Base Excision Repair

The repair of base damage is initiated by DNA repair enzymes called DNA glycosylases, which recognize specific types of damaged bases and excise them without otherwise disturbing the DNA strand.³⁹ The action of the glycosylase itself results in the formation of another type of damage observed in irradiated DNA—an apurinic or apyrimidinic (AP) site. The AP site is then recognized by another repair enzyme, an AP endonuclease, which nicks the DNA adjacent to the lesion. The resulting strand break becomes the substrate for an exonuclease, which removes the abasic site, along with a few additional bases. The small gap that results is patched by DNA polymerase, which uses the opposite, hopefully undamaged, DNA strand as a template. Finally, DNA ligase seals the patch in place.

Nucleotide Excision Repair

The DNA glycosylases that begin the process of base excision repair do not recognize all known forms of base damage, however, particularly bulky or complex lesions.³⁹ In such cases, another group of enzymes, termed structure-specific endonucleases, initiate the excision repair process. These repair proteins do not recognize the specific lesion but are thought instead to recognize more generalized structural distortions in DNA that necessarily accompany a complex base lesion. The structure-specific endonucleases incise the affected DNA strand on both sides of the lesion, releasing an oligonucleotide fragment made up of the damage site and several bases on either side of it. Because a longer segment of excised DNA—including both bases and the sugar phosphate backbone—is generated, this type of excision repair is referred to as nucleotide excision repair to distinguish it from base excision repair (described earlier), where the initial step in the repair process is removal of the damaged base only. After this step, the remainder of the nucleotide excision repair process is similar to that of base excision repair; the gap is then filled in by DNA polymerase and sealed by DNA ligase. Overall, however, nucleotide excision repair is a much slower process, with a half-time of approximately 12 hours.

For both types of excision repair, active genes in the process of transcription are repaired preferentially and more quickly. This has been termed transcription-coupled repair.⁴⁰

Strand Break Repair

Despite the fact that unrepaired or mis-rejoined strand breaks, particularly DSBs, often have the most catastrophic consequences for the cell in terms of loss of reproductive integrity,⁴¹ the way in which mammalian cells repair strand breaks has been more difficult to elucidate than the way in which they repair base damage. Much of what was originally discovered about these repair processes is derived from studies of x ray–sensitive rodent cells and microorganisms that were subsequently discovered to harbor specific defects in strand break repair. Since then, other human genetic diseases characterized by DNA repair defects have been identified and are also used to help probe these fundamental processes.

A genetic technique known as complementation analysis allows further characterization of genes involved in DNA repair. Complementation analysis involves the fusion of different strains of cells possessing the same phenotypic defect and subsequent testing of the hybrid cell for the presence or absence of this phenotype. Should two cell types bearing the defect yield a phenotypically normal hybrid cell, this implies that the defect has been “complemented” and that the defective phenotype must result from mutations in at least two different genes. With respect to DNA strand break repair, eight different genetic complementation groups have been identified in rodent cell mutants. For example, a mutant Chinese hamster cell line, designated EM9, seems especially radiosensitive owing to delayed repair of DNA single-strand breaks, a process that is normally complete within minutes of irradiation.⁴²

With respect to the repair of DNA DSBs, the situation is more complicated in that the damage on each strand of DNA may be different, and, therefore, no intact template would be available to guide the repair process. Under these circumstances, cells depend on genetic recombination (nonhomologous end joining or homologous recombination⁴³) to cope with the damage.

Nonhomologous end joining (NHEJ) is a repair mechanism that directly ligates broken ends of DNA DSBs using a heterodimeric enzyme complex consisting of the proteins Ku-70 and Ku-80, the catalytic subunit of DNA protein kinase (DNA-PK_CS), and XRCC4/ligase IV. Because no intact complementary strand is available as a template, this process is error-prone and generally results in minor deletions in the DNA sequence. NHEJ predominates during the G₁ phase of the cell cycle. Homologous recombination (HR), on the other hand, involves digestion and “cleaning up” of the broken DNA ends followed by assembly of a nucleoprotein filament that contains, among others, the proteins Rad51 and Rad52. This filament then invades the homologous DNA sequence of a sister chromatid, which becomes the template for essentially error-free repair of the broken ends by DNA synthesis. HR is active during S and G₂ phases of the cell cycle, that is, after DNA replication, when an identical copy of the DNA strand—the sister chromatid—is available.

The BRCA1 and BRCA2 gene products are also implicated in HR (and, possibly, NHEJ as well) because they interact with the Rad51 protein. Defects in these genes are associated with hereditary breast and ovarian cancer.⁴⁴

Mismatch Repair

The primary role of mismatch repair (MMR) is to eliminate from newly synthesized DNA errors such as base/base mismatches and insertion/deletion loops caused by DNA polymerase.⁴⁵ Descriptively, this process consists of three steps: mismatch recognition and assembly of the repair complex, degradation of the error-containing strand, and repair synthesis. In humans, MMR involves at least five proteins, including hMSH2, hMSH3, hMSH6, hMLH1, and hPMS2, as well as other members of the DNA repair and replication machinery.

One manifestation of a defect in mismatch repair is microsatellite instability, mutations observed in DNA segments containing repeated sequence motifs.⁴⁶ Collectively, this causes the cell to assume a hypermutable state (“mutator phenotype”) that has been associated with certain cancer predisposition syndromes, in particular, hereditary nonpolyposis colon cancer (HNPCC, sometimes called Lynch syndrome).^47,48

What Is Meant by “Cell Death”?

The traditional definition of death as a permanent, irreversible cessation of vital functions is not the same as what constitutes death to the radiation biologist or oncologist. For proliferating cells—including those maintained in vitro, the stem cells of normal tissues, and tumor clonogens—cell death in the radiobiologic sense refers to a loss of reproductive integrity, that is, the cell has been rendered unable to sustain proliferation indefinitely. This type of “reproductive” or “clonogenic” death does not preclude the possibility that a cell may remain physically intact and metabolically active and able to continue its tissue-specific functions for some time after undergoing irradiation. In fact, some reproductively dead cells can even complete a few additional mitoses before finally succumbing to death in the more traditional sense.⁵²

Apoptosis, or programmed cell death, is a type of nonreproductive or interphase death commonly associated with embryonic development and normal tissue remodeling and homeostasis.⁵³ However, certain normal tissue and tumor cells also undergo apoptosis following irradiation, including normal cells of hematopoietic or lymphoid origin or crypt cells of the small intestine and salivary gland, plus some experimental tumor cell lines of gynecologic and hematologic origin.⁵⁴ Cells undergoing apoptosis exhibit a number of characteristic morphologic (e.g., nuclear condensation and fragmentation, membrane blebbing) and biochemical (DNA degradation) changes that culminate in the fragmentation and lysis of the cell, often within 12 hours of irradiation and before the first postirradiation mitosis. Ultimately, the remains of apoptotic cells are phagocytized by neighboring cells and therefore do not cause the type of inflammatory response, tissue destruction, and disorganization characteristic of necrosis. Apoptosis is an active and carefully regulated pathway that involves several genes and gene products and an appropriate stimulus that activates the pathway. The molecular biology of apoptosis, the apoptosis-resistant phenotype noted for many types of tumor cells, and the role that radiation may play in the process are discussed in detail in Chapter 2.

