Long-term increase in uterine blood flow is achieved by local overexpression of VEGF-A165 in the uterine arteries of pregnant sheep

Increasing uterine artery blood flow (UABF) may benefit fetal growth restriction where impaired uteroplacental perfusion prevails. Based on previous short-term results, we examined the long-term effects of adenovirus vector-mediated overexpression of vascular endothelial growth factor-A165 (VEGF-A165) in the uterine artery (UtA). Transit-time flow probes were implanted around both UtAs of mid-gestation pregnant sheep (n=11) to measure UABF. A carotid artery catheter was inserted to measure maternal or fetal hemodynamics. Baseline UABF was measured over 3 days, before injection of adenovirus vector (5 × 1011 particles) encoding the VEGF-A165 gene (Ad.VEGF-A165) into one UtA and a reporter β-galactosidase gene (Ad.LacZ) contralaterally. UABF was then measured daily until term. At 4 weeks post injection, the increase in UABF was significantly higher in Ad.VEGF-A165 compared with Ad.LacZ-transduced UtAs (36.53% vs 20.08%, P=0.02). There was no significant effect on maternal and fetal blood pressure. Organ bath studies showed significantly lesser vasoconstriction (Emax 154.1 vs 184.7, P<0.001), whereas immunohistochemistry demonstrated a significantly increased number of adventitial blood vessels (140 vs 91, n=26, P<0.05) following Ad.VEGF-A165 transduction. Local overexpression of VEGF-A165 in the UtAs of pregnant mid-gestation sheep leads to a sustained long-term increase in UABF, which may be explained by neovascularization and altered vascular reactivity.


INTRODUCTION
During human pregnancy, uterine perfusion increases dramatically from 50 ml min --1 in week 10 to as much as 1300 ml min --1 at term 1 as a result of increased maternal cardiac output and trophoblastdriven modification of the uterine spiral arteries. Failure of this normal physiological process leads to fetal growth restriction (FGR) and pre-eclampsia (PET), two of the most challenging obstetric complications. FGR affects up to 8% of all pregnancies and is severe in 1:500 cases. It is associated with stillbirth, longterm neurological impairment and adult-onset cardiovascular disease. 2 There is no effective treatment. In FGR and PET, there is decreased depth and density of trophoblast invasion, 3,4 and myometrial small arteries show increased vasoconstriction and decreased endothelium-dependent vasodilatation. 5,6 Vascular endothelial growth factor-A (VEGF-A) is essential for angiogenesis during development. 7 VEGF gene or protein therapy may prove therapeutically useful in local perivascular applications by stimulating vascular protective endothelial functions. 8 VEGF vasodilates and increases blood flow in diverse vascular beds, 9,10 the effects mediated partly through its stimulation of endothelial production of nitric oxide 11 and prostacyclin. 12 The fall in uteroplacental resistance in normal pregnancy is mediated by interstitial extravillous trophoblast secretion of angiogenic and vasodilatory factors such as VEGF, to promote local blood flow to the uterus. 13,14 The invading cytotrophoblasts secrete VEGF to regulate their acquisition of an endothelial-like phenotype that allows them to replace the maternal cells that line the uterine vessels. These cells are also dependent on VEGF for their maintenance and growth. 13 In established FGR, serum levels of VEGF-A 165 are significantly diminished. 15 In PET, placental-derived sFlt-1 (soluble fms-like tyrosine kinase-1), a soluble receptor of VEGF, is upregulated, lowering circulating concentrations of free VEGF and causing endothelial dysfunction. 16 Systemic treatment of FGR resulting from uteroplacental insufficiency has given disappointing results. Sildenafil citrate (Viagra), a phosphodiesterase-5 inhibitor known to improve blood flow via cyclic guanosine monophosphate (cGMP)-mediated endothelial relaxation of blood vessels, enhanced vasodilatation of myometrial small arteries from women with FGR. 6 Oral or subcutaneous sildenafil citrate in randomized controlled trials in human FGR and PET have shown no maternal or neonatal benefit, 17 although a recent small nonrandomized study did show significantly increased abdominal circumference in severe early-onset FGR 18 but no benefit on fetal outcome. It is concerning that in pregnant ewes carrying growth-restricted fetuses, intravenous infusion of sildenafil citrate decreased uterine blood flow causing fetal hypoxia, probably because of a systemic reduction in vascular resistance and consequent blood flow 'steal' from the uteroplacental circulation to the systemic vascular circuit. 19 Similar disappointing results were obtained with nitric oxide donors such as L-arginine in randomized double-blinded control trials designed to test their efficacy at ameliorating FGR. 20 VEGF-A 121 gene or protein therapy has been found to be beneficial and alleviated some of the adverse symptoms, including elevated blood pressure and proteinurea, in murine models of PET. 21,22 Placental gene transfer of an adenovirus encoding insulin-like growth factor-1 ameliorates FGR in rats, 23 and intra-amniotic insulin-like growth factor-1 infusion improves fetal growth in FGR sheep. 24 These findings suggest that therapy aimed locally at the uteroplacental site may be needed for effective treatment of severe early-onset FGR. To date, no method of increasing uteroplacental blood flow has been developed. We hypothesized that overexpression of VEGF-A 165 in the uterine arteries (UtAs) would lead to a sustained increase in uterine artery blood flow (UABF) by beneficially altering vascular reactivity and enhancing angiogenesis. If so, this might improve growth velocity in severe FGR and lead to a novel therapeutic approach for this condition. VEGF-A 165 is considered to be the most biologically potent member of the VEGF gene family. 25 It binds to both VEGF receptor-1 (VEGFR-1) and VEGF receptor-2 (VEGFR-2), unlike placental growth factor (PlGF) that only binds to the former. VEGFR-1 is known to undergo weak or undetectable ligand-dependent phosphorylation in comparison with VEGFR-2, 26 and hence it is believed that binding to VEGFR-2 is a critical requirement to induce the full spectrum of VEGF biological responses. 27 It is for this reason that we decided to investigate the effects of VEGF-A 165 overexpression on the uteroplacental vasculature.
Previously, we showed that local adenoviral overexpression of VEGF-A 165 in the UtAs of pregnant sheep resulted in a significant increase in UABF as measured by ultrasound Doppler quantification, a subjective measurement method. We also demonstrated a significant reduction in UtA contractility and an enhancement of UtA relaxation 4 --7 days after Ad.VEGF-A 165 administration, changes that occurred concomitantly with an upregulation of VEGF in Ad.VEGF-A 165 -transduced vessels compared with Ad.LacZ-transduced vessels, as determined by enzyme-linked immunosorbent assay, immunohistochemistry and reverse transcription-PCR. 28 Using flow probes that are a more objective method of measurement, we now show a sustained effect of VEGF-A 165 gene delivery on UABF in normal pregnant sheep, not affected by uteroplacental insufficiency or FGR, lasting up to 4 weeks after adenovirus vector injection, an increase in fetal weight relative to historical controls, and we investigate the mechanism of action, including vascular reactivity, neovascularization and changes in endothelial nitric oxide synthase (eNOS) levels.

