Author + information
- Received July 31, 2006
- Accepted August 30, 2006
- Published online November 21, 2006.
- Bettina Wagner, DVM⁎,
- Martina Anton, PhD†,
- Stephan G. Nekolla, PhD⁎,
- Sybille Reder, MT⁎,
- Julia Henke, DVM†,
- Stefan Seidl, MD‡,
- Renate Hegenloh, MT§,
- Masao Miyagawa, MD⁎,
- Roland Haubner, PhD⁎,
- Markus Schwaiger, MD⁎ and
- Frank M. Bengel, MD⁎∥,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Frank M. Bengel, Division of Nuclear Medicine, Russell H. Morgan Department of Radiology, Johns Hopkins University Medical Institutions, 601 N. Caroline Street, JHOC 3225, Baltimore, Maryland 21287.
Objectives We sought to investigate the usefulness of integrated positron emission tomography (PET) and computed tomography (CT) for in vivo characterization of an angiogenesis-directed molecular intervention.
Background Controversies about the effectiveness of molecular therapies for cardiovascular disease have prompted the need for more powerful noninvasive imaging techniques.
Methods In a model of regional adenoviral transfer of the VEGF121gene to myocardium of healthy pigs, PET-CT using multiple molecular-directed radiotracers was employed.
Results Two days after gene transfer, successful transgene expression was noninvasively confirmed by a reporter probe targeting co-expressed HSV1-sr39tk reporter gene. The CT-derived ventricular function and morphology remained unaltered (left ventricular ejection fraction 57 ± 5% in adenovirus-injected animals vs. 53 ± 5% in controls; p = 0.36). Increased regional perfusion was identified in areas overexpressing VEGF (myocardial blood flow during adenosine-induced vasodilation 1.47 ± 0.49 vs. 1.14 ± 0.27 ml/g/min in remote areas; p = 0.01), corroborating in vivo effects on microvascular tone and permeability. Finally, regional angiogenesis-associated αvβ3integrin expression was not enhanced, suggesting little contribution to the perfusion increase. Fusion of CT morphology and tracer-derived molecular signals allowed for accurate regional localization of biologic signals. Findings were validated by control vectors, sham-operated animals, and ex vivo tissue analysis.
Conclusions Integrated PET-CT has the potential to dissect cardiovascular biologic mechanisms from gene expression to physiologic function and morphology. The VEGF overexpression in healthy myocardium increases myocardial perfusion without significant up-regulation of αvβ3integrin adhesion molecules early after the intervention.
Various gene- and cell-based molecular therapeutic approaches directed against cardiovascular disease have been evaluated in experimental settings, and some have reached the stage of clinical trials (1,2). Despite rapid progress, some basic principles of these therapies are still under development, and issues related to therapeutic efficacy and the best delivery of agents to myocardium are still debated. For example, early experimental results for angiogenesis therapy of myocardial ischemia were difficult to reproduce in the clinical setting (3,4).
These developments have prompted a need for powerful noninvasive imaging tools to provide specific disease-related biologic insights and facilitate the translation from experimental to clinical practice. A single cardiac imaging modality for in vivo characterization of tissue biology, physiologic function, and morphologic appearance could thus be of considerable value.
In the present study, we explored the potential of integrated positron emission tomography (PET) and computed tomography (CT) for noninvasive imaging of a myocardial molecular intervention (i.e., the regional transfer of vascular endothelial growth factor [VEGF] gene to myocardium of healthy pigs). Our primary goal was to study the feasibility of this methodology for integrated imaging of the heart from its morphology to subcellular function. As a secondary goal, we intended to get further insights into the effects of VEGF overexpression on the tissue level and on the heart as a whole.
The experimental protocol was approved by the regional governmental commission of animal protection (Regierung von Oberbayern) and is summarized in Table 1.