Most assays of radiosensitivity of cells and tissues, including those described below, use reproductive integrity, either directly or indirectly, as an endpoint. Although such assays have served the radiation oncology community well in terms of elucidating dose-response relationships for normal tissues and tumors, it is now clear that reproductive death is not necessarily the whole story. What remains unclear is whether and to what extent apoptosis contributes to our traditional measures of radiosensitivity based on reproductive integrity, and whether pretreatment assessment of apoptotic propensity in tumors or dose-limiting normal tissues is of prognostic significance. The interrelationships between these different pathways of cell death can be quite complex. Meyn,⁵⁴ for example, has suggested that a tumor with a high spontaneous apoptotic index may be inherently more radiosensitive because cell death might be triggered by lower doses than are usually required to cause reproductive death. Also, tumors that readily undergo apoptosis may have higher rates of cell loss, the net effect of which would be to partially offset cell production, thereby reducing the number of tumor clonogens. On the other hand, not all types of tumor cells undergo apoptosis, and even among those that do, radiotherapy itself could have the undesirable side effect of selecting for subpopulations of cells that are already apoptosis-resistant.

Cell Survival and Dose-Response Curve Models

As had been the case during the early years of radiation therapy, advances in radiation physics, specifically biophysics, preceded advances in radiation biology. Survival curve theory originated in a consideration of the physics of energy deposition in matter by ionizing radiation. Early experiments with macromolecules and prokaryotes established the fact that dose-response relationships could be explained, in principle, by the random and discrete nature of energy absorption, if it was assumed that the response resulted from critical “targets” receiving random “hits.”⁵⁵ With an increasing number of shouldered survival and dose-response curves being described for cells irradiated both in vitro and in vivo, various equations were developed to fit these data. Target theory pioneers studied a number of different endpoints in the context of target theory, including enzyme inactivation in cell-free systems,²⁸ cellular lethality, chromosomal damage, and radiation-induced cell cycle perturbations in microorganisms.^28,56 Survival curves, in which the log of the “survival” of a certain biologic activity was plotted as a function of the radiation dose, were found to be either exponential or sigmoid in shape, the latter usually noted for the survival of more complex organisms.²⁸

It was thought that exponential survival curves resulted from the single-hit, “all-or-nothing” inactivation of a single target, resulting in the loss of activity. A mathematical expression used to fit this type of dose-response relationship was S = . In this equation, S is the fraction of cells that survive a given dose, D, and D₀ is the dose increment that reduces the cell survival to 37% (1/e) of some initial value on the exponential portion of the curve (i.e., a measure of the reciprocal of the slope). The sigmoid survival curves, characterized by a shoulder at low doses, were consistent with target theory if one assumed that either multiple targets or multiple hits in a single target (or a combination of both) were necessary for radiation inactivation. A mathematical expression based on target theory that provided a fairly good fit to survival data was S = 1 − (1 − )ⁿ, with n being the back extrapolation of the exponential portion of the survival curve to zero dose (i.e., a measure of the width of the shoulder). Implicit in this multitarget model was that damage had to accumulate before the overall effect was registered.

It soon became apparent that some features of this model were inadequate.⁵⁷ The most obvious problem was that the single-hit, multitarget equation predicted that survival curves should have initial slopes of zero, that is, that for vanishingly small doses (e.g., repeated small doses per fraction or continuous low dose-rate exposure), the probability of cell killing would approach zero. This was decidedly not what was observed in practice for either mammalian cell survival curves or as inferred from clinical studies in which highly fractionated or low dose-rate treatment schedules were compared with more conventional fractionation. There was no fractionation schedule that produced essentially no cell killing, all other radiobiologic factors being equal.

A somewhat different interpretation of cell survival was proposed by Kellerer and Rossi⁵⁸ during the late 1960s and early 1970s. The linear-quadratic or “alpha-beta” equation, S = , was shown to fit many survival data quite well, particularly in the low-dose region of the curve, and also provided for the negative initial slope investigators had described.⁵⁷ In this expression, S is again the fractional cell survival following a dose D, α is the rate of cell kill by a single-hit process, and β is the rate of cell kill by a two-hit mechanism.

The theoretical derivation of the linear-quadratic equation is based on two very different sorts of observations. Based on microdosimetric considerations, Kellerer and Rossi⁵⁷ proposed that a radiation-induced lethal lesion resulted from the interaction of two sublesions. According to this interpretation, the αD term is the probability of these two sublesions being produced by a single event (the “intra-track” component), whereas βD² is the probability of the two sublesions being produced by two separate events (the “inter-track” component). Chadwick and Leenhouts⁵⁹ derived the same equation based on a different set of assumptions, namely that a DSB in DNA was a lethal lesion and that such a lesion could be produced by either a single energy deposition involving both strands of DNA or by two separate events, each involving a single strand.

A comparison of the features and parameters of the target theory and linear-quadratic survival curve expressions is shown in Figure 1-7.

Figure 1-7 Comparison of two mathematical models commonly used to fit cell survival curve data. A, The single-hit, multitarget model is shown with its associated parameters, D₀, n, and D_q. Although this model has since been invalidated, values for its parameters are still used for comparative purposes. D₀, dose increment that reduces the cell survival to 37% (1/e) of some initial value on the exponential portion of the curve (i.e., a measure of the reciprocal of the slope); D_q, a measure of the width of the shoulder; n, back extrapolation of the exponential portion of the survival curve to zero dose (i.e., a measure of the steepness of the shoulder). B, The linear-quadratic model and its associated parameters, α and β, form the basis for current isoeffect formulas used in radiation therapy treatment planning.

Clonogenic Assays In Vitro

As mentioned previously, it was not until the mid-1950s that mammalian cell culture techniques were sufficiently refined to allow quantitation of the radiation responses of single cells.^60,61 The Puck and Marcus²⁵ acute-dose, x-ray survival curve for the human tumor cell line HeLa is shown in Figure 1-8. Following graded x-ray doses, the reproductive integrity of single HeLa cells was measured by their ability to form macroscopic colonies of at least 50 cells (corresponding to approximately six successful postirradiation cell divisions) on Petri dishes.

Figure 1-8 Clonogenic survival of HeLa cells in vitro as a function of x-ray dose. Like many mammalian cells of tumorigenic and nontumorigenic origin, the HeLa cell survival curve is characterized by a modest initial shoulder region (n = 2.0) followed by an approximately exponential final slope (D₀ ≈ 1.0 Gy).

Adapted from Puck TT, Marcus PI: Action of x rays on mammalian cells, J Exp Med 103:653, 1956; copyright permission of The Rockefeller University Press.

A number of features of this survival curve were of particular interest. First, qualitatively at least, the curve was similar in shape to those previously determined for many microorganisms, being characterized by a shoulder at low doses and a roughly exponential region at high doses. Of note, however, was the finding that the D₀ for HeLa cells was only 96 R, some 10-fold to 100-fold less than any D₀ determined for microorganisms, and 1000-fold to 10,000-fold less than any D₀ for the inactivation of isolated macromolecules.⁵² Thus cellular reproductive integrity was found to be a much more radiosensitive endpoint for HeLa cells than for prokaryotes or primitive eukaryotes. The value of the extrapolation number, n, for HeLa cells was approximately 2.0, indicating that the survival curve did have a small shoulder, but again, much smaller than that typically observed for microorganisms. Puck and Marcus²⁵ suggested that the n value was a reflection of the number of critical targets in the cell, each requiring a single hit before the cell would be killed, and further postulated that the targets were, in fact, the chromosomes themselves. However, the potential pitfalls of deducing mechanisms of radiation action from parameters of a descriptive survival curve model were soon realized.^62,63

Survival curves for other types of mammalian cells, regardless of whether they were derived from humans or laboratory animals, or from tumors or normal tissues, have been shown to be qualitatively similar to the original HeLa cell survival curve.

