**1. Introduction**

Therapy of glioblastoma has been very problematic with disappointing results using multiple therapeutic approaches. In general, glioblastomas are considered radioresistant tumors with different radiation modalities failing to control them in the clinic. However a comprehensive and detailed analysis of the radiosensitivity of glioblastoma cells has not been performed. We now present such an analysis in this chapter seeking a better definition of patterns of radiosensitivity in glioblastomas compared to other tumor cells. These data show that some glioblastomas have unusual responses to radiation that may render them more resistant to some forms of radiotherapy but also render them amenable to exploitation by other forms of radiotherapy.

Multiple mechanisms have been proposed to be associated with radioresistance in human glioblastoma cells: Bao et al (1) have suggested increased DNA damage response. Karim et al (2) have proposed differential cyclo-oxygenase response in radioresistant glios. Brandani et al (3) have suggested HSP 70 elevation. Akuguka et al (4) have suggested increased rates in DNA double strand break rejoining association with micronuclei. Scmidberger et al (5) observed variation interferon-induced β associates with increased radiosensitivity in four out of five glioblastomas. Yao et al (6) suggest variation in cell cycle arrest, modulation of the expression of cyclin-dependent kinase inhibitors, and autophagy. Streffer et al (7) showed BCL- family proteins modulate radiosensitivity in human malignant glioma cells. Kraus et al (8) showed aberrant p21 regulation in radioresistant primary glioblastoma multiforme cells bearing wild-type p53. Haas-Kogan (9) et al showed p53 function influences the effect of fractionated radiotherapy on glioblastoma tumors. Hsiao et al (10) showed functional expression of human p21(WAF1/CIP1) gene in rat glioma cells suppresses tumor growth in vivo and induces radiosensitivity. Yount et al (11) showed cell cycle synchrony unmasks the influence of p53 function on radiosensitivity of human glioblastoma cells. Britten et al (12) showed differential level of DSB repair fidelity effected by nuclear protein extracts derived from radiosensitive and radioresistant human tumour cells. Guichard et al (13) suggest potentially lethal damage repair as a possible determinant of human tumour radiosensitivity including glioblastoma. Kal et al (14) have suggested rhabdomyosarcomas, similar to glioblastomas are sensitive to low dose-rate irradiation.

Radiobiology of Radioresistant Glioblastoma 5

We have recently proposed that patterns of inactivation in tumor cells are expressed as two general responses, the alpha response and the omega (or quadratic) response, are in fact actually comprised of four distinct components induced sequentially at increasing doses (20). Three of these responses can be well fitted to a log-linear relationship and a Poisson coefficient could be calculated over each of these log-linear responses over distinct dosesegments that represented the rate of inactivation. We will discuss these results subsequently but our data suggested that radiosensitivity of any cell line can be described as coefficients that describe four sequentially induced responses in each cell line. Some

Our studies suggest that radiosensitivity of tumor cells cannot be expressed as a single parameter or histological type, but should be analyzed on the basis of descriptors of multiple responses that are specific to human tumor cell lines. We argue that the radiosensitivity phenotype of each cell line should be defined by a set of coefficients including: 1) relative rates of inactivation over four distinct, sequentially-induced components that comprise radiosensitivity of each cell; 2) the general radiosensitivity group into which cell lines segregate non-randomly based on the values of exceptional coefficients; these groups associated with tumor cell genotype; 3) modulation of inactivation by reduced dose-rates; and 4) modulation of inactivation by the effect of ionization density of delivered

radiations; and 5) modulation by in vivo mechanisms that are particular to genotype.

