More bullets for PISTOL: linear and cyclic siloxane reporter probes for quantitative ¹H MR oximetry

Abstract

Tissue oximetry can assist in diagnosis and prognosis of many diseases and enable personalized therapy. Previously, we reported the ability of hexamethyldisiloxane (HMDSO) for accurate measurements of tissue oxygen tension (pO₂) using Proton Imaging of Siloxanes to map Tissue Oxygenation Levels (PISTOL) magnetic resonance imaging. Here we report the feasibility of several commercially available linear and cyclic siloxanes (molecular weight 162–410 g/mol) as PISTOL-based oxygen reporters by characterizing their calibration constants. Further, field and temperature dependence of pO₂ calibration curves of HMDSO, octamethyltrisiloxane (OMTSO) and polydimethylsiloxane (PDMSO) were also studied. The spin-lattice relaxation rate R₁ of all siloxanes studied here exhibited a linear relationship with oxygenation (R₁ = A′ + B′*pO₂) at all temperatures and field strengths evaluated here. The sensitivity index η( = B′/A′) decreased with increasing molecular weight with values ranged from 4.7 × 10⁻³–11.6 × 10⁻³ torr⁻¹ at 4.7 T. No substantial change in the anoxic relaxation rate and a slight decrease in pO₂ sensitivity was observed at higher magnetic fields of 7 T and 9.4 T for HMDSO and OMTSO. Temperature dependence of calibration curves for HMDSO, OMTSO and PDMSO was small and simulated errors in pO₂ measurement were 1–2 torr/°C. In summary, we have demonstrated the feasibility of various linear and cyclic siloxanes as pO₂-reporters for PISTOL-based oximetry.

Introduction

Adequate availability of oxygen is critical to the efficient functioning of many vital organs and tissues¹. Changes in oxygenation are indicative of a disruption in homeostatic conditions which are prevalent in pathologies such as tumors², wounds^3,4, ischemic heart disease^5,6 metabolic disorders^7–9 and traumatic brain injury¹⁰. The oxygen requirement changes between cells, tissues and organs and thus each tissue type exhibits a distinct normal range of oxygenation. For example, the normal tissue oxygen level in the brain is ~34 torr (mmHg) while that in the muscle is ~29 torr¹¹. The lack of adequate oxygen in cells and tissues is termed as hypoxia and could result from diminished blood flow, low blood oxygen saturation, elevated oxygen metabolism and increased cellular proliferation. Oxygen homeostasis and hypoxic stress are being recognized as important factors for development and physiology of cells and tissues. These factors also influence the pathophysiology of diseases as they regulate various intracellular signaling pathways for processes such as angiogenesis, cell proliferation and protein synthesis^12–18. Malignant tumors are known to have regions with low oxygen tension known as hypoxia which is a major driving force behind tumor progression and resistance to therapies^19–21. Hypoxia presents itself as an ideal target for the development of anti-cancer therapies due to the role that it plays in the progression of cancer²². Thus, measurement of oxygen is essential for monitoring the function of organs as well as for diagnosis, treatment planning and studying treatment response of pathologies. Consequently, there is an increased need for an oximetry technique that can facilitate repeated, non-invasive and accurate assessment of oxygen and can be translated to the clinic.

Many qualitative and quantitative oximetry techniques have been developed for oximetry such as polarographic needle electrode²³, fiber optic probes²³, Near Infrared (NIR) spectroscopy²⁴, fluorescence²⁵, immunohistochemical probes²⁶, positron emission tomography (PET)²⁷ and single photon emission computed tomography (SPECT)²⁸. Polarographic needle electrode and fiber optic probe techniques are invasive, susceptible to pressure artifacts and cannot facilitate simultaneous measurement of multiple locations and repeated measurements, while immunohistochemical hypoxia probes (pimonidazole²⁹, EF5³⁰, HIF1α²⁶) are limited to ex-vivo analysis. NIR spectroscopy is a non-invasive technique but can only detect the changes in vascular oxygen saturation and cannot distinguish between signals from oxyhemoglobin, deoxyhemoglobin, and cytochrome c oxidase³¹. PET and SPECT based techniques lack spatial resolution and cannot provide quantitative information regarding oxygenation. The current magnetic resonance imaging (MRI) based oximetry techniques can be further sub-divided into a) qualitative techniques: Blood Oxygen Level Dependent (BOLD)³², Tissue Oxygen Level Dependent (TOLD)³³, oxygen-enhanced MRI³⁴, hypoxia targeted MRI³⁵ and b) quantitative oximetry techniques: Electron Paramagnetic Resonance (EPR³⁶), ¹⁹F NMR of perfluorocarbon emulsions, Fluorocarbon Relaxometry using Echoplanar imaging for Dynamic Oxygen Mapping (FREDOM, ¹⁹F MRI of hexafluorobenzene)³⁷ and Proton Imaging of Siloxanes for mapping Tissue Oxygenation Levels (PISTOL)^38,39).

