Placental growth factor levels in quadriceps muscle are reduced by a Western diet in association with advanced glycation end products

Abstract

In peripheral artery disease (PAD), atherosclerotic occlusion chronically impairs limb blood flow. Arteriogenesis (collateral artery remodeling) is a vital adaptive response to PAD that protects tissue from ischemia. People with type II diabetes have a high risk of developing PAD and would benefit from arteriogenesis. However, arteriogenesis is suppressed in people with diabetes by a multifaceted mechanism which remains incompletely defined. Upregulation of placental growth factor (PLGF) is a key early step in arteriogenesis. Therefore, we hypothesized that metabolic dysfunction would impair PLGF expression in skeletal muscle. We tested this hypothesis in C57BL/6J and ApoE^−/− mice of both sexes fed a Western diet (WD) for 24 wk. We first assessed baseline levels of PLGF, vascular endothelial growth factor (VEGF-A), and VEGF receptor 1 (VEGFR1) protein in hindlimb skeletal muscle. Only PLGF was consistently decreased by the WD. We next investigated the effect of 24 wk of the WD on the response of PLGF, VEGF-A, VEGFR1, and monocyte chemoattractant protein-1 (MCP-1) to the physiological stimulus of vascular occlusion. Hindlimb ischemia was induced in mice by gradual femoral artery occlusion using an ameroid constrictor. Growth factor levels were measured 3–28 days postsurgery. In C57BL/6J mice, the WD decreased and delayed upregulation of PLGF and abolished upregulation of VEGF-A and VEGFR1 but had no effect on MCP-1. In ApoE^−/− mice fed either diet, all factors tested failed to respond to occlusion. Metabolic phenotyping of mice and in vitro studies suggest that an advanced glycation end product/TNFα-mediated mechanism could contribute to the effects observed in vivo.

NEW & NOTEWORTHY In this study, we tested the effect of a Western diet on expression of the arteriogenic growth factor placental growth factor (PLGF) in mouse skeletal muscle. We provide the first demonstration that a Western diet interferes with both baseline expression and hindlimb ischemia-induced upregulation of PLGF. We further identify a potential role for advanced glycation end product/TNFα signaling as a negative regulator of PLGF. These studies provide insight into one possible mechanism by which type II diabetes may limit collateral growth.

INTRODUCTION

Peripheral arterial disease (PAD) is characterized by the narrowing of arteries, resulting in reduced blood flow in the limbs. In the United States, over 30 million people were estimated to have diagnosed and undiagnosed diabetes in 2015 (8). Type II diabetes, which accounts for 90–95% of all diabetes cases (8), is a strong independent risk factor for PAD. Patients who have had type II diabetes for >6 yr are at an ~threefold greater risk of PAD than patients without diabetes (29). Overall, ~8.5 million Americans are thought to have PAD (3). Reduced blood flow in diabetic PAD leaves affected tissues at risk of ischemia, which can lead to foot ulcers and critical limb ischemia. Current treatments such as revascularization are not an option for many patients with diabetes and PAD because of the risk of surgical complications and impaired wound healing ability (49). Amputation is the final option for patients with critical limb ischemia. In 2005, >95% of non-trauma-related limb amputations in the United States were due to PAD, and ~70% of those patients had diabetes (64). Patients who are undergoing limb amputation for PAD are at high risk of having a repeat amputation, with ~25% of lower limb amputees undergoing a second amputation within a few years (64). Thus, diabetic PAD has a devastating clinical impact.

The compensatory physiological response to gradual arterial occlusion is arteriogenesis, in which remodeling of preexisting collateral arteries results in a natural bypass of the occlusion and minimizes downstream ischemia (6, 22). Therefore, enhancement of the natural adaptive response of arteriogenesis could potentially benefit patients with diabetes experiencing PAD. However, arteriogenesis is impaired in diabetes by a multifaceted mechanism which is incompletely defined (48). Successful arteriogenesis requires a coordinated response involving detection of altered blood flow by vascular cells, recruitment of circulating cells, and expression of a variety of growth factors in a temporally specific manner (19). Impaired shear stress signaling (30) and altered monocyte function (57) have been linked to poor arteriogenesis in diabetes. Likewise, levels of multiple growth factors have been reported to be altered in diabetes. However, arteriogenesis is a complex process, and attempts to rescue it via administration or overexpression of single growth factors have been disappointing overall (31). Therapies targeted at correcting specific diabetes-induced defects in arteriogenic signaling may have the best chance of success. However, the development of such therapies is hampered by a lack of knowledge regarding the mechanism(s) by which diabetes affects the expression and/or downstream signaling of key arteriogenic growth factors.

Placental growth factor (PLGF) is a VEGF-family cytokine that induces arteriogenesis primarily through a monocyte-mediated mechanism (42). PLGF overexpression induces the formation of large stable vessels (40), whereas PLGF inhibition limits collateral growth (7). In animal models of PAD, PLGF infusion at the site of vessel occlusion has been shown to be more effective than VEGF-A at inducing arteriogenesis (42, 54). Similarly, administration of a PLGF adenoviral construct improves wound healing by increasing microvasculature in the wound in the setting of impaired wound healing in diabetes (9). Furthermore, the impairment of wound healing by diabetes is associated with decreased PLGF levels (9). These findings suggest that PLGF may have a critical role in peripheral arteriogenesis and that decreased PLGF levels could contribute to poor peripheral arteriogenesis in diabetic PAD. However, mechanisms regulating PLGF expression in skeletal muscle, the tissue in which peripheral arteriogenesis occurs, have not yet been characterized. Likewise, the effect of diabetes on skeletal muscle PLGF expression has not been closely examined.

In this study, we hypothesized that metabolic conditions approximating those found in patients with type II diabetes could act to 1) decrease baseline expression of PLGF and other arteriogenic factors and/or 2) prevent upregulation of PLGF and other arteriogenic factors by the physiological stimulus of vascular occlusion. We tested these hypotheses in two mouse models of Western diet (WD)-induced metabolic dysfunction. Our in vivo results showed that both baseline and hindlimb ischemia-induced PLGF levels in skeletal muscle were consistently decreased in mice fed a WD. In contrast, the WD decreased hindlimb ischemia-induced VEGF-A and VEGF receptor 1 (VEGFR1) expression but had variable effects on their baseline levels, and monocyte chemoattractant protein-1 (MCP-1) expression was affected by mouse strain only. To begin identifying the mechanism(s) behind these effects, we conducted in vitro studies to identify the specific metabolic factor(s) influencing PLGF levels in skeletal muscle cells (SKMCs) and endothelial cells (ECs). These studies identified advanced glycation end products (AGEs) (molecules that form because of glycation of proteins and aldehydes in the presence of sugar under chronic hyperglycemic conditions) (2) and tumor necrosis factor α (TNFα) as potential mediators of this response.

Our findings show that type II diabetes-associated metabolic dysfunction impairs the regulation of both baseline and stimulated expression of PLGF, whereas it has variable effects on the other factors examined. We further provide evidence that an AGE/TNFα-mediated mechanism could potentially be a negative regulator of PLGF expression in skeletal muscle tissues. Impaired arteriogenic growth factor expression may contribute to poor wound healing and reduced capacity for collateral growth in patients with diabetic PAD. These studies lay the groundwork for detailed mechanistic investigations of these effects, which may lead to the development of novel therapeutic strategies to improve patient outcomes in diabetic PAD.

MATERIALS AND METHODS

Baseline studies.

