Overcoming Limitations in Nanoparticle Drug Delivery: Triggered, Intravascular Release to Improve Drug Penetration into Tumors

A major challenge in chemotherapy is to change the biodistribution of drugs to reduce free drug toxicity and favor tumor accumulation. Attempts to overcome this challenge have focused on nanoparticles, such as liposomes (1, 2) and other self-assembling systems (3–5). The hyperpermeability of tumor vasculature led to the hypothesis that liposomes and other particles could accumulate more specifically in tumor tissue than normal tissue by passive extravasation; the process was termed the enhanced permeability and retention (EPR) effect (6, 7). As an extension, additional tumor specificity could be achieved with active ligand targeting. While these methods have been shown to increase the amount of drug delivered, there is growing evidence (6, 8–10) that many nanoparticle delivery systems are too large to extravasate in human tumors. Even when tumor vasculature is permeable to 100-nm liposomes, the distribution pattern is extremely heterogeneous and susceptible to large inter- and intratumoral variances in vascular permeability and ligand expression (9, 11). Furthermore, the relatively large size of nanoparticles limits their penetration depth to 1 to 2 cell layers from blood vessels (6, 8–10, 12).

Nanoparticles also release drug slowly; thus, tumor cells may not be exposed to concentrations high enough to result in cell death (9, 13). Consequently, a gap exists between shielding normal tissue from drug and delivering the drug at therapeutic concentrations to all tumor cells. One solution would be to initiate local drug release from a nanoparticle in the tumor's bloodstream, where it can diffuse into the tissue, down its concentration gradient. This would avoid the reliance on particle extravasation.

We developed thermally sensitive liposomes (Dox-TSL) that rapidly release doxorubicin in response to 40°C to 42°C heat (14–20). Preclinical work showed tumor drug levels up to 30 times higher than free drug and 3 to 5 times greater than traditional liposomes; the enhanced delivery was linked to superior antitumor efficacy (15, 21). Interestingly, timing of liposome injection versus application of hyperthermia altered antitumor efficacy in rat fibrosarcomas (22). When Dox-TSLs were administered after tumors were preheated, tumor drug concentrations were doubled as compared with Dox-TSL administration before hyperthermia. We speculated that the increased drug delivery in the first case was due to intravascular release during peak plasma concentrations of liposomal drug.

This result led to the current hypothesis, the mechanism of drug delivery from Dox-TSL + heat is due to perivascular drug penetration as a result of rapidly triggered intravascular liposomal drug release. This method maintains the toxicity benefits of sequestering drug until it reaches the tumor where it releases drug with a localized trigger. Intravascular release from TSLs overcomes heterogeneity in vascular permeability and limited penetration associated with the EPR effect. In this context, TSLs serve as a continuous intravascular infusion of drug, which originates at the tumor site. This creates high intravascular drug concentrations that drive drug uptake by cells and increase drug penetration further from vessels.

While the intravascular release hypothesis has existed since the early work on TSLs in the 1970s (23–25), it has not been shown in vivo; the enhancement in drug delivery has only been shown by an increase in overall tumor drug concentrations, not drug penetration. Herein, we present the first in vivo evidence of intravascular drug release using intravital confocal microscopy and illustrate that intravascular drug release improves drug penetration to reach more tumor cells than either the EPR effect with pegylated liposomes or with free drug.

Human squamous cell carcinoma (FaDu) cells were grown as monolayers in tissue culture flasks containing minimal essential Dulbecco's Modified Eagle's Medium supplemented with 10% heat-inactivated FBS, penicillin, and streptomycin (Gibco). Cell cultures were kept at 37°C with 5% CO₂ in air. B16BL6 melanomas were transplanted to window chambers from donor animals. Both cell lines were obtained from the American Type Culture Collection.

All animal experiments were carried out in accordance with Duke University (Durham, NC) or Erasmus Medical Center's (Rotterdam, The Netherlands) Institutional Animal Care and Use Committee guidelines. Either nude athymic mice (FaDu tumor model) or eNOS-GFP transgenic mice (B16BL6 melanoma model) were used. The eNOS-GFP transgenic mouse model contains the eNOS-GFP fusion protein expression, restricted to the endothelial cells. Mice were anesthetized and then underwent dorsal window chamber implantation as described previously (26). Additional details are in Supplementary Materials and Methods.

