1. Introduction
Proton beam therapy is becoming increasingly available due to its potential to deliver maximal doses on tumour while minimizing irradiation of surrounding healthy tissues/organs at risk. In fact, its ballistic precision has always been regarded as the main physical pillars of such a treatment modality [
1]. However, in recent years, with growing numbers of patients treated and longer follow-up periods, concerns about the potential side effects of protontherapy have arisen [
2,
3]. The potential of a given particle irradiation to induce a greater level of specific biological effects compared to photon irradiation, used as a reference, is expressed in terms of its Relative Biological Effectiveness (RBE) [
2]. Nowadays, in the proton-therapy clinical practice, a fixed RBE of 1.1 is assumed [
3,
4,
5]. This reflects the known relationship between RBE and Linear Energy Transfer (LET), as in the tumour volume low-LET protons can be generally found. Despite this, data emerging from various radiobiological studies indicate that RBE of protons exceed the accepted value, at least in the distal part of the clinical spread out dose distribution, where high-LET slowing-down protons become predominant. Hence, a variable RBE should be used to more precisely describe the radiobiological effects on the tissues and organs involved in the treatment [
6,
7,
8,
9,
10].
Successful treatment planning largely depends on the accuracy of biophysical models. To assure their precision they need to be validated against data from in vitro studies. Therefore, development of the database with experimental results obtained on various cell lines is useful for RBE model validation as well as for comparing different models [
11].
While a vast literature exists reporting on the clinical results of eye proton therapy [
12,
13,
14,
15,
16], no in vitro data evaluating the cellular radioresponse of uveal cancer cells along a clinical proton SOBP are available. There is only scarce data regarding the response of human melanoma cell lines along with the proton spread-out Bragg peak. According to these data, at the distal end of a 65 MeV proton SOBP, an RBE of 1.27 and even higher for the more resistant melanoma cell lines were recorded [
17,
18,
19].More recently, work carried out on uveal melanoma Mel270 cells at the mid-SOBP position of a clinical high-energy proton beam showed an increased anti-migratory effect compared to photon irradiation [
20]. Currently, many commercial Treatment Planning Systems (TPSs) for proton therapy incorporate the capability of calculating LET distributions and RBE estimations, but there remain challenges on how to deal with uncertainties resulting from potential RBE variations. Future TPS that would potentially be able to take into account RBE variability, could reduce the toxicity and the incidence of later-occurring induced morbidities, thus making it possible to exploit the full potential of proton therapy [
2]. The main aim of this work was to study the RBE values at the clinical SOBP of the CATANA proton therapy facility. Radiation-induced biological effects measured in terms of clonogenic cell death (see below) were predicted by the LEM II (Local Effect Model) [
21], Monte Carlo approaches [
22], and semi-empirical models based on LET experimental measurements [
23]. All adopted methods were then compared with in-vitro experimental results. These were evaluated using the human 92.1 uveal melanoma cell line, a well-known model to investigate the aggressive behaviour of such types of neoplasia. In order to compare the effects induced by proton therapy in a non-tumorigenic cell line the human normal retinal pigment epithelial ARPE-19 cell line was also irradiated. The normal cells can easily recover from damage induced by ionizing radiation due to a more efficient damage response machinery with respect to the cancer cells. Moreover, choosing the normal retinal pigment epithelial ARPE-19 cell line allows obtaining insights into the magnitude of biological effects induced by an erroneous dose distribution during a proton therapy treatment plan. The retina is the organ at risk in uveal melanoma treatment and despite the usually excellent precision granted by protontherapy, the anatomical structure of the eye, with dimensions in the order of the millimeter is such that lack of a rigorous dose contouring is possible in actual clinical scenarios, leading to unwanted dose deposition. Such information is therefore of paramount importance to provide an estimate of the possible adverse effects deriving by beam forward straggling.
Microdosimetry experimental spectra acquired to estimate the proton LET in different positions along the SOBP, were measured in the same experimental conditions of the biological sample irradiation. Two different detectors were used to this scope: a mini-TEPC [
24,
25], developed at the Legnaro National Laboratories of the National Institute for Nuclear Physics (LNL-INFN) and a Silicon On Insulator (SOI) microdosimeter with 3D Sensitive Volumes (MicroPlus-Bridge) [
26,
27]. The Loncol’s weighting function was then applied to assess the microdosimetric RBE and compare it with the experimental one. The biological damage was estimated also adopting a computational method which couples Geant4 with the LEM-II model as well as the pure LEM-II model. In such a way, it was possible to compare the well-established LEM with a Monte Carlo method and a semi-empirical one and to establish an approach that best reproduces the biological data at a depth of 24 mm, the mid-position of the spread-out Bragg peak.
