Nanoparticles as multimodal photon transducers of ionizing radiation

Abstract

In biomedical imaging, nanoparticles combined with radionuclides that generate Cerenkov luminescence are used in diagnostic imaging, photon-induced therapies, and as activatable probes. In these applications, the nanoparticle is often viewed as a carrier inert to ionizing radiation from the radionuclide. However, certain phenomena such as enhanced nanoparticle luminescence and generation of reactive oxygen species cannot be explained by only Cerenkov luminescence interactions with nanoparticles. Herein, we report methods to examine the mechanisms of nanoparticle excitation by radionuclides, including interactions with Cerenkov luminescence, β particles, and γ radiation. We demonstrate that β scintillation contributes appreciably to excitation and reactivity in certain nanoparticle systems and that excitation of nanoparticles composed of large atomic number atoms by radionuclides generates X-rays, enabling multiplexed imaging through single photon emission computed tomography. These findings demonstrate practical optical imaging and therapy using radionuclides with emission energies below the Cerenkov threshold, thereby expanding the list of applicable radionuclides.

Ionizing radiation sources such as linear accelerators and more commonly radionuclides are often combined with nanoparticles (NPs) for medical applications^1–5. Radiolabelled NPs have found use in clinical^{6, 7} and preclinical sentinel lymph node imaging^{4, 8}, long-term NP integrity studies⁹, and tumour imaging through active targeting and the enhanced permeability and retention mechanisms^{10, 11}. While radiolabelled NPs have been typically detected using single photon emission computed tomography (SPECT) or positron emission tomography (PET)¹² there is increasing interest in utilizing Cerenkov luminescence (CL)^{13, 14}. CL is the visible blue-weighted light that emanates when a charged particle, such as a β particle, travels faster than the phase velocity of light in a particular medium. Since first shown in preclinical in vivo systems in 2009¹⁵, Cerenkov luminescence imaging (CLI) is being rapidly translated to the clinic^16–21.

Imaging and therapeutic applications using ionizing radiation combined with NPs have generally neglected direct interactions between the ionizing radiation and the NP. Research has focused on CL generated in the medium interacting with fluorescent NPs such as quantum dots for secondary Cerenkov-induced fluorescence imaging (SCIFI)²². CL research has extended into TiO₂ NP-based photodynamic therapy (PDT) with both radionuclides³ and external beams²³. These studies explained the mechanism of PDT as CL from the ionizing radiation source exciting the NP-based PDT agent, generating reactive oxygen species. However, the very low CL photon fluence predicted from radionuclides (nJ cm⁻²) compared to external beam radiation (mJ cm⁻²)²⁴ strongly suggests that additional NP excitation mechanisms are necessary for therapy.

Ionizing radiation results in the excitation and ionization of the surrounding medium, in this case the NP, as energy is dissipated to reach thermal equilibrium²⁵. One mechanism by which NPs can relax to the ground state is through radiative processes, such as scintillation. γ-scintillation (photon production from γ-emitters) has been studied in quantum dots²⁶ and rare-earth NPs^27–31 (Fig. 1a). The contribution of β energy on NP systems has been neglected as β⁺ radionuclides in biomedical imaging (such as ¹⁸F and ⁶⁸Ga) also ultimately emit γ photons from positron annihilation in addition to CL. Common β⁻ (i.e. electron) emitting radionuclides such as ⁹⁰Y and ³²P also emit CL, thereby obscuring each mechanism contribution. Previous studies relied on physical shielding to block specific types of interactions (CL, γ, β) between a source and target. This approach led to the conclusion that Eu₂O₃ NPs were not excited by β interactions from the ¹⁸F positron²⁷. However, due to the short positron range of ¹⁸F in water³² (β_avg =0.25 mm, β_max = 2.0 mm) most positrons annihilated prior to interacting with the NPs, limiting the photons observed.

Figure 1.

Interactions between ionizing radiation and nanoparticles. a, gamma and visible photon sources are generally considered the excitation mechanisms of NPs by ionizing radiation that produces visible light through gamma scintillation or by Cerenkov radiation energy transfer, respectively. b, A more complete view of mechanisms resulting in excitation and ionization of NP from both β and γ radionuclides c, the fuller understanding of the consequences of NP excitation by ionizing radiation enables both optical and SPECT imaging with high atomic number (Z) NPs.

Here, we report new assays using radionuclides that do not emit CL to determine β and γ interactions with NPs. We demonstrate that β interactions significantly contribute to excitation of widely used nanoparticles such as Eu₂O₃ and TiO₂, investigated for imaging²⁷ and therapy³, respectively. We furthermore demonstrate that ionizing radiation in the presence of NPs containing high atomic number (Z) elements results in emission of characteristic X-rays in the 1–100 keV range. These X-rays allow for a new strategy in biomedical imaging using a SPECT scanner for multiplexed, NP-specific in vivo imaging. Our investigation into the various interactions between ionizing radiation with NPs better explains previous imaging and photodynamic therapy results in addition to a new imaging paradigm.

