Skip to main content

Multidisciplinary evaluation of the cancer risk from neutrons relative to photons using stem cells and the induction of second malignant neoplasms following paediatric radiation therapy

Final Report Summary - ANDANTE (Multidisciplinary evaluation of the cancer risk from neutrons relative to photons using stem cells and the induction of second malignant neoplasms following paediatric radiation therapy)

Executive Summary:
The ANDANTE project was funded under Euratom FP7 and ran for 4 years from 2012 to 2014. The main purpose of the project was to investigate the biological effectiveness of neutrons to cause cancer in humans for comparison with the weighting factors usually applied to evaluate risk from exposure to neutrons for radiation protection purposes. The project plan incorporated independent studies in the disciplines of physics, radiobiology and epidemiology, with comparison and combination of the results to give an end result.
In the physics component, Monte Carlo simulations and measurements were used to characterise the neutron beams to be used in radiobiological experiments: broad spectrum, narrow quasi-monoenergetic spectra, and scattered from a water phantom by 190 MeV protons to simulate scattered neutrons produced during proton radiotherapy. The same techniques also provided data for the development of an analytical method for calculating scattered neutron doses received during proton therapy for later use in the epidemiological study. The characteristics of secondary particles generated by neutrons was used as input to a Monte Carlo track structure model used to simulate the initial damage to DNA molecules and score the statistics of damage clusters in terms of neutron energy. The results showed that DNA damage caused by neutrons, when compared to photons, reaches a maximum at a neutron energy of around 1 MeV, in general agreement with the conventional radiation protection weighting factors. To our knowledge, this activity has led to the establishment of the first neutron RBE model for DNA damage induction, purely based on first principle calculations.
Stem cells were isolated from thyroid, salivary, and mammary gland tissue, and given exposures to the neutron beams, over a range of doses up to 2 grays, as well as to 220kV x-rays in order to determine the RBE. The irradiated cells were assayed in vitro for a number of indicators of damage and possible carcinogenesis. The results showed a clear dose-response relationship for clonogenic cell survival and γH2AX assays, but equivocal results for other markers. The results were combined with the track-structure studies to derive a functional relationship of RBE with neutron energy. Irradiated cells were also transplanted into mice to investigate the formation of radiation-induced tumours but it was not possible to demonstrate a carcinogenic response during the project.
The neutron dose calculation method and the RBE values produced by the project were combined to give a method for estimating the risk from clinical data of a second cancer being caused to an individual patient by the exposure to scattered neutrons during proton therapy. The method was tested using sample anonymised patient data in preparation for a prospective epidemiological study to compare second cancer rates following paediatric proton therapy in comparison with conventional radiotherapy in order to verify the RBE values generated by the project. Initial plans were developed for an international multi-centre database of radiotherapy treatments and outcomes. Expressions of interest were positive, and possible funding streams for the study were investigated. A power calculation on cohort size and timescale indicated that it is likely to take several decades of epidemiological follow-up time to demonstrate an effect of the neutrons on the secondary cancer rates. However the setting up of an international registry of childhood cancers has the potential to give insight into the link between radiation exposure to proton (and photon) therapy and the subsequent risk of second cancers linked to such treatments.

Project Context and Objectives:
The context of the ANDANTE project

Neutrons are an unwanted by-product of the use of radiation in many essential industries. In the nuclear power industry and in the nuclear fuels processing and transportation industries, shielding from harmful neutron exposure is an overhead cost of both fission and fusion power. Also, there is a significant neutron exposure of radiotherapy patients undergoing treatment with high energy photons (greater than 10 MV) from medical linear accelerators or with charged hadrons (protons, carbon ions), as well as a potential occupational and public exposure in the vicinity of the therapy machines. A third relevant exposure path is from the neutrons generated in the upper atmosphere by cosmic rays; these contribute to the natural background radiation at high altitudes, in particular during air travel. It is estimated that about half of the risk from radiation exposure caused by cosmic rays at typical long-haul flight altitudes come from neutrons (UNSCEAR 2006).

The hazards to human health from neutrons of all energies are well acknowledged. The ability to quantify the risks from exposure to neutrons has important implications for radiation protection, including but not limited to:
- design of shielding in nuclear power plants and industries that generate neutrons;
- allocation of resources and work scheduling for radiation protection of exposed workers;
- producing better-informed guidelines for air crew exposure (particularly the problem of pregnant air crew members);
- estimation of the risks of second malignant neoplasms (SMN) following proton therapy, particularly to paediatric patients, who have a greater risk of radiation-induced cancers and longer life expectancy than adults.

At present the harmful effects on humans from neutrons are still not well understood. Because the secondary charged particles generated by interactions with energetic neutrons have a higher linear energy transfer (LET) than photons or electrons, the micro- or nano-scopic clusters of ionisation produced within tissue, and particularly within cells, can instigate quite different damage processes, leading to different effects and risks per absorbed dose.

In the current radiation protection system (ICRP 103), a simplifying assumption is made that the relative effectiveness of each radiation type, or quality, can be represented by a specified radiation weighting factor (wR), which is used to convert the physical absorbed dose in a tissue into the “equivalent dose”, or “effective dose”, taking into account the effectiveness of different ionizing radiations at low dose. The same radiation weighting factors are, for simplicity, used irrespective of damage endpoint, tissue type, dose rate, mode, and heterogeneity of exposure. The value of wR for neutrons is given as a function of energy (see figure below). It varies over nearly an order of magnitude, with a sharp peak at around 1MeV. The evaluation of wR in a particular case is always complex because neutron fields are never monoenergetic.

Figure: Radiation weighting factor wR as a function of neutron energy, reproduced from ICRP 103.

The ICRP concept of the radiation weighting factor is developed from the concept of “relative biological effectiveness” (RBE). Values of the RBE for neutrons of different energies have largely been derived from in vitro cell-killing, from chromosome aberration studies, and from animal experiments (ICRP 92). Brenner and Hall (2008) quote RBE values based on carcinogenesis studies with fission neutron spectra in mice in the range of 6 – 59. There is clearly variability and also substantial uncertainty.

Cell and animal studies do provide estimates of the value of RBE, but this approach begs the question of whether the results are applicable to humans quantitatively, or only qualitatively, or indeed not at all. This question can only be answered by epidemiological investigations on human populations, and these are very difficult to find. Initial estimates of the risks from neutrons were made from the Japanese atomic bomb data, but these are somewhat uncertain because of critical dependence on assumptions about the nature of the detonations, atmospheric conditions, geography, location of each individual, etc. In a recent reassessment (DS86 and DS02) of the radiation dosimetry associated with the atomic bombings, the earlier estimated doses of neutrons in both cities were reduced, particularly in Hiroshima, where the new values are only about 10% of the previously estimated level. The estimated neutron doses are now so small (only 1–2% of the total dose in Hiroshima and less in Nagasaki) that direct estimates of the risk for cancer associated with exposure to neutrons from the atomic bomb survivors are no longer reliable (IARC 2000) Two independent groups have estimated the most likely RBE at low doses for neutron-induced carcinogenesis in humans, respectively, as 100 [95% CI: 25–400] for solid-cancer mortality (Kellerer et al 2006), and as 63 [95% CI: 0–275] for overall cancer incidence (Little 1997).

The other significant cohorts exposed to neutrons are either occupationally exposed individuals working in the nuclear power/fuels industry, members of flight crew, or radiotherapy patients. Only in the case of radiotherapy patients is there the possibility of accurate dosimetry with a full knowledge of the dose distribution, and neutron energy spectra. The ANDANTE project is the first study to exploit this valuable cohort to develop biology-based quantitative models of the difference in risks of carcinogenesis between low- and high-LET radiation.

Objectives of the project

The overall objective of the ANDANTE project was to integrate the disciplines of radiation physics, molecular biology, systems biology modelling, and epidemiology in order to investigate the relative risk of induction of cancer from exposure to neutrons compared to photons (the RBE).

The project focused on three specific cancers that may be detected as second malignant neoplasms following paediatric photon radiotherapy: salivary gland (Boukheris et al 2008), thyroid gland (Bhatti et al 2010), and breast tissue (Travis et al 2005). Stem cells from each of the types of tissue were to be exposed to well-characterised beams of neutrons and reference x-rays, and biological markers of possible tumorigenesis were used to develop RBE models.