Eleven young domestic pigs (30 to 40 kg; Versuchsstation, Thalhausen, Germany) underwent left thoracotomy under anesthesia with mechanical ventilation (10 mg/kg/h propofol intravenously, fentanyl intravenously as required). Pericardium was opened, and gene transfer was performed in study group animals (n = 8) using doses of 1 × 1010plaque-forming units (pfu) of adenovirus (total volume 1 ml, aliquoted into 5 syringes). Those were consecutively injected into 2 separate myocardial areas of approximately 2 × 2 cm. Control group animals (n = 3) received saline injection following the same protocol. Both injection sites were marked with titanium clips (LIGACLIP 20/20, Ethicon, Norderstedt, Germany).
Two days later, all pigs were anesthetized using the same regimen and underwent PET-CT. Animals were killed after imaging (60 mg/kg sodium pentobarbital intravenously). Hearts were excised and rinsed, and transmural tissue samples were taken from the 2 areas of injection and from a remote area (inferior wall).
Adenoviral vector preparation and injection
E1 region-deleted replication-deficient adenoviral vectors carrying transgenes under transcriptional control of human cytomegalovirus promoter were used. Vectors were double cesium chloride purified. Titers were determined by plaque assay on 293 cells. For monitoring of therapeutic gene expression by co-expression of a reporter gene, adenovirus expressing the mutant herpesviral thymidine kinase (HSV1-sr39tk) reporter gene and the human VEGF121gene independently in 2 expression cassettes (AdTk-VEGF) (5) was applied in all 8 pigs. The injection site (basal or apical) was varied to control for interfering effects (Table 1). For injection into the second site, control adenovirus expressing only 1 transgene, either the HSV1-sr39tk reporter gene (Adsr39tk; n = 4), or the VEGF121gene (AdVEGF; n = 4), was used.
PET-CT imaging protocol
Imaging was done on a Siemens Biograph Sensation 16 scanner (Siemens Medical Solutions, Malvern, Pennsylvania), a hybrid system incorporating a 16-slice X-ray CT and a 3D LSO crystal-equipped PET component. [18F]fluoro-hydroxymethylbutyl-guanine (FHBG) and [18F]galacto-RGD were synthesized as previously described (6,7).
All animals were scanned in supine position under anesthesia with mechanical ventilation. Heart rate, blood oxygen saturation, and electrocardiogram were recorded continuously. Imaging sessions started with an X-ray topogram for orientation, followed by a low-dose CT (tube voltage 120 kV, tube current 14 mA) of the cardiac region for subsequent correction for photon attenuation. Ventilation was temporarily stopped for CT to obtain images in an intermediate breathing position for improved alignment between CT and PET.
Subsequently, PET emission data were obtained. Myocardial perfusion was assessed by use of [13N]ammonia. At rest, 185 to 240 MBq were injected intravenously, and a 10-min dynamic acquisition in list mode, immediately followed by a static image of 10 min, was acquired. After a break of 30 min to allow for decay, pharmacologic vasodilation was induced by infusion of adenosine (0.14 mg/kg/min) for 5 minutes. Two minutes into the adenosine infusion, a second dose of [13N]ammonia was injected and another set of images was obtained. In the 8 animals that received adenoviral vectors, 160 to 260 MBq of [18F]FHBG were then injected, and a 20-min static image was obtained 45 min after injection.
After PET, the session was concluded with contrast-enhanced CT. Imaging of a test bolus of 20 ml contrast agent (Imeron300, Altana Pharma, Konstanz, Germany) was performed. Then mechanical respiration was stopped again, 70 ml contrast agent was injected at 4 ml/s, and a volume data set was acquired (tube voltage 100 kV, tube current 550 mA), covering the entire cardiac region with continuous electrocardiogram co-registration.
Four animals underwent a second PET-CT session at day 3 to 9 after surgery. This session consisted of CT, a [13N]ammonia perfusion measurement at rest, as described in the preceding, and injection of 170 to 215 MBq [18F]galacto-RGD and a 30-min static image after 60 min.
