Minibeam radiotherapy with small animal irradiators; in vitro and in vivo feasibility studies

Minibeam radiation therapy (MBRT) delivers an ultrahigh dose of x-ray (⩾100 Gy) in 200–1000 µm beams (peaks), separated by wider non-irradiated regions (valleys) usually as a single temporal fraction. Preclinical studies performed at synchrotron facilities revealed that MBRT is able to ablate tumors while maintaining normal tissue integrity. The main purpose of the present study was to develop an efficient and accessible method to perform MBRT using a conventional x-ray irradiator. We then tested this new method both in vitro and in vivo. Using commercially available lead ribbon and polyethylene sheets, we constructed a collimator that converted the cone beam of an industrial irradiator to 44 identical beams (collimator size  ≈  4  ×  10 cm). The dosimetry characteristics of the generated beams were evaluated using two different radiochromic films (beam FWHM  =  246  ±  32 µm; center-to-center  =  926  ±  23 µm; peak-to-valley dose ratio  =  24.35  ±  2.10; collimator relative output factor  =  0.84  ±  0.04). Clonogenic assays demonstrated the ability of our method to induce radiobiological cell death in two radioresistant murine tumor cell lines (TRP  =  glioblastoma; B16-F10  =  melanoma). A radiobiological equivalent dose (RBE) was calculated by evaluating the acute skin response to graded doses of MBRT and conventional radiotherapy (CRT). Normal mouse skin demonstrated resistance to doses up to 150 Gy on peak. MBRT significantly extended the survival of mice with flank melanoma tumors compared to CRT when RBE were applied (overall p  <  0.001). Loss of spatial resolution deep in the tissue has been a major concern. The beams generated using our collimator maintained their resolution in vivo (mouse brain tissue) and up to 10 cm deep in the radiochromic film. In conclusion, the initial dosimetric, in vitro and in vivo evaluations confirmed the utility of this affordable and easy-to-replicate minibeam collimator for future preclinical studies.


Introduction
Normal tissue toxicity is the dominant dose-limiting side effect of radiotherapy (RT) (Hendry et al 2006). Applying therapeutic doses of ionizing radiation to radiosensitive tissues such as in the brain almost always produces a certain level of radiation side effect. This effect not only occurs after conventional RT (CRT), but also following intensity-modulated radiotherapy (IMRT) and proton therapy (Armoogum and Thorp 2015). In contrast, spatially fractionated methods of RT have shown an increased preservation of healthy tissue (Dilmanian et al 2003, Dilmanian et al 2007, Yuan et al 2015, Nolan et al 2017. Relative to the homogeneous dose delivery in CRT, spatially fractionated modalities apply physically separated doses of radiation to the target. Of the types of spatially fractionated techniques, microbeam radiotherapy (MRT), which delivers quasi-parallel lines, less than 100 µm wide, of a single high dose (hundreds Gys) of irradiation (peaks), separated by wider non-irradiated regions (valleys), is a promising new approach (figure 1). Preclinical studies have consistently demonstrated the selective tumoricidal and normal tissue sparing effects of this method , Smyth et al 2016. This suggests that a potential advantage of MRT is not only reduced radiation-related normal tissue toxicity, but also improved tumor control rates. This property would be of special benefit to pediatric and previously irradiated recurrent tumor patients, where radiation dose considerations are paramount. Ultra-high intensity parallel x-ray photons produced by synchrotron sources have the ideal characteristics for generating thousands Gys micro-beams. Unfortunately, such synchrotron sources are rare. At present, only three active synchrotron facilities in the world are running preclinical MRT studies. Aside from the difficulty in accessing these facilities, there is also a lack of specialized hospitals near them. Thus, MRT has not yet been clinically applied, mainly due to a lack of sufficient preclinical data.