Visible photon output from radionuclide-NP pairings

Since clinical radionuclides decay through multiple high-energy emissions (Table 1), we first aimed to delineate NP excitation mechanisms that result in output of visible light. Radionuclides that emit only β⁻ or γ were selected as excitation sources and added to a library of NPs that are of interest for biomedical imaging (Table 2), including amorphous silica NP (SNP), TiO₂, HfO₂, Eu₂O₃, Gd₂O₃, and YAG:Ce (cerium-doped yttrium aluminium garnet). The pure excitation sources were combined directly with either the NP suspension or an H₂O control solution. The pure β⁻ emitters ³⁵S and ³H were used since the maximum energies (β_max) are below the CL threshold in water. The γ emitter ^99mTc was used to assess contributions from γ excitation. Although ^99mTc and ³⁵S do not result in CL, ionization and excitation of H₂O³³ can produce photons but are several orders weaker than CL.

Table 1.

Radionuclides used with corresponding decay mechanisms and energies. Information populated from the Nuclear Science References database⁴⁹. γ± denotes positron 511 keV photons upon annihilation.

Radionuclide	β decay	β energy (average keV)	γ emission	γ energy (average)	In vivo imaging modality
3H	β−	5.68	-	-
35S	β−	48.76	-	-
177Lu	β−	149.35	γ	208	SPECT, CL
32P	β−	695.03	-	-
18F	ε+β+	249.8	γ±	-	PET, CL
89Zr	ε+β+	395.5	γ, γ±	909	PET, CL
68Ga	ε+β+	836.02	γ, γ±	1077	PET, CL
90Y	β−	933.7	γ	1761	CL
99mTc	-	-	γ	140	SPECT

Table 2.

NP characterization information. Dynamic light scattering of NPs in 60% glucose (Z_avg), absorbance of NPs at 633 nm at 1E10 particles per mL. K_edge values reproduced from NIST X-ray transition Database³⁸.

Size, Absorbance, NP Characterization Table
Nanoparticle	DLS (nm)	PDI	Absorbance (633 nm)	Kedge (keV)	Vendor
SNP (SiO2)	163.7±4.8	0.039	0.0725	1.85	Synthesised
TiO2 (Anatase)	12.8±1.5	0.082	0.192	4.98	American Elements
HfO2	34.9±0.7	0.192	0.189	65.35	American Elements
Eu2O3	90.7±3.6	0.432	0.443	48.52	Sigma-Aldrich
Gd2O3	75.7±6.8	0.503	0.357	50.25	Sigma-Aldrich
YAG:Ce	36.6±4.0	0.401	0.149	40.45	Nanostructured & Amorphous Materials Inc.
Bi2O3	201.2±4.4	0.315	0.242	90.54	Sigma-Aldrich
AuNP	85.8±0.009	0.092	0.200	80.72	Synthesised

When 30 µCi of ³H, the lowest energy pure β⁻ emitter, were combined with the NP panel, radiance enhancements over ³H in H₂O were observed: SNP (1.2-fold), TiO₂ (55-fold), HfO₂ (4-fold), Eu₂O₃ (179-fold), Gd₂O₃ (301-fold), and YAG:Ce (376-fold) (Fig. 2a). Since the short trajectory of the low energy β emitted for ³H may only interact with a few NPs, we next tested the radiance enhancements with more energetic radionuclides. ³⁵S yielded a greater photon flux enhancements for NP combinations: SNP (2-fold), anatase TiO₂ (39-fold), HfO₂ (58-fold), Eu₂O₃ (2750-fold), Gd₂O₃ (1390-fold), and YAG:Ce (690-fold). The addition of ^99mTc, a pure γ-emitter, to NPs in the panel yielded similar enhancements: anatase TiO₂ (268-fold), HfO₂ (251-fold), Eu₂O₃ (2760-fold), Gd₂O₃ (366-fold), and YAG:Ce (775-fold). Thus, both β and γ emitting radionuclides combined with NPs led to scintillation and appreciable increases in radiance. Importantly, the radiances from Eu₂O₃, Gd₂O₃, and YAG:Ce were ca. 4-fold greater with ³⁵S (β) compared to ^99mTc (γ) demonstrating that, contrary to previous reports²⁷, β scintillation is a crucial mechanism in NP excitation (Fig. 2b). Furthermore, the enhancement of NP radiances is dependent on the amount of radioactivity present, (Fig. S1). Both activity titrations and tissue attenuation studies were conducted to confirm linearity (Fig. S2). While both β and γ radionuclides were shown to excite NPs without involving CL, it remained to be determined if one type of excitation would dominate or if they are additive. Therefore, ³⁵S and ^99mTc were added sequentially to each NP well in both orders, (i.e. β then γ radionuclide added and vice versa). The addition of the second radionuclide yielded additive radiances thus the order of addition did not significantly matter (Fig. 2c). To assess the radiance dependence on the NP structure rather than the metal ion, an equimolar solution of europium cations was tested with 100 µCi ³⁵S and found to yield only 13% of the radiance of Eu₂O₃ NP containing the same amount of Eu (Fig. 2d). Radiances observed from pure β⁻ emitting and γ emitting radionuclides with NPs (Figs. 2a and 2b) are summarised and plotted in Fig. 2e (standard deviation shown alone in Fig. S3). The effect of NP size on enhancement was also determined by using three sizes of Eu₂O₃ NPs normalised to particle number, volume or surface area. When mixed with ¹⁸F-FDG, the enhancement for the different sized Eu₂O₃ NPs varied less than threefold (Fig. 2f) while each is nearly 100-fold brighter than ¹⁸F-FDG alone, suggesting the enhancement is more dependent on the molar amount of europium within NPs and less dependent on particle size in the range tested. While certain NPs are more luminescent than others, the addition of sub-CL energy radionuclides results in increased luminescence of up to nearly 3000 times that of the radionuclide alone (i.e. ³⁵S with Eu₂O₃ NPs) and distinct from chemiluminescence (Fig. S4).

