Tissue pO2 distributions in xenograft tumors dynamically imaged by Cherenkov-excited phosphorescence during fractionated radiation therapy

Hypoxia in solid tumors is thought to be an important factor in resistance to therapy, but the extreme microscopic heterogeneity of the partial pressures of oxygen (pO2) between the capillaries makes it difficult to characterize the scope of this phenomenon without invasive sampling of oxygen distributions throughout the tissue. Here we develop a non-invasive method to track spatial oxygen distributions in tumors during fractionated radiotherapy, using oxygen-dependent quenching of phosphorescence, oxygen probe Oxyphor PtG4 and the radiotherapy-induced Cherenkov light to excite and image the phosphorescence lifetimes within the tissue. Mice bearing MDA-MB-231 breast cancer and FaDu head neck cancer xenografts show different pO2 responses during each of the 5 fractions (5 Gy per fraction), delivered from a clinical linear accelerator. This study demonstrates subsurface in vivo mapping of tumor pO2 distributions with submillimeter spatial resolution, thus providing a methodology to track response of tumors to fractionated radiotherapy.

L ow levels of partial pressure of oxygen (pO 2 ) in tissue, commonly referred to as hypoxia, are known to reduce the potential for therapeutic radiation damage, inducing tumor resistance and facilitating tumor cells escape from treatment [1][2][3] . Meta-analyses reveal that improvements in regional control of cancer are primarily observed in patients with well-oxygenated tumors and not in patients with hypoxic tumors 4,5,6 . Attempts have been made in the past in clinical trials to reduce hypoxia by administration of carbogen and nicotinamide 7 or to provide radiation boost in the hypoxic areas 8 , although no methodology has yet been widely adopted in clinics. A part of the problem is that the progress in understanding and clinically addressing consequences of hypoxia has been limited by the lack of methods for non-invasive, fast, and repeatable mapping of pO 2 distributions in tumor tissues. Multiple-track oxygen electrode measurements 9 or ex vivo tissue immunohistochemistry 10 suffer from invasiveness, which limits their clinical utility. Macroscopic assessment of tumor oxygenation in humans has been performed by positron emission tomography (PET) with fluoromisonidazole (FMISO) hypoxia tracer 11,12 and by blood oxygen level dependent magnetic resonance imaging (BOLD MRI) 13,14 . However, these macroscopic imaging modalities are not able to fully sample the heterogeneity of the tumor microenvironment, while both chronic and transient hypoxia are documented to be microregional 15 . Additionally, measurements by BOLD MRI report on changes in the vascular pO 2 , whereas in tissues with abnormally developed vasculature, such as tumors, blood oxygen levels may be only poorly correlated with extravascular "tissue" pO 2 , and tissue oxygen may be misrepresented. Hence, measurements of intratumoral pO 2 between capillaries and assessment of the heterogeneity of the pO 2 distributions were the key motivating factors for this study.
Oxygen-dependent quenching of phosphorescence 16,17 has been used in the past to map oxygen distributions in subcutaneous tumors in animal models [18][19][20][21] both macroscopically and microscopically, depending upon the magnification of the imaging method used. These measurements have revealed correlations between tumor pO 2 levels and radiation sensitivity 22,23 . However, truly high-resolution depth-resolved imaging by phosphorescence is possible only at relatively shallow depths, not exceeding 0.5-0.7 mm, by means of two-photon microscopy [24][25][26][27] , while volumetric phosphorescence lifetime tomography 28 suffers from low resolution similar to other diffuse tomographic methods.
Recently, a new approach has been developed that combines phosphorescence quenching with excitation by the Cherenkov light generated within tissues that are subjected to high-energy radiation during fractionated radiotherapy [29][30][31] . This method, termed Cherenkov excited luminescence imaging (CELI), is used in the present study with the goal to directly image tumor pO 2 distributions precisely at the time of radiation delivery. It is well known that hypoxic changes can cycle on the timescale of minutes, which is the typical time of a single radiation dose fraction [32][33][34] . Thus, the ability to image intratumoral pO 2 at the actual time of the radiation delivery can be highly informative for optimization of clinical radiation therapy. In broad-beam CELI luminescence can be imaged that originates from the axial depths of up to 3-6 mm within the tumor, while the lateral spatial resolution may be as high as several hundred microns, depending on the luminescent source depth 31 . These features make CELI intrinsically capable of much higher depth resolution than alloptical tomographic imaging methods.
