393). Alexafluor conjugated goat anti-rat and goat anti-rabbit secondary antibodies were from ThermoFisher. Randomly selected areas of the myocardium were imaged with a Nikon Eclipse Ti2 inverted microscope with 40X plan fluor objective (Nikon Instruments, Melville, NY).

2.12. Statistics

Data (mean ± SE) underwent either pre-planned 2-sided t-tests or ANOVA (p<0.05 = significant; see legends for tests and n-sizes). Where possible, data were collected by an investigator blinded to treatment group.

3. RESULTS

3.1. Systemic hypoxia induces coronary collateral formation

We employed a previously described protocol that included a gradual acclimation period to decreased inspired oxygen (FiO₂) (Figure 1A) to minimize physiological stress [16]. Adult B6 mice were either maintained under normoxia (“normoxia” mice) or acclimated to the desired level of hypoxia, which was then maintained for 14 days (“hypoxia mice”) followed by gradual recovery to 21% FiO₂ (“recovery mice”). Cage levels of CO₂ were continuously measured in a preliminary study that confirmed that chamber conditions did not result in exposure to high FiCO₂ (Figure S1). As expected, heart weight was similar for normoxia and hypoxia mice at harvest (161.5 ± 7.8 mg, 162.2 ± 6.8 mg). Also as expected, mice exposed to hypoxia experienced less weight-gain and significant polycythemia that was still present 7 but not 28 days after recovery in normoxia (Figure S2).

Figure 1. Hypoxia induces coronary collateral formation.

A, C57BL/6J mice were exposed to 14 days (d) of 7% FiO₂ (“hypoxia”) after gradual reduction of FiO₂. “Hypoxia” mice were assessed immediately after 14d of hypoxia, and “recovery” mice 7d after gradual return to normoxia (21% FiO₂). B,C, Representative images of hearts and summary data, respectively, of mice subjected to hypoxia protocol in panel A, or not (“normoxia”), followed by in situ LAD ligation (star, site of ligation) and infusion of Microfil (scale bar: 1 mm). Area-at-risk (AAR) is demarcated. Microfil did not fill the AAR (LAD tree) in normoxia mice due to absence of native (pre-existing) collaterals, confirming previous results [4]. Hypoxia mice showed robust retrograde filling of the AAR, indicating hypoxia-induced formation of coronary collaterals which persisted after 7 days of recovery. Inset in B, a representative collateral (scale bar: 50 um). Percentage of the AAR filled with Microfil after LAD ligation is proportional to the number and diameter of collaterals formed [4]. Percent of AAR filled was normalized to vascular density in remote zone of left ventricle defined as at least 3 mm away from the border zone delineated by Microfil. D, Hypoxic increase in vascular density in the remote zone persisted after 7 days of recovery. E, Images from representative videos obtained to determine retrograde fill-time (RFT). RFT is the time required to back-fill the LAD tree with Microfil distal to the ligation and is inversely proportional to collateral number and diameter [4]. Lack of filling in normoxia mice due to absence of native collaterals, while hypoxia mice have robust filling indicating collateral formation. F, Normoxia mice showed no retrograde filling (infinite RFT, ∞), whereas collaterals induced by hypoxia regressed after 28 days of recovery in normoxia. Mice exposed to 10% FiO₂ did not show collateral formation. Data in this and subsequent figures are mean ± SEM, individual points are denoted by circles, n-size refers to number of animals unless otherwise specified, and brackets indicate comparison groups. N-size for left-right bars in panel C: 7, 10, 6; D: 7, 10, 6; F: 10, 12, 7, 5. C, D: pre-planned 2-sided t-tests; p<0.05 (*), 0.01 (**), 0.001 (***) or not significant (NS).

As described previously [20], we measured an index of coronary collateral conductance, ie, the amount of retrograde filling with Microfil of the downstream area-at-risk (AAR) of the LAD tree after acute ex vivo ligation at its midpoint (Figure 1B,C). No collateral-dependent filling of the AAR occurred in normoxia mice. This confirmed our previous study that used this and other methods in mice of various strains and ages and found that, unlike larger rodent species also examined, mice have no native coronary collaterals [4,13,20,26]. Maintenance of mice at 7% FiO₂ induced collaterals to form, as indicated by filling of the LAD’s AAR. These newly formed collaterals persisted since the extent of retrograde filling remained unchanged when measured 7 days after return to 21% FiO₂.