Figure 2.

Radiative output of nanoparticles with radionuclides. a, Well plate images of NPs with and without addition of low energy pure β⁻ emitter ³H show increased radiances in wells containing HfO₂, Eu₂O₃, and Gd₂O₃ (representative images, n=3 per NP). b, Head to head comparison of γ and β scintillation. Addition of 30 µCi (1.11 MBq) ^99mTc (right) or ³⁵S (left) to glucose, Eu₂O₃, Gd₂O₃ and YAG:Ce shows ³⁵S to be brighter while both radionuclides do not appreciably produce light in the suspending medium(top row) (representative well images, n=3 per NP and radionuclide). c, Computed radiance of wells from Fig 2b including subsequent addition of the other radionuclide from Fig. 2b to the same plate to assess order of addition to radiance shows minimal difference when compared to the summation of the individual radiances by ^99mTc and ³⁵S (n=3 independent samples, column centre=mean, error bars=SD) d, Radiance with Eu₂O₃ NPs and radionuclides is dependent on the nanostructure of the particle and not the ion content when compared with a solution containing ³⁵S and an equal molar solution of EuCl₃ in glucose (n=8 independent samples, representative images above figure, column centre=mean, error bars=SD). e, summary of radiance enhancements seen with addition of either pure gamma emitter ^99mTc or pure β emitting ³H and ³⁵S with sub CL energy (in H₂O) also result in enhanced radiance (n=3 independent samples, grey bar=min/max per NP, bar=SD). Grey bars represent min/max radiance for each particle for visual grouping. f, Three sizes of Eu₂O₃ NP were tested and normalised by number, volume, and surface area (n=3, independent samples grey bar=min/max per NP normalisation for visual grouping, error bars=SD). g, CL radiance of select nuclides in water per unit of activity and volume (n=8, independent samples line=mean, error bars=SD). Radionuclides that do not meet the CL threshold in water were omitted. Grey bars represent min/max radiance for each radionuclide for visual grouping. h, Radiance of NPs tested with positron-emitters ¹⁸F, ⁸⁹Zr, and ⁶⁸Ga and β⁻ emitters ¹⁷⁷Lu and ⁹⁰Y at 30 µCi (1.11 MBq) per well. Particles without activity were included as a reference (n=3 independent samples, grey bar=min/max per clinical radionuclide, error bars=SD).

To determine overall changes in the photon flux due to NP interactions, clinically used radionuclides from Table 1 were next explored with NPs in Table 2. Clinical β-emitting isotopes emit CL in water according to the Frank-Tamm equation (1) showing that CL increases with particle energy (Fig. 2g) and the refractive index of the medium. The clinical isotopes ¹⁸F, ⁶⁸Ga, ¹⁷⁷Lu and ⁹⁰Y isotopes with high-energy emission were used to observe total radiance enhancements by NPs. Radiance enhancement for high-energy electron and γ emitting ⁹⁰Y and ¹⁷⁷Lu, and 511 keV γ emitting ¹⁸F, ⁶⁸Ga and ⁸⁹Zr were measured (Fig. 2h). Similar results to ⁹⁰Y were observed with the addition of the pure high-energy β emitter ³²P (Fig. S5). The 511 keV γ from positron annihilation was also tested with the NPs from Table 2, and were not found to produce radiances as large as direct addition of ^99mTc (Figs. S6 and S21), in agreement with previous results²⁷.

Equation 1: Frank-Tamm formula defining photon output over a distance (dN/dx), where α is the fine structure constant, β is the velocity of the particle relative to the speed of light in a vacuum, n is the refractive index of the medium, and λ is the wavelength of interest in the ultraviolet-visible region.

$$\frac{\mathit{\text{dN}}}{\mathit{\text{dx}}}=2\pi \alpha (1-\frac{1}{{\beta }^2{n}^2}){\int }_{{\lambda }_1}^{{\lambda }_2}\frac{1}{{\lambda }_2}d\lambda$$ (1)

Of the NPs tested, Eu₂O₃ and Gd₂O₃ exhibited the largest enhancement in radiance. Enhancements using ¹⁷⁷Lu, ¹⁸F, ⁶⁸Ga and ⁹⁰Y with Eu₂O₃ were found to be 180, 27, 5.7 and 5.2 times that of H₂O respectively. Radiance enhancement ratios were greatest for low energy radionuclides since the control H₂O radiances are the weakest. NP radiance enhancements did not correlate with NP refractive index (Figs. S7–S8), demonstrating the effect resulted from mechanisms other than CL. Overall, the increased radiance from clinical isotopes must arise from multiple mechanisms, where photon contributions are based on the type and energy distribution of radiation emitted as well as the NP composition (Fig. S9)³⁴.

