Various angiogenic growth factors, such as the fibroblast growth factors FGF-2 (1) and FGF-1 (2), vascular endothelial growth factor isoforms (VEGF)₁₆₅(3) and VEGF₁₂₁(4) are currently being evaluated in clinical trials as potential therapeutic agents in patients with chronic myocardial ischemia (see Ware and Simons for review 5). In prior studies, we and others have shown that continuous periadventitial (6), atrial (7), intracoronary (IC) (8- 9) or single-bolus intracoronary (9- 10) administration of VEGF₁₆₅ in animal models of chronic myocardial ischemia improves regional blood flow and regional left ventricular (LV) wall motion in the ischemic territory. However, recently completed Phase I clinical trials of intracoronary and intravenous (IV) VEGF infusions demonstrated severe hypotension at relatively low doses, and a Phase II trial employing a combination of intracoronary and intravenous VEGF delivery failed to show any benefit with regard to myocardial function and perfusion at the doses used (3). The negative results of this study raise a number of questions with regard to VEGF₁₆₅ efficacy as a therapeutic angiogenic agent and the suitability of various modes of its delivery.

Systemic hypotension is the major acute side effect of VEGF₁₆₅ administration (10- 11). Hypotension is primarily mediated by VEGF-induced release of nitric oxide (NO) (10- 12) and therefore might be circumvented effectively by pretreatment with an NO synthase inhibitor. On the other hand, NO is essential for VEGF-induced angiogenesis, and nitric oxide synthase (NOS) blockade may inhibit angiogenic effects of VEGF therapy (13- 15). Therefore, the present study was designed to determine angiogenic efficacy of intravenous or intracoronary VEGF delivery and to explore the effect of pretreatment with an NO synthase inhibitor, L-NAME, on hemodynamic and angiogenic effects of VEGF therapy.

Forty-two 30 to 35 lb male Yorkshire pigs (Parsons Farm, Massachusetts) were selected for the study. The study consisted of three phases as previously described (6,9,16- 17). For all procedures the animals were anesthetized with ketamine 20 mg/kg and sodium thiopenthal 10 mg/kg IV, intubated, mechanically ventilated and further anesthetized with 1.5% to 2.5% isoflurane in room air. Postoperatively, all animals received antibiotics for 48 h and narcotic analgesics (buprenorphine 0.02 mg/kg i.m.) for 24 h and, after that, oral ketaprophen b.i.d. as needed. All animals were cared for according to the National Institute of Health Guidelines for the Care and Use of Laboratory Animals, and the protocol was approved by the BIDMC Institutional Animal Care Committee.

In the initial phase, an ameroid occluder (I.D. 2 to 2.5 mm, Research Instruments SW, Escondido, California) was placed on the proximal left circumflex coronary artery (LCX). Three weeks later, the animals were brought back for cardiac catheterization using a standard femoral access. A second 8F sheath (Avanti, Cordis, Miami, Florida) was introduced in the right femoral vein for atrial pacing. Heparin 100 U/kg was used for systemic anticoagulation. Intra-arterial pressure and electrocardiogram were continuously monitored and recorded. Selective angiography of the left and right coronary arteries was performed to confirm closure of the ameroid and to assess baseline angiographic parameters. Baseline regional blood flow was measured by intra-atrial colored microspheres. Two-dimensional echocardiographic recording was obtained at rest and after 2 min of atrial pacing at 170 beats/min to quantify global and regional cardiac function. Following these studies, the animals were randomly assigned to one of the following five groups (n = 8–9 for each group): 1) controls (saline infusion with IV 40 U/kg heparin bolus), 2) IV rhVEGF₁₆₅ infusion 0.25 μg/kg/min, 3) IV rhVEGF₁₆₅ infusion 0.05 μg/kg/min for 200 min, 4) IC rhVEGF₁₆₅ infusion 0.25 μg/kg/min and 5) IC rhVEGF₁₆₅ infusion 0.25 μg/kg/min, after 10 mg/kg IV bolus of L-NAME (Sigma, St Louis, Missouri). Except for group 3, infusions lasted for 40 min. Hemodynamic parameters were recorded for 40 min following the completion of infusion.