Recently, there have been some efforts to apply this promising method using non-synchrotron irradiators (Hong et al 2015, Bartzsch et al 2016. Nevertheless, the radiation profiles of these approaches suffered from the non-uniformity of the beams and valleys, and also the variability of the peak-to-valley dose ratios (PVDR). These are the critical parameters in the normal tissue sparing and therapeutic efficacy of the MRT . In addition, there are multiple clinical limitations for microbeams as described before by Dilmanian et al, namely, loss of beam spatial resolution in vivo as a result of cardiorespiratory induced motion in the target organs (Dilmanian et al 2006). Therefore, slightly wider 'minibeams' (200 µm ⩽ FWHM < 1 mm) have been proposed as a promising clinical future for this technique. Synchrotron-generated minibeams have also exhibited a normal tissue sparing effect for beams up to 0.68 mm in FWHM (Dilmanian et al 2006, Prezado et al 2015, Deman et al 2012. Babcock et al designed a collimator to mount near an industrial source to create minibeams (Babcock et al 2011). Although their method demonstrated the feasibility of utilizing a conventional x-ray tube to implement MBRT, their resultant beams were 1 mm wide. A recent study by Brönnimann et al revealed that beams ⩾400 µm may induce an unfavorable microvascular response in normal tissue (Brönnimann et al 2016).
Our group developed the first desktop device for applying 300 µm wide beams using a multi-beam source, based on a field emission carbon-nanotube x-ray tube (Hadsell et al 2013). Our first-generation device was able to apply an image-guided, physiologically gated beam with an entrance dose rate of 21.7 mGy s −1 , and in a long run behavioral study, we found that the generated beams using our device were able to spare normal mouse brain tissue (Chtcheprov et al 2014, Zhang et al 2014. Since our first-generation device was not ready to generate an ultrahigh dose of hundreds Gy used in MBRT studies in a reasonable time frame, we sought to investigate the possibility of applying MBRT using a conventional irradiator and a simple and inexpensive collimator, to facilitate preclinical studies on MBRT. We found that this collimator can convert the cone beam of a small animal irradiator to 44 identical beams (collimator size ≈ 4 cm × 10 cm). The dosimetry characteristics of the generated beams were investigated by the simultaneous use of two radiochromic films with different dose sensitivity ranges. Clonogenic assays revealed the induction of radiobiological cell death in two murine cell lines and initial in vivo studies demonstrated normal tissue sparing and the therapeutic advantage of our method relative to CRT. Although the loss of beam spatial resolution deep in the tissue is one of the major concerns in utilizing divergent industrial irradiators produced photons, the beams generated using our collimator kept their resolution in vivo and up to 10 cm deep in the radiochromic film. Here we demonstrate the properties of the collimator and the effectiveness of the device both in vitro and in vivo animal studies.

Irradiation
An industrial small animal irradiator (X-RAD 320, PXi, North Branford, CT) was utilized as our radiation source. The tube specifications are presented in table 1 and compared to a clinical orthovoltage tube. This irradiator incorporates an oil-cooled anode, which enables running the device for multiple hours to generate hundreds Gy doses for the MBRT peaks. It also has an integrated plane parallel transmission chamber (PTW 7862, PTW, Freiburg, Germany), which can be cross-calibrated to an ionization chamber (we used MDH 1015, Radcal, Monrovia, CA; the sensitive area of detector ≈ 1 cm × 2m) at the desired focus-surface distance (FSD), to measure the dose rate and the total dose on-time. For all experiments, the tube was driven at 160 kVp and 25 mA to match our prior setting (Hadsell et al 2013). The beam was filtered with an additional 2 mm Al and the target was placed at 37 cm FSD.

Collimation
The minibeams were produced by a custom multi-slit collimator placed in close contact with the target (proximal collimation). The collimator consists of 5 mm thick lead ribbons, which block 99.999% of the primary photons in the highest energy spectrum range. To develop the collimator, a 0.6 mm thick, 5 mm wide lead ribbon was cut into 10 cm long pieces. A sandwich of 46 lead pieces was made by alternating 300 µm thick polyethylene sheets as spacers (figure 2(B)). The resulting collimator was 4 cm wide and 10 cm long (figure 2(A)).

