Detection and prediction of drug delivery to the tumor interstitium are of critical importance in cancer chemotherapy. Prediction of drug delivery derived from standard pharmacokinetic models is frequently inadequate because of the complex nature of tumor blood flow and the microenvironment. Although drug concentrations can be directly sampled with microdialysis or in biopsy samples, we currently lack methods capable of detecting and/or predicting drug delivery to tumors noninvasively. In this study, we describe a novel magnetic resonance (MR) technique to directly detect the drug, and we present the correlation between delivery of drug and the delivery of MR contrast agents to the tumor. Experiments were performed with tumor xenografts in severe combined immunodeficient mice. Three-dimensional maps of the drug distribution within the tumors were obtained with 13C spectroscopic MR imaging with a spatial resolution of 2 × 2 × 2 mm, using signals of the 13C-labeled anticancer agent phenylacetate. Three-dimensional maps of uptake of gadolinium-diethylenetriaminepentaacetic acid (GdDTPA) contrast agent were obtained for the same tumors using dynamic MR imaging. Experimental data were analyzed for correlation between delivery of the drug and the contrast. Histological analysis was performed for excised tumors. Experimental data demonstrated a significant spatial correlation (r = 0.59 with P < 0.001) between the parameter representing delivery of the contrast to tumor interstitium, determined from the kinetic curves of GdDTPA, and integral tissue drug concentrations for two different tumor models. The method is designed to probe extravasation of the drug molecules from the bloodstream into the tumor interstitium. Although therapeutic efficiency of the drug will also depend upon drug retention in the tumor and the ability of the molecules to cross cellular membranes, inefficient drug transfer from plasma to tissue can be a major impediment in achieving effective tumor chemotherapy. The results of this study demonstrate that the uptake kinetics of a low molecular weight MR contrast agent can be used to predict delivery of drug molecules of similar size to the interstitium of solid tumors.

The effectiveness of chemotherapy of tumors can be severely impaired by limited delivery of the drug to the tumor because of low and heterogeneous blood flow (1) and for large molecules because of elevated intratumoral pressure (2). Thus, drugs with high in vitro efficiency often fail to exhibit sufficient activity against solid tumors in vivo(3) because of inefficient transfer of drug from plasma to the tumor interstitium (4). Drug pharmacokinetics within the tumor environment depend upon a combination of physiological and drug-related parameters such as degree of tumor vascularization, efficiency of drug delivery by blood flow, and drug stability during transport to and within the tumor physiological environment. Drug transport across the cellular membrane depends on the hydrophobic properties of the drug and, for small charged molecules, on pH gradients across the membrane (5).

Direct measurements of intratumoral drug concentrations are difficult to perform in vivo, and traditional pharmacokinetic models cannot account for the complexities of the tumor microenvironment and blood flow. Although tissue drug concentrations within the tumor can be sampled with a microdialysis system (6), noninvasive alternatives for detection or prediction of drug delivery to tumors are yet to be developed. PET3 imaging of radiolabeled drugs can provide a possible solution to the problem but is limited by the low spatial resolution typical for PET scans and the high costs involved in synthesis of labeled drugs and imaging studies (7, 8). MR spectroscopy is the only truly noninvasive technique, other than PET, which can detect a compound of interest in the tissue. With recent advances in MR techniques, we can now noninvasively measure the distribution of certain anticancer agents within a solid tumor and simultaneously characterize, with contrast enhanced MRI, the vascular parameters upon which drug delivery is dependent. We used this approach to test the ability of paramagnetic contrast agent-enhanced MRI to predict the delivery of drugs to solid tumors, using parameters of tumor microcirculation derived from the MRI data for experimental animal tumor models.

An inherent problem of MR spectroscopy is its low signal:noise ratio of detection, in comparison with nuclear medicine methods. However, recent results, obtained by our group and others, show that certain anticancer agents administered at clinically relevant doses can be detected in solid tumors in vivo by MR spectroscopy (1, 9, 10). These drug pharmacokinetic studies have been performed with 19F, 13C, and 1H MR spectroscopy and imaging. 19F MR spectroscopy in vivo experiments demonstrated that 5-fluorouracil and its major metabolites and catabolites can be detected in animal and clinical experiments (11, 12, 13). Proton MR spectroscopy was performed for iproplatin using double quantum coherence selection for editing drug signals from the overlapping lipid resonance (10, 14). Labeling of drug molecules with a 13C isotope is an important alternative that enables their MR detection in vivo(15) with minimal modification of the chemical and biological properties of the drug. Several technical approaches can significantly increase the sensitivity of detection of 13C, making the method feasible for in vivo application (15, 16, 17).

