## Introduction

Considering physical dose enhancement only, high-Z nanoparticles are a natural choice since their photoelectric absorption cross-section, scaling approximately with Z4, is significantly higher than those of soft tissue or water9. However, the photoelectric tissue contrast is also strongly dependent on the energy of the incoming photons (~E−3). Hence, and in contrast to kV X-rays, only limited dose enhancement would be expected at higher energies (MV X-rays)9,10. In fact, at energies exceeding 500 keV, physical interactions are dominated by Compton scattering events with cross-sections linearly proportional to Z11. Therefore, it has been suggested that chemical and biological effects play pivotal contributing roles in the nanoparticle dose enhancement found in vitro and in vivo with MV photons7,9,12. There is increasing experimental and clinical evidence supporting nanoparticle-based radiotherapy enhancement in terms of safety and efficacy for both kV and MV photons8,13,14,15,16. Most notably, HfO2 nanoparticles marketed by Nanobiotix as NBTXR3/Hensify® have recently gained approval for the European market16. These HfO2 nanoparticles obtained European CE Mark approval in April 2019 for the treatment of locally advanced soft tissue sarcoma via intratumoral injection with photon radiotherapy and are being investigated for treatments of other cancers17.

While surprisingly little is known about nanoparticle radioenhancement mechanisms with clinical MV photon beams, nanoparticle dose enhancement using protons has been even less investigated. Protons can also be used as alternatives to photons in treating cancers and exhibit better dose conformation. As positively charged subatomic particles, protons interact differently with matter, leading to a distinctly different dose–depth profile compared to those of noncharged photons18,19. While the dose deposition of photons in depth is continuous and goes beyond the tumor, resulting in an “exit dose”, protons lose the majority of their energy in the range of the Bragg peak, after which they are stopped completely19. In the Bragg peak region, protons are slowed down, leading to an increase in the interaction probability with orbital electrons and the number of ionization events in the tissue. Finally, the proton is absorbed in a charge-changing process18. The position of the Bragg peak can be tailored to the location and size of the tumor volume, to minimize the dosage delivered to the surrounding tissue19. With regard to nanoparticle proton therapy enhancement, several Monte Carlo studies have investigated the influence of proton energy20, nanoparticle size and coating21, or nanoparticle clustering22 on the dose-enhancement mediated by Au nanoparticles. Additionally, in vitro dose-enhancement studies for Au nanoparticles used with proton irradiation have been performed, indicating dose enhancements in the range of 0–44%23,24,25,26,27. Moreover, Fe-based or Au nanoparticles in combination with proton irradiation were successfully used to achieve complete tumor remission in vivo28.

While the above studies show generally promising results for dose enhancement by nanoparticles, they also suggest a strong dependence on nanomaterial composition and the type of irradiation. This has also been indicated by the findings of Smith et al. (2015), who compared the dose enhancement by Au nanoparticles during X-ray (kV and MV) and proton irradiation in alanine electron paramagnetic resonance (EPR) dosimeters, showing an approximately 60% enhancement for kV, 10% for MV and 5% for proton irradiation29. Despite individual successes, including the clinical use of HfO2 nanoparticles, we lack a mechanistic understanding and thus fail to select nanoparticle materials for optimal dose enhancement under various irradiation settings (photons, protons) in an evidence-based manner.

Here, we present a conceptual investigation which disentangles key contributors to nanoparticle dose enhancement as a function of nanomaterial composition and beam characteristics. We synthesized a portfolio of nanoparticles made from different core materials (including TiN, TiO2, WO3, and HfO2, as well as Au and SiO2 as commercially available reference materials) and assessed their physical, chemical, and biological dose-enhancement capabilities in consistent settings for which observed differences can be directly attributed to nanomaterial properties as the only variable in the system. We demonstrate that Au nanoparticles show excellent dose enhancement in kV irradiation settings, that vanish almost completely in MV photon and proton therapy settings. In contrast, HfO2 nanoparticles retain some dose-enhancement capability in MV beam therapy settings, even though greatly reduced compared to kV irradiations. Most interestingly, catalytically active TiO2 nanoparticles retain their dose-enhancement effects in all three irradiation settings.

## Results and discussion

### Synthesis and physicochemical characterization of nanoparticle radioenhancer

Figure 1 shows transmission electron microscopy images (TEM) of the flame spray pyrolysis (FSP) synthesized (TiO2, TiN, WO3, HfO2) and commercially available (SiO2, Au) nanoparticles used for radioenhancement investigations. Primary particles had spherical morphologies, with the exception of TiN and WO3, which presented slightly elliptical shapes. The mean primary particle diameters based on TEM images (dTEM) were ~5 nm for TiO2 and HfO2, 10–15 nm for WO3 and TiN and 50 nm for Au nanoparticles (Table 1).

Crystal sizes obtained from XRD measurements (dXRD) were all in agreement with the primary particle diameter values found in TEM studies (dTEM, Table 1). Nitrogen adsorption measurements (BET) revealed specific surface areas (SSAs) and particle sizes (dBET), which were, again, in good agreement with the primary particle sizes dXRD and dTEM, indicating only limited sinter-neck formation or particle aggregation (Table 1). The hydrodynamic diameters (Z-average) of all flame-made nanoparticles in water were comparable with values between 100 and 130 nm (Table 1). The citrate stabilized Au nanoparticles showed a Z-average close to their primary particle size and comparable to those of the FSP-synthesized nanoparticles. Organic residues on the nanoparticle surface originating from nanoparticle synthesis were estimated using thermogravimetric analysis (TGA). All nanoparticles showed surface residues of ≤2 wt% (weight percent). Only TiO2 showed organic contents of up to 4 wt%, which could be further reduced by post-annealing. All particle types were well-dispersible in water.

### Cytocompatibility and cellular uptake

All oxide nanoparticles were generally well tolerated by human sarcoma cells (HT1080). Sham-irradiation (0 Gy) experiments revealed lethal LC50 values well above 160 µg/mL for all nanoparticles and exposure times of 5 days, except TiN, where the LC50 was 157 µg/mL (Supplementary Fig 1). Cellular uptake was investigated for the aforementioned nanoparticles with electron microscopy and elemental analysis. Figure 2 shows scanning electron micrographs of HT1080 cells after 24 h of incubation with the indicated nanoparticles. Nanoparticles were taken up by cells, most likely via an endocytic pathway30, and formed intracytoplasmic agglomerates, well in line with earlier studies of flame-made nanoparticles31. Few hundred nanometer up to micrometer-sized nanoparticle agglomerates were distributed within the cell cytoplasm (in vesicles or endosomes). In the >100 cells analyzed per nanoparticle type, no evidence for nanoparticle uptake into the nucleus was found, even though uptake overall, and nanoparticle accumulation in the nucleus, might be particle and cell type dependent32,33. Nanoparticle uptake was comparable for all types of oxides except for WO3 nanoparticles, for which only very few nanoparticle agglomerates were found intracellularly (Fig. 2d).