Characteristics of NP photon generation

To further understand the optical output of the NP radionuclide combinations, spectra for each of the NPs with β and γ emitting radionuclides were measured from 250–840 nm in a plate reader (Fig. 3a) and in smaller but discrete step values from 500–840 nm on an IVIS optical imaging system (Fig. 3b). Disregarding material specific absorption, pure CL has the characteristic 1/λ² wavelength intensity profile (Fig. S10), whereas discrete ionizations and excitations would exhibit known atomic transitions for scintillation or NP photoluminescence. The combination of Gd₂O₃ and Eu₂O₃ with ⁶⁸Ga and ¹⁸F, respectively, produced distinct visible emission peaks (Figs. 3a, b) using a 540 nm narrow band emission filter³⁵ for Gd³⁺, and characteristic f-f transitions for Eu³⁺ using 620 and 700 nm filters³⁶. The less luminescent NPs such as HfO₂, TiO₂, Bi₂O₃, and YAG:Ce produced weaker and broader emission peaks in the 400–840 nm window (Figs. 3a, S10). The emission peak for HfO₂ at 495 nm could not be resolved with the factory installed filters³⁷ but is visible on the well plate scan (Fig. 3a). YAG:Ce has a photoluminescence peak at 540–560 nm³⁸ but is weak compared to Gd₂O₃ and Eu₂O₃. No differences were observed between traditional laser excitation and emission spectra of these particles compared to spectra collected with radionuclides (Fig. S11). Spectral peaks were identical on the IVIS when excited by α emitters ²²⁵Ac or ²²³Ra (Fig. S12). The f-f transitions appear with both β and γ radionuclides, suggesting both as sources for scintillation with Eu₂O₃, Gd₂O₃ and YAG:Ce.

Figure 3.

The visible spectrum of radionuclides with nanoparticles between 250–840 nm on well plate and IVIS imaging systems. a, Normalised spectra of NPs and suspending medium using 6 mCi (222 MBq) ⁶⁸Ga per well from 250 nm to 840 nm measured with a well plate reader. b, NPs with specific spectral emissions such as Eu₂O₃ and Gd₂O₃ with 30 µCi (1.11 MBq) ¹⁸F per well (n=3 independent samples, error bars=SD) on the IVIS imaging system. c, Diagram of in vivo phantom setup using pure β-emitting ³²P. d, Open filter and e, 620 nm filter only. n=6 Matrigel phantoms per mouse (n=3) TL: Matrigel only, TR: Matrigel + ³²P, ML:Eu₂O₃, MR: Bi₂O₃, BL: Eu₂O₃ + ³²P, BR: Bi₂O₃ + ³²P. Fig. S14, radiance spectrum for ³²P phantoms showing characteristic emission peak of Eu³⁺ identical to in vitro emission peaks (n=3, error bars=SD). f, Diagram of in vivo phantom setup for mouse containing Matrigel phantoms with 750 µCi (27.75 MBq) ³²P (clockwise from top left) Matrigel only, Eu₂O₃, YAG:Ce, and Gd₂O₃ (3×10⁶ – 3×10⁷ p/s/cm²/sr) (n=3) g, open filter. h, Emission specific window for Gd₂O₃ at 540 nm followed by similar emission peaks at 620 nm and 700 nm windows in i–j, to that seen in b. p/s/cm²/sr = photons/second/cm²/steradian

The signature emission peaks were retained in vivo (Fig. S14), where matrigel phantoms (Fig. 3c–e) showed nearly 5-fold enhanced radiance with Eu₂O₃ compared to ³²P alone using the open filter (Fig. 3d). Using a 620 nm filter, only the emission from Eu₂O₃ phantom in the presence of activity was visible (Fig. 3e). A significant advantage is that the red photons from the europium emission spectrum are much less attenuated in vivo than the blue weighted light from CL¹³. β nuclides such as ³²P can be used to produce selective optical emission both in vitro (Fig. 3b) and in vivo (Fig. 3e–j). Similarly, a comparison of Eu₂O₃ and Gd₂O₃ matrigel plugs with ³²P in vivo (Fig. 3f–j) show the emission for Gd³⁺ using the 540 nm filter (Fig. 3h), and for both particles at longer wavelengths using 620 nm and 700 nm filters, respectively (Fig. 3i–j).

NP-multiplexed SPECT imaging

β particles can interact with matter via mechanisms that include CL, electron excitation, ionization, bremsstrahlung, and annihilation (β⁺ only) while γ particles can interact with matter through the photoelectric or Compton effects, coherent scattering, or pair production (Fig. S9). During ionization, characteristic X-rays are generated through an L shell electron occupying the vacancy of a previously ejected K shell electron. As the ejected X-ray represents a specific energy loss from an element, these X-rays are commonly used to determine elemental composition of materials based upon X-ray fluorescence (XRF)^{39, 40}. Therefore, the NP panel was combined with β- or γ-emitting radionuclides, and evaluated for X-rays originating from the various NPs below 100 keV. Each of the β⁻ radionuclides tested (except ³H, which has energy emissions too low to eject K shell electrons) showed a broad bremsstrahlung emission profile as expected in addition to characteristic X-rays, denoted Kα and Kβ, that were dependent on the NP composition. Fig. 4a presents the typical X-ray spectra for each of the NPs from the pure β⁻ emitter ³²P on top of a background of bremsstrahlung. Characteristic X-ray peaks were observed for Ge, YAG:Ce, Eu₂O₃, Gd₂O₃, HfO₂, Au, and Bi₂O₃, while no X-ray peaks were observed for glucose, silica, and anatase TiO₂. Characteristic X-ray production rates increase with both energy of the β radionuclide and atomic number⁴¹, explaining the lack of observed characteristic X-ray peaks for lower Z materials like silica, TiO₂, and glucose. Here, we observe that in addition to scintillation occurring to produce visible light (Figs. 2 and 3), NPs are ionised producing characteristic X-rays in addition to and bremsstrahlung by β⁻ emitting nuclides. The ionization peaks were identical using ^99mTc (Fig. 4b). The efficiency of X-ray peak generation by the radionuclides was not calculated; however, an overlay of α, β⁺, γ and β⁻ emitting nuclides tested with Eu₂O₃ all show the formation of characteristic X-rays (Fig. S15).