Most MR studies directly detecting chemotherapeutic agents in tumors have been performed with agents that can be administered at high doses to allow detection by MR (1). These agents are, therefore, by necessity less toxic and mostly consist of agents that are categorized as cytostatic or preventive. Differentiating agents (18) and specific and nonspecific nonsteroidal anti-inflammatory drugs that have recently attracted significant attention as alternatives for prevention and/or treatment of various cancers (19) fall under these categories. However, most of the clinically used cytotoxic drugs are delivered at doses where tissue concentrations are well below MR detection limits of ∼1 mm. Of these, 5-fluorouracil is the only chemotherapeutic agent that has been extensively studied in patients using MR (11, 12, 13). For such low concentration drugs, it may still be possible to predict drug pharmacokinetics and spatial distribution with high temporal and spatial resolution from the pharmacokinetics and spatial distribution of a paramagnetic MR CA using dynamic MRI (20, 21, 22). The constraints for such an approach would be to ensure that the chemotherapeutic agent and the paramagnetic MR CA have similar transport parameters, such as water/lipid solubility, molecular radius, and macromolecular binding properties, as well as the same route of delivery. In this study, we have demonstrated the feasibility of such an approach by comparing the intratumoral pharmacokinetics of the 13C-labeled differentiating agent PA, measured directly by three-dimensional MRSI, with the delivery of the low molecular weight contrast agent GdDTPA, measured with dynamic T1 MRI and analyzed using Larsson’s model for contrast tissue uptake (23). These studies were performed for two different tumor models with significantly different proliferation rates and doubling times. The experimental results show a significant spatial correlation between the uptake and distribution of the contrast agent and the drug within the tumor interstitium. Such an approach is easily translated to the clinical setting by using MRI of GdDTPA and may be of significant use in determining the delivery of a chemotherapeutic agent to a tumor during the course of chemotherapy.

### Animals.

Male and female severe combined immunodeficient mice (4–6 weeks of age) were used for experiments with prostate and breast tumor xenografts, respectively. All animal experiments were performed in accordance with institutional guidelines. For the MR studies, mice were anesthetized with a mixture of ketamine/acepromazine (40 mg/kg and 4 mg/kg, respectively, in 0.9% NaCl solution) administered i.p. At the end of the study, animals were sacrificed, and the tumors were excised for histology.

### Tumors.

The estrogen-independent human breast cancer cells, MDA-MB-435, were inoculated into the upper thoracic mammary fat pad of female mice (106 cells/0.05 ml of HBSS) and grown for a period of ∼5 weeks. Androgen-independent rat prostate cancer cells, MatLyLu, were inoculated in the flank of male mice (106 cells/0.05 ml of HBSS) and allowed to grow for a period of 10–12 days. The average tumor volume used in this study was ∼500 mm3. MDA-MB-435 tumors reached this size within 30–35 days of inoculation, whereas MatLyLu tumors reached this size within 12 days.

### Chemicals.

13C-labeled PA ([2-13C]PA 99%; Cambridge Isotopes, Inc.) was dissolved in 0.9% NaCl solution with the pH adjusted to 7.2 with NaOH. The final concentration of the filtered injection solution was adjusted to 0.3 mm. Low molecular weight GdDTPA CA (Magnevist; Berlex Laboratory, Wayne, NJ) was used for all of the contrast-enhanced dynamic MRI studies.

### MR Experiments.

All MR studies were performed with a GE Omega-400 instrument equipped with shielded imaging gradients (52-mm room bore diameter) and a home made probe consisting of a double-tuned 1H/13C detection coil (diameter, 13 mm; Ref. 15). Anesthetized animals were placed in the probe with the tumor positioned in the coil. The tail vein of each animal was catheterized with two separate lines to infuse the drug and the CA. A Cole-Parmer single channel animal infusion pump was used to first administer the CA and subsequently the drug. Animal body temperature was stabilized by the flow of preheated air through the body of the probe.

### Dynamic MRI of Contrast Uptake.

Contrast uptake studies were performed with a Turbo-FLASH saturation recovery technique (24, 25). Three-dimensional multislice T1 relaxation maps of the tumors were obtained with two relaxation delays of 250 and 500 ms after the saturation prepulses, with a time resolution of 12 s/three-dimensional map (64 × 64 × 8 matrix size, 1-mm slice, and 16 × 16-mm field of view). An M0 map was measured once at the beginning of the experiment with a relaxation delay of 8 s. Quantitative T1 maps were generated with linear regression analysis from M0 and two relaxation maps (250 and 500 ms) for all pixels. Experiments were performed before and for 9 min after administration of GdDTPA injected as an i.v. rectangular bolus (0.2 mmol/kg over 10 s).

### MRSI of Drug Uptake.

A three-dimensional 13C spectroscopic imaging sequence with proton NOE signal enhancement and broadband proton decoupling was used for acquisition of three-dimensional maps of drug distribution in tumors. The pulse sequence was derived from a standard three-dimensional CSI experiment (26) with a single excitation pulse. The flip angle of the pulse was set to the Ernst value (27), and the phase encoding time was chosen to be 500 μs to optimize sensitivity of signal detection. Three-dimensional spectroscopic imaging maps were acquired for a field of view of 16 × 16 × 16 mm with 8 × 8 × 8-phase encoding steps. Spectroscopic imaging was performed with time averaging of the signals, i.e., the complete map of three-dimensional phase encoding was completed for each scan, and the entire experiment was repeated eight times for a total of 8 × 8 × 8 × 8 = 4096 scans. This mode of acquisition produces spectral intensities proportional to the integral of the measured signal over the time of the experiment. Drug infusion was initiated at the beginning of experiment and was continued for 7 min for a total injected dose of 650 mg/kg. With an interpulse delay of 0.5 s, this resulted in a total experimental time of ∼40 min.

