Evaluations of an early change in tumor pathophysiology in response to radiotherapy with Oxygen Enhanced Electron Paramagnetic Resonance Imaging (OE EPRI)

Abstract

Purpose

Electron Paramagnetic Resonance Imaging (EPRI) can image the partial pressure of oxygen (pO₂) within in vivo tumor models. We sought to develop Oxygen Enhanced (OE) EPRI that measures tumor pO₂ with breathing gases of 21% O₂ (pO₂^21%) and 100% O₂ (pO₂^100%), and the differences in pO₂ between breathing gases (ΔpO₂). We applied OE EPRI to study the early change in tumor pathophysiology in response to radiotherapy in two tumor models of pancreatic cancer.

Procedures

We developed a protocol that intraperitoneally administered OX071, a trityl radical contrast agent, and then acquired anatomical MR images to localize the tumor. Subsequently, we acquired two pO₂^21% and two pO₂^100% maps using the T1 relaxation time of OX071 measured with EPRI and a R₁-pO₂ calibration of OX071. We studied 4T1 flank tumor model to evaluate the repeatability of OE EPRI. We then applied OE EPRI to study COLO 357 and Su.86.86 flank tumor models treated with 10 Gy radiotherapy.

Results

The repeatability of mean pO₂ for individual tumors was ±2.6 torr between successive scans when breathing 21% O₂ or 100% O₂, representing a precision of 9.6%. Tumor pO₂^21% and pO₂^100% decreased after radiotherapy for both models, although the decreases were not significant or only moderately significant, and the effect sizes were modest. For comparison, ΔpO₂ showed a large, highly significant decrease after radiotherapy, and the effect size was large. MANOVA and analyses of the HF10 hypoxia fraction provided similar results.

Conclusions

EPRI can evaluate tumor pO₂ with outstanding precision relative to other imaging modalities. The change in ΔpO₂ before vs. after treatment was the best parameter for measuring the early change in tumor pathophysiology in response to radiotherapy. Our studies have established ΔpO₂ from OE EPRI as a new parameter, and have established that OE EPRI is a valuable new methodology for molecular imaging.

Introduction

Hypoxia is an important biomarker in oncology [1]. In particular, hypoxia causes resistance to radiotherapy applied to solid tumors [2-4]. High-energy irradiation can either directly ionize DNA causing DNA damage, or cause production of reactive oxygen species (ROS) through radiolysis of water. These resulting ROS interact with DNA, causing DNA damage and ultimately cell death if not sufficiently repaired. In addition, irradiation converts oxygen molecules (O2) into singlet oxygen free radicals, which then inhibits repair of damaged DNA and other biomolecules in cells, leading to cell death. Therefore, evaluating the hypoxic status of tumors can aid in predicting the effectiveness of radiotherapy, and can improve ‘dose painting’ where a higher dose of radiotherapy is applied to more hypoxic tumor regions. Furthermore, a change in hypoxic status can be used to evaluate the early change in tumor pathophysiology in response to radiotherapy [5].

Evaluating tumor hypoxia with molecular imaging has been an elusive objective. Blood oxygenation level dependent magnetic resonance imaging (BOLD MRI) assesses tumor hypoxia by detecting the presence of deoxyhemoglobin, which is a paramagnetic molecule that increases the R₂* relaxation rate of surrounding water molecules [6]. Tissue oxygenation level dependent (TOLD) MRI is sensitive to the concentration of O₂, a paramagnetic species that increases the R₁ relaxation rate of water [7]. However, BOLD and TOLD MRI are affected by other physiological conditions so that the oxygen-induced changes in R₂* and R₁ can be difficult to interpret [8]. Relaxation-based exogenous MRI contrast agents have been developed that are sensitive to oxygen [9]. However, the MR image contrast produced by some agents is also dependent on the concentration of the agent, confounding hypoxia assessments [10]. ¹⁹F MRI has been used to measure the partial pressure of oxygen (pO₂) by measuring the R₁ relaxation rate of a fluorinated agents that are sensitive to oxygen tension [11-13]. Although most of these agents are administered intratumorally, a perfluorocarbon has been intravenously administered for measurements of pO₂ in tumors [14].

