|Home | About | Journals | Submit | Contact Us | Français|
The lack of information regarding the metabolism and pathophysiology of individual tumors limits, in part, both the development of new anti-cancer therapies and the optimal implementation of currently available treatments. Magnetic resonance [MR, including magnetic resonance imaging (MRI), magnetic resonance spectroscopy (MRS), and electron paramagnetic resonance (EPR)] provides a powerful tool to assess many aspects of tumor metabolism and pathophysiology. Moreover, since this information can be obtained non-destructively, pre-clinical results from cellular or animal models are often easily translated into the clinic. This review presents selected examples of how MR has been used to identify metabolic changes associated with apoptosis, detect therapeutic response prior to a change in tumor volume, optimize the combination of metabolic inhibitors with chemotherapy and/or radiation, characterize and exploit the influence of tumor pH on the effectiveness of chemotherapy, characterize tumor reoxygenation and the effects of modifiers of tumor oxygenation in individual tumors, image transgene expression and assess the efficacy of gene therapy. These examples provide an overview of several of the areas in which cellular and animal model studies using MR have contributed to our understanding of the effects of treatment on tumor metabolism and pathophysiology and the importance of tumor metabolism and pathophysiology as determinants of therapeutic response.
The development of new anti-cancer therapies and the optimal implementation of currently available treatments are limited, in part, by the lack of information regarding the metabolism and pathophysiology of individual tumors. Availability of this metabolic and physiologic information for individual tumors could be used in a variety of ways including the following 1) It could guide the development of new strategies to manipulate tumor physiology and/or metabolism for therapeutic advantage, 2) It could allow the effects of treatment on tumor pathophysiology to be considered in scheduling multi-course or multi-modality treatment, 3) It may allow use of changes in tumor pathophysiology evident, prior to changes in tumor size, to detect tumor response to therapy in individual patients early in the course of treatment. One promising method to non-invasively assess tumor physiology and metabolism, which can be applied to cell and animal models and is also already widely available clinically, is magnetic resonance (MR).
Magnetic resonance imaging (MRI), magnetic resonance spectroscopy (MRS), and electron paramagnetic resonance (EPR) can non-invasively provide a wealth of information regarding tumor metabolism and pathophysiology. Moreover, because MR provides information regarding in vivo tumor metabolism and pathophysiology non-destructively, pre-clinical results from cellular or animal models are often easily translated into the clinic. This review presents selected applications to give the reader an overview of several of the areas in which cellular and animal model studies using MR have contributed to our understanding of the effects of treatment on tumor metabolism and pathophysiology and the importance of tumor metabolism and pathophysiology as determinants of therapeutic response. A companion review  focuses on the use of MR to investigate tumor biology and physiology in the absence of treatment.
Apoptosis, or programmed cell death, is a physiological, genetically controlled and morphologically defined process of cell suicide. It is required for the normal development of multicellular organisms, essential, for example, in embryonic morphogenesis or lymphocyte selection. First described on the basis of morphological observations , the precise mechanism and genetic regulation of apoptosis are still under investigation (Ref.  and references therein). In the context of cancer, however, it is now clear that cell death following chemotherapeutic treatment will occur, in most cases, by apoptosis . Conversely, resistance to chemotherapy can be affected by the expression of genes regulating the apoptotic pathway. Non-invasively identifying MR signals that are correlated with programmed cell death could therefore be useful both in contributing to the understanding of apoptosis and in helping assess tumor response in vivo. To date, however, most MR studies of apoptosis have investigated in vitro cell systems.
End-stage apoptosis is characterized by cell shrinkage, blebbing and shedding of apoptotic bodies, as well as alterations in the distribution of phospholipids within the plasma membrane. Cell shrinkage can be monitored by using diffusion-weighted 1H MRS to assess changes in intracellular volume [5,6]. Other 1H MRS studies have also identified increases in cellular fatty acid signals (resonating at 1.3, 2.8 and 5.4 ppm). These have been attributed either to a decrease in membrane microviscosity during apoptosis or to an increase in intracellular triglycerides and free fatty acids following activation of phospholipase A2 during cell death [7,8]. Further changes in cellular lipid metabolism during apoptosis have been identified using 31P MRS. A significant drop in phosphocholine (PCho) has been observed in several different cell lines following apoptosis induced by various chemotherapeutic treatments [8–13]. In HL-60 cells, the drop in PCho was also accompanied by an accumulation of CDP-choline and an inhibition of the CDP-choline: 1,2-DAG cholinephosphotransferase, possibly due to cellular acidification during apoptosis . No accumulation of CDP-choline was observed in other cell lines. In MCF-7 cells, the drop in PCho was due to inhibition in choline uptake followed by an increase in activity of the CTP:PCho cytidyltransferase . In contrast to these observations, an increase in PCho signal was observed in TNF-induced apoptosis in neutrophils . PCho could be generated as the by-product of sphingomyelin breakdown to produce ceramide — a possible second messenger in TNF as well as fas-induced apoptosis. However, in fas-induced apoptosis of Jurkat cells, no increase in PCho content or drop in sphingomyelin could be observed , further contributing to the current controversy regarding the involvement of ceramide in apoptotic signaling.