In this study, time-resolved CELI in combination with the phosphorescent probe Oxyphor PtG4 ( Supplementary Fig. 1) is used to track oxygen changes in mouse xenograft tumors. Oxyphor PtG4 belongs to the family of dendritic oxygen probes 21,35 .
Its calibration parameters are stable in biological environments, and its phosphorescence lifetime reports selectively on absolute pO 2 levels in tissues. Cherenkov light generated within the tumors, subjected to pulsed therapeutic megavolt (MV) X-ray irradiation, served as an internal light source to excite the probe's phosphorescence. A time-gated intensified charge coupled device (ICCD) camera, synchronized with the excitation pulses, captures emission at different delay times after each pulse ( Supplementary  Fig. 2a). The images of phosphorescence at different delays are used to construct an image of phosphorescence lifetimes by applying pixel-wise exponential fitting ( Supplementary Fig. 2b). Notably, the tissue pO 2 values are obtained from the phosphorescence lifetimes, as opposed to intensities, therefore being unaffected by optical heterogeneities of the medium and insensitive to the non-uniformity of the probe distribution. The practical usefulness of our imaging approach is evaluated by combining it with the hypofractioned radiation therapy scheme, whereby a single intravascular injection of Oxyphor PtG4 allows us to track oxygen dynamics in tumors during the course of the entire treatment.

Results
Bio-distribution of Oxyphor PtG4. To evaluate the biodistribution of Oxyphor PtG4 its phosphorescence was imaged ex vivo at different time points using a standard wide-field optical imaging system (IVIS 200 Spectrum CT, Perkin-Elmer, USA) after intravenous (IV) injection into the tail vein. The phosphorescence of Oxyphor PtG4 reached its peak within 5 min in the heart, lung, liver, spleen, kidneys and brain after the injection, while the highest signal in the tumor was detected after 24 h (Fig. 1a). The ex vivo images of the phosphorescence in individual organs at the 24 h time point are shown in Fig. 1b along with the quantitative image analysis (Fig. 1c). Consistent with the ex vivo results, the images obtained in vivo (Fig. 1d), also at the 24 h time point, show that the phosphorescence from the tumors is 2 times brighter than that from the normal tissue (Fig. 1e).
Long-term retention of Oxyphor PtG4 in tumors. Four mice were imaged using the IVIS instrument as well as by CELI every day for 5 days after IV injection of Oxyphor PtG4. The radiation for CELI was delivered using a beam of a large diameter, such that the images encompassed the entire abdomen area, including both tumors, as shown in the images of the excitation Cherenkov light (Fig. 2b). The phosphorescence images acquired using either IVIS or CELI indicated that sufficient amounts of Oxyphor PtG4 lasted in the tumor for at least 5 days, which provided an opportunity to image oxygenation in these tumors during multiple-fraction radiotherapy upon a single Oxyphor PtG4 injection (Fig. 2a, c). The phosphorescence intensity of Oxyphor PtG4 in the tumor and the normal tissue decreased monotonously from day 1 to day 5 (Fig. 2d, e). Since the signal intensity both in the normal tissue and tumor decreased gradually during the 5 days, the ratio "tumor signal:normal tissue signal" showed no significant decline from day 1 to day 5 (Fig. 2f). Notably, the overall intensity of the signals measured using optical excitation was much higher than that in CELI, because the intensity of the light used for excitation in the IVIS instrument is much higher than that of the Cherenkov light. However, the Cherenkov radiation is generated within the tissue, and therefore the excitation has virtually no depth limitation, while the external optical excitation attenuates exponentially with depth ( Supplementary Fig. 3a, b). As a result, CELI is capable of sampling oxygenation in much deeper tissue layers than all-optical phosphorescence imaging ( Supplementary  Fig. 3c, d).
In vivo pO 2 imaging after local injection of Oxyphor PtG4. To demonstrate the ability to image pO 2 in vivo by CELI in healthy (muscle) and tumor tissues pre-and post-euthanasia, Oxyphor PtG4 was locally injected into the tumor and into a thigh muscle. CELI was performed in four mice, first before ( Fig. 3a) and then 30 min after (Fig. 3b) euthanasia. The data processing followed the standard protocol for wide-field phosphorescence lifetime imaging 21 . The accuracy of the lifetime measurements by CELI was verified independently by performing ex vivo measurements using standard lifetime measurement instrumentation (Supplementary Fig. 4). The pO 2 images (Fig. 3c, d) were calculated from the lifetime images using the Stern-Volmer equation and the calibration constants obtained in independent measurements. In live animals the pO 2 values in the tumors were significantly lower than in the muscle tissue (Fig. 3c, e). As expected, the pO 2 values both in the tumors and muscle rapidly dropped to near zero when mice were euthanatized (Fig. 3d, f). There was a significant variation between the median pO 2 values for the tumor and muscle tissues in the four mice before euthanasia, but after euthanasia the pO 2 was uniformly near zero in all animals ( Fig. 3g).