**1.2 Coefficients that describe four sequentially-induced responses in each human** 

We have measured clonogenic inactivation in multiple cell lines over different dosesegments (20). **T**hese studies showed that response of tumor cells to radiation in vitro can be resolved into two general responses, termed the alpha response and omega response (or quadratic response) that represent the overall rate of inactivation over the dose-segment from 0 to circa 2 Gy (alpha response) and over doses greater than circa 3 Gy (omega response). These two general responses can be approximated by the linear and quadratic components of the linear-quadratic model. These two general responses vary not only between glioblastoma cell lines and other tumor types but also vary between different

Our data show these two general responses are actually comprised of four more specific responses induced sequentially in each cell line. Thus one goal of this chapter is to measure coefficients that describe these four responses in tumor cells and to suggest their relative

As stated above we have defined four sequentially-induced responses in ten tumor cell lines by increasing doses from 0.0 to 10.0 Gy. These responses are common to all tumor cells, induced at the same doses but vary in the rates of inactivation over these common dose-

**The hypersensitive (H) response** is observed over the dose-segment from 0.0 to 0.10 Gy and is characterized by highest rates of clonogenic inactivation observed in tumor cells. This

radiosensitivity phenotypes of tumor cells to genetic or epigenetic properties.

In the next several paragraphs we identify those coefficients that together define the "radiosensitivity phenotype" of human tumor cells. One overall goal is to equate

glioblastoma cells, not others, show specific values for these four responses.

**1.1 Concept of a "radiosensitivity phenotype"** 

**tumor cell** 

glioblastoma lines.

importance in clinical radiotherapy.

segments. These four responses are:

These studies used multiple types of glioblastoma cells, but they did not define glioblastoma cells based on essential cellular response mechanisms. We will identify classes of glioblastoma cells that exhibit distinct mechanisms of radiosensitivity in vitro and in vivo. In general there is an overall correlation between radiosensitivity of tumor cells in vivo, radiosensitivity of xenograft tumors as measured in the laboratory and radiosensitivity of tumors in the clinic, although it is clear more studies between these three forms of radiosensitivity is needed. One purpose of this article is to provide such data for radioresistant glioblastoma

Radiosensitivity as assayed by clonogenic inactivation is a precise and accurate endpoint measurable over a wide range of inactivation levels (circa 105) induced by a wide range of radiation doses (circa 103). Over this dynamic range, clonogenic inactivation can be measured with acceptable variation. Further, clonogenic inactivation is a dichotomous endpoint based on whether individual cells are either clonogenically inactivated or not. Mathematically, this enables the application of Poisson statistics to estimate the probability of inactivation for each increment of dose. Since 1956 when Puck and Marcus (15) published the first "survival curve" such patterns have been examined to discern underlying mechanisms that produce cellular inactivation. Although "hit-target" theory did not identify exact "hits" defined as patterns of ionizations or exact "targets" that once induced inactivates cell, these studies demonstrated a continuing concept that improves such estimates over specific dose-segments. The observation of a log-linear relationship over a specific dose-segment, e.g. logarithm of cells inactivated are a linear function of dose, indicates a constant rate of inactivation over that dose segment.

From these early data a common pattern of inactivation was usually observed for tumor cells: a low rate of cell inactivation below circa 2 to 3 Gray (Gy) followed by increased rates of inactivation at higher doses. Further the patterns of inactivation at higher doses could be approximated as a log-linear response and the slope of such a dose-segment could be calculated as a Poisson probability of inactivation, usually expressed as the parameter Do, the dose needed to inactivate a single cell.

Results from the huge empirical data base obtained in the clinic for the relative effect of different doses and protocols that induced both tumor regression and normal tissue toxicity, clearly demonstrated radiotherapy of some tumors was more successful when multiple doses below 3 Gy were used. Mathematical models were proposed to explain the rate of inactivation of tumors at lower doses circa 1 to 3 Gy and the most successful model was the "linear-quadratic (LQ) model" as proposed by Hendry ( 16) and by Fowler ( 17, 18). The LQ model was based on the concept that tumor cell inactivation was induced at a linear rate at lower doses (the alpha response) but reflected a quadratic component at higher doses determined by a coefficient beta times the square of the dose (the beta response). These efforts failed to identify specific targets or hits, which in retrospect, was partially due to the complex processes involved in cellular inactivation. However one basic observation from these analyses is useful: identification of a dose-segment over which inactivation is loglinear, is still valid in identifying the rate of inactivation over such a dose-segment, represented by a single Poisson coefficient.