RESULTS
Experimental outcome UABF was measured successfully for long term in 10 (6 singletons and 4 twins) out of 11 pregnant ewes (Table 1). In two sheep where UABF data acquisition failed (n ¼ 3 UtAs), flow probe cables were defective or were damaged during implantation. In all other cases, UABF data were available within 48 --72 h of flow probe placement. In those fetuses that underwent telemetric fetal hemodynamic monitoring (n ¼ 4), two fetal deaths occurred 8 and 12 days after fetal catheter placement. There were no other cases of maternal or fetal morbidity or mortality. One ewe gave birth to a healthy lamb on the day of scheduled post-mortem examination, 2 days earlier than expected. The delivery was uncomplicated without any postpartum hemorrhage or retained placenta. The ewe and lamb were monitored for 4 months until weaning and showed no adverse development or abnormal growth velocity. Gross examination at the time of post mortem and microscopic histological examination of ewes, fetuses and the one lamb did not reveal any pathology. The UtAs did not show any evidence of edema, leukocyte infiltration or inflammation.
There were no detectable changes in hematological and biochemical profiles or liver enzyme function when compared with baseline analysis in the mother at 1 week or 5 weeks after vector injection (n ¼ 3), or in the fetal sheep after vector injection when compared with controls (n ¼ 3).
Fetal weight Fetal weights and fetal liver weights (n ¼ 6) from singleton pregnancies undergoing long-term UABF monitoring were measured at post-mortem examination and compared with a historical singleton fetal control group from the same sheep breed (n ¼ 9). The gestational age of the two groups was not statistically different (mean 139.0±3.1 days vs mean 137.8±3.9 days, P ¼ 0.61, unpaired t-test). The mean fetal weight in the experimental group was significantly higher than that in the control group (5990±951 vs 4698±1004 g, P ¼ 0.03, unpaired t-test). Mean fetal liver weight was also significantly increased (146 ± 42 vs 106 ± 21 g, P ¼ 0.01). The maternal weights of the Ad.VEGF-A 165injected ewes at post-mortem examination were not significantly different than the maternal weights of the sheep used as historical controls (77.2 ± 6.4 vs 78.4 ± 3.8 kg, P ¼ 0.69).
Umbilical artery Doppler examination Umbilical artery Doppler pulsatility index was measured at midgestation (before vector injection) and at term (4 --6 weeks after vector injection) in fetal sheep in the uterine horn that received Ad.VEGF-A 165 injection (n ¼ 4), Ad.LacZ injection (n ¼ 4) or phosphate-buffered saline (PBS; n ¼ 2). There were no significant differences in the change in pulsatility index with gestation between any of the groups examined.
The long-term effect of Ad.VEGF-A 165 on UABF There was a slight but nonsignificant fall in UABF from baseline directly after vector injection that recovered completely by days 5 --7 in both the Ad.VEGF-A 165 -and Ad.LacZ-injected UtAs. As anticipated, UABF increased as gestation advanced. Considering singleton and twin gestations together, at 28 days post vector injection, the mean increase in blood flow in the UtAs injected with Ad.VEGF-A 165 (n ¼ 10) was significantly higher when compared with UtAs injected with Ad.LacZ vector (n ¼ 9; þ 36.53%±s.e.m. 4.51% vs þ 20.08%±s.e.m. 5.28% respectively, P ¼ 0.02, Figure 1 and Table 2). The difference in UABF was apparent and significantly different as early as 7 days after vector injection, although there was only a trend at the 14-and 21-day post-injection time points ( Table 2). The magnitude of increase in UABF was maximum when Ad.VEGF-A 165 was injected on the nongravid side in singleton pregnancies (Table 3a).
Using general linear model (GLiM) analysis we analyzed the significance of the contribution of the vector and the gravid state of the injected uterine horn to the changes in UABF observed. The mean gradient of percentage increase in UABF, defined as the slope of the percentage increase in UABF with respect to time, was significantly higher in the Ad.VEGF-A 165 -transduced vessels at all time points examined, that is, 7, 14, 21 and 28 days after vector injection ( Table 2). Only the Ad.VEGF-A 165 vector type significantly contributed to the higher gradient in Ad.VEGF-A 165 -transduced vessels (n ¼ 10). Thus, the more rapid increase in UABF seen in the Ad.VEGF-A 165 -injected UtAs was not significantly related to whether the vessel was supplying a gravid or nongravid uterine horn.

