FLASH beam generation
Our modified linac setup was based on Lempart et al’s. [18], with some changes to simplify its implementation. These changes reduced the equipment needs while ensuring a stable dose output. We used the analog signal detection port of the beam generation system reserved for engineers to identify and count the electron beam pulses. Unlike in Lempart's work [18], we, in principle, simplified the signal path of pulse identification to achieve a higher count accuracy by analyzing the picked signals. The test signal could be the magnetron current (MI) or the modulator's pulse forming network (PFN) with a sharp falling edge for each pulse, as seen in Fig. 1a. The test signal passed through a voltage comparator. At a preset number of pulses, the comparator's output was reversed and transmitted to the microcontroller to send a trigger to a high-speed relay, which was connected to the Function Keypad's (FKP) interrupt port. In doing this, we realized the beam termination at a single or any number of pulses as set. To set the delivered pulse number, an 8-bit timer was connected to a microcontroller unit (MCU, AT89C52, Atmel Corporation, San Jose, California, USA) (Fig. 1b), similar to the electrical control unit reported by Lempart et al. [18].
For mice irradiation, the "treatment head" of the linac, or the beam limiting device (BLD), was lifted and removed with the gantry set at 180 degrees. The sample was placed a short distance from the source to increase the dose output and the BLD safety interlocks bypassed. The filament current, magnetic field, and charging current of the magnetron were adjusted to increase the radio frequency’s (RF) power. The electron gun’s output was fine-tuned while the RF’s power was gradually increased so that the starting pulses were stable (Fig. 1a). The energy beam was slightly altered from the clinical 6 MeV beam.
To test if the dose output of each pulse was stable, a repeated single-pulse dose measurement was performed with a PTW Advanced Markus ion chamber (Type 34045, PTW-Freiburg) placed at a distance from the irradiation sample. In order to determine whether the output was stable, a single-pulse beam was repeated 25 times with 10 second intervals. Film measurement was compared with the chamber’s measurement for consistency checks.
Till now, the optimum FLASH parameters are uncertain. Here, we tested two strategies to obtain an ultrahigh dose rate by either a manual repeated single pulse beam with a trigger interval of 20 seconds in between (Fig. 1c upper) or multiple pulses with a 10 millisecond (ms) interval triggered once. The pulse width was 3.3 microsecond (µs) and the PRF was 12.5 Hz for single pulse delivery and 100 Hz for multiple pulse delivery (Fig. 1c lower). Verification experiments were performed before irradiating the mice.
Film dosimetry
Gafchromic EBT3 films [GafchromicTM EBT3-1417, Lot number 06191801 (expiration date June 2020)] and EBT XD films [Ashland Inc., Bridgewater, NJ 08807, USA, lot number 08221901 (expiration date February 2022)] were used for dose measurement. A study by Jaccard et al. [19] demonstrated response differences to EBT3 film between four electron beams with energies between 6 and 16 MeV was within 0.5%, so we decided to perform optical density (OD) calibration with a 10 MeV electron beam (Elekta Synergy linac) with a 40 mm x 40 mm field size, 100 cm SSD, at the maximum dose depth. The dose points for calibration were from 1 to 15 Gy with 0.5 Gy increments.
Two calibration curves were generated, one for quick reading to ensure the correct dose was to be delivered to the samples, and one for reading the film scanner (Epson Expression 11000XL, Seiko Epson Corporation, Nagano, Japan) 24-hours post-irradiation. Beam percentage depth dose curve (PDD) and absolute dose at the maximum dose depth of the FLASH beam were measured by two films positioned perpendicularly in the mice applicator (depth 16 mm) (Fig. 2a).
Irradiation films were read out with the film scanner 24-hours post-irradiation and analyzed with MEPHYSTO mc² (Medical Physics Tool) (MEPHYSTO mcc 3.3, PTW-Beijing).
Mice irradiation
This study included 69 3 to 5 week old female FvB mice (Jackson Laboratory, Sacramento, CA). This study was approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University Cancer Center with the approval number of L102042019080P.
The FVB mice were anesthetized by isoflurane and immobilized in the prone position in a custom-designed applicator, which consisted of lead blocks to position the head, 15 mm of thick silica gel and a mouse holding plate (Supple. Fig. 1a). Films were placed between the mice and the silica gel to calculate the dose irradiated to the mouse’s ventral skin.
For FLASH irradiation, the applicator was placed 2 cm above the frame of the ionization chamber (SSD 15 cm). For conventional irradiation, the SSD was 95 cm in order to uniformly irradiated two to three mice at a time. The beam first traversed the plastic applicator’s 1mm thick base and the silica gel build-up before reaching the mice, which ensured that the mice’s ventral skin was positioned at the maximum of the depth dose curve (Supple. Fig. 1b).
Serum inflammatory cytokine quantification
Blood was collected from the tail-vein or eyeball using EDTA-coated tubes 6, 18 and30 days post-irradiation. Cells were removed from plasma by centrifugation for 10 min at 3000 revolutions per minute (rpm) at 4◦C. Supernatant was then collected and stored at −80◦C. Tumor necrosis factor-α (TNF-α), interferon-γ (IFN-γ), interleukin-6 (IL-6) and IL-10 was tested using commercial enzyme linked immunosorbent assay (ELISA) kits (Fankew, Shanghai FANKEL Industrial Co., Ltd). All assays were conducted according to the manufacturer’s instructions.
Flow cytometry analysis
Erythrocytes were lysed using a red blood cell lysis buffer (BD Bioscience, 555899). Cell suspensions were filtrated through 70-mm cell strainers (Fisherbrand, 22363548), and then washed and resuspended in a staining buffer (PBS with 2% FBS). Immune cell expression markers CD3 (Cat#100204, Biolegend), CD4 (12-0041-82, Biolegend), CD8 (100712, Biolegend) and CD45 (103132, Biolegend) were determined by flow cytometry analyses after surface staining with anti-mouse specific antibodies conjugated with FITC, PE, APC or PerCP-cy5.5. All stained cells were analyzed on a CytoFLEX Flow Cytometer (Beckman Coulter), and the data analyzed using CytExpert software v2.4.
Statistical analysis
Single dose per pulse, mean dose per pulse, mean dose rate, intra-pulse dose-rate and total delivery time were calculated as follows:
Single dose per pulse = 
Mean dose per pulse = 
Mean dose rate = mean dose per pulse x pulse repetition frequency (mean dose rate for FLASH treatment was modulated by the set PRF)
Intra-pulse dose-rate = 
Total delivery time = pulse width x pulse number + pulse interval time x (pulse number-1)
Data were analyzed by using GraphPad Prism. Student’s t-test or ANOVA were used to compare the differences between two groups or among more groups. The data were presented as mean ± standard error. Statistical significance was denoted by P values. Degrees of significance were *p ≤ 0.05, **p ≤ 0.01, and ***p ≤ 0.001.
