Radiotherapy (RT) remains the major treatment modality for multiple cancers, yet it still has limitations, including dose-limiting toxicity to healthy tissues and lack of efficacy against hypoxic tumors (Barnett et al 2009, Wang and Tepper 2021, Verginadis et al 2025). New RT modalities are being developed that may overcome several disadvantages of conventional (CONV) RT.

Spatially fractionated RT (SFRT) produces non-uniform dose profiles (PF) with alternating high (‘peaks’) and low (‘valleys’) dose regions. Toxic effects on tumors are maintained and sometimes enhanced compared to standard RT. SFRT is classified in three broad categories based on the dimension of the irradiated patterns. GRID and Lattice RT, with beam width and spacing in the cm range, are used mostly in palliative care (Duriseti et al 2021, Iori et al 2023), whereas microbeam (Slatkin et al 1995) and ‘minibeam’ (Dilmanian et al 2006) RT use finer geometries and are still in development. Microbeam RT (MRT) typically consists of arrays of planes or cylinders (‘pencils’) 25–100 µm in width separated by 50–500 µm valleys. MRT is highly tolerated by normal tissue and effective at controlling tumors in preclinical models (Fernandez-Palomo et al 2020). The approach is, however, difficult to implement in the clinic since it requires ultra-high dose rates of low-energy beams generated by large synchrotrons.

Minibeam RT (MBRT) is easier to translate to cancer patients and—although less studied than other SFRT modalities—seems to retain the desirable effects of MRT, namely immunomodulation and tissue sparing (Dilmanian et al 2006, Prezado 2022). MBRT is also less prone than MRT to blurring from patient’s motions (e.g. thoracic movements from respiration). A typical MBRT geometry consists of planar beams ⩾ 500 µm thick separated by ⩾ 1 mm valleys, which can be generated with conventional x-rays sources (Bazyar et al 2017, Prezado et al 2017). The unique action of MBRT may involve alteration to the vasculature, the bystander effect, and immunomodulation (Moghaddasi et al 2022, Prezado 2022, Lukas et al 2023, Prezado et al 2024). Another advantage of SFRT is its ability to induce abscopal effects, thereby improving RT efficacity via control or elimination of distant tumors. This effect is well known with GRID (Asur et al 2015, Johnsrud et al 2020) and MRT (Fernandez-Palomo et al 2020) and was also observed with MBRT, in particular when combining RT and immune checkpoints inhibition (Rivera et al 2025).

MBRT has shown promising results in preclinical and clinical studies. In rodents, MBRT induced a strong infiltration of T cells and a robust antitumor response (Bazyar et al 2022, Bertho et al 2023a). Treatment of dog brain tumors with planar beams characterized by a full width at half maximum (FWHM) of 1 mm resulted in better tumor control and less brain damage than standard RT (Kundapur et al 2022). Moreover, a phase 1 study is ongoing at the Mayo Clinic, supported by promising results from the first implementation of MBRT in human patients (Grams et al 2024). In some studies, the antitumor effects of MBRT were inferior to those of CONV-RT (Iturri et al 2025). These results need to be interpreted with caution since the CONV doses caused severe damage to normal tissues and would not be translatable to human patients.

A consensus on treatment modalities needs to be reached for clinical translation of MBRT (Zhang et al 2020, Li et al 2024). To this end, dosimetric parameters must maximize antitumor responses. SFRT has a complex parameter space defining its efficacy. Parameters including the overall geometry (planar vs pencil beams), dose distribution (in particular peak and valley doses and their ratio), and dose rate, have not been optimized. While complex multiparametric problems can be addressed efficiently with stochastic mathematical models (Alfonso et al 2021, Bekker et al 2022, 2024), these models require realistic biological parameters that can be derived from preclinical studies.

SFRT’s impact on the immune system occurs within hours after irradiation, as revealed by transcriptomic analyses (Bouchet et al 2013). The DNA damage response, which is likely determinative for the RT outcome, is also extremely rapid, as is the generation of reactive oxygen species by ionizing radiation. It is therefore important to study biological responses to SFRT with a high temporal resolution.