PET imaging can monitor a higher retention of radionuclide agents in hypoxic tumor cells vs. normoxic cells [15,16]. For example, ¹⁸F-labeled fluoromisonidazole ([¹⁸F]F-MISO) [17] and other agents [18] can accumulate within hypoxic cells and therefore enable the detection of tumor hypoxia with positron emission tomography (PET). However, it is challenging to obtain absolute values of intratumoral pO₂ from this intracellular accumulation of PET agent, and the accumulation of the radiotracer in tumor regions that only demonstrate a moderate level of hypoxia is not optimal [18].

Bioluminescence imaging (BLI) can evaluate tumor hypoxia based on the detection of light produced by the enzyme activity of luciferase that is expressed under the control of hypoxia-activated transcription [19]. However, hypoxia would need to occur at an extraordinary low level to inhibit the luciferase catalyzed emission of light, which seems impractical for in vivo studies. Optical imaging can evaluate tumor hypoxia using oxygen-sensitive fluorophores and phosphores [20,21]. However, these optical imaging methods only provide a qualitative assessment of tumor hypoxia. More recently, photoacoustic imaging (PAI) has been developed that can detect oxy-and deoxyhemoglobin within tumor vasculature [22,23], which is analogous to BOLD MRI. The ratio of oxyhemoglobin to total hemoglobin, also known as the oxygen saturation (%sO₂), is a quantitative measurement of “vascular hypoxia” in tumors that has been used to monitor tumor response to radiation [24]. Fluorescence imaging and PAI are limited to pre-clinical studies and peripheral tumors in patients.

Electron Paramagnetic Resonance Imaging (EPRI) can measure the partial pressure of oxygen (pO₂) [25,26]. EPRI can perform a standard inversion-recovery electron spin echo (IRESE) image acquisition method to measure the T₁ relaxation time of an exogenous EPR agent, OX071 [27]. This deuterated hydroxyethyltetrathiatriarylmethyl radical has a T₁ relaxation time that is dependent on pO₂ while having almost no dependence on the concentration of the agent [26]. The pO₂ measurements with EPRI have been partially validated with in vitro studies [28-30] and in vivo studies [31,32] (with the caveat that conclusively validating pO₂ measurements is a daunting task). More recently, EPRI with OX071 has been developed for imaging mouse models with solid tumors, providing 3D pO₂ maps that quantitatively measure tumor hypoxia and support pre-clinical radiotherapy studies [33,34]. This hydrophilic agent does not permeate cells during the EPRI experiment, and therefore resides in the intravascular and extravascular extracellular compartments in the tumor [35]. The weighted average of agent accumulation is considered to be largely in the extravascular extracellular compartment, and therefore pO₂ measurements largely reflect extracellular extravascular hypoxia. One aim of our research study was to evaluate the repeatability of pO₂ measurements with EPRI.

Oxygen Enhanced (OE) imaging has been developed as a variation of hypoxia imaging. This technique acquires hypoxia-sensitive images using two breathing gases, such as 21% O₂ (medical grade air) and 100% O₂, and then measures the difference in imaging information acquired with each breathing gas. For example, OE TOLD MRI measures R₁^21% and R₁^100% relaxation rates while BOLD MRI measures R₂*^21% and R₂*^100%, to determine the difference between these measurements, ΔR₁ and ΔR₂* [36,37]. OE PAI measures %sO₂^21% and %sO₂^100% to determine the differential, ΔsO₂ [38-40]. Analogously, a second aim of our research study was to develop a procedure for OE EPRI with 21% O₂ and 100% O₂ breathing gases during a single scan session, measure pO₂^21% and pO₂^100% in the tumor microenvironment, and also evaluate the difference between these two pO₂ measurements, ΔpO₂. We hypothesized that ΔpO₂ can improve evaluations of an early change in tumor pathophysiology in response to radiotherapy in tumor models, either relative to pO₂^21% and pO₂^100% or in combination with these other measurements.