PET data analysis
Attenuation-corrected transaxial images were reconstructed by an iterative algorithm (OSEM, 4 iterations, 8 subsets, 128 × 128 matrix, slice thickness 3 mm, no overlap). Using in-house-developed software (Munich Heart) (8), volumetric sampling was performed to identify 460 myocardial segments in the resting perfusion study of each session. For precise regional co-localization, segments were then transferred to images of stress perfusion and [18F]FHBG and [18F]galacto-RGD distribution. Polar maps of left ventricular myocardial uptake for each tracer were generated and normalized. Regional tracer uptake was expressed as percentage of the maximum in 9 myocardial areas (anterior, lateral, septal, and inferior wall, divided in basal and distal areas; apex as separate area). The 2 injection areas along with a remote area were defined according to the experimental protocol and according to localization of titanium clip markers at CT.
Absolute quantification of myocardial blood flow (MBF) was performed by use of dynamic [13N]ammonia datasets, based on a validated three-compartment model (9). List mode data were rebinned to 18 consecutive dynamic image datasets (12 × 10 s, 4 × 60 s, 2 × 120 s). Arterial input function was defined by a small region of interest in left ventricular cavity. Motion-corrected time activity curves were extracted for the 460 myocardial segments previously defined from static images. The MBF in ml/g/min was calculated for each segment by compartment-model fitting and depicted in a polar map.
CT image analysis
For contrast-enhanced CT, overlapping transaxial images were reconstructed using a medium-smooth convolution kernel (B30f, 512 × 512 matrix, slice thickness 1 mm, increment 0.7 mm). Image reconstruction was retrospectively gated to the electrocardiogram. The reconstruction window was varied with a 10% increment, and 10 datasets covering the cardiac cycle in different phases were obtained. Artifact-free data sets were chosen for volume-rendered display and software-based overlay with PET. Data sets for all 10 phases were re-angulated perpendicular to the left ventricular long axis. Using in-house-developed software (Munich Heart), endocardial and epicardial contours were drawn in end-systolic and end-diastolic short-axis slices. The left ventricular ejection fraction (LVEF) was obtained, and regional end-diastolic and end-systolic wall thickness and thickening were calculated for clip-marked injection sites and a remote area in inferior wall. The PET-CT fusion images were generated following manual readjustment, using standard software provided by the manufacturer.
Measurement of plasma VEGF concentration
Blood samples were taken before surgery and 6, 30, and 48 h after the surgical intervention. Plasma levels of human VEGF gene product were measured using a VEGF enzyme-linked immunosorbent assay kit (Calbiochem, Schwalbach, Germany).
Histology and immunohistochemistry
Formalin-fixed myocardial specimens were embedded in paraffin. For qualitative histologic examination of vascular/capillary density, paraffin sections (2 μm) were stained according to van Gieson. For immunohistochemistry, sections were deparaffinized, dehydrated, and pressure-cooked in citrate buffer, followed by blocking of endogenous peroxidase (3% H2O2/methanol).
Immunostaining for transgene products used streptavidin-horseradish-peroxidase technique with incubation of slides with polyclonal rabbit anti–HSV1-tk or anti-humanVEGF antibody and consecutively with ready-to-use anti-rabbit antibody (Dako ChemMate Detection Kit K5001; DakoCytomation, Hamburg, Germany).
Immunostaining for microvascular density used the APAAP technique (DAKO Chem Mate Detection Kit APAAP, Mouse). Primary antibody against alpha smooth muscle actin (DAKOCytomation Clone 1A4, 70 mg/l, 1:50 diluted) was used. For quantification, 10 high-power fields (HPF) from a representative slice of vector-injected regions were microscopically analyzed, and small vascular structures were counted.
Data were analyzed with StatView5.0 software (SAS Institute, Cary, North Carolina). Values are expressed as mean ± SD. Differences were assessed by paired/unpaired ttest as appropriate or (for multiple comparisons) by 1-way analysis of variance with the Fisher penalized least square of differences post hoc test. A 2-sided p value of <0.05 was defined as significant.