Vascular reactivity
Organ bath experiments on UtA segments 30 --45 days post vector injection showed a significantly reduced mean contractile response to phenylephrine in the Ad.VEGF-A 165 -transduced vessels when compared with Ad.LacZ vessels (Table 3b and Figure 2). In twin pregnancies wherein each UtA supplied a pregnant horn and had similar baseline blood flows thereby serving as an appropriate internal control (n ¼ 3), the maximum contractility (E max ) in the Ad.VEGF-A 165 -transduced arteries was 155.7 ± 13.65 vs 180.7 ± 10.39 in the Ad.LacZ-transduced vessels (P ¼ 0.0003, two-way analysis of variance). We observed no significant difference in the relaxation response to bradykinin between the Ad.VEGF-A 165 -and Ad.LacZ-transduced sides, in twins (n ¼ 3, Figure 2) and singletons (n ¼ 5).
Treatment with the NOS inhibitor NG-nitro-L-arginine methyl ester (L-NAME) significantly reduced the relaxation response to bradykinin in UtA segments exposed to either vector (DE max À20.64% in Ad.VEGF-A 165 -transduced vessels vs DE max À15.07% in Ad.LacZ-transduced vessels, n ¼ 4), but there was no significantly different relaxation response in the presence of L-NAME between the Ad.VEGF-A 165 -and Ad.LacZ-injected segments. Treatment with the cyclooxygenase inhibitor Indomethacin alone did not change the endothelium-dependent relaxation (results not shown). The remaining relaxation was significantly reduced by pretreatment with the small conductance Ca 2 þ -activated K þ (SK) channel blocker Apamin, in the presence of L-NAME and Indomethacin (DE max À48.26% in Ad.VEGF-A 165 -transduced vessels vs DE max À43.21% in Ad.LacZtransduced vessels, n ¼ 4). There was no significant difference in the small amount of residual relaxation that remained after NOS, cyclooxygenase and endothelium-derived hyperpolarizing factor inhibition between the Ad.VEGF-A 165 -and Ad.LacZ-transduced sides.
The contractile and relaxation responses of the umbilical arteries from twin pregnant animals were not significantly different between fetal sheep that developed in the Ad.VEGF-A 165 -and Ad.LacZ-transduced sides (data not shown).