In the parallel physics approach, measurements and Monte Carlo simulations would identify the spectra of particles interacting with the cells. This information would be used to by the Monte Carlo track-structure model PARTRAC (Friedland et al. 2008) to simulate the initial damage to DNA, in order to generate independent models of the potential biological harm produced by high-LET particles compared to low-LET particles. The combination of these and the radiobiology results would lead to a hypothetical RBE model, specific to the tissues studied, that could be used to estimate the risk of SNMs following proton therapy.

The third parallel approach called upon epidemiology to first characterise the SNM risk following conventional external beam radiotherapy (caused by low-LET scattered photons) as a function of radiation dose. Then, the RBE model generated by the project would be used to produce a method for estimation of the risk of SNM to the thyroid, salivary gland, or breast, as a function of the dose and energy of scattered neutrons during proton therapy. The project would set up all the tools and planning for an epidemiological study to test the risk predictions based on the RBE model. A method would be devised and validated for reconstruction of the neutron doses from patient clinical data, and hence estimation of the SNM risk to the individual patient. This would be tested on real patient data to establish that a future study would be feasible. A retrospective study during the project was never an option because the tools would only be ready at the end, and there are not yet sufficient clinical data available to achieve statistically significant results. The proposed prospective study would use data from paediatric treatments because of the potentially long follow-up time.

Finally, in order to achieve maximum impact from the project results, planning would be initiated to set up a multi-national multi-centre database of proton therapy data that would include all the information needed to reconstruct scattered neutron exposures to the relevant tissues, and to start off the process of accumulating the clinical and follow-up data. After sufficient time and data have accumulated, it will be possible to re-evaluate the SNM risk from proton therapy and RBE as a risk weighting factor for use in radiation protection.

Bhatti P, Veiga LH, Ronckers CM, Sigurdson AJ, Stovall M, Smith SA, et al. Risk of second primary thyroid cancer after radiotherapy for a childhood cancer in a large cohort study: an update from the childhood cancer survivor study. Radiat Res. 2010,174:741-52.

Boukheris H, Ron E, Dores GM, Stovall M, Smith SA, Curtis RE. Risk of radiation-related salivary gland carcinomas among survivors of Hodgkin lymphoma: a population-based analysis. Cancer. 2008, 113:3153-9.

Brenner, D. J., and E. J. Hall, 2008), Secondary neutrons in clinical proton radiotherapy: A charged issue, Radiotherapy and Oncology 86 (2008) 165–170.

Friedland W. Paretzke H. G., Ballarini F., Ottolenghi A., Kreth G. Cremer C., First steps towards systems radiation biology studies concerned with DNA and chromosome structure within living cells, Radiat Environ Biophys (2008) 47:49–61

Kellerer A M, Ruhm W and Walsh L 2006 Indications of the neutron effect contribution in the solid cancer data of the A-bomb survivors Health Phys. 90 554–64

Little MP. Estimates of neutron relative biological effectiveness derived from the Japanese atomic bomb survivors. Int J Radiat Biol 1997;72:715–26.

IARC 2000: International Agency for Research on Cancer Monographs on the Evaluation of Carcinogenic Risks to Humans, Vol. 75, Part 1, IARC, Lyon.

ICRP Publication 92: Relative Biological Effectiveness (RBE), QualityFactor (Q), and Radiation Weighting Factor (wR), Annals of the ICRP, 33 (4), 2003.

ICRP Publication 103, , Recommendations of the International Commission on Radiological Protection ICRP Ann. ICRP 37 (2-4), 2007.

Travis LB, Hill D, Dores GM, Gospodarowicz M, van Leeuwen FE, Holowaty E, et al. Cumulative absolute breast cancer risk for young women treated for Hodgkin lymphoma. J Natl Cancer Inst. 2005, 97:1428-37.

UNSCEAR 2006: United Nations Scientific Committee on the Effects of Atomic Radiation, Report to the General Assembly with Scientific Annexes, United Nations, New York.

Project Results:
The organisation of this part of the report follows the logical structure of the project:

- The first part details the modelling and measurements undertaken to characterise the radiation exposures that would be used on the biological samples, and would later be received by proton therapy patients.

- The second part takes the detailed neutron-induced charged-particle fields from the modelling in the previous section to simulate the corresponding damage to DNA and formulate an RBE based on the differences in damage caused by neutrons and photons;

- The third details the isolation of stem cells from salivary gland, thyroid gland, and breast tissue, and the results of irradiating samples with neutrons and x-rays to measure RBE values;

- The fourth takes the results of all the previous sections to construct the tools needed for an epidemiological test of the values of RBE;

- The fifth presents the development of plans to set up a prospective epidemiological study, and the data base that will be needed.

1. Characterisation of neutron and x-ray exposures

1.1 Neutron facilities used for cell irradiation

In order to investigate the dependence of the RBE of neutrons relative to photons on neutron energy, the project required a range of different neutron beams as well as a reference source of photons.

All the photon irradiations were done using 220 kV x-rays as the reference low LET radiation. Both UMCG and UROS have an XStrahl200 machine for this purpose.

The neutron irradiations were carried out at:

- PTB (Physikalisch-Technische Bundesanstalt, Braunschweig, Germany), using a broad beam (mean energy 5MeV and dose-rate in the range of 3 mGy/min to 100 mGy/min) and quasi mono-energetic beams (energy 0.565 MeV and 1.2 Mev and dose-rate 3-5 mGy/min.)

- KVI-CART - Center for Advanced Radiation Technology, Groningen, Netherlands, using a field of neutrons scattered from 190 MeV protons in a water phantom, generating a neutron spectrum similar to that produced during proton therapy.

The experimental beams and sample exposures were all simulated by CUT (the work moved to TUW in the last year of the project) using the Monte Carlo radiation-transport code PHITS. Simulated neutron sources (starting from the interactions of the source proton beam with water in the KVI case) were first verified to be suitable for the project, both in terms of energy and spatial uniformity.

Following test simulations with PHITS, a sample exposure arrangement was designed using a PMMA ring holder containing 7 separate cylindrical PMMA containers 3 mm thick and 2 cm in diameter. The PMMA holder was placed on a wooden support allowing the holder to be rotated slowly to keep the contained cells uniformly mixed. The set-up is shown, together with the broad-beam geometry at PTB, in Figure 1.1. This arrangement was used for the broad beam at PTB and the scattered field at KVI. For the narrow spectrum beams at PTB, because of the small field-size the cylindrical containers were exposed individually. Given the low dose rate, three containers were exposed in a row, which delivered at the same time different dose points following a dose distance relationship verified with simulations. Containers were always kept in rotation thanks to a dedicated support. This setup is also shown in Figure 1.1.

Figure 1.1 PHITS geometry for PTB irradiations with: broad beam, upper panel; mono energetic beams, lower panel. Pictures of the setups and of the PMMA ring holder with 7 containers are also shown.

A sample of the results achieved for the characterisation of secondary charged particles from the PTB broad beam is given in Figure 1.2. Data on the secondary field obtained with simulations were used as input to the track structure modelling described in Section 2 below.

Figure 1.2 Simulation results (PHITS) for the fluence (upper panel) and relative doses (lower panel) inside the containers placed at three different distances from the neutron source of
1.2 MeV.

1.2 Characterisation of the neutron scatter fields from proton therapy beams

Measurements and modelling with Monte Carlo (MC) simulations of neutron fluence and energy spectra outside primary therapeutic proton beams were an important component of the ANDANTE project. In addition to establishing a measurement system with liquid and sold organic scintillators, an investigation of neutron and secondary particle track structures based on single-ion detection was performed. These efforts contributed to characterization and reconstruction of 3D neutron radiation fields of passively scattered and actively scanned proton beams generated by representative clinical proton fields, namely those for which retrospective paediatric proton data would be available for testing the tools developed for the pilot epidemiological study.