In vivo pO 2 imaging after IV injection of Oxyphor PtG4. Three mice were IV injected with Oxyphor PtG4 (200 μL of 200 μM solution) through the tail vein, so that the final concentration of the probe in the blood was~20 µM, and CELI was performed 24 h later (Fig. 4a). The pO 2 images (Fig. 4b) and their histograms (Fig. 4c) revealed a significant degree of hypoxia in the tumors versus the surrounding tissue. Statistically significant differences (P < 0.005, n = 3, two-sample t-test) were observed between the median pO 2 values (Fig. 4d) in the tumor (11.3 ± 0.8 Torr, mean ± SE, n = 3) and normal tissue (20.4 ± 1.1 Torr, mean ± SE, n = 3). The pO 2 distributions in the tumor regions were found to be highly heterogeneous (Fig. 4e). Analysis of locally high and low pO 2 sub-regions revealed that they are characterized by significantly different pO 2 histograms (Fig. 4f). An arbitrary threshold of 10 Torr was applied to divide the pO 2 image into . In vivo images of the phosphorescence (d) and the phosphorescence intensities in the tumors and regions of the normal tissue (marked by a black-dashed circle) (e) at 24 h after injection of Oxyphor PtG4. The mean signal from the tumors (1.97 ± 0.16 × 10 4 counts, mean ± SE) was twice as high as that from the normal tissue (0.98 ± 0.09 × 10 4 counts, mean ± SE) (n = 3). Error bars represent standard error of the mean. Significant difference was analyzed by two-sample t-test.
In vivo pO 2 imaging during multiple-fraction radiotherapy. Six mice with subcutaneous MDA-MB-231 or FaDu tumors were imaged while undergoing 5 days-long multiple-fraction radiotherapy. The radiation was delivered using a 6MV X-ray beam with 5 Gy/fraction daily, with the total radiation dose of 25 Gy. Oxyphor PtG4 was IV injected 24 h before the first treatment, and CELI was performed during each fraction of radiation. The pO 2 distributions in both MDA-MB-231 and FaDu tumors revealed a high degree of heterogeneity 36,37 . Importantly, the hypoxic regions decreased in size from one fraction to the next, and the decrease was much more pronounced in the case of the MDA-MB-231 tumors (Fig. 5a). In both tumor types, the pO 2 values were below 20 Torr on the first day, and they increased up to~30 Torr for MDA-MB-231 and up 25 Torr for FaDu tumors (Fig. 5b). The median pO 2 values in the MDA-MB-231 tumors increased markedly as the radiotherapy progressed, while no obvious changes were seen in the case of FaDu tumors (Fig. 5c), implying that more hypoxic and radio-resistant tumors allowed for less local control, which presumably resulted in a higher rate of their survival 38 (Fig. 5d). The response of the tumor volume was delayed relative to the response of the local pO 2 , i.e. MDA-MB-231 tumors started to decrease in size on day 5 and FaDu tumors only on day 9 from the start of the therapy (Fig. 5e).
Overall, the changes in the pO 2

Discussion
There is a need to measure pO 2 distributions in tumors with high spatial resolution [39][40][41] in order to identify hypoxic fractions, which are known to be less sensitive to radiation. CELI in combination with Oxyphor PtG4 may help to identify such hypoxic regions and thus help navigating and optimizing clinical radiation therapy. The fact that CELI can be performed simultaneously with fractionated therapy makes it particularly clinically relevant. In this version of CELI the limiting factor is the optical perspective of view, i.e. clear optical access to the tumor is required. In addition, the method is capable of sensing pO 2 only in a few millimeters thick layer of subcutaneous tissue. However, future realizations of the technique could be adapted for imaging of intracavitary or near-surface tumors. Most importantly, while this imaging method presently can be utilized only in pre-clinical studies, it is uniquely capable of non-invasive sampling of pO 2 across tumors with submillimeter spatial resolution exactly at the time of the radiotherapy delivery.