Apoptosis is an energy-requiring process. At the same time, it is now clear that cytochrome C release from the mitochondria is an important step in the apoptotic cascade, possibly diminishing the cell's ability to generate ATP. How cellular ATP stores evolve during apoptosis is therefore an interesting question and 31P MRS could be useful in providing an answer. Some researchers have observed an initial transient increase in cellular nucleoside triphosphates (NTP, predominantly ATP) content during apoptosis [16,17] possibly explained by upregulation of bcl2 in some cell lines following treatment. In contrast to these results, several studies have observed a steady decrease in cellular ATP following induction of cell death (Refs. [9,11–13]; see Figure 1). This drop in ATP has been correlated in most cases with a reduction in cellular NAD/H, and it remains unclear whether ATP is primarily consumed by the cell to replenish its NAD/H stores or whether the drop in high-energy phosphates reflects their utilization in apoptosis. Finally, in several cell systems, the drop in NAD/H content during apoptosis was accompanied by an accumulation of glycolytic intermediates, in particular fructose-1,6-bisphosphate (see Figure 1; Refs. [9,12,13]). This is due to activation of the enzyme poly(ADP-ribose) polymerase (PARP) following chemotherapy-induced DNA damage. Activated PARP consumes NAD, the depletion of which leads to inhibition of the glycolytic enzyme glyceraldehyde-3-phosphate dehydrogenase leading to the accumulation of fructose-1,6-bisphosphate as well as other intermediates.
In spite of some promising in vivo results , the usefulness of MR techniques in identifying apoptosis in patients still needs to be confirmed. In vivo, not only is the spectral resolution lower, but apoptosing cells are rapidly phagocytosed. In most cases, only a small proportion of apoptotic cells is present at any one time (e.g., see Ref. ). The spectra recorded from a treated tumor would therefore include a relatively small contribution from the apoptotic fraction. Nonetheless, MR techniques can clearly be useful both in monitoring changes such as cell shrinkage, already known to occur during apoptosis, and in identifying metabolic changes that had not been previously reported.
Traditional methods for detection of therapeutic response generally rely on a gross decrease in tumor size (often measured radiographically using two orthogonal dimensions). Although these methods are useful for assessing response at the end of treatment, little information is available early in the course of treatment (i.e., success or failure of therapy generally takes 2 to 4 months to become apparent). Consequently, patients with non-responsive tumors receive a full-course of toxic therapy without benefit. Hence, the need for early predictors of tumor responsiveness applicable to commonly used therapeutic modalities has long been recognized. Such a predictor would not only allow patients with non-responsive tumors to avoid the side effects and toxicity of a full-course of therapy, they could aid in the selection and evaluation of patients participating in clinical trials of new anti-tumor therapies.
One approach being evaluated as an early-response predictor is to assess tumors before and during treatment in an attempt to detect changes in the pathophysiology of tumors responding to the initial therapy. Because of the ability of MR to non-invasively evaluate tumor metabolism and physiology , several MR-based methods have been evaluated for their ability to detect treatment-induced changes occurring within the tumor prior to a decrease in tumor size. Ideally, one would like to be able to detect such changes at the highest possible spatial resolution to permit intra-tumor heterogeneity to be assessed. This section reviews two imaging-based methods with sufficient spatial resolution to assess macroscopic heterogeneity, measurement of water diffusion and dynamic contrast-enhanced MRI (DCE-MRI), which probe different aspects of tumor pathophysiology.
Water self-diffusion measured by pulsed-field gradient MR  is influenced by the restriction of diffusion due to the limited permeability of cell membranes to water . Hence, the water self-diffusion coefficient measured by MR is referred to as the apparent diffusion coefficient (ADC) and is sensitive to biophysical characteristics of tissue, including the fraction of water in the extracellular space . Commonly used anti-cancer agents are cytotoxic, resulting in tumor cells shrinking (apoptosis) or rupturing (necrosis) — both processes that should increase the extracellular water fraction. Since both radiation  and chemotherapy  have been shown to substantially increase the fraction of extracellular water, tumor water ADC should be altered by treatment. If these changes are evident prior to a decrease in tumor size, diffusion MRI could provide a clinically applicable response predictor since diffusion MRI has been applied to human brain tumors [24,25], breast , liver , pancreas , and kidney .
Treatment response studies thus far have employed rodent tumor models. A dose-dependent, reversible increase in tumor water ADC was observed in RIF-1 tumors (a murine fibrosarcoma) within 2 days of treatment with cyclophosphamide . Even in these RIF-1 tumors treated with doses of cyclophosphamide producing an estimated tumor cell kill of 67%, tumor water ADC increases substantially at a time when there is essentially no change in tumor volume. In rat intracranial 9L gliomas treated with a dose of BCNU resulting in an estimated 90% cell kill, tumor water ADC increased substantially within a week of treatment, while the tumor volume doubled over that time . Ganciclovir treatment of BT4C rat gliomas sensitized by thymidine kinase-mediated gene therapy (see Gene Therapy section) increased tumor water ADC within 4 days after treatment, although tumor growth continued over that period [32,33]. Taxol treatment of MCF-7/s (taxol-sensitive) and MCF-7/d40 (taxol-resistant) human breast cancer xenografts growing in mice increased tumor water ADC only in the drug-sensitive MCF-7/s tumors . For chemically induced rat mammary tumors treated with 100 mg kg-1 5-fluorouracil (5-FU), although all tumors decreased significantly in volume over the first 7 days after treatment, tumors with a high initial ADC started to regrow after 7 days . Collectively, these results provide strong support for the potential of tumor water ADC measurements to provide a sensitive early indicator of response to cytotoxic therapy.
Since each patient's tumor is unique, for ADC measurements to be clinically useful as an early response indicator, the relationship between drug efficacy and the change in ADC (the effective “dose-response” relationship) should not depend on the tumor. Because experimental conditions that can alter the dependence of ADC on the tumor biophysical characteristics vary among laboratories, it is difficult to compare results between laboratories. Consequently, the effects of 5-FU on tumor water ADC measured using the same experimental parameters were examined in three different murine tumors [RIF-1 and two colon adenocarcinomas, Colon-26 and Colon-38 ]. A comparison of the therapeutic response [assessed by the specific tumor growth delay, STGD ] and corresponding maximum ADC changes indicates a strong “dose-response” relationship for all three tumors (see Figure 2). Considering the RIF-1 cyclophosphamide data with these results, it appears the relationship between the maximum change in ADC and therapeutic effect may be independent of the cytotoxic agent, as well as the tumor. These observations support the hypothesis that changes in ADC induced by cytotoxic therapy are associated with changes in tumor pathophysiology due to cell death and are directly related to a therapeutic response. Hence, MR measurements of tumor water ADC for early detection of response to cytotoxic therapy should be generally applicable.