C57BL/6J and ApoE^−/− mice of both sexes were fed either a WD or a control diet (CD). The diet regimen was started when mice were 6–8 wk of age and was continued for 24 wk. Male and female mice were studied separately. Therefore, we studied a total of 8 groups: C57BL/6J males fed the CD (n = 9), C57BL/6J females fed the CD (n = 8), C57BL/6J males fed the WD (n = 15), C57BL/6J females fed the WD (n = 7), ApoE^−/− males fed the CD (n = 6), ApoE^−/− females fed the CD (n = 6), ApoE^−/− males fed the WD (n = 6), and ApoE^−/− females fed the WD (n = 6). All mice had access to food and water ad libitum and were housed in a temperature-controlled facility with a 12-h dark-light cycle. At the conclusion of the study period, mice were euthanized by excision of the heart under 5% isoflurane anesthesia. Two sets of skeletal muscles were isolated from the hindlimb: the quadriceps femoris muscle (from the thigh) and the gastrocnemius-plantaris-soleus (GPS) muscle group (from the calf). Tissues were immediately immersed in liquid nitrogen and stored at −80°C until use. All animal studies were approved by the Institutional Animal Care and Use Committee at Oklahoma State University.

Diets.

The WD (TD 88137, Harlan) supplied 42% kcal from anhydrous milk fat and cholesterol, 42.7% kcal from carbohydrate, and 15.2% kcal from protein. The CD (TD 120336) supplied 13% kcal from fat, 67.9% from carbohydrate, and 19.2% from protein. The WD also had three times more sucrose by weight compared with the CD. The source of the relatively high carbohydrate content in the CD is cornstarch and maltodextrin, which are not readily available sources of sugar. Therefore, the WD used in this study is characterized by both high-fat and high-sucrose content.

Metabolic phenotyping.

Body weight of each mouse was recorded weekly. Fasting (6 h) blood samples were obtained from all mice at the end of the 24-wk diet regimen for metabolic measurements, as described below.

Cholesterol and triglycerides.

Total plasma cholesterol was determined using the Amplex Red cholesterol assay (Molecular Probes), and total plasma triglycerides were measured using a triglyceride determination kit (Sigma-Aldrich) following the manufacturers’ protocols.

Assessment of glucose homeostasis.

Glucose homeostasis was assessed by intraperitoneal glucose tolerance testing (IPGTT) under fasting (6 h) conditions in all mice just before euthanization. IPGTT was performed on nonanesthetized mice by administering a 1 g/kg ip d-glucose bolus (~30 mg/mouse in 300 μL water). Venous blood (0.05 mL·mouse⁻¹·time point⁻¹; a total of 0.25 mL) was drawn from the facial vein for measurement of glucose and insulin at 0, 15, 30, 60, and 120 min after glucose administration. Plasma glucose level was measured using a portable glucose meter (Life Scan). Insulin was measured using a mouse insulin ELISA kit (EMD Millipore, Germany).

Oxidative stress.

Plasma 8-isoprostane was measured as a biomarker for oxidative stress using a competitive enzyme immunoassay (Cayman Chemical) according to the manufacturer’s protocol.

Protein isolation.

Skeletal muscle tissue was homogenized in 1 mL of radioimmunoprecipitation assay lysis buffer containing protease inhibitors (PMSF, sodium orthovanadate, leupeptin, benzamidine hydrochloride, aprotinin, and pepstatin A) using a Bio-Gen Pro200 homogenizer (PRO Scientific). The homogenates were centrifuged at 700 g for 5 min, and the supernatants were transferred to clean microcentrifuge tubes.

Enzyme-linked immunosorbent assay.

PLGF, VEGF-A, VEGFR1, MCP-1, and TNFα protein levels were quantified by ELISA (Mouse Quantikine and Mouse and Human DuoSet ELISAs, R&D Systems) following the manufacturer’s protocols. All samples were assayed in duplicate. Assay plates were read using a BioTek Synergy HT plate reader (BioTek Instruments) at 450 nm. In some mice, the AGE level in skeletal muscle was also measured by ELISA (OxiSelect Advanced Glycation End Product Competitive ELISA kit, Cell Biolabs). All data were normalized to total protein concentration, as determined by bicinchoninic acid assay (Pierce), except in the case of the in vitro AGE studies (in which the additional protein being added would skew the analysis).

Hindlimb ischemia studies.

Based on the results of the baseline studies, which showed limited sex-dependent differences between groups, male and female mice were pooled together for these experiments to limit the number of animals required. Diets were administered as described above. Therefore, we studied 4 groups: C57BL/6J mice fed the CD (n = 52), C57BL/6J mice fed the WD (n = 58), ApoE^−/− mice fed the CD (n = 42), and ApoE^−/− mice fed the WD (n = 45). After 24 wk of the diet program, mice were subjected to gradual femoral artery occlusion by placement of an ameroid constrictor (0.25-mm inner/0.5-mm outer diameter, Research Instruments SW; Fig. 3). For ameroid placement, mice were anesthetized using continuous 2% isoflurane by inhalation. The hair over the thigh was shaved, and the area was disinfected with Betadine. A 1–2-cm incision was made in the medial aspect of the thigh, and the femoral artery and vein were exposed. The vein, artery, and nerve were isolated from each other by blunt dissection under a microscope (Leica M651 surgical stereomicroscope). An ameroid constrictor was then placed around the femoral artery to produce gradual [~7–10 days postsurgery (dps)] occlusion. The skin was closed with VetBond (3M Animal Care Products) tissue adhesive and wound clips, and the incision site was treated topically with 2% lidocaine ointment and bacitracin-polymyxin-neomycin antibiotic ointment. The right (unoperated) limb served as a paired control. Mice were returned to cages for recovery and closely observed for signs of discomfort or altered limb function. Wound clips were removed 7 dps if still present. Mice were returned to the same diet until euthanization at 3, 5, 7, 10, 14, 21, or 28 dps (n = 4–7·group⁻¹·time point⁻¹) followed by tissue collection and analysis as described above.

Fig. 3.

An ameroid constrictor was used to induce hindlimb ischemia. FA, femoral artery.

Sham procedures.

In a different set of mice, a sham surgery was performed (n = 4·group⁻¹·time point⁻¹). The vessels were exposed and isolated as described above, with the exception of placing the ameroid constrictor on the femoral artery, and the wound was closed as above. The right limb served as a paired control for these animals as well.

Assessment of blood flow.

In a separate group of mice, hindlimb ischemia was induced following 28 wk of diet treatment as described in the previous section. Hindlimb adductor muscle blood flow was assessed at 21 dps (n = 6/group) by infusion of fluorescent microspheres. Under isoflurane anesthesia, a PE-10 catheter was inserted into the right carotid artery proximal to the aortic bifurcation. Catheters were filled with saline containing 100 IU/mL heparin and 1 mg/mL adenosine. Fluorescent microspheres (red-orange, excitation/emission 565/580 nm, 15 μm diameter, 5 × 10⁵ microspheres, Molecular Probes) were infused through the carotid catheter at a rate of 100 μL/5 s. Three minutes after microsphere infusion, mice were euthanized under 5% isoflurane. Hindlimb thigh skeletal muscles (adductor, quadriceps, and biceps femoris) were collected from each hindlimb separately for analysis of the accumulation of microspheres. Kidneys were also collected to verify even distribution of microspheres to both sides of the body. Even distribution of microspheres in each side of the body was ensured by comparing the fluorescence accumulation between the kidneys. Data were discarded if the kidney readings demonstrated >10% coefficient of variation. Tissue samples were weighed and then digested in 5 mL of 2.3 M ethanolic KOH with 0.5% Tween 80 overnight in a shaker bath at 60°C. The samples were centrifuged at 2,000 g for 20 min, and the supernatant was removed. The tissue pellet was resuspended twice in distilled water with 0.5% Tween 80 and centrifuged, and the supernatant was discarded. The pellet was then resuspended in methanol, centrifuged, and decanted. The remaining methanol was removed by evaporative drying, and the microsphere residue was dissolved in 2 mL of 2-ethoxyethylacetate. Fluorescence was measured by flow cytometry (Accuri C6; BD) using the recommended excitation and emission wavelengths for red-orange microsphere fluorescence. Mean fluorescence counts from the occluded and nonoccluded leg were normalized to tissue weight and used to calculate the ratio of flow in the occluded limb to the nonoccluded limb as an index of arteriogenesis.