This study used 2 different liposome formulations, the primary TSL formulation, used for evaluating drug kinetics and penetration, and a second formulation prepared in Rotterdam, Netherlands to corroborate intravascular release of drug in eNOS-GFP mice. The first liposome preparation consisted of 99.9 mol% of dipalmitoylphosphatidylcholine (DPPC), 1-stearoyl-2-hydroxy-sn-glycero-3-phosphocholine (MSPC), and 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (DSPE-PEG2000) with corresponding mole percentages of 85:9.8:5.2, along with 0.5 mol% fluorescein N-(fluorescein-5-thiocarbamoyl)-1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine triethylammonium salt (DHPF). TSLs were similarly prepared at the Erasmus Medical Center but with the mole percentages 90:10:4 (DPPC:MSPC:DSPE-PEG2000). Doxorubicin loading was achieved by the remote pH gradient method (27). Further details are provided in Supplementary Methods.

Nude mice with dorsal window chambers were anesthetized [Nembutal; 85 mg/kg intraperitoneally (i.p.)] and positioned on a custom-designed microscope stage and heating device for localized heating to the window chamber between 40.7°C to 41.8°C (or 34°C–36°C in unheated controls; ref. 12). Core body temperature was maintained with a warming pad set at 37°C. The tail vein was cannulated and 0.1 mL of 10 mg/mL 2-MDa rhodamine-labeled dextran (Molecular Probes) was injected i.v. A z-stack of images was collected with a LSM 510 laser-scanning confocal microscope (Zeiss) through approximately 50 to 100 μm of tissue, using a 543 excitation laser and LP 560 emission filter. A plane approximately halfway through the z-stack image was chosen for continual sequential imaging of either doxorubicin and dextran (when free doxorubicin was injected) or doxorubicin and fluorescein-labeled liposomes (when Dox-TSL was injected); both at 6 mg/kg. Images were obtained every 5 seconds for 20 minutes, including 20 seconds of background images. Four treatment groups were evaluated doxorubicin + heat (n = 5), doxorubicin − heat (n = 5), Dox-TSL + heat (n = 6), and Dox-TSL − heat (n = 4).

For visualization of the intravascular doxorubicin release in the eNOS-GFP mice, animals were anesthetized with isofluorane (Nicholas Piramal (I) Limited) and placed on a thermal stage at 37°C. Heating of the chamber to target tumor temperatures of 41°C was accomplished by an external circular resistive electric heating coil attached to the glass of the back side of a glass coverslip (28). The doxorubicin dose was 5 mg/kg. Representative pictures out of 4 independent experiments applying Dox-TSL + heat are shown. Additional details are provided in Supplementary Methods.

Data were preprocessed to segment vascular and extravascular spaces, and doxorubicin fluorescence was normalized and corrected for bleed-through. Additional analysis details are provided in Supplementary Methods. Drug penetration depth was determined by converting the 3D vascular mask from the rhodamine–dextran images into a distance map that related the distance of each pixel from the nearest vascular structure in 3D space. A distance of 0 corresponded to the end of the intervascular space. Boundary effects were removed by truncating the dataset by 50 pixels on all sides in the x–y plane. Fluorescence intensities for pixels located at each distance from the nearest vessel were then averaged.

Athymic nude mice were injected with 1 × 10⁶ FaDu cells into the right leg. Tumors were 8 to 12 mm in diameter when used (ulcerated tumors excluded) in 4 treatment groups Dox-TSL + heat (n = 9), Dox with (n = 7) and without heat (n = 5), and Doxil + heat (n = 8). Mice were anesthetized (85 mg/kg Nembutal i.p.) and fitted with a tail vein catheter and a rectal temperature monitoring probe to measure core body temperature during water bath heating (see Supplementary Methods).

Tumor-bearing legs were preheated for 10 minutes before injection to reach thermal steady state. Mice were then injected i.v. with 15 mg/kg Dox-TSL, free doxorubicin, or Doxil at 2 mg/mL to ensure an injected volume per mouse less than 0.2 mL. Following injection, the animals were heated for an additional 20 minutes, then removed and allowed to cool for 10 minutes before sacrifice (Doxil animals sacrificed 24 hours posttreatment). Tumors were snap frozen over liquid nitrogen.