Initial simulations and measurements of neutron fields were performed at the research beamline at Loma Linda University Medical Centre (LLUMC) in order to test and optimize the instruments and detectors used. We were also successful in building a detailed MC model of Gantry 1 at PSI. In addition, we obtained measured data for Gantry 2, the pencil-beam scanning (PBS) beam line at PSI where most children will be treated in the future. Eventually other proton treatment centres in USA were contacted in 2014 and additional measurements were performed at the Northwestern Medical Chicago Proton Centre (NMCPC) and the MD Anderson Cancer Centre (MDACC). We successfully incorporated details of the NMCPC beam line into the TOPAS (TOol for PArticle Simulation) platform, a Geant4-based MC simulation tool specifically developed for proton therapy treatment planning simulations

Extensive MC simulations were performed using different tissue-equivalent materials with the MC simulation tool Geant4 and TOPAS. There were relatively small differences in neutron dose and energy spectra for the investigated materials. In particular, soft tissue and bone materials provided by the company CIRS were representative materials for standard tissues with respect to out-of-field neutron dosimetry in proton radiation therapy and were selected for the measurements on clinical beamlines using liquid scintillator technology. The comparisons between simulated and measured neutron doses and spectra indicated that MC simulation reproduces the scattered neutron fields with sufficient accuracy for patient dosimetry purposes. An example of MC simulation of the scattered neutrons generated in a proton treatment using a stylised phantom is given in Figure 1.3.

LLU and PSI performed additional research in the development and optimization of a single-ion detector for track structure characterization. Neutrons create low-energy recoil ions, in particular protons when interacting with hydrogen in tissues, that are known to cause large clusters of ionization on the DNA scale. It is expected that this development will further enhance the development of neutron dose related risk models.

Figure 1.3: MC simulation geometry of a 3-field proton treatment plan on of a prostate within the stylized phantom. Primary protons are shown in light blue, while neutrons are shown in yellow.

2. Track structure modelling of initial damage to DNA

The results of calculations performed with the transport code PHITS, as described above in Section 1.1 have provided the characterization of the mixed field of secondary charged particles generated by neutron interactions in the experimental setups used for irradiations. The aim of this part of the project was to use the simulated information as an input to track structure modelling with the biophysical code PARTRAC, in which simulation of particle tracks is coupled to a realistic representation of the cellular target, and in particular of nuclear DNA. DNA damage is obviously not exhaustive of cellular/late effects, but it still remains a simple and powerful tool to measure radiation clustering properties and to obtain a quantitative evaluation of biological effectiveness. The objective was therefore to obtain the patterns of damage to DNA associated with different neutron fields in experimental setups, with the aim of further correlating such theoretical results with experimental endpoints.

The work strategy for this task was to first establish a general approach to calculation of neutron-induced DNA damage, coupling neutron transport and charged-particle track structure simulations. This was achieved and presented in two different publications (Baiocco et al. 2015 and 2016). The general coupling approach is based on the condensation of the wealth of information which can be obtained with transport calculations when characterizing the neutron-induced secondary particle field, into two representative quantities, namely: the relative dose contribution of each secondary species to the total neutron dose; and a dose-average indicator of their linear energy transfer in the cellular target. Track structure calculations deliver the evolution of a chosen DNA damage endpoint with linear energy transfer of the charged species. The damage associated with a given species having a given linear energy transfer can then be weighted by the contribution of such species to the total neutron dose. When neutron dose-weighted damage is summed up for all secondary particles in the field, the overall neutron-induced damage of the same type is obtained. Using a reference X-ray field it is possible to evaluate neutron relative effectiveness in terms of damage induction, i.e. to calculate a neutron RBE for the specific endpoint under investigation.

We have found in particular that neutron RBE for the induction of DNA-DSB clusters (where a cluster is defined as a DNA lesion containing 2 or more DSBs on a genomic length of 25 base pairs) show an evolution with neutron energy over a large range of values, which is perfectly consistent with what is known from current standard of neutron radioprotection factors, which are in turn agreed upon on the basis of the latest knowledge on experimental RBEs. The main source for the evaluation of neutron radiation weighting factors is experimental RBE data from in vitro cell-killing, chromosome aberration studies and animal experiments. Neutron effectiveness is found to be maximal at energies of about 1 MeV, then decreasing before coming to a second peak at around 20 MeV. In particular radioprotection factors are based on maximal RBE values extrapolated at low doses, and absolute values for the DSB cluster RBE are generally intermediate between the international (ICRP) and the USA (US.NRC) standards, if not in quantitative agreement. This comparison has served as a test of the good quality of the coupling approach in terms of its capability to deliver information on neutron radiobiological effectiveness. This is shown in Figure 2.1. To our knowledge, this activity has led to the establishment of the first neutron RBE model for DNA damage induction, purely based on first principle calculations.

When the same approach is applied to experimental setups, a setup-specific neutron RBE for DSB cluster induction can be obtained. Such values are found to be consistent with the ICRP standard if the average energy of neutrons in the experimental beams is considered. Correlations between predicted RBEs for DSB cluster induction and experimental RBE measured on irradiated cells are described below.

Figure 2.1: Trend of calculated RBE for DNA cluster damage as a function of neutron energy, compared with ICRP 103 and US NRC wr ’s (details in Baiocco et al, to be published).

3. Radiobiological experiments with stem cells

3.1 Introduction

The design of the radiobiological experiments, which aimed at determining the RBE of neutron exposures to stem cells of critical organs relevant to various long-term effects potentially related to radiation carcinogenesis, was based on the protocol developed in Toronto by the team of Clifton in the 1970s (Clifton et al 1978, Gould et al 1977). This in turn is a modification of the bone marrow / spleen colony stem cell assay by Till and McCulloch (1961). In short, stem cells were isolated from cell suspensions freshly prepared from organ tissue samples making use of their ability to grow as spheroids under specific culture conditions. After establishing them in culture by a few passages in special stem cell media, stem cells were irradiated with the following doses: 0, 0.1Gy 0.25Gy 0.5 Gy, 1 Gy, and 2 Gy, using either 220 kV X-rays, broad beam neutrons, quasi-monoenergetic neutrons or secondary neutrons from a primary proton beam, with a neutron spectrum similar to that generated in a therapeutic proton facility. (See Section 1.1.) Irradiated stem cells were propagated in vitro for many passages, (e.g. thyroid stem cells for 15 passages) or they were transplanted into suitable sites of isogenic or SCID mice to investigate radiation-induced changes after many months of in vivo growth.

For all organ-specific stem cells, the same radiation doses, radiation energies, dose rates and late radiation response criteria were used.
In vitro the experimental endpoints were:
- Spheroid formation ability or colony forming ability as functional test of stemness;
- Induced and un-repaired double-strand breaks using the γH2AX foci assay;
- Ability to differentiate and form tissue/organ-specific structures;
- Expression of stem cell markers or organ-specific cancer markers;
- Expression of 84 genes (mRNA) related to 9 different metabolic pathways which are potentially related to carcinogenesis.

In vivo the experimental endpoints were:
- Formation of organ-specific histological structures from transplanted stem cells;
- Organ-specific functional activity related to the tissue of origin of the grafted stem cells;
- Histopathological evidence of metaplasia or malignancy.
Two partner laboratories worked closely together using identical radiation doses and neutron beams, and the same experimental endpoints and assays, but investigating stem cells isolated from different organs in the respective laboratories by the partner researchers, namely thyroid and salivary glands in Groningen (UMCG) and human mammary glands in Rostock (UROS).

3.2. Carcinogenesis of salivary gland and thyroid gland stem cells

3.2.1 Isolation of salivary gland stem cells

Submandibular glands were dissected from mice. Tissue was mechanically and enzymatically digested and cells were put in culture medium. After 3-5 days spheroids (multicellular 3D structures) containing stem cells were obtained in non-adherent cultivation of cell suspensions of salivary gland tissue in serum-free conditions. These spheroids were digested into single cells and passaged to expand the number of cells. At the end of passage 1 irradiation experiments were performed.

3.2.2 Isolation of thyroid gland stem cells

In addition we developed a stem cell assay for the thyroid gland. To isolate cells from the thyroid gland tissue we tested several protocols. The best isolation method was mechanically and enzymatically digesting the tissue and incubating the cells in tissue specific culture medium.

A qPCR for thyroid-specific markers thyroglobulin, transcription factor NK2 homeobox 1 (NKX2-1), transcription factor Paired box gene 8 (pax8), thyroid stimulating hormone receptor (TSHR), sodium/iodide symporter (NIS) and thyroid peroxidase (TPO) confirmed that the cells that were isolated were derived from the thyroid gland. Both salivary gland cells and fibroblasts did not show any expression of these markers.