Cell culture.

For in vitro studies, human skeletal muscle cells (hSKMCs; Lonza), mouse skeletal muscle cells (mSKMCs; ATCC), human coronary endothelial cells (hCECs; Lonza), and RAW 264.7 murine macrophages (ATCC) were grown to confluence at 37°C in a humidified incubator containing 5% CO₂. hSKMCs, mSKMCs, and hCECs were maintained in SKMC growth medium (SkBM, Lonza), DMEM (ATCC), and endothelial cell growth medium (EGM-2MV, Lonza), respectively, supplemented with growth factors according to the protocols provided by the vendors. RAW cells were cultured in DMEM containing 10% fetal bovine serum and 1% penicillin/streptomycin. Equal numbers of cells were seeded on each well of 6- or 12-well culture plates for experiments. hSKMCs were used at passage 9, mSKMCs were used at passage 17, hCECs were used at passage 5, and RAW cells were used at passage 10 for all experiments.

Treatment of cells.

Prior to experimental treatments, confluent hSKMCs and mSKMCs were serum starved in 4% serum media for 24 h. hCECs were serum starved in 2% serum media for 24 h. To mimic hyperlipidemia, hyperglycemia, hyperinsulinemia, and oxidative stress, cells were treated with 200 µg/mL LDL (Alfa Aesar), 200 µg/mL oxidized LDL (Alfa Aesar), 5–20 mM glucose (Sigma), 1,000–1,500 µg/mL AGEs (both glucose- and glyceraldehyde-derived; see below), 100–500 mU/L insulin (Sigma), 0.1–0.3 mM hydrogen peroxide (H₂O₂; VWR), or 100 ng/mL TNFα (Sigma) for 24 h. Treatments were added directly to the cell medium. The medium was collected at 4, 8, and 24 h posttreatment in fresh tubes containing protease inhibitors for measurement of the PLGF protein level. Along with treated samples, medium from untreated cells was also collected and used as a control. After the medium collection, cells were washed with PBS, trypsinized (1% TrypLE, Life Technologies), collected, and lysed using radioimmunoprecipitation assay buffer containing protease inhibitors for VEGFR1 quantification. Following a 1-h incubation of cells in lysis buffer on ice, samples were centrifuged at 14,000 g for 5 min to pellet cell debris. The supernatants were then collected for ELISA.

AGE preparation.

AGE was prepared through glycation of bovine serum albumin (BSA) in the presence of d-(+) glucose or dl-glyceraldehyde. To produce glucose-derived AGEs, 50 mg/mL BSA (Sigma-Aldrich) was incubated with 45 mg/mL d-(+) glucose (Sigma-Aldrich) in PBS (pH 7.4). The solution temperature was maintained at 37°C for 8 wk to fully glycate albumin and create irreversibly formed AGEs. A control solution was prepared by incubating 50 mg/mL BSA in PBS only at 37°C for 8 wk. Glyceraldehyde-derived AGE was prepared by incubating 25 mg/mL BSA with 20 mM dl-glyceraldehyde (Sigma-Aldrich) in PBS at 37°C for 1 wk. A control solution was prepared by incubating 25 mg/mL BSA in PBS at 37°C for 1 wk.

After the incubation periods described above, both the control and glycated samples were dialyzed extensively in fresh PBS at 4°C. The solution was changed every 2 h for a total of 4 changes. The total protein concentration of the AGE preparation and the control BSA preparation was quantified using a bicinchoninic acid assay kit (Pierce). The endotoxin level of the AGE and control BSA preparations was tested using the LAL Chromogenic Endotoxin Quantitation Kit (Life Technologies) and was found to be below 1.0 EU/mL.

Statistical analyses.

For baseline studies, differences in mouse metabolic phenotypic parameters and mouse PLGF, VEGF-A, and VEGFR1 protein expression between the diets, strains, and sexes studied were analyzed by three-way analysis of variance (ANOVA) followed by post hoc testing using SAS Version 9.4 (SAS Institute, Cary, NC). For hindlimb ischemia studies, differences in mouse PLGF, VEGF-A, VEGFR1, and MCP-1 expression across the diets and strains studied were analyzed by two-way ANOVA (Excel). Values are presented as means ± SE. To identify possible relationships between metabolic parameters and PLGF expression, linear regression and correlation analysis were performed using SAS Version 9.4 (SAS Institute, Cary, NC). Effects of different treatment conditions on PLGF protein levels for in vitro experiments were assessed using one-way ANOVA (Excel). If significance was found, further analysis was conducted through multiple comparisons using the Student-Newman-Keuls post hoc method (Primer of Biostatistics software, McGraw Hill). Differences were considered significant at P ≤ 0.05.

RESULTS

Metabolic characterization of WD-fed mouse models.

To induce a spectrum of metabolic dysfunction encompassing the range of metabolic profiles found in type II diabetics, C57BL/6J and ApoE^−/− mice were fed either a CD or a WD for 24 wk. Male and female C57BL/6J mice fed the CD represent the control groups for this study. At the conclusion of the diet protocol, we performed metabolic phenotyping (Table 1).

Table 1.

Metabolic phenotyping data

	C57BL6/J				ApoE−/−
	Male		Female		Male		Female		P value
	CD	WD	CD	WD	CD	WD	CD	WD
Plasma cholesterol, mg/dL	104.94 ± 8.43†	338.30 ± 10.92*†‡	77.61 ± 9.70†	260.09 ± 19.26*†‡	614.20 ± 9.27†	1207.00 ± 16.70*,†	604.13 ± 18.97†	1242.73 ± 18.24*†	*†‡<0.0001
Plasma triglycerides, mg/dL	94.38 ± 9.40†	92.75 ± 10.74†	103.81 ± 11.23†	133.79 ± 12.27†	204.72 ± 35.96†‡	581.72 ± 15.94*†‡	365.31 ± 18.57†‡	733.31 ± 85.90*†‡	*<0.0001 †<0.05 ‡<0.01
Plasma 8-isoprostane, pg/mL	84.17 ± 10.90†	298.40 ± 13.09*†‡	60.90 ± 5.85†	194.93 ± 15.20*†‡	275.13 ± 21.39†	492.70 ± 17.58*†‡	302.88 ± 26.73†	574.28 ± 50.10*†‡	*†<0.0001 ‡<0.05
Fasting glucose week 0, mg/dL	137.33 ± 10.64	136.60 ± 6.72	117.50 ± 6.62‡	121.43 ± 7.60	135.83 ± 5.06	136.67 ± 10.11	125.33 ± 4.14	128.67 ± 7.20	‡<0.05 *†NS
Fasting glucose week 24, mg/dL	131.67 ± 3.00†	193.80 ± 3.63*†	121.00 ± 2.23†	194.71 ± 5.09*	155.17 ± 4.83†	175.33 ± 4.94*†	164.50 ± 4.09†	181.67 ± 6.24	*†<0.05 ‡NS
Plasma insulin, ng/mL	1.36 ± 0.49	7.33 ± 0.43*†	1.61 ± 0.38	6.93 ± 0.78*†	0.54 ± 0.02	5.54 ± 0.45*†	0.79 ± 0.06	5.15 ± 0.33*†	*<0.0001 †<0.001 ‡NS
Area under the glucose curve	26,852 ± 1,126‡	39,959 ± 742*†‡	23,003 ± 890‡	44,731 ± 2,317*†‡	23,524 ± 1,344	29,876 ± 1,161*†	24,654 ± 1,078	27,245 ± 2,806†	*<0.01 †<0.0001 ‡<0.05
HOMA-IR index	10.74 ± 1.01	60.31 ± 1.81*	10.29 ± 1.17	65.40 ± 7.94*†	4.51 ± 0.22	53.28 ± 4.34*	4.50 ± 0.30	42.66 ± 4.92*†	*†<0.0001 ‡NS

Values are means ± SE. CD, control diet; HOMA-IR, homeostatic model assessment of insulin resistance; NS, not significant; WD, Western diet.