EPR and MR oximetry (¹⁹F and ¹H) techniques are minimally invasive and provide quantitative oxygenation information via measuring the change in linewidth or spin lattice relaxation time, respectively, of an exogenously administered paramagnetic spin probe as it interacts with the molecular oxygen. They allow for non-invasive and repeated measurement of oxygenation at multiple locations. Some of the EPR probes are lithium phthalocyanine (LiPc), lithium naphthalocyanine (LiNc), Fusinite, Gloxy, India Ink and triarylmethyl (TAM), of which India Ink is approved for clinical use^36,40. The ¹⁹F MR oximetry uses exogenous perfluorocarbon reporters such as perfluoro-15-crown-5-ether (15C5⁴¹) or hexafluorobenzene^37,42,43 as oxygen reporters and uses the linear relationship of flurocarbon spin-lattice relaxation rate R₁ with oxygenation. Exploiting the same concept, our group has previously shown the feasibility of accurate and repeated measurements of oxygen using hexamethyldisiloxane (HMDSO) in thigh muscle and tumor regions in vivo using ¹H MR oximetry^38,39. We have also shown the ability of siloxane based nanoemulsions for tissue oximetry⁴⁴ as well as cellular oximetry^45,46. Siloxanes can be synthesized in a variety of forms (linear or cyclic, increasing chain length, with or without functional groups) and have been used in various applications such as biomedicine, cosmetics, and food processing⁴⁷. While HMDSO has been shown to be a reliable pO₂ reporter and has a large dynamic range and high pO₂ sensitivity³⁸, the values of spin lattice relaxation time T₁ (= 1/R₁) under hypoxic conditions can be as long as 11 s, leading to long measurement times. This raises the question whether any of the other siloxanes could be used as pO₂ reporter molecules and how chain length and structure (linear versus cyclic) influence the pO₂ sensitivity and dynamic range of T₁ exhibited under different oxygenation conditions. In this study, we have characterized the calibration curves of various low molecular weight linear and cyclic siloxanes and assessed their utility as pO₂ sensing reporter molecules for use with ¹H MR oximetry. The siloxanes investigated here are: linear siloxanes HMDSO, octamethyltrisiloxane (OMTSO), decamethyltetrasiloxane (DMTSO), dodecamethylpentasiloxane (DDMPSO), trimethylsiloxy-terminated polydimethylsiloxane (PDMSO, M.W. 410) and cyclic siloxanes octamethylcyclotetrasiloxane (OMCTSO) and decamethylcyclopentasiloxane (DMCPSO). These siloxanes are commercially available, inexpensive and have a single ¹H resonance around 0.1 ppm⁴⁸. Further, field and temperature dependence of the pO₂ calibration curves of HMDSO, OMTSO and PDMSO were also studied.

Theory

Quantitative MR oximetry exploits the Fermi contact interactions between paramagnetic oxygen and reporter molecules⁴⁹ that leads to a O₂ concentration (and hence pO₂) dependent relaxation of the nuclear spins. Probes used for in vivo MR oximetry must preferably possess the following characteristics: high oxygen solubility, hydrophobicity (so that diffusion of aqueous ions is restricted), should have a single resonance so that there are no chemical shift artifacts in the MR images and minimal dependence of R₁ on temperatures. Due to paramagnetic nature of molecular oxygen, it tends to shorten the nuclear relaxation times and relaxes the nuclear spins faster thereby increasing the spin-lattice (longitudinal) and spin-spin (transverse) relaxation rate R₁ and R₂ respectively of the reporter molecule. The principle is based on the linear dependence of pO₂ on the spin-lattice relaxation rate of the probe.

