The mechanism of action of radiosensitization of conventional chemotherapeutic agents☆

5-fluorouracil and fluorodeoxyuridine 

5-fluorouracil (5FU) has been used extensively with radiation (for other reviews see1, 2). 5FU has both DNA-directed (through the inhibition of thymidylate synthase) and RNA-directed (through incorporation into the 3 species of RNA) effects. Although the disruption of either RNA or DNA synthesis can produce cytotoxicity, a substantial body of evidence suggests that radiosensitization is a result of inhibition of thymidylate synthase (summarized in1).

There are a number of mechanisms by which 5FU could increase radiation sensitivity at the cellular level. First is through the killing of S phase cells, which are relatively radioresistant.3 This does not account for all of the increase in radiation sensitivity produced by the drug because noncytotoxic concentrations can also increase radiation sensitivity. Radiosensitization under noncytotoxic conditions occurs only when cells are incubated with drug before radiation. Evidence that has been accumulated over the last 6 years has supported the hypothesis that increased radiation sensitivity occurs in cells that have inappropriate progression through S phase in the presence of drug (ie, from a disordered S-phase checkpoint). It was initially discovered that radiosensitization by fluorodeoxyuridine (FdUrd) occurred in HT29 human colon cancer cells, which express activated G1/S cyclins in the presence of drug, but not SW620 cells, which did not show activated cyclins under identical drug treatment conditions.4 This conclusion is supported by studies showing that blockade of S-phase entry (by producing an arrest in G1)5 or inhibition of progression into S (by treatment with aphidicolin (a DNA alpha polymerase6) blocks sensitization. Furthermore, SW620 cells, which are resistant to FdUrd-mediated sensitization and arrest at the G1/S boundary, show sensitization by FdUrd when transduced with the viral protein HPVE6.7 This viral protein inactivates the retinoblastoma protein, releasing E2F and other S-phase transcription factors and driving cells through S phase.8, 9 Note that the crucial events producing sensitization appear to occur after the classic G1 checkpoint in the cell cycle, which is consistent with the lack of dependence on p53.10

Given the mechanism of thymidylate synthase inhibition, the short half-life of 5FU and FdUrd in plasma, and the relatively short half-life of intracellular phosphorylated metabolites, these laboratory studies have suggested that 5FU and FdUrd should be given continuously during a course of fractionated radiation if radiosensitization of most fractions is to be achieved. (Note that this differs from gemcitabine, described later, which forms phosphorylated intracellular metabolites that are stable within the cell for days.) Indeed, the use of protracted venous infusion 5FU has become a standard therapy, if not the preferred therapy, for rectal11 and pancreatic cancer.12 However, protracted venous infusion over a 5- to 6-week period is relatively complex, requiring specialized pumps and long-term venous access which makes patients susceptible to infection or thrombosis. Thus, the introduction of oral forms of 5FU, such as the prodrug capecitabine (which is now approved for use in the United States) and UFT (which is not yet approved) will make protracted concurrent treatment with drug and radiation far easier in the clinic.

Capecitabine is converted in a multistep process that depends on the enzyme thymidine phosphorylase. Therefore, capecitabine has the potential to interact differently from protracted venous infusion 5FU based on preclinical studies suggesting that radiation can induce thymidine, an inhibitor of phosphorylase in tumor tissue.13 In these studies, a single 2.5 or 5 Gy fraction resulted in a significant increase in the enzyme at both 6 and 9 days after exposure in 4 of the 5 xenograft lines studied. An increase in tumor levels of tumor necrosis factor α preceded the rise in thymidine phosphorylase. Interestingly, whole-body irradiation (5 Gy) of mice bearing a human colon cancer xenograft resulted in no increase in the activity of thymidine phosphorylase in the liver but produced a 9.4-fold increase in enzyme level in the tumor. Finally, the effect of capecitabine and irradiation on tumor growth was investigated. The WiDr human colon cancer xenograft line was selected because it is refractory to fluoropyrimidines because of low-level thymidine phosphorylase expression. As predicted, the tumor regrowth delay after irradiation in combination with capecitabine appeared to be more than additive. In contrast, tumor regrowth delay after 5FU and irradiation was less than additive. These data provide evidence that radiation therapy may increase the therapeutic index of capecitabine further, based on the lack of thymidine phosphorylase upregulation in normal tissue and the ability to exclude all but a small amount of normal tissue using conformal radiotherapy. We have attempted to extend these results to other cell lines, with mixed results. We found that radiation did appear to induce increased levels of thymidine phosphorylase in HT29 human colon cancer cells (Fig 1) but not in RKO colon cancer cells or PANC-1 pancreatic cancer cells (not shown).