Elemental analysis based on inductively coupled plasma mass spectrometry (ICP MS) was employed to quantify the metal mass normalized to the cell number. For nanoparticle concentrations of 80–160 µg/mL, an uptake of ~1 ng of metal mass per HT1080 cell was found for most particles. ICP-MS results confirmed the observations from TEM images, and revealed that intracellular metal concentrations were two orders of magnitude lower for tungsten than for all other metals (Fig. 2g and Supplementary Fig. 2). Additional investigations confirmed the dissolution of WO3 nanoparticles in acidic, lysosome-mimicking conditions (Supplementary Fig. 3).

### Physical contributions to dose enhancement

Following the physicochemical characterization of the synthesized nanoparticle radio-enhancer candidates and cellular uptake studies, we performed Monte Carlo simulations using TOPAS/Geant4 to estimate the contribution that can be expected from physical dose enhancement as a function of core material and type of irradiation (150 kVp photons, 6 MV photons or 100 MeV protons). We investigated the dose deposition within and around an individual, nanoparticle filled vesicle, as well as within the cytoplasm and nucleus of a single cell. We extracted physical dose-enhancement factors (DEFs) by building the ratios of the dose scored to the cytoplasm, nucleus, vesicle, or water shells, respectively, in the presence of the nanoparticles to that with no nanoparticles (water). The geometries were built to match cellular uptake scenarios as closely as possible, with ~400 nm nanoparticle agglomerates distributed only within the cytosol (see also Fig. 2)31. As nanoparticle uptake into the nucleus was not observed experimentally, it was considered negligible also for the simulations.

### Nanoparticle-filled vesicle model—nanoscopic physical dose enhancement

Nanoscopic dose enhancement stems from secondary electron emission after ionization of a nanoparticle, with Auger electrons contributing to low range (~10 nm) and electrons of higher energy (Photo- or Compton electrons) contributing to micrometer range enhancements34,35. Figure 3a, b show the simulated dose enhancements for different nanoparticles within a nanoparticle-filled vesicle with 200 nm radius and within 50 nm water shells surrounding the vesicle. A DEF value >1 indicated additional dose deposition by nanoparticles compared to water, while a DEF value equal to 1 meant that no additional dose deposition was observed. For lower energy X-rays (150 kVp source), the highest dose deposition and a clear impact of atomic number were observed. This is in line with the different mass energy absorption cross-sections caused by the photoelectric effect for high-Z metals. The dose-enhancement factors within a nanoparticle-filled vesicle reached values of DEF = 30–40 for Au nanoparticles and DEF = 10–20 for HfO2 and WO3 nanoparticles at the highest reached nanoparticle content of 32.4 vol% (volume percent) in the vesicle (Fig. 3a). This packing fraction is also reasonable for biological scenarios. For instance, nanoparticle volume fractions of 35 ± 16% per vesicle have been reported in cells for 30-nm-sized Au nanoparticles36, and exposure conditions similar to the ones used in our study. Low-Z nanoparticles, such as TiO2, TiN, and SiO2 showed no nanoscopic dose increase at all. The dose enhancement decay from the filled vesicle surface followed a 1/r–type decay, and the DEF converged to DEF = 1 within one micrometer of the cytoplasm (Fig. 3b). For MV X-rays, dose enhancement within and around nanoparticle-filled vesicles was only found for Au nanoparticles (Supplementary Fig. 4). This enhancement was found to be even more localized, converging to DEF = 1 within 100 nm from the vesicle surface. Nanoscopic physical enhancement of proton irradiation was negligible for all nanoparticles (Supplementary Fig. 4b).

### Cell model—microscopic physical dose enhancement

The total DEF within the cytoplasm or nucleus increased linearly with increasing nanoparticle filling. This is in line with the calculation of the macroscopic DEF

$${{{{{\rm{DE}}}}}}{{{{{{\rm{F}}}}}}}_{{{{{{\rm{macroscopic}}}}}}}=1+{{f}}_{{Z}}\cdot {\left(\frac{{{\mu }}_{{{{{{\rm{en}}}}}}}\left({E}\right)}{{\rho }}\right)}_{{Z}}\bigg/{\left(\frac{{{\mu }}_{{{{{{\rm{en}}}}}}}({E})}{{\rho }}\right)}_{{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}}$$
(1)

where fZ is the atomic number (Z) mass fraction in the system and $${{{{{{\rm{\mu }}}}}}}_{{{{{{\rm{en}}}}}}}({E})/{\rho }$$ the mass energy-absorption coefficient at a monoenergetic photon energy, E37. Mass and volume fraction (fVol) are related via a constant density ratio of the materials. We therefore used a linear fit to describe the dose-enhancement efficiency of a nanoparticle (NP) per cellular nanoparticle volume fraction, $${{{{{{\rm{\chi }}}}}}}_{{{{{{\rm{NP}}}}}}}$$, within the cytoplasm and nucleus, with the relationship $${{\chi }}_{{{{{{\rm{NP}}}}}}}=({{{{{\rm{DEF}}}}}}-1)/{{f}}_{{{{{{\rm{Vol}}}}}},{{{{{\rm{NP}}}}}}}$$ (Supplementary Fig. 5).

It is evident that for Au, HfO2, and WO3 considerable additional doses were deposited in the cytoplasm and in the nucleus, but only at kV energies. In the cytoplasm, the DEF reached values of approximately 10.5, 5.0, and 3.7, while in the nucleus it reached 5.7, 2.9, and 2.3 per nanoparticle volume fraction percent of Au, HfO2, or WO3, respectively. Using the 6 MV X-ray spectrum, the dose-enhancement efficiency per nanoparticle volume fraction was roughly 10 times reduced, compared to the 150 kVp source. We found DEF values of 1.9 and 1.4 for Au, 1.3 and 1.2 for HfO2 and 1.2 and 1.1 for WO3 nanoparticles per volume percent in the cytoplasm and nucleus, respectively. For low-Z nanoparticles such as TiO2, TiN, or SiO2, no physical nanoparticle enhancement was detected, since no considerable additional dose was deposited in the cytoplasm, nucleus, or nanoparticle-filled vesicle for any type of irradiation source. No dose-enhancement effects were found for any of the investigated nanoparticles for the proton beam source (DEF = 1).