Drug detection experiments were performed immediately after completing the contrast uptake studies, without changing the position of the animal in the probe, to facilitate coregistration of data sets. A schematic diagram of the pulse sequences used in the study are shown in Fig. 1.

### Data Processing.

After completion of MR experiments, data processing was performed off line on a Silicon Graphics Octane workstation using IDL software (Research Systems, Inc., Boulder, CO).

### Contrast Uptake.

Tissue concentrations of the paramagnetic CA GdDTPA C(t) were determined from the multislice T1 maps obtained before and for 9 min after administration of the contrast using the equation

$$\mathit{C(t)}\ {\propto}\ {[}\mathit{T}_{1}^{{-}1}\mathit{(t)}\ {-}\ \mathit{T}_{1}^{{-}1}\ (0)$$
⁠, where fast exchange conditions are assumed. Concentration versus time curves were analyzed using the equation for Larsson’s model (23):

$C(t)\ {=}\ k^{PS{\rho}}_{in}\frac{{{\int}_{0}^{t}}}{}C_{a}({\tau}){\times}exp(k^{PS{\rho}}_{out}({\tau}{-}t)/v_{e})d{\tau}$

where kPSρin is the influx volume transfer constant [min−1], kPSρout is the “outflux constant,” and ve is the extravascular extracellular space per unit volume of the tissue. In a simplified model of contrast uptake (20) the volume transfer constants are assumed equal, kPSρin = kPSρout. In tumors where delivery of the CA can be flow limited, this simple relation may not hold; also, parameter ve is not known. Therefore, we used a two-parameter fit of the equation (1) and determined two independent parameters k1 and k2 proportional to the volume transfer constants kPSρin and kPSρout, respectively. In the rest of the paper, parameters k1 and k2 will be used as indices of contrast uptake and clearance. The arterial input function Ca(t) used in the model was independently measured in vitro using samples of arterial blood from the catheterized carotid artery and averaged for three animals. Three-dimensional maps for the parameter k1 were generated on a pixel by pixel basis for the entire data set and were interpolated to a 16 × 16 × 16 data set corresponding to a spatial cube with a side length of 16 mm.

### Drug Delivery.

Three-dimensional MRSI data were filtered with a Gaussian window in the spatial domains and an exponential window in the time domain and Fourier transformed using absolute value calculation. The integral of the [2-13C]PA peak at 45 ppm was calculated for each voxel of the three-dimensional data set, and the final three-dimensional image of integral drug concentrations was interpolated to the 16 × 16 × 16 matrix corresponding to the 16 × 16 × 16-mm cube. The drug uptake data set was registered with the data set for the contrast uptake parameter k1 for all animals, using translation in three-dimensional space. Because MR data sets were acquired without disturbing the position of the animal, the maximum translation length was limited to ± 2 data points in the X, Y, or Z direction.

### Histology.

After completing MR studies, the animals were sacrificed, and tumors were excised, fixed in 10% buffered formalin at pH 7.4, dehydrated, and embedded in paraffin blocks. Up to 10 5-μm sections were cut to provide a uniform coverage of the whole tumor. All sections were deparaffinized and stained with H&E to determine the fraction of viable and necrotic cells. Optical images were digitized at high resolution with a CCD camera (Sanyo, 1/3″) attached to an Olympus microscope.

### Statistical Analysis.

Correlation between drug delivery and contrast uptake data sets was determined with Pearson’s correlation. The significance of correlation was evaluated using a two-tailed significance test. Association between delivery of the drug and the CA was assessed with linear regression analysis using Student t test with (n − 2) degrees of freedom for the regression parameter b.

Typical results for the curve fitting of the CA to Eq. A in a single voxel of a three-dimensional image are shown in Fig. 2. The arterial input curve of the CA used in the convolution integral is also shown on the plot. The best fit was produced with a Powell nonparametric routine implemented in the IDL language. The appearance of the [2-13C]PA peak in the nonlocalized spectra of the tumor, during the course of drug infusion, is shown in Fig. 3. Broadband decoupled NOE 13C spectra were acquired with 128 scans/spectrum with a 1-s repetition delay. Reconstructed images of the contrast uptake parameter k1 and integral drug concentration for a MatLyLu and an MDA-MB-435 tumor are shown in Fig. 4 for eight slices obtained with identical spatial localization. Spatial registration of the proton MR images corresponding to the contrast uptake and 13C spectroscopic images for drug delivery was performed using an interactive three-dimensional volume rendering routine using small adjustments of linear offsets. A color presentation of three-dimensional data sets of contrast uptake and drug concentration within the MatLyLu tumor is shown in Fig. 5 using the green channel for the CA and the red channel for drug. On the composite image, yellow regions correspond to areas where high contrast uptake spatially correlates with high drug concentrations. A high degree of spatial correlation between tissue concentration of the drug and contrast uptake is evident in the image.