Dosimetry
SpekCalc version Pro was used to simulate the x-ray spectra (Poludniowski 2007, Poludniowski and Evans 2007, Poludniowski et al 2009. GAFCHROMIC ™ MD-V3 (peak) and EBT3 (valley) films (Ashland Advanced Materials, Bridgewater, NJ, US) were utilized for dosimetry and evaluating dose profiles. The key technical features of GAFCHROMIC ™ films that make them suitable for this purpose include the minimal response difference over a wide photon energy range, and high spatial resolution (25 µm or higher) (GAFCHROMIC ™ Dosimetry Media, Type EBT-3; G GAFCHROMIC ® MD-V2 Dosimetry Film). Consequently, several MRT and MBRT studies have used them for dosimetry evaluations (Crosbie et al 2008, Anderson et al 2010, Hadsell et al 2013b, Hong et al 2015. The films were cross-calibrated to the integrated plane parallel transmission chamber without any collimator in place and scanned 24 h later with 1200 dpi resolution (spatial resolution ≈ 20 µm) as previously described ( A custom dose phantom was also created to evaluate the dose depth effects of the system. Ten layers of 4 cm × 4 cm PMMA, with a thickness of 2 mm each, were sandwiched with 11 layers of GAFCHROMIC ™ films of the same area to construct a phantom, as shown in the supplementary data figure 1 (stacks.iop.org/PMB/62/8924/mmedia). Data from this phantom were used to calculate the percentage dose drop (PDD) with and without the collimator, peakto-valley-dose ratio (PVDR) and beam-width (full width half maximum = FWHM) of beams generated using the collimator. Since kilovolt energy photons were used in our experiments, the reference point (MU) was defined at the phantom surface. The field size (4 cm × 4 cm) was matched to the clinical orthovoltage irradiator applicator at FSD = 30 cm (Xstrahl 200 and 300 2017). The aluminum first half-value-layer (HVL Al ) was evaluated following AAPM's TG-61 for kilovoltage x-ray beam dosimetry protocol setup (table 1 and   In vitro dose responses of B16-F10 and TRP cell lines were evaluated using clonogenic assays with delayed plating after treatment protocol as shown in figure 3 (Franken et al 2006). In short, an appropriate number of cells were seeded in 12.5 mm 2 cell culture flasks, grown to ≈90% confluency, and irradiated with different doses of MBRT. The flasks were filled with the complete media and placed upside down so that the collimator could be placed in close contact with the growth surface. To mimic the subcutaneous tumor dose, a 2 mm thick PMMA sheet was placed between the flask and the collimator. The control flasks were placed outside of the incubator inverted for an equal duration of time, to account for the effect of cell death due to detachment (anoikis) or an unfavorable environment (low temperature, humidity and pH). Six hours following irradiation, a single cell suspension was obtained and an appropriate number of cells was counted and seeded in each well of a 6-well plate. Colonies were fixed with 70% ethanol and stained with crystal violet two weeks later. After scanning, the colonies were counted using ImageJ (NIH, Bethesda, MD). The minimal pixel size of the particles was defined by measuring the size of a colony consisting of 25 cells for each cell line. The surviving fraction was calculated by the following equation: Plating efficiency (PE) = Number of Counted Colony Number of Platted Cells × 100 Surviving fraction (SF) = Irradiated Sample PE Control Sample PE .

In vivo feasibility
Five-week old female C57BL/6J mice were acquired from Jackson Laboratory (Bar Harbor, ME) and allowed to acclimate for a week before study initiation. The mice were housed in the UNC-CH Division of Laboratory Animal Medicine pathogen free designated environment and cared for in accordance with the United States Department of Health and Human Services Guide for the Care and Use of Laboratory Animals; all procedures were approved by UNC-CH Institutional Animal Care and Use Committee.