Here, we present a simple platform based on a portable x-ray source for minibeam SFRT combined with live-cell microscopy. The approach enables rapid exploration of dosimetric parameters and exquisite time resolution for live-cell analyses. We used Monte Carlo (MC) simulations of dose distribution and measured early DNA damage responses with different collimation geometries. We anticipate that this approach will yield a better mechanistic understanding of SFRT for parameter optimization.

2.1. X-ray dosimetry and collimation

The radiation source was a miniature kV x-ray tube system (Mini-X, 4 W; Amptek) with an Au target. Dosimetry was performed at the maximum current setting of the source (50 kV, 79 µA). The beam output was determined at 10 cm using the ‘in air’ method of the AAPM TG-61 protocol (Ma et al 2001) and a soft x-ray ionization chamber (PTW 34013) calibrated in a beam of half-value layer (HVL) equivalent to 0.11 mm Al, very similar to our Mini-X beam HVL of 0.107 mm Al (Prajapati et al 2021). The ratio for water-to-air of the mean mass energy-absorption coefficients averaged over the incident photon spectrum and backscatter factor were 1.044 and 1.019, respectively (Ma et al 2001). EBT4 Gafchromic films (Ashland, lot #07062302) were calibrated (RGBs, 2nd degree polynomial curves) at the reference condition, at the surface of RW3 water-equivalent phantom slabs. Dose levels were between 0 and 14.3 Gy, with 9 intermediate calibration points. Film scanning was performed 24 h after irradiation, following 5 blank warm up scans on an Epson 10000XL scanner at 400 dpi and in transmission mode. The RGB channels were extracted, and the multichannel method (Micke et al 2011) was applied using the ImageJ software (https://imagej.nih.gov/ij/; version 1.52 v). Film measurements uncertainty was evaluated at 6.1%, by combining with a quadratic sum the uncertainty of the reference beam output measurement of 3.6% (Ma et al 2001) and specific films uncertainties provided by the AAPM TG-61 report (Niroomand‐rad et al 2020). Brass (CW612N; 8.4 g cc−1) collimators were milled from 1 mm thick disks, 6.4 mm in diameter, by the Usinage du Loir (Seiches sur le Loir, France). Collimator geometries consisted of vertical stripes (planar minibeam SFRT) or arrays of spots (mini-GRIDs), as described in the results section. The different extrusion patterns were verified by bright-field imaging with a 4x objective (supplementary figure S1). EBT4 films were irradiated to define the spatial distribution of dose with the different collimators. The films were placed under a 2 mm thick layer of water, with the tip of the collimator holder positioned just above the water layer, as for cell irradiation. To define peak and valley doses for the planar collimator designs, intensities on EBT4 films were averaged along 2 mm lines positioned on 3 central peaks and on 2 adjacent valleys. For the collimated spots, film intensity was measured from circular regions of interest (ROI; 0.5 mm in diameter). Valleys of doses were measured by averaging intensities along 0.5 mm lines corresponding to the bisector between the central spot and the nearest neighbor. Dose PF were generated using rectangular ROI (0.1 × 7 mm) positioned across the SFRT patterns. Measurements were done in triplicates i.e. for three independent film irradiations.

2.2. RW3 measurements and MC model

GATE10 (Krah et al 2024) MC code was used to create the model of the Mini-X with the QGSP_BERT_LIV physics list. Production cut and maximal step size were fixed at 0.01 mm in targets. Model parameters such as photons energy spectrum, focal spot and beam sizes were adjusted from theoretical data using film measurements as reference. Simulation parameters are given in table 1. Percent depth dose (PDD) and PF were acquired with EBT4 films at several depths in RW3 slabs for an open field of 5.7 mm in diameter, corresponding to the opening of the Mini-X collimator holder. For these measurements, the collimator holder (‘nozzle’) was attached to the source, without SFRT collimator. PF were also measured for a planar collimator (five 0.5 mm wide stripes in steps of 1 mm) in the same conditions. To facilitate the comparison between simulations and measurements, the Mini-X nozzle was placed in contact with the RW3 slabs. Films were scanned as described above with a spatial resolution of 0.127 mm (200 dpi) for the open field and 0.0635 mm (400 dpi) for the SFRT collimator. Scans were converted to dose using the calibration curves obtained previously. Once validated, the model was used in realistic condition to evaluate the dose rate in biological samples considered as water (table 1).