To confirm the above findings using a second method, in separate groups of mice we estimated relative collateral resistance as the transit time required for onset of retrograde filling of the distal LAD tree with low-viscosity Microfil after ex vivo LAD ligation (“retrograde-fill time”, RFT) [4]. Consistent with the above findings that mice lack native coronary collaterals, normoxic mice had infinite RFT (Figure 1E,F). In contrast, RFT in hypoxic mice was 16.3 ± 2.3 sec (Figure 1E,F). This value agrees with what we obtained 5 days after LAD ligation, which induced collateral formation (16.5 ± 1.7 sec, n = 13) and is similar to 19 ± 1 sec that we reported previously for ligation-induced collateral formation [4]. In contrast to collateral resistance remaining unchanged in hypoxia mice after 7 days of recovery (Figure 1C), RFT had returned to infinity after 28 days of recovery, presumably due to pruning away or loss of patency of the newly formed collaterals (Figure 1F). This indicates that collaterals stimulated to form by hypoxia do not persist when normoxia is reestablished. To assess whether hypoxic collateral formation is induced by smaller reductions in FiO₂, we exposed mice to 5 days of acclimation followed by 10% FiO₂ for 2, 4 or 8 weeks. Surprisingly, no collaterals formed (RFT was infinite at each time-point, n = 5, 4, 4 respectively, data not shown). Likewise, RFT was also infinite after exposure to 8.5% FiO₂ for 2 weeks (n = 12, data not shown). This is unlike in brain where exposure to 12, 10, 8.5 and 7% FiO₂ stimulated a dose- and duration-dependent formation of new collaterals that persisted unchanged when examined after 6 weeks of return to normoxia [16]. Hematocrits for these groups are given in Figure S2B.

3.2. Hypoxia-induced collateral formation provides myocardial protection

To assess the expected effect of hypoxic collateral formation to reduce infarct size after coronary artery ligation, mice were acclimated to hypoxia, maintained for 14 days, and then recovered to 21% FiO₂ whereupon the LAD was ligated (Figure 2A). We previously found that collateral formation induced by ligation requires 48h to become detectable [4]. Therefore, we measured infarct area 24h after ligation to examine the effect of collaterals induced by hypoxia per se. Infarcts were 23% smaller than in normoxia mice (Figure 2B,C). As expected, this was not due to a difference in anatomic territory of the LAD’s AAR (Figure 2D). Hypoxia mice that received more-distal ligation with smaller AARs had infarcts that were 50% smaller than normoxia mice (Figure S3A). Since hypoxia stimulates erythropoiesis (Figure S2A) and conductance is inversely proportional to blood viscosity which is primarily determined by hematocrit, we hypothesized that collateral blood flow is impeded by increased viscosity. Consist with this, reduction of hematocrit by phlebotomy in hypoxia mice further reduced infarct area (Figure S3B).

Figure 2. Hypoxia-induced coronary collateral formation protects myocardium at risk after acute myocardial infarction.

A, Experimental protocol. After recovery to 21% FiO₂, mice underwent LAD ligation or hearts were harvested for histology. In this figure, “hypoxia” refers to mice exposed to hypoxia followed by recovery. B, Representative images of heart slabs stained with 1% triphenyl tetrazolium chloride 24 h after ligation in hypoxia mice and mice not exposed to protocol in panel A (“normoxia”). Red, non-infarcted tissue; white, infarcted tissue. C, Smaller left ventricle (LV) infarctions in hypoxia mice due to coronary collateral formation. D, Area-at-risk (AAR, phthalocyanine staining) comparable for normoxia and hypoxia mice. E, Hypoxia increased capillary density but not diameter (anti-CD31 (endothelial cell) immunofluorescence). F, No effect of hypoxia on density or size of LV cardiomyocytes (anti-laminin (extracellular membrane) immunofluorescence). N-size (number of animals) for left-right bars in panel C: 10/bar; D: 5/bar; E-F: 5,5,4,4. C-F: pre-specified 2-sided t-tests; p<0.05 (*), 0.01 (**), 0.001 (***) or not significant (NS).

To further characterize this model of hypoxia-induced coronary collateral formation, we also examined capillary and cardiomyocyte density and size. Hypoxia increased capillary density in the left ventricle (LV) without a significant change in capillary width (Figure 2E), although a shift towards a larger number of smaller vessels was evident (Figure S4A). Interestingly, hypoxia decreased capillary density in the right ventricle (RV) without changing width (Figure S5A). Hypoxia did not alter the number or size of LV cardiomyocytes (Figure 2F).