While no comprehensive study is available in the literature, data for Au nanoparticles are available from Rudek et al. for a comparable cell model. They reported DEFs of 2 and 3.5 in the nucleus and cytoplasm, respectively, for 100 kV monoenergetic X-rays and 5 wt% randomly distributed Au nanoparticles38. In our study these values were 2.2 (nucleus) and 3.5 (cytoplasm) for 0.26 vol% Au, which relates to 5 wt% Au, and are thus well in line with their data. Furthermore, we agree with their finding, that intracellular physical dose enhancement is rather negligible for protons.

### Chemical contributions to dose enhancement

For these measurements of chemical dose enhancement, the total metal content at all nanoparticle concentrations tested was below 0.3 wt% or below 0.05 vol%. At such low percentages, physical dose enhancement is negligible, especially for nanoparticles other than Au and for MV X-rays or protons (DEF < 1.2, see Fig. 3d–f). The ROS dose enhancement (DEFROS) was at least 10–100 times higher than the physical dose-enhancement effect per nanoparticle volume percent (Supplementary Fig. 5). Therefore, we concluded that the nanoparticle effects observed in this assay were based primarily on catalytic surface effects. To exclude effects from synthesis-related organic residues on the surface of TiO2 nanoparticles (Table 1), we conducted an annealing study using temperatures of up to 500 °C to remove such residues. DEFROS remained unaffected by annealing (Supplementary Fig. 6), and thus, we concluded that surface organic residues had no measurable influence on ROS generation under irradiation in our setting.

Catalytic processes have previously been postulated to explain the chemical enhancement, which was found to be several magnitudes higher (e.g., up to 2000 times)41 than what could be expected from purely absorption-related processes during irradiation of nanoparticle suspensions39,41. One such proposed catalytic mechanism involves the formation of a structured water layer around nanoparticles leading to weakened H–OH bonds and thereby to more efficient water radiolysis in the vicinity of nanoparticles39. Such a mechanism would be nanoparticle material independent. In our study, ROS amplification for comparable surface exposures was dependent on the material composition, indicating that the catalytic process at hand rather involves a charge transfer process specific to the nanoparticle composition. Semiconductor particles, with or without band-gap engineering, have been shown to possess photocatalytic properties, which can be harnessed, for example, to decompose organic molecules, such as methylene blue or ethylene42,43,44. TiO2 and WO3 are thought to be particularly effective for production of •OH radicals and photodegradation of organic molecules due to the relative position of their charge and valence band potentials45. Therefore, we concluded that ROS generation with TiO2 and WO3 was superior to that with HfO2 nanoparticles for all irradiation conditions (kV/MV photons and protons), due to their more beneficial energy band potentials.

For HfO2 nanoparticles, a clear impact of the radiation type on ROS enhancement was observed with enhancement ratios scaling roughly as 1:2:4 for (MeV protons):(MV X-rays):(kV X-rays) (Fig. 4e, f). The different catalytic efficiencies for kV versus MV X-rays may be related to higher absorption cross-sections at kV energies, suggesting also a higher excitation probability at such energies. Nanoparticle excitation leading to electron–hole pair generation is the fundamental process behind photocatalytic surface reactions46. We thus concluded that (i) proton irradiation led to lower nanoparticle excitation compared to X-rays, and (ii) the energy spectrum (150 kVp vs. 6 MV) of the X-rays had an influence on the electron–hole pair generation for high-Z nanoparticles (HfO2 and WO3) but not for TiO2 nanoparticles. Au nanoparticles also possess the ability to generate ROS under X-ray41,47,48 or proton49 irradiation. Using 150 kVp X-rays and acellular ROS assays, Au NPs were found to generate additional •OH radicals48. Physical and catalytic surface effects as well as an inverse size dependence have been suggested to play roles in the chemical enhancement by Au NPs41,47,50,51. Due to strong assay interference we were, however, not able to confirm these results for Au nanoparticles in our setting.

Taken together, the above investigations indicated a strong role for radio-catalytic processes in the chemical stage of nanoparticle radio-enhancement.

### In vitro effects of dose enhancement

There is a scarcity of studies comparing different irradiation sources on the same setup. Literature data exist exclusively for Au nanoparticles. Available studies for Au nanoparticles hint at strongly attenuated radioenhancement for MV X-ray and proton irradiation conditions, compared to that for kV X-rays. Our results are in line with those of Smith et al. (2015), who showed dose enhancements at 5 Gy of ~60% for 80 kVp X-rays, 10% for 6 MV X-rays and <5% for 150 MeV protons using alanine wax impregnated with 3 wt% of 5 nm Au nanoparticles29. Similarly, Chithrani et al. (2010) found enhancement effects for 50 nm Au nanoparticles of 66% with 105 kVp, which dropped to 17% with 6 MV X-ray irradiation using in vitro clonogenic assays52. Based on the literature, it is evident, that reported in vitro dose-enhancements observed for Au nanoparticles are most commonly in the range of 0–100% for low energy (kV) X-rays and well below 50% for MV X-rays12,53. Our results for 50 nm Au nanoparticles validate these observations and showed a very pronounced drop in radioenhancement efficacy in clinically relevant MV radiotherapy and proton therapy settings. These results further indicate the clear limitations of 50 nm Au nanoparticles and highlight the urgent need for high-performance nanoparticle radio-enhancers with rational core material selection and particle design for clinically relevant high-energy irradiation. These findings also validate the clinical use of HfO2 nanoparticle radio-enhancers and indicate even stronger benefits with catalytically active materials, such as TiO2, in combination with a high surface area per volume.

### Unraveling the dominant mechanisms under different irradiation conditions

In order to gain mechanistic insights, we correlated the in vitro dose-enhancement effects to the nanoparticle volume fraction (Supplementary Fig. 9) to compare relative nanoparticle performance and enable contextualization of our physical and chemical dose enhancement results. For kV X-ray irradiation, the relative trend in DMR can be very well predicted with the physical DEF simulations, well in line with earlier findings31. For this energy regime, the atomic number plays a crucial role contributing to physical effects (i.e. increased energy deposition within the cytoplasm and nucleus). Consequently, for kV photons, Au was the most efficient radio-enhancer, followed by HfO2, followed by TiO2 and TiN.

For MV photon beam irradiations, the atomic number dominance is overruled by the chemical enhancement activity. Highest dose enhancements under 6 MV X-ray irradiation were reached for TiO2 and HfO2. Strikingly, our results indicated that 50 nm Au nanoparticles were less efficient than the oxides, even at comparably higher volume fractions.