3. Results and Discussion
One of the aims of this work is the comparison of three methods to calculate the RBE in a clinical setting: microdosimetric-based, Monte Carlo and LEM II model. All of the proposed approaches were compared with experimental data. Radiation-induced cell death was measured by clonogenic assay and RBE values were calculated using the Linear-Quadratic model. The clonogenic survival curves of two specific cell lines were experimentally measured at 24 mm, the mid-position of the 62 MeV clinical proton SOBP. Additionally, at the SOBP entrance, the survival curves were also estimated from the experimental measurements of the microdosimetric spectra. The estimation of the survival curves with these three approaches permitted the estimation of the RBE values in the centre of the SOBP region. The radiobiological effects of protons were experimentally evaluated using the human 92.1 uveal melanoma cell line. In order to evaluate the damage induced by proton therapy in a typical tissue belonging to the same anatomical site, the human normal retinal pigment epithelial ARPE-19 cell line was also irradiated. The assessment of the detrimental effects in healthy tissues due to radiotherapy is pivotal, especially for the treatment of cancers affecting the eye since even slight impairment of the normal ocular tissue would lead to a loss of function of the entire organ [
42]. The results of the clonogenic curves with the fitted parameters, for the two cell lines irradiated at the proton mid-SOBP and under a reference gamma beam, are summarised in
Table 1. Table also report the corresponding derived values of RBEs.
Two fundamental radiobiological parameters were contextually obtained, the alpha and beta values, together with their ratio. Within the LQ model context, the alpha and beta coefficients of the clonogenic survival curves represent the single-hit, lethal damage, generally not repairable, and the multiple-hit damage, associated with sub lethal damage, respectively. Sublethal damage is responsible for the shouldered part of the clonogenic curve and can be usually repaired [
43]. The
ratio is an useful indicator of radiosensitivity, hence, its calculation is of paramount importance to evaluate the cell response to the radiation therapy (RT) treatment [
44,
45]. The clonogenic endpoint has, in fact, shown a lower radiosensitivity exhibited by the ocular cancer cells with respect to the normal cells as expected. In addition, for both cell lines, an increased RBE with proton beam irradiation was obtained. The capability of the two microdosimeters to describe not only the RBE but also the survival curve at the mid-SOBP was then investigated.
Figure 2 shows the acquired microdosimetric spectra from which the RBEs values and the parameters of the corresponding clonogenic curves were derived. Distributions are presented as
yd(
y) versus
y on a logarithmic
x-axis to facilitate visual interpretation. In this representation, the area under the curve between two values of
y is proportional to the fraction of physical absorbed dose due to the events in that interval.
Specifically, in
Figure 2 the two positions (entrance and mid of the SOBP) and the two detectors (mini-TEPC and MicroPlus Bridge) are considered. In the same figures, the Loncol’s biological weighting function
r(
y) and the corresponding weighted-dose curves
yr(
y)
d(
y) are also reported. The observed increase of the area under the red curve (i.e., the
yr(
y)
d(
y) curve) for the spectra acquired at the cell position (bottom row of
Figure 2), reflects an increase in the RBE. It can be observed, in fact, that at the entrance position the biologically weighted distributions overlap almost perfectly with the un-weighted distributions, because for
y < 10 keV/
m the weighting function
r(
y) almost equals almost equals 1. Different behaviour can be observed at the mid-SOBP, where the weighting function augments the contribution of events above 10 keV/
m leading to an increase of the
(Equations (
8)–(
11)). The RBE calculated at the mid-SOBP with the mini-TEPC, (
-TECP = 1.18), resulted slightly higher than that calculated with the Microplus Bridge (
-MicroPlus = 1.12). With respect to the entrance position, the
-TECP at the mid-SOBP increases by about 17%, and the
-MicroPlus increases by about 13%. The reason for this underestimation of the MicroPlus with respect to the TEPC is the larger simulated site size (about 17
m with respect to 1
m of the TEPC): low energy protons stop inside the silicon thickness, therefore the largest proton events (proton-edge) are at about 40 keV/
m for the Microplus, while they are at about 140 keV/
m for the mini-TEPC. The
values, corrected by a constant factor to have
= 1.1 at the entrance position of the SOBP, and the linear parameters
of the linear-quadratic model, calculated using Equation (
11), are reported in
Table 2. The
was assumed to be equal to
(see
Section 2.4).