Correlation analysis of the spatial distribution of contrast and drug uptake was performed for 6 MatLyLu and 4 MDA-MB-435 tumors. Individual results are presented in Table 1 with averaged values of the Spearman’s correlation coefficient of RMLL = 0.62 for MatLyLu tumors and RMDA = 0.55 for MDA-MB-435 tumors. Both results are statistically significant with P = 0.001. Data sets for both drug and contrast uptake contain significant number of zero points, which can produce erroneously high correlation when taken into account. Therefore, zero points were excluded from the correlation analysis of the experimental data.

The significant correlation between parameters of contrast and drug uptake allowed us to analyze the experimental data using linear regression analysis. Results for the individual tumors are summarized in Table 1. Averaged values of regression coefficients for the two tumor models studied are: bMLL = 156 ± 20 and bMDA = 89 ± 13, measured in relative units, where the errors represent the SE. Scatter plots of the data and the corresponding linear regression line are shown in Fig. 6 for representative data sets for a MatLyLu tumor and an MDA-MB-435 tumor. As for the correlation analysis, zero points were excluded from the linear regression analysis.

A comparison of contrast and drug uptake maps with a histological section obtained from this region for an MDA-MB-435 tumor is shown in Fig. 7. Both contrast and drug uptake were significantly reduced in the central region of the tumor, which corresponded to a region of central necrosis in the histological section.

The data demonstrated a significant correlation between the delivery of low molecular weight MR CA and the integral concentration of a small molecular weight drug, irrespective of the origin, inoculation site, and growth rate of the tumor. Tissue uptake of the CA was assessed by the parameter k1, which is a complex function of the vascular characteristics of the tissue determined as the index of GdDTPA uptake. For fast water exchange across the vascular wall typical for tumors, the parameter k1 is proportional to

$$\mathit{E\ {\cdot}\ F}\ {=}\ \mathit{F}\ (1\ {-}\ \mathit{exp\ (PS/F)})$$
⁠, where F[ml/g min] is the capillary blood flow per unit mass of tissue, P[cm/min] and S[cm2/g] are permeability and surface area per gram of tissue, respectively, and E is the extraction fraction (the fraction of tracer that leaves the blood and enters the tissue during one pass of blood through the capillary bed; Ref. 20). Generally, for a given plasma concentration, drug delivery to target tissue and its clearance from the interstitium are primarily dependent on tissue perfusion and the micropermeability of the vasculature to the drug molecules (28). Interstitial pressure gradients that may exist within the solid tumors are of lesser importance for the delivery of low molecular weight substances, because the transport of molecules from the capillaries to the interstitium is mainly by diffusion, for compounds with molecular weight Mr <2000 (2, 29). Thus, for a chemically stable drug with a defined arterial input function, the accumulation of drug in the tissue will be highly dependent upon the product of regional blood flow rate F and the extraction fraction of the drug E, which is directly proportional to the index of contrast uptake k1. The pharmacokinetics at later time points or retention of the drug in the tissue may differ from that of GdDTPA because of differences in the volume of distribution because drug molecules can traverse the cellular membrane, whereas GdDTPA is an extracellular agent. Differences in different clearance rates and possible metabolic conversion of drug molecules may also lead to differences in pharmacokinetics between GdDTPA and the drug molecules at later time points. The contrast uptake index k1 relates to the initial uptake rate of the CA and therefore does not depend on its clearance k2. Parameter k1 is a measure of the transport of small molecules across the vascular wall into the tumor interstitium. It can be measured with high spatial resolution using contrast enhanced dynamic MRI and can therefore be used as a “pharmacoangiographic” marker to predict the delivery of drug molecules.

To prove the feasibility of this approach, in our study the uptake of the CA was compared directly with MR measurements of tissue concentrations of [2-13C]PA drug detected with high sensitivity in the same tumor. PA belongs to a group of aromatic fatty acid compounds known to induce differentiation in a wide range of human tumors (18, 30, 31, 32). The mechanisms by which the agent reduces cell proliferation and induces re-expression of genes silenced in cancer are still not clear. Some of the often synergistic mechanisms proposed include: (a) depletion of serum glutamine by conjugating PA (33); (b) inhibition of the mevalonate pathway which interferes with synthesis of sterols and isoprenoids; (c) inhibition of isoprenylation of ras and related proteins (34); (d) inhibition of histone deacetylase (35) and DNA methylation (36); and (e) activation of human peroxisome proliferator-activated receptors (37, 38). Phase I clinical trials revealed that doses of PA >20 g/day (250 mg/kg) are well tolerated by the patients (39). The drug concentration used in our studies with mice (650 mg/kg) was well tolerated, and the animals completely recovered after the experiments, without signs of neurosuppression. The concentration used provided sufficient MR signal to enable in vivo drug detection with a spatial resolution of 8 mm3.