Clonogenic Assays In Vivo

To bridge the gap between the radiation responses of cells grown in culture and in an animal, Hewitt and Wilson⁶⁴ developed an ingenious method to assay single cell survival in vivo. Lymphocytic leukemia cells obtained from the livers of donor CBA mice were harvested, diluted, and inoculated into disease-free recipient mice. By injecting different numbers of donor cells, a standard curve was constructed that allowed a determination of the average number of injected cells necessary to cause leukemia in 50% of the recipient mice. It was determined that the endpoint of this titration, the “50% take dose” or TD₅₀, corresponded to an inoculum of a mere two leukemia cells. Using this value as a reference, Hewitt and Wilson then injected leukemia cells harvested from gamma-irradiated donor mice into recipients and again determined the TD₅₀ following different radiation exposures. In this way, the surviving fraction after a given radiation dose could be calculated from the ratio of the TD₅₀ for unirradiated cells to that for irradiated cells. Using this technique, a complete survival curve was constructed that had a D₀ of 162 R and an n value close to 2.0, values quite similar to those generated for cell lines irradiated in vitro. For the most part, in vivo survival curves for a variety of cell types were also similar to corresponding in vitro curves.

A similar trend was apparent when in vivo survival curves for nontumorigenic cells were first produced. The first experiments by Till and McCulloch65,66 using normal bone marrow stem cells were inspired by the knowledge that failure of the hematopoietic system was a major cause of death following total body irradiation and that lethally irradiated animals could be “rescued” by a bone marrow transplant. The transplanted viable bone marrow cells were observed to form discrete nodules or colonies in the otherwise sterilized spleens of irradiated animals.

Subsequently, these authors transplanted known quantities of irradiated donor bone marrow into lethally irradiated recipient mice and were able to count the resulting splenic nodules and then calculate the surviving fraction of the injected cells in much the same way as was done for in vitro experiments. The D₀ for mouse bone marrow was 0.95 Gy.⁶⁶ Other in vivo assay systems based on the counting of colonies or nodules included the skin epithelium assay of Withers,⁶⁷ the intestinal crypt assays of Withers and Elkind,^68,69 and the lung colony assay of Hill and Bush.⁷⁰ During the late 1960s and early 1970s, it also became possible to do excision assays, in which tumors irradiated in vivo were removed and enzymatically dissociated and single cells from the tumors were plated for clonogenic survival in vitro, thereby allowing more quantitative measurement of survival and avoiding some of the pitfalls of in vivo assays.⁷¹

Non-Clonogenic Assays In Vivo

Unfortunately, some normal tissues and tumors are not amenable to clonogenic assays. Thus new assays were needed that had clinical relevance yet did not rely on reproductive integrity as an endpoint. Use of such assays required one leap of faith, however, namely that the endpoints assessed would have to be a consequence of the killing of clonogenic cells, although not necessarily in a direct, one-to-one manner. Because nonclonogenic assays do not directly measure cell survival as an endpoint, data derived from them and plotted as a function of radiation dose are properly called dose- response curves rather than cell survival curves, although such data are often analyzed and interpreted similarly.

Historically, among the first nonclonogenic assays was the mean lethal dose or LD₅₀ assay, in which the (whole body) radiation dose to produce lethality in approximately 50% of the test subjects is determined, usually at a fixed time after irradiation, such as 30 (LD_50/30) or 60 days (LD_50/60). Clearly, the LD₅₀ assay is not very specific in that the cause of death can result from damage to a number of different tissues.

Another widely used nonclonogenic method to assess normal tissue radioresponse is the skin reaction assay, originally developed by Fowler and colleagues.⁷² Pigs were often used because their skin is similar to that of humans in several key respects. An ordinate scoring system was used to compare and contrast different radiation schedules and was derived from the average severity of the skin reaction noted during a certain time period (specific to the species and whether the endpoint occurred early or late) following irradiation. For example, for early skin reactions, a skin score of “1” might correspond to mild erythema, whereas a score of “4” might correspond to confluent moist desquamation over more than half of the irradiated area.

Finally, two common nonclonogenic assays for tumor response are the growth delay/regrowth delay assay⁷³ and the tumor control dose assay.⁷⁴ Both assays are simple and direct, are applicable to most solid tumors, and are clinically relevant. The growth delay assay involves the periodic measurement of a tumor’s dimensions as a function of time after irradiation and a calculation of the tumor’s approximate volume. For tumors that regress rapidly during and after radiotherapy, the endpoint scored is typically the time in days it takes for the tumor to regrow to its original volume at the start of irradiation. For tumors that regress more slowly, a more appropriate endpoint might be the time it takes for the tumor to grow or regrow to a specified size, such as three times its original volume. Dose-response curves are generated by plotting the amount of growth delay as a function of radiation dose.

The tumor control assay is a logical extension of the growth delay assay. The endpoint of this assay is the total radiation dose required to achieve a specified probability of (local) tumor control—usually 50% (TCD₅₀)—in a specified period of time after irradiation. The TCD₅₀ value is obtained from a plot of the percentage of tumors locally controlled as a function of total dose. The slope of the resulting dose-response curve may be used for comparative purposes as a measure of the tumor’s inherent “radiosensitivity” and/or its degree of heterogeneity. More heterogeneous tumors tend to have shallower dose-response curves than more homogeneous ones, as do spontaneous tumors relative to experimental ones maintained in inbred strains of mice.

Repair in Tissues

When considering repair phenomenon in intact tissues, it is important to remember that both the magnitude of the repair (related to the shape of the shoulder region of the single-dose survival curve and the dose delivered) and the rate of the repair can influence how the tissue behaves during a course of radiation therapy. For example, a particular tissue—normal or tumor—may be quite capable of repairing most damage produced by each dose fraction, but if the interfraction interval is so short as to not allow all the damage to be repaired prior to the next dose, the tolerance of that tissue will be less than otherwise anticipated. Second, the sparing effect of dose fractionation for both normal and tumor tissues can be explained largely by sublethal damage recovery between fractions, but at sufficiently small doses per fraction the degree of sparing will reach a maximum below which no further sparing occurs, all other radiobiologic factors being equal. This is a reflection of the fact that some radiation damage is necessarily lethal and not modifiable by either further fractionation or changing postirradiation conditions.

Methodology

Several techniques were subsequently developed for the collection of synchronized cells in vitro. One of the most widely used was the mitotic harvest or “shake-off” technique first described by Terasima and Tolmach.86,87 By shaking cultures, mitotic cells, which tend to round up and become loosely attached to the culture vessel’s surface, can be dislodged, collected along with the overlying growth medium, and inoculated into new culture flasks. By incubating these flasks at 37° C, cells begin to proceed synchronously into G₁ phase (and semisynchronously thereafter). Thus, by knowing the length of the various phase durations for the cell type being studied and then delivering a radiation dose at a time of interest after the initial synchronization, the survival response of cells in different phases of the cell cycle can be determined.

A second synchronization method involved the use of DNA synthesis inhibitors such as fluorodeoxyuridine⁸⁸ and, later, hydroxyurea⁸⁹ to selectively kill S phase cells yet allow cells in other phases to continue cell cycle progression until they become blocked at the border of G₁ and S phase. By incubating cells in the presence of these inhibitors for times sufficient to collect nearly all cells at the block point, large numbers of cells can be synchronized. In addition, the inhibitor technique has two other advantages, namely that some degree of synchronization is possible in vivo⁹⁰ as well as in vitro and that by inducing synchrony at the end of G₁ phase, a higher degree of synchrony can be maintained for longer periods than if synchronization had been at the beginning of G₁. On the other hand, the mitotic selection method does not rely on the use of agents that could perturb the normal cell cycle kinetics of the population under study.