If x is the molar fraction of oxygen the net spin lattice relaxation rate R₁ is given by⁴⁹

$$R_1=\left(1-x\right)\ast R_{1d}+x\ast (R_{1d}+R_{1p})=R_{1d}+x\ast R_{1p}$$ 1

Where R_1d = diamagnetic or anoxic component of the relaxation rate and

R_1p = paramagnetic component of the relaxation rate due to the contribution of oxygen.

As per Henry’s law, the dissolved mole fraction x is directly related to the partial pressure of oxygen.

$$p{O}_2=k\ast x$$ 2

k is a constant that determines the solubility of oxygen in the agent and is different for different agents. Thus, net relaxation rate becomes

$$R_1=A\prime +B\prime \ast p{O}_2$$ 3

where A′ = R_1d and B′ = R_1p/k

Since longitudinal relaxation rate is a function of temperature we assume a linear dependence of constants A′ and B′ on temperature (for relevant physiological range) which empirically can be defined as

$$A\prime =A+C\ast T$$ 4

$$B\prime =B+D\ast T$$ 5

Substituting value of A′ and B′ in Eq. [3] results in a temperature-dependent model for net relaxation rate

$$R_1=A+B\ast p{O}_2+C\ast T+D\ast T\ast p{O}_2$$ 6

Inverting the above equation, quantitative pO₂ levels can be determined more accurately and reliably if temperature is also monitored.

$$p{O}_2=\frac{R_1-A-C\ast T}{B+D\ast T}$$ 7

Therefore, estimation of these parameters (A, B, C and D) is crucial for accurate pO₂ measurement. Equation [7] also allows us to relate the errors in pO₂ estimation at a particular oxygenation level and temperature to the potential errors or uncertainty in temperature measurement, particularly for in vivo applications. Also, for a given R₁ measurement, the relative error in pO₂ determination per 1 °C error in temperature estimation at a particular temperature T and oxygenation level pO₂ can be derived as:

$$\frac{\mathrm{\Delta }p{O}_2}{\mathrm{\Delta }T}=\frac{\left|C+D\ast p{O}_2\right|}{|B+D\ast T|}$$ 8

While comparing signals from same volumes of various siloxanes (as typically a fixed volume of reporter probe would be used in vivo), the relative theoretical signals, α^siloxane, of each (compared to HMDSO) can be computed by accounting for the density differences and the mole fraction of protons in a mole of the siloxane.

$${\propto }^{siloxane}=\frac{{\rho }^{siloxane}\ast N_H^{siloxane}\ast M{W}^{HMDSO}}{{\rho }^{HMDSO}\ast N_H^{HMDSO}\ast M{W}^{siloxane}}$$ 9

where, ${\rho }^{siloxane}$ is the density of the siloxane under consideration (0.764 g/ml for HMDSO at 25 °C), $N_H^{siloxane}$ is the number of H atoms per molecule (18 for HMDSO) and $M{W}^{siloxane}$ is the molecular weight of the siloxane (162 g/mol for HMDSO). The theoretical values of α^siloxane are listedin Table 1.

Table 1.

Summary of calibration constants and α^siloxane of the various linear and cyclic siloxanes at 4.7 T (37 °C).

Siloxane	Molecular wt. (g/mol)	Intercept A′ (s−1)	Slope B′ (s*torr)−1	η (X10−3) = B′/A′	Relative Signal (αsiloxane)
HMDSO*	162.4	0.1125 ± 1.38 × 10−3	0.0013 ± 2.09 × 10−5	11.6	1.00
OMTSO	236.5	0.1597 ± 7.50 × 10−3	0.0012 ± 8.38 × 10−5	11.4	0.98
DMTSO	310.7	0.1780 ± 3.30 × 10−3	0.0015 ± 3.60 × 10−5	8.4	0.97
DDMPSO	384.8	0.2062 ± 6.70 × 10−3	0.0017 ± 7.44 × 10−5	8.2	0.97
OMCTSO	296.6	0.2827 ± 3.50 × 10−3	0.0016 ± 3.87 × 10−5	5.6	0.91
DMCPSO	370.8	0.3169 ± 3.10 × 10−3	0.0015 ± 3.47 × 10−5	4.7	0.92

*HMDSO data from ref. ³⁸ (Kodibagkar et al., Magn Reson Med 55, 743-748) reproduced here for comparison.

