Dicarboxylic Acids as pH Sensors for Hyperpolarized ¹³C Magnetic Resonance Spectroscopic Imaging

Abstract

Imaging tumoral pH may help characterize aggressiveness, metastasis, and therapeutic response. We report the development of hyperpolarized [2–¹³C,D₁₀]diethylmalonic acid, which exhibits a large pH-dependent ¹³C chemical shift over the physiologic range. We demonstrate that co-polarization with [1–¹³C,D₉]tert-butanol accurately measures pH via ¹³C NMR and magnetic resonance spectroscopic imaging in phantoms.

TOC Image

Hyperpolarized [2-¹³C,D₁₀]diethylmalonic acid demonstrates long-lived signal enhancements for high-accuracy pH imaging via ¹³C magnetic resonance spectroscopic imaging.

Introduction

Interstitial acidification, one of the hallmarks of numerous human cancers,¹ has significant impact on the tumor microenvironment. Upregulation of aerobic glycolysis leads to proton export from tumor cells and extracellular acidification,² leading to reduced tumor uptake of chemotherapeutics,³ decreased antitumor immune cell function,⁴ and tumor invasion and metastasis.^5,6 Interestingly, interstitial pH heterogeneity within a tumor may contain important information about tumor behavior, especially considering that tumor cells tend to grow and migrate predominantly along gradients of decreasing pH.⁶ These findings suggest that pH imaging approaches may provide valuable information for clinicians wishing to grade and effectively treat tumors.

Many techniques exist for the measurement of interstitial pH in vivo,⁷ including fluorescence,^6,8 positron emission tomography,^9–12 and magnetic resonance (MR) based approaches.^13,14 The two pH imaging modalities best able to capture intratumoral pH heterogeneity with potential for clinical implementation are ¹H chemical exchange saturation transfer (CEST) MRI and hyperpolarized (HP) ¹³C magnetic resonance spectroscopic imaging (MRSI).⁷ HP ¹³C MRSI, enabled by MR signal enhancement on the order of 10⁴–10⁵ via dynamic nuclear polarization (DNP),¹⁵ has enabled the study of several metabolic and transport processes relevant to cancer, and it has been applied to human prostate cancer imaging in phase I clinical trials.¹⁶ To date, the primary HP agent for measuring interstitial pH is ¹³C-bicarbonate, which represents a ratiometric approach to calculating pH. Because the conjugate acid (H₂CO_3, in rapid equilibrium with CO₂) and base (HCO₃⁻) exhibit distinct MR resonances, the ratio of bicarbonate and CO₂ MR signal intensities can be measured in each volume element (voxel) to calculate a pH map using a modified Henderson-Hasselbalch equation.¹⁷ However, the spatial resolution is limited in part by the low signal-to-noise ratio (SNR) of CO₂, which is typically at a concentration an order of magnitude lower than bicarbonate at physiologic pH values (pK_a = 6.17 at 37 °C).^17,18 Recently, a new class of chemical shift (CS) pH probes has been reported, in which the protonated and deprotonated forms of the molecule give rise to a single MR resonance rather than two. Such HP molecules, which include ¹⁵N-pyridine derivatives,¹⁹ imidazole-¹⁵N₂,^{20 and 13}C-N-(2-acetamido)-2-aminoethanesulfonic acid (ACES),²¹ may circumvent low SNR concerns regarding the quantification of two peak intensities.

Some dicarboxylic acids are known to have second pK_a values in the physiologic range,²² as well as carbon nuclei with long T₁ relaxation time constants, making them suitable for pH imaging via HP ¹³C MRSI. Therefore, the goal of this work was to identify a dicarboxylic acid that could be hyperpolarized and used for accurate pH measurement with ¹³C MRSI.

Experimental

Full experimental details can be found in the Electronic Supplementary Information (ESI).