As described in the companion article , information on several aspects of tumor vascularity (e.g., blood volume, perfusion, permeability) can be derived from analysis of the intra-tumor kinetics of MR contrast agents obtained using DCE-MRI. Since this technique is non-invasive, the tumor can be monitored longitudinally over a period of time to study the changes in tumor vascularity occurring during growth and alterations induced by various kinds of therapy. Initial results in the clinic using the low molecular weight agent, Gd-DTPA, indicate that DCE-MRI may be valuable for both diagnosis [38,39] and prognosis [40,41] of cancer. While these results suggest that there are substantial physiologic differences (i.e., between benign and malignant, or between non-responsive and responsive tumors), the clinically approved low molecular weight agent does not distinguish between changes in perfusion and permeability. Hence, larger molecular weight contrast agents, which can provide an independent measure of vascular permeability, may provide additional critical information. Since they are not yet clinically approved, the ability of these large molecular weight contrast agents to provide an early indication of treatment response has been studied in rodent tumor models.
DCE-MRI with large molecular weight contrast agents indicates changes in tumor vascularity early after treatment for several treatment modalities. In R3230 mammary adenocarcinomas, vascular permeability to albumin-GdDTPA is elevated significantly in irradiated tumors compared to control non-irradiated tumors [42,43]. In these same tumors, the vascular volume measured by a macromolecular agent gadomer-17 (molecular weight ~35 kDa) stays constant over 2 weeks in control tumors despite its rapid growth . After treatment with a combination of mitomycin-C and flavone acetic acid, the mean vascular volume decreased by 42% in responders after a week and further down to 56% in week 2 .
In another study, DCE-MRI was used to examine the effects of using cytokine therapy to modulate the allogenic rejection of C6 glioma cells implanted subcutaneously into Wistar rats (see Table 1). Adenoviruses expressing mouse interleukin 1-α (IL1-α), mouse interferon γ (IFN-γ), and human transforming growth factor β (TGF-β) were intratumorally injected into three rats, one from each kind. Although the tumor size in both control and rIL-1α-infected animals increased at day 4 (+72% and +140%, respectively), the vascular volume of control tumors decreased by 62% (compared to baseline), while the rIL-1α-inoculated animal had an increase of 41%. These changes of vascular volume at day 4 were predictive of tumor growth several days later. At day 7, the size of control tumors had decreased by 40% while that of the rIL-1α-infected tumors increased by 159%. The predictive value was valid for both unmodulated allogenic tumor rejection and when there was modulation of an immune response (the rIL1-α-treated animal), resulting in a delayed onset of tumor rejection. Immune modulation that increased the immune response (rTGF-β and rIFN-γ) resulted in an accelerated tumor rejection which prevented the evaluation of vascular volume as a predictive tool. These studies support the use of DCE-MRI as a practical tool in the assessment of tumor vascularity to evaluate therapeutic response. As macromolecular contrast agents become more widely available for clinical use, non-invasive evaluation of changes in vascular volume following cancer therapy can be performed.
Cell kill for most commonly used chemotherapeutic agents is thought to occur by damage to DNA ; historically, most chemotherapeutic agents have focused on inhibition or poor fidelity of DNA reproduction. Recent studies have suggested that other targets for anti-tumor agents, such as the tumor vasculature, microtubular structure, cell membrane, intracellular milieu, pH, or metabolic pathways, could enhance tumor destruction or inhibit proliferation. A novel approach to treating tumors is to use traditional chemotherapeutic agents in combination with drugs aimed at interfering with tumor metabolism, i.e., agents that inhibit glycolysis, aerobic metabolism, etc. These agents often have limited, if any, activity by themselves, but can potentiate the effect of other drugs or radiation. Examples of these agents include 2-deoxyglucose (2-DG), lonidamine (LON), and 6-aminonicotinamide (6-AN).
6-AN, an analogue of niacin, is currently being evaluated in pre-clinical trials as part of a combination chemotherapy and as a radiosensitizer [46–49]. Although 6-AN was evaluated in clinical trials about 40 years ago and lacked efficacy , in vitro studies have shown 6-AN to sensitize cells to radiation  cis-platinum  and 1,3 Bis (2-chloroethyl)-1-nitrosourea (BCNU) . 6-AN acts by competition with niacin in pathways utilizing NAD (P), being metabolized to 6-ANAD (P). 6-ANAD (P), in turn, can act as a competitive inhibitor of NAD (P) requiring processes. 6-ANADP is a particularly potent inhibitor of the PPP enzyme, 6-phosphogluconate (6-PG) dehydrogenase, which is an important step in the synthesis of NADPH and ribose units required for biosynthesis and DNA repair. Inhibition of this enzyme by 6-AN leads to accumulation of 6-PG which inhibits glycolysis. Therefore, 31P MRS studies of tumor metabolism should detect a decrease in NTP and phosphocreatine (PCr) due to inhibition of glycolysis and possibly detect 6-PG, which is distinguishable from naturally occurring phosphoethanolamine.