Developments in the early 1970s provided what are now considered some of the most valuable tools for the study of cytokinetic effects: the flow cytometer and its offshoot, the fluorescence-activated cell sorter.⁹¹ These have largely replaced the aforementioned longer and more labor intensive cell cycle synchronization methods. Using this powerful technique, single cells are stained with a fluorescent probe that binds stoichiometrically to a specific cellular component, DNA in the case of cell cycle distribution analysis. The stained cells are then introduced into a pressurized flow cell and forced to flow single file and at a high rate of speed through an intense, focused laser beam that excites the fluorescent dye. The resulting light emission from each cell is collected by photomultiplier tubes and recorded. Then a frequency histogram of cell number as a function of relative fluorescence, is generated, with the amount of light output being directly proportional to DNA content of each cell examined. Accordingly, cells with a “1X” DNA content would correspond to cells in G₁ phase, cells with a “2X” DNA content, in G₂ or M phase, and cells with DNA contents between “1X” and “2X,” in the S phase of the cell cycle. By performing a mathematical fit to the DNA histogram, the proportion of cells in each phase of the cell cycle can be determined, the phase durations can be derived, and differences in DNA ploidy can be identified. DNA flow cytometry is quite powerful in that not only can a static measure of cell cycle distribution be obtained for a cell population of interest but, also, dynamic studies of, for example, transit through the various cell cycle phases, or treatment-induced kinetic perturbations, can be monitored over time (Figure 1-13). A flow cytometer can be outfitted to include a cell-sorting feature; in this case, cells analyzed for a property of interest can be collected in separate “bins” after they pass through the laser beam and, if possible, used for other experiments.

Figure 1-13 The analytic technique of flow cytometry has revolutionized the study of cell cycle kinetics by allowing rapid determination of DNA content in cells stained with a fluorescent dye that binds stoichiometrically to cellular DNA. A, Frequency distribution for a population of exponentially growing cells. The large and small peaks correspond to cells with G₁ (1×) and G₂/M (2×) phase DNA content, respectively; the cells in S phase have an intermediate DNA content. B to D, DNA histograms for a cell population synchronized initially in mitosis and then allowed to progress into G₁ (B), S, and G₂/M (C and D).

Age Response Through the Cell Cycle

Results of an age response experiment of Terasima and Tolmach,⁸⁷ using synchronized HeLa cells, are shown in the lower panel of Figure 1-14. Following a single dose of 5 Gy of x rays, cells were noted to be most radioresistant in late S phase. Cells in G₁ were resistant at the beginning of the phase but became sensitive toward the end of the phase, and G₂ cells were increasingly sensitive as they moved toward the highly sensitive M phase. In subsequent experiments by Sinclair,^92,93 age response curves for synchronized Chinese hamster V79 cells showed that the peak in resistance observed in G₁ HeLa cells was largely absent for V79 cells. This is also illustrated in Figure 1-14 (upper panel). Otherwise, the shapes of the age response curves for the two cell lines are similar. The overall length of the G₁ phase determines whether the resistant peak in early G₁ will be present; in general, this peak of relative radioresistance is only observed for cells with long G₁ phases. For cells with short G₁ phases, the entire phase is often intermediate in radiosensitivity. An analysis of the complete survival curves for synchronized cells^93,94 confirms that the most sensitive cells are those in M and late G₂ phase, in which survival curves are steep and largely shoulderless, and the most resistant cells are those in late S phase. The resistance of these cells is conferred by the presence of a broad survival curve shoulder rather than by a significant change in survival curve slope (Figure 1-15). When high LET radiations are used, the age response variation through the cell cycle is significantly reduced or eliminated, since survival curve shoulders are either decreased or removed altogether by such exposures (see also “Relative Biologic Effectiveness,” discussed later). Similar age response patterns have been identified for cells synchronized in vivo.⁹⁰

Figure 1-14 Cell cycle age response for sensitivity to radiation-induced cell killing in a representative rodent cell line (V79, upper panel) having a short G₁ phase duration and in a representative human cell line (HeLa, lower panel) having a long G₁ phase duration. Both cell lines exhibit a peak of radioresistance in late S phase and maximum radiosensitivity in late G₂/M phase. A second trough of radiosensitivity can be discerned near the G₁/S border for cells with long G₁-phase durations.

Adapted from Sinclair W: Dependence of radiosensitivity upon cell age. In Proceedings of the carmel conference on time and dose relationships in radiation biology as applied to radiotherapy. BNL Report 50203, Upton, NY, 1969, Brookhaven National Laboratory.

Figure 1-15 Cell survival curves for irradiated populations of Chinese hamster cells synchronized in different phases of the cell cycle (left panel), illustrating how these radiosensitivity differences translate into the age-response patterns shown in the right panel (see Fig. 1-14).

The existence of a cell cycle age response for ionizing radiation provided an explanation for the unusual pattern of SLDR observed for cells maintained at 37° C during the recovery interval (see Figure 1-9). In Elkind and Sutton’s experiments, exponentially-growing cells were used, that is, cells that were asynchronously distributed across the different phases of the cell cycle. The cells that survived irradiation tended to be those most radioresistant. As such, the remaining population was enriched with the more resistant cells and partially synchronized. For low LET radiation, those cells that were most resistant were the cells that were in S phase at the time of the first radiation dose. However, at 37° C, cells continued to progress through the cell cycle, so those surviving cells in S phase at the time of the first dose may have moved into G₂ phase by the time the second dose was delivered. Thus the observed survival nadir in the SLDR curve was not due to a loss or reversal of repair but rather, because the population of cells was now enriched in G₂ phase cells, which are inherently more radiosensitive. For even longer radiation-free intervals, it is possible that the cells surviving the first dose would transit from G₂ to M, and back into G₁ phase, dividing and doubling their numbers. In this case, the SLDR curve again shows a surviving fraction increase because the number of cells has increased. None of these cell cycle–related phenomena occur when the cells are maintained at room temperature during the radiation-free interval because movement through the cell cycle is inhibited under such conditions; in that case, all that is noted is the initial survival increase due to prompt SLDR.

Radiation-Induced Cell Cycle Blocks and Delays

Radiation is capable of disrupting the normal proliferation kinetics of cell populations. This was recognized by Canti and Spear⁹⁵ in 1927 and studied in conjunction with radiation’s ability to induce cellular lethality. With the advent of mammalian cell culture and synchronization techniques and time-lapse cinemicrography, it became possible for investigators to study mitotic and division delay phenomena in greater detail.

Mitotic delay, defined as a delay in the entry of cells into mitosis, is a consequence of “upstream” blocks or delays in the movement of cells from one cell cycle phase to the next. Division delay, a delay in the time of appearance of new cells at the completion of mitosis, is caused by the combined effects of mitotic delay and any further lengthening of the mitosis process itself. Division delay increases with dose, and is, on average, about 1 to 2 hours per Gray,⁸⁷ depending on the cell line.

The cell cycle blocks and delays primarily responsible for mitotic and division delay are, respectively, a block in the G₂-to-M phase transition and a block in the G₁-to-S phase transition. The duration of the G₂ delay, like the overall division delay, varies with cell type, but for a given cell type is both dose and cell cycle age dependent. In general, the length of the G₂ delay increases linearly with dose. For a given dose, the G₂ delay is longest for cells irradiated in S or early G₂ phase and shortest for cells irradiated in G₁ phase.¹⁰¹ Another factor contributing to mitotic and division delay is a block in the flow of cells from G₁ into S phase. For x-ray doses of about 6 Gy and above, there is a 50% decrease in the rate of tritiated thymidine uptake in exponentially growing cultures of mouse L cells. Little⁹⁶ reached a similar conclusion from G₁ delay studies using human liver LICH cells maintained as confluent cultures.