View Large Image Download to PowerPoint

Fig. 1. Effect of radiation on thymidine phosphorylase levels in HT29 human colon cancer cells. HT29 cells were irradiated with 4 Gy. They were assessed for up to 18 days later for thymidine phosphorylase content by immunoblotting.

Intensive efforts have been directed toward the development of assays that will predict response to 5FU as a chemotherapeutic agent. High levels of thymidylate synthase appear to confer resistance to treatment,14 as do high levels of the 5FU catabolic enzyme dihydropyrimidine dehydrogenase.15 It seems likely that these levels will also predict the ability of 5FU and FdUrd to act as radiation sensitizers, although clinical data are lacking at this time.

Gemcitabine 

The activity of gemcitabine against a variety of solid tumors combined with evidence that it affected deoxynucleotide triphosphate (dNTP) pools suggested that it might be a radiosensitizer. Initial studies showed significant enhancement of radiation-induced cell killing at both noncytotoxic and cytotoxic concentrations.16 There was no evidence of radiosensitization when the cells were irradiated before gemcitabine exposure, whereas the greatest enhancement ratio was observed when cells were incubated for 24 hours before irradiation. Subsequent studies explained these results by revealing that maximum sensitization was produced under conditions in which cells were both redistributed into S phase and were depleted of phosphorylated deoxynucleotides, especially deoxyadenosine triphosphate (dATP). These conditions could be produced both by continuous exposure to a low concentration (10-30 nmol/L) of gemcitabine or 8 to 48 hours after a 2-hour exposure to 100 nmol/L (noncytotoxic) or 3 μmol/L (cytotoxic) concentrations.17 This latter condition more closely resembled the clinical setting, in which the drug is administered over 30 minutes to 1 hour, producing plasma levels of >10 μmol/L.18, 19 Additional studies have revealed that cells derived from pancreatic cancer, head and neck cancers, adrenal cancer, and breast cancer are sensitized by clinically achievable concentrations of gemcitabine.20, 21, 22

Substantial evidence suggests that inhibition of ribonucleotide reductase is a key step in producing sensitization. First, the time course of dATP pool depletion (which is a result of ribonucleotide reductase inhibition) matches the time course of radiosensitization.16, 20 Second, cells with widely varying endogenous dNTP pools and sensitivity to gemcitabine were similarly sensitized when similar dATP pool depletion and S-phase redistribution were produced.20 To attempt to obtain direct evidence that inhibition of ribonucleotide reductase is responsible for radiosensitization, we studied human KB cells that overexpress ribonucleotide reductase through the expression of the M2 subunit. These cells show a 3-fold increase in expression of the enzyme and are relatively resistant to hydroxyurea. Control cells were transfected with the M1 submit, which does not affect ribonucleotide reductase activity.23 We found that ribonucleotide reductase overexpressing cells were resistant to gemcitabine-mediated radiosensitization compared with control cells (Fig 2).

View Large Image Download to PowerPoint

Fig. 2. Gemcitabine-mediated radiosensitization of ribonucleotide reductase overexpressing cells. Cells transduced with the active M2 subunit (showing increased activity of ribonucleotide reductase) and the inactive M1 subunit (as a control) were the kind gift of Y Yen. Cells were treated with gemcitabine for 24 hours before irradiation. They were then assessed for clonogenic survival.

