Cellular irradiations with laser-driven carbon ions at ultra-high dose rates

Carbon ions were accelerated using the Astra GEMINI laser system of the Central Laser Facility at the Rutherford Appleton Laboratory, UK, that delivered ∼12 J, 40 fs single pulses onto ultrathin carbon foil targets, at intensities ∼6 × 10²⁰ W cm⁻², by an f/2 parabola. We used a double plasma mirror system for pulse contrast enhancement. To generate the carbon ion beam, we used 15 nm thick targets, which were irradiated by linearly polarized pulses at normal incidence. Under these conditions, the acceleration of ions from the target bulk takes place under the influence of multiple acceleration mechanisms. The carbon ions initially undergo radiation pressure acceleration in the Light Sail mode (Henig et al 2009, Macchi 2010) until the target becomes transparent to the relativistically intense laser pulse - during this second phase, efficient coupling of the laser pulse into electrons, as it propagates through the dense target plasma, leads to an enhancement of the accelerating fields which boosts the ion energy(Henig et al 2009, Higginson et al 2018) We note that the highest carbon energies can generally be obtained by using circularly polarized laser pulses, which reduce target heating and delay the onset of transparency(Scullion et al 2017, McIlvenny et al 2021) However, linear polarization was chosen in our experiment as it allowed the production of a more stable and uniform beam at the chosen ion energies of 10 MeV n⁻¹.

The proton and carbon-ion energy spectra were measured by using a Thomson Parabola Spectrometer (TPS) placed along the target-normal direction, coupled with BAS-TR Image Plates (IP). Figure 1(A) shows the energy spectra measured in a single shot for protons and carbon ions (C6+). Under the conditions of the experiment, we typically obtain comparable cutoff energies/nucleon for carbon ions and protons (Scullion et al 2017) (with carbon ions slightly more energetic than protons in the data shown in figure 1(A)).

Zoom In Zoom Out Reset image size

Given the multi-species nature of the beam, and the comparable energies/nucleon of the two main species, a challenge of the experiment was designing and implementing an experimental setup that allows irradiating a cell sample with a pure carbon beam, minimizing the proton contribution to the final delivered dose at the cell plane. This was achieved by placing a 0.9 T dipole magnet at about 50 mm from the target to select in energy and spatially disperse the ion beam accelerated along the target normal direction (figure 2 part (C)) (see also (Milluzzo et al 2020)). A 500-μm width slit was used to reduce the particle divergence captured by the magnet as well as the energy spread at the cell position. The cell holder was then placed in air at about 1 cm from a 50 μm Kapton beam exit window and at total distance of about 300 mm from the target. Such distance was chosen to ensure a sufficient spatial dispersion of protons and carbon ions at the cell position as well as to spatially separate the proton and carbon beams, while ensuring a sufficient ion flux reached the cells. A Geant4- simulation reproducing the experimental setup and the condition was performed to predict the energy-spatial dispersion of protons and carbon ions. Figure 1(B) shows the spatial distribution of protons and carbon ions after the magnetic dispersion at the cell position as calculated from the Geant4 simulations. The particle energy per nucleon is also shown in the color scale bar together with the particle fluence distribution (protons and carbon ions in red and black respectively) in the vertical direction. As seen in the figure, protons and carbon ions are spatially separated at the cell location and carbon ions with an energy per nucleon on the cells around 10 MeV u⁻¹ can be selected for the irradiation of cells without any contribution from protons. The short drift distance after the magnet also allows maintaining a sufficient high particle flux to reach Gy-level doses at the cell plane. The Geant4 model output of the experimental setup was validated by comparing the experimental spatial-energy distribution of carbon ions at the cell plane with the simulation predictions as shown in figure 1(C). The measurement has been performed by using RCF covered by different thicknesses of aluminum absorbers able to stop different carbon ion energies. By recording the signal cut-off on the RCF corresponding to the energy cut off caused by the different aluminum absorbers, it was possible to establish a correlation between the carbon ion energies and their position on the RCF. Figure 1(C). shows the comparison between the simulated energy dispersion (black point) and 3 experimental points obtained using 117, 177 and 229 um thick aluminum absorbers corresponding to an energy cut-off of 6.6, 8.6 and 10 MeV u⁻¹ respectively.