A possible role for DNA damage and its repair in the etiology of division delay was bolstered by the finding that certain cell types that either did not exhibit the normal cell cycle delays associated with radiation exposure (such as AT cells)³⁸ or, conversely, were treated with chemicals (such as caffeine) that ameliorated the radiation-induced delays⁹⁷ tended to contain higher amounts of residual DNA damage and to show increased radiosensitivity.

It is now known that the radiation-induced perturbations in cell cycle transit are under the control of cell cycle checkpoint genes, whose products normally govern the orderly (and unidirectional) flow of cells from one phase to the next. The checkpoint genes are responsive to feedback from the cell as to its general condition and readiness to transit to the next cell cycle phase. DNA integrity is one of the criteria used by these genes to help make the decision whether to continue traversing the cell cycle or to pause, either temporarily or, in some cases, permanently. Cell cycle checkpoint genes are discussed in Chapter 2.

Redistribution in Tissues

Because of the age response through the cell cycle, an initially asynchronous population of cells surviving a dose of radiation becomes enriched with S phase cells. Because of variations in the rate of further cell cycle progression, however, this partial synchrony decays rapidly. Such cells are said to have “redistributed”⁹⁸ with the net effect of sensitizing the population as a whole to a subsequent dose fraction (relative to what would have been expected had the cells remained in their resistant phases). A second type of redistribution, in which cells accumulate in G₂ phase (in the absence of cell division) during the course of multifraction or continuous irradiation because of a buildup of radiation-induced cell cycle blocks and delays, also has a net sensitizing effect. This has been observed during continuous irradiation by several investigators,⁹⁹ and in some of these cases, a net increase in radiosensitivity is seen at certain dose rates. This so-called “inverse dose rate effect,” where certain dose rates are more effective at cell killing than other, higher dose rates, was extensively studied by Bedford and associates.¹⁰⁰ The magnitude of the sensitizing effect of redistribution varies with cell type depending on which dose rate is required to stop cell division. For dose rates below the critical range that causes redistribution, some cells can escape the G₂ block and proceed on to cell division.

Linear Energy Transfer (LET)

The total amount of energy deposited in biologic materials by ionizing radiation (usually expressed in units of keV, ergs, or joules per gram or kilogram) is in and of itself insufficient to describe the net biologic consequences of those energy deposition events. As such, 1 Gy of x rays, although physically equivalent in terms of energy imparted per unit mass to 1 Gy of neutrons or alpha particles, does not produce equivalent biologic effects. It is the microdosimetric pattern of that energy deposition, that is, the spacing or density of the ionization events, that determines biologic effectiveness; this quantity—the average energy deposited locally per unit length of the ionizing particle’s track—is termed its linear energy transfer or LET.

LET is a function both of the charge and the mass of the ionizing particle. For example, photons set in motion fast electrons that have a net negative charge but a negligible mass. Neutrons on the other hand give rise to recoil protons or α-particles that possess one or two positive charges, respectively, and are orders of magnitude more massive than electrons. Neutrons therefore have a higher LET than photons and are said to be densely ionizing, whereas x or gamma rays are considered sparsely ionizing. The LET concept is illustrated in Figure 1-16 for both densely and sparsely ionizing radiations. For a given ionizing particle, the rate of energy deposition in the absorbing medium also increases as the particle slows down. Therefore, a beam of radiation can only be described as having an average value for LET.

Figure 1-16 Variation in the density of ionizing events along an incident particle’s track for radiations of different values of linear energy transfer (LET). The more closely the ionizing events are spaced, the more energy will be deposited in the target volume and, to a point, the more biologically effective per unit dose the type of radiation will be.

Representative LET values for types of radiations that have been used for radiation therapy include 0.2 keV/µm for cobalt-60 (⁶⁰Co) gamma rays; 2.0 keV/µm for 250 kVp x rays; approximately 0.5 to 5.0 keV/µm for protons of different energies; approximately 50 to 150 keV/µm for neutrons of varying energy; 100 to 150 keV/µm for alpha particles; and 100 to 2500 keV/µm for “heavy ions.” (In general, for a given type of radiation, the LET decreases with increasing energy.)

Relative Biologic Effectiveness

Insofar as the “quality” (LET) of the type of radiation influences its biologic effectiveness, two questions immediately come to mind. First, why do seemingly subtle differences in microdosimetric energy deposition patterns lead to vastly different biologic consequences? Second, how is this differing biologic effectiveness manifest in terms of the commonly used assays and model systems of classical radiobiology, and how can this difference be expressed in a quantitative way?

Because high LET radiations are more densely ionizing than their low LET counterparts, it follows that energy deposition in a particular “micro” target volume will be greater, and, therefore, more severe damage to biomolecules would be anticipated. In this case, the fraction of cell killing attributable to irreparable and unmodifiable DNA damage increases in relation to that caused by the accumulation of sublethal damage. As such, a number of radiobiologic phenomena commonly associated with low LET radiation are decreased or eliminated with high LET radiation. For example, there is little, if any, sublethal⁵² or potentially lethal damage recovery.¹⁰¹ This is manifest as a reduction or loss of the shoulder of the acute-dose radiation survival curve, little or no sparing effect of dose fractionation or dose rate, and a corresponding reduction in the tolerance doses for normal tissue complications, particularly for late effects.¹⁰² Variations in the age response through the cell cycle are reduced or eliminated for high LET radiation,⁹⁰ and the oxygen enhancement ratio (OER), a measure of the differential radiosensitivity of poorly oxygenated versus well-oxygenated cells (discussed later), decreases with increasing LET.¹⁰³ The dependence of OER on LET is illustrated in Figure 1-17; at an LET of approximately 100 keV/µm, the relative radioresistance of hypoxic cells is eliminated.

Figure 1-17 Relative biologic effectiveness (RBE) is demonstrated on the left y-axis as a function of linear energy transfer (LET) for a number of biologic endpoints, including production of chromosomal aberrations, cell kill, and tissue reactions. The RBE rises to a maximum corresponding to an LET of approximately 100 KeV/µm and then decreases as the LET continues to rise. Shown below the x axis are the ranges of LET for photons plus several different types of particulate radiations that have been used clinically. Dependence of the oxygen enhancement ratio (OER) on LET is shown on the y axis.

In light of these differences between high and low LET radiations, the term relative biologic effectiveness (RBE) has been used to compare and contrast two radiation beams of different LETs. RBE is defined as the ratio of doses of a known type of low LET radiation (historically, 250 kVp x rays were the standard, but ⁶⁰Co gamma rays also can be used) to that of a higher LET radiation, to yield the same biologic endpoint. However, RBE does not increase indefinitely with increasing LET but reaches a maximum at approximately 100 keV/µm and then decreases again, yielding approximately a bell-shaped curve.

One interpretation of why there is an apparent maximal LET of 100 keV/µm in terms of biologic effectiveness is that at this ionization density, the average separation between ionizing events corresponds roughly to the diameter of the DNA double helix (approximately 2 nm). As such, radiation characterized by this LET has the highest probability of producing DSBs in DNA, the putative lethal lesion, by the passage of a single charged particle. Lower LET radiations have a smaller likelihood of producing such “two-hit” lesions from a single particle track and are therefore less biologically effective. Radiation beams of higher LET than the optimal are also less biologically effective because some of the energy is wasted as more ionization events than minimally necessary are deposited in the same local area. This phenomenon has been termed the “overkill effect.”