For proton irradiation, HfO2 and Au showed lowest radiation enhancement efficacies. Dose-enhancement efficacy of HfO2 nanoparticles with protons was 5 and 2 times lower compared to the ones with kV and MV photons, respectively, in line with chemical enhancement results (see Supplementary Fig. 4). Proton dose-enhancement efficacy of Au nanoparticles was comparable to the one with MV X-ray irradiation and ~20 times lower than the one with kV X-ray irradiation. Surprisingly, titanium-based materials outperformed the other materials showing higher proton enhancement than suggested by the physical or catalytic mechanisms presented here. This strong dose enhancement was likely caused by another mechanism specific to proton interactions with those elements. When natural titanium (48Ti) is bombarded with protons it can undergo a nuclear reaction via 48Ti(p,x)48V to produce the positron emitting radioisotope 48V54,55. Similarly, proton bombardment on nitrogen targets can yield the positron emitters 13N or 11C via 14N(p,pn)13N or 14N(p,α)11C reactions56. While positron emitters, as opposed to alpha emitters, are generally useful for diagnostics using positron emission tomography, recent studies have proposed that positrons might also be useful for cancer therapy57. Whether nuclear reactions or other biological mechanisms under proton irradiation enhanced Ti- and N-containing materials more than under X-rays in our case should be investigated further. Literature on proton enhancement using TiO2 or TiN nanoparticles is largely missing.

### Deciphering catalytic contributions

Ionizing radiation-induced cell damage is the result of cooperative contributions from indirect and direct cell stress mechanisms, including ROS generation, DNA damage, and subcellular organelle damage and autophagy58. Indirect actions are thought to be mediated mostly by hydroxyl radicals (•OH)59 and are estimated to contribute up to 90% to DNA damage or cell lethality60,61. For charged particles such as protons, a slightly more significant contribution of direct action to cell killing has been described, since they are more densely ionizing than X-rays27. •OH radicals can be quenched efficiently using dimethylsulfoxide (DMSO) to protect cells from X-ray and proton damage27,60,62. A maximum degree of protection was found for concentrations of approximately 1 M DMSO61. Here, we investigated the radioenhancement mechanism of nanoparticles by using DMSO in different concentrations to quench hydroxyl-mediated damage mechanisms.

Figure 6 shows the effects of radical quenching during kV X-ray irradiation. Radiation treatment in combination with DMSO and/or nanoparticles revealed very interesting effects. Nanoparticle concentrations were chosen such that radioenhancement effects were expected, while sham irradiation (0 Gy) effects were negligible (Supplementary Fig. 10). Indeed, all nanoparticles revealed dose-enhancement effects, which can be expressed as monodose nanoparticle radiation enhancement ratio (NER or RER63). We give this ratio as

$${{{{{\rm{NER}}}}}}=\frac{{{{{{\rm{S}}}}}}{{{{{{\rm{F}}}}}}}_{6{{{{{\rm{Gy}}}}}},{{{{{\rm{no\; NP}}}}}}}}{{{{{{\rm{S}}}}}}{{{{{{\rm{F}}}}}}}_{6{{{{{\rm{Gy}}}}}},{{{{{\rm{NP}}}}}}}}$$
(2)

where SF6Gy is the normalized cell surviving fraction at a dose of 6 Gy. For WO3 nanoparticles, the NER was ~1.6 with or without DMSO (Fig. 6f). For all other nanoparticles, increasing DMSO concentrations decreased the nanoparticle radiation-enhancement effect until effect saturation occurred. The complete protection from nanoparticle radioenhancement effects was observed with TiN nanoparticles and 0.1 M DMSO or more, showing a NER of 1.3 without DMSO and a NER close to 1.0 with DMSO. For TiO2, HfO2 and Au nanoparticles, the NERs decreased from 2.0, 2.1, and 1.6 (no DMSO) to 1.3, 1.5, and 1.4 at >0.5 M DMSO, respectively. In other words, addition of DMSO suppressed the total nanoparticle radiation-enhancement effects by up to 73% for TiO2, 52% for HfO2 and 34% for Au nanoparticles (Supplementary Fig. 11). These percentages reflect the cell damage enhancement mediated by •OH radicals. Short treatment with up to 1 M DMSO alone did not show negative effects on cell growth (Supplementary Fig. 10a). For nanoparticle-free samples, cell survival fractions of 70–80% were found after 6 Gy X-ray treatment, which were increased to 90% in presence of DMSO at concentrations of 0.1 M or higher (Fig. 6). These findings are in agreement with previous reports which amounted the percentage of indirect action to 63–89%60,61,62,64. The remaining, not DMSO suppressed nanoparticle dose enhancement must stem from other physical, chemical, or biological damage mechanisms.

TiO2 has been shown to activate cellular stress and reduce metabolic capacity65, elevate the levels of intracellular ROS, inflammation-related genes and apoptosis, and reduce GSH levels66. Increased DNA damage and micronucleus formation, a probable mechanism of genotoxicity, has also been reported67. Nevertheless, our results indicated that most (>73%) of the nanoparticle enhancement was due to production of ROS, especially •OH, during irradiation with ionizing radiation. Youkhana et al. (2017) also attributed the in vitro dose enhancement of TiO2 to increased ROS generation based on aqueous DCFDA results68. Here, we were able to show conclusively that this oxidative stress increased during irradiation most likely due to the catalytic surface effects of TiO2, since no physical dose enhancement was found in our nanoscopic and microscopic models.

DMSO addition only partly attenuated nanoparticle enhancement by high-Z nanoparticles HfO2 and Au. Thus, direct DNA damage might occur at kV X-ray irradiation for high-Z nanoparticles, which is consistent with our simulations of microscopic dose-deposition within the cell nucleus. Additionally, •OH radical formation in the cytosol might then be a combination of physical and surface catalytic processes for these high-Z materials at kV X-ray energies. Interestingly, cytoplasmic processes leading to the disruption of organelles, such as mitochondria or lysosomes, may play a major role in nanoparticle-mediated radioenhancement69,70. Most recently, it was also shown that even very low concentration of 10 nm Au nanoparticles can have an effect on cell cycle phase, the proportion of radiosensitive G2 cells as well as the DSB repair kinetics71. Thus, the nanoparticle-mediated radiation response is complex with sensitization of cancer cells as well as dose enhancement both contributing to the overall response.

The only nanomaterial for which DMSO showed no effect on the radioenhancement efficacy was WO3. Since in our case WO3 nanoparticles dissolved prior to irradiation, their action must be governed by low concentrations of free tungsten ions. Using soluble sodium tungstate, it has been shown that tungsten can alter cell cycle progression, stimulate cytokine profiles and increase DNA damage and apoptosis72,73,74. Furthermore, it has been reported that tungstate modulates the activation of the central double strand break (DSB) signaling kinase ATM and that sensitized cells to DNA DSB-inducing agents, such as ionizing radiation75. Such a biological sensitization mechanism could explain our findings that (dissolved) WO3 nanoparticles sensitized HT1080 cells under X-ray or proton irradiation to similar extents without increasing oxidative (•OH radical) stress.