To determine the presence of stem cells in the thyroid gland sphere cultures, we studied their self-renewal capacity. For the self-renewal assay, after 1 day, thyropheres (passage 0) were dissociated to single cells using trypsin, were seeded into matrigel (20.000 cells/gel), and were passaged every 7 days. During each passage, cell numbers and the number of spheres formed were counted. During the first two passages the number of spheres that are formed is very low, but after the third passage there is an increase in the percentage of sphere formation. Thyroid gland cells can be cultured for more than 15 passages. The capacity of the cells to form secondary spheres indicates that stem cells are present in these cultures.

Figure 3.1. Images of primary thyroid gland spheres at increasing magnification.

In addition to their self-renewal capacity, stem cells should be able to differentiate into organ-specific cell types. For this we used a hypothyroid mouse model. I-131 was injected (10 MBq) and four weeks after injection decreased T4 levels were found in I-131-treated mice, indicating that these mice suffered from hypthyroidism. One week later, thyroid gland cells of passage 4 were transplanted underneath the kidney capsule. Eight weeks after transplantation, thyroid follicles were present under the kidney capsule which grew in size in time. These follicles contained thyroid specific cell types, visualized by thyroglobulin and thyroid transcription factor-1, indicating that adult stem cells isolated from murine thyroid tissue can generate thyroid tissue in athyroid mice.

3.2.3 In vitro experiments on thyroid and salivary gland cells

At the end of passage 2, irradiation experiments were performed. The cells were irradiated in small containers in a rotator to guarantee equal distribution and prevention of attachment. Cells were irradiated with 220 kV X-rays at high dose-rate (0.52 Gy/min). The same irradiation set-up was used for broad beam neutron irradiations performed at PTB. Both salivary gland and thyroid gland cells were irradiated with a high dose-rate (0.1 Gy/min) with single doses of 0, 0.1 0.25 0.5 1 and 2 Gy. The quasi-monoenergetic irradiations at Physikalisch-technische Bundesanstalt (PTB) Braunschweig were performed with a motor-driven single cuvette holder for three containers placed in a row with intervals of 2 cm between the first and the second and 3 cm between the second and the third container. The distance between the first cuvette and the irradiation tube was 5 cm. This installation ensures doses of 1 Gy for the first container, 0.5 Gy for the second and 0.25 Gy for the third cuvette. For the irradiation with secondary neutrons the same setup was used as for the x-ray irradiations. Secondary neutron irradiation was performed at the KVI-CART facility in Groningen with thyroid gland cells.

To study the sphere formation capacity, salivary gland stem cells and thyroid gland stem cells were replated at a density of 20,000 cells per well. One week after irradiation, spheres were counted and the dose-dependent decrease in the ratio of cells that formed spheres compared to unirradiated cells was determined (Fig. 3.2). Interestingly, after x-ray irradiation low-dose hypersensitivity to radiation was observed in the stem cells of both glandular organs, which was not present after neutron irradiation. Only at doses above 0.5 Gy do broad beam neutrons have a more pronounced effect on cell survival when compared to x-rays.

Figure 3.2. Surviving Fraction (SF) as a function of dose for thyroid gland stem cells. Best fit curves are obtained: with the Linear Quadratic model for SF as a function of radiation dose D: SF = exp (-αD-βD2), with α and β free fit parameters for neutron irradiations; with a modified LQ model (Joiner et al., 1988) for X-ray irradiations. The dotted line is the fit to X-rays survival data with a LQ model, given for reference.

DNA double-strand breaks (DSB) were quantified in salivary gland and thyroid gland cells at 30 minutes and 24 hours following x-ray and neutron irradiation using immunofluorescent microscopy for γH2AX. Thirty minutes post x-ray and broad beam neutron irradiation (initial damage) a dose-dependent increase in the percentage of γH2AX positive cells (≥3 foci) was observed. At 24 hours post x-ray irradiation (residual damage) the percentage of positive cells went back to base levels (same as unirradiated samples).
Salivary gland cells were irradiated at passage 1 and thyroid gland cells at passage 2 with x-rays or neutrons (Figure 3.3).

Figure 3.3. Extra yield (over the 0 Gy value) of foci as a function of dose scored in thyroid gland stem cells 30 minutes after irradiation. Best fit curves obtained with a linear dose-response model are also shown.

Expression of stem cell markers; CD24, CD29, CD44 and SCA were analysed after passage 3, passage 5, passage 10, and passage 15. No differences were observed in marker expression between cells irradiated with the different doses. RNA was extracted from salivary gland and thyroid gland cells. qPCR analysis was performed for four genes (TNKS, SNAI1, CCND2) that were selected from a qPCR array. Samples were analysed at passage 2, 4, 9 and 14. Some differences in marker expression were found between unirradiated and irradiated samples, however not sufficient to correlate them to transformation.

3.2.4 In vivo experiments on thyroid and salivary gland cells

To test the tumorigenic potential of irradiated cells we used two different mouse models, one model for salivary gland cells and one for thyroid gland cells.

Salivary glands were locally irradiated with a single dose of 10 Gy which is known to induce sufficient damage to allow stem cells to regenerate the tissue without compromising the general health of the animals. Thirty days after irradiation, both submandibular glands of irradiated mice were injected with 5000 unirradiated or irradiated cells into each gland. The animals were followed up for 1 year to allow tumour formation. At the time of tumour appearance or at the latest after 1 year, the mice were sacrificed and the salivary gland and tumour analysed. Three experiments have been performed using this model with transplantation of different treated cells into 35 mice. One year after transplantation the mice were sacrificed and the salivary glands were dissected. There were no differences in the weights of the glands between the different mice. No macroscopic tumours were identified in the glands. Tissue sections were stained with hematoxilin and eosin but no differences in tissue morphology was observed between the glands that were transplanted with irradiated or unirradiated cells.

To study whether tumour formation will occur after transplantation of irradiated thyroid gland cells, we used a different model - a subcutaneous transplantation model. As a positive control we used a cell line that is known to form tumours after subcutaneous transplantation - human FTC (follicular thyroid cancer) cells. Seven weeks after transplantation, tumours were formed in these mice, indicating that the model is successful. The samples that we transplanted were: unirradiated thyroid gland cells (n=12), thyroid gland cells that were irradiated with 1 Gy of x-rays (n=12) and thyroid gland cells that were irradiated with 1 Gy of broad beam neutrons (HDR) (n=12). Until the end of the project (12 weeks after transplantation), no macroscopic tumours have been formed, but these mice will be monitored for up to 1 year after transplantation.

3.3. Carcinogenesis of breast tissue cells

3.3.1 Isolation of breast tissue cells

In compliance with ethical guidelines, breast tissue was removed during reduction mammoplasty and glandular tissue was mechanically isolated. Several digestion and filtration steps were performed before cultivation. Primary mammospheres were formed after 7-10 days in a floating 3D cell culture system. Since primary mammary gland stem cells did not survive in vitro long enough for the planned experiments, established non-tumorigenic MCF10A cells originating from the human breast and which contain a subpopulation of progenitor cells were used for in vitro experiments. Breast cells were irradiated and plated for colony formation and DNA damage assays. The plating efficiency of these cells was 35%, on average.

3.3.2 In vitro experiments on breast cells

The cell survival curves fitted with the linear quadratic formalism are shown in figure 3.4.

Figure 3.4 Surviving fraction as a function of dose for mammary gland cells. Best fit curves obtained with the LQ model as in figure 3.2 are also shown.

DNA damage in breast cells was determined 24 hours following the different irradiations by counting the number of radiation-induced residual γH2AX foci as an indication of unrepaired DNA double-strand breaks (DSB). (Figure 3.5)

Figure 3.5. Extra yield (over the 0 Gy value) of foci, as a function of dose, scored in mammary gland cells 24 hours after irradiation. Best fit curves obtained with a linear dose-response model are also shown.

In conclusion, the RBE values based on the colony-forming assay and DNA-DSB have the same qualitative characteristic, but are not comparable quantitatively.

The terminal duct lobular unit (TDLU) assay was established to investigate differentiation in vitro. Freshly isolated breast cells were cultivated in an extracellular matrix together with human mammary fibroblasts (HMF) for up to 10 days. After incubation with different supplements, terminal duct structures could be observed. Yet, after x-ray irradiation no structural changes of TDLUs were observed and the quantification of TDLU number or degree of differentiation was very difficult. For this reason it was decided, to cancel further TDLU experiments with neutrons.

72 hours after IR the expression of MUC-1 and EpCAM, which are selective markers for stem and progenitor cells, were analysed in breast cells. Both surface markers showed variable responses to radiation but conclusions could not be made due to high uncertainty in the measurements.