Factors that Influence Relative Biologic Effectiveness

Relative biologic effectiveness is highly variable and depends on several irradiation parameters, including the type of radiation, total dose, dose rate, dose fractionation pattern, and biologic effect being assayed. As such, when quoting an RBE value, the exact experimental conditions used to measure it must be stated. Because increasing LET differentially reduces the shoulder region of the radiation survival curve compared with its exponential or near-exponential high-dose region, the single-dose RBE increases with decreasing dose (Fig. 1-18). Second, the RBE determined by comparing two isoeffective acute doses is less than the RBE calculated from two isoeffective (total) doses given either as multiple fractions or at a low dose rate. This occurs because the sparing effect of fractionation magnifies differences in the initial slope or shoulder region of cell survival or tissue dose response curves (Fig. 1-19).

Figure 1-18 Theoretical cell survival curves for x rays and neutrons, illustrating the increase in relative biologic effectiveness (RBE) with decreasing dose. This occurs because higher linear energy transfer (LET) radiations preferentially decrease or eliminate the shoulder on cell survival curves.

Adapted from Nias A: Clinical radiobiology, ed 2, New York, 1988, Churchill Livingstone.

Figure 1-19 Increase in the relative biologic effectiveness (RBE) of neutrons relative to x rays when comparing single radiation doses with fractionated treatment. For a given level of cell kill (or other approximately isoeffective endpoint), the more highly fractionated the treatment, the higher the RBE.

Mechanistic Aspects of the Oxygen Effect

A more rigorous analysis of the nature of the oxygen effect is possible with cells or bacteria grown in vitro. Historically, oxygen had been termed a “dose-modifying agent,” that is, that the ratio of doses to achieve a given survival level under hypoxic and aerated conditions was constant, regardless of the survival level chosen. This dose ratio to produce the same biologic endpoint is termed the oxygen enhancement ratio, or OER, and is used for comparative purposes (Fig. 1-21). The OER typically has a value of between 2.5 and 3.0 for large single doses of x or gamma rays, 1.5 to 2.0 for radiations of intermediate LET, and 1.0 (i.e., no oxygen effect) for high LET radiations.

Figure 1-21 Representative survival curves for cells irradiated with x rays in the presence (aerobic) or virtual absence (hypoxic) of oxygen. The oxygen enhancement ratio (OER) is defined as the ratio of doses under hypoxic-to-aerobic conditions to yield the same biologic effect, which in this case is a cell surviving fraction (SF) of 0.05.

Increasingly, there is evidence that oxygen is not strictly dose-modifying. Several studies have shown that the OER for sparsely ionizing radiation is lower at lower doses than at higher doses. Lower OERs for doses per fraction in the range commonly used in radiotherapy have been inferred indirectly from clinical and experimental tumor data and more directly in experiments with cells in culture.^108,109 It has been suggested that the lower OERs result from an age response for the oxygen effect, not unlike the age responses for radiosensitivity and cell cycle delay.²⁷ Assuming cells in G₁ phase of the cell cycle have a lower OER than those in S phase, and because G₁ cells are also more radiosensitive, they would tend to dominate the low-dose region of the cell survival curve.

Although the exact mechanism(s) of the oxygen effect are likely complex, a fairly simplistic model can be used to illustrate our current understanding of this phenomenon (Fig. 1-22). The radical competition model holds that oxygen acts as a radiosensitizer by forming peroxides in important biomolecules (including, but not necessarily limited to, DNA) already damaged by radiation exposure, thereby “fixing” the radiation damage. In the absence of oxygen, DNA can be restored to its preirradiated condition by hydrogen donation from endogenous reducing species in the cell, such as the free radical scavenger glutathione (a thiol compound). In essence, this can be considered a type of very fast chemical restitution or repair. These two processes, fixation and restitution, are considered to be in a dynamic equilibrium, such that changes in the relative amounts of either the radiosensitizer, oxygen, or the radioprotector, glutathione, tip the scales in favor of either fixation (more damage, more cell killing, greater radiosensitivity) or restitution (less damage, less cell killing, greater radioresistance).

Figure 1-22 Schematic representation of the proposed mechanism of action for the oxygen effect. The radical competition model holds that oxygen acts as a radiosensitizer by forming peroxides in DNA already damaged by radiation, thereby fixing the damage. In the absence of oxygen, DNA can be restored to its preirradiated state by hydrogen donation from endogenous reducing species in the cell, such as the free radical scavenger glutathione. An oxygen-mimetic, hypoxic cell radiosensitizer may be used to substitute for oxygen in these fast, free radical reactions, or an exogenously supplied thiol compound may be used to act as a radioprotector.

Consistent with this free radical-based interpretation of the oxygen effect is the finding that, for all intents and purposes, oxygen need only be present during the irradiation (or no more than about 5 ms after irradiation) to produce an aerobic radioresponse.^110,111 The concentration of oxygen necessary to achieve maximum sensitization is quite small, evidence for the high efficiency of oxygen as a radiosensitizer. A sensitivity midway between a fully hypoxic and fully aerobic radioresponse is achieved at an oxygen tension of about 3 mm of mercury, corresponding to about 0.5% oxygen, more than a factor of 10 lower than partial pressures of oxygen usually encountered in normal tissues. This value of 0.5% has been termed oxygen’s “k-value” and is obtained from an oxygen “k-curve” of relative radiosensitivity plotted as a function of oxygen tension¹¹² (Fig. 1-23).

Figure 1-23 An oxygen k-curve illustrates the dependence of radiosensitivity on oxygen concentration. If a fully anoxic cell culture is assigned a relative radiosensitivity of 1.0, introducing even 0.5% (3 mm Hg) oxygen into the system increases the radiosensitivity of cells to 2.0. By the time the oxygen concentration reaches about 2%, cells respond as if they are fully aerated (i.e., radiosensitivity ≈ 3.0). The green-shaded area represents the approximate range of oxygen concentrations encountered in human normal tissues.

Reoxygenation in Tumors

After the convincing demonstration of hypoxic cells in a mouse tumor,¹⁰⁶ it was assumed that human tumors contained a viable hypoxic fraction as well. However, if human tumors contained even a small fraction of clonogenic hypoxic cells, simple calculations suggested that tumor control would be nearly impossible with radiation therapy.¹¹³ Because therapeutic successes obviously did occur, some form of reoxygenation must take place during the course of multifraction irradiation. This was not an unreasonable idea, because the demand for oxygen by sterilized cells would gradually decrease as such cells were removed from the tumor and a decrease in tumor size, a restructuring of tumor vasculature, or intermittent changes in blood flow could make oxygen available to these previously hypoxic cells.

The reoxygenation process was extensively studied by van Putten and Kallman,¹¹⁴ who serially determined the fraction of hypoxic cells in a mouse sarcoma during the course of a typical multifraction irradiation protocol. The fact that the hypoxic fraction was about the same at the end of treatment as at the beginning of treatment was strong evidence for a reoxygenation process. Reoxygenation of hypoxic, clonogenic tumor cells during an extended multifraction treatment would increase the therapeutic ratio, assuming that normal tissues remained well-oxygenated. This is thought to be another major factor in the sparing of normal tissues relative to tumors during fractionated radiation therapy.