For all animal studies, contrast uptake measurements were performed before 13C CSI of the drug. 13C spectroscopic imaging studies were commenced within 15 min of completing GdDTPA-enhanced MRI, without altering the position of the animal and/or tumor in the radiofrequency coil, to ensure complete registration of the data sets. Our reasons for this particular sequence of experiments were: (a) no interference of residual GdDTPA with 13C detection of signals of the drug was observed in the experiments; and (b) i.v. administration of high dose of the drug could cause changes in tumor perfusion affecting delivery of the CA to the tumor. Contrast uptake was obtained from quantitative pixel-by-pixel T1 maps of the tumor after administration of the CA. In comparison with steady-state type experiments (T1 weighted imaging), this approach provides a linear dependence of the parameter on the tissue concentration of the contrast agent, and the method is not sensitive to changes in T2 and T2* caused by the contrast agent. The trade-off of this technique is the longer acquisition time and lower spatial resolution. To improve the temporal resolution of the method, we developed a saturation recovery Turbo-FLASH sequence for fast acquisition of three-dimensional quantitative T1 maps, which is extensively used in our laboratory for the study of experimental tumors (40). By using saturation recovery, the long relaxation delays between consecutive single-shot acquisitions are not required. Furthermore, T1 maps measured with nonselective saturation of the entire tumor are not prone to fresh blood inflow artifacts. Reduced spatial resolution of a single-shot T1 imaging is not relevant for our studies because it still remains significantly higher than the spatial resolution of 13C spectroscopic imaging.

Two tumor lines with significantly different growth rates and different origins were chosen for these experiments. MDA-MB-435, a hormone-independent, highly metastatic human breast cancer line with relatively slow proliferation rate, was inoculated orthotopically in the mammary fat pad of female mice. On the other hand, MatLyLu, a rapidly growing aggressive rat prostate cancer cell line, was inoculated s.c. Despite these widely different factors, which can result in a significantly different tumor microenvironment, the regression analysis did not reveal statistically significant changes between delivery of the drug and the index of contrast uptake for both tumor lines. A positive intercept of the regression line with the Y axis (drug uptake), corresponding to a measurable drug delivery to the region of the tumor with negligibly low contrast uptake, was observed in most of the model tumors. This result can be rationalized from the facts that: (a) the low spatial resolution of 13C spectroscopic imaging of drug distribution is prone to significant volume averaging effects (41); and (b) the drug uptake maps were recorded with time averaging for a period of ∼40 min. During this time, the diffusion of the drug molecules within the tumor interstitium may give rise to local drug concentrations within areas of otherwise low delivery. As seen from Fig. 6, this contribution is limited to not more than 25% of the maximum drug integral intensity within the tumor.

Histological analyses of regions of low drug and contrast agent concentration demonstrated necrosis in these areas, supporting the possibility that the delivery of even low molecular weight chemotherapeutic agents may be nonuniform as well as limited in solid tumors.

In conclusion, in this study we have demonstrated, for the first time, that the delivery of a low molecular weight drug to a solid tumor can be approximated by measuring the delivery of a similarly sized MR CA. The dependence was established using PA, which can be administered at a high clinical dose and therefore is directly detectable with MR, allowing correlation analyses for individual voxels within the tumor. The close agreement of regression analysis for two widely different tumor models suggests that the uptake and distribution of an MRI CA, such as GdDTPA, can be used to quantitatively predict intratumoral drug concentration and distribution. The limitations of the technique are that: (a) the drug should obey the same tracer characteristics as the contrast agent (GdDTPA in this case); and (b) the method can only predict efficiency of drug delivery to the interstitium of solid tumors and not the drug retention in the tissue. The latter is determined by combination of unique properties of drug molecules such as lipophilicity and possible metabolic modification of the chemical structure of drug molecules. The application of such an approach can be extended even further by developing surrogate nontoxic, GdDTPA-based MRI contrast agents that “mimic” the size and lipophilicity of chemotherapeutic agents used in a particular treatment regimen. The ability to predictively visualize drug delivery to regions of the tumor will play a significant role in evaluating the effectiveness of various chemotherapeutic agents currently in use, because it can be used to rule out uncertainties in treatment outcome attributable to failure of the drug to reach the tumor. Such an approach may also significantly alleviate normal tissue toxicities associated with chemotherapy, because it will be possible to determine whether poor response is attributable to poor delivery rather than resistance of the cells to the drug, in which case it would be more appropriate to improve delivery rather than escalate the dose.

Fig. 1.

A, saturation recovery Turbo-FLASH fast imaging. The saturation block is repeated three times to completely suppress initial magnetization. Phase encoding is rearranged to shift the central phase encoding steps with low gradient values to the beginning of the acquisition block. B, three-dimensional spectroscopic imaging 13C pulse sequence with NOE signal enhancement and BB proton decoupling during acquisition. Flip angle of the radiofrequency pulse is optimized according to the Ernst’s principle.

Fig. 1.

A, saturation recovery Turbo-FLASH fast imaging. The saturation block is repeated three times to completely suppress initial magnetization. Phase encoding is rearranged to shift the central phase encoding steps with low gradient values to the beginning of the acquisition block. B, three-dimensional spectroscopic imaging 13C pulse sequence with NOE signal enhancement and BB proton decoupling during acquisition. Flip angle of the radiofrequency pulse is optimized according to the Ernst’s principle.

Close modal
Fig. 2.

A representative data set demonstrating the fitting of the contrast uptake curve for a typical voxel within the tumor to Larsson’s model. ———, the fitted curve; ----, the independently measured arterial input curve for the contrast agent; •, measured tissue concentration of the CA.