The RNA of all samples of breast cells was isolated 72 hours following neutron and photon IR exposure. Three genes which were known to be related to breast cancer were analysed. All genes were measured by qRT PCR using the house-keeping gene GAPDH and normalized to 0 Gy control. Breast cells irradiated with 0.25 Gy – 1 Gy of 1.2 MeV monoenergetic neutrons showed clearly significant decreases in the expression of RB1, p27 and p53. The expression of p53 was also decreased following 0.5 Gy and 1 Gy of 0.565 MeV monoenergetic neutrons and after the exposure to secondary neutrons from 0.19 Gy up to 0.76 Gy, (Figure 3.6 right). Furthermore the expression of p27 is significantly reduced after 0.76 Gy of scattered secondary neutrons.

Figure 3.6. Fold difference in expression of tumour suppressors (RB1, p27, p53) relative to GAPDH 72 hours following 1.2 MeV monoenergetic neutrons (left), 0.565 MeV monoenergetic neutrons (middle) and scattered secondary neutrons (right); * p ≤ 0.05 ** p ≤ 0.01; mean ± SEM

Alongside the analysis of mRNA expression of p53, RB1 and p27, the RT2 Profiler Array was performed for X-ray (0 and 1 Gy) and monoenergetic neutrons (0 and 1 Gy) with 0.565 MeV to obtain an overview of 84 genes which are possibly affected by irradiation. 42 genes were influenced by X-rays and 17 genes by neutrons, 14 of these genes by both radiations. The affected genes are distributed over all functional gene groups.

3.3.2 In vivo experiments on breast cells

For in vivo experiments, freshly isolated human mammary gland stem cells were irradiated and transplanted the following day into prepared SCID/NOD mice. The fat pad of three-week-old female mice was cleared and then humanized with human mammary fibroblasts. Two weeks later 50.000 human stem cells were transplanted into the cleared fat pads. In order to produce a hormone level in mice that is comparable to that of breast development during pregnancy of a human woman, sham pregnancies of the mice were induced.

Altogether 73 mice were successfully grafted with a total of 145 transplants. All animals were regularly checked for evidence of tumours in the transplanted sites. Animals were sacrificed when evidence of tumour was found or when they showed signs of poor health or at the termination date of 6 months after transplantation (i.e. the mean life expectancy of the SCID/NOD mice). The mammary fat pads containing the grafts were removed and prepared for histological examination. 145 samples were histologically analysed in total. 39 of the 145 tissue samples showed pathological features, 21 of 145 (14 %) of all analysed samples contained mammacarcinomas, 3 showed fibrosarcomas, 3 presented with nerve sheath tumours. In addition there were 3 samples with hyperplasia and 9 showing signs of chronic inflammation. The mammacarcinomas were diagnosed in six different radiation groups: 220 kV X-rays HDR (8 of 45, 18 %), 220 kV X-rays LDR (2 of 16, 12.5 %), broad beam neutrons HDR (4 of 27, 15 %), broad beam neutrons LDR (3 of 11, 27 %), 1.2 MeV monoenergetic neutrons (1 of 10, 10 %) and scattered secondary neutrons (3 of 24, 12.5 %).

Table 3.1. Expression of p27 in histological analysed tumours

There was no indication of dose-dependence within the various radiation qualities, nor in all experiments combined. The incidence rate of mammcarcinomas in the dose groups 0 – 0.5 Gy varied between 17 and 25%, only 1 breast cancer was observed in 48 grafts of human breast cells which had been exposed to 1 Gy. On the basis of histological staining of p27, a score of expression could be determined. There was no dependency of p27 expression on radiation dose or quality.

3.4 Relating radiobiology experiments to track structure modelling

The results of the project task on modelling neutron effects provided RBE values for the DSB cluster endpoint associated with neutron experimental beams (1.2 and 0.56 MeV mono energetic beams at PTB, neutrons from the broad beam spectrum at PTB and secondary neutrons from a 190 MeV proton irradiated water phantom at KVI) used for the cell irradiations. The aim of this part of the project was then to investigate possible correlations and functional relationships (particularly in terms of dependence on neutron energy) between predictions and measurements of RBE obtained with selected endpoints.

Two specific endpoints have been identified for this analysis, namely: γH2AX foci scored in irradiated cells at late time points (residual damage at 24 hours) after irradiation, and clonogenic cell survival. Cells used for measurements were mammary gland cells (UROS) and thyroid gland stem cells (UMCG).

The reason why residual DNA damage has been chosen is as follows: when we look at damage of sufficient complexity (such as DSB clusters), we can argue that the DNA repair system of the cell will be found still active at long times after the exposure. The RBE for DSB cluster induction can then in principle be correlated to experimental RBE obtained from the scoring at 24 hours of radiation-induced foci (defined as the yield of foci at a given dose minus the yield of foci scored in the 0 Gy condition). If a linear dependence of such extra yield of residual foci with increasing dose is found, then the RBE is simply given by the ratio of the slope parameters of the linear fits to neutron- and X ray-induced foci data. This ideal situation is complicated by several factors: (i) a simple one to one DSB cluster - focus relation cannot be assumed, since the genomic length of interest in the definition of a DSB cluster (25 bp) is much shorter than that of interest for the formation of the focus (of the order of the Mbp); (ii) parameters such as the resolution of images used for foci counting, or other foci characteristics as their dimensions or brightness should be considered with the aim of quantitatively reproducing the yield of radiation-induced foci; (iii) the observed experimental extra yield of foci versus dose relationships appears not to be fully linear, but with a saturation effect at higher doses. As a consequence of all these points, correlations between the predicted and measured RBEs can be obscured, and we would expect RBE values obtained from foci induction to be subject to large uncertainties. The easiest correlation between data and model predictions can be looked for in a simple ranking of effectiveness: no agreement is found in this sense, since RBE for DSB cluster induction is higher for the mono energetic beams at PTB than those from KVI and from the broad PTB spectrum, while secondary neutrons at KVI always look to be the most effective for foci induction at 24 hours. (See figures 3.3 and 3.5.) All other experimental RBE values for the same endpoints are compatible with 1 if uncertainties are taken into account.

Concerning clonogenic survival data, from the modelling it is known that the DSB cluster endpoint, as a function of charged particle LET, is found to correlate with the dependence of survival-related RBE on LET. This should therefore remain true for neutron-induced DSB clusters. Survival data for mammary gland cells can be fitted with a Linear Quadratic expression for the surviving fraction as a function of the dose (figure 3.4) and the related RBE can be extracted at all surviving fractions. For thyroid gland stem cells, on the contrary, hypersensitivity at doses lower than about 0.2 Gy is found. (See figure 3.2.) This requires a modified version of the LQ model such as the one proposed by Joiner et al. (1988) to best fit data after x-ray irradiation. This means that low-dose RBE values are necessarily lower than one, which cannot be reproduced by modelling DNA damage induction but requires a much more complex modelling of the cell as a biological system responding to a radiation insult. When looking for correlations between predicted and measured RBEs at the low dose limit, we now find agreement in terms of ranking of effectiveness of the different neutron beams for mammary gland cells: the two monoenergetic setups are found to be more effective than the broad beam, as for the predictions of DNA damage induction, but again neutrons at KVI have the highest effectiveness, which is not compatible with model predictions. For thyroid gland cells a ranking of effectiveness at doses >0.5 Gy sees again neutrons from KVI as the most effective radiation, in disagreement with model predictions, while 1.2 MeV neutrons are generally more effective than neutrons from the broad beam spectrum at PTB, which is compatible with model predictions.

We have finally tried to investigate possible correlations between foci scoring and survival data for mammary gland cells: possible ways to quantify effectiveness in this case are to look at the survival chance after a given average residual damage, or at the level of residual damage necessary to correlate with the same survival chance. In particular, if we look at survival after a given average residual damage, e.g. when cells have on average 1 focus, we found that the probability of clonogenic survival is much lower if such a focus is induced by 1.2 MeV or 0.56 MeV neutrons, than by secondary neutrons at KVI, neutrons from the PTB broad beam or x-rays. This might suggest that, even if the focus we are looking at equally indicates the presence of non repaired DNA damage after 24 hours in all cases, such damage has different implications on clonogenic cell survival depending on the radiation producing it. This in turn suggests that single foci in different situations are the result of different levels of DNA damage, depending on the radiation inducing them. This is fully supported by DNA damage calculations: a single event induces more DSB clusters for 1.2 MeV or 0.56 MeV neutrons than from secondary neutrons at KVI, neutrons from the PTB broad beam or from x-rays. The ranking of effectiveness which can be now inferred from different SF when a single focus per cell is counted on average, is now the same as predicted by DSB cluster induction. In particular, the maximal effectiveness of secondary neutrons at KVI does not hold any more, and KVI neutrons are as effective as neutrons from the broad beam at PTB in terms of survival at equal residual damage (figure 3.7).