What physiologic characteristics would lead to tumor reoxygenation during a course of radiotherapy, and at what rate would this be expected to occur? One possible cause of tumor hypoxia, and by extension, a possible mechanism for reoxygenation, was suggested by Thomlinson and Gray’s pioneering work.¹⁰⁵ The type of hypoxia that they described is what is now called chronic hypoxia or diffusion-limited hypoxia. This results from the tendency of tumors to outgrow their blood supply, and given that oxygen is readily consumed by tumor cells, it follows that natural gradients of oxygen tension should develop as a function of distance from blood vessels. Cells situated beyond the diffusion distance of oxygen would be expected to be dead or dying, yet in regions of chronically low oxygen tension, clonogenic and radioresistant hypoxic cells may persist. Should the tumor shrink as a result of radiation therapy or if the cells killed by radiation have a decreased demand for oxygen, it is likely that this would allow some of the chronically hypoxic cells to reoxygenate. However, such a reoxygenation process could be quite slow—on the order of days or more—depending on how quickly tumors regress during treatment. The patterns of reoxygenation in some experimental rodent tumors are consistent with this mechanism of reoxygenation, but others are not.

Other rodent tumors reoxygenate very quickly, on a time scale of minutes to hours.¹¹⁵ This occurs in the absence of any measurable tumor shrinkage or change in oxygen utilization by tumor cells. In such cases, the model of chronic diffusion-limited hypoxia, and slow reoxygenation does not fit the experimental data. During the late 1970s, Brown and colleagues¹¹⁶ proposed that a second type of hypoxia may exist in tumors, an acute perfusion-limited hypoxia. Based on the growing understanding of the vascular physiology of tumors, it was clear that tumor vasculature was often abnormal in both structure and function secondary to abnormal angiogenesis. If tumor vessels were to close transiently from temporary blockage by blood cell components, vascular spasm, or high interstitial fluid pressure in the surrounding tissue, the tumor cells in the vicinity of those vessels would become acutely hypoxic almost immediately. Then, assuming blood flow resumed in minutes to hours, these cells would reoxygenate. However, this type of hypoxia can also occur in the absence of frank closure or blockage of tumor vessels, from, for example, vascular shunting or decreased red blood cell flux or overall blood flow rate.¹¹⁷ Because of this, the name “perfusion-limited” hypoxia is perhaps misleading; a better moniker might be “fluctuant” or “intermittent” hypoxia. Although intermittent hypoxia would explain the rapid reoxygenation observed for some tumors, it does not preclude the existence of chronic diffusion-limited hypoxia.

Intermittent hypoxia with rapid reoxygenation has since been demonstrated unambiguously for rodent tumors by Chaplin and colleagues.¹¹⁸ It is still not clear how many human tumors contain regions of hypoxia (although many do—see later discussion), what type(s) of hypoxia is present, whether this varies with tumor type or site, and whether and how rapidly reoxygenation occurs. However, the knowledge that tumor hypoxia is a diverse and dynamic process opens up a number of possibilities for the development of novel interventions designed to cope with, or even exploit, hypoxia.

Measurement of Hypoxia in Human Tumors

Despite prodigious effort directed at understanding tumor hypoxia and developing strategies to combat the problem, it was not until the mid-1980s that these issues could be addressed for human tumors because there was no way to measure hypoxia directly. Before that time, the only way to infer that a human tumor contained treatment-limiting hypoxic cells was by using indirect, nonquantitative methods. Some indirect evidence supporting the notion that human tumors contained clonogenic, radioresistant, hypoxic cells includes the following:

In 1988, one of the first studies showing a strong association between directly measured oxygenation status in tumors and clinical outcome was published by Gatenby.¹²⁴ An oxygen-sensing electrode was inserted into the patient’s tumor, and multiple readings were taken at different depths along the probe’s track. The electrode was also repositioned in different regions of the tumor to assess intertrack variability in oxygen tension. Both the arithmetic mean PO₂ value for a particular tumor as well as the tumor volume-weighted PO₂ value directly correlated with local control rate. A high tumor oxygen tension was associated with a high complete response rate, and vice versa. In a similar, prospective study, Höckel and colleagues^4,125 concluded that pretreatment tumor oxygenation was a strong predictor of outcome among patients with intermediate- and advanced-stage cervical carcinoma (Figure 1-24).

Figure 1-24 The disease-free survival probability of a small cohort of cervical cancer patients stratified according to pretreatment tumor oxygenation measured using an oxygen electrode.

Adapted from Höckel M, Knoop C, Schlenger K, et al: Intratumoral PO₂ predicts survival in advanced cancer of the uterine cervix, Radiother Oncol 26:45, 1993.

Today, oxygen electrodes are still considered the “gold standard” against which all other methods for measuring tumor oxygenation are compared. Unfortunately, the method has its limitations. One weakness is that relative to the size of individual tumor cells, the electrode is large, averaging an outer diameter of 300 µm, a tip recess of 120 µm, and a sampling volume of about 12 µm in diameter.¹²⁶ Thus, not only is the oxygen tension measurement regional but the insertions and removals of the probe no doubt perturb the oxygenation status. Another problem is that there is no way to determine whether the tumor cells are viable or not. If such cells are hypoxic but not clonogenic, they would not be expected to affect the radiotherapy outcome.

A second direct technique for measuring oxygenation status takes advantage of a serendipitous finding concerning how hypoxic cell radiosensitizers are metabolized. Certain classes of radiosensitizers, including the nitroimidazoles, undergo a bioreductive metabolism in the absence of oxygen that leads to their becoming covalently bound to cellular macromolecules.127,128 Assuming that the bioreductively bound drug could be quantified by radioactive labeling¹²⁹ or tagged with an isotope amenable to detection using positron emission tomography¹³⁰ or magnetic resonance spectroscopy,¹³¹ a direct measure of hypoxic fraction can be obtained.

Another approach to detecting those cells containing bound drug was developed by Raleigh and colleagues.^132,133 This immunohistochemical method involved the development of antibodies specific for the already-metabolized, bound nitroimidazoles. After injecting the parent drug, allowing time for the reductive metabolism to occur, and taking biopsies of the tumor and preparing histopathology slides, the specific antibody is applied to the slides and regions containing the bound drug are visualized directly. This immunostaining method has the distinct advantages that hypoxia can be studied at the level of individual tumor cells,¹³⁴ spatial relationships between regions of hypoxia and other tumor physiologic parameters can be assessed,¹³⁵ and the drug does not perturb the tumor microenvironment. However, the method remains an invasive procedure, is labor intensive, does not address the issue of the clonogenicity of stained cells (although such cells do have to be metabolically active), and requires that multiple samples be taken because of tumor heterogeneity.