Results

Calibration curves of linear and cyclic siloxanes at 4.7 T

R₁ of the linear and cyclic siloxanes (Fig. 1) OMTSO, DMTSO, DDMPSO, OMCTSO and DMCPSO were measured as a function of pO₂ at 4.7 T and 37 °C and fit to the Eq. [3] to yield the calibration constants A′ and B′. At a fixed temperature (37 °C), the R₁ of all the siloxanes showed a linear dependence on pO₂ (R² > 0.99) (Fig. 2). In the linear siloxanes, it was observed that with increasing molecular weight, the intercepts of the linear fits increased (ranging from 0.11–0.32 s⁻¹) but the slopes were almost similar (ranging from 1.3 × 10⁻³ −1.7 × 10⁻³ s⁻¹.torr⁻¹). The cyclic compounds had higher R₁ than the linear compounds at each oxygen concentration and showed a similar trend with increasing molecular weight as the linear siloxanes. The recovery curves for all siloxanes showed a monoexponential behavior (Supplementary Fig. S1). Table 1 lists the values of the calibration constants A′ and B′ for all the linear and cyclic siloxanes at 4.7 T and 37 °C along with the α^siloxane values. HMDSO calibration constants at 4.7 T were included from our previously published work³⁸ for comparison.

Figure 1.

2D structures of the various linear and cyclic siloxanes characterized in this study. Linear siloxanes: (A) hexamethyldisiloxane (HMDSO), (B) octamethyltrisiloxane (OMTSO), (C) decamethyltetrasiloxane (DMTSO), (D) dodecamethylpentasiloxane (DDMPSO), (E) polydimethylsiloxane (PDMSO). Cyclic siloxanes: (F) octamethylcyclotetrasiloxane (OMCTSO) and (G) decamethylcyclopentasiloxane (DMCPSO).

Figure 2.

Dependence of spin lattice relaxation rate R₁ on pO₂ (at 37 °C and 4.7 T) for linear and cyclic siloxanes of different chain lengths. The cyclic compounds displayed higher longitudinal relaxation rates than the linear compounds at all oxygen concentrations. *Data from ref 38 (Kodibagkar et al., Magn Reson Med 55, 743–748) reproduced here for comparison.

Temperature and field dependence of calibration of siloxanes

The dependence of R₁ of HMDSO on pO₂ was determined at 7 T (Fig. 3A,B) and 9.4 T (Fig. 3C,D) as a function of temperature and was observed to be linear at both fields at all temperatures measured. Constants A′ and B′ were then plotted with respect to variations in temperature, in the physiologically relevant range of 17 °C to 48 °C, to yield characterization parameters A, B, C and D (Eqs. [4] and [5]) at 7 T (Supplementary Fig. S2A,B) and 9.4 T (Supplementary Fig. S2C,D). These are listed in Table 2. Relative pO₂ error (Eq. [8]) was computed based on these constants and was found to be between 0.6–1 torr/°C in the physiologically relevant pO₂ range 0–50 torr at 7 T (Supplementary Fig. S2E). In contrast, the relative pO₂ error was ~ 1 torr/°C at 9.4 T showing very small variation between 0–50 torr (Supplementary Fig. S2E). E.g. for a pO₂ of 5 torr, resulting error in pO₂ determination per degree change in temperature at 37 °C was ~ 0.7 torr/°C at 7 T and ~ 1 torr/°C at 9.4 T.

Figure 3.

Dependence of spin lattice relaxation rate R₁ of HMDSO at 7 T (A,B) and 9.4 T(C,D) on pO₂ (A,C) and temperature (B,D).

Table 2.

Summary of temperature dependence of calibration constants of siloxanes measured at different fields.