VEGF expression
At the end of gestation (30--45 days after vector injection), human VEGF-A 165 was not detectable by enzyme-linked immunosorbent assay in any UtA, uterine wall or placentome samples of all pregnant ewes injected with Ad.VEGF-A 165 (n ¼ 10). Similarly, when compared with blood samples obtained from the ewes before vector injection, samples obtained from the ewes and fetuses at post-mortem examination did not show any detectable human VEGF-A 165 expression using enzyme-linked immunosorbent assay.
Immunohistochemical analysis for VEGF expression in the UtAs showed specific staining of VEGF around the adventitial blood vessels on both the Ad.VEGF-A 165 -and Ad.LacZ-treated sides, but there was no detectable quantitative difference in the level of expression between the sides (data not shown).
Seminested PCR detected a band specific for the Ad.VEGF-A 165 transgene only in the Ad.VEGF-A 165 -transduced vessels (n ¼ 3), but not in Ad.LacZ-transduced vessels or in other tissues ( Figure 4).   Abbreviations: ANOVA, analysis of variance; NA, not applicable; E max , maximum contractile response; UABF, uterine artery blood flow; UtA, uterine artery; VEGF, vascular endothelial growth factor. The second and third branches of the main uterine arteries from each animal were examined in quadruplicate.

eNOS and VEGF receptor expression
We were unable to detect any difference in the levels of eNOS between the Ad.VEGF-A 165 -and Ad.LacZ-transduced vessels by western blotting 30 --45 days after gene transfer. However, in transduced vessels prepared for western blotting 4 --7 days after vector injection, we found levels of eNOS to be upregulated (Figures 5a and b). Immunostaining of VEGF receptors demonstrated expression of VEGFR-1, VEGFR-2 and neuropilin-1 in luminal and adventitial vessel endothelium in Ad.VEGF-A 165 -, Ad.LacZ-and PBS-injected vessels. There was an upregulation of VEGFR-2 in the UtAs transduced with Ad.VEGF-A 165 compared with Ad.LacZ, at 4 --7 days after injection ( Figure 6). There was no detectable difference in the level of any of the receptors at 1 month post injection (data not shown).
Vessel enumeration and intima/media ratios The average number of positively stained adventitial blood vessels counted in the main UtA and its three branches of twin animals (n ¼ 4) was significantly greater in the UtAs transduced with Ad.VEGF-A 165 compared with Ad.LacZ (140.6±15.1 vs 91.7±9.9 respectively, n ¼ 26, Po0.05, Figure 6). The number of vessels in PBS-injected UtAs (86.4 ± 9.4, n ¼ 16) was not significantly different from Ad.LacZ-transduced UtAs. We also observed a reduction in the intima/media ratio in Ad.VEGF-A 165 -transduced vessels compared with Ad.LacZ-transduced vessels on hematoxylin and eosin-stained sections (twin animals, n ¼ 4; Figure 5c).
Effect of Ad.VEGF-A 165 on maternal and fetal hemodynamics Maternal blood pressure (BP) and heart rate (HR) were monitored in 5 ewes. At 5 days after vector injection, the maternal mean arterial pressure had fallen from 92.78±6.34 mm Hg at baseline to 90.41 ± 8.05 mm Hg (n ¼ 5, P ¼ 0.21), which is very similar to our observation in the sham-injected control ewes. Maternal HR increased slightly from 106 beats per min (b.p.m.) to 108 b.p.m.