Finally, the survival curves relative to the mid-SOBP position and derived from the microdosimetric data, experimentally measured from cell irradiation experiments and calculated with the described Monte Carlo and using the Survival software [
37], are reported in
Figure 3 (92.1 cell line) and
Figure 4 (ARPE-19 cell line). Both figures also show the survival curve measured for the
gamma-ray irradiation. The survival experimental curves were fitted using the double exponential function (Equation (
1)) and the
and
derived from it were reported in
Table 1. The survival fraction related to the experimental data and applyed approaches for the two investigated curves is reported in the
Table 3 (92.1 cell line) and
Table 4 ARPE-19 cell line).
The agreement between the three investigated methods, with the experimental radiobiological data, was statistically evaluated by applying the
test. Results of the test (
p-values included) are reported in
Table 5 and
Table 6: the
values are calculated taking into account the data reported in the
Table 3 and
Table 4.
In almost all the investigated cases, the resulting
p-value is well above 0.25, showing that the models are able to well reproduce the experimental data. The
p-value calculated applying the Loncol’s function and using the microdosimetric spectra measured with the MicroPlus probe is lower than 0.05 as shown in
Table 5. This disagreement could be related to the higher sensitive volume of the adopted detector with respect to Mini-TEPC.
4. Conclusions and Perspectives
In this study, we showed that the LEM-II model, the Monte Carlo-based LEM model and a semi-empirical method based on microdosimetric spectra, can adequately reproduce experimentally derived survival curves from in vitro cell irradiations performed at the mid-SOBP of the CATANA 62 MeV clinical proton beam.
For both cell lines, an RBE greater than 1.1 was measured and reproduced applying all of the investigated approaches. This points out that the biological damage of both healthy and tumorigenic cells is underestimated if a fixed RBE of 1.1 is assumed. Microdosimetric spectra were measured at the entrance and mid-SOBP position with both a mini-TEPC, simulating a 1 m water site size, and the MicroPlus microdosimeter (17 m water equivalent thickness). Despite the markedly different microdosimetric spectra obtained with these two detectors, when the Loncol’s weighting function is applied to the dose distributions both lead to an assessment of the RBE10 that is in good agreement with radiobiological data for ARPE-19 cells obtained in the same beamline. The higher discrepancy obtained with the 92.1 cell line and the solid-state detector could be due to a lower detector sensitivity related to the higher SV volume. The remarkable obtained result showed the potentiality of using a microdosimetric spectrum to predict not only the RBE but also the full survival curve, at least at the mid-position of the 62 MeV proton SOBP. Currently, different commercial treatment planning softwares for proton therapy incorporate tools for calculating LET distributions and RBE prediction. However, while the dose prescriptions calculated by the treatment planning are routinely verified with certified ionization chambers, there is no commercial equipment currently available to perform routine verification of LET or RBE distributions calculated by the TPS. The Local Effect Model is the computational radiobiological approach most widely used in hadrontherapy. The LEM model was fully integrated into the analytical treatment planning system TRiP98 and successfully tested for clinical application. In this work, the good agreement between the Monte-Carlo-based LEM model and LEM-II model with the experimental data clearly evidence how a TPS based on an RBE and LET prediction could improve the estimation of the radiobiological response of the treated neoplasia as well as the surrounding healthy tissues. Moreover, the pretty good accordance between the semi-empirical methods based on microdosimetric spectra with the experimental data shows that the LET estimation procedure should be successfully inserted in the clinical routine to optimize the RBE distribution. The current study identifies a clinical and critical aspect of the uveal melanoma protontherapy treatment. The correct evaluation of the damage induced by ionizing radiation during RT treatment, also thanks to Monte Carlo-based simulations, represents a valuable and potent tool to foresees the impact of ionizing radiation on cancer tissues. Moreover, studying both healthy and tumorigenic cells allows obtaining fundamental information about the biological effects induced by an erroneous dose distribution during a proton therapy ocular treatment plan. The biological response is therefore of paramount importance to provide an estimate of the possible adverse effects deriving by beam forward straggling. A correct prediction of the damage is a key point for a more personalized TPS.