^*Diet effect by strain/sex;

^†strain effect by sex/diet;

^‡sex effect by strain/diet (3-way ANOVA followed by post hoc testing).

Among the WD-fed groups, ApoE^−/− mice were significantly less obese than C57BL/6J mice (data not shown). Plasma cholesterol was increased in both C57BL/6J mice and ApoE^−/− mice, relative to the matched-strain/sex group on the CD. Significant effects of both sex and strain were also observed (P < 0.0001). In the C57BL/6J strain, the WD had a relatively greater effect on plasma cholesterol in males than in females. No sex differences were observed in the ApoE^−/− strain. However, plasma cholesterol levels were several times higher in all ApoE^−/− groups than in the corresponding C57BL/6J groups.

The WD had no effect on plasma triglyceride levels in C57BL/6J mice of either sex but significantly elevated triglyceride levels in ApoE^−/− mice of both sexes (P < 0.0001). Female ApoE^−/− mice had higher triglyceride levels than male ApoE^−/− mice on either diet (P < 0.01). There was a clear strain difference, with significantly higher triglyceride levels in all ApoE^−/− mice relative to the corresponding C57BL/6J groups (P < 0.05).

Plasma 8-isoprostane was measured to characterize oxidative stress. Plasma 8-isoprostane levels were significantly increased in all groups by the WD (P < 0.0001). There was a significant strain effect, with all ApoE^−/− mice demonstrating higher oxidative stress than the corresponding C57BL/6J groups (P < 0.0001). A significant effect of sex on the effect of the WD on 8-isoprostane levels was also seen in both strains. Interestingly, the effect varied in opposite directions by strain. C57BL/6J females fed the WD had lower 8-isoprostane levels than C57BL/6J males, whereas ApoE^−/− females fed the WD had higher 8-isoprostane levels than ApoE^−/− males (P < 0.05).

Fasting blood glucose levels were significantly elevated in WD-fed C57BL/6J mice of both sexes from weeks 3 to 4 on the diet onward (data not shown). In contrast, WD-fed ApoE^−/− mice did not demonstrate any impairment in fasting blood glucose until week 13, followed by inconsistent impairment (data not shown). Glucose clearance as measured by IPGTT and presented as area under the curve (AUC) indicated that both male and female WD-fed C57BL/6J mice had impaired glucose tolerance (P < 0.01). In contrast, male WD-fed ApoE^−/− mice had slightly but significantly impaired glucose tolerance (P < 0.01), whereas female WD-fed ApoE^−/− mice showed no impairment as assessed by this measurement. These results are somewhat surprising, since ApoE^−/− mice have previously been reported to exhibit impaired glucose tolerance following shorter periods of WD feeding, for example, 7 wk (51). However, the homeostatic model assessment of insulin resistance index was markedly increased in all WD-fed groups. Therefore, feeding of the WD disrupted glucose homeostasis, albeit with some strain- and sex-dependent differences.

All WD-fed mice (both strains and sexes) were hyperinsulinemic (P < 0.0001). A significant (P < 0.001) strain difference existed, with WD-fed ApoE^−/− mice of both sexes showing lower insulin levels than their C57BL/6J counterparts. However, the fold increase in insulin in WD-fed ApoE^−/− mice was greater than that seen in C57BL/6J mice.

Baseline PLGF, VEGF-A, and VEGFR1 expression in quadriceps muscle.

The muscles of the thigh are the primary location for arteriogenesis in the hindlimb because of the presence of preexisting collaterals. We found that PLGF, VEGF-A, and VEGFR1 protein levels in the quadriceps muscle were affected differently by the WD (Fig. 1). PLGF protein was significantly decreased by the WD in all groups except C57BL/6J females (P < 0.001). In contrast, the WD had a differential, strain-dependent effect on VEGF-A protein levels in quadriceps muscle: VEGF-A was increased by the WD in the C57BL/6J strain (P < 0.0001) but decreased in the ApoE^−/− strain (P < 0.0001). A clear sex-dependent difference in the response of VEGFR1 to the WD was observed in quadriceps muscle, with WD-fed females of both strains showing a diet-dependent increase in VEGFR1 protein (P < 0.05), whereas males showed no response. A strain-dependent difference in VEGFR1 expression was also noted, with overall higher expression levels in ApoE^−/− females versus C57BL/6J females (P < 0.05).

Fig. 1.

Left: quadriceps femoris placental growth factor (PLGF) protein levels were significantly reduced by a Western diet (WD) in all groups except C57BL/6J females. Middle: in contrast, VEGF-A showed strain-dependent, opposing responses to the diet, with increased VEGF-A protein in WD-fed C57BL/6J mice, but decreased VEGF-A protein in WD-fed ApoE^−/− mice. Right: VEGF receptor 1 (VEGFR1) showed a clear sex-dependent effect of the diet treatment, with increased VEGFR1 protein found in WD-fed females of both mouse strains, but no effect of the diet in males.

In these studies, we assessed PLGF, VEGF-A, and VEGFR1 protein rather than mRNA. Previous studies by our group have demonstrated that PLGF mRNA and protein are not well correlated under some circumstances and that regulation of PLGF expression has an important posttranscriptional component; therefore, the significance of mRNA results may be difficult to interpret (53).

To understand the possible mechanism(s) behind the WD-induced reduction of PLGF in quadriceps muscle, we performed Pearson correlation analysis of each individual animal’s quadriceps muscle PLGF protein level with its plasma cholesterol, isoprostane, triglyceride, insulin, glucose, AUC, and homeostatic model assessment of insulin resistance index values. Quadriceps muscle PLGF protein was significantly and negatively correlated with cholesterol, 8-isoprostane, insulin, and glucose in both C57BL/6J and ApoE^−/− mice (Table 2).

Table 2.

Correlation of metabolic indices with PLGF levels in skeletal muscle

	Cholesterol		Isoprostane		Insulin		Glucose (30 min)
	Quad	GPS	Quad	GPS	Quad	GPS	Quad	GPS
C57BL/6J
r value	−0.49	−0.61	−0.54	−0.52	−0.59	−0.50	−0.45	−0.48
P value	<0.01	<0.01	<0.001	<0.01	<.0001	<0.05	<0.01	<0.05
n	39	24	39	24	38	24	39	24
ApoE−/−
r value	−0.78	−0.61	−0.55	−0.49	−0.76	−0.57	−0.45	−0.15
P value	<0.0001	<0.01	<0.01	<0.05	<0.0001	<0.01	<0.05	NS
n	24	21	24	21	24	21	24	21

GPS, gastrocnemius-plantaris-soleus; NS, not significant; PLGF, placental growth factor; Quad, quadriceps; n = number of samples.

Baseline PLGF, VEGF-A, and VEGFR1 expression in GPS muscle.

In contrast to the quadriceps muscle, in which arteriogenesis (collateral growth) is the predominant type of vascular remodeling following femoral occlusion (19), the GPS (calf) muscles are primarily a site of angiogenesis (capillary proliferation). To determine whether a WD affects growth factor expression differently in tissue that is primarily angiogenic rather than arteriogenic, we examined PLGF, VEGF-A, and VEGFR1 expression in the GPS muscle group (Fig. 2). PLGF expression in the GPS muscle group tended to be reduced by the WD, but this decrease was significant only in C57BL/6J males and ApoE^−/− females (P < 0.05). VEGF-A protein expression in the GPS muscle group was decreased by the WD in both strains (P < 0.05) in contrast to the diet’s opposing effects on VEGF-A expression in quadriceps muscle from the two strains. The WD had a limited effect on VEGFR1 expression in the GPS, decreasing it only in ApoE^−/− males (P < 0.05).

Fig. 2.