The pioneering work of Joiner and his colleagues in identifying "low-dose hyperradiosensitivity" (19) demonstrated that at lower doses (< 0.5 Gy) there were additional changes in rates of clonogenic inactivation that could not be well explained by the linearquadratic model.

These studies used multiple types of glioblastoma cells, but they did not define glioblastoma cells based on essential cellular response mechanisms. We will identify classes of glioblastoma cells that exhibit distinct mechanisms of radiosensitivity in vitro and in vivo. In general there is an overall correlation between radiosensitivity of tumor cells in vivo, radiosensitivity of xenograft tumors as measured in the laboratory and radiosensitivity of tumors in the clinic, although it is clear more studies between these three forms of radiosensitivity is needed. One purpose of this article is to provide such data for

Radiosensitivity as assayed by clonogenic inactivation is a precise and accurate endpoint measurable over a wide range of inactivation levels (circa 105) induced by a wide range of radiation doses (circa 103). Over this dynamic range, clonogenic inactivation can be measured with acceptable variation. Further, clonogenic inactivation is a dichotomous endpoint based on whether individual cells are either clonogenically inactivated or not. Mathematically, this enables the application of Poisson statistics to estimate the probability of inactivation for each increment of dose. Since 1956 when Puck and Marcus (15) published the first "survival curve" such patterns have been examined to discern underlying mechanisms that produce cellular inactivation. Although "hit-target" theory did not identify exact "hits" defined as patterns of ionizations or exact "targets" that once induced inactivates cell, these studies demonstrated a continuing concept that improves such estimates over specific dose-segments. The observation of a log-linear relationship over a specific dose-segment, e.g. logarithm of cells inactivated are a linear function of dose,

From these early data a common pattern of inactivation was usually observed for tumor cells: a low rate of cell inactivation below circa 2 to 3 Gray (Gy) followed by increased rates of inactivation at higher doses. Further the patterns of inactivation at higher doses could be approximated as a log-linear response and the slope of such a dose-segment could be calculated as a Poisson probability of inactivation, usually expressed as the parameter Do,

Results from the huge empirical data base obtained in the clinic for the relative effect of different doses and protocols that induced both tumor regression and normal tissue toxicity, clearly demonstrated radiotherapy of some tumors was more successful when multiple doses below 3 Gy were used. Mathematical models were proposed to explain the rate of inactivation of tumors at lower doses circa 1 to 3 Gy and the most successful model was the "linear-quadratic (LQ) model" as proposed by Hendry ( 16) and by Fowler ( 17, 18). The LQ model was based on the concept that tumor cell inactivation was induced at a linear rate at lower doses (the alpha response) but reflected a quadratic component at higher doses determined by a coefficient beta times the square of the dose (the beta response). These efforts failed to identify specific targets or hits, which in retrospect, was partially due to the complex processes involved in cellular inactivation. However one basic observation from these analyses is useful: identification of a dose-segment over which inactivation is loglinear, is still valid in identifying the rate of inactivation over such a dose-segment,

The pioneering work of Joiner and his colleagues in identifying "low-dose hyperradiosensitivity" (19) demonstrated that at lower doses (< 0.5 Gy) there were additional changes in rates of clonogenic inactivation that could not be well explained by the linear-

indicates a constant rate of inactivation over that dose segment.

the dose needed to inactivate a single cell.

represented by a single Poisson coefficient.

quadratic model.

radioresistant glioblastoma

We have recently proposed that patterns of inactivation in tumor cells are expressed as two general responses, the alpha response and the omega (or quadratic) response, are in fact actually comprised of four distinct components induced sequentially at increasing doses (20). Three of these responses can be well fitted to a log-linear relationship and a Poisson coefficient could be calculated over each of these log-linear responses over distinct dosesegments that represented the rate of inactivation. We will discuss these results subsequently but our data suggested that radiosensitivity of any cell line can be described as coefficients that describe four sequentially induced responses in each cell line. Some glioblastoma cells, not others, show specific values for these four responses.