We postulate the following mechanisms of action for the nanomaterials used in this study:

Silicon dioxide: SiO2 is a biologically relatively inert nanomaterial that does not produce relevant ROS or cell sensitization or lead to observable dose enhancement effects. It can be used as a control for further radioenhancement studies.

Titanium nitride: No physical dose enhancement in the cytoplasm or nucleus was found for TiN nanoparticles. TiN might reduce the antioxidant capacity and sensitize cancer cells to ionizing radiation. With low concentrations of DMSO, the dose enhancement effect was diminished completely. Under proton irradiation, however, nuclear effects of 48Ti and 14N atoms may play additional roles leading to increased proton dose enhancements observed in vitro.

Titanium dioxide: The in vitro dose enhancement by TiO2 nanoparticles under X-ray irradiation did not show an energy dependence. Any physical dose enhancement in vesicles, cytoplasm or nucleus can be excluded. Ionizing radiation generated cellular oxidative stress via catalytic reactions, which were higher under X-ray than under proton irradiation. An additional biological effect (<30%) might play a significant role, since most but not all dose enhancement was quenched by DMSO. High availability of nanoparticles per cell were necessary to show high dose enhancement effects. This was allowed by the low toxicity profile. Under proton irradiation, nuclear effects may play an additional role in further enhancing cell damage.

Tungsten oxide: WO3 nanoparticles can generate physical dose enhancement under kV X-ray irradiation and surface catalytic ROS under all ionizing irradiation sources used. However, nanoparticle dissolution—likely in lysosomal compartments—led to low tungsten content per cell and to a rather biological than oxidative sensitization of the cells toward X-ray and proton irradiation.

Hafnium dioxide: HfO2 nanoparticles enhanced the dose within the cell nucleus and enhanced cellular oxidative stress via dose deposition in the cytoplasm as well as via surface catalytic reactions under kV X-rays. Catalytic processes were reduced under MV X-ray irradiation and were lowest under proton irradiation. In addition to a very low toxicity profile, HfO2 nanoparticles showed proof of in vitro dose enhancement governed by physical and chemical phenomena under photon and proton irradiation.

Gold: 50 nm Au nanoparticles showed high dose enhancement effects at kV energies, which can be justified partly (<35%) by indirect (•OH radical mediated) and mostly by direct (dose to nucleus or cytoplasm) dose enhancement effects. These effects were strongly reduced with ionizing radiation at MV energies. The applicable nanoparticle surface of a 50 nm nanoparticle is 100 times less than that of a 5 nm nanoparticle, which reduces possible dose enhancement effects based on catalytic surface processes compared to all other nanoparticles.

Taken together, the results of this comprehensive study allow extraction of important general design principles for nanoparticle radioenhancers for kV and MV photon and proton therapies based on quantitative and comparative data. While nanoparticle uptake, cellular toxicity, and radio-enhancement are cell-line dependent, previous work on different cancer cell lines has indicated that the relative trends in radioenhancer effectiveness hold true, albeit with slightly different absolute values31. Future investigations should focus on further exploration of the materials design space based on such comparative, well-standardized settings, and validation of the radioenhancement performance in animal models and with different cell types. Potential effects of nanomaterial surface functionalization on dose-enhancement should be investigated carefully, including potential ROS quenching by antioxidant molecules (such as dopamine), as well as potentially synergistic effects leading to augmented ROS generation, e.g. by porphyrins. Eventually, the effective nanoparticle concentration reached in the cancer cells will govern dose enhancement. While experimental research in mouse models on intravenously injected nanoparticles has indicated that only a small fraction of the injected nanoparticles may accumulate inside cancer cells, intratumoral administration of HfO2 nanoparticles (NBTXR3) appears to partially overcome the delivery problem and shows convincing therapeutic effectiveness in preclinical and clinical settings. This effectiveness may, however, be even further improved by optimizing radioenhancer designs, in parts based on the insights provided by this work.

## Methods

### Monte Carlo simulations

To estimate the impact of nanoparticles on energy deposition, physical stage simulations were performed using the TOPAS-nBio76,77 extension and the TOPAS78,79 toolkit, which is based on the Geant480 Monte Carlo simulation system. We used the default Geant4-DNA track structure physics list81 for the simulation of particle transport within the cell including the full Auger de-excitation cascade process35 with fluorescence, Auger electron production, and particle-induced X-ray emission turned on. The energy range minimum was set to 10 eV. Since this physics module supports only transport in H2O, we used the condensed-history EMStandardOption4 physics list in all nanoparticle regions, including nanoparticles made out of water, and specified a 1 nm range cut for all particles. The dose-enhancement factor (DEF) was defined as the ratio of the dose scored to the cytosol, nucleus, or to water shells in the presence of the nanoparticles of interest to that with water nanoparticles. Simulations using X-rays were performed on a MacOS Catalina distribution using TOPAS version 3.5 using triplicates of 108 histories. The X-ray spectrum of the 150 kVp source was simulated using the XRayGUI (Version 1.4.2.0, BAM Federal Institute for Materials Research and Testing, Berlin, Germany) matching experimental conditions as closely as possible. For the X-ray spectrum of the 6 MV source, already published particle fluence within a water phantom from a Varian Linac was extracted from Choi et al. (2019)82. Proton simulations with a 100 (±1%) MeV beam were performed on a CentOS7 Linux distribution on the hpc Euler cluster of ETHZ using TOPAS version 3.7 and triplicates of 3.33 × 105 histories.

For the nanoscopic simulations, a simple vesicle model consisting of a 400 nm sized water sphere and randomly distributed 50 nm nanoparticles was centered in a water cube of 20 µm side length (X-rays) or 6 µm side length (protons). The dose inside the vesicle was scored as well as the dose around the vesicle within 100 nm water shells up to a distance of 1 µm. The beam size was 4 µm, ensuring an even irradiation field also around the vesicle. 50 nm sized nanoparticles were chosen, since for physical stage simulations smaller-sized nanoparticles only have a comparably small influence on the surrounding energy deposition38, however, they increase computation time tremendously. The materials of the nanoparticles were defined by using the metal and oxygen or nitrogen mass content as given in Supplementary Table 1. For proton simulations it was important to only score and compare the dose in the nanoparticle-free space. For the microscopic dose-enhancement simulations, a simple cell geometry consisting of an outer water sphere with a diameter of 6 µm and an inner water sphere with a diameter of 3 µm was modeled representing the cytosol and the nucleus, respectively. This was close to the cytosol/nucleus ratio of 1.7 as used by Lin et al.83 and Rudek et al.38. For X-ray simulations the cell was centered in a cubic water box of 20 µm side length, while for proton simulations it was centered in a smaller, 6 µm long cubic water box to reduce computation time. In the cell model all nanoparticles were placed in vesicles of 400 nm diameter at a pre-defined position and a volume fraction of 32.4%. Different amounts of such nanoparticle filled vesicles were then placed in the cytoplasm only, because metal oxides or Au nanoparticles enter cells predominantly by endocytotic pathways and are clustered within roughly 300–500 nm-sized vesicles within the cytoplasm, rarely entering the cell nucleus31,84. The particle beam (X-ray or proton) had the same diameter as the cell.