Zoom In Zoom Out Reset image size

In this study, we used patient-derived radioresistant glioblastoma stem like cells (E2 cells) that expressed the most common stem cells biomarkers such as Nestin and Sox-2 (Gomez-Roman et al 2020). The cells were cultured in Advanced DMEM-F12 medium supplemented with B27, N2, L-Glutamine, heparin, epidermal growth factor and basal fibroblast growth factor (Thermo Fischer Scientific, Loughborough, UK). Cells were grown as monolayers on top of 3 μm thick Mylar film mounted in the custom-made stainless-steel dishes (figure 2(A)). Before seeding cells, the dishes were thoroughly sterilized using 70% ethanol, rinsed in PBS, dried and then coated with 2.5% Matrigel dissolved in cell culture medium. The cells were seeded 24 h before irradiation and maintained in 5% CO₂ at 37°C inside cell culture incubators until irradiated. Before irradiation, the top part of the cell culture dish was sealed with a piece of Mylar and mounted with another ring plate to hold the Mylar in place. Later this top piece of Mylar was cut as semi-circle to prevent any spillage of medium during irradiation when the dish is held vertical. At this stage, for each sample an individual RCF holder (figure 2(B)) was fitted firmly under the dish to account for the shot-to-shot dose variation.

This whole cell dish and RCF holder assembly was then placed inside the sample irradiation chamber (figure 2(D)) which provided a closed system acting as a barrier between the cell medium inside the dishes and the outside air to prevent any bacterial contamination. This sample irradiation chamber was then transported to the Target Area and mounted next to the Kapton window for irradiation with laser-driven carbon ions.

After irradiation the cell irradiation system assembly was transported back to the biolab, RCFs were removed and the bottom of the mylar was marked with a permanent marker to highlight the corresponding irradiation field along with the direction of the energy gradient. These markings together with the RCF scans profile were later used to precisely score the cells for foci in uniform dose fields. Comparative low LET irradiations were performed with 225 kVp x-rays using an X-Rad 225 (Precision x-ray, Connecticut, USA), x-ray generator with a dose rate of 0.59 Gy min⁻¹.

The dosimetry procedure employed in the experiment has been reported in (Milluzzo et al 2020). We summarize here its main features. As mentioned earlier, the cell monolayer was grown within a stainless-steel assembled dish, which was specifically designed to reduce the material thickness crossed by carbon ions before reaching the cells. The dose delivered to the cells was measured for each shot using unlaminated EBT3 RCF placed just behind the cell monolayer using a suitable holder as shown in (figure 2(B)). The unlaminated EBT3 RCF is a customized design, which consists of an active layer overlaid on a single 125 μm matte polyester layer (in the standard EBT3 design, the active layer is sandwiched between two matte polyester layers). The use of unlaminated EBT3 enabled the detection of the dose deposited by low energy carbon ions with a higher accuracy, which would otherwise be stopped before reaching the active layer of standard EBT3. Moreover, the absence of one of the two substrate layers allows a close contact between the RCF active layer and cells to enable more reliable dose measurements with a maximum difference of ∼2% between the dose released to the cells and the one delivered to the EBT3 active layer. These unlaminated EBT3 RCFs were previously calibrated for dose response with conventionally accelerated 6 MeV n⁻¹ carbon ions at the LNS of the Institute of Nuclear Physics, Catania, Italy. The calibration was then used to convert the raw RCF images acquired to dose delivered in water for each shot.

P53 binding protein −1 (53BP1) is a widely used surrogate marker of DNA double strand break (DSB) damage, which migrates to sites of a DNA DSB within the cellular nucleus (Schultz et al 2000) and mediates further cell signalling to activate DNA repair pathways. This localisation can be readily detected through immunofluorescence microscopy in the form of bright foci. The number of foci correlates to the amount of DNA DSB damage and has been extensively used to study radiation induced DNA damage and repair (Marková et al 2007, Groesser et al 2011). This technique has also been used successfully to quantify DNA DSB damage and repair in laser-driven experiments (Raschke et al 2016, Hanton et al 2019). After irradiation, as described in section 2.2, the cells were fixed in 4% paraformaldehyde at various time points ranging from 0.5, 1, 6 and 24 h and stored until further processing. Later the cells were rinsed in room temperature phosphate buffered saline (PBS), blocked in 10% Goat serum in PBS for one hour at 37 °C and stained for 53BP1 immunofluorescence assay.

