Combined ion beam irradiation platform and 3D fluorescence microscope for cellular cancer research

To improve particle radiotherapy, we need a better understanding of the biology of radiation effects, particularly in heavy ion radiation therapy, where global responses are observed despite energy deposition in only a subset of cells. Here, we integrated a high-speed swept confocally-aligned planar excitation (SCAPE) microscope into a focused ion beam irradiation platform to allow real-time 3D structural and functional imaging of living biological samples during and after irradiation. We demonstrate dynamic imaging of the acute effects of irradiation on 3D cultures of U87 human glioblastoma cells, revealing characteristic changes in cellular movement and intracellular calcium signaling following ionizing irradiation.


Background
Over 50% of cancer patients are treated with radiation therapy (RT) [1,2].Charged particles (protons and carbon ions) are a preferred choice for RT modalities due to their localized physical dose with minimal lateral scattering into surrounding healthy tissues [3].Radiation types are usually characterized by their linear energy transfer (LET), i.e., the energy a particle deposits in the tissue along its path.Amongst charged particles, the radiation dose of high-LET (>10 keV/µm) particles in the tumor is more concentrated around the ion track compared to low-LET (<1 keV/µm) particles, thus controlling toxicity of normal tissue [4].There is now increasing clinical data on the success of Heavy Ion Radiotherapy (HIRT) in treating cancers that are typically refractive to conventional low-LET therapy, including cancers such as pancreas, rectum and sarcomas [5].However, little is known about the underlying molecular mechanisms that drive this efficacy.There has been some discussion [6][7][8][9][10] that as well as producing local responses to the tumor, particle RT may induce long-range systemic anti-cancer effects involving the immune system, and that these effects may be responsible for the overall success of the modality.
Therefore, developing a comprehensive understanding of cancer's molecular and cellular basis by characterizing how cells in the tumors respond to ionizing radiation is critical for radiation oncology.To achieve this goal, researchers need a combination of state-of-the-art tools that can deliver ionizing radiation to living cells in 3D cultures at clinically-relevant doses, while simultaneously acquiring high-speed, functional time-lapse imaging at cellular resolution.

Swept confocally-aligned planar excitation (SCAPE) microscopy
Cellular properties ranging from cell shape and size, to motility and migration behaviors can provide read-outs of cellular health and function.Fluorescent ion indicators such as calcium sensitive dyes, as well as genetically-encoded fluorescent proteins, can also permit optical assessment of functional cell signaling dynamics.
Confocal microscopy has been widely used by biologists to investigate cellular structure dynamics, while two-photon microscopy is a popular tool for capturing calcium activity of neurons in the brain [11].However, point-scanning microscopy methods such as these have limited volumetric imaging speeds and can also cause associated photodamage in sensitive samples.Light sheet microscopy can enable high-speed volumetric imaging of 3D tissues while also providing low phototoxicity [12][13][14].Among the different methods for light sheet microscopy, swept confocally-aligned planar excitation (SCAPE) microscopy has a valuable single-objective geometry, which leaves the other side of the sample unobstructed, permitting integration with sample perturbations [13] such as radiation.SCAPE also offers flexibility to image from 100 volumes-per-second (VPS) to longitudinal time lapse imaging over hours [15,16].SCAPE's compact and relatively simple design made it possible to integrate the system into our vertical beamline.

Prior studies with microbeam irradiation
Charged particle microbeams (micrometer or sub-micrometer diameter of radiation cross-section) are an excellent radiological research tool for delivering controlled doses to living cells [17].Columbia University's Radiological Research Accelerator Facility [18] (RARAF) is one of only a handful of ion microbeam facilities worldwide that is equipped for radiobiological research.
The very small size of microbeams generally necessitates the use of integrated microscopes to target the beam to specific cells, or even sub-regions of a single cell and to observe subsequent cellular responses.For example, cell staining combined with CCD-based epifluorescence microscopy has been used for online cell recognition for radiation treatment, in concert with microbeam irradiation setups at Columbia RARAF [19], Gray Cancer Institute (London, UK) [20,21], GSI (Darmstadt, Germany) [22], and JAERI (Takasaki, Japan) [23] .To avoid cell staining and exposure to ultraviolet light, phase contrast optical microscopy was adopted at SNAKE (Munich University, Germany) [24], and INFN-LNL (Padova, Italy) [25].Unstained cell recognition was further improved by incorporating quantitative phase microscopy to cell irradiation platforms at Columbia RARAF [26] and GSI [26,27].
However, these prior combined platforms would not meet the needs of our current application, which requires both larger-scale irradiation, and volumetric imaging of 3D tissues and cell cultures with functional time-lapse capabilities and low phototoxicity.
The irradiation platform developed here was designed to deliver a range of beam sizes depending on experimental need, overcoming limitations on translating from micrometer-scale radiation delivery to millimeter-scale radiation delivery with similar dose deposition.The vertical configuration of our irradiation platform also permits biological sample dishes to be placed horizontally for irradiation and SCAPE imaging.

Environmental control
The ability to particle beam irradiate in vitro 3D tissue models, which recapitulate disease and tissue processes better than cellular monolayers [28], will advance our understanding of the role of cell signaling and tissue remodeling in radiation responses [29].However, it is imperative to maintain the viability of cells for long-term experiments (>30 min), requiring maintenance of a controlled culture environment (37 °C; 5% CO 2 ; and >97% relative humidity) [30].Despite their small form factor, commercially-available benchtop cell culture environmental chambers cannot be practically accommodated in a microbeam system [15].Therefore, it was crucial to design a custom environmental control system around the vertical microbeam system, which not only allows the maintenance of cultures outside an incubator over an extended period, but also permits integration of the vertical microbeam system with high-resolution and high-speed cellular imaging.

Micro-tumors and invasiveness
The invasion and metastasis of tumor cells requires cell migration [31].Cell migration is a cyclic process involving the repetitive extension of invadopodia/lamellipodia at the leading edge, the formation of adhesion sites, contraction of the cell body, and the release of trailing adhesion sites.These processes, together with the activity of proteases such as metalloproteinases, eventually lead to the migration of invading tumor cells into normal tissues (tumor growth and metastasis).
The cyclic morphological and adherence changes observed during cell migration are accompanied by repetitive Ca 2+ signals, which take the form of Ca 2+ spikes or oscillations.In the U87 glioblastoma cell line the generation of Ca 2+ oscillations have been linked to cell migration [32].In this study, we demonstrate our system's ability to follow both cell motility and migration in 3-dimensions and examine the effects of charged particle radiation on the movement dynamics and Ca 2+ fluctuations and oscillations in 3D microtumor models made from U87 glioblastoma cells.