3.3. Expression of Rabep2 is upregulated by hypoxia, and Rabep2 and Vegfa are required for hypoxic stimulation of collateral formation

Hypoxia-inducible factor-1α (HIF-1A) is a ubiquitous transcription factor that regulates oxygen balance in metazoan species and promotes vascular growth by induction of VEGF-A and other angiogenic growth factors to improve oxygen delivery during development and in ischemia [27–29]. We previously showed that Rabep2 is upregulated in brain of hypoxic mice [16] and is required for formation of the native collateral circulation in brain and hindlimb during development [5,30]. RABEP2 is a RAB5 effector that forms a complex with the RAB5 guanine-nucleotide exchange factor RABEX5 to regulate endosomal trafficking of VEGF-receptor 2 which is required for maintained VEGF-A signaling [31,32]. To begin to examine signals involved in coronary collateral formation induced by hypoxia, we measured Rabep2, Vegfa and Vegfr2 expression in mouse heart and found that all three increased with exposure to hypoxia (Figure 3A). Hypoxia also increased mRNA and protein expression of RABEP2 in human brain microvascular endothelial cells (HBMVEC) (Figure 3B–D).

Figure 3. Hypoxia stimulates Rabep2, Vegfa and Vegfr2 expression.

A, Acclimation to 7% FiO₂ (same time-course as in Figure 1A) and exposure to subsequent hypoxia for the indicated durations increased Rabep2, Vegfa and Vegfr2 mRNA (heart, C57BL/6J) relative to mice not exposed to hypoxia (“normoxia”). Mice did not receive LAD ligation. B, 24 or 48 h exposure of human brain microvascular endothelial cells (HBMVEC) to 5% or 2% O₂ increased Rabep2 mRNA. C, 24 h exposure to 5% O₂ increased RABEP2 protein (immunoblot). D, Representative anti-RABEP2 and anti-B-ACTIN immunoblot of total cell lysate from HBMVEC maintained at 21% O₂ or after 24 h 5% O₂. N-size (animals or cell culture experiments) for left-right bars in panel A: 3, 3, 4, 3 (2 for last Vegfr2 bar due to technical sample loss); B: 4, 6, 4, 3, 5, 4; C: 3/bar. A, B: one-way ANOVA; C: 2-sided t-test; p <0.01 (**), p <0.001 (***).

Given the above and our recent findings that systemic hypoxia induces additional collaterals to form in brain beyond the native population by a mechanism that requires VEGF-A and RABEP2 [16], we tested for involvement of both proteins in hypoxic collateral formation in heart. Hypoxia did not significantly upregulate Vegfa in hearts of B6.Vegfa^Fl/Fl;CAG-CreERT^+/− mice, demonstrating the effectiveness of knockdown (Figure S6). We then exposed B6.Vegfa^Fl/Fl;CAG-CreERT^−/−, B6.Vegfa^Fl/Fl;CAG-CreERT^+/−, B6 and B6.Rabep2^−/− mice to hypoxia. Knockdown of Vegfa or deletion of Rabep2 inhibited collateral formation as indicated by significant increase in RFT/coronary collateral resistance (Figure 4A–C). For comparison, we also examined collateral formation induced by experimental acute MI which we had reported previously [4]. Interestingly, deletion of Rabep2 did not inhibit collateral formation induced by ischemia following LAD ligation (Figure S7A). This indicates that different mechanisms drive collateral formation induced by hypoxia and ligation.

Figure 4. Hypoxic coronary collateral formation requires Vegfa and Rabep2.

A,B, Retrograde fill-time (RFT) was prolonged in B6.Vegfa^Fl/Fl;CAG-CreERT^+/− and B6.Rabep2^−/− mice exposed to reduction in FiO₂ (protocol as in Figure 1A). C, Images from representative videos showing robust back-filling (RFT) of the LAD tree distal to the ligation (star) in C57BL/6J (B6) mice is significantly delayed in B6.Rabep2^−/− mice (arrowhead shows filling beginning to occur). D, Rabep2^−/− mice have increased density but not width/diameter of capillaries (anti-CD31 immunofluorescence). E, Hypoxia increased capillary width but not density in Rabep2^−/− mice. F, B6 and Rabep2^−/− mice have similar capillary to cardiomyocyte ratio in normoxia. Rabep2^−/− mice do not show an increase in capillary to cardiomyocyte ratio after hypoxia. G, Representative time points for scratch-wound closure (edges delineated in black) in HBMVEC monolayers transfected with negative control or Rabep2 siRNA. H, Rate of wound area closure (normalized to initial area) is reduced by Rabep2 depletion. Mean ± SEM are for 14 wounds/siRNA per time point. N-size (number of animals) for left-right bars in panel A: 6, 6; B: 8, 11; D-E: 5/bar; F: 5, 5, 4, 5. A,B,D-F: pre-planned 2-sided t-test; p<0.05 (*), 0.01 (**), 0.001 (***) or not significant (NS).