Although dATP pool depletion appears to be necessary, it is not sufficient for sensitization to occur. For instance, treatment of HT29 cells with high concentrations of gemcitabine for a short time (<4 hours) produces only moderate sensitization despite near maximal depletion of dATP pools. Maximum sensitization appears to require simultaneous redistribution into S phase along with dATP pool depletion.24 It has also been proposed that radiosensitization of log phase cells occurs by selective sensitization of radioresistant S-phase cells.25

The initial efforts to investigate mechanisms of radiosensitization focused on the period after exposure. Gemcitabine affects neither radiation-induced DNA damage nor repair, as measured by pulsed-field gel electrophoresis, under conditions known to produce radiosensitization.17, 26 In further efforts to explore the role of DNA repair in gemcitabine-mediated radiosensitization, we assessed the role of mismatch repair. For these experiments, we used HCT-116 cells, which contain a mutation in the mismatch repair gene MLH1, and HCT-116 cells transduced with chromosome 3, which regain mismatch repair capabilities. Mismatch repair defective cells were resistant both to the direct cytotoxic and radiosensitizing effects of gemcitabine compared to mismatch repair competent cells (Fig 3).

View Large Image Download to PowerPoint

Fig. 3. Effect of MLH1 mutation and mismatch repair defect on cytotoxicity and radiosensitization of HCT-116 human colon cancer cells. HCT-116 cells were obtained from the American Type Culture Collection. HCT-116 cells containing chromosome 3 (mismatch repair competent) or chromosome 2 (mismatch repair defective, as are the parental cells, as a control for containing an extra chromosome) were exposed to the indicated concentrations of drug for 24 hours. They were then were assessed for clonogenic survival (left) or for radiation sensitivity (right). Radiation sensitivity is expressed as the ratio of the mean inactivation dose (area under the cell survival curve) of the control divided by that of the drug-treated cells.

How does this DNA damage produce cell death? We have found that apoptosis may play a role in cells which show the greatest radiosensitization by gemcitabine, but modest sensitization can still occur even in cells which do not die by programmed cell death.30 Taken together, these findings suggest that gemcitabine may radiosensitize cells that progress into S phase by depleting dATP pools, leading to misincorporation and misrepair of incorrect bases after radiation. These lesions produce both apoptotic and nonapoptotic cell death.

In vivo studies have been conducted to provide insight into the role of the schedule of gemcitabine administration on the therapeutic index. In 1 study, a single 50 mg/kg intraperitoneal dose of gemcitabine was given at various times before or after a single 25 Gy fraction.31 The longest regrowth delays occurred when gemcitabine was administered 24 to 60 hours before irradiation, in agreement with the predictions made by in vitro studies.17 It is important to note that a single 5 mg/kg intraperitoneal dose of gemcitabine, which alone had minimal effect on tumor regrowth, enhanced the response to radiation.32 This confirms the in vitro finding that minimally cytotoxic concentrations of gemcitabine can radiosensitize.

We have investigated the therapeutic index in a head and neck model to refine our initial clinical trial combining gemcitabine with radiation for locally advanced head and neck cancer. In that study, we found that even 50 mg/m2 given once a week throughout the course of treatment (70 Gy in 2 Gy fractions) was intolerable.33 We compared this once weekly with twice weekly delivery, selected in an attempt to maximize radiosensitization based on the in vitro data indicating radiosensitization for only 48 to 72 hours after drug exposure. We defined equitoxic regimens using a lip erythema assay for once weekly or twice weekly gemcitabine combined with 27.5 Gy in 5 fractions radiation and then assessed these equitoxic combinations for efficacy in flank tumor regrowth delay studies.34 We found that tumors treated with twice weekly gemcitabine with radiation underwent a significantly greater growth delay than tumors treated with once weekly drug plus radiation, confirming the greater therapeutic index predicted by the in vitro studies. A clinical trial based on these results is now accruing patients.