Volumetric light sheet microscopy with SCAPE
SCAPE microscopy captures high-speed 3D images by illuminating samples using a laterally swept oblique light sheet [13,14].The system incorporated into the accelerator beamline followed the design of system D Z in Voleti et al [14] with modifications to include both, 488 nm and 561 nm excitation, and to facilitate objective positioning (see Fig. 1).
Briefly, in SCAPE merged blue and yellow laser light (488 nm and 561 nm Coherent OBIS lasers) passes through a Powell and cylindrical lenses and is then reflected by a multi-band dichroic filter and deflected by a galvanometer mirror through a telescope that is focused towards the edge of the back focal plane of O1, a primary 1.0 NA 20x water immersion objective lens (Olympus, XLUMPLFLN20XW).This beam path forms an obliquely angled light sheet at the sample.Fluorescent light emitted from the sample is collected through the same objective lens (O1) and passes back through the optical path, reflecting off the galvanometer mirror and passing through the dichroic into an additional telescope to reach the back of an air objective (O2; Nikon CFI Plan Apo Lambda 20x/0.75NA).With a magnification of 1.33 between O1 and O2 (corresponding to the ratio of the refractive index of water and air) this second objective forms a 3D oblique image of the illuminated plane at the sample.A third objective lens (O3; Nikon CFI Plan Apo Lambda 10x/0.45NA) is positioned obliquely such that its focal plane aligns with the oblique plane image.Note, for baseline tests without irradiation, measurements of cellular motility (Fig. 4) were acquired using an offline SCAPE system with O1, O2 and O3 all water-coupled (Olympus, XLUMPLFLN20XW).Light received into O3 is then passed through a dual-color image splitter (spectrally splitting at 561 nm) and focused using a 70 mm tube lens onto a high-speed sCMOS camera (Andor Zyla 4.2+).The image on the camera corresponds to an oblique optical section of the sample, where the focal plane of the camera is aligned with the plane illuminated by the oblique light sheet [33].High-speed 3D images are acquired by moving the system's galvanometer mirror from side to side.This scanning changes the angle of the beam entering the back of O1, translating the oblique light sheet from side to side.Light coming from the sample is de-scanned by the same galvanometer mirror, in a similar way to a confocal microscope, causing the oblique light sheet image between O2 and O3 to remain stationary and thus continually focused on the camera.A full volume can thus be acquired by a single sweep of the galvo mirror, with images on the camera corresponding to focused images of the oblique light sheet as it sweeps laterally across the sample.This mechanism permits very efficient collection of oblique 'optical sections' at high speeds.The camera can read 100 rows (corresponding to 100 depth pixels) at over 2000 frames per second, such that an image spanning 200 lateral steps can be acquired at 10 VPS.3D images are acquired using only one moving part -the galvanometer moving at the volume rate -while the sample and primary objective lens remain stationary.Dual color data was collected simultaneously with one or both lasers on, using the image splitter (splitting at 561 nm) between O3 and the camera tube lens.Split green and red images are placed side by side High-speed 3D images are acquired by moving the system's galvanometer mirror from side to side.This scanning changes the angle of the beam entering the back of O1, translating the oblique light sheet from side to side.Light coming from the sample is de-scanned by the same galvanometer mirror, in a similar way to a confocal microscope, causing the oblique light sheet image between O2 and O3 to remain stationary and thus continually focused on the camera.A full volume can thus be acquired by a single sweep of the galvo mirror, with images on the camera corresponding to focused images of the oblique light sheet as it sweeps laterally across the sample.This mechanism permits very efficient collection of oblique 'optical sections' at high speeds.The camera can read 100 rows (corresponding to 100 depth pixels) at over 2000 frames per second, such that an image spanning 200 lateral steps can be acquired at 10 VPS.3D images are acquired using only one moving part -the galvanometer moving at the volume ratewhile the sample and primary objective lens remain stationary.Dual color data was collected simultaneously with one or both lasers on, using the image splitter (splitting at 561 nm) between O3 and the camera tube lens.Split green and red images are placed side by side onto the camera chip such that no additional rows need to be acquired for dual-color acquisition, thus maintaining the highest possible speed of imaging.
The single-objective nature of the SCAPE system enables samples to be imaged from above while the irradiation beam snout can access the sample from below.The simplicity of SCAPE's optical design made it compatible with proximity to the magnetic environment of the RARAF microbeam, while its compact form factor permitted mounting of the entire microscope above the beam line (Fig. 1).However, modifications to the ion beam focusing system were also required to expand the ion beam and ensure that ions could accurately reach the sample dish as detailed further below.