Three weeks after initiation of therapy, the animals were brought back for final evaluation with repeat coronary angiography, myocardial blood flow determination and rest and pacing stress echocardiography. After these studies, animals were put to death, and the hearts were removed for microsphere analyses.

Selective right coronary and left coronary angiography was performed using a 7-FR JR4 diagnostic angiography catheter (Cordis, Miami, Florida) and Renografin (Squibb Diagnostics, Princeton, New Jersey) contrast and were recorded using cine angiography (General Electric, LU). The films were read in a blinded manner by two experienced angiography readers. Discordant readings were adjudicated by consensus (10% of cases). Collateral index was rated using a semiquantitative approach as previously described (18): 0 = no visible collaterals, 1 = faint filling of side branches of the main epicardial vessel without filling the main vessel, 2 = partial filling of the main epicardial vessel, 3 = complete filling of the main vessel.

Colored microspheres (Dye Trac, Triton Technologies, San Diego, California) were used for blood flow determinations. Microspheres were injected through a 7F JL4 diagnostic angiography catheter that was advanced in a retrograde manner into the left atrium, with its position confirmed by recording of an atrial pressure wave.

At midstudy, 3 ×10⁶ blue microspheres and at final study 5 × 10⁶ yellow (rest regional blood flow) and red (stress flow) microspheres were injected over 30 s while a reference blood sample was withdrawn from the femoral artery at 4 ml/min for 2 min. At the end of the study, the heart was cut in five or six cross sections. From these cross sections, photographic images were taken and digitized, and infarct areas were morphometrically measured using Optimas 6.0 software. A 1-cm-thick cross section of the LV was taken at mid-papillary level for microsphere analysis and was divided in eight radial segments of equal size. Three segments representing the LCX territory were further subdivided in endocardial and epicardial sections. The tissue samples and the reference blood samples were processed, and myocardial blood flow was calculated as previously described (9).

Transthoracic echocardiograms were performed in all animals at rest and after 2 min of right atrial pacing stress at 170 beats/min (7F bipolar pacing electrode NIBH, CR BARD, Billerica, Massachusetts). Echocardiograms were performed with a Hewlett-Packard Sonos 5500 (Hewlett-Packard, Andover, Massachusetts) sector scanner equipped with a 4 MHz phased-array transducer. Two-dimensionally guided M-mode tracings were recorded. Anterior and posterior wall thickness and LV internal dimensions were measured in standard fashion on at least three consecutive beats and averaged using an off-line analysis system (Cardiac Workstation, Freeland Systems, Louisville, Colorado) by one observer blind to study group. Regional function was assessed by systolic wall thickening in the posterior (LCX territory) and anterior (left anterior descending [LAD] territory) walls calculated as [%Th = (systolic thickness−diastolic thickness)/diastolic thickness × 100]. %Th in LCX territory at each time point was normalized to that in the LAD territory, and changes over time were compared. Global function was assessed using fractional shortening calculated as [FS = (diastolic LV diameter−systolic LV diameter)/diastolic diameter × 100].

Microvascular studies on vessels were prepared from the epicardial myocardium of the LCX- and LAD-dependent territories (80 to 170 μm in internal diameter and 1 to 2 mm in length) as previously described (19). Responses to endothelium-dependent receptor-mediated vasodilatory substances (ADP, 1 nm to 10 μm) and endothelium-independent agent sodium nitroprusside were tested after potassium chloride preconstriction.

The total rhVEGF₁₆₅ (Genentech, Inc, San Francisco, California) dose delivered by IC infusion (pump: model 901, Harvard Instruments, Dover, Massachusetts) was equally divided between the RCA and the proximal LCX arteries (20 min in each artery) using a 3F infusion catheter (Cordis Corp.). In those animals in which the proximal LCX was too short to stabilize the infusion catheter, rhVEGF₁₆₅ was infused into the proximal part of the LAD. IV rhVEGF₁₆₅ was infused through the ear vein. In the L-NAME group, 10 mg/kg L-NAME IV (Sigma, St Louis, Missouri) in saline was injected 5 min before rhVEGF₁₆₅.