Fig. 2.

A representative data set demonstrating the fitting of the contrast uptake curve for a typical voxel within the tumor to Larsson’s model. ———, the fitted curve; ----, the independently measured arterial input curve for the contrast agent; •, measured tissue concentration of the CA.

Close modal
Fig. 3.

In vivo13C spectra of [2-13C]PA, acquired during the course of i.v. drug infusion at a total dose of 650 mg/kg, in a MatLyLu tumor.

Fig. 3.

In vivo13C spectra of [2-13C]PA, acquired during the course of i.v. drug infusion at a total dose of 650 mg/kg, in a MatLyLu tumor.

Close modal
Fig. 4.

Matched slices extracted from three-dimensional images demonstrating similarities in the distribution of contrast and drug uptake for a MatLyLu tumor (tumor volume, 0.9 cm3) and MDA-MB-435 tumor (tumor volume, 0.61 cm3).

Fig. 4.

Matched slices extracted from three-dimensional images demonstrating similarities in the distribution of contrast and drug uptake for a MatLyLu tumor (tumor volume, 0.9 cm3) and MDA-MB-435 tumor (tumor volume, 0.61 cm3).

Close modal
Fig. 5.

Three-dimensional images of contrast uptake (green) and drug uptake (red) for the MatLyLu tumor shown in Fig. 4, fused and displayed using volume rendering software. The yellow regions in the image correspond to the tumor areas where high contrast uptake spatially correlates with high drug concentrations.

Fig. 5.

Three-dimensional images of contrast uptake (green) and drug uptake (red) for the MatLyLu tumor shown in Fig. 4, fused and displayed using volume rendering software. The yellow regions in the image correspond to the tumor areas where high contrast uptake spatially correlates with high drug concentrations.

Close modal
Fig. 6.

Scatterplot for contrast/drug uptake extracted from registered three-dimensional data sets for the MDA-MB-435 and MatLyLu shown in Fig. 4. Zero points were excluded from calculation of the linear regression line.

Fig. 6.

Scatterplot for contrast/drug uptake extracted from registered three-dimensional data sets for the MDA-MB-435 and MatLyLu shown in Fig. 4. Zero points were excluded from calculation of the linear regression line.

Close modal
Fig. 7.

Maps of contrast uptake (A), drug uptake (B), and H&E stained histological section (C) of the slice from an MDA-MB-435 tumor with a region of central necrosis.

Fig. 7.

Maps of contrast uptake (A), drug uptake (B), and H&E stained histological section (C) of the slice from an MDA-MB-435 tumor with a region of central necrosis.

Close modal

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.

1

The study was supported by American Cancer Society Institutional Research Grant (to D. A.), and in part by NIH ROI CA73850.

3

The abbreviations used are: PET, positron emission tomography; MRI, magnetic resonance imaging; MRSI, magnetic resonance spectroscopic imaging; CA, contrast agent; GdDTPA, gadolinium-diethylenetriaminepentaacetic acid; NOE, nuclear Overhauser effect; PA, phenylacetate.

Table 1

Pearson’s correlation coefficients and linear regression parameters for drug/contrast delivery for the individual tumors

Tumor typeCorrelation RaRegression b
MatLyLu
I 0.62 173 ± 6
II 0.56 106 ± 4
III 0.60 90 ± 3
IV 0.49 157 ± 7
V 0.78 177 ± 4
VI 0.66 68 ± 3
MDA-MB-435
I 0.55 94 ± 4
II 0.51 51 ± 2.2
III 0.61 89 ± 3
IV 0.54 46 ± 2
Tumor typeCorrelation RaRegression b
MatLyLu
I 0.62 173 ± 6
II 0.56 106 ± 4
III 0.60 90 ± 3
IV 0.49 157 ± 7
V 0.78 177 ± 4
VI 0.66 68 ± 3
MDA-MB-435
I 0.55 94 ± 4
II 0.51 51 ± 2.2
III 0.61 89 ± 3
IV 0.54 46 ± 2
a

All values represent significant correlation with P ≤ 0.001 (n ≥ 40).

We thank G. Cromwell for transplanting the tumors.