Optimizing ion beam focusing within the irradiation platform
Columbia University's RARAF facility has been designing, developing, and using focused ion beams for biological irradiation for more than 30 years [17,19,[34][35][36][37][38][39][40][41][42][43][44][45].The focusing systems have used quadrupole focusing elements in electrostatic triplets and quadruplets [35,43] and permanent magnets [37].We recently upgraded our ion focusing system to a 7 T superconducting solenoid (Cryomagnetics Inc, Oak Ridge, TN), designed to enable focusing of a wide range of particles (5 MeV H + to 30 MeV C 6+ ) to spot sizes ranging from 5 mm down to 2 µm in diameter.Our microbeam is centered on this new focusing modality.The use of a solenoid results in fewer aberrations, and a cylindrically symmetric focusing of the ions into the focal spot [46,47].This new microbeam ion beam optics design uses the same object and focusing distances as used in previous RARAF ion beam designs [43], thus requiring only minor facility modifications for support of the 7 T magnet.The largest change is the relocation of the limiting aperture from the bottom of the double triplet lens to the two-thirds location in the solenoid.This change greatly increases the acceptance of the solenoid system, allowing for the same particle fluence over larger spot sizes.
Figure 2(A) shows a model of the solenoid focusing system generated using SIMION [48], a software for simulations of ion beam trajectories.The magnetic field was constructed using the Poisson SUPERFISH [49] field model and imported into SIMION.Note the limiting aperture at the two-thirds point in the solenoid.The Si 3 N 4 vacuum window is placed ≈100 µm below the focus point to minimize the distance in air that the ion beam travels before irradiating the sample.
Two apertures are used to define the beam: the object aperture of the focusing system and the limiting aperture defining the beam envelope.The object aperture (Technisches Bernd Fisher, Darmstadt, Germany); also used as the object aperture for our previous microbeam systems [43] consists of two pairs of precision ground tungsten rods, held apart to form a very controlled wedge from 10 µm to 150 µm.The two pairs are mounted orthogonally to each other and provide semi-rectangular apertures between 10 µm and 150 µm.The crossed wedges are pictured in Fig. 2(B).The limiting aperture is a precision-machined aperture held by a compressive assembly at the two-thirds point in the solenoid field on a custom mounting system designed and built in our machine shop.The mounting system guarantees alignment with the solenoid and with the Si 3 N 4 exit window.This assembly is shown in a schematic in Fig. 2(C).The limiting aperture is an interchangeable brass disc with center holes ranging from 400 µm up to 4 mm depending on the spot size required for the experiment being performed.For the experiments described here we used a 2 mm limiting aperture.
The ion beam is characterized for size and fluence through our standard dosimetry systems [50][51][52].Briefly, the primary beam size measurement is performed using the knife-edge occlusion method for small (<250 µm) beams [43].This method uses a sharp edge that is longer than the beam width, stepped through the diameter of the beam while the percentage of the beam passing the edge is measured at each step.This percentage is then plotted out from the 10%-90% The single-objective nature of the SCAPE system enables samples to be imaged from above while the irradiation beam snout can access the sample from below.The simplicity of SCAPE's optical design made it compatible with proximity to the magnetic environment of the RARAF microbeam, while its compact form factor permitted mounting of the entire microscope above the beam line (Figure 1).However, modifications to the ion beam focusing system were also required to expand the ion beam and ensure that ions could accurately reach the sample dish as detailed further below.

Figure 2. Expansion and extension of ion beam.
A: SIMION model of solenoid (green area is solenoid field) focusing of 30 MeV C 6+ ions to 2 mm beam spot size.Note the limiting aperture in the green field area.This provides very large beam acceptance for ion fluence effective for irradiation over all our spot sizes.The inset (bottom left) shows trajectories of few particles through the solenoid field.Bottom: The object aperture (B) and the limiting aperture schematic (C) of the microbeam.The object is a pair of crossed ground tungsten cylinder wedges giving semi rectangular apertures from 10 µm to 150 µm.The limiting aperture is held at the two-thirds point in the solenoid field and is aligned by construction to the solenoid and the Si 3 N 4 window.The limiting aperture is a brass disc held in place by compression with a center hole ranging from 400 µm to 4 mm.

Optimizing ion beam focusing within the irradiation platform
Columbia University's RARAF facility has been designing, developing, and using focused ion beams for biological irradiation for more than 30 years [17,19,[34][35][36][37][38][39][40][41][42][43][44][45].The focusing systems have used quadrupole focusing elements in electrostatic triplets and quadruplets [35,43] and permanent magnets [37].We recently upgraded our ion focusing system to a 7T superconducting solenoid (Cryomagnetics Inc, Oak Ridge, TN), designed to enable focusing of a wide range of particles (5 MeV H + to 30 MeV C 6+ ) to spot sizes ranging from 5 mm down to 2 µm in diameter.Our microbeam is centered on this new focusing modality.The use of a solenoid results in fewer aberrations, and a cylindrically symmetric focusing of the ions into positions and linearly fitted to 0% and 100% through those locations for the measured beam diameter.Dosimetry for small beam sizes is achieved by counting particles with a solid-state detector that measures the fluence of the beam in the current beam spot.For larger beam sizes (250 µm to 2 mm), EBT3 Gafchromic film (Ashland Advanced Materials, Bridgewater, NJ) with an exposed active layer is used for measuring the beam spot size through observation of the darkening of the film.The film darkening is also used as the dosimetry of the beam for a given spot size and examined for uniformity across the whole irradiation spot.For all spot sizes, the beam is controlled in an on/off manner delivering the required fluence for the experimental dose requested.
For all beam sizes and particle types, LET is verified using a tissue equivalent gas proportion counter (TEPC) [51].The TEPC has been used for more than 40 years as our LET verification system for dosimetry on our Track Segment Facility [51,52].

Cell-culture environmental control
Our imaging and irradiation geometry requires access to both sides of the sample for simultaneous irradiation and imaging.However, this geometry poses a challenge for standard environmental control methods.To address this challenge, we developed a custom cell-culture environmental In our design, gas (5% CO 2 ; Balance Air) flows over a cell culture after it is pushed through Hanks' Buffered Saline Solution (Gibco TM ) in a gas-washing bottle (PYREX™ 31770250EC) placed on a hot plate (Scilogex MS-H280-Pro).This setup minimizes disruption to the culture medium as the aluminum shroud covers the culture dish, ensuring constant gas flow (inset in Figure 1).The availability of incubators in the RARAF microbeam room and the ability to swiftly move the shroud in the vertical direction allow the cell culture dish to be positioned for experiments in less than a minute after being removed from the incubator.
To maintain a controlled temperature (37 °C ± 0.5 °C) in the cell culture medium for extended periods (≥2 h), the aluminum sample holder, aluminum shroud, and the water immersion objective lens (Olympus N20X-PFH 20X) are heated to ≈40 °C.For the objective lens, we used a 2-channel temperature controller (Okolab H401-T-Controller).The shroud and sample holders were heated using a home-built system based on Inkbird ITC-106VH PID Temperature Thermostat Controllers, with polyimide thermofoil heaters (Minco Inc, Minneapolis, MN) and RTD temperature sensors (Minco).

System Integration
As shown in Figure 1, the final integrated system includes the SCAPE microscope mounted above the vertical ion beam's focal point.SCAPE's primary objective (O1) was mounted on a rotation stage to enable the sample dish to be easily positioned above the ion beam snout.Fine repositioning of the sample was achieved by placing the sample holder on a 3D translation stage assembled using Thorlabs XRP50 stages providing 10 µm reproducibility.SCAPE's graphical user interface (Matlab™, Mathworks) was modified to enable both fast imaging and time-lapse acquisition, as well as triggering to enable time synchronization with radiation delivery.