Keniry et al.  used 31P MRS to study 6-AN in vitro in combination with gossypol, another metabolic inhibitor, and found that it abolished the metabolic effects of gossypol. They were unable to resolve 6-PG in the whole cell spectra . Subsequent studies have investigated the effect of 6-AN in combination with 6-methyl mercaptopurine riboside (MMPR) plus N-(phosphonacetyl)-L-aspartate (PALA) in vivo. PALA decreases pyrimidine synthesis , MMPR decreases purine synthesis , and 6-AN inhibits the PPP and glycolysis and therefore, this combination should inhibit multiple metabolic processes and enhance tumor cell kill by anti-neoplastic agents. The combination of these three drugs decreases the NTP to inorganic phosphate (Pi) ratio and the PCr to Pi ratio at 3, 10 and 24 hours . Based on these findings, it was hypothesized that depletion of energy at 10 hours posttreatment with the three-drug combination would enhance response to radiation. This was evaluated by treating mice bearing first-generation transplants of the spontaneous CD8F1 mammary carcinoma with three cycles of either PALA, MMPR, and 6-AN, followed by three 15 Gy fractions, or PALA, MMPR, 6-AN, and 15 Gy given three times. This tumor model, which has a 100% correlation with human breast cancer with regards to response to chemotherapy , had never been cured by any previous treatments. Neither the chemotherapy nor radiation alone induced any long-term complete remissions (CR); only one brief CR was noted. However, the combination of PALA, MMPR, 6-AN, and 15 Gy given three times resulted in a 65% complete response rate with 25% of the tumor-bearing mice without recurrence at greater than 1 year, representing a likely cure of these tumors. Subsequent investigations of the effect of single-agent 6-AN as a potential radiation sensitizer in vivo detected 6-PG and demonstrated efficacy, although inferior to the three-drug combination [48,49].
Mechanistic studies of the effect of 6-AN on tumor metabolism were conducted in RIF-1 cells . Using 13C and 31P MRS, perfused RIF-1 cells were studied after treatment with 40 µM 6-AN. 6-PG was detected by both 31P and 13C MRS (see Figure 3A and B). Figure 3A demonstrates 31P MR spectra indicating a decrease in PCr and NTP and an increase in 6-PG after treatment with 6AN. The 13C MRS studies showed that after 6-AN, 6-PG and 6-phosphoglucono-d-lactone (6-PGL) were detected (Figure 3B), while glucose consumption and lactate production were decreased (not shown). Enhancement of radiation only occurred when 6-AN was given before radiation and not the reverse sequence, indicating that the metabolic effect of 6-AN was necessary for radiation enhancement. It was not clear, however, whether inhibition of the pentose phosphate pathway (PPP) or glycolysis induced radiosensitization. 1H MRS quantitation of the relative inhibition of glycolysis and PPP indicated that 6-AN primarily inhibited glycolysis. These studies reflect the translational potential of MR in studying metabolic inhibitors as potential anti-neoplastic agents or sensitizers  and as potentially valuable tools for clinical and preclinical studies.
LON is a metabolic inhibitor which has been shown to enhance radiation and hyperthermia. Ben-Horin et al.  have used 31P and 13C MRS to study perfused MCF-7 cells. Lonidamine induced intracellular acidification accompanied by ATP depletion, decreased glucose accumulation, and increased lactate accumulation. They concluded that the mechanism of action of this drug is via inhibition of lactate transport. Ben-Yoseph et al.  studied 9L glioma cells and also noted acidosis and decreased NTP without any effect on tumor blood flow after treatment with LON. They concluded that LON inhibits lactate efflux leading to cellular acidification.
2-DG is a glucose analogue that competitively inhibits both transport and metabolism of glucose. It is phosphorylated intracellularly and is thought not to undergo further metabolism and inhibits glycolysis. Karczmar et al.  have used 2-DG to deplete ATP and found that it did not affect brain bioenergetics. Kaplan et al. [61,62] studied multi-drug resistance using 2-DG. 13C and 31P MRS data showed that multidrug-resistant (MDR) cells accumulate more 2-DG and at a faster rate than wild-type (WT) cells. They suggested that this difference could be used therapeutically and subsequently showed that 2-DG was more toxic to MDR cells than wild type cells. In contrast, Rasmussen et al.  have also used 2-DG to study glycolytic rates and have shown that MDR Ehrlich ascites cells do not show enhanced glycolysis compared to wild type, suggesting that further studies are necessary to resolve this issue which may have therapeutic significance. Further studies to address the critical questions of selectivity and toxicity are necessary.
Tumors have long been known to exhibit aerobic glycolysis, generating lactic acid even in the presence of oxygen . As a consequence of this phenomenon, tumors have traditionally been thought of as acidic tissues. Direct measurement of pH in a variety of solid tumors by use of microelectrodes has strengthened this viewpoint . As discussed in the companion article , in vivo 31P MRS permits the simultaneous non-invasive measurement of intracellular pH (pHi) and extracellular pH (pHe) in tumors using endogenous inorganic phosphate (Pi) and exogenous 3-aminopropylphosphonate (3-APP), respectively. Measurement of pHi and pHe by this method in solid tumor xenografts has shown that the pHi of tumor cells is neutral to alkaline while the pHe of the same tumors is acidic [66,67]. This pH gradient can impact the effectiveness of chemotherapy. The plasma membrane of cells is more permeable to the uncharged forms of ionizable drug molecules compared to their charged forms, especially for lipophilic species. Thus, while the uncharged form of the drug tends to equalize in concentration across both sides of the plasma membrane, the charged form of a weak-base species exists in higher concentration on the more acidic side of the membrane, leading to an accumulation of the drug in the more acidic compartment. This phenomenon is sometimes referred to as “ion-trapping” . Cells in a tumor can maintain acid-outside plasmalemmal pH gradients of up to 0.5 pH units . Some commonly prescribed anti-neoplastic agents, such as the anthracyclines, have acid dissociation (pKa) constants of 7.5 to 8.5. For such drugs, acid-outside plasmalemmal pH gradients of 0.5 pH units can lead to a 50% reduction in intracellular drug levels as compared to extracellular drug levels, conferring a form of “physiological drug resistance” upon the tumor [70,71]. Abolishing or reversing the direction of the plasmalemmal pH gradient in tumors in vivo therefore should be an attractive means of enhancing the effectiveness of weak-base chemotherapeutic agents.