Siloxane	Field	A (s−1)	B (s*torr)−1	C (s*°C)−1	D (s*torr*°C)−1	Intercept A′ (s−1) 37 °C	Slope B′ (s*torr)−1 37 °C	η (X10−3) = B′/A′
HMDSO	4.7 T*	0.1479 ± 2.8 × 10−3	0.0018 ± 5 × 10−5	−9.5 × 10−4 ± 8.1 × 10−5	−1.23×10−5 ± 1.3 × 10−6	0.1125 ± 1.38 × 10−3	0.0013 ± 2.09 × 10−5	11.6
HMDSO	7 T	0.1472 ± 1.34 × 10−3	0.0015 ± 2.4 × 10−5	−9.75 × 10−4 ± 3.9 × 10−5	−1.12 × 10−5 ± 6.9 × 10−7	0.1105 ± 1.05 × 10−3	0.00108 ± 1.16 × 10−5	9.77
HMDSO	9.4 T	0.1598 ± 8.78 × 10−4	0.0013 ± 1.75 × 10−5	−1.3 × 10−3 ± 2.8 × 10−5	−1.41 × 10−6 ± 5.67 × 10−7	0.1107 ± 5.04 × 10−4	0.00125 ± 5.57 × 10−6	11.3
OMTSO	4.7 T	0.2145 ± 1.28 × 10−2	0.002 ± 1.65 × 10−4	−1.34 × 10−3 ± 3.6 × 10−4	−2.84 × 10−5 ± 4.65 × 10−6	0.1597 ± 7.59 × 10−3	0.0012 ± 8.38 × 10−5	7.5
OMTSO	7 T	0.2030 ± 1.95 × 10−3	0.0017 ± 1.32 × 10−5	−1.49 × 10−3 ± 5.73 × 10−5	−1.17 × 10−5 ± 3.87 × 10−7	0.1484 ± 3.24 × 10−3	0.0013 ± 3.58 × 10−5	8.76
OMTSO	9.4 T	0.1990 ± 1.76 × 10−3	0.0016 ± 1.72 × 10−5	−1.37  × 10−3 ± 5.7 × 10−5	−8.37 × 10−6 ± 5.58 × 10−7	0.1477 ± 4.42 × 10−3	0.0013 ± 4.88 × 10−	8.8
PDMSO	7 T	0.2943 ± 2.53 × 10−3	0.0015 ± 2.61 × 10−5	−0.0023 ± 7.43 × 10−5	−7.9 × 10−6 ± 7.68 × 10−7	0.2085 ± 1.37 × 10−3	0.0012 ± 1.511 × 10−5	5.75

*HMDSO data from ref. ³⁸ (Kodibagkar et al., Magn Reson Med 55, 743–748) reproduced here for comparison.

The dependence of R₁ of OMTSO on pO₂ was determined at 4.7 T (Fig. 4A,B), 7 T (Fig. 4C,D) and 9.4 T (Fig. 4E,F) and was also observed to be linear at both fields and all temperatures studied. The temperature dependence of constants A′ and B′ was determined by a linear fit of the constants at different temperatures at 4.7 T, 7 T and 9.4 T (Supplementary Fig. S3A–F). Relative pO₂ error for OMTSO ranged between 0.85–1.12 torr/°C in the physiologically relevant pO₂ range at 9.4 T and was observed to be between 0.65–1.35 torr/°C at 4.7 T and ~ 0.87–1.2 torr/°C at 7 T (Supplementary Fig. S3G). E.g. relative pO₂ error was calculated to be ~ 0.7 torr/°C, 0.9 torr/°C and 0.88 torr/°C at fields of 4.7 T, 7 T and 9.4 T, respectively, for a pO₂ of 5 torr. The dependence of R₁ of PDMSO on pO₂ was determined at 7 T (Fig. 5) and was also found to be linear at all temperatures, as observed for HMDSO and OMTSO. Similarly, a linear fit of A′ and B′ at different temperatures yielded constants A, B, C and D for PDMSO at 7 T (Supplementary Fig. S4, Table 2). Based on these constants, the relative pO₂ error was found to be between 1.5–1.8 torr/°C in the physiological pO₂ range (Supplementary Fig. S4C). E.g. the relative error was calculated to be ~ 1.5 torr/°C for a pO₂ of 5 torr at 7 T. With a view to compare the predicted pO₂ values from the calibration constants with the actual pO₂, we computed pO₂ maps from the calibration data at 7 T and 37 °C (using the corresponding constants shown in Table 2) for HMDSO, OMTSO and PDMSO. These are shown in Fig. 6 and match the target pO₂ (0, 5, 10 and 21% O₂ corresponding to 0, 38, 76 and 160 torr respectively) used to bubble the siloxanes in the corresponding tubes.

Figure 4.

Dependence of spin lattice relaxation rate R₁ of OMTSO at 4.7 T (A,B), 7 T (C,D) and 9.4 T (E,F) on pO₂ (A,C,E) and temperature (B,D,F).

Figure 5.

Dependence of spin lattice relaxation rate R₁ of PDMSO on (A) pO₂ and temperature at (B) 7 T.