Left: placental growth factor (PLGF) protein levels in the gastrocnemius-plantaris-soleus (GPS) muscle group tended to be reduced by the Western diet (WD), although the effect was not significant for all groups. Center: VEGF-A was consistently reduced by WD feeding. Right: VEGF receptor 1 (VEGFR1) was decreased by the WD in ApoE^−/− mice but was unaffected by diet in C57BL/6J mice.

Pearson correlation analysis was performed to compare each individual animal’s PLGF protein level in the GPS muscle group with metabolic variables as described for the quadriceps muscle. Similar to our observations in the quadriceps, PLGF levels in the GPS were negatively correlated with plasma cholesterol, isoprostane, and insulin in both C57BL/6J and ApoE^−/− mice. Glucose was negatively correlated with PLGF only in the C57BL/6J strain (Table 2).

In keeping with the inconsistent responses of VEGF-A and VEGFR1 to the diet, no consistent pattern of correlations for these factors could be identified across the diet groups/muscle types (data not shown).

Stimulated PLGF, VEGF-A, VEGFR1, and MCP-1 expression in quadriceps muscle.

After 24 wk of the diet program, a second group of mice was subjected to gradual left femoral artery occlusion by placement of an ameroid constrictor as described by Yang et al. (61), and the arteriogenic process was allowed to proceed for up to 28 dps (Fig. 3). In this model of hindlimb ischemia, closure of the ameroid occurs at ~7 dps, and flow recovers to ~70% of presurgery values from 21 dps onward (61). To study the temporal expression pattern of the arteriogenic growth factor response in the tissue in which arteriogenesis is occurring, we harvested thigh skeletal muscle from both limbs at 3, 5, 7, 10, 14, 21, or 28 dps and measured the levels of PLGF, VEGF-A, VEGFR1, and MCP-1 protein. Data are presented as the ratio of each protein between the left (occluded) limb and the right (nonoccluded) limb, which served as a paired control.

In C57BL/6J mice fed the CD, PLGF expression was significantly elevated beginning 3–10 dps and returned toward control levels thereafter, in agreement with our previous observations of PLGF mRNA in rat hindlimb collateral arteries (45). Critically, upregulation of PLGF by occlusion was both delayed and reduced in magnitude in C57BL/6J mice fed the WD, in which peak expression did not occur until up to 21 dps (Fig. 4A). Sham operation did not significantly affect PLGF (data not shown).

Fig. 4.

Placental growth factor (PLGF) (A), VEGF-A (B), VEGF receptor 1 (VEGFR1) (C), and monocyte chemoattractant protein-1 (MCP-1) (D) expression displayed varying time courses in response to hindlimb ischemia in C57BL/6J mice. In mice fed the Western diet, expression of all factors except MCP-1 was significantly reduced, delayed, or abolished.

VEGF-A was also upregulated by femoral occlusion in C57BL/6J mice fed the CD. The time course of VEGF-A expression was delayed relative to PLGF, peaking at 14 dps. VEGF-A upregulation by occlusion was absent in C57BL/6J mice fed the WD (Fig. 4B). Sham operation did not affect VEGF-A levels (data not shown).

VEGFR1 expression also displayed a unique time course, with an increase beginning at 10 dps in the C57BL/6J control group (P < 0.0001), that was maintained until the conclusion of the study. This increase was abolished by the WD (Fig. 4C). Sham operation did not significantly increase VEGFR1 expression (data not shown).

Finally, we examined the time course of MCP-1 expression, since MCP-1 is known to be upregulated early in arteriogenesis. MCP-1 protein was already elevated at 3 dps in C57BL/6J mice, similarly to our previously reported results for MCP-1 mRNA in rat hindlimb skeletal muscle (34). Surprisingly, the WD had no effect on MCP-1 upregulation in C57BL/6J mice in contrast to its clear effect on PLGF, VEGF-A, and VEGFR1 (Fig. 4D). MCP-1 in the operated limb of sham controls was slightly increased by the sham surgery (to 3.8 ± 0.58-fold of the unoperated limb; n = 4), but this increase was significantly lower than the increase seen in the limbs of animals receiving the ameroid implant (27.8 ± 5.5-fold of the unoperated limb; n = 4, P < 0.01 vs. sham).

We also assessed the effect of hindlimb ischemia on PLGF, VEGF-A, VEGFR1, and MCP-1 in the ApoE^−/− strain. Strikingly, the response of all 4 factors to ameroid occlusion was essentially absent in all ApoE^−/− mice, regardless of diet (Fig. 5, A–D), suggesting that the level of metabolic dysfunction present in this strain is sufficient to suppress growth factor expression without the added insult of a WD.

Fig. 5.

The response of placental growth factor (PLGF) (A), VEGF-A (B), VEGF receptor 1 (VEGFR1) (C), and monocyte chemoattractant protein-1 (MCP-1) (D) expression to hindlimb ischemia induced by ameroid occlusion was almost completely absent in ApoE^−/− mice fed either the control diet or the Western diet.

Relationship of growth factor expression to metabolic parameters.

Mice in hindlimb ischemia studies were also metabolically phenotyped as described for mice used in baseline studies. To begin identifying the mechanism(s) by which a WD affects upregulation of arteriogenic factors in response to hindlimb ischemia, we performed Pearson correlation analysis of metabolic indices versus growth factor expression at each time point. Although this analysis did not conclusively rule in/out any single metabolic variable in the growth factor responses we observed, some clear patterns emerged.

Overall, PLGF displayed the strongest association with isoprostane and insulin levels, factors which were also significantly negatively correlated with baseline expression. The most significant effect of isoprostane and insulin on upregulation of PLGF by hindlimb ischemia was observed at 7–10 dps, around the time of expected ameroid closure, and around the peak of PLGF expression in the control group (isoprostane R = −0.57, P < 0.01; insulin R = −0.64, P < 0.01; n = 19–21).

VEGF-A levels at 14 dps (peak of VEGF-A expression in the control group) were negatively associated with isoprostane and the AUC from the IPGTT test (isoprostane R = −0.55, P < 0.05; AUC R = −0.49, P < 0.05; n = 18).

VEGFR1 levels continued to climb until at least 21 dps. At 14 and 21 dps, VEGFR1 was most strongly associated with isoprostane (R = −0.60, P < 0.01 at 14 dps; R = −0.46, P = 0.05 at 21 dps; n = 18–19).

MCP-1 protein was sharply upregulated immediately following occlusion and rapidly returned toward baseline levels in the control group, in agreement with our previous results for MCP-1 mRNA in rat hindlimb (45). Accordingly, the most significant associations of metabolic variables with MCP-1 were observed early in the time course. Interestingly, MCP-1 was most strongly negatively associated with cholesterol, isoprostane, and triglycerides at 3 dps (cholesterol R = −0.65, P < 0.01; isoprostane R = −0.63, P < 0.01; triglycerides R = −0.54, P < 0.05; n = 16–17).

Hindlimb blood flow in control and WD-fed mice.

To verify that the ameroid placement induced hindlimb ischemia, and to determine whether flow recovery was affected by diet and/or strain, a separate group of mice with ameroid-induced hindlimb ischemia was used for measurement of hindlimb blood flow using the fluorescent microsphere method. The perfusion of the hindlimb adductor muscles was measured in both hindlimbs of each mouse at 21 dps, and the ratio of blood flow in the occluded and nonoccluded limb was calculated. The flow ratio in C57BL/6J mice fed the CD was 0.80 ± 0.03 (n = 6), indicating good recovery of blood flow to the limb in this group consistent with the reported recovery in this model (61). In WD-fed C57BL/6J, the flow ratio was slightly but nonsignificantly decreased to 0.67 ± 0.06 (n = 6, P = 0.06). It is possible that the effect of the diet in C57BL/6J mice would have reached significance if blood flow had been measured at an earlier time point. The flow ratio was significantly reduced in both ApoE^−/− groups relative to their C57BL/6J counterparts, indicating impaired blood flow recovery in this strain, and was further decreased by the WD (ApoE^−/− CD, 0.65 ± 0.02, n = 6, P < 0.01; ApoE^−/− WD, 0.41 ± 0.06, n = 6, P < 0.05) (Fig. 6). ApoE^−/− mice have previously been shown to have impaired limb blood flow recovery relative to C57BL6/J mice (11).