Sodium bicarbonate (NaHCO3) has been used in humans to treat metabolic acidosis resulting from renal failure  or to alleviate exercise-induced acidosis , and has been shown to effect modest increases in blood pH. For alkalosis to result in improved effectiveness of a weak-base drug, the subject should be able to tolerate doses of NaHCO3 sufficient to produce meaningful increases in tumor pHe. MCF7 tumor-bearing SCID mice tolerated ad libitum water containing 200 mM NaHCO3 for periods of up to 90 days continuously and gained weight at rates similar to control mice . Representative 31P spectra of MCF-7 tumors grown in SCID mice treated or untreated with NaHCO3 are shown in Figure 4. For the tumor in the control mouse, the pHe measured by 3-APP is 6.93 and the pHi measured by Pi is 7.29. In the tumor in the bicarbonate-treated mouse, the 3-APP resonance is shifted relative to controls, indicating a significant alkalinization of pHe, while all other resonances (Pi, PME, NTP) are statistically indistinguishable from the control. In this example, the measured pHe was 7.76. The maximum measured pHe in normal hind leg tissue in bicarbonate-treated SCID mice was 7.67 compared to an average measured pHe of 7.39 in control mice. Thus, chronic oral administration of NaHCO3 resulted in significantly greater alkalinization of tumor tissue than normal tissue, and this provides a method to increase the chemotherapeutic index of weak-base drugs.
According to the measured plasmalemmal pH gradients, the theoretical cytosolic-to-extracellular partition ratio for doxorubicin in tumors should be 1.7 in bicarbonate-treated animals, and 0.55 in control animals, leading to a 3.1-fold increase in tumor sensitivity to doxorubicin . Similarly, the calculated doxorubicin partition ratio for normal tissue is 1.9 in bicarbonate-treated animals and 1.6 in control animals, leading to a 1.2-fold increase in sensitivity of normal tissue to doxorubicin. Thus, bicarbonate treatment can potentially yield a 2.6-fold increase in the chemotherapeutic index of doxorubicin. Indeed, combination therapy of MCF7 tumor-bearing SCID mice with doxorubicin and 200 mM NaHCO3 resulted in a significantly reduced tumor growth rate in those mice as compared to tumors in mice treated with doxorubicin alone, while bicarbonate alone had no effect on the tumor growth rate . Prolonged treatment with sodium bicarbonate in the drinking water did not have significant effects on animal well being. In mice, the lifetime of free drug is on the order of 1 hour following intravenous delivery . Therefore, acute metabolic alkalinization with sodium bicarbonate lasting 1–2 hours should enhance drug uptake into tumor cells while sparing the host animal the deleterious effects of chronic denial of fresh water. For the very same reasons, acute alkalinization should also have greater clinical utility than chronic treatment with sodium bicarbonate.
Measurements of tumor oxygenation and changes in tumor oxygenation are crucial to an understanding of tumor physiology and response to therapy. The oxygen tension in tumors is a key determinant of the sensitivity of tissues to ionizing radiation and to certain chemotherapeutic agents . For many years, the occurrence of hypoxia in tumors has been postulated to be a limiting factor in the response to radiation therapy  and recent results using the Eppendorf electrode system to measure pO2 have confirmed that this can occur in human tumors [78,79]. In addition to quantitative oxygen measurements made using oxygen electrodes, a variety of spectroscopic methods have been used. Some of these are based on the use of fluorescent and phosphorescent probes which are sensitive to oxygen tension. Others are based on spectroscopic differences between deoxyhemoglobin and oxyhemoglobin. Although these methods offer some advantages, none of them is used routinely and their invasiveness and/or poor spatial resolution has encouraged increased emphasis on the development of magnetic resonance methods sensitive to oxygen tension.
Some magnetic resonance methods are quantitative, directly measure tumor pO2, and may allow identification of hypoxic tumor regions and accurate measurement of changes in pO2 during therapy. One very promising method is based on 19F MRI of signals from perfluorocarbons after either intravenous or intratumoral injection [80,81]. As described in the companion review , the T1 relaxation time of the fluorine nucleus in these molecules is sensitive to local oxygen tension, and since relatively large amounts of fluorine-containing molecules can be injected, images with reasonable spatial resolution can be obtained. Another approach, in vivo electron paramagnetic resonance (EPR) oximetry with particulate materials, can provide repeated non-perturbing measurements from the same site . EPR oximetry is based on the fact that the presence of molecular oxygen affects the linewidth of paramagnetic materials. This effect is especially useful in lines that are intrinsically narrow or have mechanisms that involve factors in addition to those mediated simply by changes (decreases) in relaxation times. The effects of oxygen on linewidths of carbon-based particulates derived from coals or chars are especially large at low concentrations of oxygen, making these materials particularly suitable for oximetry in tumors where the pO2 values of most interest are quite low (in the range of 0 to 10 Torr).
Another MR approach, which makes use of proton (water) MRI, relies on blood-oxygen-level-dependent (BOLD) contrast. The large paramagnetic susceptibility of deoxyhemoglobin generates large magnetic field gradients in and around blood vessels which effect the linewidth (or T2* relaxation time) of the water proton signal. This does not allow quantitative measurement of the oxygen level, except in large veins, because there are many other important influences on the water proton signal. However, BOLD MRI can provide a qualitative indication of changes in tumor oxygenation. Equipment used to obtain BOLD MRI measurements is widely available both for clinical measurements in cancer patients and for studies of animal models.