CT identifies cardiac intervention sites and preserved structure and function
Titanium clip markings were reliably identified in all animals (Fig. 1A),allowing for fusion and co-localization with PET-derived biologic images. Global LVEF was normal in all animals. There was no difference between adenovirus-injected animals and saline-injected control animals (57 ± 5% vs. 53 ± 5%; p = 0.36). Consistently, regional contractile function was homogeneous. There was no difference for regional wall thickening between adenovirus-injected, saline-injected, or noninjected areas (p = 0.73) (Fig. 1C). Absolute end-diastolic wall thickness was also not different between regions (p = 0.70) (Fig. 1B).
PET reporter gene imaging and PET-CT image fusion identify successful gene transfer and regional transgene expression
All sites injected with adenoviral vector encoding for HSV1-sr39tk reporter gene (AdTk-VEGFor Adsr39tk) revealed significantly elevated uptake of FHBG compared with AdVEGF-injected or remote areas (p < 0.01) (Fig. 2A).Increased regional uptake co-localized with titanium clip markings (Figs. 3Aand 3B). in vivo results were confirmed by ex vivo immunohistochemistry. Expression of HSV1-sr39tk gene product was identified in cytosol and nuclei of myocardial cells in areas injected with AdTk-VEGFor Adsr39tk. Immunostaining for VEGF was positive in the cytosol of myocardial cells in areas injected with AdTk-VEGFor AdVEGF(Fig. 4).
Further support for successful transgene expression was obtained by the appearance of the secretable gene product VEGF121in the systemic circulation. Human VEGF was not detectable in saline-injected animals, nor was it detectable in blood samples before adenoviral injection. After gene transfer, the gene product appeared in the blood of all vector-injected animals, peaked (755 ± 946 pg/ml) at 6 h, declined (385 ± 783 pg/ml) by 30 h, and was detectable in only 1 animal at 48 h after gene transfer.
PET-CT identifies enhanced relative and absolute perfusion after VEGF gene transfer
There was a significant increase of relative uptake of [13N]ammonia in VEGF-transduced areas after injection of either AdTk-VEGFor AdVEGF(p < 0.01 vs. AdVEGFand remote areas) (Figs. 2B, 2C, 3A, and 3B). No differences were found between Adsr39tk-injected and noninjected control areas. All saline-injected control animals showed regionally homogeneous perfusion, ruling out procedure-related influences (Fig. 3C).
Because regionally increased relative uptake of [13N]ammonia in static PET images may occur because of increased vascular permeability without a true increase of absolute tissue perfusion, we additionally explored quantitative MBF from dynamic images. At intraindividual comparison, MBF at rest in study animals was not significantly different between areas injected with AdTk-VEGFand remote areas (0.85 ± 0.22 vs. 0.78 ± 0.13 ml/g/min; p = 0.19). During adenosine vasodilation, MBF was significantly higher compared with remote areas (1.47 ± 0.49 vs. 1.14 ± 0.27 ml/g/min; p = 0.01). Overall, MBF in vector-injected animals was in the range of MBF measured in respective regions of sham-operated animals (0.75 ml/g/min at rest and 1.31 during adenosine), when compared interindividually.
Postmortem analysis revealed mildly, but significantly, enhanced microvascular density in areas injected with VEGF-expressing vector (5.57 ± 2.15 microvessels/HPF vs. 2.17 ± 1.11 for Adsr39tk-injected areas; p = 0.03) (Fig. 5).
VEGF-induced regional perfusion increase is not associated with up-regulation of αvβ3integrin
The second PET-CT session at 3 to 9 days after the intervention focused on measurement of αvβ3integrin expression. Perfusion tracer uptake was again measured at rest, and was still significantly elevated in regions overexpressing VEGF (80 ± 6% vs. 63 ± 8% in remote areas; p = 0.002), confirming persistence of vascular effects.