Fig. 6.

Fluorescent microsphere analysis demonstrated that hindlimb skeletal muscle blood flow recovery following induction of hindlimb ischemia was significantly reduced in ApoE^−/− mice compared with their C57BL/6J counterparts and that the Western diet had a significant effect to reduce blood flow recovery in ApoE^−/− mice (and a similar but nonsignificant effect in C57BL/6J mice).

Effects of individual metabolic factors on PLGF expression in SKMCs in vitro.

To begin unraveling the molecular mechanism(s) by which a WD affects PLGF expression in vivo, we next conducted in vitro studies to individually test the effects of cholesterol, oxidative stress, glucose, and insulin (the factors most strongly correlated with baseline PLGF levels in vivo) on PLGF expression in cultured mSKMCs and hSKMCs. Previous studies in our laboratory have demonstrated that ECs are a major source of PLGF protein (59). We compared PLGF expression in ECs and SKMCs to determine whether SKMCs are also a significant source of PLGF. Although ECs produced more PLGF protein than SKMCs (588.63 ± 76.44 pg/mg total protein vs. 167.27 ± 18.90 pg/mg total protein; P < 0.05), SKMCs were nevertheless a considerable source of PLGF. Given that skeletal muscle is overwhelmingly composed of SKMCs (>80% of all cells), with ECs representing a very small fraction of the tissue (<1% of all cells) (60), we reasoned that most PLGF produced within skeletal muscle likely arises from SKMCs, despite the higher level of expression in ECs. Therefore, we examined the effect of exogenous metabolic factors on PLGF expression in SKMCs. Treatment with 200 μg/mL LDL or 200 μg/mL oxidized LDL reduced PLGF expression in mSKMCs after 24 h of treatment (P < 0.05). However, no significant effect of cholesterol on PLGF protein expression by hSKMCs was detected (Fig. 7). Similarly, mild oxidative stress (treatment with 0.3 mM H₂O₂) reduced PLGF expression by mSKMCs but not hSKMCs (Fig. 8). Glucose treatment (0–20 mM) dose-dependently increased PLGF protein in mSKMCs but slightly decreased it in hSKMCs (Fig. 9, A and C). Insulin slightly stimulated PLGF protein expression in both mouse and human SKMCs at 100 mU/mL but had no effect at higher concentrations (Fig. 9, B and D). Thus, exogenous cholesterol, oxidative stress, glucose, and insulin all failed to consistently reproduce the in vivo effect of the WD to reduce baseline PLGF expression. Similarly, VEGFR1 (PLGF receptor) protein expression was unaffected by the various metabolic treatment conditions (data not shown).

Fig. 7.

Treatment with LDL and oxidized LDL (ox-LDL) reduced placental growth factor (PLGF) protein expression by cultured mouse skeletal muscle cells (mSKMCs) after 24 h (top) but had no effect on human skeletal muscle cells (hSKMCs; bottom).

Fig. 8.

Treatment with 0.3 mM H₂O₂ to mimic oxidative stress reduced placental growth factor (PLGF) protein expression by cultured mouse skeletal muscle cells (mSKMCs) after 24 h (top) but had no effect on human skeletal muscle cells (hSKMCs, bottom).

Fig. 9.

A and C: placental growth factor (PLGF) protein was dose-dependently increased by glucose in mouse skeletal muscle cells (mSKMCs, A) but was slightly decreased by glucose in human skeletal muscle cells (hSKMCs, C). B and D: insulin slightly but significantly increased PLGF protein expression in both mSKMCs (B) and hSKMCs (D) at 100 mU/mL but had no effect at higher concentrations.

Effect of AGE on PLGF expression in SKMCs in vitro; tissue AGE levels.

Since reduced baseline PLGF expression in skeletal muscle was associated with markers of impaired glucose tolerance, such as hyperglycemia and hyperinsulinemia, but treatment of SKMCs with exogenous glucose and insulin did not closely mimic the effect of the WD on PLGF levels, we next tested whether long-term consequences of hyperglycemia such as AGE formation might contribute to the reduction in PLGF we observed in vivo. We first verified that AGE levels were increased by the WD treatment in our animal study by analyzing tissue AGE levels in quadriceps femoris and GPS muscle samples. AGE deposition within skeletal muscle of both strains was significantly elevated by the WD (Fig. 10; P < 0.05). We then treated mSKMCs and hSKMCs with AGE derived from either glucose or glyceraldehyde. Similar to our results with other exogenous metabolites, AGE moderately reduced PLGF in mSKMCs but not in hSKMCs (Fig. 11).

Fig. 10.

Advanced glycation end products (AGEs) are elevated in both quadriceps femoris (A) and gastrocnemius-plantaris-soleus (GPS; B) muscles of C57BL/6J and ApoE^−/− mice fed a Western diet. n = 9–10 per group.

Fig. 11.

Advanced glycation end products (AGEs) derived from glucose did not affect placental growth factor (PLGF) levels in human skeletal muscle cells (hSKMCs; top) but moderately inhibited PLGF protein production in mouse skeletal muscle cells (mSKMCs) after 24 h (bottom).

Potential indirect effect of AGEs to reduce PLGF via TNFα in vitro.

Our in vitro studies were unable to demonstrate direct effects of the WD-associated metabolites tested to reduce PLGF expression in SKMCs (above). A series of parallel studies in ECs (not presented here) were similarly inconclusive. Therefore, we last considered the possibility that one or more metabolites could act indirectly to suppress PLGF expression in either SKMCs or ECs. It is well established that AGE contributes to the vascular complications of diabetes by inducing the expression of inflammatory mediators such as TNFα in macrophages and other cell types (28, 38, 47). Therefore, we hypothesized that this pathway could affect PLGF expression. As an initial test of this hypothesis, we treated RAW 264.7 murine macrophages with AGE. AGE sharply induced TNFα production by macrophages (Fig. 12). We then tested the effect of TNFα on PLGF expression in hSKMCs and hCECs. Exogenous TNFα did not affect PLGF expression in hSKMCs (data not shown) but did significantly inhibit PLGF expression in hCECs (Fig. 12). Further studies are needed to determine whether AGE-induced TNFα production by macrophages contributes to the WD-induced reduction in skeletal muscle PLGF levels that we observed in the present study.

Fig. 12.

Advanced glycation end products (AGEs) derived from glucose sharply upregulated TNFα protein production by RAW 264.7 murine macrophages (top). Exogenous TNFα significantly inhibited placental growth factor (PLGF) protein production by human endothelial cells (hCECs; bottom).

DISCUSSION

Effect of a WD on baseline growth factor expression.

The primary goal of this study was to investigate the effect of metabolic conditions approximating type II diabetes on the expression and regulation of PLGF and related growth factors and receptors in skeletal muscle tissues. One challenge in investigating arteriogenic growth factors in the setting of metabolic disease is that many arteriogenic factors can also be pro-atherogenic (23). We chose to focus on skeletal muscle since its vasculature is mainly a site of arteriogenesis in PAD, not atherosclerosis (which occurs in larger upstream vessels in PAD). Therefore, we believe the responses described here are primarily relevant to arteriogenesis. It is possible that other potentially opposite responses of the factors studied to the diet may be occurring in the macrovasculature or in circulating cells (26).

Long-term feeding of a WD is a physiologically relevant model for human type II diabetes and its associated cardiovascular complications, including PAD (5, 58). Our metabolic data demonstrate that the WD induced different, complex metabolic phenotypes in the two mouse strains examined in this study. C57BL/6J mice fed a WD demonstrated hypercholesterolemia, oxidative stress, hyperglycemia, and hyperinsulinemia. ApoE^−/− mice fed the WD also demonstrated hypercholesterolemia (extreme), oxidative stress (extreme), and hyperinsulinemia; key qualitative differences from the C57BL/6J strain were the additional presence of hypertriglyceridemia and delayed/inconsistent hyperglycemia. Overall, these results are consistent with those reported by others (16, 25, 41, 43, 44), although the lack of impaired glucose tolerance in the ApoE^−/− mice was surprising given other reports (51).