A dose of radiation kills oxygenated cells more efficiently, leaving a greater proportion of hypoxic cells that tend to reoxygenate in many animal tumors and human xenografts . It has been suggested that one of the reasons fractionated radiation is successful is that tumor reoxygenation occurs after each dose. Consequently, the tumor is well-oxygenated at the time of subsequent doses, while the effects on normal tissues are not altered because they already are fully oxygenated in terms of responses to radiation . Different tumor types and different individual tumors, however, may fail to reoxygenate or may have altered timing of post-radiation changes in pO2. It then follows that if methods were available to assess tumor oxygenation status repeatedly and accurately, more effective radiation treatments could be carried out by timing radiation doses to take advantage of reoxygenation. Those tumors that fail to reoxygenate and thus are unlikely to respond to irradiation could be treated with another modality without having to wait for the current endpoint, which is a failure to respond adequately to the full course of therapy.
Using EPR oximetry, the time course of pO2 changes after an initial radiation dose in animal tumors has been followed over periods of several weeks . The typical time pattern after irradiation has been an initial decrease post-irradiation, followed by a subsequent increase in tumor pO2. The biological significance for these changes was tested by delivering split-dose radiation with the second dose timed to coincide with the measured high or low point of pO2 post-irradiation (see Figure 5). As shown in the figure, the measured values of pO2 were predictive of the biological effect. In fact, the effects of the split-dose regime delivered at the time when the pO2 was at its high point post-irradiation was at least as effective as if the radiation had been delivered in a single dose, even though usually splitting a dose of radiation lowers its efficacy.
These results suggest that EPR oximetry may be a useful tool for optimizing the scheduling of radiation therapy by enabling the therapist to deliver radiation doses at times where the maximum effects would occur in the tumors. Because at normal (physiological) levels of oxygen the radiation response of cells does not change over a broad range of oxygen concentrations and surrounding normal tissues usually do not have hypoxic areas, the response of the normal tissues adjacent to the tumor is not affected by radiation-induced changes in the pO2. Therefore, by optimizing the scheduling of radiation to coincide with post-irradiation increases of oxygenation, the therapeutic ratio should be increased.
While these results are quite promising, additional studies in animals and then in human patients will be required to demonstrate the efficacy of this approach. Other techniques that could provide similar data (e.g., 19F MRI with perfluorocarbons) would be equally applicable; however, their capability to assess reoxygenation has not been established. The EPR technique described above can be extended to report on the pO2 in several locations simultaneously by the use of suitable gradients of the magnetic field . Using soluble stable free radicals, it may be possible to obtain much more detailed maps of pO2 in tumors using EPR imaging .
Although tumor reoxygenation may lead to effective radiotherapy in some tumors, in others, treatments designed to increase tumor oxygenation may be required. Because the efficacy of tumor-oxygenating treatments likely varies among tumors, non-invasive methods to assess their impact are needed to guide their development and clinical use. Several laboratories have used BOLD MRI to detect large and reproducible T2* increases in a number of rodent tumor models during inhalation of pure oxygen and carbogen [88–93]. In at least one tumor model, the R3230 mammary adenocarcinoma growing in Fisher rats changes in T2* during tumor-oxygenating treatments averaged over the whole tumor strongly correlates with changes measured with an oxygen microelectrode (Ref. ; see Figure 6). Moreover, in BA1112 tumors, BOLD MRI correctly ranks the effects of three tumor-oxygenating treatments [carbogen alone, perfluorocarbon (PFC) emulsion injected intravenously, and the combination of carbogen and PFC] on tumor hypoxic fraction . This experimental evidence supports the hypothesis that changes in T2* reflect changes in the oxygen saturation of blood in tumors caused by tumor-oxygenating treatments — and that this is strongly related to changes in extravascular pO2 and hypoxic fraction.
MRI allows evaluation of spatial and temporal heterogeneity within tumors in their response to oxygenating treatments. In fact, MRI has provided some evidence that some tumor regions respond paradoxically to oxygenating treatments. Although there may be an increase in average oxygenation, tumor-oxygenating treatments frequently decrease blood oxygenation in a small but significant fraction of tumor volume as indicated by increased water signal linewidth . This may mean that some tumor regions become more radioresistant during tumor-oxygenating treatments, and cells in these regions may survive radiation and repopulate the tumor following therapy. Depending on the mechanism underlying these effects, MRI could guide the formulation of tumor-oxygenating treatments or combinations of tumor-oxygenating treatments which maximally increase oxygenation throughout the entire tumor volume.
Oxygen challenges cause contrast changes in MRI images which may be useful for diagnostic purposes. Decreases in water signal linewidth in tumors caused by tumor-oxygenating treatments are much more significant than changes in surrounding normal tissues [Refs. [88,89]; e.g., see Figure 6]. This may be because tumors are initially hypoxic while normal tissue is not. Conversely, Vexler et al.  showed that decreases in T2* in rodent hepatic tumors due to ischemia are much larger than T2* decreases in surrounding normal tissue; perhaps because the tumors have insufficient blood supply. Thus, changes in T2* in response to increases or decreases in blood oxygenation may help to identify tumor regions which are hypoxic or on the verge of hypoxia. In general, changes in image contrast in response to a variety of benign challenges, which affect oxygenation and blood supply, may help to identify tumors and characterize their physiology.
Use of EPR and MRI to measure tumor oxygenation and changes in tumor oxygenation is a relatively new field that is developing rapidly. The methods that are now being developed and validated may have a significant impact on the diagnosis and treatment of solid tumors in the coming decade.
Gene therapy offers the possibility of curing disease by correcting the genetic defects that underlie its pathogenesis. As a genetic disease caused by multiple mutations in growth-regulating genes in a single cell, cancer does not seem amenable to this form of “corrective” gene therapy. However, there has arguably been more success in cancer gene therapy than any other type. There are currently over 200 cancer gene therapy protocols enrolling patients. The majority of these novel therapeutic approaches do not aim to correct the host of somatic mutations that lead to oncogenesis, but rather aim to create therapeutically exploitable phenotypic differences between normal and malignant cells.