Cell culture and labelling
To test the integrated system, we used a series of 3D tumor models developed at the Center for Radiological Research [53][54][55].Human U87 Glioblastoma cells were chosen as an aggressive tumor model type known to grow rapidly in the 3D model system.These cells are representative of a hard-to-treat tumor that is a prime candidate for HIRT.As described previously [56], to form 3D cell cultures, human U87 Glioblastoma cells were grown as a monolayer, trypsinized, injected into the gel matrix and incubated for 24-48 h before imaging.Depending on the cell type, these injected cell boluses can form either a solid tumor core or a more dispersed cell distribution.U87's high motility and potential for tissue invasion form a more dispersed cell bolus.control system to ensure constant temperature, stable pH levels, and minimal evaporation of the culture medium during irradiation and imaging (Fig. 3).
In our design, gas (5% CO 2 ; Balance Air) flows over a cell culture after it is pushed through Hanks' Buffered Saline Solution (Gibco) in a gas-washing bottle (PYREX 31770250EC) placed on a hot plate (Scilogex MS-H280-Pro).This setup minimizes disruption to the culture medium as the aluminum shroud covers the culture dish, ensuring constant gas flow (inset in Fig. 1).The availability of incubators in the RARAF microbeam room and the ability to swiftly move the shroud in the vertical direction allow the cell culture dish to be positioned for experiments in less than a minute after being removed from the incubator.
To maintain a controlled temperature (37 °C ± 0.5 °C) in the cell culture medium for extended periods (≥2 h), the aluminum sample holder, aluminum shroud, and the water immersion objective lens (Olympus N20X-PFH 20X) are heated to ≈40 °C.For the objective lens, we used a 2-channel temperature controller (Okolab H401-T-Controller).The shroud and sample holders were heated using a home-built system based on Inkbird ITC-106VH PID Temperature Thermostat Controllers, with polyimide thermofoil heaters (Minco Inc, Minneapolis, MN) and RTD temperature sensors (Minco).

System integration
As shown in Fig. 1, the final integrated system includes the SCAPE microscope mounted above the vertical ion beam's focal point.SCAPE's primary objective (O1) was mounted on a rotation stage to enable the sample dish to be easily positioned above the ion beam snout.Fine repositioning of the sample was achieved by placing the sample holder on a 3D translation stage assembled using Thorlabs XRP50 stages providing 10 µm reproducibility.SCAPE's graphical user interface (Matlab, Mathworks) was modified to enable both fast imaging and time-lapse acquisition, as well as triggering to enable time synchronization with radiation delivery.

Cell culture and labelling
To test the integrated system, we used a series of 3D tumor models developed at the Center for Radiological Research [53][54][55].Human U87 Glioblastoma cells were chosen as an aggressive tumor model type known to grow rapidly in the 3D model system.These cells are representative of a hard-to-treat tumor that is a prime candidate for HIRT.As described previously [56], to form 3D cell cultures, human U87 Glioblastoma cells were grown as a monolayer, trypsinized, injected into the gel matrix and incubated for 24-48 h before imaging.Depending on the cell type, these injected cell boluses can form either a solid tumor core or a more dispersed cell distribution.U87's high motility and potential for tissue invasion form a more dispersed cell bolus.
We tested the combined imaging and irradiation system with a range of cell labelling strategies: Cyto-Red (Biosettia) and CellTracker Red (Thermo Fisher) for whole cell labeling, and Oregon Green BAPTA-1 AM (Thermo Fisher), a calcium sensitive fluorescent dye as a functional reporter cellular Ca 2+ dynamics.We also tested human U87 cells transfected with the pGreenFire 2.0 NFkB reporter (SBI).

Experimental data acquisition
Experimental parameters were established during preliminary testing using both online and offline SCAPE systems.Dye concentrations, laser powers, and imaging durations were selected to maximize signal to noise while minimizing phototoxicity effects that could confound interpretation of irradiation effects.
For calcium imaging experiments (Figs.5-6), 3D cultures formed from human U87 Glioblastoma cell cultures were incubated with CellTracker Red and Oregon Green BAPTA-1 AM for 2 hours, then placed in phenol red free growth media clean buffer for 30 minutes prior to imaging.Samples were imaged using the integrated irradiation + SCAPE system with environmental control, collecting a 350 × 500 × 320 x-y-z pixel in 1 second, repeated every 12.5 s for 30 minutes (imaged at 1.4 µm × 2 µm × 1.1 µm voxels sampling = 700 µm × 700 µm × 350 µm xyz).Samples were irradiated (as detailed further below) for 30 seconds after around 10 minutes of baseline imaging.Control samples were imaged in the same way but without irradiation.

Data analysis
Resulting SCAPE dual-color, 3D time-series data were color-registered and visualized in 2D (Matlab and ImageJ) and 3D (Imaris) to assess cell motility (Fig. 4).For quantitative analysis of intracellular calcium datasets (Fig. 5 and Fig. 6), Matlab was used to first segment 3D volumes (combined red + green channels) using adaptive thresholding (imbinarize) and connected component analysis (bwconncomp), , defining segmented regions with >15 voxels as putative cells.Guided by segmentation of the first volume, segmentation of subsequent volumes was used to enable 3D motion tracking of each cell for the duration of the recording.The time-dependent xyz centroid position of each cell's segmented region of interest (ROI) over time was then used as a measure of the cell's movement dynamics (quantifying the rate of change of smoothed Euclidean displacement after subtraction of mean motion across all cells corresponding to physical sample drift).Red and green fluorescent signals were extracted from a 10 × 10 × 10 voxel cube around each time-varying xyz centroid.Calculation of the green / red fluorescence ratio can account for motion and sample-volume related signal changes to provide a more robust measure of calcium-dependent changes [57].De-skewing corrections [14] were applied to raw data to generate Imaris renderings in Fig. 4, Fig. 5(A) andVisualization 1 and Visualization 2, along with segmented data to generate 3D renderings of the resulting 3D ROIs corresponding to each cell (Fig. 5C(i) and D(i)).
Individual and population-level analysis could then be performed on 500 -1,200 segmented cells per sample to evaluate the simultaneous real-time dynamics of both calcium signaling and cell motility in relation to irradiation.This analysis included quantification of calcium oscillatory patterns using the spectrogram of ratiometric calcium traces as illustrated in Fig. 6(A).