Figure 3.7. Predicted RBE values for DSB cluster induction for the different experimental setups as a function of neutron energies (average neutron energies for the broad beam and the KVI setups are respectively 5 and 17 MeV), compared to different experimental RBE values from Surviving Fraction (SF) vs. extra yield of residual foci correlations in mammary gland cells: RBEs are obtained as SF ratios at fixed ΔY (RBE from Sf in the legend) or as residual damage ratio at fixed SF (RBE from ΔY in the legend).

To conclude, when looking at functional relationships between predicted and measured RBE values and average neutron energies in the different irradiation scenarios (average energies for the broad beam and the KVI setups are respectively approximately 5 and 17 MeV), the decreasing trend of neutron effectiveness as a function of neutron energy is consistently found both in calculated and experimental RBE up to the average energy of the broad beam irradiations. Experimental RBEs from foci induction and survival data separately tend to indicate a maximal effectiveness for neutrons in the KVI irradiations, which is not consistent with calculations, but when looking at survival at a given residual damage (1 focus), secondary neutrons at KVI are found to be less effective (as affective as the broad beam in experimental data), in accordance with expectations from the modelling.

4. Developing the tools to reconstruct patient dose and risk

4.1 Calculation of neutron dose from patient data

The neutron dose calculation method was developed using the pencil beam scanning (PBS) proton therapy on Gantry 1 at the Paul Scherrer Institute (PSI). The method enabled spatially localized neutron doses to be calculated using a Monte Carlo-based parameterization of neutron dose and neutron spectra kernels for proton pencil beams. The parameterization was developed in order to calculate the neutron component of the delivered dose distribution for each treated patient. We have focused on neutron dose from PBS in this task as, for a passive beam line with closed aperture models were already developed (e.g. Schneider et al 2015).

Monte Carlo simulations were provided as described in section 1.2. The simulations were performed in a water phantom for single proton beams. The geometry of the nozzle at Gantry 1 at PSI including the treatment room walls were incorporated into the simulations with as much detail as possible. The beam characteristics were included into the Monte Carlo simulations and verified with measurements. The quantities determined by the Monte Carlo simulations included the three-dimensional total-dose distributions within the phantom, the dose induced by neutrons and their secondary particles and the energy spectra of the neutrons produced inside the phantom material.

The neutron dose parameterization consists of two steps. First the neutron dose on the central axis of the proton pencil beam was modelled absolutely in Gy per proton and then the lateral neutron dose was modelled relative to the central axis dose. Mechanistic models were used for both steps. The complex dose distribution in a patient is then calculated by summing up the single proton pencil beams including the neutron dose model in the dose kernel. The neutron dose was modelled by assuming that each point on the proton pencil beam acts as a neutron point source. Three physical processes were considered for describing the dose distribution from a neutron point source: dose build-up, inverse-square law, and exponential attenuation in the material.

The radially symmetric neutron dose kernels were computed as a function of water equivalent range and radial distance from the central axis of a single pencil beam. Such kernels were calculated for each of the nominal energies used for treatments on Gantry 1 at PSI (i.e. 138, 160 or 177 MeV). Depending on the nominal energy used for any field, the appropriate kernel is applied at the position of each individually applied proton pencil beam of the field (typically many thousand per field) and the neutron dose to all dose calculation points in the calculation matrix from this single pencil beam calculated. The results, compared to MC calculations are shown in Figures 4.1 and 4.2. The neutron dose calculation algorithm was integrated into the treatment planning system at PSI, so that neutron fields can be calculated on any proton patient treated. An example is given in Figure 4.3.

Within the same calculation framework, analytical models for the spatially localized neutron energy and possible energy dependent neutron RBE models were obtained. In particular, the model of neutron RBE for DSB cluster induction obtained within the project (Fig.2.1) was used.

Figure 4.1. Modelled neutron dose in Gy/proton on the proton pencil beam axis as the solid lines shown together with Monte Carlo simulated data for the proton energies 138 MeV (diamonds), 160 MeV (triangles) and 177 MeV (squares).

Figure 4.2. Modeled relative lateral neutron dose as the solid lines shown together with Monte Carlo simulated data for the proton energy 177 MeV in a proximal and in b distal to the Bragg Peak.

Figure 4.3. Proton dose distribution for a pediatric ependymoma planned with 160 MeV protons in a. The corresponding neutron dose distributions in mGy per fraction are shown in b-d.

4.2 Estimation of risk of SMN from neutron dose

The development of the SMN model proceeded in two steps. First the risk of developing an SNM in salivary, thyroid, and breast tissue following conventional radiotherapy was estimated from published data. Then this was modified by a weighting factor to take into account the RBE proposed by the project for neutrons compared to the low-LET radiation produced by medical linear accelerators. The result was an algorithm that enabled the estimation of the risk to a proton therapy patient from the reconstructed neutron exposure information obtained as described in the previous section.

The basic SMN risk model was derived from a combination of the Japanese A-bomb survivor data with secondary cancer data from of Hodgkin’s patients from a Western population. The issue of risk transfer between Japanese and Western populations was addressed by transfer of the risk according to ICRP 103 (2007), by establishing a weighting of ERR (excess relative risk) and EAR (excess absolute risk) that provides a reasonable basis for generalizing across populations with different baseline risks.

Thus in the limit of small dose all proposed dose-response relationships approach the LNT model and the initial slope can be obtained from the most recent data for solid cancer incidence. Here the data for a follow-up period from 1958 to 1998 was used from a publication of Preston et al.2007. For greater doses, a recently developed mechanistic model for carcinoma which accounts for cell killing and fractionation effects was used. An example of this for the salivary gland is shown in Figure 4.3.

Unfortunately not all organ specific risks could be fitted by using the Hodgkin RT data. It was therefore decided to fit directly from epidemiological data on second cancer risk for thyroid and female breast (in addition to the Hodgkin- model), since these are of particular importance for the ANDANTE project. For these organs the data from Travis et al (2003) and Bhatti et al (2010) provide dose data at the location of the origin of the second tumour. It was therefore possible to combine these data with the initial slope of the A-bomb survivors at low dose to obtain dose response relationships. An example for breast tissue is given in Figure 4.4.

Figure 4.3. Plot of excess absolute carcinoma risk for cancer of the salivary glands per 10,000 persons per year as a function of point dose in the organ. The 3 curves correspond to different assumptions of the amount of repopulation/repair. The fits are presented for age at exposure of 30 years and attained age of 70 years.

Figure 4.4. Plot of excess absolute carcinoma risk for cancer of the female breast per 10,000 persons per year as a function of point dose in the organ. Fully modelled dose-response relationship is depicted by the solid line and the symbols represent the epidemiological data of Travis et al (2003). The fits are presented for age at exposure of 30 years and attained age of 70 years.

5. Planning for a prospective epidemiological study

5.1 Pilot study using retrospective data

In preparation for the projected large-scale prospective study, the ANDANTE project planned a pilot study using real clinical data, to demonstrate that all components required were practical and feasible, and that the tools developed for dose reconstruction and SMN risk prediction would work. This would use the retrospective patient data retrieved from the paediatric proton patient database at Loma Linda University Medical Centre (LLUMC). However clinical and beam line data from LLUMC were eventually not available, and an alternative plan had to be developed, which was consequently followed. We were able to identify a small retrospective cohort of 18 patients treated with cranio-spinal irradiation at the MD Anderson Cancer Centre proton therapy centre in Houston, TX. These patients had previously been selected for Monte Carlo-based neutron organ dose modelling and second cancer risk evaluation for research undertaken at MDACC. Using these data to validate the risk model developed in the previous section (4.2) was further discussed and a plan outlined.