Two of the most widely studied cancer gene therapy paradigms rely on expression of non-mammalian enzymes that convert non-toxic prodrugs into cytotoxic metabolites. Local expression of such prodrug-converting enzymes in tumor tissue results in localized chemotherapy, aiming to improve outcomes by minimizing the dose-limiting toxicity associated with systemic chemotherapy. The first of these “chemosensitization” gene therapy approaches is based on the preferential phosphorylation of the anti-herpetic drug, ganciclovir, by the Herpes Simplex Virus thymidine kinase (HSV-tk) gene . The other widely employed strategy results in the production of the antimetabolite, 5-FU, from the antimycotic agent, 5-fluorocytosine (5-FC), by the enzymatic action of microbial cytosine deaminase (CD) .
The ability to non-invasively measure the distribution, magnitude and duration of therapeutic transgene expression could have a profound impact on the development of cancer gene therapy. Such assays would allow subject-by-subject correlation of therapeutic efficacy with levels of transgene expression, as was recently accomplished using radionuclide imaging , and would be useful tools in the development of vector technology. A recent study has demonstrated the potential of MR spectroscopy for visualizing transgene expression in cancer cells . In this study, the enzymatic activity of yeast CD expressed in subcutaneous tumors was quantified with 19F MRS to observe the conversion of 5-FC to 5-FU. This study was the first direct confirmation that an enzyme/prodrug gene therapy paradigm results in local chemotherapy (see Figure 7). Future studies may provide data on the feasibility of obtaining 19F MR images of the metabolites within the tumor tissue that would allow for regional assessment of CD delivery for correlation with therapeutic response.
While a variety of delivery vehicles have been utilized for delivering transgenes to tumor cells, viral vectors have been the most widely used vector systems. Viral vectors have similar dimensions as monocrystalline iron oxide (MION) particles which are used as an MRI contrast agent. MR has been reported useful for imaging the distribution of viral vectors as they have been shown to have essentially the same tissue distribution as MION particles . Additional strategies for using MRI to detect gene delivery have been demonstrated in vitro using specific marker genes such as melanin  and the human transferrin receptor .
Small animal tumor models are used for the pre-clinical evaluation of the anti-neoplastic efficacy of experimental gene therapy paradigms. The use of MRI and MRS for evaluating the effects of gene therapy on tumor physiology, biology, biochemistry, and morphology has been reported. Studies using orthotopic tumor models have benefited from the use of MR, which provides the ability to demarcate the location and size of the tumor prior to gene delivery. This has been used for the guiding the delivery of both retroviral  and adenoviral  vectors into intracerebral glioma models. Changes in tumor volumes and growth rates following gene therapy protocols in orthotopic tumor models have also proven beneficial for assessing therapeutic efficacy [96,103–106]. In addition to using MR to assess changes in morphometry, recent studies using diffusion MRI (see Early Detection of Therapeutic Response section) have shown that changes in tumor water ADC can be observed following gene therapy [32,33]. Finally, changes in MRS-observable proton metabolites following HSV-tk gene therapy have also been reported [32,103]. Taken together, these studies reveal that a variety of significant gene therapy-induced effects can be non-invasively observed in tumor models using MRI/S.
Anti-vascular cancer therapy is here defined as a treatment acting primarily via effects on development or function of tumor blood vessels. Its potential advantages include the following. 1) Damage to a small number of vascular endothelial cells could result in the death of a large number of tumor cells which depend on them for the supply of nutrients. 2) It is easier to deliver drugs to vascular endothelial cells than to cells in solid tumors. 3) Tumor vasculature is a new target so agents should have a new spectrum of toxicities. 4) Inhibition of formation of new blood vessels (anti-angiogenic therapy) may slow or prevent metastatic spread or growth of micro-metastases. Because MR can non-invasively assess tumor vascularity and energetics, it provides a powerful tool for the pre-clinical development of anti-vascular therapy that can be readily translated into clinical use.
Several vaso-active agents, in particular hydralazine, have been shown to result in a selective reduction in tumor perfusion in transplanted animal tumors and, in some cases, also in primary tumors in animals [e.g. Ref. [ 107]]. 31P MRS has been widely applied in such studies for non-invasive monitoring of the metabolic consequences of an acute reduction in tumor blood flow (fall in NTP, increase in Pi and often a decrease in pH). The mechanism of action is generally believed to involve vasodilatation of normal peripheral blood vessels, resulting in a reduction in arterial blood pressure and a “steal effect” of blood away from the tumor (tumor vessels may not respond normally because of a lack of smooth muscle and/or appropriate receptors). The high interstitial pressure of tumors may also contribute to the collapse of some vessels when blood pressure falls. Since the blood flow and metabolic effects of this approach are generally short-lived and do not normally cause permanent damage to tumor blood vessels, they are not normally included as anti-vascular agents.
Vascular damage has been shown to be an important feature of treatments with the cytokines TNF-α  and IL-1α . These cytokines resulted in hemorrhage (determined histologically) and an attenuation of tumor energetic status (31P MRS). Flavone acetic acid (FAA) was probably the first anti-cancer drug shown to act predominantly via tumor vascular damage. Evelhoch et al.  demonstrated that FAA caused a reduction in bioenergetic status (31P MRS measurement of NTP/Pi) and in tumor blood flow (2H MRS measurement of D2O clearance). Su et al.  measured the uptake of Gd-DTPA and a macromolecular MR contrast agent (gadomer-17, 35 kDa) in a rat adenocarcinoma before treatment and after a combination of Mitomycin-C and FAA. The kinetics of Gadomer-17 was used to calculate tumor vascular volume, which was reduced by 42% 4 days after treatment in responding tumors. Unfortunately, the pre-clinical success with FAA was not translated into the clinic, possibly because FAA acted via causing a release of TNF in the mouse but did not lead to TNF release in humans. A more potent analogue of FAA, dimethylxanthenone acetic acid (DMXAA), has been developed which can cause TNF release in humans. 31P MRS has been used to follow the time course and dose-dependency of DMXAA effects in murine C3H mammary tumors  while DCE-MRI studies (uptake of GdDTPA) have been included in phase I clinical trials of DMXAA and preliminary results showed a reduction in both the rate and magnitude of Gd-DTPA uptake 24 hours post-treatment in 7 out of 10 patients .