The mice underwent irradiation under anesthesia with 1%-2.5% isoflurane in medicalgrade oxygen at a flow rate of 0.8-1 L min −1 . Except for the irradiation field, the rest of the animal body was shielded with 1 cm thick lead (figure 4). The anesthetized mice were laid prone on an in-house designed mouse-holder and the head was fixed using ear bars and nose cone and their right thigh was fixed on the designated holder using medical tape.
To evaluate the effect of our method on normal skin, the right hind limbs of mice were shaved and irradiated with graded doses of CRT or MBRT (5 mice per group) and observed three times a week for the appearance of acute skin effect up to one month post-irradiation. The contralateral legs served as the control. Upon appearance, the skin effect was scored based on scoring system, illustrated in table 2 (Pommier et al 2004).
To investigate the spatial resolution of the beam in vivo, the mid and posterior part of a mouse brain was irradiated using the collimator with 100 Gy beams. Brain tissue from the animal was collected 6 h post-irradiation.
The mouse model of melanoma was prepared by subcutaneous injection of 1 × 10 5 B16-F10 cells in the right hind limb of the mice. The cells were prepared and injected based on the Overwijk et al protocol (Overwijk and Restifo 2001). One week later, mice were randomly assigned to three groups, and mice in CRT and MBRT groups received radiation on the inoculation site (MBRT = 150 Gy on peaks, CRT = 15 Gy). The tumor diameters were measured every other day and tumor volume was calculated using the formula L × W 2 × 0.52, where L is the longest dimension and W is the perpendicular dimension in mm. The mice were humanely sacrificed when the tumor burden reached the protocol specified volume of 1.2 cm 3 . The data was analyzed by SPSS (Ver. 22, IBM Corp., Armonk, NY). The P-value was calculated using a log-rank test.

Immunohistochemistry
A whole mouse brain was fixed in formalin for 48 h, processed, embedded in paraffin and serially sectioned at 5 µm thickness. Sectioned slides were used for γ-H2AX For all experiments, the anesthetized mouse was fixed on an in-house mouse holder and all of the body (except the irradiation field) was covered with 1 cm of lead. (A) CAD drawing of the irradiation setting. Note that the collimator is not shown all the way to the left to enable one to see the underneath shielding and mouse holder; (B) the radiation field was 1.5 cm × 1.5 cm to cover the entire mouse thigh; (C) two pieces of Gafchromic ® MD-V2 films were placed at the entrance and exit plans for dosimetry purposes.
(Double-DNA-Strand-Break marker) immunofluorescence (IF) staining immediately. If staining was not done immediately, the unstained slides were stored in a nitrogen gas chamber. A rabbit monoclonal anti-phospho-ser 139-H2AX antibody was obtained from Cell Signaling Technology (Cat# 9718, Danvers, MA). IF was carried in the Bond Autostainer (Leica Microsystems Inc., Norwell, MA). All solutions were from Leica Microsystems (Norwell, MA). Slides were deparaffinized in Bond dewax solution (AR9222) and hydrated in Bond wash solution (AR9590), antigen retrieval of γ-H2AX was performed for 20 min at 100 °C in Bond-epitope retrieval solution 1 pH-6.0 (AR9961), then a protein blocking reagent (PV6122, Leica) was added for 10 min. After pretreatment, the slides were incubated for 2 h with γ-H2AX (1:2000). Detection was performed using the BondT Polymer Refine Detection system (DS9800) and Tyramide-Cy5 reagent (Perkin Elmer, SAT705A001EA). Slides were counterstained with Hoechst 33258 (H3569, Invitrogen, Carlsbad, CA) and mounted with a ProLong Gold antifade reagent (P36934, Life Technologies). Positive and negative controls (no primary antibody) were included. High-resolution acquisitions of IF slides in the DAPI and Cy5 channels were performed in the Aperio ScanScope FL (Leica) using 20× objective. Nuclei were visualized in a DAPI channel (blue) and γ-H2AX in Cy5 (red).