One hypoxia marker based on the immunohistologic method, HypoxyProbe-1, is commercially available. It detects reductively-bound pimonidazole and has been used in experimental and clinical studies around the world.^136,137 Applying HypoxyProbe-1 to human tumor specimens yields a range of hypoxic fractions similar to that noted for experimental rodent tumors and a mean value of approximately 15%.^133,138 This marker can also be used to probe disease states other than cancer that may have the induction of tissue hypoxia as part of their etiology, such as cirrhosis of the liver^139,140 and ischemia-reperfusion injury in the kidney.¹⁴¹

There is also considerable interest in endogenous markers of tissue hypoxia,^142,143 which could reduce to some extent the procedural steps involved in, and the invasive aspects of, detecting hypoxia using exogenous agents. Among the endogenous cellular proteins being investigated in this regard are the hypoxia-inducible factor-1 alpha (HIF-1α, a transcription factor),¹⁴⁴ the enzyme carbonic anhydrase IX (CA-9 or CAIX),¹⁴⁵ glucose transporter-1 (GLUT-1),^146–148 and osteopontin.^149,150

Radiosensitizers

Radiosensitizers are loosely defined as chemical or pharmacologic agents that increase the cytotoxicity of ionizing radiation. “True” radiosensitizers meet the stricter criterion of being relatively nontoxic in and of themselves, acting only as potentiators of radiation toxicity. “Apparent” radiosensitizers still produce the net effect of making the tumor more radioresponsive, yet the mechanism is not necessarily synergistic nor is the agent necessarily nontoxic when given alone. Ideally, a radiosensitizer is only as good as it is selective for tumors. Agents that show little or no differential effect between tumors and normal tissues do not improve the therapeutic ratio and therefore may not be of much clinical utility. Table 1-2 summarizes some of the classes of radiosensitizers (and radioprotectors—discussed later) that have been used in the clinic.

TABLE 1-2 Selected Chemical Modifiers of Radiation Therapy

Bioreductive Drugs

In the wake of the failure of most hypoxic cell radiosensitizers to live up to their clinical potential, a new approach to combating the hypoxia problem emerged: the use of bioreductive drugs that are selectively toxic to hypoxic cells. Although these agents kill rather than sensitize hypoxic cells, the net effect of combining them with radiation therapy is apparent sensitization of the tumor due to the elimination of an otherwise radioresistant subpopulation. Such drugs have been shown to outperform the nitroimidazole radiosensitizers in experimental studies with clinically relevant fractionated radiotherapy.¹⁷² To the extent that hypoxic cells are also resistant to chemotherapy because of tumor microenvironmental differences in drug delivery, pH, or the cell’s proliferative status, complementary tumor cell killing might be anticipated for combinations of bioreductive agents and anticancer drugs as well.¹⁷³

Most hypoxia-specific cytotoxic drugs fall into three categories: the nitroheterocyclics, the quinone antibiotics. and the organic nitroxides.¹⁷⁴ All require bioreductive activation by nitroreductase enzymes such as cytochrome P450, DT-diaphorase, and nitric oxide synthase, to reduce the parent compound to its cytotoxic intermediate, typically an oxidizing free radical capable of damaging DNA and other cellular macromolecules. The active species is either not formed or else is immediately back-oxidized to the parent compound in the presence of oxygen, which accounts for its preferential toxicity under hypoxic conditions. Examples of nitroheterocyclic drugs with bioreductive activity include misonidazole and etanidazole^175,176 and dual-function agents such as RSU 1069.¹⁷⁷ The latter drug is termed “dual function” because its bioreduction also activates a bifunctional alkylating moiety capable of introducing crosslinks into DNA. Mitomycin C and several of its analogs (including porfiromycin and EO9) are quinones with bioreductive activity that have been tested in randomized clinical trials in head and neck tumors.¹⁷⁸

The lead compound for the third class of bioreductive drugs, the organic nitroxides, is tirapazamine (SR 4233)157,158,172 (see Table 1-2). The dose-limiting toxicity for single doses of tirapazamine is a reversible hearing loss; other effects observed include nausea and vomiting and muscle cramps.¹⁷⁹

Tirapazamine is particularly attractive because it both retains its hypoxia-selective toxicity over a broader range of (low) oxygen concentrations than the quinones and nitroheterocyclic compounds¹⁸⁰ and its “hypoxic cytotoxicity ratio,” the ratio of drug doses under hypoxic versus aerobic conditions to yield the same amount of cell killing, averages an order of magnitude higher than for the other classes of bioreductive drugs.¹⁷³ Laboratory and clinical data also support a tumor-sensitizing role for tirapazamine in combination with the chemotherapy agent cisplatinum.¹⁷⁹

To date, randomized phase II and III clinical trials with tirapazamine combined with radiotherapy and/or chemotherapy (particularly, cisplatinum) have yielded mixed results.179,181 Although the drug has in no way become a staple of clinical practice, its activity against some tumors warrants further investigation, as does the search for more effective agents from the same or a similar chemical class.

Normal Tissue Radioprotectors

Amifostine (WR 2721) (see Table 1-2) is a phosphorothioate compound developed by the United States Army for use as a radiation protector. Modeled after naturally occurring radioprotective sulfhydryl compounds such as cysteine, cysteamine, and glutathione,¹⁸² amifostine’s mechanism of action involves the scavenging of free radicals produced by ionizing radiation, radicals that otherwise could react with oxygen and “fix” the chemical damage. Amifostine can also detoxify other reactive species through the formation of thioether conjugates, and, in part because of this, the drug can also be used as a chemoprotective agent.¹⁸³ Amifostine is a pro-drug that is dephosphorylated by plasma membrane alkaline phosphatase to the free thiol WR 1065, the active metabolite. As is the case with the hypoxic cell radiosensitizers, amifostine need only be present at the time of irradiation to exert its radioprotective effect.

In theory, if normal tissues could be made to tolerate higher total doses of radiation through the use of radioprotectors, then the relative radioresistance of hypoxic tumor cells would be negated. However, encouraging preclinical studies demonstrating radioprotection of a variety of cells and tissues notwithstanding,^184,185 radioprotectors such as amifostine would not be expected to increase the therapeutic ratio unless they could be introduced selectively into normal tissues but not tumors. The pioneering studies of Yuhas and colleagues^154,186,187 addressed this issue by showing that the drug’s active metabolite reached a higher concentration in most normal tissues than in tumors and that this mirrored the extent of radioprotection or chemoprotection. The selective protection of normal tissues apparently results from tumors being both less able to convert amifostine to WR 1065 (because of lower concentrations of the required phosphatases) and to transport this active metabolite throughout the cell.

Dose-reduction factors (DRFs) (the ratio of radiation doses to produce an isoeffect in the presence versus the absence of the radioprotector) in the range of 1.5 to 3.5 are achieved for normal tissues, whereas the corresponding DRFs for tumors seldom exceed 1.2. Those normal tissues exhibiting the highest DRFs include bone marrow and gastrointestinal tract, liver, testicular, and salivary gland tissues.¹⁵⁴ Brain and spinal cord tissues are not protected by amifostine, and the oral mucosa is only marginally protected.¹⁵⁴ Comparable protection factors are obtained for some chemotherapy agents, including cyclophosphamide and cisplatin.^188,189 The dose-limiting toxicities associated with the use of amifostine include hypotension, emesis, and generalized weakness or fatigue.¹⁹⁰

Amifostine is currently indicated for the reduction of renal toxicity associated with repeated cycles of cisplatin chemotherapy in patients with advanced ovarian and non–small cell lung cancer. It is also approved for use in patients receiving radiotherapy for head and neck cancer, in the hopes of reducing xerostomia secondary to exposure of the parotid glands.

Finally, just as there are apparent radiosensitizers there are also apparent radioprotectors that have the net effect of allowing normal tissues to better tolerate higher doses of radiation and chemotherapy but through mechanisms of action not directly related to the scavenging of free radicals. Various biologic response modifiers, including cytokines, prostaglandins (such as misoprostol),^191,192 anticoagulants (such as pentoxifylline)^193,194 and protease inhibitors are apparent radioprotectors because they can interfere with the chain of events that normally follows the killing of cells in tissues by, for example, stimulating compensatory repopulation or preventing the development of fibrosis. Finally, there is also growing interest in the use of biologic agents that inhibit apoptosis as normal tissue radioprotectors.^195,196 Such agents have little or no effect on tumor cells, most of which are already apoptosis-resistant.