All data are given as mean ± SEM unless indicated otherwise. As this study was not powered for direct between group comparisons, paired t tests on continuous variables measured at mid- and final study were performed for each experimental group. For baseline variable comparison, analysis of variance (ANOVA) (SPSS 7.5) with a Bonferroni correction was applied. In addition, in the case of multiple measurements per animal over time, such as the hemodynamic parameters, a multivariate analysis of variance (MANOVA) with repeated measures and with treatment as the second independent variable was used. Ordinal data, such as collateral index scale, were analyzed using Kruskal Wallis and by the Wilcoxon test for differences within the group over time (i.e., final study vs. midstudy). A p value <0.05 was considered statistically significant.

Fifty-six animals underwent ameroid surgery, 14 (25%) died before randomization for treatment. Forty-two animals (control n = 9, IV VEGF n = 8, IC VEGF n = 8, IC VEGF + L-NAME n = 9, IV VEGF slow infusion n = 8) completed the study. The LCX was occluded at mid-study in all animals.

Both IV and IC VEGF infusions caused a transient drop in mean arterial blood pressure (MAP) that peaked at 10 to 15 min after the start of infusion and ranged from 15.3 ± 4.0 mm Hg in the 0.25 μg/kg/min IC group to 22.3 ± 3.0 in the 0.05 μg/kg/min IV group (26% reduction in MAP) (Table le1). Thus, reducing the rate of infusion to 0.05 μg/kg/min did not prevent hypotension. Pretreatment with L-NAME, however, effectively inhibited the decrease in MAP without significantly augmenting the initial (pre-L-NAME) blood pressure. In all groups, MAP returned to pretreatment values in approximately 20 min after the end of the infusion (data not shown). These changes in blood pressure were not accompanied by any significant changes in the heart rate.

Complete angiography data were missing from two animals. The collateral index at baseline was similar for all groups (Kruskal Wallis, p = 0.897). All groups demonstrated a trend towards increased collateralization as expected in this model. However, a significant increase in left-to-left collateralization was observed only in both IC groups (Figure 1). At the same time, no right-to-left (RCA → LCX) collateralization remained unchanged (not shown). Thus, administration of VEGF into the left and right coronary system resulted in a directional increase in collateral circulation (left-to-left but not right-to-left collaterals), which was not inhibited by a single administration of L-NAME.

Regional blood flow to the LCX and LAD territories at rest before the start of VEGF infusion, and at rest and during stress three weeks after treatment, was measured by colored microspheres. Baseline flow to the normal part of the heart was similar in the treatment groups (ANOVA, p = 0.210) and did not change significantly over time, with p-values ranging from 0.10 for the IC VEGF group to 0.67 for the IV VEGF 0.05 μg/kg/min group (Table le2). Rest flow increased significantly in the IC VEGF group ((Figure 2), paired t test, p = 0.006) and showed an upward trend in the IC VEGF + L-NAME group (p = 0.122). Stress flow during the final study was higher in the IC VEGF/L-NAME group (p = 0.003) than in controls and showed a trend in the IC VEGF group (p = 0.06). The IV groups were no different from controls with respect to change in rest flow over time or to flow at stress.

To assess the effect of VEGF therapy on global and regional LV function, M-mode and two-dimensional echo recording were obtained at rest and during pacing stress at mid-study and at the final study. Left ventricular fractional shortening (FS) did not change in any of the VEGF treated groups (Figure 3). In contrast, in the control group the LV FS decreased significantly (paired t test, p = 0.002). At stress, the FS showed an upward trend in the IC VEGF + L-NAME group only (paired t test, p = 0.067). No differences were found in FS over time at rest (MANOVA with repeated measures, p = 0.623 for the time effect and 0.640 for the time by treatment effect) or during stress (p = 0.243 for the time effect and 0.149 for time by treatment).

Regional ventricular function was measured as percent systolic wall thickening in the ischemic (LCX, target) area normalized for the septal (normal) area (TWT/NWT). At baseline at rest, TWT/NWT ranged from 0.25 ± 0.09 in the IV VEGF 0.25 μg/kg/min group to 0.43 ± 0.08 in the VEGF+L-NAME group (ANOVA, p = 0.614). During pacing, TWT/NWT ranged from 0.04 ± 0.11 in the control group to 0.33 ± 0.10 in the IC VEGF group (ANOVA, p = 0.282). No significant improvement was noted from mid-study to final study in any group at rest or stress (Figure 4). Using MANOVA with repeated measures, no differences in regional LV function were found between the treatment groups at rest (p = 0.955 and p = 0.853 for time effect and time by treatment effect respectively) or during stress (p = 0.165 and p = 0.891 for time effect and time by treatment effect).