Results
At a FSD of 37 cm, XRAD 320 produced a homogenous 14.5 cm × 14.5 cm radiation field (data not shown). The mean dose rate in the air at this distance was ≈2.9 Gy min −1 (4.8 cGy s −1 ) after the filters described in the table 1 (supplemental figures 2(a) and (b)). The simulated x-ray spectrum of the device can be found in the supplemental figure 1(c). The mean energy of the x-ray spectrum was 61 KeV and the 1st HVL Al is 7.99 ± 0.41 (supplemental figure 1(d)). The mounted transmission chamber measured the dose within 1.25 ± 0.08 percentage difference compared to the cylindrical ion chamber.
The dosimetric characteristics of the collimator are shown in table 3. The collimator (figure 2) was able to convert the cone beam of this irradiator to 44 minibeams. The generated beam profile is shown in figure 5. In the y-direction (parallel to the beams), the peak did not decrease as the distance to the central axis increased ( figure 5(B)). As shown in figure 5(C), the profile was uniform in the x-direction, and homogenous valleys were generated.
To study the behavior of the minibeams deep in the tissue, we measured the PDD of peak, valley and integral dose and PVDR at different depths using the PMMA phantom, and compared it with CRT (see supplemental figure 1 for detailed phantom measurements and setting). The results are shown in figure 6. It should be noted that the thickness of the radiochromic films (EBT3 = 278 µm; MD-V3 = 255 µm) were added to the PMMA sheet thickness (2 mm) to measure the depth of each point in the phantom. The entire doses in figure 6(A) were normalized to the mean entrance dose in CRT and the mean entrance integral dose in MBRT. Interestingly, the CRT and MBRT integral dose demonstrated a similar pattern in depth. The peak dose dropped to its half value at 19.61 ± 0.04 mm depth, while the valley dose increased in the first 12.7 mm and then started to decrease. As a result of peak and valley dose behavior in the tissue, PVDR decreased exponentially and reached a plateau at a depth of almost 23 mm (figure 6(B)).
To investigate the cell response to the MBRT, the surviving fractions of the two cell lines after different doses of MBRT were calculated (figure 7).
One major concern in using non-synchrotron irradiators (divergent photons) to apply MBRT is the loss of spatial resolution deep in the tissue, as well as the potential formation of a homogeneous high dose radiation field that may cause tissue damage at the exit plane. The beam profile dependence on the depth of PMMA is shown in figure 6(C). The beams maintained their resolution at least up to the depth of our phantom (≈2.53 cm), sufficient for small animal studies, where the thickest portion of the body (head) in the prone position is around 2 cm. To further investigate the depth dependence of the beam, a 4 cm × 10 cm EBT3 film was placed along the beam path (the longest dimension was parallel to the beams) between two 2 mm thick PMMA sheets and irradiated with 25 Gy (figures 8(A) and (B)). Interestingly, the peaks and valleys were distinguishable at a depth of 10 cm (figure 8(C)). To evaluate the spatial resolution of the beams in vivo, a mouse brain was irradiated with 100 Gy minibeams. The mouse was sacrificed 6 h post-irradiation and its brain tissue was stained with γ-H2AX,  a DNA double-strand break marker. The mini-beams maintained their spatial resolution in vivo where physiological movements (heartbeat and respiration) were two major confounding factors ( figure 8(D)).
To investigate the effect of MBRT on normal tissue and compare it with CRT, the right legs of normal mice were irradiated with graded doses of either irradiation modality (MBRT and CRT) and screened for the appearance of acute-skin effects (see table 2 for scoring detail). We observed that the highest doses that did not induce radiation side-effects in mice were 150 Gy and 15 Gy in MBRT and CRT, respectively (figure 9).