3D tumor model functional responses to radiation
To explore the acute effects of charged particle radiation on cell motility and intracellular calcium dynamics, human U87 Glioblastoma 3D culture models labelled with CellTracker Red and Oregon Green BAPTA-1 AM were imaged using our beam-line integrated SCAPE microscopy system as detailed above.3D images were acquired at 1 VPS at 12.5 s intervals for 10 mins before irradiation and for 20 mins following irradiation, with the irradiation taking ≈30 s (2-3 time points of acquisition).The samples were irradiated with 16 ± 1 Gy of protons at an average LET of 12 keV/µm uniformly in a spot size that circumscribed the imaging area (1.1 mm in diameter, solenoid current: 35 A, limiting aperture 2 mm, object aperture 50 µm x 50 µm).Control measurements were acquired in the same way, but without irradiation.
Figure 5(A) shows a 3D rendering of de-skewed raw SCAPE data (single volume) showing co-labelling of cells with red CellTracker and green calcium sensitive fluorophores.Figure 5(B)(i) shows a smaller volume of the same sample, with specific cells highlighted and tracked (as detailed above).Time-courses of the indicated cells are plotted in Fig. 5(B)(ii), showing the ratiometric green/red fluorescence dynamics (corresponding to intracellular calcium) along with the motion dynamics of the same cell.
These temporal traces represent the range of different dynamic intracellular calcium signals typically observed in this system, and vary between slow trends, sharp increases and decreases, transient flashes and the onset and cessation of periodic oscillatory activity at a range of frequencies.Cells were also observed to transiently initiate and cease moving, with some cells moving back and forth vigorously (motility), and others exhibiting sudden relocations (migration) as well as cell division.Visualization 2 and Visualization 3 highlight these dynamics, and especially the behavior of cells 10 and 11 (also seen in the sequence of frames shown in Fig. 5(B)(iii)) which spontaneously separate synchronously with several other cells, while cell 8 exhibits a sudden relocation during a transient increase in calcium.
These complex dynamics and interactions highlight the challenge of comprehensively evaluating the irradiation response of a population of cells.The sample shown here (Dish 1) was irradiated at 10 minutes (as indicated by dotted vertical lines in Fig. 5(B)(ii)), and although this irradiation coincides with changes in several cells (e.g. the onset of oscillations in cell 1), there are other cells whose oscillations onset prior to irradiation (e.g.cell 4).Irradiation effects may also cause delayed changes in cellular function and behavior making causal effects challenging to disambiguate from unrelated events.
Comparisons of pilot datasets acquired in samples irradiated with both 8 Gy and 16 Gy, as well as un-irradiated controls, revealed heterogeneous patterns of cell motility, mobility and intracellular calcium dynamics.One major variable was found to be the initial health state of each 3D culture, along with progressive changes in cell health likely relating to environmental regulation and imaging effects.Fortunately, our imaging platform's ability to simultaneously monitor 100's or even 1000's of cells within each sample permits detailed analysis and comparison of a wide range of population-level effects.Here, we present results from 3 samples: Dish 1 and Dish 2 were irradiated with 16 Gy protons, and were found to exhibit similar radiation responses, while Dish 3 is an un-irradiated control.All 3 samples were prepared from the same cell lines, were kept under careful environmental control, and all imaging of these control and irradiated samples was performed on the same day.5(D) shows data from the un-irradiated control (Dish 3) for comparison (from which over 600 cells were segmented).A clear feature observed in Dish 1 (but not the control) is a transient decrease in the ratiometric calcium signal during irradiation.Figure 5(E)(i) plots the average of the normalized red, green and ratiometric signals and confirms that a similar dip is observed in irradiated Dish 2, but not in control Dish 3.This representation also demonstrates that the dip is primarily seen in the green (calcium-dependent) channel, and not in the red signal, consistent with a physiological rather than a physical effect.The relatively uniform baseline of the ratiometric signal demonstrates the value of correcting the green signal with red.The average motion of cells in all 3 dishes is also plotted in Fig. 5(E)(ii).Although there is a transient increase in movement in Dish 1, this effect starts before irradiation and is not seen in Dish 2 or the control (Dish 3), and so is inconclusive.
Although these average fluorescence responses in Fig. 5(E)(i) confirm an irradiation effect, to further leverage the power of our approach, we developed analysis methods to explore single-cell and sub-population level effects as shown in Fig. 6.First, the presence of oscillations in each ratiometric timecourse was inferred from spectrogram power in the 0.007-0.016Hz range (Fig. 6(A)) (using detrended, max-normalized green fluorescence data).Each cell's average movement and oscillation power was then calculated for periods prior to, and after irradiation (averaging over 0 to 8 mins (pre) and 11-30 mins (post) for oscillations, and 2 to 8.7 mins (pre) and 12.6 to 26 mins (post) for movement).The same time windows were used on the un-irradiated control.
Scatter plots in the top row (Fig. 6(B)) show the change in oscillation and movement for every cell, before (pre) and after (post) irradiation for Dishes 1 and 2, and for the same time-periods for the control dish (Dish 3).Plots Fig. 6(c)-(f) below shows scatter plots of subsets of cells, selected based on differences in their collective behavior: Fig. 6(C) Cells whose mean oscillation power increased after irradiation (by more than 0.2), Fig. 6(D) Cells whose mean oscillation power decreased after irradiation (by more than 0.15), Fig. 6(E) Cells whose movement decreased by more than 0.05 and Fig. 6(F) Cells whose movement increased after irradiation (by more than 0).The % of total cells in each group is also noted on each plot.Example timecourses of cells in each of these groups, for the two irradiated dishes (1 and 2) and the control (dish 3) are also shown.
Although patterns are difficult to discern when plotting all cells together in Fig. 6(B), several trends are clear from the sub-population plots.For example, in (Fig. 6(C)) we can see that in both of the irradiated samples, cells whose calcium oscillations increase after irradiation show a clear shift from higher to lower levels of movement.Although cells in the control also show that increased calcium oscillations correspond to less movement, cells within this group transition to both higher and lower levels of movement with little net overall movement change.We infer that oscillatory states generally correspond to reduced cell motion (consistent with [58]), but that this behavior after irradiation may represent transition to a (potentially irreversible) impaired state that is accelerated / exacerbated by irradiation.Our observation that Dish 1 had a higher number of oscillatory cells initially, could imply that these cells were at a lower initial health state than cells in dishes 2 and 3.Although dishes 2 and 3 had very few cells with clear initial calcium oscillations (Fig. 6(D) inset plots), to examine whether cells ever recover from being in an oscillatory state, we examined cells whose calcium oscillations decreased over time (Fig. 6(D)).We found that almost all cells in this group also decreased their movement after irradiation.We interpret decreased movement in cells transitioning out of an oscillatory state as progressing to a worsened state of cell damage.Isolating the cells exhibiting the strongest decreases in movement (Fig. 6(E)), we note that a higher fraction of cells make this transition in the irradiated dishes (14.7% and 21.1% compared to 7.5% in the control), further supporting that these cells have To analyze subpopulations more clearly, we plot in C-F scatter plots and histograms of sub-populations meeting different criteria, while providing example fluorescence traces from individual cells within each group for each dish (left).C. Shows cells with the largest increase in oscillations after irradiation, D: largest decrease in oscillations, E: any decrease in movement and F: largest increase in movement.The % of overall cells within each subgroup is indicated on the Y axis.
sustained irradiation-related damage.Examining cells whose movement levels increased after irradiation (Fig. 6(F)), we find that almost 60% of cells in the control exhibit higher movement levels later in the acquisition period, whereas only 22.4% and 34% of cells in irradiated dishes 1 and 2 respectively show increases in movement, also consistent with irradiation-related damage.We note that these movement effects were detectable within sub-populations, but were much harder to distinguish in overall average movement plots shown in Fig. 5(E)(ii).Time courses plotted for each subgroup (Fig. 6(C)-(F)) are for individual cells, each verified by eye in the raw data as accurate representations of calcium signal and movement changes.They highlight the wide variety of trends, patterns and oscillation frequencies observed.In rare cases, we also observed cells whose calcium levels (green fluorescence) abruptly decreased to near-zero, although insufficient cells were observed to infer irradiation-dependence.As visible in Supplemental videos, there were also a range of (apparent) cells of different sizes, including small and medium-sized cells which exhibited rapid 'wriggling' (motility) versus translation (mobility / migration), a property that may not have been accurately quantified or distinguished by our relatively simple 'movement' metric.In some cases, we observed large movements preceding transition to an oscillatory state.Some movements were also associated with calcium transients (such as rapid fluctuations, or bursts / flashes) that could be considered as distinct from more periodic oscillations observed in cells that tended to be stationary.These properties may not have been entirely disambiguated by our 'oscillation' and 'movement' metrics quantified within 'pre' and 'post' time-periods.However, our observations of hese more complex features of individual cell behaviors highlight the complexity of this kind of data, and the myriad of different forms of quantitative analysis that could be performed with our integrated system to understand the effects of perturbations at the level of single cells and populations within 3D cell cultures over diverse timescales.