Methods

In vivo Tumor Models

All experiments involving mice were conducted according to the protocols approved by the Institutional Animal Care and Use Committee at the UT MD Anderson Cancer Center. A total of 30 female athymic nude mice (IMSR_JAX:002019; Jackson Laboratory, Bar Harbor, ME), 6-8 weeks of age, were housed in a pathogen-free facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care. The 4T1 murine mammary carcinoma cell line (CVCL_0125 CRL-2539) and the SU.86.86 pancreatic ductal adenocarcinoma cell line (CVCL_3881 CTL_1837, ATCC) were purchased from ATCC, Manassas, VA. The COLO 357 metastatic pancreatic adenocarcinoma cell line (CVCL_0221) was obtained from an internal inventory at UT MD Anderson Cancer Center. For tumor implantation, 1.0 x 10⁶ 4T1 cells, 5.0 x 10⁶ SU.86.86 cells, and 5.0 x 10⁶ COLO 357 cells were suspended in 50% phosphate buffered saline and 50% Matrigel^™ (Corning, Inc.) injected subcutaneously in the right lower flank when the mice reached an age of 6-8 weeks. MRI and EPRI were performed on the mice when the tumors reached 7-14 mm in diameter, which required 10 days for 4T1 tumors, 5-6 weeks for COLO 357 tumors and 7-10 weeks for SU.86.86 tumors. Despite the large diameter of some tumors, anatomical MR imaging showed no evidence for necrosis within the tumors.

EPRI with anatomical MRI

For each EPRI scan, we prepared a dose of 300 μL of 72.2 mM OX071 at pH 7.3-7.4 (O2M Technologies, LLC). Our MRI-EPRI imaging protocol was 87 minutes long (Fig. 1A). To accommodate this timeframe, we intraperitoneally (IP) administered the agent because IP administration of OX071 allowed for a longer retention in the tumor relative to intravenous or intratumoral injections. The mouse was then anesthetized with 1.5-2.0% isoflurane and secured to a customized cradle in a right decubitus position, positioning the tumor above the center portion of the three sagittal fiducials and one slanted fiducial that contained degassed and air-sealed samples of 2 mM OX071 (Fig. 1B).

Figure 1. Oxygen Enhanced Electron Paramagnetic Resonance Imaging (OE EPRI).

A) The imaging protocol consisted of preparing the mouse for imaging with an intraperitoneal injection of OX071, anatomical MRI scans, and Inversion Recovery Electron Spin Echo (IRESE) EPRI scans to measure pO₂ with 21% O2 and 100% O2 breathing gases. b) The tumor was placed at the center of 3 horizontal fiducials and one slanted fiducial (marked with yellow arrows in the photo insert). c) The mouse was imaged with a 7T Biospec MRI instrument (Bruker Biospin, Inc) located near the EPRI instrument. d) The mouse was then imaged with a JIVA-25 EPRI instrument (O2M Technologies, LLC).

Anatomical MRI was performed with a Biospec MRI scanner with a 35-mm-diameter quadrature transceiver coil and a 7 T superconducting magnet (300 MHz proton resonance frequency; Bruker Biopsin Inc., Billerica, MA), which was located within 20 meters of the EPRI instrument (Fig 1C). A multislice 2D Rapid Acquisition with Relaxation Enhancement (RARE) acquisition method was used with the following parameters: 2.499 sec TR; 5.42 msec TE; RARE factor of 12; 128x128 matrix size; 38.4 mm² field of view, 300 μm² in-plane resolution; 1 mm slice thickness, 36 contiguous slices in an axial orientation; 1 average; total acquisition time of 25 sec.

After the MRI acquisition, the imaging cradle containing the mouse was transferred to a JIVA-25^™ EPRI scanner with a 35 mm-diameter linear transceiver coil and a 25 mT electromagnet (720 MHz electron resonance frequency; O2M Technologies, LLC, Chicago, IL, USA; Fig. 1D). During the EPR imaging process, the isoflurane gas was controlled to maintain a light anesthetic to avoid effects of anesthesia on pO₂ measurements [41]. Importantly, initial EPR images were acquired with 21% O₂ breathing gas (21% oxygen and 79% nitrogen, Matheson Tri-gas Inc., Irving, TX), and we allowed the mouse to equilibrate for 12 minutes before the pO₂^21% measurements. A 3D electron spin-echo (ESE) spatial sequence was acquired with the following parameters: 60 ns pulse length; 450 ns spin-echo delay; 45 μs repetition time; 67 baseline projections; 654 projections spaced at equal solid angles; a 16-step phase-cycle scheme with FID suppression; 1.5 G/cm gradient; 4.24 cm³ FOV; 0.66 mm³ spatial resolution; and an overall scan duration of 194 sec.