Entire tumor sections were imaged for doxorubicin and stained and imaged for CD31 using a Zeiss Axioscope 2 scanning stage with ×20 objective and Metamorph image software (see Supplementary Methods).

Data are presented as means with SDs. Statistical differences between groups were computed using the Wilcoxon rank-sum test. AUC_0–20 values are expressed as median ± SDs of the product %I_vasc,max × seconds.

Free drug administration was compared with intravascular release of drug from Dox-TSL. Heat showed little influence on the accumulation of free doxorubicin (red signal), which diffused out of the vascular space (green signal) and into the extravascular space immediately after injection (Fig. 1 and Supplementary Movies S1 and S2). However, declining vascular concentration caused the blood vessel compartment to switch from a drug source to a drug sink, resulting in reabsorption of drug into the vasculature more quickly than it can diffuse and enter cells. This leads to limited tumor cell exposure. Thus, extravascular drug levels quickly declined, with a precipitous drop in apparent drug accumulation within 5 minutes of injection, in line with previous observations (13, 29). In Fig. 1, free drug reabsorption into the vasculature is reflected by a decrease in drug (shown in red) in the extravascular region. As the reabsorption occurs more slowly than the blood flow rate in the nearby vessels, any drug that is reabsorbed into the vascular compartment is immediately removed from the tumor area, explaining why the vessels do not seem to have more drug during reabsorption.

Dox-TSL's prolonged plasma half-life maintains high intravascular concentrations, resulting in increased accumulation of drug with time in the interstitial space (Fig. 1 and Supplementary Movies). In the absence of heat, doxorubicin remains contained within the liposome and confined to blood vessels. To investigate intravascular release, fluorescently labeled liposomes were imaged to determine the contribution of extravasation versus intravascular release (Fig. 2). Overlap between green lipid label and red fluorescence of doxorubicin is seen as yellow. The intravascular signal remains strongly yellow throughout the 20 minutes, whereas the red signal increases in the interstitial space. In contrast, there is very little yellow color in the interstitial space during heating. The lack of overlap between lipid and doxorubicin fluorescence in the interstitial space proves that drug accumulation in the interstitial space is due to release from liposomes still confined in the bloodstream. Intravascular release of doxorubicin and subsequent redistribution into tumor tissue is further illustrated with B16 melanoma using the eNOS-GFPtg mouse model (Fig. 3 and Supplementary Movie S2).

For each method of drug delivery, vascular clearance and extravascular accumulation of drug were quantified with image analysis (Fig. 4). Intravascular release results in longer tumor cell exposure at higher drug concentrations as compared with free drug. Free drug administration, ±heat, results in peak extravascular drug levels within 1 minute. Thereafter, drug levels in the interstitial space decline as vascular concentration decreases. The drop in vascular concentration is a reflection of elimination and distribution of drug to other organs. Within 20 minutes, residual drug levels in tumor were reduced to 9.6% ± 2.0% of the original peak intravascular drug levels (I_vasc,max) in heated tumors and 13.7% ± 5.7% in nonheated tumors. In contrast, the high intravascular drug levels maintained by the rapidly releasing liposomes allow for sustained drug accumulation in the extravascular space. At 20 minutes, drug is continuing to accumulate, with extravascular levels at 160% ± 47% of the peak vascular drug level, 11- to 17-fold higher than free drug.

The area under the concentration–time curve (AUC) for the extravascular space provides a measure of the cumulative exposure of tumor cells to drug (12, 13). Within the first 20 minutes, AUC_0–20 values of free drug with and without heat are not statistically different, 156.8 ± 41.7 and 185.2 ± 160.8, respectively (P value = 0.6905). The AUC_0–20 value for heated nanoparticles was 9-fold higher than free drug + heat, 1700 ± 863.1 (P value = 0.0317). After 20 minutes of heating, the AUC value for the Dox-TSL treatment group was still increasing due to continual accumulation of drug, whereas AUC increases in the free drug + heat group with each additional minute are minimal due to very low intravascular drug levels. This AUC_0–20 value is only indicative of the first 20 minutes of drug delivery during a typical 1-hour heat treatment. Thus, these differences would be magnified for a 1-hour exposure. The ability to maximize gains in AUC will be dependent on the local liposome concentration in the blood supply, underscoring the importance of heating before liposome injection. Starting heat after liposome injection would be expected to result in suboptimal AUC, due to declining vascular levels of Dox-TSL caused by clearance from the bloodstream (30). Maximum AUC could also be affected by the circulation kinetics and stability of the TSL formulation in plasma.