Combretastatin is a tubulin-binding agent with anti-angiogenic and anti-vascular properties. Beauregard et al.  used 31P MRS and several MRI approaches to study the effects of combretastatin (100 mg kg-1) on subcutaneous sarcoma F tumors in mice. They observed a fall in energetic status (reported as a two-fold increase in Pi/NTP) 150 minutes after treatment and a fall in tumor pH. The rate of uptake of Gd-DTPA into the tumor center was decreased 2.5-fold at 160 minutes with a smaller effect reported for the tumor periphery. Changes in T2-weighted images (i.e., slight increases in the area of low signal intensity regions) 180 minutes after treatment were ascribed to hemorrhage, consistent with histological changes in these tumors. The tumor water ADC was measured before and up to 1.8 hours after treatment but no significant changes were detected. Maxwell et al.  used MRS and MRI techniques to study the effects of combretastatin (100 mg kg-1) on C3H mammary tumors implanted on the mouse foot. The 31P MRS results were similar to those of Beauregard et al.  (marked fall in NTP/Pi) although an additional measurement at 24 hours after treatment showed full recovery of NTP/Pi. Single-voxel 1H MRS measurements of lactate did not show a consistent increase, possibly due to lipid contamination and/or a reduction in glucose supply associated with blood flow reduction. No increase in low signal intensity regions of T2-weighted MR images was detected within 3 hours of combretastatin treatment. In the case of the C3H mammary tumor, histology showed no significant increase in tumor necrosis 3 hours after treatment although an increase in necrosis from 2% to 16% was observed at 24 hours. The extent of hemorrhagic necrosis may vary between anti-vascular therapies and tumor types but these data imply that hemorrhage is not a pre-requisite for combretastatin-induced tumor blood flow reduction.
Given the availability of MR methods relevant to the actions of anti-vascular drugs, it is both feasible and desirable to include MR end points as part of the preclinical and clinical evaluation of such agents. However, several complicating factors should be taken into account. 1) The analysis of Gd-DTPA kinetics may be complicated by drugs which act both to decrease perfusion and to increase vascular permeability. Freely diffusible tracers (e.g. D2O for animal studies) and blood-pool contrast agents may give more specific information. 2) Conflicting effects of hemorrhagic necrosis and edema may be seen in MR images. 3) Lacate-edited 1H MRS is relatively sensitive and specific but lactate changes may be attenuated by limited glucose supply. 4) 31P MRS is relatively insensitive but has been valuable for monitoring the effects of vaso-active and vascular-damaging agents in animal tumors. There may be a threshold of blood flow reduction and/or a delay before effects on energy metabolism are observed. 5) A range of MRI tools for studying angiogenesis has been introduced [e.g., Abramovitch et al. , Sipkins et al. ] and may be useful for evaluating anti-angiogenic drugs. High spatial resolution will be needed and micro-metastases may be difficult or impossible (at the present time) to detect.
Magnetic resonance can non-invasively provide a wealth of information regarding tumor metabolism and pathophysiology. As a consequence, it has contributed to our understanding of both the effects of treatment on tumor metabolism and pathophysiology and the importance of tumor metabolism and pathophysiology as determinants of therapeutic response. MR has made substantial contributions in these areas including several discussed in this review. MR can readily monitor changes such as cell shrinkage, already known to occur during apoptosis, and previously unappreciated metabolic changes associated with upstream steps in the apoptotic pathway have been identified. MR measurable changes in tumor water ADC induced by cytotoxic therapy, which appear to be associated with changes in tumor pathophysiology due to cell death, may be generally applicable for early detection of response to cytotoxic therapy. DCE-MRI is a practical tool for non-invasive evaluation of changes in tumor vascularity following cancer therapy and may provide an early indication of therapeutic response. The metabolic information provided by MR allows evaluation of the pharmacodynamics of metabolic inhibitors that can help to optimize their use as potential anti-neoplastic agents or sensitizers. 31P MRS measurement of pHe and pHi simultaneously demonstrates for the first time in vivo that raising the plasmalemmal pH gradient with sodium bicarbonate in drinking water leads to significant improvements in the therapeutic effectiveness of doxorubicin against MCF-7 xenografts. Non-invasive assessment of tumor oxygenation with EPR and MRI allows the effects of modulating hypoxia to be considered in the timing of subsequent doses of radiotherapy. 19F MRS quantitation of the enzymatic activity of yeast CD expressed in subcutaneous tumors by observing the conversion of 5-FC to 5-FU is the first direct confirmation that an enzyme/prodrug gene therapy paradigm results in local chemotherapy. Because in vivo tumor metabolism and pathophysiology can be assessed non-destructively by MR, pre-clinical results from cellular or animal models are often easily translated into the clinic. Hence, these MR techniques, which have distinct advantages over other proposed and currently implemented techniques, have the potential to revolutionize treatment monitoring.
1USPHS grant CA43113 (J.L.E.); USPHS grant CA83041 and the Flinn Foundation (R.J.G.); USPHS grant CA75476 (G.S.K.); DAMD17-98-1-8153 (J.A.K.); UC Biotechnology Program grant 97-08 and California BCRP grant 3IB-0028 (O.N.); USPHS grant CA77575 (N.R.); UKCRC grant SP1780/0103 (S.M.R.); USPHS grants GM51630 and RR11602 (H.M.S.).