In this work we showed that Oxyphor PtG4 accumulates preferentially within tumors and remains there for at least 5 days. This localization is likely due to the enhanced permeability and retention (EPR) effect. Localization studies ( Supplementary  Fig. 7) revealed that the probe distributes across the entire interstitial space of the tumor. There may be slightly higher sequestration into necrotic areas at later time points, but there is also more necrosis at these time points, making this further sequestration pattern hard to assay with certainty. Importantly, the probe is present in the tumors throughout the entire time course of the studies, and stronger quenching of the phosphorescence at higher oxygen concentrations, characteristic of normal tissues, results in the phosphorescence from hypoxic tumors being significantly brighter (Fig. 1e), providing a natural signal selection mechanism. A single injection of Oxyphor PtG4 allowed pO 2 imaging over multiple days during fractionated radiotherapy. Since pO 2 is estimated from the phosphorescence lifetime, the decrease in the concentration of the probe due to its clearance from the body does not affect the ability to quantify oxygen, as long as enough probe is present for acquiring adequate signals. Longer-term residence of Oxyphor PtG4 remains to be investigated as well as potential long-term toxicity issues; however, cell metabolic activity measurements using a standard MTT assay showed that the cell viability was above 90% when the cells were incubated with Oxyphor PtG4 at the concentration of 40 µM for 24 h (Supplementary Fig. 8), that is two times higher than in all in vivo experiments performed in this study. A similar result has been obtained previously in the toxicity tests of the probe Oxyphor PdG 4 42 , which is structurally almost identical to Oxyphor PtG4. Hematoxylin and eosin staining assay did not show differences between tissues of the main organs of animals infused with Oxyphor PtG4 vs phosphate buffer saline (PBS) ( Supplementary  Fig. 9). Likewise, blood chemistry tests (Supplementary Table S1) revealed no difference between treatments with Oxyphor PtG4 vs PBS. Overall, our results suggest that Oxyphor PtG4 has no toxic effects on the physiological function of cells, and its clearance from the body occurs presumably by the renal process.
The excitation source in CELI is the Cherenkov light generated along the pathways of MV X-ray beams. Therefore, excitation occurs at all depths where the radiation dose is deposited, including deep within tissues. Nevertheless, the detection of the phosphorescence emission is still limited by the tissue absorption and scattering. Considering the limitations of imaging sensors and the moderate brightness of the partially quenched phosphorescence, the maximal imaging depth is currently estimated to be~2 cm (Supplementary Fig. 3). The depth can be increased by employing brighter phosphorescent probes, such as recently developed Oxyphor 2P 43 and/or using cameras with sensitivity optimized specifically for the phosphorescence spectral range (750-850 nm). The scope of potential applications of CELI then can be extended on imaging of near-surface tumors, such as head and neck tumors, as well as on pre-clinical assessments of oxygenation during therapy in laboratory animals. It is also possible that intracavity measurements could be achieved with appropriate catheter-based cameras.
This study was specifically designed to use clinically relevant stereotactic hypofractionated radiotherapy doses and 6MV X-rays from a clinical LINAC system. In our recent work we demonstrated the possibility of high-resolution CELI during standard clinical intensity-modulated radiotherapy in a wholebreast geometry. We do not anticipate technical difficulties in adapting the latter method for fractionated treatment 44 . Compared to the current clinical pO 2 measurement modalities, the CELI approach is capable of significantly higher spatial resolution and allows image acquisition simultaneously with radiotherapy delivered in multiple fractions. One current limitation of the technique is that only two-dimensional pO 2 images can be obtained. Our previous studies have shown that threedimensional (3D) tomographic imaging of phosphorescence using light sheet excitation and diffuse optical tomography image reconstruction methods makes it possible to obtain 3D pO 2 maps with submillimeter resolution in vivo 29 . However, the tomography technique relies on thin sheet-like X-ray beams, which are not routinely used in the current clinical practice. Furthermore, scanning of the entire tissue volume for tomographic reconstructions may be difficult to incorporate into protocols of clinical therapy. In contrast, the simplified method reported here can be easily added to clinical protocols, enabling evaluation of tumor pO 2 histograms at the time of the radiation dose delivery, which is a long-sought goal in tumor therapy.