Doxorubicin's poor penetration characteristics have been highlighted in vivo with mouse studies and in patients with breast cancer (31, 32). To determine if intravascular release resulted in enhanced drug penetration in addition to accumulation, we first used image analysis to quantify relative drug concentrations with distance away from blood vessels (Fig. 5). Administration of free drug exhibits a characteristic decline in drug concentration with distance from vessels; decreasing intravascular concentrations quickly cause drug to diffuse back toward the vasculature, resulting in relatively low drug retention outside the microvessels. At distances of 35 μm from vessels (3–4 cell layers), the maximum amount of drug delivered within the first few minutes was only 31% ± 7% of the peak intravascular drug level (I_vasc,max), and at 20 minutes after injection, the amount of drug remaining at this distance declined to 10% ± 2%. Drug penetration for free drug without heat was similar, 28% ± 5% at maximum and 14% ± 4% at 20 minutes.

Drug delivered with heat-triggered liposomes was continuous throughout the period of heating, resulting in increased levels of drug out to 35 μm. This penetration was the furthest possible depth measureable in the window chambers, given the heterogeneity of vessels across all groups and tumor regions. In contrast to the 10% residual drug at 35 μm from vessels after 20 minutes with free drug, Dox-TSL drug levels at this same distance were 138% ± 45% of I_vasc,max (P value = 0.0087). In addition, drug accumulation at the endothelial cell barrier seemed to exhibit saturation kinetics; the drug level present at the endothelial cell edge did increase, but to a lesser extent than at further distances from microvessels. Drug accumulation at the endothelial cell edge changed from 88% ± 12% of I_vasc,max at time 0 to 126% ± 32% of I_vasc,max at 20 minutes. However, at distances 10 μm from the vasculature, the increase was more dramatic, from 41% ± 16% at time 0 to 180% ± 46% at 20 minutes. Thus, the sustained intravascular release of drug may aid in saturating endothelial and tumor cell drug transport mechanisms, providing a method to enhance perivascular penetration. Maintaining high intravascular drug levels led to extravascular drug concentrations that exceeded maximal plasma levels. This is most likely due to cellular uptake of the drug, which removes drug from the interstitial space, thus maintaining the concentration gradient. Indeed, nuclear accumulation of drug is readily seen in the confocal images and is consistent with our prior observation of increased DNA-bound drug for the Dox-TSL formulation, compared with free drug or nonthermally sensitive liposomes (15). This ability to locally deliver drug concentrations that exceed vascular concentrations cannot be achieved with free drugs or nonthermally sensitive pegylated liposomes, where the extravasation is dependent entirely upon pharmacodynamics and/or the EPR effect, respectively (8, 33, 34).

To further define the enhanced drug penetration from Dox-TSL, we conducted histologic analysis of flank FaDu tumors on mice that were given Dox-TSL + heat, free drug ± heat, or Doxil + heat. Tumors were harvested 30 minutes after heating for all treatment groups except the Doxil + heat group, which was harvested 24 hours posttreatment to directly compare the Dox-TSL penetration distance with the maximum nanoparticle accumulation typically seen with the EPR effect (8, 33, 34). Tissue sections were imaged for the vascular marker CD31 and doxorubicin, and then analyzed using custom Matlab code to determine the median penetration distance of drug from vessels over the entire tumor section. With free drug, median drug penetration was 29 ± 10 and 55 ± 6 μm, without and with heat, respectively (Fig. 6), which is comparable with published reports for this drug (32). However, the concentration delivered at all distances with free drug was much lower than with either Dox-TSL or Doxil, with maximum fluorescence of free drug being 16% to 25% of the maximum fluorescence closest to vessels with Dox-TSL. A direct comparison was also made between TSL + heat and the traditional liposome Doxil + heat. Dox-TSL + heat resulted in a penetration distance double that achievable with Doxil + heat; drug penetrated 78 ± 18 μm compared with 34 ± 9 μm, respectively (P value = 0.0106). In addition, at the point where drug levels from Dox-TSL + heat started to decrease below the level immediately proximal to the vasculature, drug levels had already declined to 36% of their initial value for the Doxil + heat group. Thus, the confocal and histology data together show that liposome-mediated intravascular drug release not only exposes tumor cells to a higher drug concentration for a longer time period but also improves tumor exposure to drug at every distance from vessels. Most importantly, the overall achievable drug penetration is doubled with intravascular release as compared with free drug or the EPR effect seen with Doxil + heat.