Figure 4.

High energy spectra and imaging of NP ionization from radionuclides. a, Characteristic X-ray spectra with the pure β⁻ emitter ³²P showing qualitative abundance and increasing Kα and Kβ energy with increasing atomic number (Z) from 1–100 keV. ×10 magnification above YAG:Ce trace represents characteristic X-rays of the cerium dopant upon longer measurement (representative spectra shown). b. Identical characteristic X-rays generated using ^99mTc from 1–100 keV. For Eu₂O₃, characteristic X-ray peaks were produced using clinical radionuclides ⁹⁰Y, ⁶⁸Ga, ¹⁸F, and ¹⁷⁷Lu (Fig. S15). c, X-ray generation curves from ^99mTc, ³²P and ³⁵S in Eu₂O₃ NPs by SPECT based upon region of interest drawn over phantom cylinder (n=1 phantom per activity level per radionuclide, centre value=median intensity reported from 1cc volumetric cylinder, error bars=SD). d, Proof of concept in vitro using Eu₂O₃, HfO₂, and Bi₂O₃ NPs with ^99mTc to produce a multicolour image based upon characteristic X-rays captured by SPECT. e, Diagram of mouse phantom used for dual spectrum imaging of either europium or bismuth particles with ³²P. f, Characteristic X-rays imaging on SPECT produced from Eu₂O₃ with enlarged phantom regions to highlight signal presence and absence g, Bi₂O₃ NPs (1×10¹¹NPs) with ³²P show signal localization to the Bi₂O₃ NP by SPECT and do not show signal from either NP alone nor activity alone.

While rare earth NPs have been excited in vivo using exogenous X-ray beams to produce photons^{30, 42}, X-rays emitted from the NP by ionization from endogenous radionuclides can be used for multicolour NP specific imaging. With an adjustable keV photon energy window, a SPECT instrument represents an ideal imaging platform for this modality. For NP composed of higher Z atoms like europium, gadolinium, gold, and bismuth, the characteristic keV photons generated from ionization enables direct detection by X-ray imaging, and the different X-ray energy emitted (Table 2) enables multimodal detection. A titration curve was produced for Eu₂O₃ using ³⁵S, ³²P, and ^99mTc, where ^99mTc showed the greatest signal of europium X-rays per unit of activity (Fig. 4c) and demonstrated that imaging on a preclinical platform was both feasible and quantitative (Fig. S16). To test the multicolour capability of this system, a well plate was prepared containing either Eu₂O₃, HfO₂, or Bi₂O₃ with ^99mTc and X-ray emissions were detected at the energies corresponding to the predominant metal in each particle (Fig. 3d). The top panel in Fig. 3d shows the 140 keV photopeak from ^99mTc activity, while the subsequent panels show X-ray emission for europium, hafnium, and bismuth. A corresponding study was also done using mixtures of these NP (Fig. S17) demonstrating that X-rays from mixtures can be resolved with an X-ray signal dependent on the composition of the mixture. To test the multiplexing in vivo with a pure β⁻ emitter, matrigel plugs containing Eu₂O₃ or Bi₂O₃ NPs were implanted into a nude mouse with or without ³²P (Fig. 4e–g). The matrigel phantom containing both ³²P and the Eu₂O₃ show a distinct signal for europium that is not seen in any other phantom (Fig. 4f). Similarly, the phantom containing Bi₂O₃ and ³²P shows the highest intensity for bismuth characteristic X-rays (Fig. 4g) and is specific to the detection channel used (Fig. S18) provided the detector can resolve the X-ray energy difference.

Clinical relevance and new imaging technology

Beyond SiO₂, TiO₂, and AuNP, most NPs investigated face a large barrier for clinical translation. In particular, there is known toxicity for uncoated rare earth oxides in lysosomes⁴³. Rare earth NP toxicity must be mitigated by increasing surface biocompatibility and facilitate biological clearance. However, our results highlight that NPs in proximity to clinical radionuclides provide enhanced photon radiance and generate characteristic X-rays that can be used in several modes of imaging. The high energy of ⁹⁰Y makes this new SPECT mode of imaging immediately clinically relevant with Theraspheres® or other ⁹⁰Y sources. Theraspheres® are glass microspheres containing ⁹⁰Y and are FDA approved for radioembolization of primary liver tumours when delivered by catheter directly to the tumour via the hepatic artery branches. Imaging their location in vivo has proved challenging. Therefore, we assessed our imaging technique on the combination of Theraspheres® and bismuth or europium NPs. This is advantageous over e.g. mixing the spheres with small molecule radiotracers which could quickly dissipate from the injection side (e.g. ^99mTc-MDP). Tumour bearing mice received intratumoural injections of Theraspheres® or Theraspheres® co-injected with either bismuth or europium NPs. In the europium and bismuth energy windows (~10 keV width total), Theraspheres® alone showed no appreciable signal (Fig. 5a and 5b) compared to Theraspheres® co-injected with Eu₂O₃ or Bi₂O₃ NPs (Fig. 5c and 5d). A modest CT contrast was observed with Theraspheres®, and enhanced with either europium or bismuth NPs (Fig S19) for visual location recognition. Here, select high Z NPs served as an imaging agent for Therasphere® detection by SPECT, opening up the possibility for multimodal detection with other NP agents simultaneously.