Overall myocardial uptake of the αvβ3integrin-targeted tracer [18F]-galacto-RGD was very low, and no focal uptake was detected in regions of VEGF overexpression showing increased perfusion (Fig. 6).Relative regional uptake was not different from remote areas (72 ± 5% vs. 75 ± 9%; p = 0.45).
In summary, the present study demonstrates that integrated PET-CT can be used to get detailed in vivo insights into biologic mechanisms of the heart. Changes induced by molecular intervention were characterized from gene expression over physiologic changes to morphologic appearance. Successful adenoviral transfer and subsequent expression of target gene (VEGF121) in myocardium was specifically identified by PET imaging of a co-expressed reporter gene, which co-localized with clip-marked myocardial injection sites detected by CT. With the help of CT co-localization, specific effects of target gene overexpression on myocardial microcirculation at rest and during vasodilation were then characterized by quantitative perfusion imaging. The same technique also suggested that αvβ3integrins do not play a role in the early phase of this VEGF-induced increase of myocardial perfusion in healthy myocardium. Finally, co-registration of CT further added to PET-determined biologic data by defining morphology and confirming absence of impairments of contractile function. These findings establish an imaging modality with substantial future potential in development and monitoring of cardiovascular therapy.
Integrated PET-CT has reportedly been useful for clinical imaging of myocardial perfusion and viability (10–13). The present study goes into further detail and demonstrates the feasibility of molecular characterization of the heart by integrating information from multiple PET tracers and contrast-enhanced CT. It is suggested that PET-CT has substantial potential as a translational tool that can be applied in experimental models as well as in clinical practice. For future clinical application, issues related to radiation exposure by combined nuclear and X-ray techniques need to be considered, and short-lived PET tracers will still require an on-site cyclotron for production. But it can be expected that localization and interpretation of the often weak specific signal from molecular-targeted PET probes will be facilitated by co-registration with high-resolution cardiac CT. Although fusion of PET and CT data had to be visually realigned in this study owing to potential influences of respiratory and cardiac motion, algorithms that allow for registration of respiratory motion during imaging are under development. These are expected to improve the precision of spatial co-registration in the future (14) and may help to circumvent problems with faulty attenuation correction of PET data from potentially misaligned CT scans to extend the application of PET-CT beyond myocardial imaging toward biomorphologic imaging of coronary vessels in the future.
Although proof of principle for the usefulness of PET-CT as a cardiac molecular imaging tool was the primary goal of this study, we chose a model of VEGF myocardial gene transfer, which allowed for using the novel imaging approach to get further insights into the mechanisms and consequences of VEGF overexpression. This model was chosen because it has been considered promising for angiogenesis therapy of myocardial ischemia (2). Difficulties with this therapy were encountered when trying to reproduce experimental results in clinical trials (3,4), emphasizing the need for a more detailed understanding of therapeutic mechanisms and for more specific techniques to monitor success. To avoid the complexity and interindividual heterogeneity associated with animal models of myocardial ischemia, we chose for this initial methodologic evaluation to study effects of VEGF gene transfer in a stable model of surgical injection into healthy myocardium. Our results show that VEGF, when successfully overexpressed in normal heart muscle, increases regional myocardial perfusion. The CT morphology helped in accurately identifying the area of intervention and in localization of perfusion increase. It is tempting to speculate that lack of morphologic information to accurately identify injection sites may have resulted in a lack of observation of physiologic in vivo effects of gene transfer in a previous study with stand-alone microPET (15).
Interestingly, expression of αvβ3integrins did not seem to play a role in the VEGF-induced increase of perfusion early after gene transfer in our study. αvβ3Integrin is a group of adhesion molecules which has been targeted previously by molecular imaging as playing a role in post-myocardial infarction angiogenesis (16). The tracer used in the present study has been extensively validated as a marker of αvβ3integrin expression in previous studies (7,17,18).