DISCUSSION
We have studied the long-term effects of local adenovirusmediated overexpression of VEGF-A 165 in the UtAs of pregnant sheep and found a significantly higher increase in blood flow in vessels transduced with Ad.VEGF-A 165 compared with those transduced with the control vector Ad.LacZ. A statistically significant increase in UABF is apparent as early as 7 days after Ad.VEGF-A 165 vector injection and is maintained for at least 28 days, where it reaches a 37% increase over baseline compared with 20% on the Ad.LacZ-injected side (P ¼ 0.02). This increase was concomitant with a significant reduction in UtA contractility and a significant increase in the number of adventitial blood vessels in Ad.VEGF-A 165 -treated UtAs. These findings are significant because the increase in UABF is maintained over at least 4 weeks, a critical finding that supports a possible therapeutic potential for this intervention. Importantly, the use of the 'gold standard' transit-time flow probes in this study, instead of Doppler ultrasonography 28 as used in our previous short-term study, provides a far more accurate and reliable estimate of the true increase in UABF. 29 To our knowledge, this is the first report of a single intervention that leads to a sustained increase in UABF within or outside of pregnancy. There are currently no other interventions that have successfully achieved a long-term increase in UtA perfusion. The cGMP-specific phosphodiesterase inhibitors, such as sildenafil citrate, have been suggested as a possible therapy in FGR, but have been found to have the opposite effect on growth-restricted sheep pregnancies. 19 Additionally, this is the first report of long-term vectormediated VEGF expression in the uteroplacental circulation.
In this study we measured blood flow in both the gravid and nongravid UtAs to enable comparison of UABF in the same animal. Previous studies have examined blood flow in vessels supplying only the gravid horn 30 or total uterine blood flow (gravid þ nongravid UtA). 31 We observed that there was a larger increase in UABF when Ad.VEGF-A 165 was injected into the UtA supplying the nongravid rather than the gravid horn. This could be because the vessels supplying the nongravid horn have undergone less vascular remodeling than those supplying the gravid horn, 32 which may make them more susceptible to the effects of VEGF overexpression. During normal pregnancy, the uteroplacental vascular bed develops and the vascular channels dilate to facilitate the maximal supply of substrates and oxygen to the developing fetus. 33 It is possible that a larger effect might be observed in pregnancies compromised by FGR related to uteroplacental insufficiency, where there is evidence of reduced trophoblast invasion of the spiral arteries and reduced perfusion.
Fetal weight in late gestation was increased in our experimental group as compared with a historical control cohort. The maternal weights of the ewes in both the groups were not significantly different. Although our experiments were not designed to investigate fetal growth, these results are encouraging, and suggest that the increased UtA perfusion that resulted from VEGF expression may result in increased nutrient and oxygen transport to the fetuses. Fetal growth however is highly variable between  The drop in UABF observed during the first few days after vector injection was limited to B10% and was probably caused by the vessel occlusion and trauma during injection. UABF recovered in all treated sheep by days 5 --7 after vector injection. This is still a point of concern, however, if this gene transfer technique is to be applied in the clinical setting, where growth-restricted fetuses are frequently hypoxemic. 34 We observed a significant reduction in the UtA contractile response but no difference in relaxation response 30 --45 days after Ad.VEGF-A 165 transduction. This is different to our findings on the short-term effects of Ad.VEGF-A 165 delivery, showing that in addition to a significant reduction in the contractile response, the relaxation response was significantly increased in Ad.VEGF-A 165transduced vessels. 28 This is most likely to be because at term, the uteroplacental blood vessels are maximally dilated, 35 and VEGF overexpression, which is known to bring about vasodilatation, may be unable to further enhance the relaxation of UtAs. Using differential blockade of the endothelium-mediated vasorelaxant pathways, we observed that it is primarily the NOS/endotheliumderived hyperpolarizing factor pathway that plays an important role in relaxing the UtAs of pregnant sheep at term, as also noted in our short-term studies. 28 The Ad.VEGF-A 165 -induced upregulation of eNOS in the UtAs observed up to 7 days was not evident after long-term gene transfer. This suggests two possible mechanisms: either continued significant VEGF expression is needed for long-term eNOS upregulation or, alternatively, eNOS level at term and in normal pregnancy is already at its peak and cannot be upregulated further by VEGF overexpression. We are currently performing studies in pregnant sheep UtA endothelial cells to elucidate the underlying mechanism.
Blocking the NOS pathway with L-NAME resulted in a significantly enhanced contractility of the Ad.VEGF-A 165 -but not Ad.LacZ-transduced UtAs. Thus, although the response to bradykinin was not significantly different in the presence of NOS inhibitors between the two sides, NOS/NO appeared to be playing an important role in the diminished contractility of Ad.VEGF-A 165transduced vessels. We speculate that VEGF may be conditioning the UtA smooth muscle to become more sensitive to NO and thereby generate more cGMP.
The vascular reactivity of the umbilical arteries in the long term was no different between the Ad.VEGF-A 165 -and Ad.LacZtransduced sides. In experiments using in vitro isolated perfused lobules of the human placenta from normal pregnancies and those affected by PET, Brownbill et al. 36 observed that VEGF-165 applied to the fetal-side circulation was a potent vasodilator of the fetoplacental vasculature. However, we did not observe any difference in umbilical artery Doppler parameters in these longterm or in previous short-term experiments (4 --7 days post vector injection). 28 This suggests that the transgenic VEGF protein that exists in the human placenta as a monomer of 38K and a dimer of 76K 37 is too large to cross the epitheliochorial placental barrier in any significant degree. It is also possible that free VEGF on the maternal side is sequestered by the excessive syncytiotrophoblast release of soluble VEGFR-1 into the intervillous space.