Then two sets of the IRESE acquisition method were acquired for pO₂ imaging with the onset of the first IRESE sequence starting at 27 minutes after the injection of the OX071 agent, with the following parameters: 60 ns pulse length; 400 ns spin-echo delay; 10 delays from 410 ns to 40 μs, 55 μs repetition delay; 67 baseline projections; 654 projections spaced at equal solid angles; a 16-step phase-cycle scheme with FID suppression; 1.5 G/cm gradient; 4.24 cm³ FOV; 0.66 mm³ spatial resolution; and an overall scan duration of 12.03 minutes for one IRESE scan, and a total of 24.06 min for both IRESE scans. The breathing gas was then switched to 100% O₂ (Matheson Tri-gas Inc., Irving, TX). We waited for 12 minutes after the gas switch to ensure that tissue oxygenation reaches equilibrium. Then another two sets of 3D IRESE acquisitions were performed under 100% O₂ breathing gas to measure pO₂^100% using the same parameters listed above.

All pO₂ measurements were performed using SpecMan4EPR. Image reconstruction, processing and registration was performed using a MATLAB-based toolbox (Mathworks Inc., Natick, MA) developed at the University of Chicago [42,43]. The T₁ relaxation curves from IRESE images were fitted with a single exponential model to extract the T₁ relaxation times for each voxel. The R₁ (=1/T₁) rates were converted to pO₂ values using a predetermined calibration of pO₂ vs. 1/T₁ [26]. The 2D MR images were registered with the 3D pO₂ maps using the fiducial markers. The region-of-interest (ROI) of the tumor in the MR images was used to localize the tumor in the pO₂ maps.

Radiation Therapy

Radiation therapy was performed for all mice with COLO 357 tumors and SU.86.86 tumors one day after the pre-treatment EPRI scan, and followed by a post-treatment EPRI scan one day after treatment. A single fraction of 10 Gy was delivered to the mice using 225 kVp X-rays in AP/PA geometry from a X-RAD SmART small animal irradiator (Precision X-Ray Inc., Madison, CT). This instrument was calibrated monthly according to the manufacturer’s instructions. Before delivering the treatment dosage, a set of scout images were acquired using cone-beam CT for accurate localization of the tumor. A beam filter of 2.0 mm Al was used during the imaging process, and a filter of 0.3 mm Cu was used during the irradiation process. A 15 mm collimator was used for treating tumors with a diameter smaller than 1.2 cm, and a 25 mm collimator was used for treating larger tumors. The dose rate for the treatment beams was 400 cGy/min with both collimators.

Biostatistics Analyses

The ROI of the tumor was identified in the multislice 2D MR images, and transferred to the 3D EPR image. All reported tumor pO₂^21% and pO₂^100% values are the average values over all voxels in the 3D ROI of a tumor. The ΔpO₂ values were calculated by subtracting the average pO₂^21% of a tumor from the average pO₂^100% of the same tumor. All reported standard deviations are the standard deviations of the average pO₂^21%, pO₂^100%, or ΔpO₂ values in a cohort with the same tumor type and at the same timepoint in the study. The hypoxic fraction of the tumor ≤ 10 torr (HF10) was calculated based on all voxels that had calculated pO₂ values within the tumor ROI [44].

We used a pairwise, two-tailed Student’s t-test with equal variances for comparisons of each single parameter between repetitions with the same breathing gas, between 21% O₂ vs. 100% O₂ breathing gases, and before vs. after radiotherapy. We used an unpaired two-tailed Student’s t-test with equal variances for comparisons of each single parameter between tumor models. In addition, we used Multivariate Analysis Of Variance (MANOVA) to evaluate the statistical significance of changes in both pO₂^21% O₂ and pO₂^100% before vs. after treatment. Our MANOVA evaluations used Pillai’s trace as a test statistic to approximate an F statistic, and then used the F distribution to calculate the p-value, using R for statistical computing [45]. For all of our evaluations, a value of p < 0.05 was considered to be moderately significant, and a value of p < 0.01 was considered to be highly significant.