Table 1. GATE10 parameters used for simulations.

  Number ofMean statisticalSimulationVoxel size (mm) X/Y/Zinitial particlesuncertainty [Range]PDD and biological simulations3.0/3.0/0.11092.3% [0.2–5.1]PF open field1.0/0.127/0.11010In field 0.3% [0.2–0.6] Penumbra 0.7% [0.3–1.4] Out of field 6.4% [0.7–20.0]2.3. Integration of the x-ray source on a microscope and targeting

For x-ray irradiation combined with live-cell imaging, Mini-X was positioned on the stage of an inverted fluorescence microscope using a custom aluminum holder, as described (Prajapati et al 2021). The holder maintains the source at a set vertical distance, with the beam orthogonal to the microscope stage. With this setup, the source moves coherently with the microscope stage (and therefore the sample). The stage can be moved to collect images from different regions of the irradiated sample. The cell culture dish was positioned based on the localization of the collimator, visualized in bright field by de-focusing the 4x objective. The exact position of peaks and valleys of dose was determined at the endpoint of live-cell imaging experiments by taking bright field images from a piece of EBT4 film glued below the coverslip. For lower doses (< 5 Gy), we used a similar strategy, with a piece of XR-RV3 Gafchromic film (Ashland) placed in the culture medium, approximately 0.5 mm above the cells. These XR-RV3 films were only used for targeting but not for dosimetry.

2.4. Cell lines and cell culture

Mouse Panc02 pancreatic cancer cells were cultured in DMEM/F12 supplemented with 10% fetal bovine serum (FBS), 1 mM sodium pyruvate, 0.1 mM NEAA (non-essential amino acids; Corning), and penicillin/streptomycin at 37 °C, 5% CO2 in a humidified incubator. In initial MBRT experiments, cells were grown directly on EBT4 films. The films were laser-cut to 18 mm disks. Edges were sealed with nail polish which effectively prevented culture medium from leaking into the active layer (as verified visually), and the films were coated with a solution of Type I collagen (3 mg ml−1 for 1 h; PureCol, Advanced BioMatrix). The films were placed in 12-well culture plates and cells were seeded on top. For subsequent MBRT tests, cells were seeded on 18 mm coverslips (150 000 cells/well). For clonogenic assays, cells were seeded in 12-well plates (50 000 cells/well). For live-cell imaging experiments, human U2OS osteosarcoma cells stably expressing mCherry fused to the c-terminus of 53BP1 (mCh-53BP1ct) (Locatelli et al 2022) were cultured in DMEM/F12 with 10% FBS. The cells were seeded in 35 mm glass-bottom dishes (MatTek; 50 000–100 000 cells/dish) and cultured for 24 h until irradiation. In all experiments, the volume of culture medium was adjusted before irradiation to a constant height of 2 mm above the cells, by calculating the necessary volume of medium based on the geometry of the cell culture device. To assess dose attenuation, U2OS_mCh-53BP1ct cells were mixed in 150 µl of collagen I (6 mg ml −1 final collagen concentration, Advanced BioMatrix) at a dilution of 300 000 cells ml −1 and seeded in 35 mm glass-bottom dishes to form a cylinder 14 mm in diameter and 1 mm in height. Cultures were regularly tested for mycoplasma contamination; all tests were negative.

2.5. Immunofluorescence

Cells were fixed in 10% formalin, permeabilized with TX-100, washed in PBS-glycine, and incubated 2 h in blocking buffer (10% goat serum in immunofluorescence buffer [IF; 130 mM NaCl, 13.2 mM Na2HPO4, 3.5 mM NaH2PO4, 0.1% bovine serum albumin, 0.05% NaN3, 0.2% Triton X-100, and 0.05% Tween 20]). Antibodies were diluted in blocking buffer and incubated on samples overnight at 4 °C. Samples were washed three times with IF buffer, incubated with fluorescently labeled secondary antibodies (1 h at room temperature), washed again with IF and stained with 4′,6-Diamidino-2-phenylindole dihydrochloride (DAPI; Invitrogen, 0.5 µg ml−1 ; 10 min). Samples were mounted with coverslips using ProLong Gold Antifade reagent (Invitrogen). Antibodies were against 53BP1 (Abcam, cat# Ab36823; 5 µg ml−1), γH2AX (Millipore, clone JBW301, cat# 05–636; 2 µg ml−1), and P-p53 Ser 15 (Cell signaling, cat# 16G8; 1.9 µg ml−1). Secondary antibodies conjugated with Alexa Fluor dyes (AF488, AF568, or AF647; ThermoFisher) were used at 1:500 dilutions.