### Materials

The following metal precursors and solvents were used: Titanium (IV) isopropoxide (Sigma-Aldrich, 97%), hafnium (IV) isopropoxide isopropanol adduct (Alfa Aesar, 99%), ammonium (meta)tungstate hydrate (Sigma-Aldrich, ≥85.0%), 2-ethylhexanoic acid (2-EHA) (Sigma-Aldrich, 99%), xylene (Sigma-Aldrich, ≥98.5%), ethanol absolute (VWR chemicals, ≥99.8%), diethylene glycol monobutyl ether (Sigma-Aldrich, ≥98.0%). 50 nm Au nanospheres (Citrate, BioPureTM) in USP purified water (1 mg/mL) were purchased from nanoComposix Inc (San Diego, USA). Hydrophilic fumed silicon dioxide nanoparticles (Aerosil®, A380) were obtained from Evonik Industries AG (Essen, Germany).

### Nanoparticle synthesis

TiO2, HfO2, and WO3 nanoparticles were synthesized using flame spray pyrolysis, as described in detail elsewhere31,85,86. The precursor to produce TiO2 nanoparticles was made by stirring titanium isopropoxide and xylene at a metal concentration of 0.16 M for 30 min. Likewise, the WO3 precursor solution was prepared by mixing ammonium (meta)tungstate hydrate with diethylene glycol monobutyl ether and ethanol (1:1) at a metal concentration of 0.2 M. In the case of HfO2, hafnium isopropanol adduct was first dissolved in 2-EHA by stirring at 120 °C under reflux cooling for several hours at 0.96 M. After cooling down, it was diluted 1:5 with xylene, reaching a final metal concentration of 0.16 M. The liquid precursor solutions were fed through a capillary with flow rates of 3 mL/min (HfO2, TiO2) or 5 mL/min (WO3) and dispersed by O2 (PanGAS, purity > 99%) with a flow rate of 5 L/min into fine droplets. The pressure drop at the capillary was kept constant at 1.6 bar. A premixed ring-shaped CH4/O2 flame (1.5 L min−1/3.2 L min−1) ignited and stabilized the spray flame. Particles formed in the gas phase were collected on a glass fiber filter (Type GF6, Hahnemühle FineArt GmbH) with the aid of a vacuum pump (Busch Mink MM 1202 AV). The collected nanoparticle powder from the filter was subsequently sieved (mesh size = 250 μm) to remove any filter residues. Titanium nitride (TiN) nanoparticles were made from FSP TiO2 nanoparticles following a previously established nitridation method87. Titania particles were nitrided on quartz wool in a U-shaped quartz reactor under pure ammonia flow (75 mL/min) and 700 °C heat treatment. The first heating rate up to 600 °C was 20 °C/min, followed by a second heating rate of 3 °C/min until the target temperature of 700 °C was reached and held for 30 h. After heat treatment the powder was cooled down to room temperature with a rate of 40 °C/min and soft oxidized (using 5% oxygen in argon) at room temperature.

### Nanoparticle characterization

Transmission electron microscopy (TEM) imaging was performed on a JEOL 2200FS TEM operated at 200 kV. TEM samples were prepared by drop-casting dispersions of 100 μg/mL in ultrapure water or EtOH onto carbon-coated grids (200 mesh copper, EM Resolutions). X-ray diffraction patterns were obtained with a Bruker 2D Phaser with a step size of 0.01°. Rietveld refinement was done using Profex88 (Version 4.2.5). XRD patterns as well as refinement results are plotted in Supplementary Fig. 12. The particles’ specific surface area (SSA) was measured based on N2 adsorption at 77 K using the Brunauer–Emmett–Teller (BET) (Tristar II Plus, Micrometrics) method. Prior to the measurement, all samples were degassed at 150 °C in N2 for at least 45 min. The primary particle diameter, dBET, was estimated using equation $${{d}}_{{{{{{\rm{BET}}}}}}}[{{{{{\rm{nm}}}}}}]=\frac{6000}{{{{{{\rm{SSA}}}}}}\left[{{{{{{\rm{m}}}}}}}^{2}{{{{{{\rm{g}}}}}}}^{-1}\right]{{{{{\rm{\rho }}}}}}[{{{{{\rm{g}}}}}}\; {{{{{\rm{c}}}}}}{{{{{{\rm{m}}}}}}}^{-3}]}$$. Bulk material densities are given in Table 1. Hydrodynamic size measurements were acquired via dynamic light scattering (DLS) on a Zetasizer Nano ZS90 (Malvern Instruments Ltd., Worcestershire, UK) at a 90° scattering angle and a concentration of 0.1 mg/ml in ultrapure water. All nanoparticle suspensions were bath sonicated for at least 15 min prior to the measurement. To assess organic surface content, thermogravimetric analysis (TGA) was performed with a NETZSCH TG 209 F1 instrument (NETZSCH-Gerätebau GmbH, Selb, Germany) with a heating rate of 10 °C/min under nitrogen flow. The TGA as well as their first differential (DTG) can be found in Supplementary Fig. 13.

### Cell lines and culture conditions

Human fibrosarcoma HT1080 cells (ATCC®CCL121TM) were cultured in minimum essential medium Eagle (MEM, Sigma-Aldrich or Gibco) supplemented with 10% fetal bovine serum (FBS, Sigma-Aldrich), 1% l-glutamine (Sigma-Aldrich), 1% penicillin–streptomycin (PS, Sigma-Aldrich), 1% non-essential amino acids (NEAA, Sigma-Aldrich), and 1 mM sodium-pyruvate at 37 °C under a humidified atmosphere containing 5% CO2. Subculturing was conducted at 70–80% confluency by treatment with 0.5% Trypsin-EDTA (Sigma-Aldrich).