The area within the dish over which the cells were scored was selected on the basis of the information provided by the RCF, by means of a PMMA window with a small rectangular cut which was overlayed with a RCF region where the dose delivered by the ∼10 MeV n⁻¹ carbon was sufficiently uniform. This defined the chosen irradiation field, which was then marked on the rear of the cell's Mylar substrate using the same PMMA mask. This procedure was repeated for each irradiated cell dish. In this field, at least 150–200 cells were scored for each datapoint in duplicate providing a reasonable sample size using a Carl Zeiss Axiovert 200 M Fluorescence microscope equipped with excitation and emission filters suitable to image the nuclear foci stained with Alexa fluor 488 and counter stain nuclei stained with DAPI. Foci were counted manually using a 40X magnification Plan Apochromat objective with a numerical aperture of 0.95 and a free working distance of 0.25 mm, without any immersion medium. As the cellular nucleus thickness was comparable to the objective depth of focus, no Z stack analysis was required. However, to ensure all foci were counted in any of the Z plane of the cells, foci counting was done by varying the focal plane manually and the images were stored a.s. ZVI images for further analysis including foci size measurements.

Foci area measurement offers an insight into the extent of the damage, has been previously used to study the impact of radiation quality of DNA DSB damage (Ward et al 2003, Marková et al 2007), and has been shown to reasonably correlate to the LET of the ions (Costes et al 2006). 53BP1 foci area measurements were done using the particle counter function of the Image J image Analysis program (Schindelin et al 2012, Rueden et al 2017). Briefly, the length in pixels on an image was calibrated in micrometres by using the scale function in Image J and option 'global' was checked so that same setting could be applied to all images. Images with 53BP1 foci (FITC channel only) were threshold-adjusted so that only the foci were visible in the image. After adjusting the threshold, the 'Analyze particles' options was used to check each focus. The total number of foci analysed was at least 100 for each dataset. We then compared the 53BP1 foci area measurements in the cells exposed to laser-driven carbon ions and 225 kVp x-rays.

RBE, which is defined as the biological effectiveness of any particular type of radiation relative to reference radiation such as gamma or x-rays, has been shown to be a very useful parameter for optimizing ion beam radiotherapy (Carabe et al 2013, Jones et al 2018). While the cell-killing RBE is mainly utilized in such studies, here, we have used an alternative quantity named relative foci induction (RFI), which is simply the ratio of the values of carbon ions- induced 53BP1 foci to the x-rays induced foci at any given time after irradiation. This provides a comparative overview of the repair of DSB damage induced by x-rays and carbon ions.

Monitoring the dose in each shot was essential in our set-up, due to shot-to-shot variations of the dose delivered to the sample. This arises from variations of the ion beam parameter associated to small changes in the interaction conditions (e.g. laser energy, intensity, contrast, target homogeneity) currently outside our control given the complexity of the set-up. Across the experiment (about 90 shots for which we carried out dosimetric characterization), the average dose on the cells was 1 Gy with a standard deviation of 0.6 Gy. By down-selecting the irradiations (and using about half of the shots), the standard deviation was reduced to 0.2 Gy for the irradiations where the cells were processed.

The experimental setup was simulated using the Geant4 Monte Carlo code version10.05 as previously in (Milluzzo et al 2020). The experimental carbon energy spectrum shown in figure 1(A) was used as input of the simulation, in order to predict the energy dispersion at the cell position. This calculation enabled to fix the vertical position of the cells corresponding to about 10 MeV u⁻¹ carbon ions (see figure 3(A)) at about 50 mm from the 0 axis. Considering this, as shown in figure 3(A), a region of interest (ROI) with an area of 4 × 4 mm (light blue square) was selected on the RCF placed behind the cells corresponding to the carbon ion energy of interest. In the shot shown in (figure 3(A)), the cells were at the center on the RCF holder (as shown in figure 2(B), the RCF was placed under the cell monolayer which was adherent on 3 μm Mylar), while the beam was slightly off center. The shift was considered to correctly quantify the dose delivered at the position of the cells. Figure 3(B) shows the transverse dose profile obtained along the white line shown in figure 3(A) and averaged over 4 mm in the vertical direction. An average dose of 1 ± 0.1 Gy (10% variation) is estimated considering the total transverse profile (black line in figure 3(B)). In contrast, if we restrict the evaluation to the region where the cell were located (light blue square shown in figure 3(A)), an average dose of 1 ± 0.02 Gy (2% of dose variation) is evaluated for this shot. This analysis was performed for each shot. The Monte Carlo simulations were also used to accurately evaluate the final energy distribution of carbon ions reaching the cells within the 4 × 4 mm² ROI shown in figure 3(A). As discussed in (Milluzzo et al 2020), a final Gaussian energy distribution is obtained within the ROI, centered at 9.5 ± 0.8 MeV with a total FWHM of 1.6 MeV and an energy spread ∆E/E~15%. The estimation of the final energy spread at the position of the cells, considering also the transverse dimension of the sample, together with the shot-to-shot measurement of the dose delivered at the cells, enabled us to evaluate an important parameter, i.e. the average dose-rate employed for the cell irradiation.