Normoxic Rabep2^−/− mice had increased capillary density in the LV (Figure 4D) and RV (Figure S5B). Unlike in B6 mice (Figure 2E), hypoxia did not increase LV capillary density in Rabep2^−/− mice (Figure 4E). Hypoxia modestly increased LV capillary width in Rabep2^−/− mice (Figure 4E, S4B), but had no effect on RV capillary density or width in Rabep2^−/− mice (Figure S5C). Rabep2^−/− mice had an increased number of myocytes compared to B6 mice but there was no difference in myocyte size (Figure S8A,B). Capillary to myocyte ratio was similar between B6 and Rabep2^−/− mice in normoxia, however, Rabep2^−/− mice did not show an increase in capillary to myocyte ratio in hypoxia (Figure 4F). Similar to B6 mice, myocyte number and size was not affected by hypoxia in Rabep2^−/− mice (Figure S8C). Hematocrit in Rabep2^−/− mice was comparable to normoxic and hypoxic B6 mice (Figure S7B), confirming our previous finding that erythropoiesis is not altered by absence of RABEP2 [16].

The above findings show that RABEP2 and VEGFA are required for hypoxic coronary collateral formation. RABEP2 also negatively regulates capillary density in both ventricles at baseline and is required for increased density in the LV induced by hypoxia. To determine how its absence might impede these processes, we examined siRNA knockdown in HBMVECs. Knockdown of Rabep2, which was effective (Figure S9), inhibited migration following monolayer wounding (Figure 4G,H).

4. DISCUSSION

The study yielded several notable findings. Systemic hypoxemia stimulated formation of collaterals between the LAD and adjacent arterial trees. This was accompanied by smaller infarctions following ligation of the LAD. Hypoxia upregulated expression of Rabep2, Vegfa and Vegfr2. Deletion of Rabep2 inhibited collateral formation induced by hypoxia, as did knockdown of Vegfa. Knockdown of Rabep2 in HBMVEC impaired endothelial cell migration, providing—along with the above increases in expression—insight into how RABEP2 might contribute to hypoxia-induced collateral formation. These results demonstrate that hypoxia, alone, ie, in the absence of acute or chronic arterial obstruction, is sufficient to induce collaterogenesis in the adult heart.

An unexpected finding was that collateral formation only occurred in response to severe hypoxemia, ie, acclimation to 7% FiO₂ over two weeks followed by maintenance at 7% for an additional two weeks. No formation occurred after one or two weeks of acclimation to 10% or 8.5% FiO₂, respectively, followed by two weeks at these more moderate levels of hypoxia. By contrast, we previously found that 7, 8.5 and 10% FiO₂ stimulated new collaterals to form in mouse brain in a hypoxia “dose” and duration-dependent manner [16]. In the present study, no collateral formation occurred when exposure to 8.5% was extended to eight weeks. It is possible that longer exposure to one of the above more moderate levels of hypoxia may induce collateral formation in heart, since we found in brain that 12% FiO₂ for eight weeks caused the same amount of collateral formation induced by two weeks of 10% FiO₂ [16]. Why sensitivity to hypoxic collateral formation differs in the two tissues could arise from several factors: 1) Brain has a greater dependence on oxidative metabolism. Since recent evidence suggests that collaterals are capable of optimizing oxygen delivery (see below), this greater dependence may result in brain having a more robust induction of hypoxic collaterogenesis. 2) The location and anatomic width of the watershed regions where collaterals reside (ie, the “collateral zones”) differ in the two tissues. In brain, the middle (MCA), anterior and posterior cerebral artery trees and the collateral zones situated between them are arrayed in 2-dimensions within the pial membrane that separates the parenchyma of the cerebral cortex from the subarachnoid space. Thus, the collateral zones are located at a distance from the brain’s capillary beds. By contrast, the watershed zones in heart reside within the myocardium between the branches of the 3-dimensionally arrayed coronary artery trees and are thus in close proximity to the capillary beds. 3) Under normoxia, oxygen levels in the cerebral watershed regions are lower than elsewhere in brain (since the watersheds are “furthest from the aorta”). This makes brain tissue under the watersheds especially vulnerable to ischemia if severe hypotension occurs, explaining the “watershed stroke” phenomenon [33,34]; oxygen levels in the coronary watersheds have not been measured]. 4) The cerebral (pial) watersheds appear to be at least 2-fold wider (average ~700 μm among 15 strains of mice) [3], than the watersheds between the LAD and surrounding arterial trees [4], resulting in a longer path length for oxygen diffusion in brain. In fact, the particularly narrow watersheds in the small heart of the mouse, and thus ready diffusion of oxygen from the surrounding terminal arterioles and capillaries, has been suggested as an explanation for why native collaterals do not form during development in heart, unlike in brain and other tissues of this species [4,35]. The differences listed above favor a lower level of oxygen at baseline in the watersheds of murine brain compared to heart. This could result in brain being more sensitive to induction of collaterogenesis when oxygen levels in the watersheds decline further during hypoxemia.