#### **1.1 Concept of a "radiosensitivity phenotype"**

Our studies suggest that radiosensitivity of tumor cells cannot be expressed as a single parameter or histological type, but should be analyzed on the basis of descriptors of multiple responses that are specific to human tumor cell lines. We argue that the radiosensitivity phenotype of each cell line should be defined by a set of coefficients including: 1) relative rates of inactivation over four distinct, sequentially-induced components that comprise radiosensitivity of each cell; 2) the general radiosensitivity group into which cell lines segregate non-randomly based on the values of exceptional coefficients; these groups associated with tumor cell genotype; 3) modulation of inactivation by reduced dose-rates; and 4) modulation of inactivation by the effect of ionization density of delivered radiations; and 5) modulation by in vivo mechanisms that are particular to genotype.

In the next several paragraphs we identify those coefficients that together define the "radiosensitivity phenotype" of human tumor cells. One overall goal is to equate radiosensitivity phenotypes of tumor cells to genetic or epigenetic properties.

#### **1.2 Coefficients that describe four sequentially-induced responses in each human tumor cell**

We have measured clonogenic inactivation in multiple cell lines over different dosesegments (20). **T**hese studies showed that response of tumor cells to radiation in vitro can be resolved into two general responses, termed the alpha response and omega response (or quadratic response) that represent the overall rate of inactivation over the dose-segment from 0 to circa 2 Gy (alpha response) and over doses greater than circa 3 Gy (omega response). These two general responses can be approximated by the linear and quadratic components of the linear-quadratic model. These two general responses vary not only between glioblastoma cell lines and other tumor types but also vary between different glioblastoma lines.

Our data show these two general responses are actually comprised of four more specific responses induced sequentially in each cell line. Thus one goal of this chapter is to measure coefficients that describe these four responses in tumor cells and to suggest their relative importance in clinical radiotherapy.

As stated above we have defined four sequentially-induced responses in ten tumor cell lines by increasing doses from 0.0 to 10.0 Gy. These responses are common to all tumor cells, induced at the same doses but vary in the rates of inactivation over these common dosesegments. These four responses are:

**The hypersensitive (H) response** is observed over the dose-segment from 0.0 to 0.10 Gy and is characterized by highest rates of clonogenic inactivation observed in tumor cells. This

Radiobiology of Radioresistant Glioblastoma 7

radiosensitivity Williams et al (21, 22). The values of the coefficients that describe the alpha response derived over doses from 0 to circa 2.5 Gy segregated all cell lines non-randomly into four distinct, statistically-valid radiosensitivity groups. Each radiosensitivity group is inactivated over the alpha response at different rates and the statistical variation is significant. While the alpha response, that is in reality the combined effects to the H, R and alpha\* responses, is related to the omega response in the four radiosensitivity groups, measurements of the general alpha response as will be shown subsequently, correlates with the alpha\* response as measured using the linear-quadratic model. Thus the radiosensitivity groups listed below are dependent on the values of the alpha\* response. Our work showed these four radiosensitivity groups segregate with specific genotypes, one of which was a group of some glioblastoma cells that as stated earlier express what we refer to as the "glio"

All tumor types we examined segregated into only four cellular radiosensitivity groups: A **VS (very sensitive) radiosensitivity group** was comprised of only a single hypersensitive tumor line that was mutated in the ataxia telangiectasia mutated (ATM) gene. This cell line is hypersensitive to radiation on the basis of clonogenic inactivation, expression of