Figure 6.

Representative MR images and T₁ maps from the pO₂ calibration along with predicted pO₂ maps for HMDSO (a–d), OMTSO (e–h) and PDMSO (i–l) at 37 °C and 7 T. In each figure the tubes from left to right were bubbled with 0%, 5%, 10% and 21% O₂ (balance N₂), respectively. All the images were analyzed using MATLAB R2018b (MathWorks, https://www.mathworks.com/products/matlab.html).

Discussion

HMDSO has been previously characterized as a pO₂ reporter molecule for quantitative oximetry using the PISTOL ¹H MR oximetry technique^38,39 and HMDSO based nanoemulsions have also been used to report oxygenation in vivo and in vitro at 4.7 T^44,45. Furthermore, PDMSO nanoemulsions have been used for cell labelling and oximetry of neural stem/progenitor cells⁴⁶ at 7 T. In this study we had several goals. Firstly, we aimed to expand the utility of the PISTOL technique for MR oximetry by identifying other siloxanes for use as pO₂ reporter molecules. Temperature-dependent calibration of HMDSO was conducted at 7 T and 9.4 T to study the effect of magnetic field on calibration of HMDSO and to extend its previously demonstrated utility as pO₂ sensor (at 4.7 T) to higher fields (7 T and 9.4 T). Based on initial observations at 4.7 T, the promising pO₂ reporter OMTSO was further evaluated at higher fields (7 T and 9.4 T) and calibration curves were characterized as a function of temperature. Finally, given the prior application of PDMSO at 7 T for cell labelling and oximetry⁴⁶, its calibration curve at 7 T was characterized as a function of temperature.

At a given temperature and magnetic field the linear relationship between pO₂ and temperature (defined by Eq. 3) determines the sensitivity of R₁ to changes in pO₂. The intercept A′ represents the relaxation rate observed under anoxic condition (the diamagnetic, oxygen independent contribution R_1D) and its inverse represents maximum T₁ displayed by the probe. The slope B′ represents the sensitivity of the probe’s relaxation rate to the changes in oxygenation and is a ratio of the paramagnetic contribution (R_1p) of oxygen to the relaxation rate of the probe and the solubility of oxygen in the probe. The ratio η = slope/intercept (B′/A′) is a parameter that helps in determining and comparing the sensitivity index of different MRI pO₂ reporter molecules. A larger slope B′ and smaller intercept A′ represent greater sensitivity to changes in pO₂ but also indicates longer imaging times. Since a smaller A′ implies a larger maximum T₁ (observed under anoxic or hypoxic conditions which are usually of interest in studying pathologies), an adequate sampling of the recovery curves would require the use of longer recovery time TR for imaging. We characterized the relaxation behavior of the linear and cyclic siloxanes by comparing the magnetization recovery curves after bubbling with N₂ (0% O₂) vs 21% O₂ (Supplementary Fig. S1). Bi-exponential T₁ behavior was not observed in any of the magnetization recovery curves of the evaluated linear and cyclic siloxanes suggesting that the availability of oxygen to all the protons (e.g end chain vs backbone for linear siloxanes) was unhindered. We observed a decrease in η with respect to an increase in chain length of the linear siloxanes and the η values ranged from 8.2–11.6 × 10⁻³ torr⁻¹ with only small changes in pO₂ sensitivity. Another important observation was that anoxic relaxation rate (A′) increased with increasing chain length of the siloxanes and was higher for cyclic siloxanes than the linear siloxanes. This is again consistent with the observations for perfluoroalkanes⁵⁰ and alkanes⁵¹ and is a consequence of the reduction of molecular tumbling rate for larger chain lengths leading to an increase in relaxation rate. The higher anoxic relaxation rate indicates a shorter maximum T₁ and hence potentially less time needed for T₁ mapping for the larger siloxanes which can be further exploited to map tissue oxygenation faster than HMDSO. The B′ value remained similar between the linear and cyclic siloxanes indicating that the solubility of oxygen and its proximity to the methyl protons remains similar between the siloxanes of different chain lengths and structures.