Based on measurement of collateral conductance, the collaterals induced to form by hypoxia persisted after 7 days of recovery but were pruned away or lost patency after 28 days. This is different from brain, where the increase in collateral density induced by 10% or 12% FiO₂ remained unchanged 6 weeks after return to normoxia, as did the amount of reduction in infarct volume seen at this time following MCA ligation [16]. Why durability of the newly formed collaterals differs in the two tissues is unclear, but may relate to the fact that cerebral collaterals reside in the pial membrane where they protrude into the subarachnoid space and therefore do not occupy space in the brain parenchyma. In contrast, coronary collaterals are located within the myocardium, not in the pericardial membrane which is in a location similar to the pial membrane. Thus, efficiencies of space and contractile function may cause the newly formed collaterals in heart to be removed once the hypoxic stimulus is no longer present.

Augmentation of collateral conductance was once thought to occur exclusively via outward remodeling of native collaterals in response to vascular obstruction and the resulting pressure drop and increased fluid shear stress induced across them [9,13]. However, we recently reported that hypoxia alone, in the absence of obstruction, induced new collateral vessels to form in brain [16], and show in the present study that the same occurs in heart. We hypothesize that a primary driving force for this is an increase in the oxygen gradient within the watershed zone between arterial trees [16]. This would result in localized activation of HIF1A, the major driver of Vegfa expression and other oxygen-sensitive genes that likely include Rabep2, although the latter remains to be determined. Over-expression of HIF1A throughout the myocardium resulted in disorganized “lakes of capillaries” and enlarged epicardial coronary vessels with collective increase in coronary blood flow [36]. It is possible that systemic hypoxemia causes local activation of HIF1A in the coronary watersheds and contributes to activation of the complex process and multiple cell types undoubtedly involved in the formation of collateral vessels. Shear stress is a well-known driving force for outward remodeling of native collaterals in pathological arterial obstruction. However, whether it is involved in the formation of new collateral vessels in adult heart and brain, as described in the current and our previous studies [16], is not known. We recently proposed a model for how collaterals might form de novo in the adult in response to systemic hypoxia [16].