Discussion
Recent investigations have suggested that cancer cells killed by charged particles may be more immunogenic than those killed by photons [59,60] and thus charged particles have a potential to revolutionize radiotherapy [7].It is known that radiation affects calcium signaling which, in turn, is involved in several key cellular responses, including being one of many signals for cell death mechanisms, e.g., apoptosis [61].The calcium signaling pathway has also been associated with immunosuppression of gliomas [62].Therefore, to better understand, and thereby improve, particle radiotherapy, it is crucial to develop a comprehensive appreciation of local and non-local effects of charged particles on cellular function and physiology.
The technology developed in this work allows simultaneous evaluation of targeted and nontargeted effects of charged particle radiation.We demonstrated that real-time imaging of the exposure of a 3D tumor cell culture to irradiation allows evaluation of the time-dependent modulation of intracellular signaling and movement dynamics in almost every cell in the population (Fig. 6(A)).
Disruptions (mostly increases) in spontaneous calcium oscillations in 2D cell cultures have previously been observed post X-ray irradiation [63,64].However, our ability to simultaneously measure and unambiguously extract both calcium signals and movement from 100 -1,000's of cells throughout an intact 3D tumor model adds significantly richer information to these prior studies.In our case, analysis revealed the importance of looking at sub-populations of cells to account for the stochastic nature of irradiation responses, the widely ranging states of cells within a population, as well as considering the confounding effects of irradiation-independent changes in cell health and function during long imaging sessions.
Our analysis confirms that there are important relationships between cellular calcium dynamics and cell mobility and motility, and particularly highlights the occurrence and properties of intracellular calcium oscillations in living cells.Consistent with previously proposed models of calcium-modulated growth cone dynamics [58] in many cases we observed a reduction in movement during calcium oscillations (Fig. 6(C)).Notably, our data (Fig. 6(D)-(F)) suggests that in irradiated samples, cells were less likely to recover from this oscillatory state and recommence movement, a finding which could be relevant for understanding the effects of irradiation on tumor growth, spread and metastasis.Further studies are required to more fully explore this phenomenon to better account for the effects of initial cell health, and to improve quantification of cellular behavior to better stratify types of motion and calcium dynamics.
Extensions of this work could include comparing radiation types, dose responses and assessment of different types of normal and malignant cell types in different kinds of 3D culture and organoid systems, or even in intravital animal tumor models.Observations could be extended to follow cellular responses over multiple hours or even days following irradiation.Imaging assessments could also include monitoring the irradiation-response of range of other fluorescent markers of interest, including markers of the structure and function of cellular organelles (e.g., mitochondria), altered gene expression and DNA damage response proteins.
In summary, we have developed and demonstrated a versatile radiation platform to deliver micrometer to mm diameter radiation beams with similar dose deposition.This new irradiation platform has been used to irradiate 3D tumor models based on a human cancer cell line.The integrated SCAPE microscopy system and environmental control system permits exploration of mechanistic effects of charged particle radiotherapy at the cellular and tissue levels.We demonstrated that this integrated irradiation and imaging system enables observation of immediate radiation effects, in this example, effects on motility and spontaneous calcium signaling in 3D samples before, during, and after irradiation over extended periods.

Figure 1 .
Figure 1.An integrated system for irradiation and fast 3D microscopy.A: Photograph of the RARAF microbeam room with the integrated SCAPE [13] system (to the right) and the cell culture environmental setup (to the left).B: Closer view of the sample holder, cell culture dish, and the shroud around the objective lens.C: SCAPE microscope optical layout and sample geometry (inset).