Performing MR oximetry at higher magnetic field strengths has the following advantages: 1) resonances have larger chemical shift separation between them which aids in selective excitation of the siloxane resonance as well as the suppression of the water and fat resonances, 2) increase in net magnetization, leading to improved signal-to-noise, 3) improved dynamic range in T₁ over the physiological pO₂ range resulting in more accurate pO₂ measurements. On the other hand, the relaxation times tend to increase at higher magnetic fields^52,53 which might result in an increase in the total imaging time for pO₂ mapping. Our goal was to evaluate the relationship of T₁s of HMDSO and OMTSO at 7 T and 9.4 T and also help in determining the choice of siloxane for applications requiring higher temporal resolution. Our results suggest that at 37 °C for HMDSO, T₁ was constant at 9 s at pO₂ = 0 torr and ranged from to 3.5–3.2 s at pO₂ = 160 torr on changing the field strength from 7 T to 9.4 T, which differed by ~ 2% from T₁ obtained at 4.7 T³⁸. Also, the calculated relative error in pO₂ determination as given by Eq. [8] at 37 °C was ~ 0.7 torr/°C at 7 T and ~ 1 torr/°C at 9.4 T when the actual pO₂ value was 5 torr. Since the maximum T₁ (and hence potentially imaging time) at 7 T and 9.4 T was same as for 4.7 T with no significant increase of temperature-fluctuation induced error in pO₂ determination, PISTOL oximetry using HMDSO would be improved at higher magnetic field strengths. Similarly, for OMTSO the T₁ ranged from 6.3–6.8 s at pO₂ = 0 torr and around 2.8 s at pO₂ = 160 torr over the range of magnetic fields strengths studied. Thus, the changes in the fields strength will not result in a substantial increase in the imaging time for OMTSO and the η and relative signal (α^siloxane) are similar to that of HMDSO (Table 1). Further, the calibration of OMTSO and PDMSO demonstrated that longitudinal relaxation rate of both the siloxanes varied linearly with respect to changes in pO₂ at temperatures in the physiological range, demonstrating the potential of OMTSO and PDMSO to measure dynamic changes in tissue pO₂. At a temperature of 37 °C OMTSO and PDMSO had an oxygen sensitivity similar to HMDSO (B′ values ranging from 0.0011 to 0.0013 s⁻¹ torr⁻¹) at all three fields. It should be also noted that boiling point of OMTSO (153 °C) is higher than the boiling point of HMDSO (101 °C).This suggests that it maybe be more advantageous to use OMTSO for generating nanoemulsions for cell labelling applications than HMDSO (used previously⁴⁵) as it would be less volatile during the emulsification process⁵⁴. The simulated errors in pO₂ determination due to temperature fluctuations for OMTSO as well as PDMSO were found to be in the same range as HMDSO at 4.7 T, 7 T and 9.4 T.

We have previously demonstrated the feasibility of in vivo pO₂ mapping following intra-tissue injection of ‘neat’ HMDSO as well as HMDSO based nanoemulsions^39,44,55. Dilution of the siloxane in a solvent can potentially affect the pO₂ calibration curve (and hence T₁ or R₁) in a “concentration-dependent” manner by changing the intercept due to changes in the dipole-dipole interactions of the siloxane protons with the solvent protons, although this was not tested here. Jamrogiewicz et al. studied the dependence of ¹H relaxation times of linear HMDSO on dilution using a mixture of carbon tetrachloride with deuterated benzene and found that T₁ did not significantly depend on the analyte concentration in the sample or the mutual ratio of the solvents used⁴⁸. Dilution of siloxanes in solvents can also affect the slope of the calibration curve as the oxygen solubility may change based on solvent used and the siloxane concentration. Dilution in tissue by use of less siloxane administered per tissue volume or by diluting a siloxane emulsion is unlikely to affect the calibration as the siloxane is restricted to a local partitioned environment consisting solely of other siloxane molecules and dissolved gasses in either case. We recommend using undiluted probes for pO₂ mapping applications in order to maintain the highest signal to noise ratio.