### Cell sample preparation for STEM imaging

To image intracellular nanoparticles, 1.4 × 105 cells were seeded in T25 flasks in 8 mL medium and allowed to attach overnight. 700 μL of a 1 mg/mL nanoparticle suspension in ultrapure water were added. Following an incubation time of 24 h, the cells were washed twice with PBS, detached with 1 mL of Accutase (Sigma-Aldrich), washed with PBS and fixed overnight with 2.5% glutaraldehyde and 4% paraformaldehyde in PBS. The cell pellets were then washed with 0.1 M cacodylate buffer for 20 min and stained with 2% osmium tetroxide and 1.5% potassium ferricyanide for 1 h. Gradual dehydration via an ethanol gradient (50%, 70%, 80%, 90%, 100%, 100%, 100%) was followed by embedding into an epoxy resin (EPON 812 kit, Sigma-Aldrich). This was accomplished by mixing and adding resin and ethanol 1:2 (5 days), 1:1 (3 h), 1:0 (3 h), 1:0 (24 h) to the pellets. The resin blocks were then hardened in an oven at 60 °C for 48 h. Afterwards, thin sections of ~100 nm were cut from the blocks using an ultramicrotome (Scope M). The sections were imaged using a Hitachi S-4800 scanning electron microscope operated in transmission mode at 20 or 30 kV.

### Cell viability assessment and number of cell quantification

The viability of cells was using the CellTiter-Glo (CTG) assay (CellTiter-Glo® Luminescent Cell Viability Assay, Promega, G7571) according to the manufacturer’s specifications but adapted to our set-up. CTG buffer and substrate were thawed and mixed, and 300 µL of cell medium (from the experimental wells) was replaced by 200 µL CTG reagent. After incubation in the dark on a shaker for 20 min and equilibration for 30 min in the dark, luminescence (integration time 1 s) was measured with a microplate reader (Mithras LB 943 Multimode). An in-house made black titanium adapter was applied to eliminate crosstalk between the wells. This method was established by comparing the luminescence from 20,000 cells per well with and without adapter (Supplementary Fig. 14a). It was found that the adapter eliminated luminescence cross-talk, which led to an even distribution of the luminescence signal on the plate. A luminescence standard deviation of 5.0% with adapter compared to 13.9% without adapter was achieved. A cell standard curve with N = 3 independent biological experiments with duplicates was performed and with this we were able to convert the luminescence into number of cells (#cells) via: $${{{{{\rm{\#cells}}}}}}=x=\frac{y{b}}{{a}-{b}}$$, where y is the luminescence signal and a = 3.872 × 106 and b = 3.431 × 105 are fitting parameters (R2 = 0.991) (for more details see Supplementary Fig. 14b).

### Acellular ROS assay and chemical dose-enhancement quantification

The DCF assay was adapted from Zhao and Riediker (2014)89. 2′,7′-dichlorodihydrofluorecein diacetate (H2DCF–DA) powder (Sigma-Aldrich) was dissolved in dimethyl sulfoxide (DMSO, Sigma-Aldrich) to reach a concentration of 5 mM. Aliquots of this were stored at −20 °C for further use. The thawed H2DCF-DA stock solution was mixed with NaOH (10 mM) 1:4 and incubated in the dark for 30 min. Thereafter, it was diluted with Tris–HCl buffer (0.1 M, pH = 7.4) to 8 μM. The final H2DCF working solution was kept on ice in the dark at all times. All nanoparticles were weighted, dispersed and bath sonicated in Tris–HCl buffer reaching 10× stock concentrations. Prior to irradiation, 0.1 mL Tris–HCl buffer, 0.1 mL nanoparticle-suspension and 0.8 ml H2DCF working solution were added to the wells of two transparent 24-well plates identically. Final nanoparticle concentrations were of 36.8, 368, and 1840 cm2/mL. One plate served as 0 Gy reference plate, the other plate was irradiated in the phantom with 12 Gy. Following irradiation, the liquid from the wells was centrifuged in microtubes and triplicates of the supernatant were transferred to transparent 96-well plates. The fluorescence signal was measured using a microplate reader (485 nm excitation, 535 nm emission, Tecan infinite 200Pro or Mithras LB 943 Multimode). The average 0 Gy fluorescence values, Fl0Gy, of each particle or blank were subtracted from the 12 Gy averages, Fl12Gy. The dose-enhancement factors (DEFROS) were then calculated using the fluorescence intensity (FI) as follows:

$${{{{{\rm{DC}}}}}}{{{{{{\rm{F}}}}}}}_{{{{{{\rm{ROS}}}}}}}=\frac{{{{{{\rm{F}}}}}}{{{{{{\rm{I}}}}}}}_{12{{{{{\rm{Gy}}}}}}\;{{{{{\rm{with}}}}}}\; {{{{{\rm{NP}}}}}}}-{{{{{\rm{F}}}}}}{{{{{{\rm{I}}}}}}}_{0{{{{{\rm{Gy}}}}}}\;{{{{{\rm{with}}}}}}\;{{{{{\rm{NP}}}}}}}}{{{{{{\rm{F}}}}}}{{{{{{\rm{I}}}}}}}_{12{{{{{\rm{Gy}}}}}}\;{{{{{\rm{without}}}}}}\;{{{{{\rm{NP}}}}}}}-{{{{{\rm{F}}}}}}{{{{{{\rm{I}}}}}}}_{0{{{{{\rm{Gy}}}}}}\;{{{{{\rm{without}}}}}}\;{{{{{\rm{NP}}}}}}}}$$
(3)

### Cell irradiation and in vitro dose-enhancement quantification

2000 HT1080 cells in 300 µL cell medium were seeded in 48 well plates and led to adhere for 24 h. After this, cells were incubated with 200 µL of different nanoparticle or control solutions to reaching final nanoparticle concentrations of 0–320 µg/mL. To achieve different nanoparticle concentrations for cell incubation, solutions were prepared from a fresh and 30 min bath sonicated 3.2 mg/mL nanoparticle-milliQ stock solution which was added in different volumes to a fixed amounts of cell medium. MilliQ was used to adjust the solutions to achieve the right concentration and a fixed amount (10%) of water in every solution. Nanoparticle free solutions consisted also of cell medium and 10% milliQ water. After 24 h cells were washed twice using 250 µL PBS. 500 µL cell medium was added thereafter. For the transport to and from the different irradiation facilities cells were kept in a cooled box. Irradiation was performed at room temperature. After the irradiation procedure, cells were incubated at standard conditions and medium was changed 2 days after. On the 5th day after irradiation, cell viability was assessed using the CellTiter-Glo® assay. One biological repeat per nanoparticle concentration consisted of triplicates (n = 3), and for control cells of sextuplicates (n = 6). To quantify in vitro dose enhancement, first the luminescence from the viability assessment was translated into number of cells (#cells) as stated above. From that the surviving fraction, SF, of cells following the irradiation dose responses of 0, 2, 4, 6, and 8 Gy, was calculated as

$${{{{{\rm{SF}}}}}}(x{{{{{\rm{Gy}}}}}})=\frac{{{{{{\rm{\#cell}}}}}}{{{{{{\rm{s}}}}}}}_{x{{{{{\rm{Gy}}}}}},{{{{{\rm{with}}}}}}\;{{{{{\rm{or}}}}}}\;{{{{{\rm{without}}}}}}\;{{{{{\rm{NPs}}}}}}}}{{{{{{\rm{\#cell}}}}}}{{{{{{\rm{s}}}}}}}_{0{{{{{\rm{Gy}}}}}},{{{{{\rm{with}}}}}}\;{{{{{\rm{or}}}}}}\;{{{{{\rm{without}}}}}}\;{{{{{\rm{NPs}}}}}}}}$$
(4)