An **S (sensitive) radiosensitivity group** was comprised of 17 cell lines all but one expressing wild type tumor protein 53 (wtTP53). The coefficients that describe the alpha response for these cell lines were intermediate between the VS cell line and other more resistant lines.

cell lines that expressed a mutant form of TP53 although the exact form of mutation that renders cells more resistant has not been defined. R cells are intermediate in their

A **VR (very radioresistant) radiosensitivity group** was identified that was comprised of only three human glioblastoma cells. For descriptive purposes, we will refer to the factor or gene that leads to this exceptional resistance as "glio". This is in contradistinction to the S

We also defined rates of clonogenic inactivation of glioblastoma cell lines to low dose-rates (0.25 Gy per hour) compared to high dose rate (circa 50 Gy per hour) and again find significant differences in some radioresistant glioblastoma cell lines 21, 22. Importantly, these data suggest that some glioblastoma cell lines are distinctly different in their response to low dose-rate irradiation compared to their resistance to radiation delivered at higher doses. We will show analysis of these data subsequently below, but they do demonstrate that some glioblastoma cell lines that are very resistant as measured over the alpha\* responses have a unique response to low dose-rate radiation that perhaps can be exploited in the clinic. Therefore a broad assessment of the radiosensitivity phenotype of a human

**1.5 Coefficients that represent the susceptibility of radiosensitivity to differences in** 

While we will publish data elsewhere on the effects of dose-rate and ionization density, here we can make the general statement here that these data show the H and R responses are

and R groups that contain tumor cells that derive from multiple histological types.

**1.4 Coefficients that describe radiosensitivity of tumor cells to low dose-rate** 

tumor cell should include response to protracted irradiation.

apoptosis, cell cycle progression and susceptibility to chromosomal aberrations.

An **R (resistant) radiosensitivity group** was comprised predominantly of

radiosensitivity between S and R cells.

response.

**irradiation** 

**ionization-density** 

response is related to low-dose radio-hypersensitivity as described by Joiner and his colleagues (19). In each cell line in which the H response is observed, it is expressed over the dose segment from 0.0 and 0.10 Gy with a very low threshold if any but ends at 0.10 Gy. The survival level at 0.10 Gy can be used to calculate the slope of the H response between 0.0 Gy and 0.10 Gy as α (SF.1). The H response varies strongly in different types of tumor cells with some genotypes not expressing this response. Thus each cell line can be classified as to whether the H response is induced and the slope of the rate of inactivation. Interestingly, some glioblastoma cells express the H response at a high rate but are very radioresistant at higher doses.

**The resistant (R) response** is observed over the dose-segment from 0.1 to 0.2 Gy and is characterized as increased resistance to clonogenic inactivation exhibited in each cell line that expresses the H response. The R response appears to be induced at circa 0.10 Gy in all cells and persists until it is terminated when the alpha\* response is induced. The R response is coupled strongly to the H response. In our studies shown below the R response is not always expressed as a log-linear response across the dose-segment from 0.0 to 0.1 Gy and hence the change in rates of inactivation over the R response varies strongly with genotype. The expression of the H and R responses do not correlate with the expression of the alpha\* and omega\* responses.

**The alpha\* (repair) response** is induced at 0.20 Gy in all cells and once induced extends to all higher doses. The alpha\* response is a protective response preventing excessive loss of irradiated somatic cells. The rate of inactivation over the alpha response is more resistant than rates evidenced over the H response and also more resistant than increased sensitivity observed subsequently when the omega\* response is induced. The alpha\* response is characterized by transiently suppressed apoptosis, induction of repair responses and perturbation of cell cycle progression. We refer to this induced response specifically as the alpha\* (α\*) response and it is the only response determining inactivation between 0.2 and 2.0 Gy. The slope of the alpha\* response is a correlate of the general alpha response observed between 0.0 and circa 3 Gy measured either by the slope of inactivation estimated between 0.0 and 2.0 Gy, α (SF2) or by linear component of the linear-quadratic model α (LQ). We will show examples of this relationship subsequently.