Given that gemcitabine produces radiosensitization through prolonged inhibition of ribonucleotide reductatase, it is of interest to note that hydroxyurea, a pure inhibitor of ribonucleotide reductase, has been used as a radiosensitizer in the treatment of head and neck and cervix cancer. However, a daily schedule does not produce continuous enzyme inhibition. The use of continuous infusion hydroxyurea has been proposed as a radiosensitization agent and is undergoing phase 1 testing.35 It would be of interest to compare this schedule of hydroxyurea with once or twice weekly gemcitabine.

Platinum analogs 

Platinum analogs, and specifically cis-diammino-platinum (II) (cisplatin), cis-diammine-1, 1-cyclobutane dicarboxylatoplatinum(II) (carboplatin), and more recently cis-[(1R,2R)-1,2-cyclohexanediamine-N,N'] [oxalato(2-)-O,O'] platinum (oxaliplatin), are being used clinically in combination with radiation in the treatment of a variety of solid tumors. It is not surprising that the mechanism of the interaction continues to be intensely investigated and is the focus of a number of reports.36, 37, 38 Douple and colleagues have reported several potential mechanisms associated with cisplatin and carboplatin-mediated radiation potentiation under both oxic and hypoxic conditions.39, 40, 41, 42 The enhanced cell killing observed when platinum analogs are administered before or after radiation is believed to be mediated through a variety of mechanisms, including (but not limited to) enhanced formation of toxic platinum intermediates in the presence of radiation-induced free radicals,43 a radiation-induced increase in cellular platinum uptake,44 inhibition of DNA repair,45 and cell-cycle arrest.46, 47

Clinical data confirming that radiation delivered concurrently with cisplatin/carboplatin results in an overall improved clinical outcome are expanding, with supporting results observed in trials involving non–small-cell lung cancers, cancers of the head and neck, and more recently in advanced gynecologic malignancie.48, 49, 50

Oxaliplatin is a recently developed third-generation cisplatin analogue with a 1,2-diaminocyclohexane (DACH) carrier ligand. Oxaliplatin reacts with DNA, forming mainly platinated intrastrand crosslinks with 2 adjacent guanines or adjacent guanine/adenine residues.51, 52, 53 It has been reported that the DACH-platinum adducts formed by oxaliplatin are more effective at inhibiting DNA synthesis52, 54 and are more cytotoxic than cis-diammine-platinum adducts formed from cisplatin and carboplatin.53, 54, 55, 56, 57 The 1,2-DACH carrier ligand affects the rate of mono- and di-adduct conversion58 and the ability of cells to tolerate unrepaired platinum-DNA adducts.57, 59 Oxaliplatin has also been shown to possess activity in cisplatin-resistant cell systems.55, 60, 61, 62, 63, 64, 65, 66, 67 As reviewed by Raymond et al,53 a number of studies have shown that alterations in mismatch repair activity may confer intrinsic resistance to cisplatin and carboplatin but not alter the sensitivity to oxaliplatin. The mismatch repair protein complex appears to be prevented from binding to oxaliplatin adducts because of particular conformational distortions in the region of the adduct.68 The latter effects may be responsible for differentiating the antitumor activity of oxaliplatin from carboplatin and/or cisplatin.

Studies examining the radiation sensitizing properties of oxaliplatin are limited. Blackstock et al69 were the first to observe and report the radiation sensitizing properties of oxaliplatin; a sensitizer enhancement ratio after a 24-hour preirradiation exposure to oxaliplatin of 2.7 at 10% survival was observed in an HT29 human colon carcinoma model and was dependent on drug exposure duration and concentration.69 Preliminary xenograft studies from these same investigators resulted in data suggesting that the combination was at least additive and perhaps synergistic. The time interval between the oxaliplatin administration and the radiation appeared important in these studies, likely because of the slow biotransformation of oxaliplatin when compared with other platinum analogs. Although data confirming these preliminary result come from Cividalli et al,70 showing a similar radiation enhancement in a CH3/TIF mammary adenocarcinoma tumor model, these investigators did not observe a sequence or time interval dependence. Table 1 reflects reported and ongoing trials combining radiation with oxaliplatin in the treatment of rectal cancer.