2.6. Clonogenic assay

Irradiated cells were left to recover for 24 h at 37 °C, then detached with 0.25% trypsin-EDTA and seeded in 10 cm dishes (1000 cells/dish). Seven days later, cells were fixed for 20 min with 10% formalin then washed with PBS. Colonies were stained with 0.4% crystal violet (20 min), washed with water to remove excess coloration and left to dry. Images were taken with a Fusion Solo S imager (Vilber). Colony count was performed using FIJI (https://fiji.sc). Images were cropped to remove colonies on the rim of the plate. The 8-bit images were binarized, using a threshold value of 130 and the Adjustable Watershed function was applied with a tolerance of 0.9 to separate touching colonies. Colonies were counted with the analyze Particle function, using a size range adapted to the median colony size and a lower boundary for circularity of 0.05. Plating efficiencies and surviving fractions were calculated as described (Franken et al 2006). Assays were done in triplicates for each dose, from three successive cell passages.

2.7. Microscopy

Images of fixed cells were taken with an inverted IX83 Olympus microscope equipped with a qCMOS camera (ORCA-Quest, Hamamatsu), using a 40x air objective (N.A. = 0.95). Overview images were taken at 4x or 10x magnification, using the multiple image alignment function of the CellSens acquisition software for tiling. Excitation light was from an X-Cite TURBO illumination system (Excelitas). The 385 nm LED was used for DAPI, the 475 nm LED for AF488, the 575 nm LED for AF568, and the 630 nm LED for AF647. Emitted light was filtered using 460/50 nm (DAPI), 525/50 nm (AF488), 600/37 nm (AF568), and 700/75 nm (AF647) filters. A 20x air objective (N.A. = 0.7) was used for live-cell imaging of DNA repair foci (mCherry-53BP1ct) and a 4x objective (N.A. = 0.13) was used to follow cell counts, divisions, and motion in time-lapse experiments. 20x magnification images were taken every 5 min over 30 min whereas 4x images were recorded with 30 min intervals over 24 h. Cells were maintained at 37 °C and in 5% CO2 atmosphere with a stage-top incubator (TokaiHit). A custom incubator lid with a 1.5 cm central hole accommodated the nozzle of the Mini-X source. Laser-based drift correction (IX3-ZDC2) was engaged to maintain the focus during time-lapse acquisitions. To assess dose attenuation, the top and bottom layers of a cylinder of cells in collagen was imaged at 40x magnification. The sample was overlaid with a coverslip post fixation, resulting in a coverslip/cell-collagen/coverslip sandwich, which was flipped to take top and bottom images with identical optical parameters.

2.8. Image analysis

Automated foci count in fixed cell images was performed using a custom macro in FIJI (https://fiji.sc), as described (Gaber et al 2025). Briefly, cell nuclei were segmented with DAPI signals using CellPose3 (Stringer and Pachitariu 2025) and DNA damage foci were detected with a Laplacian of Gaussian filter. A semi-automated approach was used for the double-strand break (DSB) foci analyses in live imaging movies. Individual cell nuclei were cropped, intensity was normalized using the Enhance Contrast function of FIJI, and images were smoothed using Gaussian blur (sigma = 1). Foci were identified using the find maxima function and enumerated across the image stack (i.e. for each time point). Cell motions were analyzed using the Mosaic Particle Tracking plugin in FIJI. Fold change in cell number was determined after segmenting cell nuclei (Otsu) or cytosolic GFP signals (CellPose3). Frequency distributions were calculated, to retrieve cell numbers relative to the positions of the minibeams (using 0.3 mm bins). Fold change in cell number was calculated by dividing the distribution at 24 h by the distribution from the first frame (pre-irradiation). Mitotic events were identified visually, recording cell position and time post-irradiation (according to frame number).