**The omega\* (triage) response** is induced at circa 3.0 Gy in all cells and extends to all higher doses. We refer to this induced response as the omega\* (ω\*) response and consider it a triage response that results is increased inactivation of damaged cells. It can be approximated mathematically by the quadratic response of the linear-quadratic model. The omega response is determined by linear regression of data above circa 3.0 Gy and its slope designated as omega (ω). The omega response is the combined effect of the omega\* response induced as circa 3 Gy and an extension of the alpha\* response. The omega\* response is characterized by increased rates of clonogenic inactivation and chromosomal aberrations but decreased rates of tumorigenesis, carcinogenesis and mutation. Thus the increased inactivation of cells at doses above 3 Gy preferentially removes cells more likely to express mutation or cancer and thus can be considered a triage process that preferentially eliminates cells by detecting radiationinduced properties and eliminating cells by post-repair apoptosis.

#### **1.3 Coefficients that describe the rates of inactivation over the alpha and omega responses segregate human tumor cells into only four statistically-valid cellular radiosensitivity groups**

In a broad survey of clonogenic inactivation in multiple tumor cells we observed that the rate of inactivation over the alpha response was the major determinant of overall cellular

response is related to low-dose radio-hypersensitivity as described by Joiner and his colleagues (19). In each cell line in which the H response is observed, it is expressed over the dose segment from 0.0 and 0.10 Gy with a very low threshold if any but ends at 0.10 Gy. The survival level at 0.10 Gy can be used to calculate the slope of the H response between 0.0 Gy and 0.10 Gy as α (SF.1). The H response varies strongly in different types of tumor cells with some genotypes not expressing this response. Thus each cell line can be classified as to whether the H response is induced and the slope of the rate of inactivation. Interestingly, some glioblastoma cells express the H response at a high rate but are very radioresistant at

**The resistant (R) response** is observed over the dose-segment from 0.1 to 0.2 Gy and is characterized as increased resistance to clonogenic inactivation exhibited in each cell line that expresses the H response. The R response appears to be induced at circa 0.10 Gy in all cells and persists until it is terminated when the alpha\* response is induced. The R response is coupled strongly to the H response. In our studies shown below the R response is not always expressed as a log-linear response across the dose-segment from 0.0 to 0.1 Gy and hence the change in rates of inactivation over the R response varies strongly with genotype. The expression of the H and R responses do not correlate with the expression of the alpha\*

**The alpha\* (repair) response** is induced at 0.20 Gy in all cells and once induced extends to all higher doses. The alpha\* response is a protective response preventing excessive loss of irradiated somatic cells. The rate of inactivation over the alpha response is more resistant than rates evidenced over the H response and also more resistant than increased sensitivity observed subsequently when the omega\* response is induced. The alpha\* response is characterized by transiently suppressed apoptosis, induction of repair responses and perturbation of cell cycle progression. We refer to this induced response specifically as the alpha\* (α\*) response and it is the only response determining inactivation between 0.2 and 2.0 Gy. The slope of the alpha\* response is a correlate of the general alpha response observed between 0.0 and circa 3 Gy measured either by the slope of inactivation estimated between 0.0 and 2.0 Gy, α (SF2) or by linear component of the linear-quadratic model α (LQ). We will

**The omega\* (triage) response** is induced at circa 3.0 Gy in all cells and extends to all higher doses. We refer to this induced response as the omega\* (ω\*) response and consider it a triage response that results is increased inactivation of damaged cells. It can be approximated mathematically by the quadratic response of the linear-quadratic model. The omega response is determined by linear regression of data above circa 3.0 Gy and its slope designated as omega (ω). The omega response is the combined effect of the omega\* response induced as circa 3 Gy and an extension of the alpha\* response. The omega\* response is characterized by increased rates of clonogenic inactivation and chromosomal aberrations but decreased rates of tumorigenesis, carcinogenesis and mutation. Thus the increased inactivation of cells at doses above 3 Gy preferentially removes cells more likely to express mutation or cancer and thus can be considered a triage process that preferentially eliminates cells by detecting radiation-