Using the treatment planning and CT data, it is possible to reconstruct the neutron fields at the sites of interest such as breast and thyroid, using the methodology described in section 4.1 and to implement the SMN risk model described in section 4.2 to predict individual SMN risks for breast and thyroid in these paediatric patients. A practical way for how to accomplish this was discussed with Dr. Newhauser and Dr. Choonsik Lee from the Radiation Epidemiology Branch of the NCI.

During the ANDANTE project period, we established a close working relationship with Dr. Choonsik Lee from the Radiation Epidemiology Branch of the NCI. Dr. Lee has made considerable contributions to the development of anthropomorphic phantoms for applications ranging from radiation protection to organ dose calculations from medical exposures. He was involved in the creation of the NCI-University of Florida computational anthropomorphic phantom family. Figure 4.5 shows a schematic representation of these computational phantoms, which are still being developed. A main advantage of these phantoms is that they are constructed from non-uniform rational basis splines (NURBS) and polygon mesh surfaces representing reference new-born, 1-year-old, 5-year-old, 10-year-old, 15-year-old male and female, and adult male and female humans. Coupled with Monte Carlo simulations or, eventually validated analytical models, these phantoms can be used to estimate patient organ doses.

Figure 4.5. The NCI-UF family of anthropomorphic phantoms (courtesy Dr. Choonsik Lee).

Further derivations from of the original NCI-UF paediatric computational phantoms will be selected according to parameters such as age, height, and body mass, and will be morphed to the CT scans of the MDACC patient cohort. This is an active field of research of Dr. Lee, but is believed to lead to relatively accurate organ doses. This work is foreseen as a follow-up of ANDANTE as a collaboration between Dr. Schulte, Dr. Lee, Dr. Newhauser, and Dr. Schneider with support from a graduate student at LSU.

Since the clinical proton beam line at MDACC was among those for which we obtained measured out of field doses (reported in 1.2) it is appropriate to use MCNPX and Geant4/TOPAS (in comparison) for the evaluation of organ doses in the retrospective patient cohort from MDACC. For the PSI patients, one can use the analytical PBS model already developed. While the beam line geometry is available for MCNPX MC simulations, we will also consider extending this work and develop an analytical dose model for the passive scattering MDACC beam line. This will be also done as a follow-up of this project in collaboration by Dr. Newhauser and Dr. Schneider.

Once organ doses become available, application of the initial risk model developed in the project to calculate the life-time risk of SMNs in different organs will be a relatively straight-forward task. For the purpose of achieving a demonstration of feasibility of the methodology, the work done by the project gives sufficient confidence to proceed with the planning for a prospective study, as described below.

5.2 Cohort size and follow-up time

The first step in planning a prospective study is to estimate how large the cohort needs to be and how long the study must continue for, to have any chance of achieving statistically significant results for the risk of SMNs linked to the radiation therapy for primary cancers.

In order to determine the required study size and length of follow-up, two approaches were taken: an approach that takes summary information from a literature review into account, and a more conventional power calculation, done to ascertain if an effect of the neutrons on the secondary cancer rates could be detected within a follow-up time of two decades. The power calculations were based on published risks of all solid cancer per unit neutron dose (Walsh et al 2013) and various assumptions about the baseline SMN risk and neutron absorbed organ doses from scanned and scattered modalities of proton therapy treatment.

In order to estimate the necessary duration of a potential epidemiological follow-up, previous studies on the risk of SMNs after radiotherapy in childhood were systematically reviewed and 58 studies from 2001 until present were found. For the estimation of the required follow-up period, 17 of the 58 papers reported links between radiation therapy for primary tumours and secondary tumour occurrences.

The averaged follow-up times associated with the occurrences of different types of SMNs, calculated using information from the reviewed papers were found to be:
- Brain: 12.2 years
- Thyroid: 18.9 years
- Oral cavity & pharynx: 17.1 years
- Liver, pancreas: 18.9 years
- Bladder: 20.4 years
- Ovary: 18.9 years
- Breast: 17.0 years
- Bone: 16.4 years
- Soft tissue: 18.9 years
- Melanoma: 18.1 years
- Lung: 21.9 years
- Uterus, colon: 25.2 years

With regard to the fact that brain and/or nervous system cancer, breast cancer, and thyroid cancer have the shortest times to occurrence after radiotherapy, a length of follow-up of a of minimum 20 to 25 years should be planned for, in order to minimize the potential influence of biases in the resulting risks.

Power calculations were done assuming that a radiation effect would not become discernible if the power from a statistical test to detect a difference between a disease rate in the exposed sub-population and the corresponding rate in the general population (assuming a significance level of 0.05) is less than 80%. The statistical tests applied here assumed the following ideal conditions:
The baseline risks are known (rates of disease for Europe for the age/gender subpopulation); health effects in the subpopulation can be detected without any loss of follow-up (this is unrealistic – but this gives an upper limit to the power); and cancer registries are complete. The excess relative risk (ERR) estimates for all solid cancer per unit organ averaged mean neutron dose (ERR/Gy) from the Japanese A-bomb survivors Life Span Study (LSS) (Walsh et al 2013) were applied in the power calculations.

The results of the power calculations are that powers of over 80% are only achievable, with a 20-year follow-up, if:
1. An average of 2000 persons per year enter the cohort (i.e. for example 1000 from Europe and 1000 from USA) over the 20 year follow-up period.
2. The organ average absorbed neutron dose from scattered proton therapy is 10 mGy
3. The solid cancer ERR/ Gy, organ averaged mean neutron dose, is 24 (i.e. at the upper 95% confidence interval of the range indicated by analyses based on the LSS data).
Since solid cancer risks per unit neutron dose are unlikely to be centred at the upper 95% level indicated by the LSS, the power calculations indicate that an effect of the neutrons on the secondary cancer rates is unlikely to be detected within a few decades of epidemiological follow-up time.

5.3 Planning for a prospective study

The database envisaged by the ANDANTE-project involves many proton therapy centres in different countries. The first step was the identification of candidate proton therapy centres and, for this purpose, contact was made with large national and international organizations (e.g. National Cancer Institute (NCI), USA, or WHO-IARC) for expressions of interest and initial data collection. Information was collected in detail from 17 European proton therapy centres and estimates of the numbers of paediatric patients per year treated in USA proton therapy centres were made.

BfS concentrated on directly collecting data from the European proton therapy centres. Eighteen currently existing proton centres in Europe were listed and contacted. After exclusion of those proton therapy centres which have not opened so far, 17 operating European proton therapy centres remained. It was possible to obtain preliminary information, on annual numbers of paediatric patients treated, typical indications for proton therapy and names of staff members interested in involvement in future work on the prospective study, from eight of these centres. Estimates of the number of paediatric proton therapy patients treated annually at USA proton therapy centres were provided by the USA partner (LLU).

Information and results from the LLU and MDACC feasibility studies as well as from the various proton therapy centres provide an excellent basis for the further development of the database for a future epidemiological study. It is recommended that an international registry of childhood cancers treated with protons (and photons) should be established, with the aim of investigating all potential late side effects of proton (and photon) therapy, including second cancers. This should be done in close cooperation with existing childhood cancer survivor registries in Europe and the USA. Currently, plans along these lines are being pursued in close collaboration between USA, European and International institutions, and should be supported. The primary aim of such an international registry should not be the determination of neutron RBE, since the ANDANTE power calculations (made in task 4.4) do not indicate that an effect of the neutrons on the secondary cancer rates could be detected within a few decades of epidemiological follow-up time. However, if the registries are established and become functional in providing medical information on overall and specific treatment outcomes, the scientific questions related to the aims of the ANDANTE project could be addressed at a later date, depending on the further development of proton treatment concepts and technologies. Ultimately, the far-reaching goal is to enhance our understanding of the link between radiation exposure to proton (and photon) therapy and the subsequent risk of SMNs linked to such treatments.


Baiocco et al. (2015) Reaction mechanism interplay in determining the biological effectiveness of neutrons as a function of energy Rad. Prot. Dos. (2015), Vol. 166, No. 1–4, pp. 316–319

Baiocco et al (2016) The origin of neutron biological effectiveness as a function of energy. Under submission

Bhatti et al. 2010: “Risk of Second Primary Thyroid Cancer after Radiotherapy for a Childhood Cancer in a Large Cohort Study: An Update from the Childhood Cancer Survivor Study”, Radiation Research 174 (2010).