Results of immunohistochemical staining showed expression of VEGFR-1, VEGFR-2 and neuropilin-1 in the luminal and adventitial endothelium, indicating that all three VEGF receptors could be involved in mediating VEGF activity in UtAs. Although VEGFR-1 and VEGFR-2 expression has been reported in ovine UtA endothelial cells, 38  injection, but no detectable difference in the level of any of the receptors at the longer time points examined, when transgene and protein expression had declined. Reduced VEGFR-2 expression is a characteristic of atherosclerotic arteries, and may be associated with the development of vasculopathies and their complications. 39 There is also a significant reduction in the maternal serum levels of soluble VEGFR-2 in women whose pregnancies are complicated by FGR. 40 As VEGFR-2 is primarily derived from the vascular endothelial cells, it is possible that the enhanced endothelial expression of VEGFR-2 observed in this study may therapeutically benefit pregnancies affected by FGR. Our findings suggest that upregulation of eNOS in the first week after vector injection is responsible for the initial increase in UABF. The long-term increase in UABF, however, may also be related to enhanced UtA vascularization reflected by an abundant adventitial blood supply. Further investigation into the nature of these adventitial blood vessels is needed. The vasa vasorum is a microvascular network that originates primarily in the adventitia of large arteries and supplies nutrients and oxygen to the outer layers of the arterial wall. 41 Thus, proliferation of the vasa vasora may augment the function of the UtA thereby enhancing uterine perfusion. In animal experiments, VEGF gene transfer is capable of inducing therapeutic angiogenesis in diverse tissues and organ systems. Overexpression of VEGF using viral vectors stimulates significant neovascularization and supraphysiological increase in perfusion in healthy and ischemic skeletal muscles and myocardium because of increased angiogenesis and capillary enlargement. 42 The reduced intima/media ratio observed in the Ad.VEGF-A 165 -transduced UtAs is consistent with previous findings that VEGF gene delivery inhibits arterial neointima formation, in part via increased production of NO. 11,43 Other studies suggest that disrupting the function of the vasa vasorum results in neointimal thickening. For example, femoral artery ligation in pigs occludes blood flow to the vasa vasora and the consequent hypoxia results in intimal hyperplasia. 44 Hence, intimal hyperplasia and periadventitial blood flow may be inversely related. Intimal hyperplasia observed in the UtAs of premenopausal women is accompanied by impaired cGMP production, and altered arginase and NOS activities. 45 There exists a positive correlation between intima/ media ratios and endogenous NOS inhibitors such as L-NMMA and asymmetric dimethylarginine in endothelial cells, as well as endothelin-1 within the vessel wall of perimenopausal human UtAs. 46 The reduction in intima/media ratios that we observed in our study may confer an arterioprotective effect on the UtAs.
There are theoretical concerns that overexpression of VEGF might lead to adverse maternal or fetal hemodynamic effects. We did not observe any maternal or fetal hemodynamic changes other than a small fall in maternal BP toward the end of term, which is normally observed in sheep. 47 At the same time, no long-term expression of VEGF-A 165 could be detected in maternal or fetal tissues, which provides reassurance against long-term toxic effects.
However, application of our therapeutic approach to severely growth-restricted human fetuses is particularly challenging as these fetuses are commonly hypoxic, and reductions in UABF during UtA occlusion may actually lead to a rapid decompensation into fetal acidemia and thus a significantly higher chance of fetal/neonatal complications. A minimally invasive technique such as transfemoral UtA catheterization with temporary balloon occlusion of the vessel lumen, as is used to treat massive obstetric hemorrhage, 48 may significantly decrease UtA trauma and postinjection constriction. Further experiments to determine optimal vector dose and the mode of delivery are required. Another issue is the timing of gene delivery. Fetuses with advanced growth restriction and cardiovascular compensation through brain sparing have a significant degree of hypoxemia but are not acidemic until abnormal precordial venous Dopplers signal decompensation. 49 This should be considered when deciding on the best timing for vector application, and especially when the longitudinal progression of early severe FGR is well defined. 50 In addition, those fetuses with early FGR and very high umbilical artery resistance at presentation (44 s.d. above the mean for gestation) are at a very high risk of morbidity and mortality. 51 These fetuses would represent the target population for a phase I trial of Ad.VEGF-A 165 , and application of the vector in this case would be preferable at a stage where significant hypoxemia has not yet developed (that is, before brain sparing occurs) to prevent acute deterioration during UtA occlusion. Application at this stage would also be helpful when one considers that Ad.VEGF-A 165 will need some time before a substantial increase in UABF is noticed. If applied late, the time delay in observing any beneficial changes might be enough for a fetus with an already established brain sparing phenomenon to reach a preterminal stage. 52 CONCLUSION This report shows that local adenovirus-mediated VEGF-A 165 overexpression in the pregnant sheep UtAs at mid-gestation results in a significant long-term increase in UABF, a reduction in UtA vascular contractility, increased adventitial angiogenesis and increased fetal weight. It gives hope that VEGF gene therapy has potential to reverse the impaired uteroplacental perfusion found in the majority of cases of FGR. Vector administration to the mother appears to be safe, leading to no undesirable vector expression, change in maternal or fetal hemodynamics or pathology. Studies in growth-restricted small and large animals, optimization of the delivery technique and further safety evaluation will be required before clinical application could be contemplated.