**1.3 Coefficients that describe the rates of inactivation over the alpha and omega responses segregate human tumor cells into only four statistically-valid cellular** 

In a broad survey of clonogenic inactivation in multiple tumor cells we observed that the rate of inactivation over the alpha response was the major determinant of overall cellular

higher doses.

and omega\* responses.

**radiosensitivity groups** 

show examples of this relationship subsequently.

induced properties and eliminating cells by post-repair apoptosis.

radiosensitivity Williams et al (21, 22). The values of the coefficients that describe the alpha response derived over doses from 0 to circa 2.5 Gy segregated all cell lines non-randomly into four distinct, statistically-valid radiosensitivity groups. Each radiosensitivity group is inactivated over the alpha response at different rates and the statistical variation is significant. While the alpha response, that is in reality the combined effects to the H, R and alpha\* responses, is related to the omega response in the four radiosensitivity groups, measurements of the general alpha response as will be shown subsequently, correlates with the alpha\* response as measured using the linear-quadratic model. Thus the radiosensitivity groups listed below are dependent on the values of the alpha\* response. Our work showed these four radiosensitivity groups segregate with specific genotypes, one of which was a group of some glioblastoma cells that as stated earlier express what we refer to as the "glio" response.

All tumor types we examined segregated into only four cellular radiosensitivity groups:

A **VS (very sensitive) radiosensitivity group** was comprised of only a single hypersensitive tumor line that was mutated in the ataxia telangiectasia mutated (ATM) gene. This cell line is hypersensitive to radiation on the basis of clonogenic inactivation, expression of apoptosis, cell cycle progression and susceptibility to chromosomal aberrations.

An **S (sensitive) radiosensitivity group** was comprised of 17 cell lines all but one expressing wild type tumor protein 53 (wtTP53). The coefficients that describe the alpha response for these cell lines were intermediate between the VS cell line and other more resistant lines.

An **R (resistant) radiosensitivity group** was comprised predominantly of

cell lines that expressed a mutant form of TP53 although the exact form of mutation that renders cells more resistant has not been defined. R cells are intermediate in their radiosensitivity between S and R cells.

A **VR (very radioresistant) radiosensitivity group** was identified that was comprised of only three human glioblastoma cells. For descriptive purposes, we will refer to the factor or gene that leads to this exceptional resistance as "glio". This is in contradistinction to the S and R groups that contain tumor cells that derive from multiple histological types.

#### **1.4 Coefficients that describe radiosensitivity of tumor cells to low dose-rate irradiation**

We also defined rates of clonogenic inactivation of glioblastoma cell lines to low dose-rates (0.25 Gy per hour) compared to high dose rate (circa 50 Gy per hour) and again find significant differences in some radioresistant glioblastoma cell lines 21, 22. Importantly, these data suggest that some glioblastoma cell lines are distinctly different in their response to low dose-rate irradiation compared to their resistance to radiation delivered at higher doses. We will show analysis of these data subsequently below, but they do demonstrate that some glioblastoma cell lines that are very resistant as measured over the alpha\* responses have a unique response to low dose-rate radiation that perhaps can be exploited in the clinic. Therefore a broad assessment of the radiosensitivity phenotype of a human tumor cell should include response to protracted irradiation.