Clifton, K.H. et al.: Thyroid gland formation from inocula of monodispersed cells: early results on quantitation, function, neoplasia and Radiation effects. Int.J.Radiat. Oncol.Biol.Phys. 4: 987-990 (1978)

Gould, M.N. Biel, W.F. Clifton K.H.: Morphological and quantitative studies of Gland Formation from inocula of monodispersed rat mammary Gland cells. Exp.Cell Res. 107: 405-416 (1977)

ICRP Publication 103, Recommendations of the International Commission on Radiological Protection ICRP Ann. ICRP 37 (2-4), 2007.

Joiner et al, 1988 - M.C. Joiner and H. Johns, Radiation Research 114, 2 385 (1988)

Preston D.L. E. Ron, S. Tokuoka, S. Funamoto, N. Nishi, M. Soda,K. Mabuchi and K. Kodama K, “Solid Cancer Incidence in Atomic Bomb Survivors: 1958–1998,” Radiat. Res. 168:1-64 (2007).

Schneider C, Newhauser W, Farah J. An analytical model of leakage neutron equivalent dose for passively-scattered proton radiotherapy and validation with measurements. Cancers 2015 18;7(2):795-810.

Till,J.E. McCulloch, E.A.: A direct measurement of the radiation sensitivity of normal mouse bone marrow cells. Radiat. Res. 14: 213-222 (1961)

Travis LB, Hill DA, Dores GM, et al.: Breast cancer following radiotherapy and chemotherapy among young women with Hodgkin disease. JAMA 290(4):465-75 (2003).

Walsh L, Neutron relative biological effectiveness for solid cancer incidence in the Japanese A-bomb survivors – an analysis considering the degree of independent effects from γ-ray and neutron absorbed doses with hierarchical partitioning. Radiat. Environ. Biophys. 52(1), 29-36, 2013

Potential Impact:
Potential impact of the ANDANTE project

The ANDANTE project will have a significant impact on both the scientific community, and the wider radiation protection community and on the optimisation of particle therapy.

Short term impact

The short-term impact comes from both the confirmation of expected results and the emergence of unexpected results. Both types of result are important for taking science forwards.


Using the power of the Monte Carlo simulation method, it was possible to predict the exact combination of ionising particles impinging on biological tissues in the experimental irradiation arrangements used in the project. This was verified by taking measurements and comparing them to MC simulations of the identical geometry. The level of agreement can be added to all of the previous confirmations of the applicability of the MC method to radiation transport, and in particular the applicability of Geant4 and MCNPX to medical radiation exposure situations extended to neutrons.

Further to the agreement between measurements and simulations, was the demonstration that common phantom materials used for routine dose measurements in conventional radiotherapy (e.g. the tissue-equivalent phantom materials of soft tissue, and compact bone by the phantom company CIRS) were also tissue-equivalent for neutrons scattered by protons of typical energies used in proton therapy. Within the level of accuracy required for radiotherapy dosimetry, the scattered fields produced did not significantly differ from those simulated using the ICRU/ICRP standard human tissues. This has implications for the practical use of these materials in taking measurements to verify the levels of scattered neutrons produced by clinical proton (and high-energy photon) treatment machines.

Possibly the most interesting result to be presented to the scientific community, is a physical explanation for the apparent peak in biological effectiveness of neutrons compared to photons at a neutron energy of around 1 MeV. Up until now it has been somewhat of a mystery as to what causes this, with no apparent peak in any of the relevant physical interaction coefficients. But with the power of the Monte Carlo method it is possible to investigate levels of detail that are not possible using experimental methods, and it was shown that the explanation rests on the production of low-energy knock-on protons by the neutrons at this energy that have high stopping power with a similar effect to the Bragg peak in the proton beam. This provides a substantial endorsement of the shape of the curve used for the radiation weighting factor WR in the recommendations in ICRP 103.

Another output from the project that has wide potential use is the algorithm for calculating the 3-dimensional neutron dose distribution in energy bands from patient treatment data on a pencil-beam scanning proton machine. It was demonstrated that this algorithm can be incorporated into a proton therapy treatment planning system to enable an out-of-field scattered neutron risk estimate to be made on every patient treated. For patients with good expected cure rates and long subsequent life expectancy the SMN risk can be factored into the planning optimisation.


A method for isolation of (mouse) thyroid stem cells and an assay to confirm their identity was developed for the ANDANTE project. This is the first time that the demonstration of the existence of thyroid stem cells has been achieved, and it will add this tissue type to the range of biological material available for experimentation.

In the experiments on salivary gland and thyroid gland stem cells at UMCG, a clear hypersensitivity to the 200kV x-rays was demonstrated for the surviving fraction able to form spheres. Survival showed a minimum at around 200mGy for both cell types. This was not seen in either the exposures to neutrons, or in the experiments performed on breast cells at UROS. Hypersensitivity has been controversial since it was first observed, because so far there has been no totally plausible explanation. However, the fact that it was observed will certainly add to the debate.

The other remarkable result from the radiobiology experiments that was quite unexpected was the apparent absence of any effects of either the x-ray of neutron radiation on any of the carcinogenesis-related markers throughout the range of doses from 0.0 to 2.0 Gy. This was not what was hoped for the sake of the objectives of the project, because this gave no information at all on RBE. But it is a very interesting finding to add to the knowledge of how stem cells respond to radiation and the role that they may play in the process of carcinogenesis.


The contribution to the scientific community from the epidemiological component of the project is the planning and tools in preparation for a large-scale prospective epidemiological study to use the growing clinical data from proton treatments of paediatric cancers to gain a greater understanding of qualitative and quantitative effects differences between the effects that neutrons have on biological systems compared to low-LET radiation. The methodology and software are available for PBS proton beams and have been demonstrated on patient data. Meetings have been organised with the National Cancer Institute (NCI, USA) and International Agency for Research on Cancer (IARC, WHO) in 2016 to discuss the setting up of the database that will be required for the study. Funding sources have been identified, and the proposal has already been taken up by key personal both in Europe and the USA.

Long-term impact

Like most of the challenges in the low-dose radiation risk research area, the problem of quantifying the long-term risks from neutron exposure more accurately than currently known, either in absolute terms or in relation to photon exposure, are likely to be solved only in the long term by a combination of bottom-up systems radiobiology and top-down epidemiology.

The process involved in getting from initial molecular damage to some biological structure to an outcome of a tumour decades later is too complex to admit a fully deterministic systems model at the present stage of knowledge and computing power. Nevertheless, ANDANTE has made a contribution that will be important for future understanding of the mechanisms of damage formation and propagation, using the track-structure modelling and stem cell methods developed in the project.

The project has set up all that is needed for a prospective epidemiological study that will test the RBE model proposed, and also from the results will be able to derive directly the risks of SMN’s from out-of-field exposures received during proton therapy. But the power calculations that predict the numbers required for the conglomerate cohort and the length of follow-up indicate that the study will need to continue for many decades to achieve statistically significant results.

Thus in each of the disciplines engaged in the project, the influence on future research will continue well into the next generation of researchers.

Wider potential impact

The results from the ANDANTE project will have an impact on any facet of radiation protection where neutrons are a significant factor. The re-evaluation of RBE for neutrons will provide information of fundamental importance to the ICRP formalism of radiation protection.
This will have direct implications for any industry where neutrons are produced as a by-product, for example:

The nuclear power industry (both conventional fission reactors and potentially fusion power in the future). The safety design of work areas, and the site safety rules and procedures are all dependent on the knowledge of the risks to health from neutrons.

High energy medical linear accelerators. Any linear accelerator producing x-rays of energy more than 10MeV generates unwanted neutrons, and the bunker design must include special elements such as a long entrance maze, borated polyethylene wall linings, neutron doors, etc. The level to which safety design must reduce neutron dose rate is determined directly from the estimated risks from neutrons.

Long-haul air crew. Guidelines for scheduling of high-altitude long-haul air crew are dependent on the estimated risks from neutrons.

More specifically since the validation and evaluation of the RBE models makes use of follow-up data from paediatric proton therapy patients, there is a strong possibility that the results from ANDANTE, certainly from the prospective epidemiological study, will have a direct impact on the understanding of long term risks of second cancers following exposure to neutrons during proton therapy. This will be of considerable benefit to an increasing number of children who are being treated with this modality in assessing the safety of the treatment compared to possible alternatives.

List of Websites:

Andrea Ottolenghi
Dipartimento di Fisica
Università degli Studi di Pavia
Via Bassi 6
I-27100 Pavia, Italy