MATERIALS AND METHODS Animals
All experiments in the current investigation were carried out in normal pregnant sheep not affected by vascular placental insufficiency. A total of 17 time-mated Romney breed mid-gestation pregnant ewes (80 --102 days of gestation, term ¼ 145 days) carrying singleton (n ¼ 10) or twin pregnancies (n ¼ 7) were studied until the end of gestation (Table 1). Gestational ages were confirmed by ultrasound. 53 All work was conducted in accordance with the UK Animals (Scientific Procedures) Act (1986).

Animal surgery and vector injection
After starving overnight, pregnant ewes at 90.91 ± 2.51 days of gestation (n ¼ 11, 5 twins, 6 singleton gestations) underwent general anesthesia induced with thiopental sodium 20 mg kg --1 IV (Thiovet, Novartis Animal Health UK Ltd, Royston, UK) and maintained with 2 --2.5% isoflurane in oxygen (Isoflurane-Vet, Merial Animal Health Ltd, Harlow, UK) after intubation. Umbilical artery Doppler measurements, pulsatility index and resistance index were acquired. 28 For chronic maternal hemodynamic monitoring (n ¼ 5), a BP-sensitive PA-D70 catheter (Data Sciences International, Tilburg, The Netherlands) was inserted into the carotid artery lumen. A laparotomy was then performed, the UtAs were identified bilaterally and mobilized immediately proximal to the first bifurcation. A 6 mm 6PS transit-time flow probe (Transonic Systems Inc., Ithaca, NY, USA), which can measure blood flow with an absolute accuracy of ±10%, was placed (n ¼ 11 animals) around the main UtA on each side. The cabling from each probe was then exteriorized onto the ewe's right flank, and the skin buttons were secured to the skin as described. 29 After 4 to 10 days (at 97.82±2.37 days of gestation), the sheep underwent a second general anesthetic and laparotomy for local administration of Ad.VEGF-A 165 and Ad.LacZ vectors into the UtAs contralaterally. The viral dose administered was 5 Â 10 11 viral particles in 10 ml PBS, over a 5-min period during which the UtA was digitally occluded proximal to the site of injection. 28 The operators were blinded as to which horn of the uterus received Ad.VEGF-A 165 at the time of vector injection surgery. The ewe received standard analgesia and antibiotic prophylaxis, and the abdominal incision was closed. 29 Two ewes (1 singleton and 1 twin) received injection of PBS (10 ml) instead of adenovirus vector into each UtA to act as a sham injection control group, providing data for maternal hemodynamic assessment and tissues for organ bath experiments and other analyses. In the ewes where fetal hemodynamic monitoring was performed, UABF data were not measured (Table 1). Fetal catheter implantation was accomplished using an open fetal surgical technique at 101.25±0.25 days of gestation, where we implanted the catheter of a D70-PCTP hemodynamic telemetric device (Data Sciences International) into the carotid artery of the fetal sheep (n ¼ 4). 54 Vector injection was performed as described above 6 --8 days later.

Animal monitoring
Measurements of UABF, maternal and fetal blood pressure and heart rate were recorded continuously over the 3 days preceding and 7 days succeeding vector injection, and for 1 h daily thereafter to capture acute and chronic effects. UABF was sampled at a rate of 128 Hz and data were transmitted telemetrically via the skin buttons when they were connected to the PhysioGear I transmitter system and the PhysioView Data Acquisition Software (Transonic Systems Inc.). The data acquired from the flow probes were analyzed using Acknowledge software 3.9.1 (Biopac Systems Inc., Goleta, CA, USA). The baseline UABF was calculated as the average of three daily mean UABF measurements taken for 1 h each day before vector injection. UABF percentage change from baseline was calculated at specified time points, 7, 14, 21 and 28 days after vector injection, using the average of 3 consecutive daily mean UABF measurements on the day of and 1 day either side of the time point. A two-way GLiM was used to compare the UABF percentage change in Ad.VEGF-A 165 -and Ad.LacZ-injected UtAs at each time point and also the gradients of UABF percentage change over the length of gestation. The two factors accounted for in the GLiM analysis were whether the UtA supplied a gravid or nongravid horn and whether Ad.VEGF-A 165 or Ad.LacZ vector was injected.
Maternal and fetal BP and HR were recorded telemetrically. Fetal BP was corrected for amniotic fluid pressure. Baseline values were obtained continuously over the 3 days preceding vector injection and compared with corresponding postinjection values. Uploaded traces were analyzed using Dataquest ART 4.1 software (Data Sciences International). A twotailed paired t-test was used to compare changes in BP before and after the administration of the vector.