Figure 5.

Imaging of clinical radioembolization agent ⁹⁰Y Theraspheres® with characteristic X-rays. a,b, Representative images of 1 mCi Theraspheres® injected intratoumorally into BT-474 orthotopic breast tumours (red arrow) using characteristic X-ray energy windows for Eu₂O₃ and Bi₂O₃ NPs, respectively, shows no appreciable signal (n=3 animals, representative image shown). c,d, Imaging of Theraspheres® co-injected with Eu₂O₃ or Bi₂O₃ NPs shows selective signal near NP injection site (n=4 animals, representative image shown). Lymph node imaging of drainage from right and left hind footpads with ⁹⁰Y labelled Eu₂O₃ and AuNP respectively with e, IVIS open filter, and f, select 620nm and 700nm wavelengths to identify Eu₂O₃ selective drainage (n=3, representative single mouse sequence shown). g, SPECT-CT showing characteristic X-rays from 66.9keV ionization of gold from Au-PEG-DOTA(⁹⁰Y) NPs draining into sentinel lymph nodes 48h post intradermal tail injection (n=3 animals, representative mouse image shown). Corresponding IVIS image to (Fig. 5g) can be found in Fig. S20. h, Non-CL light based PDT. Radionuclides ¹⁸F, ³⁵S, and ^99mTc were added to cells in vitro, along with TiO₂ NP with transferrin attached (Tf) with and without titanocene (Tc). Therapeutic effects were demonstrated with radionuclides that do not emit CL, highlighting NP derived light can be as useful for PDT alongside CL (n=3 independent samples, centre value=mean, error bars=SD). **=P< 0.01, ****=P< 0.0001, ns=not significant.

Next, this technique was used to visualise lymphatic drainage and identify sentinel, locoregional draining lymph nodes, a procedure critical in oncologic surgery⁸. Eu₂O₃ NPs coated with silica and PEGylated AuNPs were synthesised to visualise drainage using chelator free⁴ and chelate methods with ⁹⁰Y respectively. NPs were injected into either hind footpad to monitor lymph node drainage by IVIS (Fig. 5e). The smaller AuNPs (80 nm vs. 240 nm) drained nearly fifteen times faster (>30 µm/s) and reached the sentinel and higher echelon nodal stations faster as compared to the larger Eu₂O₃ NPs. This method would allow analysing lymphatic functionality preclinically. Furthermore, 620 nm and 700 nm emission windows enabled selective imaging of Eu₂O₃ only drainage (Fig. 5f). Sentinel lymph node drainage of AuNPs was also seen using the X-ray energy window for gold (Fig. 5g, S20). While other means of detecting sentinel nodes exist there are no clinical approved means to determine lymphatic function both non-invasively with SPECT for pre-surgical mapping and optically during surgery using the same agent and thus imaging the same biodistribution. Further miniaturisation and targeting could allow this system to move beyond lymph node drainage to image specifically tumour distribution for preclinical mapping and intraoperative imaging.

Combining TiO₂ with sub CL radionuclides produced appreciable photon flux enhancements. Therefore, based on the photodynamic therapy previously described through CL³, we sought to evaluate and compare the therapeutic effect in vitro of sub-CL radionuclides together with TiO₂ transferrin (Tf) and titanocene (Tc) NPs. The radionuclides ¹⁸F, ³⁵S, and ^99mTc were chosen to test if CL alone, or NP derived visible light from β or γ scintillation, are sufficient to elicit a therapeutic effect. No significant difference in viability was seen between radionuclide and the addition of TiO₂Tf for all three radionuclides tested; however, the addition of TiO₂TfTc produced a significant therapeutic effect (Fig. 5h). By using the non-CL radionuclides ³⁵S and ^99mTc, we demonstrate each aforementioned mechanism (not just CL) can be a contributor to a PDT effect from radiotracers and TiO₂, as hypothesised previously.

Conclusions

Radionuclide-NP combinations are widely used in vivo because of the high specific activity, payload, and multifunctional properties achievable with NPs. The types of interactions between ionizing radiation and NP that produce photons are dependent on both the NP’s and the radiotracer’s physical characteristics, summarised in Figure 6. Low energy radionuclides such as ³H can polarise regions on a NP while higher energy β emitters can produce CL in a NP but can also ionise, causing characteristic X-ray production. Gamma emitting radionuclides operate through the photoelectric effect where an incident gamma ray is converted into electrons which could recombine or relax to generate light. Gamma radiation can also generate characteristic X-rays if a K-shell electron is ionised. Clinical radionuclides can produce several high-energy β and γ emissions, each producing light through the aforementioned mechanisms on a given NP.