The sum of the effects of VEGF on healthy myocardium observed in the present study are most likely explained as follows. Our imaging studies were performed a few days after adenoviral gene transfer, at a time when expression of transgenes is known to be highest. Tissue levels of VEGF therefore reached a peak at this early time. Vascular endothelial growth factor is known to be a potent vasodilator, because it activates endothelial nitric oxide synthase and thus increases nitric oxide (NO) levels (19). Additionally, VEGF is known to increase vascular permeability through formation of intercellular gaps, vacuoles, and fenestrations, an effect partially mediated via NO release (20). The observed increase of perfusion at rest and during pharmacologic vasodilation is therefore most likely explained by these direct effects of overexpressed VEGF on the microvasculature. Vasodilation and increased microvascular denisty have also recently been established as important NO-mediated precursors of angiogenesis (21). Significant amounts of new blood vessel are not expected to have formed at this early time, and the mild increase of histologic microvascular density may be associated with vasodilation and thus improved detectability of microvessels. Expression of integrin adhesion molecules may not yet play a role or may require the strong stimulus of ischemic myocardial damage suggested in previous studies (16) for detectability, which was not present in our setup of healthy myocardium. Although this hypothesis for an underlying mechanism of our observations is at present speculative, it will be interesting to determine its validity in future molecular imaging studies of disease models. Notably, quantitatively assessed global microvascular reactivity was low (ratio stress/rest MBF >2) compared with values obtained in other animal and human studies. But this observation was true for vector-injected and sham-operated animals. It is therefore not specific for the molecular intervention and may rather be attributed to surgical or anesthetic procedure.
In conclusion, our study provides proof of feasibility for integrated multislice PET-CT to dissect cardiac biologic mechanisms following a molecular intervention. The technique seems ready for serial application in other animal models and models of cardiac disease or new therapies. Vascular endothelial growth factor gene transfer, or novel refined techniques for angiogenesis gene therapy (e.g., more complex models of short- and long-term ischemia) can be evaluated repetitively at different stages, and the obtained biologic and morphologic data may help in refining the clinical usefulness of therapy. Similarly, other molecular therapeutic approaches in the fields of cardiac gene therapy, cell transplantation, or tissue engineering can be evaluated noninvasively using this powerful biomorphologic imaging technique.
We thank the PET center, cyclotron unit, and animal care unit of TU München for assistance in radiotracer production, image acquisition, and animal handling.
Supported by a grant from the Deutsche Forschungsgemeinschaft (Be 2217/4-1).
- Abbreviations and Acronyms
- replication defective type 5 adenovirus expressing a mutant herpesviral thymidine kinase reporter gene
- replication defective type 5 adenovirus co-expressing a mutant herpesviral thymidine kinase reporter gene and the human VEGF121gene
- replication defective type 5 adenovirus expressing the human VEGF121gene
- computed tomography
- myocardial blood flow
- positron emission tomography
- vascular endothelial growth factor
- Received July 31, 2006.
- Accepted August 30, 2006.
- American College of Cardiology Foundation
- Forrester J.S.,
- Price M.J.,
- Makkar R.R.
- Grines C.L.,
- Watkins M.W.,
- Mahmarian J.J.,
- et al.
- Kastrup J.,
- Jorgensen E.,
- Ruck A.,
- et al.
- Anton M.,
- Wittermann C.,
- Haubner R.,
- et al.
- Miyagawa M.,
- Anton M.,
- Haubner R.,
- et al.
- DiFilippo F.P.,
- Brunken R.C.
- Koepfli P.,
- Hany T.F.,
- Wyss C.A.,
- et al.
- Namdar M.,
- Hany T.F.,
- Koepfli P.,
- et al.
- Wu J.C.,
- Chen I.Y.,
- Wang Y.,
- et al.
- Pichler B.J.,
- Kneilling M.,
- Haubner R.,
- et al.
- You D.,
- Waeckel L.,
- Ebrahimian T.G.,
- et al.