1
Griffiths J. R., Glickson J. D. Monitoring pharmacokinetics of anticancer drugs: non-invasive investigation using magnetic resonance spectroscopy.
,
41
:
75
-89,
2000
.
2
Stohrer M., Boucher Y., Stangassinger M., Jain R. K. Oncotic pressure in solid tumors is elevated.
Cancer Res.
,
60
:
4251
-4255,
2000
.
3
Twentyman P. R. Predictive chemosensitivity testing.
Br. J. Cancer
,
51
:
295
-299,
1985
.
4
Kristjansen P. E., Brown T. J., Shipley L. A., Jain R. K. Intratumor pharmacokinetics, flow resistance, and metabolism during gemcitabine infusion in ex vivo perfused human small cell lung cancer.
Clin. Cancer Res.
,
2
:
359
-367,
1996
.
5
Stubbs M., Rodrigues L., Howe F. A., Wang J., Jeong K. S., Veech R. L., Griffiths J. R. Metabolic consequences of a reversed pH gradient in rat tumors.
Cancer Res.
,
54
:
4011
-4016,
1994
.
6
Nakashima M., Shibata S., Tokunaga Y., Fujita H., Anda T., Arizono K., Tomiyama N., Sasaki H., Ichikawa M. In-vivo microdialysis study of the distribution of cisplatin into brain tumour tissue after intracarotid infusion in rats with 9L malignant glioma.
J. Pharm. Pharmacol.
,
49
:
777
-780,
1997
.
7
Meikle S. R., Matthews J. C., Brock C. S., Wells P., Harte R. J., Cunningham V. J., Jones T., Price P. Pharmacokinetic assessment of novel anti-cancer drugs using spectral analysis and positron emission tomography: a feasibility study.
Cancer Chemother. Pharmacol.
,
42
:
183
-193,
1998
.
8
Fischman A. J., Alpert N. M., Babich J. W., Rubin R. H. The role of positron emission tomography in pharmacokinetic analysis.
Drug Metab. Rev.
,
29
:
923
-956,
1997
.
9
Artemov D., Bhujwalla Z. M., Maxwell R. J., Griffiths J. R., Judson I. R., Leach M. O., Glickson J. D. Pharmacokinetics of the 13C labeled anticancer agent temozolomide detected in vivo by selective cross-polarization transfer.
Magn. Reson. Med.
,
34
:
338
-342,
1995
.
10
He Q., Bhujwalla Z. M., Maxwell R. J., Griffiths J. R., Glickson J. D. Proton NMR observation of the antineoplastic agent Iproplatin in vivo by selective multiple quantum coherence transfer (Sel-MQC).
Magn. Reson. Med.
,
33
:
414
-416,
1995
.
11
Stevens A. N., Morris P. G., Iles R. A., Sheldon P. W., Griffiths J. R. 5-fluorouracil metabolism monitored in vivo by 19F NMR.
Br. J. Cancer
,
50
:
113
-117,
1984
.
12
Findlay M. P., Leach M. O., Cunningham D., Collins D. J., Payne G. S., Glaholm J., Mansi J. L., McCready V. R. The non-invasive monitoring of low dose, infusional 5-fluorouracil and its modulation by interferon-α using in vivo 19F magnetic resonance spectroscopy in patients with colorectal cancer: a pilot study.
Ann. Oncol.
,
4
:
597
-602,
1993
.
13
Wolf W., Waluch V., Presant C. A. Non-invasive 19F-NMRS of 5-fluorouracil in pharmacokinetics and pharmacodynamic studies.
NMR Biomed.
,
11
:
380
-387,
1999
.
14
He Q., Shungu D. C., van Zijl P. C. M., Bhujwalla Z. M., Glickson J. D. Single scan in vivo lactate editing with complete lipid and water suppression by selective multiple quantum coherence transfer (Sel-MQC) with application to tumors.
J. Magn. Reson. B
,
106
:
203
-211,
1995
.
15
Artemov D., Bhujwalla Z. M., Glickson J. D. In vivo selective measurement of (1–13C)-glucose metabolism in tumors by heteronuclear cross polarization.
Magn. Reson. Med.
,
33
:
151
-155,
1995
.
16
van Zijl P. C., Chesnick A. S., DesPres D., Moonen C. T., Ruiz-Cabello J., van Gelderen P. In vivo proton spectroscopy and spectroscopic imaging of [1–13C]glucose and its metabolic products.
Magn. Reson. Med.
,
30
:
544
-551,
1993
.
17
Terpstra M., Gruetter R., High W. B., Mescher M., DelaBarre L., Merkle H., Garwood M. Lactate turnover in rat glioma measured by in vivo nuclear magnetic resonance spectroscopy.
Cancer Res.
,
58
:
5083
-5088,
1998
.
18
Myers C. E. Differentiating agents and nontoxic therapies.
Urol. Clin. North Am.
,
26
:
341
-351,
1999
.
19
Fournier D. B., Gordon G. B. COX-2 and colon cancer: potential targets for chemoprevention.
J. Cell. Biochem.
,
77
:
97
-102,
2000
.
20
Tofts P. S. Modeling tracer kinetics in dynamic Gd-DTPA MR imaging.
J. Magn. Reson. Imaging
,
7
:
91
-101,
1997
.
21
Su M. Y., Muhler A., Lao X., Nalcioglu O. Tumor characterization with dynamic contrast-enhanced MRI using MR contrast agents of various molecular weights.
Magn. Reson. Med.
,
39
:
259
-269,
1998
.
22
Kovar D. A., Lewis M. Z., River J. N., Lipton M. J., Karczmar G. S. In vivo imaging of extraction fraction of low molecular weight MR contrast agents and perfusion rate in rodent tumors.