A Student’s t-test can have limitations because the p-value indicates only if there is a significant difference, and is dependent on sample size. For comparison, an effect size is an excellent indicator of how much of a significant difference exists, and is independent of sample size [46]. The effect size is defined as the change in a parameter divided by the average of the two standard deviations of the parameter. Therefore, we also determined the effect size of each parameter for two scans with the same breathing gas, when changing the breathing gas from 21% O₂ to 100% O₂, and before vs. after radiotherapy. An effect size ≤ 0.3 is considered to be small, > 0.3 and < 0.8 is modest, and ≥ 0.8 is large.

Results

Biocompatibility of OX071 and EPRI

The IP injection of OX071 was well-tolerated during the 87-minute MRI-EPRI scan session. The EPR imaging for ~75 minutes was also well-tolerated, demonstrating that potential concerns for tissue heating during EPRI are unwarranted with this instrument [47]. The automated temperature feedback control during the imaging procedure was very useful for maintaining 37.0 ± 0.5 °C during the imaging session, which added consistency to our studies [48]. Our careful control of a light level of isoflurane anesthetic maintained the breathing rate between 80 – 120 breaths per minute, which also added consistency to our studies and avoided hypoxia caused by heavier anesthesia [41]. One mouse with a SU.86.86 tumor expired 20 minutes after being placed in the EPRI scanner during the pre-radiation imaging session, most likely due to an overflow of anesthetic.

Repeatability of tumor pO₂ measurements with EPRI

Multislice anatomical MRI could easily identify the entire 3D tumor in each tumor model, which was used to identify the tumor in the 3D pO₂ maps (Fig. 2). We manually estimated that our image registration process achieved a spatial accuracy of less than two pixels in the anatomical MR and EPR images, which is less than 0.6 mm. For the 10 mice with the 4T1 tumor model, the repeatability of the tumor pO₂^21% measurements showed an average difference of 2.6 torr, and the repeatability of tumor pO₂^100% also showed the same average difference of 2.6 torr, demonstrating that the precision of repeatability was independent of the breathing gas (Table 1). These results also demonstrated an overall 9.6% repeatability of the tumor pO₂ measurements, considering that the tumor pO₂^21% averaged 24.5 torr and tumor pO₂^100% averaged 36.1 torr.

Figure 2. Parametric maps of tumor pO₂ during OE EPRI.

(a) Multislice 2D MR images and one 3D EPRI imaging volume were acquired with 21% O₂ and 100% O₂ breathing gases. The pO₂ was measured throughout the tumor volume for each model. Only one mouse is shown for each model, with 21% O₂ gas and prior to radiotherapy. (b) Anatomical MR images (green border) were acquired for tumor localization. MR and EPR images were registered using fiducials (yellow arrows). The parametric maps of pO₂^21% (blue border) and pO₂^100% (red border) were segmented to only show the tumor ROI and overlaid on the anatomical MR images. The average tumor pO₂ value is listed for each image.

Table 1.

Repeatability of pO₂ Measurements

Measurement	1st scan1	2nd Scan1	p value2	Repeatability3	Average Repeatability4
pO221%	24.8 ± 8.2 torr	24.3 ± 8.7 torr	0.67	± 2.6 torr, 10.7%	9.6%
pO2100%	36.2 ± 12.2 torr	36.1 ± 14.6 torr	0.91	± 2.6 torr, 7.2%

^1.The average and standard deviation were calculated from the average pO₂^21% and pO₂^100% values of each tumor.

^2.The p value was determined for the 1^st scan vs. the 2^nd scan.

^3.The repeatability was determined by calculating the difference of the average pO₂^21% (or pO₂^100%) for the 1^st scan vs. the 2^nd scan for each mouse, and then taking the average of these differences for the 10 mice.

^4.The average of the two repeat abilities of pO₂^21% and pO₂^100%.

The two tumor pO₂ measurements from repeatability studies with the 4T1 model were not significantly different, either with 21% O₂ breathing gas or with 100% O₂ gas (Table 1, Fig. 3a). The effect sizes were also small when comparing the two scans with each breathing gas. These results further ensured that the pO₂ measurements were reliable. Moreover, the tumor pO₂^21% measurements were significantly different from pO₂^100%, and also with the COLO 357 and SU.86.86 models (Table 2, Fig. 3b). The effect size of pO₂ when changing from 21% O₂ to 100% O₂ breathing gas was 1.08 for the 4T1 model, indicating that the change in pO₂ was large. Therefore, EPRI was indeed sensitive to the Oxygen Enhanced gas challenge.