Our results show that hypoxic collateral formation requires Vegfa and Rabep2. It is not surprising that Vegfa is required given its role in vasculogenesis, angiogenesis, arterial fate specification and arterialization, and new collateral formation in brain [16,37]. However, our findings in heart and previous report in brain [16] that Rabep2 expression increases in response to hypoxia and is required for hypoxic formation of new collaterals are novel. RABEP2 was originally identified as a RAB5 effector, which binds to its GTP-activated form to promote endocytosis [31]. Rab GTPases are key components of receptor tyrosine kinase signaling [38] and, importantly along with their effectors, are regulated by hypoxia and HIF1A [39,40]. We speculate that RABEP2’s role in hypoxic collateral formation relates to its involvement in VEGFR2 trafficking. Unlike its ligand VEGFA, expression of VEGFR2 did not increase in response to hypoxia [41]. Instead, hypoxia upregulated VEGFR2 through an increase in the number of cell surface receptors [42]. It is hypothesized that stabilization of VEGFR2 occurs by maintenance of its endosomal recycling back to the cell membrane instead of directing it to lysosomal degradation [43,44]. RABEP2 has been recently shown to play a direct role in the upregulation of VEGFR2 by promoting RAB4 and RAB5 mediated endosomal recycling of VEGFR2 [32]. Moreover, deletion of RABEP2 in vivo resulted in impaired endocytic trafficking [5]. Therefore, our finding that hypoxia increases its expression and is required for hypoxic collateral formation provides a plausible mechanism for the stabilization of VEGFR2 in the setting of hypoxia, HIF1A activation, and increased VEGFA expression. Prior studies have suggested that VEGFR2 recycling is critical for collateral remodeling of native pre-existing collaterals (arteriogenesis) and not angiogenesis [44–46]. Our observations suggest that overlapping mechanisms may exist for de novo collateral formation and remodeling that require endocytic cycling of VEGFR2. Previous studies have shown that VEGF-A contributes to formation of the collateral circulation during development[7,47] and in the adult to: collateral remodeling following arterial ligation [47], coronary collateral growth in response to repetitive coronary occlusion [48], and formation of new collaterals following acute coronary ligation [4] and in response to systemic hypoxia in brain [16].

Interestingly, although previous studies have not investigated the effect of reduced FiO₂ on the coronary collateral circulation, earlier work by Eckstein [49] and Scheel and coworkers [50] found that chronic anemia in dogs was accompanied by an increase in coronary collateral conductance. However, these studies did not distinguish between formation of new collateral vessels versus outward remodeling of native pre-existing collaterals—which would be favored by coronary dilation and increased blood. And of course given the level of understanding contemporaneous with these studies, the potential roles of signaling pathways, eg, Vegfa and Rabep2, were not examined.

We were surprised that, unlike with hypoxemia, Rabep2 was not required for collateral formation induced by LAD ligation in adult mice. This may be due to significant differences in the stimuli and activated pathways. LAD ligation causes a sudden disruption of blood flow, pressure differential between the dependent and surrounding arterial trees, anoxia and cell death, and a significant inflammatory response. By contrast, the hypoxia protocol that we used causes a gradual and then sustained reduction in tissue oxygen that likely triggers different signaling pathways. Interestingly, there is evidence that ligation of the LAD may induce collaterals to form in neonatal mice [26,51] and that sustained hypoxemia, also in young mice, re-activates the endogenous regenerative potential of cardiac myocytes [52].

To further characterize our model, we examined the density and size of capillaries and cardiomyocytes. Hypoxia (7% FiO₂) increased capillary density in the LV and decreased it in the RV, whereas capillary diameters did not change. The presumably adaptive increase in density in the LV has been described previously [53,54], whereas the decrease in density in the RV that has been reported previously [55,56] may reflect the eccentric chamber remodeling that occurs in response to pulmonary hypertension caused by hypoxemia of this magnitude/duration. Hypoxia did not alter the density or size of LV cardiomyocytes. We did not examine cardiomyocytes in the RV because their increase in number and size have been reported previously, in accordance with the well-known remodeling caused by hypoxic pulmonary hypertension [57].

Our previous studies identified Rabep2 as a novel determinant of embryonic collaterogenesis and showed that naturally occurring polymorphisms in it account for the majority of the wide variation in collateral number and diameter present in different strains of adult mice [5,30]. Those studies also found that deletion of Rabep2 had no apparent phenotype in adult mice other than for abundance of native collaterals in brain and other tissues, including no effect on formation of the general arterial-venous circulation in tissues during development nor on tumor angiogenesis or capillary density or diameter in gray and white matter of brain (capillarity was not examined in other tissues). We were therefore surprised in the present study to find a phenotype in Rabep2^−/− mice for hypoxic angiogenesis in the heart. Baseline capillary to cardiomyocyte ratio in the LV was similar between Rabep2^−/− and B6 mice, however, the increase in capillary density induced by hypoxia in B6 mice was abolished in Rabep2^−/− mice (Figure 4F). Similarly, Rabep2^−/− mice did not show the decrease in RV capillary density induced by hypoxia in B6 mice (Figure S7A–C). Our previous studies showed that newborn retinal angiogenesis, a process driven by physiological hypoxia, occurs normally in Rabep2^−/− mice [5]. Our findings in this study may reflect differences in the tissues studied in response to systemic, sustained hypoxemia. At baseline, Rabep2^−/− mice had an increased number of LV myocytes relative to B6 mice. However, hypoxia did not affect myocyte number or size in either strain. The basis for differences in myocyte density remains to be determined. As discussed above, we and others have shown that RABEP2 is involved in downstream VEGFA-VEGFR2 signaling in endothelial cells [5,32]. It is possible that cardiomyocytes have a unique requirement for RABEP2 related to VEGFA which regulates their development.