Magn. Reson. Med.
,
38
:
259
-268,
1997
.
23
Larsson H. B., Stubgaard M., Frederiksen J. L., Jensen M., Henriksen O., Paulson O. B. Quantitation of blood-brain barrier defect by magnetic resonance imaging and gadolinium-DTPA in patients with multiple sclerosis and brain tumors.
Magn. Reson. Med.
,
16
:
117
-131,
1990
.
24
Haase A., Frahm J., Matthaei D., Hanicke W., Merboldt K-D. FLASH Imaging. Rapid NMR imaging using low flip-angle pulses.
J. Magn. Reson.
,
67
:
258
-266,
1986
.
25
Hoehn-Berlage M., Norris D., Bockhorst K., Ernestus R. I., Kloiber O., Bonnekoh P., Leibfritz D., Hossmann K. A. T1 snapshot FLASH measurement of rat brain glioma: kinetics of the tumor-enhancing contrast agent manganese(III) tetraphenylporphine sulfonate.
Magn. Reson. Med.
,
27
:
201
-213,
1992
.
26
Brown T. R., Kincaid B. M., Ugurbil K. NMR chemical shift imaging in three dimensions.
,
79
:
3523
-3526,
1982
.
27
Ernst R. R., Bodenhausen G., Wokaun A. .
Principles of Nuclear Magnetic Resonance in One and Two Dimensions
,
153
Oxford Science Publications
1990
.
28
Blasberg R. G., Kobayashi T., Patlak C. S., Shinohara M., Miyoaka M., Rice J. M., Shapiro W. R. Regional blood flow, capillary permeability, and glucose utilization in two brain tumor models: preliminary observations and pharmacokinetic implications.
Cancer Treat. Rep.
,
65
:
3
-12,
1981
.
29
Jain R. K. Transport of molecules across tumor vasculature.
Cancer Metastasis Rev.
,
6
:
559
-593,
1987
.
30
Thibault A., Cooper M. R., Figg W. D., Venzon D. J., Sartor A. O., Tompkins A. C., Weinberger M. S., Headlee D. J., McCall N. A., Samid D. A Phase I and pharmacokinetic study of intravenous phenylacetate in patients with cancer.
Cancer Res.
,
54
:
1690
-1694,
1994
.
31
Ferrandina G., Melichar B., Loercher A., Verschraegen C. F., Kudelka A. P., Edwards C. L., Scambia G., Kavanagh J. J., Abbruzzese J. L., Freedman R. S. Growth inhibitory effects of sodium phenylacetate (NSC 3039) on ovarian carcinoma cells in vitro.
Cancer Res.
,
57
:
4309
-4315,
1997
.
32
Samid D., Ram Z., Hudgins W. R., Shack S., Liu L., Walbridge S., Oldfield E. H., Myers C. E. Selective activity of phenylacetate against malignant gliomas: resemblance to fetal brain damage in phenylketonuria.
Cancer Res.
,
54
:
891
-895,
1994
.
33
James M. O., Smith R. L., Williams R. T., Reidenberg M. The conjugation of phenylacetic acid in man, sub-human primates and some non-primate species.
Proc. R. Soc. Lond. B Biol. Sci.
,
182
:
25
-35,
1972
.
34
Danesi R., Nardini D., Basolo F., Del Tacca M., Samid D., Myers C. E. Phenylacetate inhibits protein isoprenylation and growth of the androgen-independent LNCaP prostate cancer cells transfected with the T24 Ha-ras oncogene.
Mol. Pharmacol.
,
49
:
972
-979,
1996
.
35
Lea M. A., Randolph V. M., Hodge S. K. Induction of histone acetylation and growth regulation in erythroleukemia cells by 4-phenylbutyrate and structural analogs.
Anticancer Res.
,
19
:
1971
-1976,
1999
.
36
Samid D. Re. Therapeutic targeting of transcription in acute promyelocytic leukemia by use of an inhibitor of histone deacetylase[see comment].
J. Natl. Cancer Inst.
,
91
:
475
-476,
1999
.
37
Pineau T., Hudgins W. R., Liu L., Chen L. C., Sher T., Gonzalez F. J., Samid D. Activation of a human peroxisome proliferator-activated receptor by the antitumor agent phenylacetate and its analogs.
Biochem. Pharmacol.
,
52
:
659
-667,
1996
.
38
Samid D., Wells M., Greene M. E., Shen W., Palmer C. N., Thibault A. Peroxisome proliferator-activated receptor γ as a novel target in cancer therapy: binding and activation by an aromatic fatty acid with clinical antitumor activity.
Clin. Cancer Res.
,
6
:
933
-941,
2000
.
39
Thibault A., Samid D., Cooper M. R., Figg W. D., Tompkins A. C., Patronas N., Headlee D. J., Kohler D. R., Venzon D. J., Myers C. E. Phase I study of phenylacetate administered twice daily to patients with cancer.
Cancer (Phila.)
,
75
:
2932
-2938,
1995
.
40
Ravi R., Mookerjee B., Bhujwalla Z. M., Sutter C. H., Artemov D., Zeng Q., Dillehay L. E., Madan A., Semenza G. L., Bedi A. Regulation of tumor angiogenesis by p53-induced degradation of hypoxia-inducible factor 1α.
Genes Dev.
,
14
:
34
-44,
2000
.
41
Hugg J. W., Maudsley A. A., Weiner M. W., Matson G. B. Comparison of k-space sampling schemes for multidimensional MR spectroscopic imaging.
Magn. Reson. Med.
,
36
:
469
-473,
1996
.