The therapeutic effect of our method was investigated by irradiating the mouse models of melanoma one week after right thigh inoculation. Using the current setting and the doses we applied, we found that tumor started to grow at a slower pace in MBRT and the mice in the MBRT group survived significantly longer than the other two groups (p < 0.001).  Interestingly, the tumor did not grow in one mouse that received MBRT treatment up to 60 d post tumor inoculation. As a result, MBRT was more robust in controlling the tumor growth rate than CRT (figure 10).

Discussion
Synchrotron generated micro-and minibeam radiation preclinical studies have shown promising results in sparing the normal tissue while inducing higher therapeutic effects than conventional radiation therapy (Dilmanian et al 2006, Deman et al 2012, Prezado et al 2015. These findings have prompted investigators to explore ways to generate MRT and MBRT using non-synchrotron irradiators, with the goal of facilitating the translation of this promising modality to the clinic. However, most of these studies have designed complicated collimators or irradiators that confined the MRT and MBRT studies to a few labs and facilities (Babcock et al 2011, Hadsell et al 2013, Hong et al 2015, Bartzsch et al 2016. Here, we have demonstrated a straightforward and affordable method for applying homogeneous MBRT over a large field of irradiation using an animal irradiator. Moreover, our simple method appears to be highly effective and the data generated can be reproduced in different facilities. When compared to previous studies, the dosimetric characteristics of our method are in great agreement. The PDD pattern of peaks, valley, and PVDR is identical to previously measured and reported dosimetric evaluations of MBRT (Siegbahn et al 2006, Gokeri et al 2010, Deman et al 2011. The PVDR in our study is also comparable to a previously reported nonsynchrotron based study (Babcock et al 2011). In fact, it should be noted the valley dose is generated due to compton scatter, and consequently, it is roughly proportional to the primary peak dose from which it originates and decreases as the number of minibeams decreases (smaller radiation field size). As a result, the fact that we generated comparable PVDR while covering a larger radiation field (4 cm × 10 cm field versus 1.75 cm circular field in the study by Babcock et al) and using a higher peak dose (100 Gy for peak and valley doses measurements versus 20-25 Gy in the study by Babcock et al) underlines the utility of our method. Furthermore, we utilized two different radiochromic films with different ranges of sensitivity to precisely measure peak and valley doses and scanned them with high resolution (1200 dpi), which enabled higher accuracy compared with previous studies (Babcock et al 2011). The calibration curves of EBT3 and MD-V3 in three-color channels are demonstrated in the supplemental figure 3. We also found that from ≈23 mm deep in the tissue, the PVDR remains almost constant within the error bars ( figure 6(B)). This pattern has been reported before (Deman et al 2011), and illustrates the normal tissue sparing effect of this method at the exit plane. The peaks in our method dropped to 50% at lower depth compared to synchrotron generated MBRT, which was expected due to the lower mean of energy used in our experiment relative to the synchrotron (61 KeV here versus 80 KeV or higher in synchrotron studies) (Deman et al 2011).
The use of an industrial irradiator as the source of radiation for MRT or MBRT introduces two major limitations as shown in figure 11: First, in MRT and MBRT, kilovolt x-ray beams are utilized in order to keep the spatial fractionation deep in the tissue. At this energy level, more than 99% of the electrons' energy would be converted to heat in the anode. In the current study, an oiled-cooled-anode irradiator with a large focal spot (8 mm) was employed that provides better heat conduction and, in addition, lessens the heel effect. So, a large homogeneous radiation field can be used to apply MRT or MBRT on a large area to minimize the duration of radiation ( figure 12(A)) and consequently, lessen the chance of smearing of the beams due to physiological movements during the long radiation time. As mentioned before, we also employed minibeams instead of microbeams to minimize this effect. As it is clear in figures 8(D) and (E) we were able to generate 100 Gy beams that maintained their resolution deep in the live mouse brain.