#### **1.5 Coefficients that represent the susceptibility of radiosensitivity to differences in ionization-density**

While we will publish data elsewhere on the effects of dose-rate and ionization density, here we can make the general statement here that these data show the H and R responses are

Radiobiology of Radioresistant Glioblastoma 9

controls to account for possible proliferation during the period between plating and irradiation. This control consisted of plating 105 cells in separate plates when replicates of cells were plated for colony formation. When irradiation was performed on the plates for colony formation, the microcolony plates were stained and the number of cells per colony measured. The average number of cells per colony was below 1.20 cells per microcolony for

Low dose rate irradiation was carried out in a specially constructed Cs-137 irradiator with temperature control and the ability to irradiate cells with constant or exponentially-

Tumors were established by subcutaneous injection of 5 million cells suspended in PBS into the upper thigh of nude mice. Each cohort included 6 to 13 tumors. Tumor growth rate was determined by measuring three orthogonal diameters of each tumor twice a week and the tumor volume estimated as π/6 [D1 x D2 x D3], when individual tumor volumes reached ~0.1-0.3 cm3, radiation treatment was initiated. Modal specific growth delay (mSGD) was measured for all cohorts in which a majority of tumors reached a volume four times the initial volume. Response was normalized to growth of unirradiated cells. We chose not to use the mean of specific regrowth delay patterns since a significant proportion of our cohorts included one or more tumors that did not regrow. Thus the mean became limited as a regrowth parameter. For cohorts for which some tumors did not regrow we estimated mSGD based on the regrowth pattern for the minority of tumors that did regrow. When we tested the sensitivity of modal to mean growth delay in selected cohorts in which all tumors regrew, the modal value always fell within one standard deviation of the mean. These methods share some characteristics of the methods described by Schwatchofer [25]. To provide an overview of the dichotomous response when some tumors regrow but some do not, we indicated such cohorts with an arrow showing this value, in terms of overall tumor

all cell lines and did not vary significantly between cell lines.

decreasing dose-rates.

**3. Analyses** 

**2.5 Regrowth delay in xenograft tumors** 

response, was the common minimum response.

subline abrogated in p21 (19S186) .

**3.1 Clonogenic inactivation of radioresistant glioblastoma cell lines** 

vary with the dose-segment over which the data are presented.

In our previous studies (21, 22) we identified three glioblastoma cell lines (U251, T98G, U-87) that were the most resistant of 39 cell lines examined as defined by comparison of clonogenic inactivation between circa 2 Gy and 10 Gy. These three radioresistant cell lines expressed two forms of TP53, with U251 and T98 expressing mutTP53 and U87 expressing wtTP53. For designation purposes we will refer to these three cell lines as expressing a VR radiosensitivity phenotype and expressing either a glio+mutTP53 genotype (U251 and T98G) or a glio+wtTP53 genotype. In figure 1 we compare clonogenic inactivation curves for these three VR (very radioresistant) glioblastoma cell lines compared to two colorectal cancer cell lines that fall into the R (radioresistant) radiosensitivity group wtTP53 (HCT116) and its

The data in figure 1 show relative radiosensitivity between the five cell lines but it is important in our interpretation of these data to show them in the context of overall radiosensitivity of human tumor cell lines. In figure 1 there are clear differences between the three glioblastoma cell lines and the two more sensitive colorectal tumors. These differences

generally not modified by either by dose-rate or ionization density. In contradistinction, the alpha\* and omega\* responses are highly susceptible to dose-rate and ionization density.

#### **1.6 Coefficients that represent the modulation of in vivo radiosensitivity by genotype and dose**

We demonstrated variation in the response of tumor xenografts to radiotherapy protocols based on genotype and dose-schedule. In these studies, Williams et al 21, we showed genotype of tumor cells influenced both in vitro radiosensitivity of tumor cells and also, by a different mechanism, influenced xenograft response in vivo. We attributed this effect, that was substantial in some cells, as an interaction between tumor genotype and the in vivo tumor microenvironment. Importantly one glioblastoma line that was in the VR cellular radiosensitivity group, expressed surprising sensitivity when irradiated as xenograft tumors in vivo.