2.9. Statistics

Statistical analyses were performed using Prism 10 (GraphPad). The D’Agostino & Pearson omnibus normality test was used to test for normality. Nonparametric tests were used if the data did not pass the normality test (at alpha = 0.05). Statistical tests are indicated in figure legends. All statistical tests were two-sided. A P value < 0.05 was considered significant.

3.1. Characterization of the x-ray source

We described previously the integration of a portable x-ray source (Mini-X) to an inverted widefield fluorescence microscope (Prajapati et al 2021). At the reference condition (see Methods), we obtained a dose rate to water of 1.78 Gy min−1 at a water phantom surface, 10 cm from the source. To irradiate cells, Mini-X is positioned on the stage of the microscope, its beam orthogonal to the cell growth surface (figures 1(A) and (B)). The tip of the collimator holder is positioned just above the culture medium, the volume of which was adjusted to a constant 2 mm layer. The distance between the source target and the coverslip with cell samples is ∼19.4 mm. In these experimental conditions, a dose rate of 22.5 Gy min−1 was evaluated at the cell layer with the MC model.

Figure 1. Characterization and modeling of the miniature x-ray source. (A) Schematic of the source (Mini-X) mounted on an inverted fluorescence microscope. (B) Picture of the system. (C) Setup for PDD and PF measurements. (D) Comparison between film measurements (red) and GATE10 simulations (green) in RW3 for open field PDD (left) and PF at five depths (right). Absolute discrepancies are provided as supplementary data (figure S6). (E) Schematic of the approach to assess dose attenuation. (F) DNA double-strand breaks detected with a DSB sensor (mCherry-53BP1ct) in U2OS cells located at the top and bottom of a collagen gel. Cells were irradiated (30 s) and left 1 h at 37 °C to recover before fixation and imaging. Foci counts are shown in the graph. *, P< 0.05 (unpaired t-test; N = 50).

Download figure:

Standard image High-resolution image 3.2. Modeling of the source

A MC approach was used to model dose deposition of the uncollimated source. An absolute conversion factor was calculated simulating the beam output reference conditions. This factor was applied to convert each simulation to Gy min−1. After adjustment of the model parameters (photons energy spectrum, focal spot size, and beam size), simulated PDD and PF were in good agreement with film measurements in RW3 slabs (as illustrated in figure 1(C)), with simulated values within measurement uncertainty (figure 1(D)).

The model predicts a 33% dose attenuation (from 11.3 Gy to 7.6 Gy; statistical uncertainty of 0.6%) within a 1 mm thick biological sample below a 2 mm layer of culture medium. To assess dose attenuation in situ, cells expressing a fluorescent sensor of DNA DSBs were embedded in collagen and cast into a 1 mm high cylinder, under a 2 mm layer of medium. Cells were irradiated and fixed after 1 h recovery—which corresponds to the maximal response amplitude for this repair factor—and DSB foci were enumerated at the top and bottom of the gel (figure 1(E)). There was a 26% decrease in the median number of DSBs at the bottom compared to the top of the collagen layer, which goes along model predictions for absorbed doses. We note that the yield of 53BP1 repair foci (i.e. foci/Gy/nucleus) decreases with increasing doses (Neumaier et al 2012), due (at least in part) to the coalescence of 53BP1 foci (Kilic et al 2019). As such, we expect to underestimate dose attenuation with this biological measure. This level of attenuation is acceptable for studies with cell monolayers and 3D models including spheroids, tissue sections, and skin. It also represents a physiological fall-off in dose within 1–5 cm of tumor and normal tissue.

3.3. X-ray collimation

Collimators were designed to produce planar minibeams in the sub-millimeter range, as done in previous preclinical studies (Prezado et al 2024), and as implemented in an ongoing clinical trial (Grams et al 2024) (figure 2(A)). Vertical slits of 1, 0.5 and 0.3 mm in width were spaced 2, 1, and 0.6 mm on center, respectively, to keep a (theoretically) constant proportion of 50% irradiated area. Two additional ‘rosette’ designs were chosen, based on equilateral lattices of 1 mm spots. Spot centers were spaced 1.3 mm for 50% surface irradiation, and 2 mm for ∼30% surface irradiation. These designs, referred to as mini-GRIDs hereafter, were inspired by SFRT modeling work by Bekker et al (Bekker et al 2024). The collimators were non-divergent, as done clinically (Grams et al 2024). Slit length and spot numbers were maximized to fit into the collimator disks. The diameter of the collimators (6.4 mm) is equal to the inner diameter of the collimator-holding ‘nozzle’ of Mini-X. The collimators are positioned at the bottom of the holder, as shown in figure 2(B), with the surface of the collimator ∼2 mm above the cells.