Tissue sampling
At the end of gestation (136 to 144 days) the sheep underwent terminal general anesthesia. The umbilical artery pulsatility index and resistance index were measured using ultrasound Doppler and were compared with preinjection values. The UtAs and their next three divisions down to the level of the uterine wall (vessel diameter 1 mm) as well as other maternal and fetal tissues were sampled as described. 28 Fetal biometry and weight were noted. Tissue samples to be used in histological and immunohistochemical studies were fixed in 4% paraformaldehyde for 24 h and then transferred into 70% alcohol to be subsequently blocked in paraffin. All other samples were snap frozen in liquid nitrogen and stored at À80 1C.

Organ bath studies for UtA reactivity
The cleared second and third UtA branches or umbilical arteries were divided into 3 mm long segments and examined on two 8-chambered organ bath setups in the absence and presence of inhibitors as described, 28 namely L-NAME (300 mM), an NO synthase inhibitor, Indomethacin (10 mM), a cyclooxygenase inhibitor and Apamin (1 mM) a blocker of SK channels that inhibits the actions of endothelium-derived hyperpolarizing factor. Additionally, the contractility of UtAs to phenylephrine was measured in the presence of L-NAME.

Histological and blood examination
Paraffinized tissue sections stained with hematoxylin and eosin were observed microscopically for histological examination. Maternal and fetal blood samples obtained before vector injection and at post mortem were tested for routine hematology, biochemistry and liver function tests.

Assessment of neovascularization
Paraffinized UtA sections were immunostained with anti-von Willebrand factor (1:400, A0082, Dako, Glostrup, Denmark) using the avidin-biotinperoxidase system (PK4000, Vector Laboratories, Peterborough, UK); 3,3 0diaminobenzidine (Sigma Aldrich, Gillingham, UK) and a light hematoxylin counterstain were used as substrates for visualization. The positively stained adventitial blood vessels with a distinct lumen were then counted under a microscope (Nikon Eclipse E600, Amstelveen, The Netherlands) by observers blind to the type of vector injection.

Measurement of VEGF expression
The quantity of human VEGF-A 165 protein in frozen samples of UtA, uterine wall and placentome from 10 pregnant sheep was measured by enzymelinked immunosorbent assay (R&D Systems, Minneapolis, MN, USA) as described. 28 Human VEGF-A 165 levels were also measured in pre-and postinjection maternal serum samples and fetal serum samples collected at post-mortem examination.
Ad.VEGF-A 165 vector expression was assessed using seminested reverse transcription-PCR in maternal and fetal tissues. The first-round PCR was performed as described previously. 28,55 The second-round PCR was carried out on the first-round PCR product, using a reverse primer that bound to a sequence internal to the complementary binding site of the first-round reverse primer (Figure 4, sequence 5 0 -GATCCGCATAATCTGCATGGT-3 0 ). The annealing was at 54 1C for 60 s, and all other conditions were kept identical to the first-round PCR.
Primers designed to amplify the endogenous sheep sequences associated with the TATA-box Binding Protein (GenBank reference L47974) were used as a positive control on sheep complementary DNA and gave no amplification signals from adenovirus DNA alone. PCR conditions used were as described. 28 PCR products were analyzed on 1% agarose gel stained with ethidium bromide. eNOS studies Protein extracts from the snap-frozen UtA tissues were used to estimate eNOS levels by western blotting, using a monoclonal mouse anti-eNOS antibody (1:3000, 610296, BD Transduction Laboratories, Oxford, UK). Samples of UtA tissues from a previous short-term cohort of sheep that were killed 4 --7 days post administration of the vector 28 were also analyzed.

CONFLICT OF INTEREST
This work was funded in part by Ark Therapeutics Ltd, London. The adenovirus vectors were supplied free of charge by Ark Therapeutics Oy, Kuopio, Finland. JM is Chief Scientific Officer, Ark Therapeutics Ltd. IZ is a consultant for Ark Therapeutics Ltd.