Table 1. Oxaliplatin/radiation trials in rectal cancer

	Investigator	Tumor Stage	Oxaliplatin	5-FU	Radiation
	Freyer et al74	T1–T4	80-130 mg/m2 day 1	Day 1-5	45 Gy
	CALGB phase I/II	T3/T4	40-60 mg/m2 weekly	Continuous infusion	50.4 Gy
	ECOG phase I/II	T3/T4	55-85 mg/m2 biweekly	Continuous infusion	50.4 Gy

The need for better laboratory-clinical interactions 

A fundamental issue in the application of conventional chemotherapeutic agents with radiation concerns determination of the molecular mechanism for the therapeutic index. In the case of the fluoropyrimidines and, perhaps, gemcitabine, it seems likely that dysfunctional S-phase checkpoints in cancer cells may permit inappropriate S phase progression, whereas normal cells arrest at the G1/S boundary and are relatively protected. The selectivity of the platinums radiosensitization remains a mystery. There is overwhelming clinical evidence from randomized trials that cisplatin increases control rates and survival in combination with radiation in a variety of malignancies. Yet, the laboratory studies have not clearly revealed a reason for a therapeutic index. It seems likely that the results of clinical trials will be improved by a better understanding of mechanism, suggesting that additional study is required.

Even in situations in which we have achieved substantial insight into the mechanism of sensitization, there is an important gap between this knowledge and the efficient development of clinical results. The experience with gemcitabine and radiation is illustrative. It was clear from cell culture studies that gemcitabine would be a potent radiation sensitizer at concentrations well below those achieved during a standard infusion of the drug administered as a cytotoxic agent (1,000-1,200 mg/m2). Thus, the toxic results of the first trial of gemcitabine and radiation in the treatment of non–small-cell lung cancer using large radiation portals and full chemotherapeutic doses of drug71 were predictable. However, extraordinarily low doses of gemcitabine (50 mg/m2) produced substantial toxicity when combined with radiation in the treatment of head and neck cancer33 and higher but only modest doses (300-400 mg/m2) were intolerable in pancreatic cancer.72 These trials are in marked contrast to the modest toxicity of full-dose dose gemcitabine and conformal radiotherapy for pancreatic tumors.73 These findings suggest that we need better model systems to help us determine not simply the mechanism of sensitization but the therapeutic index of a treatment. These models will need to include considerations for (1) the irradiated site, (2) the volume irradiated, and (3) the fraction size.

Efforts to understand how conventional chemotherapeutic agents work in combination with radiation in the clinic are hampered by the lack of data from clinical specimens. Even when biopsy studies are possible,33 they represent only a point in time. Tumors and normal tissues change during treatment. Thus, it will be important to continue to develop techniques such as magnetic resonance spectroscopy to permit the measurement of intracellular metabolites over time. The determination of 5FU, which is present at close to millimolar concentrations after therapeutic infusions, can be detected by clinical magnets (1.5-3 Tesla). Gemcitabine, FdUrd, and cisplatin, which are active in the micromolar range, will require positron emission tomography imaging. In addition to imaging the drug, it will be crucial to develop methods of measuring the target. For instance, in the case of FdUrd, one can imagine titrating the drug not to toxicity but to suppression of the target enzyme thymidylate synthase, or for gemcitabine or hydroxyurea, the inhibition of ribonucleotide reductase. In this age of molecularly targeted therapy, we should not forget that conventional chemotherapeutic agents have molecular targets too!