Figure 2. Collimation of the Mini-X source. (A) Photograph of the brass collimators. The proportion of extruded surface is indicated. (B) Schematic showing positioning of the collimators. (C) PF for planar SFRT (5-stripes collimator) measured with EBT4 films at different depths in RW3. (D) Images of EBT4 films irradiated for the indicated amount of time with the different collimators. Intensities correspond to dose (Gy) and are displayed as a heatmap. (E) Averaged dose profiles of the planar minibeams and the spot mini-GRIDs (30 s irradiation; mean ± SD, N = 3).

Download figure:

Standard image High-resolution image

We first measured the PF for the 0.5 mm planar beam geometry (5-stripes collimator) at various depth, using the setup illustrated in figure 1(C). As shown in figures 2(C), a clear separation of peaks and valleys was maintained over a depth of 7.5 mm. The peak-to-valley dose ratios (PVDRs) were 15.8, 11.8, 10.9, 8.3 and 6.4 at depths of 1.4, 2.7, 4.0, 5.3 and 7.5 mm, respectively (combined uncertainty of 8.6%). Importantly, PF measurements indicate that collimation for MBRT is possible with the source in conditions compatible with preclinical in situ cancer models.

Next, we sought to characterize the different MBRT geometries. Film irradiation is illustrated in figure 2(D). Radiochromic films, irradiated in culture plates under a 2 mm thick layer of water (reproducing conditions used to irradiate cells), were used to calculate peaks and valleys of dose, FWHM, center-to-center distance (table 2), and to generate dose PF, displayed in figure 2(E). For the 1 mm and 0.5 mm planar beams, high PVDRs were obtained (22.7 and 12.4, respectively). Hence, both 1 mm and 0.5 mm planar beam designs enable delivery of clinically relevant doses in peak regions and much lower doses in the valley regions. The PVDR for the half-millimeter planar minibeam collimator is very similar to the one obtained for a similar collimator geometry at the surface of the target for the two first patients treated with MBRT (Grams et al 2024). The thinnest minibeam design (0.3 mm in width) produced valleys with greater dose than the larger beam collimators, with only a ∼2-fold difference in dose between peaks and valleys. This penumbra effect is likely due to the narrow band spacing. Additionally, peak doses decreased with decreasing band width, reflecting lower volumes of scattering medium and lower output factors for narrower collimators. To characterize spot mini-GRIDs, intensity measurements were taken across spot centers. Doses were highly contrasted between peaks and valleys (figure 2(E)), with PVDR of 23.4 for the 30/70 design and of 8.2 for the 50/50 design (table 3). The fact that the collimators are directly attached to the tube ensures their correct alignment with the beam and the reproducibility of this alignment.

Table 2. Dosimetry of the planar minibeams.

Collimator width (mm)Peak (Gy)aValley (Gy)aPVDRbFWHM (mm)CTC (mm)115.7 ± 0.70.7 ± 0.0322.7 ± 1.41.10 ± 0.072.34 ± 0.000.513.7 ± 0.31.1 ± 0.112.4 ± 1.20.61 ± 0.051.19 ± 0.040.38.8 ± 0.13.8 ± 0.12.3 ± 0.10.49 ± 0.030.71 ± 0.03

aDoses measured after 30 s irradiation. bFilm measurement uncertainty is 6.1% and combined uncertainty for PVDR is 8.6%. CTC, center-to-center distance.

Table 3. Dosimetry of the spot mini-GRIDs.

Spot distance (mm)Peak (Gy)aValley (Gy)aPVDRbFWHM (mm)cCTC (mm)211.7 ± 0.20.5 ± 0.0323.4 ± 1.51.38 ± 0.312.49 ± 0.021.313.9 ± 0.51.7 ± 0.18.2 ± 0.61.14 ± 0.131.60 ± 0.04

aDoses measured after 30 s irradiation. bFilm measurement uncertainty is 6.1% and combined uncertainty for PVDR is 8.6%. cx/y averages.CTC, center-to-center distance.