By studying two strains of mice with varying genetic susceptibility to diet-induced metabolic dysfunction, we were able to examine arteriogenic growth factor expression across a spectrum of metabolic phenotypes. In humans, PAD most commonly occurs in older patients with type II diabetes or other metabolic abnormalities; however, most animal studies aimed at characterizing arteriogenic signaling pathways mediating collateral growth in the periphery have utilized young animals with a normal metabolic profile. Therefore, results of the current study are highly applicable to the typical clinical setting of human PAD.

Since PLGF is a key arteriogenic factor, which is one of the earliest to be upregulated in collateral arteries following upstream occlusion (45), we reasoned that dysregulation of its expression might occur in diabetic PAD, consistent with the reduced arteriogenesis found in this setting. Therefore, we tested the hypothesis that a WD would decrease PLGF levels in skeletal muscle in the mouse models described above. Our results confirmed that baseline levels of PLGF protein were indeed decreased in hindlimb skeletal muscle by the WD. PLGF levels were reduced in both the quadriceps and the GPS muscle group by the WD; however, the WD had a greater inhibitory effect on PLGF expression in the quadriceps muscle. This finding agrees well with the fact that arteriogenesis is the predominant type of vascular remodeling that occurs in the quadriceps in response to occlusion, whereas angiogenesis (which is not primarily mediated by PLGF) is the main type of remodeling in skeletal muscle of the distal limb. Of the metabolic variables analyzed, cholesterol, isoprostane, insulin, and glucose were all moderately but significantly negatively correlated with the PLGF protein level in skeletal muscle.

We observed only minor sex-dependent differences in PLGF expression, with decreased PLGF levels in quadriceps of C57BL/6J females compared with C57BL/6J males and decreased levels of PLGF in GPS of ApoE^−/− females versus ApoE^−/− males. Interestingly, these differences were not present in either strain when fed the WD. The significance of this finding remains to be determined.

The results presented here represent the most detailed investigation to date of PLGF expression in skeletal muscle and how it is altered by WD-induced metabolic dysfunction. However, our results agree well with a previous microarray study which found that PLGF gene expression is reduced in ischemic hindlimbs of Lepr^db/db mice (50).

VEGF-A and VEGFR1 were also analyzed for comparison with PLGF. The effect of the WD on the baseline expression of these factors was less consistent than the effect we observed on PLGF. However, some intriguing strain-, sex-, and muscle-dependent differences in response to the WD were observed.

Baseline VEGF-A protein levels were inconsistently affected by the WD. In the GPS muscle group, VEGF-A was decreased by the WD in mice of both strains. However, in the quadriceps muscle, the WD significantly reduced VEGF-A levels in ApoE^−/− mice but increased them in C57BL/6J mice. This strain-dependent difference was quite striking. Other studies have generally reported that diabetes either increases (21) or does not change (10) VEGF-A levels in skeletal muscle. Although the mechanism behind the difference in the VEGF-A response to the WD between the two strains examined in the current study is yet to be identified, our results suggest that the significance of VEGF-A levels across different mouse strains/models of diabetes should be interpreted with caution.

The baseline level of VEGFR1 expression in skeletal muscle was also inconsistently affected by the WD, with VEGFR1 levels increased by the WD in quadriceps muscle from some groups but decreased by the diet in the GPS muscle from some groups. The effect of diabetes on VEGFR1 expression in skeletal muscle has not been extensively studied, but one group has reported reduced VEGFR1 levels in diabetic mouse tibialis anterior muscle (21). We anticipated that there might be differences in the effect of the diet on PLGF since it has a major role in placentation in females. However, of the factors that we studied, VEGFR1 showed the clearest sex-dependent response to the WD, which increased VEGFR1 in quadriceps of female but not male mice of both strains. This WD-induced increase in VEGFR1 is potentially of interest with regards to preeclampsia as well as arteriogenesis. However, the mechanism leading to increased VEGFR1 in skeletal muscle of female mice fed a WD, and its possible significance with regards to preeclampsia, remains to be determined. Neither VEGF-A nor VEGFR1 showed a consistent association across strains and muscle types with the metabolic variables analyzed.

Effect of a WD on upregulation of growth factors by hindlimb ischemia.

Elevated shear stress is a primary stimulus for arteriogenesis. Although shear stress-dependent regulation of many angiogenic/arteriogenic factors including VEGF (13), VEGFR2 (1), PDGF (12, 37), and FGF (36) has been studied, our group was the first to provide conclusive evidence that PLGF expression is influenced by shear stress. We demonstrated that PLGF is upregulated in both cultured ECs and in isolated arterioles by exposure to high shear stress in vitro (46). In addition, we previously found that PLGF mRNA expression increases sharply in rat hindlimb collaterals immediately following induction of experimental PAD, in agreement with a key role for elevated shear stress in controlling PLGF expression (34). Similar findings have been reported by others (33). Therefore, the next question we asked is whether upregulation of PLGF and other arteriogenic factors by the physiological stimulus of vascular occlusion would also be affected by the WD. We chose the ameroid occlusion model of hindlimb ischemia for these studies since the gradual occlusion produced by this technique more closely mimics human PAD than does acute ligation of the femoral artery (61). We assessed the growth factor response in the thigh region because this is the primary site at which arteriogenesis occurs in hindlimb ischemia. The thigh region contains preexisting collateral roots in the medial thigh, quadriceps femoris, and biceps femoris muscles (32). Chronic occlusion of the femoral artery alters the pressure gradient across these collateral roots, leading to an increase in shear stress through these small vessels. Existing evidence suggests that increased shear stress is a key trigger for these small collaterals to undergo arteriogenesis (19).

We observed a sharp elevation in PLGF expression at 7–10 dps in quadriceps muscle of C57BL/6J mice fed the CD. This time course is consistent with a shear stress-mediated response since the ameroid constrictors that we used to occlude the femoral artery are known to close and thereby create maximum shear stress in collaterals around ~7 dps (61). Furthermore, in contrast to acute femoral artery occlusion, the ameroid model does not increase hypoxia-inducible gene expression in the thigh region; however, it does increase the expression of shear stress-regulated genes, including endothelial nitric oxide synthase (61). The response of PLGF to occlusion was both delayed and decreased in C57BL/6J mice fed the WD, demonstrating that both baseline and stimulated expression of PLGF are impaired by the diet. The decreased/delayed PLGF expression is consistent with the decreased/delayed arteriogenesis reported in numerous models of diabetes, as well as with reduced PLGF-induced macrophage infiltration (19). Even more strikingly, the response of PLGF to occlusion was completely absent in ApoE^−/− mice fed either diet. These results suggest that the degree of inhibition of occlusion-induced PLGF increases as the severity of the metabolic phenotype increases. Since ApoE^−/− mice had a distinct metabolic phenotype with only limited dysregulation of glucose metabolism, this result also suggests that multiple factors are likely responsible for this effect.

VEGF-A and VEGFR1 each had a distinct pattern of response to the occlusion in C57BL/6J mice, with VEGF-A peaking at ~14 dps and VEGFR1 continuing to increase until at least 21 dps. Again, both responses were abolished by the WD in C57BL/6J mice and were largely absent in ApoE^−/− mice fed either diet.

MCP-1 expression also had a unique time course, peaking at 3 dps. This immediate upregulation of MCP-1 is consistent with our previous results from rat hindlimb skeletal muscle (34). Somewhat surprisingly, this time course suggests that MCP-1 is upregulated before maximal shear stress occurring in the collaterals (since the ameroid does not fully close until ~7 dps). Most interestingly, MCP-1 was not affected by the WD in C57BL/6J mice. However, MCP-1 upregulation in response to occlusion was largely absent in ApoE^−/− mice fed either diet.

