1. Can we predict radiation carcinogenesis from first principles?
Rob Stewart, Ph.D.
School of Health Sciences
Purdue University
550 Stadium Mall Drive
West Lafayette, IN
April 11, 2003

2. Nature of “The Beast”
In many situations, cells and tissues are exposed to temporally and spatially complex radiation fields

3. Radiation field depends on particle energy
5,000 mm
25,000 mm
1 MeV e- in water
5 MeV e- in water

4. … the type of radiation 250 keV e- in water 250 keV e+ in water 500 mm

5. … and the initial direction(s) of flight
500,000 mm
(50 cm)
500,000 mm
(50 cm)
1 MeV photons in water
10 MeV photons in water

6. “Dose” is only the first step
Cancer develops through physical, chemical, biochemical and microevolutionary processes that happen over hours, days, months and even years.

7. The Infamous Double Strand Break (DSB)
Pose a major threat to integrity of the genome
Created by
Certain chemotherapeutic drugs (e.g., bleomycine)
“Spontaneously” as a by-productive of cellular processes
Oxidative metabolism
Replication fork encounters a single-strand break
Ionizing radiation (including of course cosmic rays, dental x-rays, radon, 40K, etc.)
Guanine (G)
Cytosine (C)
Adenine (A)
Thymine (T)
Complementary base pairs encode genetic information and provide opportunities for error-free repair.
Double strand break (DSB)

8. Are all DSBs lethal?
DSBs are distributed among identically irradiated cells ~ according to a Poisson distribution
45 Gy h-1
0.5 Gy h-1
0.12 Gy h-1
CHO cells
Some cells survive because they do not sustain critical DNA damage (i.e., a DSB).

9. Most DSBs are rejoined and non-lethal
Radiation creates ~ DSB Gy-1 cell-1.
Less than 4% of the initial DSBs are lethal.

10. Homologous recombination (HR)
How are DSBs rejoined?
Homologous recombination (HR)
Gene conversion
Single-strand annealing
Non-homologous end joining (NHEJ)

11. Homologous DNA Sister chromatid Homologous chromosome
Repetitive DNA sequences

12. Homologous Recombination (HR)
Requires extensive regions of homology
Allelic recombination
Sister chromatid
Homologous chromosome
Ectopic recombination
Other regions of genome with sequence homology
Holiday junction resolution is not random
Gene conversion without cross over more frequent than gene conversion with cross over.
HR has the potential to rejoin DSBs with no loss in genetic information (error-free repair)
Adapted from M. van den Bosch, P.H.M Lohman, and A. Pastink, DNA Double-Strand Break Repair by Homologous Recombination, Biol. Chem. 383, (2002)

13. Non-Homologous End Joining (NHEJ)
DSB is recognized by DNA protein kinase (DNA-PK)
KU80/KU70 heterodimer
Catalytic sub-unit DNA-PKcs
Ligase IV and XRCC4 co-factor promote ligation of the DNA break ends
NHEJ is an error prone DSB restitution pathway
Ku70 (G22P1) -- DNA end binding, Ku80 (XRCC5), XRCC4 Ligase accessory factor, LIG4 Ligase IV.
Double-strand DNA breaks (DSBs) pose a major threat to living cells, and several mechanisms for repairing these lesions have evolved. Eukaryotes can process DSBs by homologous recombination (HR) or nonhomologous end joining (NHEJ). NHEJ connects DNA ends irrespective of their sequence, and it predominates in mitotic cells, particularly during G1 (Takata et al., 1998). HR requires interaction of the broken DNA molecule with an intact homologous copy, and allows restoration of the original DNA sequence. HR is active during G2 of the mitotic cycle and predominates during meiosis, when the cell creates DSBs, which must be repaired by HR to ensure proper chromosome segregation. How the cell controls the choice between the 2 repair pathways was investigated by Goedecke et al. (1999). They demonstrated a physical interaction between the mammalian Ku70, which is essential for NHEJ (Baumann and West, 1998), and MRE11 (600814), which functions both in NHEJ and meiotic HR. Moreover, they showed that irradiated cells deficient for Ku70 are incapable of targeting Mre11 to subnuclear foci that may represent DNA-repair complexes. Nevertheless, Ku70 and Mre11 were differentially expressed during meiosis. In the mouse testis, Mre11 and Ku70 colocalized in nuclei of somatic cells and in the XY bivalent. In early meiotic prophase, however, when meiotic recombination is most probably initiated, Mre11 was abundant, whereas Ku70 was not detectable. Goedecke et al. (1999) proposed that Ku70 acts as a switch between the 2 DSB repair pathways. When present, Ku70 destines DSBs for NHEJ by binding to DNA ends and attracting other factors for NHEJ, including Mre11; when absent, it allows participation of DNA ends and Mre11 in the meiotic HR pathway.
Adapted from F. Daboussi, A. Dumay, F. Delacote, and B.S. Lopez, DNA double-strand break repair signaling: The case of RAD51 post-translational regulation, Cellular Signaling 14, (2002)