Another limitation is the fact that the industrially generated x-ray photons are divergent and, in combination with big focal size, generate a wider penumbra that increases the valley dose deep in the tissue ( figure 12(B)). Several histological studies support the idea that surviving cells in the valley regions, and consequently the valley doses, play a crucial role in maintaining tissue function (Serduc et al 2006, Laissue et al 2007, Schültke et al 2013, Benedict et al 2014. As a result, a narrower penumbra is desired. We utilized the following consideration to minimize the penumbra. 1. In contrast to previous studies (Babcock et al 2011, Hadsell et al 2013b, we used a proximal collimation (the collimator was close to the target instead of the radiation source) (see figure 12(B)). 2. As shown in figure 12(C), there is an inverse relation between the penumbra and FSD.
However, dose rate changes proportionally to 1 FSD 2 . Based on our previous studies (Zhang et al 2014a, Yuan et al 2015, we estimated the required dose to induce therapeutic responses with minimal effect on normal tissue to be 150-300 Gy, so to maintain the total treatment duration to less than 2 h; according to the approved protocol by IACUC to minimize animal anesthetic effects; the farthest FSD was chosen (FSD = 37 cm, dose rate = 2.9 Gy min −1 , collimator relative output factor = 0.84). In addition, by increasing the FSD, using a parallel-septa collimator was feasible, which eased the design and alignment of the collimator ( figure 12(D)). As a proof of principle, and to investigate the efficacy of our method in vitro, the response of two-cell lines using clonogenic assay was evaluated and the cell survival curve was generated (figure 7). This assay measures the ability of a single cell to form a colony and is based on the idea that for a tumor to be eradicated, it is only necessary to render its consisting cells unable to proliferate indefinitely (Hall and Giaccia 2012). When compared to the previously reported survival curve of these two cell lines after CRT, we found that 6 Gy of CRT will induce a comparable in vitro effect to almost 42 Gy of MBRT in TRP and B16-f10 (Twyman-Saint et al 2015, Schmid et al 2016. Future studies include evaluating the surviving fraction of the cells after different doses of each modality, for a precise in vitro comparison of the cell response to CRT versus MBRT. Due to the distinct spatial fractionation of the x-ray beam in MRT/MBRT, finding the actual equivalent dose MRT/MBRT versus CRT is convoluted and studies have used different assumptions for the physical or biological equivalent dose. For instance, Ibahim et al tried to find the in vitro equivalent dose using clonogenic and real-time cell impedance sensing assays (Ibahim et al 2014). In synchrotron based MRT studies, normal tissue toxicity is more dependent on the valley region parameters because: (1) Ultra-high doses of x-ray destroy all cells along the beam path, and (2) the beam size is approximately 25 µm-50 µm, spaced at 200 µm-400 µm on center. Consequently, most of the tissue in the radiation field receives the valley dose. In the current study, the beams to valley FWHM ratio is larger than with synchrotron microbeams, so a higher normal tissue toxicity than equivalent valley dose was expected. Since we investigated the effect on our method on SC flank tumors, the effect of different doses of MBRT and CRT on normal skin tissue in vivo was evaluated to find the optimal radiobiological equivalent dose (RBE). No acute skin toxicity was observed in mice that were irradiated with doses up to 150 Gy and 15 Gy in MBRT and CRT, respectively (figure 9). In a previous study, we demonstrated the sparing effect of MBRT on normal mice brain tissue post-irradiation, up to 8 months . Currently, we are investigating the full extent of normal tissue toxicity, ED50 and TD50 of our technique on different tissues.