Zoom In Zoom Out Reset image size

Knowing the energy spread at the cell position and the target-cell distance (300 mm) it is possible to retrieve the temporal duration of the carbon bunch reaching the cells within the ROI in figure 3(A) as shown in table 1.

Considering an average dose delivered of 1.0 ± 0.02 Gy and the pulse duration ∆t of about 500 ps obtained from the difference on the arrival times shown in table 1, a dose-rate of 2 ± 0.02 × 10⁹ Gy s⁻¹ can be estimated.

The 53BP1 foci formation assay offers a simple method to study DNA DSB yields in individual cells at doses of order 1–2 Gy. As shown in figure 4(A), the mean number of 53BP1 foci per cell per Gy induced by x-rays at 0.5 h was 25 ± 1 while the value of carbon ions induced 53BP1 foci at this time point was 17 ± 1. At 24 h post-irradiation, significant amounts of x-ray induced DNA DSB damage was repaired as seen through reduction in the 53BP1 foci yield to 3.6 ± 0.4. However, the carbon-ion induced DNA DSB damages were repaired at a slower rate with a value of 11.5 ± 0.4 53BP1 foci per cell per Gy persisting at 24 h post-irradiation.

Zoom In Zoom Out Reset image size

To obtain a better insight of the DNA DSB damage and repair process, we expressed the 53BP1 foci kinetics as percentage repair over the period of 24 h as shown in figure 4(B). Here we normalized the background-subtracted values of mean 53BP1 foci per cell per Gy at various time points to those induced at 30 min for both 225 kVp x-rays or laser-driven carbon ions. Considering the amount of induced damage at 30 min post-irradiation as 100%, the subsequent time points such as 1, 6 and 24 h show the progress of the repair process as seen through reduction in the percentage of unrepaired DNA DSB. E2 cells, due to their radioresistant nature, successfully repaired the low LET x-ray induced DNA DSB damage and at 24 h only about 6% unrepaired DNA DSB damages were observed. The carbon ion -induced complex damage persisted and at 24 h, at least 60% of the unrepaired DNA DSB damages were seen. The repair kinetics of the DNA DSB damage was fitted for both curves with an exponential two-phase decay equation (Groesser et al 2011, Plante et al 2019) using the expression

$$\begin{eqnarray*}&&Y=Plateau+SpanFast* \exp \left(-KFast* X\right)+SpanSlow* \exp \left(-KSlow* X\right),\end{eqnarray*}$$

where Y is the number of foci, Plateau is the number of foci at infinite time

SpanSlow = (Y0-Plateau)*(100-Percent Fast)*.01

SpanFast = (Y0-Plateau)*Percent Fast*.01

Y0 is the foci number at time zero and X is time in hours

K Fast and K Slow are constants accounting for the fast and slow rate of repair, respectively.

Based on the two- phase decay fitting shown in figure 4(B), the value of the x-ray induced 53BP1 foci plateau was negligible compared to the 61% value observed for the laser-driven carbon ions. The percent fast repair for x-rays induced foci was 69% while for the carbon ions this value was 52%. The complexity of the DNA lesions induced by laser driven carbon ions affected the DNA DSB repair kinetics fitting equation parameters and the variation is clearly seen in rate constant ratio as well as in other parameters, as shown in supplementary table 1.

In order to obtain an insight into the spatial features of the DNA DSB damage, we measured the area of both the x-ray and carbon ion induced 53BP1 foci at 24 h post- irradiation. We compared at least 100 53BP1 foci induced in the irradiated cells and clearly observed the impact of LET on the area of the 53BP1 foci. At 24 h post irradiation, the area of carbon-induced foci was 1.60 ± 0.07 μm² while x-ray induced foci were smaller with an area of 0.63 ± 0.02 μm². The difference between the area among the low LET x-rays and the high LET carbon ions was statistically significant with a p < 0.0001. Similar observations in the size variation of x-ray and carbon ions induced foci have been reported by other investigators (Oike et al 2016).

We further compared the effectiveness of laser-accelerated-carbon ions for foci induction relative to x-rays in terms of relative foci induction (RFI) as shown in figure 6. We observed an RFI value of 3.2 ± 0.2, which is similar to reported cell killing RBE of high LET carbon ions.