14. DSB repair in mammalian cells
HR is potentially error-free. But inappropriate HR can lead to large DNA rearrangements (chromosome aberrations).
Impaired or increased HR has been associated with a predisposition towards cancer
NHEJ is highly mutagenic but consequences are usually less severe.
NHEJ predominates in G0 (quiescent cells) and in G1/early S phase cells. HR is important in late S/G2 phase.
DSB repair is not the same in quiescent and actively dividing cells.
DSB repair is a function of cell cycle phase.
HR and NHEJ are regulated through a complex set of signaling pathways.
Overall rate and fidelity of DSB repair can be disrupted in many different ways.

15. Local damage complexity
DNA organized into a chromatin fiber x
Simple double strand break
Track x
Complex double strand break
Track

16. DSB repair may be affected by damage complexity
Not difficult to imagine that collateral damage near the site of two opposing strand breaks could impair
Resection of damaged break ends
DNA synthesis
Branch migration and Holiday junction formation
Disruptions in HR would likely depend on the spatial configuration and types of nearby damage sites.

17. Pairwise damage interaction

18. Regional Multiply Damaged Sites
Proximity effects
One radiation track can create multiple DSB.
Some DSBs may be in close spatial proximity.
Break ends in close temporal and spatial proximity are more likely to interact than ones separated in time or space.
Frequency of pairwise damage interaction increases with increasing particle LET.
Regional Multiply Damaged Sites x x x
Track

19. Repair-misrepair (RMR) model
DSBs are created and rejoined
Gy h-1 at time t
pairwise damage interaction
Repair processes convert fraction (1-a) of the initial DSBs to lethal or non-lethal mutations
Non-lethal
Lethal
C.A. Tobias, The repair-misrepair model in radiobiology: comparison to other models. Radiat. Res. Suppl. 8:S77-S95 (1985).

20. Surviving fraction
S(t) is the fraction of cells free of lethal damage at time t. Lethal damage created during a short time interval dt, whose average is dF/dt, are randomly distributed among cells without regard for which cells already have lethal damage.
Surviving fraction at time t
For a review, see R.K. Sachs, P. Hahnfeld, and D.J. Brenner, Review: The link between low-LET dose-response relations and the underlying kinetics of damage production/repair/misrepair. Int. J. Radiat. Biol. 72(4) (1997).

21. Virtual Cell (VC) Software
45 Gy h-1
0.5 Gy h-1
0.12 Gy h-1
CHO cells
Simulates the repair and misrepair of DNA damage
LPL model (Curtis 1986)
RMR model (Tobias (1985)
TLK model (Stewart 2001)
Predicts endpoints such as
Expected number of DSB as a function of time
Fraction of cells that survive irradiation
Fraction of cells that acquire genetic instability and become unstable (transformed)
Tumor control probability after radiation therapy
Expected time of tumor reoccurrence after radiation therapy