As expected, ischemia in the LCX region impaired the response of microvessels to the endothelium-dependent agent ADP compared with microvessels in the normally perfused LAD myocardium. Intracoronary infusion of VEGF fully restored the responses of the LCX region microvasculature to ADP (Figure 5). This salutary effect of VEGF administration was partially blocked by the pretreatment with L-NAME at the time of VEGF infusion. Intravenous administration of VEGF failed to restore microvascular function in the LCX myocardium (data not shown). No differences were observed in the endothelium-independent responses of the microvasculature.

Vascular endothelial growth factor is a potent angiogenic agent that might be clinically useful in the management of syndromes of myocardial ischemia. VEGF₁₆₅ promotes angiogenesis in both cell culture models (20) and in vivo models including the heart (8- 10,21). Furthermore, VEGF₁₆₅ plays a key role in embryonic vasculogenesis because disruption of even a single copy of the gene is sufficient to induce early lethality (22- 23). The mechanism of VEGF-induced angiogenic activity is not clear but is thought to involve stimulation of vessel permeability (24), release of NO (13), promotion of monocyte/macrophage chemotaxis (25) and induction of expression of a number of angiogenesis-promoting proteases (26).

This multitude of actions might contribute to its therapeutic effect but also harbors a potential for dose-limiting side effects. For example, in early preclinical and clinical trials, VEGF₁₆₅ induced an NO-mediated hypotension. Therefore, the present study was undertaken to investigate how, in the setting of chronic myocardial ischemia, therapeutic and toxic effects are affected by the mode of delivery and by inhibition of NO synthase activity. Single intracoronary or intravenous injection of VEGF would offer relative ease of use and appeal to a broader segment of the patient population than the currently used intramyocardial injection during bypass surgery. In the case of VEGF₁₆₅, a bolus intracoronary administration (10) or local intracoronary delivery (9) in a porcine model resulted in a significant improvement in myocardial perfusion and function but was associated with profound and sometimes lethal hypotension (10). Therefore, we have chosen to administer VEGF₁₆₅ by either intracoronary infusion or intravenous infusion over 40 min. To assess the rate of infusion on hemodynamics and angiogenic potency, another group of animals was given the same total amount of VEGF₁₆₅ over 200 min by an intravenous infusion.

Because VEGF₁₆₅ hypotension is predominantly NO-mediated, we have also explored the utility of short-term NO synthase inhibition with L-NAME. L-NAME is an irreversible inhibitor capable of inactivating all three NO synthases (nNOS, eNOS and iNOS) and was shown to eliminate ∼80% of VEGF-induced vasodilation (11,27). The effect of single L-NAME administration on the angiogenic efficacy of VEGF₁₆₅ has not been studied.

We chose a porcine ameroid model for this study because of its several unique benefits. First, the application of an occluder results in reproducible and gradual occlusion of the coronary artery, with reduced blood flow to the LCX territory. This reduction in coronary flow leads to a significant reduction in regional myocardial function and an equally significant reduction in microvascular responsiveness to endothelium-dependent vasodilators such as ADP. Furthermore, these changes occur in the setting of minimal (typically <5%) myocardial necrosis and remain stable over time. The latter attribute is especially important for longitudinal testing of angiogenic agents. This study was not powered to permit between-group comparisons, and we therefore cannot directly compare treatment groups. However, the chosen analysis allows us to support statements on the efficacy of each delivery method separately.

In the present study, regional myocardial blood flow as measured by the colored microsphere method and angiography was consistently improved by intracoronary VEGF₁₆₅. Moreover, pretreatment with a single dose of L-NAME sufficient to inhibit VEGF-induced systemic hypotension did not inhibit the ability of VEGF₁₆₅ to enhance regional blood flow at stress in this model. N_ω-nitro-L-arginine methyl ester hydrochloride pretreatment, however, did prevent VEGF-mediated restoration of endothelium-dependent microvascular reactivity, possibly because of the irreversible nature of NOS inhibition by L-NAME and the considerable half-life of L-NOARG, the active metabolite of L-NAME (28).