While traditional chemotherapeutic drugs, such as doxorubicin, cisplatin, and paclitaxel have dose limiting toxicities, new molecularly targeted cancer therapies more specifically interfere with cancer cell proliferation and metastasis, and have the potential to be more effective and less harmful to normal cells. Yet, the challenge for these molecular therapeutic agents is still fundamentally the same—limited delivery. Lack of drug penetration or sufficient concentration to result in tumor toxicity has been highlighted using multiple small-molecule chemotherapeutics (32, 35–37), and the newer fleet of monoclonal antibodies and small-molecule inhibitors would be expected to display these same penetration-limited pharmacodynamics (12, 36). The larger the size of the drug, the more tumor penetration becomes a fundamental limitation. Lack of drug penetration becomes even more significant when using nanoparticles, as their perivascular accumulation and slow release pose a severe hindrance to drug delivery. The results here show that even with hyperthermia, the maximum penetration of Doxil is limited to 34 μm; other studies not using hyperthermia have reported even more limited penetration of 8 to 16 μm (10).

Treatment with TSL + heat to local tumors can be thought of as the fusion of 2 traditional approaches to drug delivery: (i) continuous infusion of drug and (ii) nanoparticle encapsulation. Drug infusions maintain elevated drug levels in the bloodstream, and thus prevent drug washout from the tumor before cellular uptake (13). However, this approach does not provide any tumor selectivity. Alternatively, nanoparticles are typically designed to shield drug from normal tissue, but this strategy sacrifices drug bioavailability. Our results present the union of these 2 ideas—using a liposome to shield the normal tissue from toxicity and incorporating local release in the tumor vascular supply, effectively serving as a continuous infusion that originates at the tumor site. While this idea was previously postulated for TSL (24, 25, 38), these results are the first in vivo evidence to prove that intravascular release is the predominant method of drug delivery. Furthermore, these results show that this approach exposes tumor cells to drug over a longer time frame, delivers drug at a higher concentration at all distances from blood vessels, and doubles the maximum penetration distance of drug into tumor tissue. It is also important to note that the window chamber data for Dox-TSL + heat treatment clearly show that drug is being taken up into tumor cell nuclei for both tumor types studied. We previously showed that the percentage of DNA-bound doxorubicin with Dox-TSL is substantially greater than what can be achieved with a Doxil-type of liposome formulation (15). The Doxil formulation is not thermally sensitive; we have shown that Doxil does not exhibit enhanced drug release in the mild hyperthermia temperature range (41°C–45°C; ref. 16). Thus, we would not expect significant nuclear uptake of doxorubicin during heating with Doxil. This recapitulates the concept that Dox-TSL intravascular drug release/delivery provides a venue for bioavailable drug delivery, as compared with Doxil, where the drug remains entrapped inside the liposome after delivery, where subsequent release occurs over days to weeks.

Since the formulation and efficacy publications for Dox-TSL in 2000 (15, 16), several studies have been published, including additional preclinical studies (21, 22), a canine clinical trial (39), and pharmacokinetic data from a human clinical trial (30). A comprehensive review of these studies was recently published (40).

Dox-TSL has also been tested in 3 phase I clinical trials. Results of a phase I trial for patients with primary and metastatic tumors of the liver treated with the combination of radiofrequency ablation (RFA) and Dox-TSL have been reported (41). This dose escalation study involved a 30-minute infusion of Dox-TSL given 15 minutes before RFA. A statistically significant difference in the time to treatment failure was observed between the patients receiving at least the maximum-tolerated dose (MTD) and patients receiving less than the MTD (374 vs. 80 days, respectively). The authors concluded that the combination of RFA and Dox-TSL was safe and likely more efficacious than RFA alone. These results formed the rationale for a multinational phase III trial for treatment of hepatocellular carcinoma (NCT00617981; ref. 42). Accrual was recently completed and follow-up is in progress. A report of this trial is expected within the next year.