3.4. DNA damage responses with different SFRT geometries

As a first approach to relate spatial dose fractionation to biological effects, pancreatic cancer cells (Panc02) were seeded directly on collagen-coated EBT4 films, irradiated using the 7-spot ‘rosette’ mini-GRID collimator, and stained for phosphorylated H2AX (γH2AX), a DNA DSB marker. In these experiments, far red fluorescence was used to detect immunostaining signal. EBT4 films emit very little autofluorescence when illuminated with 630 nm light, and this excitation wavelength does not affect the active layer of the film (supplementary figure S2). It is therefore possible to image cells stained with far red dyes on radiochromic films to directly relate biological effects to dose distribution. As shown in figure 3, γH2AX signals detected 1 h after irradiation matched the ‘rosette’ pattern on the EBT4 film. Hence our MBRT model produces a spatially defined biological response and γH2AX staining is an efficient way to retrospectively identify SFRT patterns.

Figure 3. Spatially fractionated irradiation of cells on radiochromic films. (A) Schematic of the approach. (B) Overview image of an EBT4 film used for cell culture, after collimated irradiation with a 5-spot mini-GRID. (C) Low magnification (4x) images of the film/culture substrate (in bright field) and Panc02 cells stained for γH2AX 1 h after irradiation (∼7 Gy peak dose). (B) γH2AX staining at higher magnification (20x) in regions corresponding to a peak and a valley of dose (zones 1 and 2, respectively, shown in (B)). Images of single cell nuclei are displayed at the bottom.

Download figure:

Standard image High-resolution image

To characterize cellular responses to the different minibeam/GRID geometries, Panc02 cells were irradiated, adjusting irradiation times to reach a peak dose of 10 Gy with all geometries. This dose was chosen for this radioresistant cell line as it caused a ∼50% decrease in viability following whole-field irradiation (supplementary figure S3). Cells exposed to whole-field irradiation and non-irradiated cells (i.e. distal from the irradiated zones) were used as controls. To evaluate the early response to DNA damage, we examined p53 phosphorylation at serine 15 (P-p53), γH2AX, and the accumulation of 53BP1 DNA repair foci, 1 h post-irradiation. The results show a clear patterned induction of P-p53 (figure 4(A)) and γH2AX (figure 4(B)) mirroring the dose PF generated by the different collimators. In all MBRT geometries tested, P-p53 and γH2AX intensities were significantly higher in peaks compared to valleys (P< 0.0001; 2way ANOVA and Šidák’s multiple comparison test) (figures 4(C) and (D)). In higher magnification images, γH2AX damage foci were clearly visible and more numerous in peaks compared to valleys (figure 5(A)). As expected, there was a good (albeit not perfect) overlap between γH2AX and 53BP1 foci: both factors accumulate at DSBs, but their foci formation and resolution kinetics are different. γH2AX rapidly spreads on chromatin flanking DSBs, while 53BP1 accumulates later and persists longer at break sites. Moreover, 53BP1 forms molecular condensates at DSBs (Kilic et al 2019) that are very distinct and can be precisely quantified. Cells in peak regions accumulated significantly more 53BP1 foci than cells in valleys of dose (P < 0.0001; 2way ANOVA and Šidák) (figure 5(B)).

Figure 4. DNA damage response to minibeam and mini-GRID SFRT. (A) and (B) Phosphorylation of p53 (ser15) (A) and γH2AX (B), detected by immunostaining in Panc02 cells 1 h after irradiation (10 Gy peak dose). Image tiles were taken at 10x magnification. The graphs show mean nuclear intensities across SFRT patterns (red line in schematics). Each dot represents a cell. (C) Enlarged images of P-p53 and γH2AX staining of cells irradiated with minibeams (5-stripes geometry), in between peaks (zone 1) and distal from irradiation (zone 2). The corresponding zones are indicated in the overview images (A) and (B). (D) Quantification of P-p53 and γH2AX signal intensities in peaks and valleys of dose, and in non-irradiated regions. Open field irradiation (CONV) is shown as control. The data shown are representative of two independent experiments.