Figure 3. Repeatability of OE EPRI.

(a) The pairwise comparison of the tumor pO₂ values from the first and second EPRI scans of the 4T1 model showed no significant difference between scans for either breathing gas. (b) The average tumor pO₂^21% from both scans was compared with the average tumor pO₂^100% from both scans, which showed a significant difference when the breathing gas was changed for all three tumor models. (c) The ΔpO₂ is defined as the difference between pO₂^100% and pO₂^21%.

Table 2.

Comparisons of Tumor Models

Tumor	pO221%	pO2100%	P value Between gases	P value1
				vs. 4T1	vs. COLO 357	vs. SU.86.86
4T1	24.5 ± 8.2	36.1 ± 13.1	0.002	---	0.31	0.31
COLO 357	28.1 ± 7.0	34.9 ± 8.7	0.0003	0.81	---	0.83
SU.86.86	28.4 ± 7.7	36.7 ± 9.9	0.0005	0.92	0.68	---

^1.light shaded cells compare pO₂^21% and dark shaded cells compare pO₂^100%

Comparisons of pO₂ between tumor models

The average pO₂^21% was 24.5 - 28.4 torr for the 4T1, pre-treatment COLO 357, and pre-treatment SU.86.86 tumor models, which were not significantly different (Table 2). These results suggest that all three tumor models had a similar hypoxic status despite previous indications that the COLO 357 model should be more hypoxic than the SU.86.86 model [49-52]. Similarly, the average pO₂^100% was 34.9 - 36.7 torr for these tumor models, and these values were not significantly different. However, the hypoxic status of a tumor model should be judged with 21% O₂ breathing gas, so we did not use pO₂^100% to gauge the hypoxic status of each tumor model.

The average HF10 of the SU.86.86 tumor model [(22 ± 12)%] was higher than the COLO 357 tumor model [(11 ±10)%], and this difference was moderately significant (Fig. 4a). This result was surprising because previous studies have indicated that the COLO 357 model should be more hypoxic than the SU.86.86 model, as described above [49,50]. This combination of HF10 and pO₂^21% results suggests that the SU86.86 tumor model may have a broader distribution of hypoxia within each solid tumor.

Figure 4. Evaluations of the tumor fraction < 10 torr (HF10).

(a) The SU.86.86 tumor model had a higher HF10 than the COLO 357 model. Both models had a substantial range of HF10. This range provided the opportunity to compare HF10 vs. the change in ΔpO2 for each model, (b) which showed a moderate correlation for the COLO 357 model, (c) but did not show a correlation for the SU.86.86 model.

Comparisons of pO₂ before and after radiotherapy

The COLO 357 and SU.86.86 tumor models were imaged with OE EPRI one day before and one day after 10 Gy radiotherapy (Fig 1a). The average pO₂^21% decreased 3.5 and increased 3.8 torr for the COLO 357 and SU.86.86 models, respectively, although these changes were not significant (p = 0.28 and 0.21, respectively; Fig. 5a,d). The average pO₂^100% decreased 8.7 and 4.2 torr for the COLO 357 and SU.86.86 models, respectively, and this difference before and after treatment was moderately significant for the COLO 357 model (p = 0.031) and was not significant for the SU.86.86 model (p = 0.092).

Figure 5. OE EPRI monitors the early changes in tumor pathophysiology in response to radiotherapy.

Results are shown for the (a-c) COLO 357 model, and (d-f) for the Su.86.86 model. (a,d) The changes in tumor pO₂^21% and pO₂^100% after 10 Gy radiotherapy were not significant or only mildly significant for both tumor models. (b,e) The decreases in tumor ΔpO₂ after radiotherapy were highly significant for both tumor models (c,f) ΔpO₂ showed the greatest effect size in changes in tumor pathophysiology in response to radiotherapy for both tumor models.

Importantly, the average ΔpO₂ decreased by 5.2 and 8.9 torr for the COLO 357 and SU.86.86 models, respectively, and these differences were highly significant (p < 0.001; Fig. 5b,e). Furthermore, MANOVA determined that the combination of pO₂^21% and pO₂^100% showed a statistically significant change after radiotherapy for the COLO 357 and SU.86.86 models (p=0.01 and p=0.004, respectively). This multivariate analysis reinforced that OE EPRI had merits relative to EPRI with a single breathing gas.