Although half of the intracoronary VEGF dose was given into the right coronary artery, right-to-left collateralization did not improve, which is consistent with previous experiments in this model (6,9,11,29). The development of right-to-left collaterals is probably dependent on anatomic variations and is not treatment related.

Despite the improvement in myocardial perfusion, evaluation of regional LV function demonstrated continued abnormality (i.e., low target over normal wall thickening index) in both IC groups. Alternatively, this absence of functional improvement could suggest either that restoration of the coronary flow was not complete or that the echocardiographic assessment is less sensitive and/or accurate than the magnetic resonance imaging employed in our prior study of intracoronary VEGF₁₆₅(9). Because the same model was used in both of our studies with intracoronary VEGF, we suggest that echocardiographic assessment of ventricular function in our hands might not have been sensitive enough to pick up modest changes in ventricular function.

Intravenous delivery of VEGF₁₆₅, whether given as a 40 or 200 min infusion, was not effective with regard to most end points studied. Although there was a statistically significant increase in the coronary collateral index, the absence of any improvement in microsphere-measured myocardial blood flow, regional LV function or microvascular function strongly argues against effectiveness of this mode of delivery. This result is similar to findings in the study of Lazarous et al. (7) in which daily intra-atrial injections of 0.72 mg rhVEGF₁₆₅ from 10 to 16 days or 17 to 23 days after ameroid placement were also ineffective.

Both IC and IV infusions induced equally profound systemic hypotension that was not reduced by decreasing the rate of infusion. At the same time, L-NAME effectively abolished the hypotensive effect of VEGF₁₆₅. Hypotension is a serious acute side effect of systemic or intracoronary VEGF₁₆₅ administration (10- 12,27), and in a number of in vivo studies, this effect was shown to be mediated by nitric oxide. VEGF₁₆₅ induces acute NO release, and chronic VEGF₁₆₅ administration leads to induction of both eNOS (30) and iNOS (31) expression. The currently reported successful use of a single L-NAME dose prior to VEGF₁₆₅ infusion to inhibit VEGF-induced hypotension while retaining its angiogenic effect may therefore be of clinical importance as the total VEGF₁₆₅ dose that can be administered is no longer limited by this side effect.

The differences between this study and the recent Phase II clinical trial (3) might offer new insight into the pharmacokinetics and the role of the mode of VEGF₁₆₅ delivery in the coronary circulation. To avoid hypotension, the dose in the clinical trial was limited to approximately 80 μg VEGF₁₆₅ given intracoronary as opposed to 400 μg in our study. In the clinical study, intravenous VEGF₁₆₅ dosages were given 3, 6 and 9 days later, but our results show no effect of intravenously delivered VEGF₁₆₅, so they probably did not contribute to the overall effective VEGF₁₆₅ dose. As intravenous delivery is ineffective, the efficacy of intracoronary delivery probably depends on the first-pass effect, the magnitude of which is unknown for VEGF₁₆₅ in the coronary circulation. It is therefore also important that in the clinical trial, rhVEGF₁₆₅ was delivered in a relatively small part of the coronary system and not directly to the occluded regions, whereas in our study, the entire coronary circulation was covered, and half of the dose was directed at the occluded artery. Thus, it is possible that intracoronary rhVEGF₁₆₅, when given at a sufficiently high dose and directed to the target areas including the occlusions, is effective as a therapeutic angiogenic growth factor in myocardial ischemia. The hypotensive side effects of such a high dose rhVEGF₁₆₅ can be inhibited by a single L-NAME dose.

Single intracoronary infusion of 0.25 μg/kg/min of rhVEGF₁₆₅ results in a significant improvement in collateralization of chronically ischemic myocardium and concomitant increase in myocardial blood flow. N_ω-nitro-L-arginine methyl ester hydrochloride at a dose that is sufficiently high to block VEGF-induced hypotension most likely does not affect VEGF-induced angiogenesis. Intracoronary VEGF 10 μg/kg did not improve LV function. Single intravenous infusion of the same amount of rhVEGF₁₆₅ was ineffective with regard to restoration of coronary perfusion or myocardial function.