Download figure:

Standard image High-resolution image

Figure 5. MBRT-induced DNA double-strand breaks. (A) γH2AX and 53BP1 detected by immunostaining in Panc02 cells 1 h after irradiation (7–10 Gy peak dose). Images are from peak and valley regions from the different minibeam SFRT geometries. (B) Quantification of 53BP1 damage foci. Each symbol represents the mean from an experiment (N = 3), with > 200 cells analyzed per experiment and condition.

Download figure:

Standard image High-resolution image 3.5. Live-cell analyses of MBRT responses

To follow DNA damage induction in peaks and valleys of dose in real time, we used osteosarcoma cells (U2OS) stably expressing the fluorescent mCherry-53BP1ct DSB sensor for live imaging (figure 6(A)). A region of the cell culture dish located underneath the collimator was selected for time-lapse imaging. Cells were irradiated after collection of the initial frames. Precise localization of dose peaks and valleys was based on bright field images of a radiochromic film (figures 6(A-C)). Cells in peak regions accumulated clearly more foci than cells in valley regions (figure 6(D)) and this effect was apparent already 5 min post irradiation (figures 6(E) and (F); supplementary movies 1–2). In peak regions, large pre-existing 53BP1 foci—also known as 53BP1 nuclear bodies (Lukas et al 2011)—dissolved, giving way to a large number of smaller DSB foci. In contrast, in dose valleys most 53BP1 nuclear bodies remained relatively unchanged in size and shape. In control experiments where cells were not irradiated, DSB levels remained constant (supplementary figure S4), ruling out DSB induction via photodamage from our imaging conditions.

Figure 6. Live-cell imaging of DNA double-strand breaks induced by MBRT. (A) Targeting strategy for live-cell minibeam SFRT imaging. The region of interest (ROI) for imaging is selected by direct observation of the collimator in bright field. Movies including SFRT irradiation (IR) are recorded. At the endpoint, a bright field image of the radiochromic film (glued below the coverslip or placed directly above the cells) is taken and registered with cell images to define IR dose profiles. (B) Image of the film and dose intensity profile (portion of 5-stripes MBRT). Dashed lines correspond to dose peaks. Peak doses were 2.5 Gy. (C) Overview of U2OS cells expressing the mCherry-53BP1ct DSB sensor. The fluorescence image was inverted for visualization. (D) Quantification of mCherry-53BP1ct foci before irradiation (Pre-IR) and 30 min post-irradiation. Each dot represents a cell (N = 280). (E) Heat map showing the evolution of DSB foci counts as a function of time and distance across the SFRT pattern. (F) Representative images of mCherry-53BP1ct signals in U2OS cells located in a peak and valley of dose (zones 1 and 2 shown in C).

Download figure:

Standard image High-resolution image

Next, we examined cell motions using low magnification time-lapse images taken across the peak/valley profile over a 24 h period (figure 7(A)). Migration distances were constant across the SFRT region (figures 7(b) and (F)), indicating that the peak dose had no major impact on cell mobility. We do not exclude transient effects on cell migration, which may not have been captured.

Figure 7. Effect of MBRT on cell motions and proliferation. (A) Overview of U2OS cells before and 24 h after planar minibeam irradiation (5-stripes geometry; 10 Gy in peaks). Fluorescent signals correspond to mCherry-53BP1ct. Images were inverted to visualize cell nuclei. (B)–(D) Cell-based readouts relative to the distance along the minibeams peak/valley profile: cell motions ((B); travel distance in µm), fold change (FC) in cell number (C), and time/position of mitotic events (D). Cell numbers are averages for bin distances of 0.3 mm. The black dotted line in the graphs corresponds to the position of the central peak of dose. The positions of neighboring dose peaks (gray dotted lines) are based on the known inter-peak distance. (E) Image of the EBT4 film (located under the coverslip; see figure 6(A)) and corresponding dose intensity profile. (F) Illustration of cell motion traces in regions corresponding to a peak and a valley of dose. (G) Illustration of cell division events. Cell position and time post-irradiation (last frame before telophase/cytokinesis; yellow arrowheads) was recorded.

Download figure:

Standard image High-resolution image

To assess the effect of minibeam SFRT on cell proliferation, we compared cell densities before and 24 h after irradiation