The effect size of a change in pO₂^21% before vs. after radiotherapy was 0.47 and 0.51 for the COLO 357 and SU.86.86 models, respectively, indicating a modest change (Fig. 5c,f). The effect size of a change in pO₂^100% before vs. after radiotherapy was 0.98 and 0.47 for the COLO 357 and SU.86.86 models, respectively. While this effect size for the COLO 357 model is considered large, the effect size for SU.86.86 model was modest. Finally, the effect size of a change in ΔpO₂ before vs. after radiotherapy was 1.59 and 2.41 in the COLO 357 and SU.86.86 models, respectively, indicating a large difference. Therefore, the ΔpO₂ was the best parameter for evaluating an early change in tumor pathophysiology in response to radiotherapy.

Hypoxic tumors are resistant to radiotherapy, and therefore tumors with larger HF10 values are expected to have less change in ΔpO₂ after treatment. Indeed, we observed that HF10 and ΔpO₂ were moderately correlated for the COLO 357 model (R² = 0.59; Fig. 4b). For comparison, we did not observe a relationship between HF10 and ΔpO₂ in the SU.86.86 model (R² = 0.0004; Fig. 4c). Notably, the SU.86.86 model showed a larger change in ΔpO₂ and a larger effect size with this change in ΔpO₂ relative to the COLO 357 model, indicating that the SU.86.86 model had a larger change in tumor pathophysiology in response to radiotherapy.

Discussion

In this study, we have developed an OE-EPRI method with an image acquisition protocol with the OX071 agent that was well tolerated by mice. Physiological monitoring of temperature and respiration and good handling in a customized cradle for MRI and EPRI were important characteristics for studying hypoxia. The high-quality image registration between MRI and EPRI facilitated quantitative measurements of tumor pO₂. An intraperitoneal injection provided a ~75-minute window for EPRI acquisition that was useful for OE-EPRI.

Our pO₂ measurements of the 4T1 tumor model had a repeatability of 9.6%. This level of precision matched the precision of previous studies using EPR oximetry [53]. This repeatability includes “engineering” precision of the instrument as well as “biological” precision that is dependent on cycling hypoxia and other physiological fluctuations within the two 13-minute EPRI measurements of pO₂. For comparison, quantitative [¹⁸F]F-MISO PET that interrogates tumor hypoxia has an estimated precision of 37% [54]. Notably, 2-[¹⁸F]-FDG PET has a reliable coefficient of variation (COV) estimated to be 10-12%. [¹⁸F]F-MISO PET has more complicated pharmacokinetics than 2-[¹⁸F]-FDG PET, which may explain the worse precision of [¹⁸F]F-MISO PET compared to 2-[¹⁸F]-FDG PET [55]. Cancer studies with BOLD and TOLD MRI have a reported COV of 15% and 18.7% at 1.5 T magnetic field strength [56,57] and 35% COV at 3 T [56]. Therefore, our study indicates that EPRI can evaluate hypoxia with better precision than PET and MRI, although more EPR, PET, and MRI studies are warranted to validate this indication. Measurements of %sO₂ with PAI have a coefficient of variation of 8% [58], which is comparable to our EPRI study. Notably, PAI measures %sO₂ that interrogates vascular hypoxia, while EPRI measures pO₂ that interrogates extracellular, extravascular hypoxia, so that PAI and EPRI should be selected for specific biomedical purposes.

Our tumor pO₂ values were comparable to similar measurements of tumor models in previous studies [59,60]. Our average pO₂ in the SU.86.86 model was higher than measured in a previous report [49]. We carefully maintained mice with a light anesthetic that resulted in a higher breathing rate relative to the previous study, which may have avoided anesthesia-induced hypoxia. This observation attests to the need for good physiological monitoring when measuring a molecular parameter related to tumor metabolism.

Previous studies of protein expression levels associated with hypoxia indicated that the SU.86.86, COLO 357, and 4T1 tumor models are normoxic-to-hypoxic, respectively [49-52,60]. While the EPRI method showed this trend, the difference in pO₂ between models was not statistically significant. Furthermore, each cohort of 10 mice for each model had a range of average tumor pO₂ values. We contend that tumor models may not be able to be labeled as “always normoxic” or “always hypoxic”. Instead, individual mice may need to be evaluated for the status of average pO₂ in each tumor, asl also shown in a previous report [61]. EPRI may become a useful tool for evaluating individual tumors.