In general, siloxanes are considered non-toxic or minimally toxic and toxicity decreases with increasing molecular weight⁴⁷. However, individual siloxanes should be evaluated for safety before in vivo use. Previous studies have shown that HMDSO is quite inert, and exhibited minimal toxicity in rats tested for subchronic inhalation toxicity^56,57. No oral toxicity (LD₅₀ > 5 ml/kg) was found in rats and no irritation and acute toxicity was reported in Draize tests of skin or eye irritancy in a study in rabbits⁵⁸. In our previous studies, we saw no overt signs of toxicity, inflammation or discomfort after injection of HMDSO³⁹ or HMDSO nanoemulsions⁴⁴ into muscle, although no microscopic analyses were performed. Cytotoxicity analysis of HMDSO nanoemulsions showed that the half maximal inhibitory concentration (IC₅₀) at a concentration of 0.4–1% (v/v)⁴⁵ in 3T3 fibroblast cells while a IC₅₀ > 2% (v/v) was reported for PDMSO nanoemulsions in mouse neural progenitor/stem cells⁴⁶. These findings indicate that the use of siloxanes, especially longer chain siloxanes may be feasible for human applications. In particular, the use of siloxane emulsions for labelling transplanted cells and monitoring cell health is promising due to the trace amounts used. Further, the PISTOL technique used for ¹H MR oximetry in conjunction with siloxanes utilizes pulse sequence components that are readily available on clinical MRI scanners such as selective RF pulses (for excitation of the siloxane resonance and suppression of fat and water signals) and echoplanar readout (for fast T₁ mapping). This adds to the promise of clinical translation of ¹H MR oximetry using siloxanes.

In summary, we have demonstrated for the first time the feasibility of various linear and cyclic siloxanes as pO₂-sensing probes for ¹H MR oximetry. Of these OMTSO can be identified as a promising pO₂ probe which could enable faster mapping of tissue oxygenation than HMDSO without a significant drop in sensitivity. Alternatively, for applications requiring better temporal resolution or for cell labelling applications, one can use cyclic or long chain linear siloxanes, such as PDMSO, along with a recently developed pulse sequence for faster ¹H MR oximetry⁵⁵. In general, all the siloxanes studied here, with a broad range of boiling points and dynamic range of T₁’s, can be used for diversifying the applications of ¹H MR oximetry.

Materials and Methods

The linear siloxanes HMDSO, OMTSO, DMTSO, DDMPSO, and cyclic siloxanes OMCTSO and DMCPSO were purchased from Sigma-Aldrich (St Louis, MO). PDMSO (MW = 410, viscosity = 2 cSt) was purchased from Alfa Aeser (Tewksbury, MA). All the materials were used as received and all the experiments were conducted without any dilutions i.e. used ‘neat’.

For the sample preparation, each siloxane (1 ml) was placed in 4 gas-tight NMR glass tubes (Wilmad Taperlok, Buena, NJ) and saturated by bubbling for 15 minutes with varying standard of gases including 0%, 5%, 10%, and 21% O₂ (balance N₂), respectively. Gases with varying oxygen concentrations were made by mixing nitrogen and air in varying proportions in a HypoxyDial (STARR Life Sciences Corp.; Oakmont, PA). A pO₂ meter was connected in line with the output of the HypoxyDial in order to verify the accuracy of the HypoxyDial. The tubes were then sealed. For measurement of the temperature dependence of T₁, the temperature of the water pad was varied between 17 to 52 °C. A fiber optic probe (FISO Technologies Inc., Quebec, Canada) was used to measure the temperature of the tubes.

MR experiments were performed on a Varian Inova 4.7 T, Bruker BioSpec 7 T and Varian Inova 9.4 T. The tubes were placed together on a pad with circulating water and T₁ measurements were performed using previously described methods^38,39 after the tube temperature was allowed to equilibrate at the desired value for 10–20 mins using a surface or volume coil. Briefly, T₁ measurement was conducted by using pulse-burst saturation recovery with a variable TR ranging from 0.1–55 ms. T₁ data were fit to a single exponential, 3-parameter magnetization recovery equation using the Levenberg-Marquardt algorithm. The data at each temperature was then fit into the Eqs. [4–7] described earlier to obtain the corresponding calibration constants and T₁ values. MATLAB R2018b (MathWorks, Natick, MA) was used to analyze the images and compute T₁ maps and pO₂ maps. Using equation [8], the dependence of errors in pO₂ determination per 1 °C change due to temperature fluctuations was simulated for oxygenation levels in relevant hypoxic range (0 torr to 50 torr) at 37 °C.

Author contributions

S.A. performed the experiments, analyzed the data and wrote the manuscript. P.G. and U.R. performed the experiments, analyzed the data and edited the manuscript. V.D.K. designed the experiments, supervised the project, edited and reviewed the manuscript.

Competing interests

The authors declare no competing interests.

Supplementary information

is available for this paper at 10.1038/s41598-020-57889-9.