22. Split-dose Experiment
Time (h)
Dose rate (Gy h-1)
4 hour
S = 0.12
RMR parameters for CHO cells

23. External beam radiation therapy
S  (SF2)30
= 7.27 × 10-3
S = 5.4×10-3
SF2 = 0.849
A 60 Gy radiation treatment (2 Gy × 30) delivered over 6 weeks (M-F skipping weekends). The 2 Gy daily doses are delivered at 6 Gy h-1 (= 2 Gy/20 minutes).
RMR parameters for CHO cells

24. Brachytherapy
S = 3.9×10-4
A 125I seed that delivers 150 Gy in 1.1 years. Dose rate decreases exponentially with a half-life of 1,443 h (peak dose rate = 72.4 mGy h-1).
RMR parameters for CHO cells

25. Combined radiation treatments
Scombined = 7.4×10-6
Sebrt(60 Gy) = 5.4×10-3
Sbrachy(100 Gy) = 8.9×10-3
Sebrt × Sbrachy= 4.8×10-5
Hypothetical combined external beam and brachytherapy radiation treatment (160 Gy total delivered dose).
RMR parameters for CHO cells

26. RMR and LQ survival models are related
The widely used linear-quadratic (LQ) survival model may be written as
Equating S(D) and S() gives
See M. Guerrero, R.D. Stewart, J. Wang, and X.A. Li. Phys. Med. Biol. 47, 3197–3209 (2002) and RK Sachs, P. Hahnfeld, DJ Brenner. Int. J. Radiat. Biol. 72(4), (1997).

27. A mechanistic interpretation of the LQ
Accuracy of repair process
Rate of DSB rejoining {
Pairwise damage interaction process always
creates a mutation (chromosome aberration).
But not all of them are lethal.
Expect a/b ratio to increase as
rate of DSB rejoining (l) increases.

28. Prediction of LQ parameters from “first principles” – a tantalizing possibility
Small black filled symbols generated using Monte Carlo sampling methods
Large red symbols parameter values obtained from the direct analysis of measured survival data

29. Lack of a dose rate effect is insufficient evidence to infer ‘no repair’
dose rate effects
Lack of a dose rate effect implies
A small fraction of the initial damage is unrepairable (< 2%)
complex DSBs?
Rapid DSB rejoining
G → 0 (large )
(/) << 1
Rapid DSB fixation competes with 1st and 2nd order repair.
Other possibilities
90% of small-cell lung cancer (SCLC) cases are found to have an inactivated retinoblastoma (Rb) gene. Cells do not have a functional G1/S cell cycle block. In vitro experiments with SCLC (NCI-H69) performed using exponentially growing cellsDamaged G1 phase cells proceed into S phase.

30. Damage formation and repair is still only the beginning
Cancer develops through physical, chemical, biochemical and microevolutionary processes that happen over hours, days, months and even years.

31. Tumor growth kinetics Exponential cell kinetics are sometimes observed
Cell birth rate (h-1)
Cell death rate (h-1)
Doubling time

32. Radiation therapy for the treatment of prostate cancer
Prostate tumor composed of 107 tumor cells.
Wang et al. (2003) radiosensitivity parameters
JZ Wang, M. Guerrero, XA Li. How low is the alpha/beta ratio for prostate cancer? Int. J. Radiat. Oncol. Biol. Phys. 55(1), (2003).

33. Tumor control probability (TCP)
Dose in parentheses is the treatment dose that gives a TCP of 90%
Prostate tumor composed of 107 tumor cells.
Wang et al. (2003) radiosensitivity parameters
JZ Wang, M. Guerrero, XA Li. How low is the alpha/beta ratio for prostate cancer? Int. J. Radiat. Oncol. Biol. Phys. 55(1), (2003).

34. Multi-stage cancer model(s)
Through a series of mutational events, stem cells acquire minor and enhanced genetic instability and other traits.
Tumor forms through the clonal expansion of the unstable cell population.
Cell birth/death processes may change as cells progress towards malignancy.