Figure 6.

Interactions between ionizing radiation and NP. Ionizing radiation interactions with NP may result in a number of emissions. Low-energy β emissions from radionuclides like ³H with TiO₂ can cause electron hole recombination and resultant emission of photons. β emissions in the tens to hundreds of keV can result in numerous interactions with nanoparticles, such as Cerenkov emission within the NP, excitation or electron hole formation, or atom ionization in the NP of interest. High-energy photons, such as the 140 keV gamma from ^99mTc, can result in the photoelectric effect or ionization within the NP. Clinical radionuclides such as PET tracers and radiotherapeutics, can result in the all of the aforementioned interactions along with CL interactions from the NP.

The visible light emanating from β-emitting radiotracers has allowed preclinical and clinical luminescence imaging through CLI¹⁶, but the photon flux is quite low. NPs in proximity to either β or γ emitting radionuclides were shown to exhibit enhanced total photon flux and emit at discrete wavelengths in systems with luminescent metal ions. Exploring the mechanisms of NP radiance with radionuclides that emit only a single type of radiation revealed unforeseen contributions of NP luminescence from the β particle. Our study sets the stage for imaging with all β-emitting radionuclides, exploiting the advantages of an enhanced signal emanating from the NP compared to CL alone. Using radionuclides with radiosensitizing agents for photodynamic therapy is increasingly investigated,^44–46 and a greater understanding of how NPs can increase photon density for PDT should aid progress. Additionally, the heretofore unappreciated characteristic X-rays resulting from these interactions yields a new imaging modality in SPECT. These findings have broad implications for nuclear detection, optical imaging, and design of high Z NPs for imaging and therapy. Beyond the biosciences, these results can be exploited in environmental and material analytics.

Methods

Nanoparticle preparation

All NPs except SiO₂ and AuNP were purchased from either American Elements or Sigma Aldrich. SiO₂ NPs were prepared via a modified Stöber method⁴. Silicated Eu₂O₃ NPs were prepared based upon a modified gold NP method described previously⁴⁷. Nanopowders were suspended in an aqueous solution of 60% by weight glucose with the aid of a 500 W tip sonicator to create a NP suspension. Au-PEG-DOTA NPs were prepared using an established AuNP protocol followed by thiol-PEG-NHS coupling (Nanocs Inc. HS-PEG10k-NHS) and immediate conjugation with S-2-(4-Aminobenzyl)-1,4,7,10-tetraazacyclododecane tetraacetic acid (macrocyclics) under slightly basic conditions. Particles were purified and concentrated by tangential flow filtration using a Spectrum hollow fiber microkros 50kD mPES membrane and stored at 4°C until chelation with ⁹⁰Y as described previously.

Nanoparticle characterization

NP morphology and diameter were determined using both a Jeol 200 kV TEM and a Malvern Zetasizer Nano ZS dynamic light scattering. A Malvern Nanosight NS500 instrument was used to determine particle concentration. The absorbance and photo-luminescence was determined on a SpectraMax M5 plate reader via quartz cuvette using 1E10 particles per mL or lower and extrapolated when exceeding 1 OD. Excitation and emission spectra were acquired on a Cary Eclipse Fluorescence Spectrometer and the SpectraMax M5 plate reader at 1 mg per mL.

Radiotracer production (SI)

Hydrogen-3, phosphorus-32, sulfur-35, lutetium-177, and yttrium-90 were purchased from PerkinElmer as liquid solutions. When appropriate, solutions were brought to neutral pH prior to addition to NP solutions. Zirconium-89 was produced on the MSK cyclotron and purified as zirconium oxalate. This was neutralised with sodium carbonate prior to use. Gallium-68 was produced on a germanium generator and eluted in 0.3N HCl. This was neutralised with 28% ammonium hydroxide prior to use. Fluorine-18 was obtained from the MSK nuclear pharmacy in the form of ¹⁸F-fluoro-2-deoxyglucose ([¹⁸F]-FDG). Technecium-99 metastable was obtained from the MSK nuclear pharmacy in the form of ^99mTc pertechnetate. ⁹⁰Y Theraspheres® were obtained through a material trade agreement and generosity of BTG interventional medicine. ²²⁵Ac was generously gifted from the Lewis Lab at MSKCC and ²²³Ra was obtained through a material trade agreement with the Thorek lab at Johns Hopkins University.

Beta, Gamma and X-ray interactions for visible light modulation

24 well black walled plates from Greiner were used for IVIS optical imaging. NP solutions were prepared as previously stated at 1E10, 1E11 and 1E12 particles per mL. 1 mL of nanosuspension was diluted with 5 µL of the radionuclide of interest, targeting 30 µCi (1.11MBq) per well at time of measurement. Imaging was performed on an IVIS Spectrum at f=1 for durations of 60 s to 300 s with an open filter. Initially the NPs in the panel (Table 2) were measured without any radioactivity to determine background radiative output. Spectrum scans were acquired using 20 nm increment filters between 500 nm and 840 nm on the IVIS Spectrum or between 250–840nm in 10nm increments on the SpectraMax M5. For the IVIS Spectrum ROIs were drawn over each well and radiance values were reported in p/s/cm²/sr after background subtraction averaging n=3 wells and decay correction. Enhancement was normalised to the radiance of an identical activity of radionuclide in H₂O. Well plate IVIS experiments were done in triplicate with spectral measurements done over the course of three hours per plate. Replicate plates were done on subsequent weeks with at least n=3 replicate experiments for a nanoparticle radionuclide combination. Radionuclide radiances alone were determined on the IVIS using a minimum of 4 replicates per activity level per nuclide with at least 5 different activity amounts. Radiances were then normalised to the intensity per unit of activity per volume.