Correlation analyses showed that all four factors were negatively associated with isoprostane levels. PLGF and VEGF-A also displayed a negative association with markers of impaired glucose metabolism, whereas MCP-1 was negatively associated with both cholesterol and triglycerides (which were not associated with impaired upregulation of the other factors studied). This difference suggests that the extreme cholesterol levels in the ApoE^−/− mice may contribute to the lack of MCP-1 upregulation by occlusion. Overall, comparison of the correlation results from the baseline study with those from the occlusion study indicates that baseline expression and occlusion-induced upregulation of PLGF and the other factors studied could potentially be independently affected by different aspects of the metabolic phenotype.

In vitro studies on metabolic variables and growth factor expression.

To begin defining the mechanism(s) behind the WD-induced inhibition of baseline PLGF expression, we tested the effects of individual metabolic factors on PLGF expression in both mSKMCs and hSKMCs in vitro. Cholesterol and oxidative stress inhibited PLGF expression in mSKMCs, which is consistent with the results from mouse skeletal muscle; however, these findings were not replicated in hSKMCs. This discrepancy could reflect species differences in regulation of PLGF. It is well known that findings from mouse models do not always translate to human disease. Therefore, it is possible that cholesterol influences PLGF in mSKMCs and tissue but not in hSKMCs and tissue.

Treatment with exogenous glucose or insulin also failed to consistently replicate our in vivo results. Since the WD-fed mice were exposed to sustained hyperglycemia in our long-term in vivo studies, we next considered the possibility that AGEs could influence PLGF expression. It has been established in the literature that AGE interferes with shear stress-induced arteriogenesis. Freidja et al. (15) reported that arterial enlargement in response to chronically elevated shear stress in mesenteric arteries was impaired in Zucker diabetic fatty rats and that treatment with the AGE-breaking compound ALT-711 was able to rescue this response. Another study demonstrated that reduced blood flow in ischemic hindlimbs of streptozotocin-induced diabetic mice could be rescued by inhibiting AGE formation (55). Similarly, impaired arteriogenesis in ischemic hindlimbs of streptozotocin-induced diabetic rats can be partially rescued by overexpression of glyoxalase-1 (an AGE-detoxifying enzyme) (4), and knockout of the receptor for AGE (RAGE) removes the inhibitory effect of streptozotocin-induced diabetes on flow recovery in mouse ischemic hindlimb (20). Lastly, a recent mRNA profiling study of 84 angiogenesis-related genes in RAGE knockout mice with streptozotocin-induced diabetes suggested that PLGF is upregulated by femoral ligation to a greater extent in the absence of RAGE (35). Altogether, these reports are consistent with the possible existence of an AGE/RAGE-mediated mechanism that suppresses baseline and/or shear stress-induced PLGF expression in diabetes.

Since AGEs were not measured in our original metabolic phenotyping study of plasma, we performed a post hoc analysis of AGE levels in skeletal muscle tissue from control and WD-fed mice. AGE levels were significantly increased in skeletal muscle of WD-fed mice from both the C57BL/6J and ApoE^−/− strains, supporting the hypothesis that elevated AGE levels could be inhibitory for PLGF expression in skeletal muscle. However, similar to the other compounds tested in vitro, AGE treatment did not consistently decrease PLGF expression in SKMCs.

Since none of the metabolic factors tested in vitro were able to directly and consistently replicate the suppression of PLGF produced by the WD in vivo, we concluded that it is more likely that one or more of these compounds acts indirectly to influence PLGF expression. Studies by other groups have established a clear link between AGE, TNFα, oxidative stress, and endothelial dysfunction in diabetes, as discussed below; therefore, we reasoned that this pathway could also contribute to diabetes-related impaired PLGF expression and arteriogenesis.

Stimulation of TNFα expression by AGE has been reported in a variety of cell types, including ECs, vascular smooth muscle cells, and macrophages (28, 38, 47). Gao et al. (17, 18) demonstrated that AGE/RAGE signaling induces TNFα expression and oxidative stress in Lepr(db) mice. TNFα knockout impairs arteriogenesis following acute femoral artery occlusion in mice (24), suggesting that a certain level of TNFα is required. However, chronic elevation of TNFα appears to exert an inhibitory effect on arteriogenesis (63), and endothelial function was improved by inhibition of TNFα in the mouse studies noted above (17, 18).

We found that AGE treatment significantly increased TNFα levels in RAW 264.7 murine macrophages, as anticipated. Interestingly, AGE did not increase TNFα expression in either hSKMCs or hCECs. Others have previously reported that AGE increases TNFα in ECs; however, in that study, ECs were activated with IL-1α, LPS, and IFNγ before AGE treatment. When not activated in this manner, TNFα was undetectable and was not induced by AGE (47). Since diabetes is a chronic inflammatory state, it is possible that responses of ECs and/or SKMCs to AGE or other metabolic factors did contribute to the WD-mediated effect on PLGF we observed in vivo, although the quiescent cells we studied in vitro did not. This possibility remains to be explored in future studies.

Treatment of hCECs with TNFα significantly reduced PLGF protein levels, consistent with a potential role for AGE/TNFα in mediating reduced PLGF expression. The relationship of TNFα and PLGF has been only peripherally investigated. Interestingly, TNFα levels in plasma of women with preeclampsia are elevated, whereas those of PLGF are reduced (27). Moreover, a recent study of BeWo human choriocarcinoma cells demonstrated that low levels of TNFα stimulate PLGF expression, whereas higher concentrations are inhibitory (56); these findings agree well with the idea that TNF is pro-arteriogenic when acutely elevated but inhibitory for arteriogenesis when chronically elevated. Lastly, PLGF has been shown to induce TNFα production in monocytic cells in a variety of different settings (14, 39, 52, 62). This effect is consistent with the pro-arteriogenic action of PLGF. It also suggests the possibility of a negative feedback loop by which TNFα may act to inhibit PLGF under conditions of chronically elevated TNFα; again, this possibility remains to be investigated.

In conclusion, this study examined the effect of a WD on baseline levels and hindlimb ischemia-induced expression of PLGF, VEGF-A, VEGFR1, and MCP-1 in skeletal muscle. Overall, our results showed a consistent effect of the diet to decrease baseline PLGF levels, in contrast to inconsistent effects on VEGF-A and VEGFR1. In contrast, the WD strongly affected hindlimb ischemia-induced expression of all the growth factors tested. These data suggest that there is a complex and possibly separate interaction of metabolic factors with signaling pathways mediating baseline and stimulated growth factor expression in skeletal muscle. In vitro investigation identified AGE as a potential indirect mediator of the reduced PLGF levels we observed in vivo via its effects to stimulate TNFα production in macrophages (and perhaps activated SKMCs/ECs). Although this potential mechanism is intriguing, further studies are necessary to determine whether it contributes to WD-mediated suppression of PLGF expression and to identify the signaling mechanism(s) involved.

Therapies aimed at normalizing arteriogenic growth factor expression in skeletal muscle could potentially rescue impaired peripheral arteriogenesis in patients with type II diabetes. Thus, our results have important clinical implications.

GRANTS

This work was supported by the National Institutes of Health Grants R01-HL-084494 and R56-HL-084494 (to P. C. Lovern).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

P.C.L. conceived and designed research; A.T.S., F.R., and O.A.S. performed experiments; A.T.S., F.R., O.A.S., M.E.P., and P.C.L. analyzed data; A.T.S., F.R., O.A.S., and P.C.L. interpreted results of experiments; A.T.S., F.R., O.A.S., and P.C.L. prepared figures; A.T.S. and P.C.L. drafted manuscript; A.T.S., F.R., O.A.S., M.E.P., and P.C.L. edited and revised manuscript; A.T.S., F.R., O.A.S., M.E.P., and P.C.L. approved final version of manuscript.