The in vivo therapeutic effect of our method was investigated using a mouse model of melanoma by applying RBE of each modality. We observed that applying MBRT using the current setting and doses was more effective in controlling the tumor growth rate than CRT (p = 0.002) and ablated the tumor in one out of 9 mice in MBRT group (11% chance of ablation) (figure 10). It is worth mentioning that CRT is not an effective treatment for B16-F10 When compared to synchrotron studies, we successfully produced a method that mimics the spatial beam pattern of synchrotron MBRT. Although in the initial MRT experiments an ultra-high dose of x-ray (up to thousands of Gys) have been utilized (Serduc et al 2006), recent studies demonstrated the toxicity effect appears at much lower doses (Mukumoto et al 2017). Serduc et al found that the toxicity is directly correlated to beam width (Serduc et al 2009). Using our approach, 150 Gy is the maximum MBRT peak dose that did not induce skin effects. Our generated beams are almost 250 µm, which are below the threshold (beamwidth ⩾ 400 µm) that induces a different response in normal tissue microvasculature (Brönnimann et al 2016). However, it should be noted that synchrotron sources may induce a different biological response due to the extremely high dose rates.
The translation of our method to clinics may encounter several technical limitations. The low dose rate and limited heat conduction capacity of clinical irradiators are the major limitations of the current method, and, consequently, developing a high intensity kilovoltage irradiator would be advantageous. Second, with proximal collimating, aligning the collimator with the source may require a greater deal of accuracy. Designing a device to mount the collimator may ease this problem. An applicator may also be needed to restrict the radiation field to the Figure 12. (A) Comparison of the generated beam intensity between a small (left) versus large (right) focal spot irradiator; (B) the comparisons in the generated beam profile deep in the tissue in the small (left) or large (middle and right) focal spot irradiator are used and the collimator is placed near the source (left and middle) versus near the target (right); (C) the equation demonstrates in large focal spot irradiators, the penumbra has an inverse relation with the focal spot to surface distance; (D) the schematic shows that by increasing the FSD, parallel septa in the collimator can be used, and a wider radiation field can be covered. The rectangle covers an equal radiation field, far or close to the source, the trapezoid area contains beams with the same degrees.
lesion. Although normal skin demonstrated a higher resistance to MBRT in the acute phase, confining the radiation field to the lesion is always desired in order to minimize the normal tissue toxicity. Finally, beam smearing due to the long duration of therapy also remains a big consideration. One approach would be to apply MBRT by a physiologically gated irradiator (Chtcheprov et al 2014).
Our study has several limitations. One limitation was using radiochromic film for dosimetry. Although these films have been used extensively in MRT and MBRT studies (Crosbie et al 2008, Anderson et al 2010, Hadsell et al 2013, Hong et al 2015, currently the uncertainty levels for film dosimetry has been reported between 1% and 10% depending on the situation (Martisíková et al 2008, Bouchard et al 2009. DeWerd et al 2011, Palmer et al 2015. Here we followed the single scan, three channel protocol, recommended by the supplier to minimize errors (Efficient Protocols for Accurate Radiochromic Film Calibration and Dosimetry 2017). However, the protocol did not eliminate some reported source of errors, like film curvature at scanning (Palmer et al 2015). We cross-calibrated the mounted transmission chamber to an ion chamber and used the transmission plate for dose measurements. This also introduced some potential errors (1.25 ± 0.08 percentage difference between the two methods). To minimize this error, the calibration was done at a high dose (150 Gy) and checked twice. Finally, some errors were introduced from utilizing the plastic phantom (PMMA or acrylic) instead of water. Although the IAEA TRS398 code of practice approved their application for low energy x-ray dosimetry, the plastic phantom introduces error in measurements mainly due to density variation (up to 4%) in different batches, and non-homogenous thickness distribution even in one sheet (Tello et al 1995). The density of the sheet we used was 1.174 g cm −3 and we measured the entire slab and used the piece that was 2 ± 0.01 mm thick.

Conclusion
In conclusion, we have demonstrated a relatively simple and easily reproducible method for applying homogeneous MBRT using a small animal irradiator unit. We described our approach, its dosimetric characteristics, and its effectiveness in vitro and in vivo. At its current stage, our method can be used for applying MBRT in preclinical studies.