The data was then fitted using MATLAB (MathWorks Inc., MA, US) to the linear quadratic model

$${{{{{\rm{SF}}}}}}(D)={{{{{{\rm{e}}}}}}}^{-({{{{{\rm{\alpha }}}}}}D+{{{{{\rm{\beta }}}}}}{D}^{2})}$$
(5)

where D is the irradiation dose in Gy and α and β are fitting parameters. The dose-modifying factor was extracted by finding the lethal dose leading to 50% survival (LD50%) with any particle type and comparing it to the one without nanoparticles:

$${{{{{\rm{DM}}}}}}{{{{{{\rm{R}}}}}}}_{50 \% }=\frac{{{{{{\rm{L}}}}}}{{{{{{\rm{D}}}}}}}_{{50 \% ,}{{{{{\rm{without}}}}}}\;{{{{{\rm{NPs}}}}}}}}{{{{{{\rm{L}}}}}}{{{{{{\rm{D}}}}}}}_{{50 \% ,}{{{{{\rm{with}}}}}}\;{{{{{\rm{NPs}}}}}}}}$$
(6)

Sham-irradiation toxicity of nanoparticles was defined by normalizing the number of cells treated with nanoparticles to the number of control cells treated with no nanoparticles and irradiated with 0 Gy:

$${{{{{\rm{Normalized}}}}}}\; {{{{{\rm{viability}}}}}}=\frac{{{\#}{{{{{\rm{cells}}}}}}}_{0{{{{{\rm{Gy}}}}}},\,{{{{{\rm{with}}}}}}\; {{{{{\rm{NPs}}}}}}}}{{\#}{{{{{\rm{cells}}}}}}_{0{{{{{\rm{Gy}}}}}},\,{{{{{\rm{without}}}}}}\; {{{{{\rm{NPs}}}}}}}}$$
(7)

To be included in the DMR50% analysis, a sham irradiation cell viability cut-off of 60% (lowest limit) was defined for all nanoparticles. A sigmoidal curve with fitting parameters a and b in the form of

$${{{{{\rm{Normalized}}}}}}\; {{{{{\rm{viability}}}}}}=1-\frac{1}{1+{a}{{{{{{\rm{e}}}}}}}^{-{b}x}}$$
(8)

was fitted to the cell-viability data to extract nanoparticle LD50 doses (x).

### Cell digestion and nanoparticle uptake quantification using ICP-MS

To analyze nanoparticle uptake into cells, cells were seeded and treated in duplicates in the same way as in a typical irradiation experiment. After 24 h nanoparticle treatment, cells were washed twice with 250 µL PBS. Thereafter, instead of adding cell medium, cells were trypsinized (80 µL). Trypsination was stopped using 220 µL medium and cells were transferred into Eppendorf tubes and stored at −20 °C until further analysis. To acquire the cell number, three experimental wells with control cells were pooled, centrifuged at 200 × g for 5 min and counted using a hematocytometer. This was repeated twice. Prior to ICP-MS analysis, cells were digested in 1 mL HNO3, 3 mL HCl and 0.5 mL HF using a 1 h bath sonication treatment. After digestion, samples were filled up to 50 mL using MilliQ water. For ICP-MS elemental standard curves, ionic metal standards (Au, Hf, W, and Ti) were prepared in the same matrix as the samples in concentrations from 0.01 to 10 ppb. A 1 ppb quality control (IV4 or IV100, Inorganic Ventures, Christiansburg, VI, USA) served as reference and ensured correctness of the standard curve. Measurements were performed on an Agilent 7900 ICP-MS (Agilent Technologies, Santa Clara, USA). Non-spectral interference effects were corrected using 100 ppb Sc, Ge, In, and Lu mixed on-line with the sample via a t-piece at a ratio of 1:10. All elements were measured in Helium mode. Elemental masses were then normalized to cell number and further translated into nanoparticle volume fraction using the atomic mass fraction as well as the nanoparticle bulk density as specified in Table S1 of the Supplementary information. The nanoparticle volume fraction was then corrected using the sham-irradiation viability fit from Supplementary Fig. 1. The volume of a cells was defined as 2800 µm3 using the diameter of 17.5 µm as given by a coulter counter. The relationship of metal mass per cell (f(x)) and nominal nanoparticle concentration (x) was described using a non-linear fit with fitting parameters A, B, and p in the form of

$${f}\left(x\right)=\frac{{A}{x}^{{{{{{\rm{p}}}}}}}}{{{B}}^{{{{{{\rm{p}}}}}}}+{{x}}^{{{{{{\rm{p}}}}}}}}$$
(9)

### Hydroxyl radical scavenging using DMSO

$${{{{{\rm{DoP}}}}}}=1-\left(\frac{{{{{{\rm{S}}}}}}{{{{{{\rm{F}}}}}}}_{6{{{{{\rm{Gy}}}}}},\;{{{{{\rm{DMSO}}}}}},\;{{{{{\rm{no}}}}}}\; {{{{{\rm{NP}}}}}}}}{{{{{{\rm{S}}}}}}{{{{{{\rm{F}}}}}}}_{6{{{{{\rm{Gy}}}}}},\;{{{{{\rm{DMSO}}}}}},\;{{{{{\rm{NP}}}}}}}}-1\right)\Bigg/\left(\frac{{{{{{\rm{S}}}}}}{{{{{{\rm{F}}}}}}}_{6{{{{{\rm{Gy}}}}}},\;{{{{{\rm{no\; DMSO}}}}}},\;{{{{{\rm{no}}}}}}\;{{{{{\rm{NP}}}}}}}}{{{{{{\rm{S}}}}}}{{{{{{\rm{F}}}}}}}_{6{{{{{\rm{Gy}}}}}},\;{{{{{\rm{no}}}}}}\;{{{{{\rm{DMSO}}}}}},\;{{{{{\rm{NP}}}}}}}}-1\right)$$