35. Incidence of lung cancer
At background radiation levels (75 to 225 mGy), endogenous processes may account for 70 to 90% of lung cancers.
At 1 Gy, endogenous processes may account for as much as 30% of lung cancers.
Estimated lung cancer incidence with and without DNA damage caused by endogenous processes.
H. Schöllnberger, R.D. Stewart, R.E.J. Mitchel, and W. Hofmann, An examination of radiation hormesis mechanisms using a multi-stage carcinogenesis model. In progress. Abstract submitted to ICRR 2003 Brisbane, Australia (2003).

36. Induction of cellular defense mechanisms
A 3-fold low dose (rate) enhancement in DNA repair and radical scavenging would provide support for an effective threshold.
Estimated lung cancer incidence with and without low dose (rate) adaptations in radical scavenging and DNA repair.
H. Schöllnberger, R.D. Stewart, R.E.J. Mitchel, and W. Hofmann, An examination of radiation hormesis mechanisms using a multi-stage carcinogenesis model. In progress. Abstract submitted to ICRR 2003 Brisbane, Australia (2003).

37. Comments
Multi-stage models that use exponential cell growth kinetics are extremely sensitive to the selection of a net cell birth rate (a-b) and have conceptual difficulties
For lung cancer, (a-b) ~
Cell density < ~ 108 to 109 cells cm-3.
Tumor size has a finite upper bound ~ 1014 or 1015 cells.
Over extended periods of time (months or years), age, health status, etc., etc., will impact on cancer development
Cell birth/death parameters will change over time and most likely has a stochastic (chaotic) element.
Wounds or disease may temporarily alter the tissue microenvironment and accelerate the clonal expansion of aberrant cells.
Normal cells affect behavior of transformed cells and vice versa
Cell signaling (bystander) effects.

38. Cell signaling
Cells in higher animals coordinate cellular activities using hundreds of different kinds of signaling molecules
Proteins, small peptides, amino acids, nucleotides, steroids, retinoids, fatty acid derivatives, and gases such as nitric oxide (NO) and carbon monoxide
Signaling can be long range (synaptic and endocrine signaling) or short range (autocrine and paracrine signaling)
Goodbye friends. I’ve caught a virus and must leave you.
Maybe I’ve got the virus and just don’t know it yet
Direct cell-to-cell communication through gap junctions
Autocrine and paracrine signaling through excreted messengers

39. Why do we care?
Medium transfer and single-cell irradiator experiments demonstrate that radiation-damaged cells emit signals that cause radiation-like changes in nearby undamaged cells
changes in gene expression, mutations, increases in sister chromatid exchanges, induction of chromosomal instability, and cell transformation
Cell birth, differentiation and death processes are highly regulated through multiple signaling networks.

40. Growth Inhibition
Cell growth in vivo is limited (at a minimum) by the availability of space, nutrients, and growth factors
Cells compete with each other for resources
Life support capacity
Radiation killing

41. Microevolution of a tumor
Normal and transformed cells vie for resources
Normal cells
Tumor cells
Crowding effects:
Over-expression of growth factor receptors

42. Virtual Tissue Model (VTM)
System of 6 or 7 differential equations describe the in vivo cell system (i.e., the Virtual Tissue).
Normal cells and transformed cells vie for resources.
Tissue microenvironment is re-shaped as the relative number of normal and transformed cells changes.

43. Can we simulate cancer from first principles?
Yes!
No!
Maybe…
Computational issues
Mechanisms and
parameters
What’s a first principle?
If we could construct a “quantum mechanical supermodel” it would be as complicated as the experimental system
very computation and memory intensive
numerical stability of solutions (~ 1E26 atoms, 1E18 s time scale)
Initial state of the system?
At what point in time does the simulation begin? At birth? At conception?
Terms such as “first principles” mechanistic, ab initio are somewhat cliché and overused.

44. A lot more can be done http://healthsciences.purdue.edu/vc/
Acknowledgement: The Virtual Cell software development effort is supported in part by the U.S. Department of Energy's Low Dose Radiation Research Program through the Office of Science (BER), Grant Number DE-FG02-03ER63541.