High-energy photon luminescence measurements

The radionuclide of interest at identical concentrations to above was placed in 1 mL H₂O in a 24 well black walled plate upon which a poly(methyl methacrylate) plate and black paper was then placed. On top of the black paper another 24 well plate containing in triplicate wells of H₂O, 60% glucose, or NP solutions was then placed. IVIS imaging was conducted with an open filter at F1 for 300s per image. ROIs were drawn over the wells and background subtracted. See Fig. S22 for setup diagram.

X-ray and gamma ray spectroscopy

Measurement of the X-ray and gamma spectrum between 1 and 1700 keV was performed on a Canberra broad energy Ge detector. 1×10¹² particles in an Eppendorf tube were placed at a fixed height in the detector, and activity was added such that crystal dead time was between 1–5% for gamma and pure β emitters. Clinical Isotopes were added at similar activities, resulting in ~60% dead time due to additional counts from emitted γ’s. Spectra were qualitatively compared to blank tubes containing identical amounts of activity in 60% glucose only. Channel peaks were calibrated to energies with the use of an ²⁴¹Am 1.0 µCi source found in a Family Gard^® FG200 Smoke Detector. Spectra peaks were compared to atomic transitions of the NP elements using the NIST X-ray Transitions Database⁴⁸.

Modified SPECT-CT imaging using characteristic X-rays

0.5–1 mCi (18.5–37 MBq) of ³²P was mixed with nanoparticles (1×10¹¹ NPs) and matrigel and subcutaneously injected into athymic nude mice for pilot phantom studies. AuNP tail drainage experiments were done using 270 µCi (9.9 MBq) of ⁹⁰Y chelated to ~10mg of AuNPs and injected intradermally into the tail of an athymic nude mouse. Energy windows for imaging were selected based upon main Kα and Kβ transitions on a Mediso nanoSPECT/CT (Fig. S23). We constructed artificial isotopes in the Mediso isotope master file and subsequently performed energy and uniformity calibrations to prepare the nanoSPECT/CT for NP characteristic X-ray imaging. SPECT images were acquired using the standard high-energy pinhole collimators, and a scan time of 1 hour, and with 35–60% smoothing for acquired SPECT images.

In Vivo Imaging with athymic nude mice

Animal experiments were done in accordance with protocols approved by the Institutional Animal Care and Use Committee of Memorial Sloan Kettering Cancer Center and followed National Institutes of Health guidelines for animal care. Athymic male nude mice were used for lymph node drainage studies while athymic female nude mice were implanted orthotopically with BT-474 cells and injected intratumourally with 1 mCi of ⁹⁰Y Theraspheres® with and without 1E11 Eu₂O₃ or Bi₂O₃ NPs. Cells were authenticated but not checked for mycoplasma while in culture. Europium oxide nanoparticles were silicated and annealed in a 500°C oven for one hour before chelator free radiolabelling with ⁹⁰Y. Approximately 270 µCi (9.9 MBq) of ⁹⁰Y labelled NPs (10mg AuNP or 5mg Eu₂O₃) were injected into either hind food pad and imaged on the IVIS every 15 minutes through 3 hours and then subsequently at 6 and 12 hours.

In Vitro PDT using TiO₂ NPs

Experimental conditions were based off of methods described previously for TiO₂ NPs³. TiO₂ NPs from Sigma were washed with Chelex grade water and dispersed in 1× DPBS at a concentration of 1 mg/mL. Apo-transferrin from Sigma was dissolved in 1× DPBS at a concentration of 10 mg/mL and diluted 1:1 with TiO₂ NP dispersion. TiO₂Tf dispersion was placed on a cold block and tip sonicated for 120s with a 5s on, 10s off pulse sequence. Dispersion was immediately filtered with 0.45 µm filter and measured for NP absorbance relative to a known unfiltered sample. Titanocene was dissolved in DMSO and added to the TiO₂Tf NPs before incubation with HT1080 fibrosarcoma cells overnight. Cells were authenticated but not checked for mycoplasma while in culture. Radionuclides were added the following day at concentrations between 0.01–0.80 mCi/well (n=3 per condition). Plates were placed in a shielded incubator for 72h before MTS viability assay. MTS assay was read out using a M5 spectramax plate reader.

Statistical Analysis

Graphpad prism software V6.05 and V7.03 were used for data analysis and statistical calculations. A 2-way ANOVA analysis was done for the ^99mTc and ³⁵S head to head radiance analysis.

Data Availability

The data that support the plots within this paper and other finding of this study are available from the corresponding author upon reasonable request.

Code Availability

Code used to reconfigure the SPECT/CT for X-Ray imaging of select elements is available upon request. Code used requires subsequent energy and uniformity mapping for each combination before use.