Among the parameters evaluated in this study, a decrease in ΔpO₂ is best single parameter for evaluating the early changes in tumor pathophysiology in response to radiotherapy. Multiparametric analyses of combined pO₂^21% and pO₂^100% also showed a highly significant change after radiotherapy, further substantiating that OE EPRI has a strong value in evaluating changes in tumor pathophysiology in response to radiotherapy. For comparison, the changes in the single parameters pO₂^21% or pO₂^100% were not significant or only moderately significant. Furthermore, the effect size was a useful parameter for demonstrating the large decrease in ΔpO₂ relative to the smaller changes in pO₂^21% and pO₂^100%.

We offer the following interpretation of the physiologic significance of ΔpO₂. As an analogy, an automobile can have a gas tank that is ¼ full, and therefore has ¾ of unused, available tank capacity. When the tank is then filled with petroleum gas, the tank is then 100% full. Using this analogy, we suggest that pO₂^21% represents the O₂ in the “tank” of the tumor, pO₂^100% represents a “full tank” in the tumor, and ΔpO₂ represents the unused, available “tank capacity” in the tumor. Figure 5 indicates that the radiotherapy had only a mild effect on the O₂ in the tumor “tank” (pO₂^21%) and only a moderate effect on the tank volume (pO₂^100%), and had a major effect on the unused, available “tank capacity” in the tumor (ΔpO₂). This interpretation is not conclusively proven in this study, and instead we offer this interpretation to inspire future studies that may interrogate this interpretation.

Our study also shows that additional studies with OE EPRI are warranted. Future investigations can test other tumor models of human cancers, including models with immunocompetent strains that would be appropriate for radio-immunotherapy studies [62]. Future studies can also test models that avoid the potentially immature vasculature and heterotopic location of a subcutaneous tumor. Other studies can investigate the relationship between tumor size and hypoxia, both before and after treatment. Additional studies can evaluate lower radiation doses, intensity-modulated radiation therapy, or hypofractionated stereotactic body radiation therapy relative to our study that used a uniform dose of 10 Gy that is high relative to clinical practice [4,63-66]. Notably, we refer to pO₂ and ΔpO₂ as ‘parameters’ in our study, and we have been careful to describe these parameters as indicating an early change in tumor pathophysiology and not as a change in tumor viability because we did not relate these parameters to longer-term tumor response to radiotherapy. Other investigations can be performed to correlate ΔpO₂ with tumor response to treatment using ex vivo histopathology and monitoring longitudinal tumor growth and survival, which would allow pO₂ and ΔpO₂ to be considered as a ‘biomarker’ if correlations between these measurements and tumor growth delay and/or survival are identified. The noninvasive EPRI method can be exploited to interrogate longitudinal changes in tumor reoxygenation after initial radiotherapy, especially because radiotherapy can cause complex kinetics of reoxygenation [67]. Our current study was not designed to elucidate these complex kinetics of reoxygenation. In addition, future improvements in EPRI technology can improve these studies, such as the development of a hybrid EPRI / low-field-MRI instrument that would improve the workflow of our sequential MRI and EPRI workflow. Future EPRI improvements may have faster sampling of tumor pO₂ which could be employed to observe the equilibration of O₂ in the tumor microenvironment after changing breathing gases.

Finally, HF10 can be evaluated in future studies to further understand the value of pre-treatment HF10 measurements for predicting radiotherapy response. The distribution of pO₂ values can be interrogated to evaluate the spatial heterogeneity of pO₂^21%, pO₂^100%, and ΔpO₂ within the 3D parametric maps of pO₂ measurements from EPRI, potentially demonstrating the merits of molecular imaging [68]. This future analysis of the spatial heterogeneity can address oxygen heterogeneity as a microscopic phenomenon.

In conclusion, the EPRI method measures tumor pO₂^21% and pO₂^100% with outstanding precision. This non-invasive molecular imaging method can evaluate different tumor models and individual tumors within a model cohort. Importantly, OE EPRI can be used to measure a change in ΔpO₂, which is the best parameter in our study for measuring early changes in tumor pathophysiology in response to radiotherapy in our study. Furthermore, our study establishes OE EPRI as a methodology for many future studies.