Fig. 1 .
Fig. 1.An integrated system for irradiation and fast 3D microscopy.A: Photograph of the RARAF microbeam room with the integrated SCAPE [14] system (to the right) and the cell culture environmental setup (to the left).B: Closer view of the sample holder, cell culture dish, and the shroud around the objective lens.C: SCAPE microscope optical layout and sample geometry (inset).

Fig. 2 .
Fig. 2. Expansion and extension of ion beam.A: SIMION model of solenoid (green area is solenoid field) focusing of 30 MeV C 6+ ions to 2 mm beam spot size.Note the limiting aperture in the green field area.This provides very large beam acceptance for ion fluence effective for irradiation over all our spot sizes.The inset (bottom left) shows trajectories of few particles through the solenoid field.Bottom: The object aperture (B) and the limiting aperture schematic (C) of the microbeam.The object is a pair of crossed ground tungsten cylinder wedges giving semi rectangular apertures from 10 µm to 150 µm.The limiting aperture is held at the two-thirds point in the solenoid field and is aligned by construction to the solenoid and the Si 3 N 4 window.The limiting aperture is a brass disc held in place by compression with a center hole ranging from 400 µm to 4 mm.

Figure 3 .
Figure 3.The cell culture environmental scheme (not drawn to scale) in the RARAF microbeam room for extended (≥2 h) imaging and radiation exposures outside an incubator.Created with BioRender.com

Fig. 3 .
Fig. 3.The cell culture environmental scheme (not drawn to scale) in the RARAF microbeam room for extended (≥2 h) imaging and radiation exposures outside an incubator.Created with BioRender.com

Figure 4
Figure 4 show 3D renderings of dynamic SCAPE microscopy images of 3D cultures of human U87 Glioblastoma cells transfected with pGreenFire 2.0 NFkB reporter and Cyto-Red (as detailed above).Images are shown at different magnification scales showing detailed cell morphology in both red and green color channels.Panels B and C highlight the motility of different cells during the course of the time-lapse acquisition.Visualization 1 shows this full dataset, permitting the continuous motility of cells to be appreciated along with the lack of photobleaching throughout the 1.5 h imaging session.

Figure 4
Figure 4  shows 3D renderings of dynamic SCAPE microscopy images of 3D cultures of human U87 Glioblastoma cells transfected with pGreenFire 2.0 NFkB reporter and Cyto-Red (as detailed

Fig. 5 .
Fig. 5. Intracellular calcium activity and movement dynamics in human U87-glioblastoma 3D culture during 16 Gy proton irradiation.A) A single timepoint of a 3D volumetric movie of U87 Glioblastoma cells taken using the SCAPE imaging system integrated into the RARAF beam-line .Volume is 700 µm × 700 µm × 350 µm acquired in 1 s (Dish 1).B(i) Subvolume of A showing numbered cells on x-y (top) and y-z (bottom) maximum intensity projections (MIPs) with corresponding fluorescence and movement traces extracted in B(ii).B(iii) shows a series of MIP timepoints from a subregion of B(i).Stationary arrows highlight: white arrows: two cells (10 and 11 in B(ii)) that suddenly separate at the same time as the cells highlighted by the blue arrows move apart.Pink arrow highlights cell (8 in B) which increases calcium and then shifts its position downwards.See Visualization 2 and Visualization 3 for dynamic versions of A and B. C (irradiated, Dish 1) and D(i) and (unirradiated control, Dish 3) show 3D segmentation used to extract fluorescence and movement signals from 600 -1200 cells per sample.Heatmaps in C(ii) and D(ii) show (max-min) normalized ratiometric green/red fluorescence (left) and movement (right) for all cells (colors matching C-D(i)).A dark vertical stripe in C(ii), left corresponds to the irradiation period (vertical white dotted lines).E(i) shows averaged normalized green, red and ratiometric signals with standard errors for two irradiated dishes and the control with the irradiation period denoted by vertical dotted black lines.E(ii) shows the average motion signal with standard error across all cells for the same two irradiated samples and control.

Figure 5 (
Figure 5(C) shows segmented 3D regions of interest corresponding to over 1,200 (putative) cells segmented in Dish 1, and plots their ratiometric calcium signals (normalized to their maximum values) and movement signals as a kymograph.Figure 5(D)  shows data from the un-irradiated control (Dish 3) for comparison (from which over 600 cells were segmented).A clear feature observed in Dish 1 (but not the control) is a transient decrease in the ratiometric calcium signal during irradiation.Figure5(E)(i) plots the average of the normalized red, green and ratiometric signals and confirms that a similar dip is observed in irradiated Dish 2, but not in control Dish 3.This representation also demonstrates that the dip is primarily seen in the green (calcium-dependent) channel, and not in the red signal, consistent with a physiological rather than a physical effect.The relatively uniform baseline of the ratiometric signal demonstrates the value of correcting the green signal with red.The average motion of cells in all 3 dishes is also plotted in Fig.5(E)(ii).Although there is a transient increase in movement in Dish 1, this effect starts before irradiation and is not seen in Dish 2 or the control (Dish 3), and so is inconclusive.Although these average fluorescence responses in Fig.5(E)(i) confirm an irradiation effect, to further leverage the power of our approach, we developed analysis methods to explore single-cell and sub-population level effects as shown in Fig.6.First, the presence of oscillations in each ratiometric timecourse was inferred from spectrogram power in the 0.007-0.016Hz range (Fig.6(A)) (using detrended, max-normalized green fluorescence data).Each cell's average movement and oscillation power was then calculated for periods prior to, and after irradiation (averaging over 0 to 8 mins (pre) and 11-30 mins (post) for oscillations, and 2 to 8.7 mins (pre) and 12.6 to 26 mins (post) for movement).The same time windows were used on the un-irradiated control.Scatter plots in the top row (Fig.6(B)) show the change in oscillation and movement for every cell, before (pre) and after (post) irradiation for Dishes 1 and 2, and for the same time-periods for the control dish (Dish 3).Plots Fig.6(c)-(f) below shows scatter plots of subsets of cells, selected based on differences in their collective behavior: Fig.6(C) Cells whose mean oscillation power increased after irradiation (by more than 0.2), Fig.6(D) Cells whose mean oscillation power decreased after irradiation (by more than 0.15), Fig.6(E) Cells whose movement decreased by more than 0.05 and Fig.6(F) Cells whose movement increased after irradiation (by more than 0).The % of total cells in each group is also noted on each plot.Example timecourses of cells in each of these groups, for the two irradiated dishes (1 and 2) and the control (dish 3) are also shown.Although patterns are difficult to discern when plotting all cells together in Fig.6(B), several trends are clear from the sub-population plots.For example, in (Fig.6(C)) we can see that in both of the irradiated samples, cells whose calcium oscillations increase after irradiation show a clear shift from higher to lower levels of movement.Although cells in the control also show that increased calcium oscillations correspond to less movement, cells within this group transition to both higher and lower levels of movement with little net overall movement change.We infer that oscillatory states generally correspond to reduced cell motion (consistent with[58]), but that this behavior after irradiation may represent transition to a (potentially irreversible) impaired state that is accelerated / exacerbated by irradiation.Our observation that Dish 1 had a higher number of oscillatory cells initially, could imply that these cells were at a lower initial health state than cells in dishes 2 and 3.Although dishes 2 and 3 had very few cells with clear initial calcium oscillations (Fig.6(D) inset plots), to examine whether cells ever recover from being in an oscillatory state, we examined cells whose calcium oscillations decreased over time (Fig.6(D)).We found that almost all cells in this group also decreased their movement after irradiation.We interpret decreased movement in cells transitioning out of an oscillatory state as progressing to a worsened state of cell damage.Isolating the cells exhibiting the strongest decreases in movement (Fig.6(E)), we note that a higher fraction of cells make this transition in the irradiated dishes (14.7% and 21.1% compared to 7.5% in the control), further supporting that these cells have