Reduced oxygen and increased shear- and transmural wall-stress induced by vasodilation and polycythemia—acting on endothelial cells, pericytes and smooth muscle cells within the watershed zones between adjacent arterial trees—are likely the predominant stimuli promoting formation of new collaterals within the coronary microcirculation. However, it is possible that systemic sympatho-excitatory effects of chronic hypoxia on cardiac dynamics might also participate. Increased contractility, rate of contraction, and loading conditions on cardiomyocytes would favor an increase in their metabolism, further reducing tissue oxygen and releasing metabolites some of which have been shown to alter proliferation and migration of vascular wall cells. No studies have examined whether chronically altered cardiac dynamics, per se, induce or modify de novo formation of new coronary collaterals, the onset of which requires at least three days to become detectable [4]. However as noted in the Introduction, abundant coronary collaterals have been identified angiographically in individuals with sleep apnea, chronic obstructive pulmonary disease [17,18], and humans and other species residing at high altitude [19,20] As well, increased coronary collateral conductance was reported in dogs made chronically anemic and in piglets exposed to several weeks of reduced inspired oxygen [21,22]. However, these findings do not differentiate between anatomic remodeling or chronic dilation of pre-existing collaterals versus formation of new collateral vessels. Given that acute bradycardia and administration of negative chronotropic agents have been shown to increase coronary collateral blood flow [58,59], future studies are warranted to determine whether chronic presentation of these and other effectors of altered cardiac dynamics induce coronary collaterogenesis.

Our observation that hypoxic collateral formation was accompanied by a reduction in infarct size following ligation of the LAD does not prove causality. However, collateral blood flow is a well-established primary determinant of infarct volume after AMI and stroke. In accordance with this, previous studies of different strains of mice found that the amount of reduction in infarcted tissue post-ligation inversely correlated with the number of collaterals induced to form by either arterial occlusion in heart or hypoxemia in brain [4,16]. Nevertheless, other changes induced by chronic hypoxia could contribute to the decrease in infarct size observed in the present study, including adaptive changes in cardiomyocyte metabolism [60] and increase in capillary density (Figure 2E) and outgrowth of arterioles from adjacent arterial trees into the border zone surrounding the ligated territory of the LAD. Another limitation of this study is that hypoxia has systemic effects on cardiac remodeling, cardiopulmonary coupling and hemodynamics as well as changes in metabolism and gene expression. We cannot exclude the effects of increased coronary blood flow and pulmonary hypertension in our model. However, we did not observe any qualitative differences in mortality or body weight in our comparisons of B6, Rabep2^−/− and Vegfa transgenic mice which were all exposed to the same hypoxia protocols.

In conclusion, sustained hypoxemia provides a model to study formation of new coronary collaterals that avoids the complexities of sudden disruption of blood flow, cell death, inflammation, and repair that are induced by ligation models of cardiac injury. Studying hypoxic collateral formation may identify new disease pathways and markers and provide insights into how coronary collaterals form during development, how new ones form in occlusive coronary artery disease, and how the latter might be augmented therapeutically.

HIGHLIGHTS.

• Systemic hypoxemia stimulated de novo formation of coronary collaterals.

• This was associated with smaller infarctions following ligation of the LAD artery.

• Rabep2 was upregulated by hypoxia and required for hypoxic collateral formation.

• Migration was impaired in Rabep2 depleted endothelial cells—one potential mechanism.

• These findings provide a novel model for studying coronary collateral formation.

NON-STANDARD ABBREVIATIONS AND ACRONYMS:

AAR

Area at risk

AMI

Acute myocardial infarction

B6

C57BL/6J mouse strain

CAD

Coronary artery disease

FiO₂

Fractional inspired oxygen content

HBMVEC

Human brain microvascular endothelial cells

LAD

Left anterior descending coronary artery

RABEP2

Rab GTPase-effector binding protein-2

RFT

Retrograde fill-time

VEGFA

Vascular endothelial growth factor-A

VEGFR2

VEGF Receptor-2

WT

Wildtype