Methods
Ethics statement. Experimental procedures involving live animals were carried out in accordance with the protocols approved by Dartmouth Institutional Animal Care and Use Committee (Protocol Numbers 00002173). Subcutaneous and intravenous injections were performed under anesthesia that was induced and maintained with isoflurane. All efforts were made to minimize animal suffering.
Oxygen probe Oxyphor PtG4. Oxyphor PtG4 belongs to the family of dendritic oxygen probes 35 . It is a direct analog of the previously published probe Oxyphor PdG 4 21 , different only by the metal in the porphyrin: platinum vs palladium ( Supplementary Fig. 1a). The porphyrin is responsible for the key optical properties of the probe, which include two intense absorption bands (λ max = 435 and 624 nm) and strong phosphorescence (λ max = 780 nm, QY~0.07 in deoxygenated aq. solutions) occurring in the tissue near-infrared window (Supplementary Fig. 1b). The porphyrin molecule is encapsulated in a hydrophobic jacket, formed by a folded aryl-glycine dendrimer. The dendrimer protects the porphyrin from interactions with endogenous macromolecules and attenuates the rate of oxygen diffusion in the vicinity of the porphyrin, bringing the probe's sensitivity into the desired pO 2 range. The exterior of the probe is modified with multiple polyethyleneglycol (PEG) residues, making it hydrophilic and inert with respect to the endogenous macromolecules. Due to the extensive PEGylation the probe's calibration parameters remain unaltered in biological environments. The molecular weight of Oxyphor PtG4 is~35 kDa, and the approximate diameter of the molecule in an aqueous solution is~6 nm. Therefore, Oxyphor PtG4 can easily diffuse into the tumor interstitial space from the leaky vasculature: a phenomenon known as EPR effect. However, the same PEG coating prevents Oxyphor PtG4 from permeating cellular membranes. Thus, measurements using Oxyphor PtG4 give average intratumoral extravascular pO 2 .
The phosphorescence lifetime of Oxyphor PtG4 at 37°C under fully anoxic conditions is τ 0 = 45 μs. The response of the phosphorescence to oxygen was calibrated by simultaneously measuring the probe's phosphorescence lifetime and pO 2 in solution using a high-precision oxygen electrode, while gradually decreasing oxygen concentration by substituting oxygen with argon 21 . The Stern-Volmer oxygen quenching plot was linear throughout the physiological pO 2 range (0-150 Torr) with the quenching constant k q = 366 Torr −1 s −1 . In addition, the probe was calibrated using Cherenkov excitation by LINAC pulses and four acquisition  , and a 750 nm/100 nm band pass filter. The camera was contained in a homemade lead box, placed~2 m away from the imaging field to protect it from Compton scattered X-ray photons. The noise induced by X-ray photons was therefore significantly suppressed, and the signal-tonoise ratio (SNR) was more than seven times higher after shielding ( Supplementary  Fig. 10). The time-gated ICCD camera was synchronized with the radiation pulses, which were~3.25 μs long and delivered with 360 Hz repetition rate. The camera gate was turned on at 5, 10, 20, and 30 μs delay times and the phosphorescence was integrated on the CCD for 200 μs. The intensifier gain was ×100.
Longitudinal imaging study. The mice injected with Oxyphor PtG4 were continuously imaged by CELI for 5 days during their hypofractionated radiotherapy, and also imaged with the IVIS instrument for comparative purposes. All mice were under general anesthesia (isofluorane 1-3% admixed to O 2 and delivered through a nose cone) throughout imaging.  NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14415-9 ARTICLE NATURE COMMUNICATIONS | (2020) 11:573 | https://doi.org/10.1038/s41467-020-14415-9 | www.nature.com/naturecommunications diluted in PBS) were added to the wells. The samples were incubated for 24 h in a cell culture incubator, and 10 μL of 5 mg/mL MTT solution [3-(4,5-dimethylthiazolyl-2)-2,5-diphenyl tetrazolium bromide; MP Biomedical, Solon OH] was added to each well. After incubating the plate for another 6 h, the media was removed from all wells and 150 μL of acidified isopropanol was added to each well. Finally, the percentage of living cells was determined based on the measured optical density (OD) at 540 nm.
Statistical analysis. All statistical analyses for differences in pO 2 between treatment groups were done using two-sample t-test with Matlab (Mathworks, Inc., Natick, MA).
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support the findings of this study are available within the paper and its Supplementary Information. Source datasets generated and analyzed during the study are available from the corresponding author upon request.