Figure
Figure 5(C) shows segmented 3D regions of interest corresponding to over 1,200 (putative) cells segmented in Dish 1, and plots their ratiometric calcium signals (normalized to their maximum values) and movement signals as a kymograph.Figure 5(D)  shows data from the un-irradiated control (Dish 3) for comparison (from which over 600 cells were segmented).A clear feature observed in Dish 1 (but not the control) is a transient decrease in the ratiometric calcium signal during irradiation.Figure5(E)(i) plots the average of the normalized red, green and ratiometric signals and confirms that a similar dip is observed in irradiated Dish 2, but not in control Dish 3.This representation also demonstrates that the dip is primarily seen in the green (calcium-dependent) channel, and not in the red signal, consistent with a physiological rather than a physical effect.The relatively uniform baseline of the ratiometric signal demonstrates the value of correcting the green signal with red.The average motion of cells in all 3 dishes is also plotted in Fig.5(E)(ii).Although there is a transient increase in movement in Dish 1, this effect starts before irradiation and is not seen in Dish 2 or the control (Dish 3), and so is inconclusive.Although these average fluorescence responses in Fig.5(E)(i) confirm an irradiation effect, to further leverage the power of our approach, we developed analysis methods to explore single-cell and sub-population level effects as shown in Fig.6.First, the presence of oscillations in each ratiometric timecourse was inferred from spectrogram power in the 0.007-0.016Hz range (Fig.6(A)) (using detrended, max-normalized green fluorescence data).Each cell's average movement and oscillation power was then calculated for periods prior to, and after irradiation (averaging over 0 to 8 mins (pre) and 11-30 mins (post) for oscillations, and 2 to 8.7 mins (pre) and 12.6 to 26 mins (post) for movement).The same time windows were used on the un-irradiated control.Scatter plots in the top row (Fig.6(B)) show the change in oscillation and movement for every cell, before (pre) and after (post) irradiation for Dishes 1 and 2, and for the same time-periods for the control dish (Dish 3).Plots Fig.6(c)-(f) below shows scatter plots of subsets of cells, selected based on differences in their collective behavior: Fig.6(C) Cells whose mean oscillation power increased after irradiation (by more than 0.2), Fig.6(D) Cells whose mean oscillation power decreased after irradiation (by more than 0.15), Fig.6(E) Cells whose movement decreased by more than 0.05 and Fig.6(F) Cells whose movement increased after irradiation (by more than 0).The % of total cells in each group is also noted on each plot.Example timecourses of cells in each of these groups, for the two irradiated dishes (1 and 2) and the control (dish 3) are also shown.Although patterns are difficult to discern when plotting all cells together in Fig.6(B), several trends are clear from the sub-population plots.For example, in (Fig.6(C)) we can see that in both of the irradiated samples, cells whose calcium oscillations increase after irradiation show a clear shift from higher to lower levels of movement.Although cells in the control also show that increased calcium oscillations correspond to less movement, cells within this group transition to both higher and lower levels of movement with little net overall movement change.We infer that oscillatory states generally correspond to reduced cell motion (consistent with[58]), but that this behavior after irradiation may represent transition to a (potentially irreversible) impaired state that is accelerated / exacerbated by irradiation.Our observation that Dish 1 had a higher number of oscillatory cells initially, could imply that these cells were at a lower initial health state than cells in dishes 2 and 3.Although dishes 2 and 3 had very few cells with clear initial calcium oscillations (Fig.6(D) inset plots), to examine whether cells ever recover from being in an oscillatory state, we examined cells whose calcium oscillations decreased over time (Fig.6(D)).We found that almost all cells in this group also decreased their movement after irradiation.We interpret decreased movement in cells transitioning out of an oscillatory state as progressing to a worsened state of cell damage.Isolating the cells exhibiting the strongest decreases in movement (Fig.6(E)), we note that a higher fraction of cells make this transition in the irradiated dishes (14.7% and 21.1% compared to 7.5% in the control), further supporting that these cells have

Fig. 6 .
Fig. 6.Population and sub-population analysis of irradiation effects on intracellular calcium oscillations and movement.A. Illustrates the spectrogram of an oscillatory fluorescence signal showing how oscillations were quantified for each cell by averaging power between 0.007-0.016Hz (dotted white lines).B. Scatter plots and histogram distributions for the average oscillatory power and movement of all cells before (cyan) and after (pink) irradiation in two irradiated samples and one control (left to right -same dishes as in Fig.5).To analyze subpopulations more clearly, we plot in C-F scatter plots and histograms of sub-populations meeting different criteria, while providing example fluorescence traces from individual cells within each group for each dish (left).C. Shows cells with the largest increase in oscillations after irradiation, D: largest decrease in oscillations, E: any decrease in movement and F: largest increase in movement.The % of overall cells within each subgroup is indicated on the Y axis.