A phase I trial has been completed and a phase II trial has been initiated for the treatment of recurrent chest wall disease in breast cancer patients with the combination of Dox-TSL and hyperthermia (NCT00826085; ref. 43). A prior phase I study was conducted at Duke University, and a article combining these 2 studies is being prepared for submission. Because of the broad range of doxorubicin antitumor efficacy, Dox-TSL has the potential to be used to treat multiple other cancer types.

Further research into intravascular release methods will be dependent on the ability to localize the intended stimulus, in this instance, application of hyperthermia to the tumor. Advances in the application of hyperthermia over the last decade have already made this a clinical reality for localized tumors (44–46). This article presents the idea of fast-triggered release, which may not necessarily be limited to stimulus by hyperthermia, however. With further advances in nanoparticle technology, this same mechanism may equally be achieved using other trigger stimuli, including ultrasound, light, radiation, or potentially even magnetic fields. Furthermore, the ability to deliver drug to tumor at vascular penetration distances twice as far as previously achievable may result in increased benefit from microenvironmentally targeted drugs, such as hypoxic cytotoxins, which have shown dramatic in vitro results but often fail to deliver the same results in vivo due to lack of penetration into hypoxic areas (47, 48).

Drug penetration distances are important clinically; for example median intercapillary distances in cervix cancer have been reported to be 167 μm (49). These distances were also shown to be prognostically important, likely in relation to drug delivery or hypoxia. Thus, data from this work suggest that TSLs can deliver drug to most tumor cells in such patients, considering that the 78-μm drug-diffusion distance is on both sides of the capillary bed. This not only provides an argument for further investigation into various nanoparticle triggers but also allows some drugs to potentially be revisited if their efficacy has in the past been limited by drug diffusion.

D. Needham is the inventor of Dox-TSL and has ownership interest (including patents) in a patent. Dox-TSL was licensed from Duke University by Celsion Corporation. M.W. Dewhirst and D. Needham are former consultants for Celsion Corporation and M.W. Dewhirst owns stock, has a commercial research grant, and ownership interest (including patents) in the company. L.H. Lindner has ownership interest (including patents) in the patent on long-circulating thermosensitive liposomes based on phosphatidyloligoglycerols (WO 2002/064116 und WO 2004/026282) and is a consultant/advisory board member of MedTherm. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.A. Manzoor, L.H. Lindner, A.J. Simnick, S. Das, G. Hanna, G.A. Koning, D. Needham, M.W. Dewhirst

Development of methodology: A.A. Manzoor, L.H. Lindner, A.J. Simnick, M.R. Dreher, S. Das, G. Hanna, G.A. Koning, M.W. Dewhirst

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.A. Manzoor, L.H. Lindner, G. Hanna, W. Park, G.A. Koning, T.L.M. ten Hagen, M.W. Dewhirst

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.A. Manzoor, L.H. Lindner, M.R. Dreher, S. Das, G.A. Koning, T.L.M. ten Hagen, D. Needham, M.W. Dewhirst

Writing, review, and/or revision of the manuscript: A.A. Manzoor, L.H. Lindner, C.D. Landon, J.-Y. Park, M.R. Dreher, S. Das, A. Chilkoti, G.A. Koning, T.L.M. ten Hagen, D. Needham, M.W. Dewhirst

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.A. Manzoor, C.D. Landon, J.-Y. Park

Study supervision: A.A. Manzoor, G.A. Koning, D. Needham, M.W. Dewhirst

The authors thank Pavel Yarmolenko and Andrew Fontanella for their help with various aspects of the Matlab coding.

This work was supported by funding provided by grants from the NIH/National Cancer Institute CA42745-21, 22, 23, NIH/National Center for Complementary and Alternative Medicine 1F31ATT006644-01, and Department of Defense W81XWH-09-1-0115. Financial support was also provided by the Center for Interventional Radiology, the intramural research program of the NI H, and by a research grant from the “Dr. Mildred Scheel Stiftung fuer Krebsforschung (Deutsche Krebshilfe e.V.).”

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.