Dicarboxylate screening

Eleven natural abundance dicarboxylates were initially screened to measure their pH-dependent ¹³C chemical shifts (Figure 1). Aqueous solutions of these compounds were prepared containing 250 mM dicarboxylate and 250 mM urea (CS standard), and the pH was carefully adjusted with HCl or NaOH to either 6.5 or 7.4 using a standard laboratory pH meter. ¹³C NMR spectra were acquired at 11.7 T and 37 °C, referenced to urea at 163.7 ppm, and the CS change between these two pH values was measured.

Figure 1.

Investigation of dicarboxylates as ¹³C MR pH sensors. The downfield CS migration from pH 6.5 to 7.4 is listed near each labelled ¹³C nucleus. Two molecules with large CS migration over this pH range are highlighted in yellow: diethylmalonic acid (top right) and cyclopropane-1,1-dicarboxylic acid (lower left). Literature pK_a values for these molecules can be found in the ESI.

Synthesis of [2–¹³C,D₁₀]diethylmalonic acid and [2–¹³C,D₄]cyclopropane-1,1-dicarboxylic acid

Based on the pH-dependent ¹³C chemical shifts obtained, enriched syntheses of both [2–¹³C,D₁₀]diethylmalonic acid (DEMA) and [2–¹³C,D₄]cyclopropane-1,1-dicarboxylic acid (CPDA) were performed (Figure 2). Brief synthetic routes are described below, based on previously described methods.²³ [2–¹³C,D₁₀]diethylmalonic acid: [2–¹³C]diethylmalonate was alkylated with [D₅]bromoethane and saponified using NaOH. [2–¹³C,D₄]cyclopropane-1,1-dicarboxylic acid: Similar to above, but [D₄]1,2-dibromoethane was used in place of [D₅]bromoethane. All compounds were characterized via standard methods, as described in the ESI.

Figure 2.

Synthesis schemes and representative HP ¹³C T₁ decay curves at 11.7 T for (a) [2–¹³C,D₁₀]diethylmalonic acid (DEMA) 3, and (b) [2–¹³C,D₄]cyclopropane-1,1-dicarboxylic acid (CPDA) 5. The measured T₁ values at 11.7 T for DEMA and CPDA were 105.6 ± 5.2 s and 70.2 ± 4.5 s, respectively (n = 3 each).

Hyperpolarization and characterization of ¹³C dicarboxylate pH sensors

Enriched ¹³C dicarboxylate sensors were hyperpolarized via the dynamic nuclear polarization (DNP) technique and their solution-state T₁ time constants were determined. ~3.8 M DEMA in N,N-dimethylacetamide was prepared with 15 mM OX063 trityl radical and 2 mM Gd-DOTA and co-polarized with tert-butanol (tBuOH), which was formulated with OX063 in glycerol as previously described.²⁴ ~4 M CPDA in dimethyl sulfoxide was prepared with 15 mM OX063 trityl radical. After dissolution and NaOH titration (pH 6.6 - 7.5, both compounds), the HP solution-state T₁ values were determined via dynamic ¹³C MRS (5° hard pulses, flip angle correction, TR = 3 s, n = 3) at 11.7 T and 37 °C.

Titration curve for ¹³C-enriched DEMA

Based on the T₁ data obtained for ¹³C DEMA, we obtained a NMR titration curve for this compound in preparation for imaging studies. 5 mM solutions of [2–¹³C,D₁₀]DEMA and [1–¹³C,D₉]tBuOH were prepared ranging from pH 2.5 - 8.8. The CS difference between the labeled carbons was measured at 11.7 T and 37 °C, plotted versus pH, and fitted to a sigmoidal curve¹³ to obtain an MR titration curve. This MR titration curve was used to calculate pH for HP spectroscopy and phantom experiments using the ¹³C Δppm.

pH imaging phantom

Phantom studies were performed to investigate the use of HP DEMA for pH imaging. HP DEMA and tBuOH were diluted to ~5 mM each and titrated in five separate tubes to various pH values at about 37 °C. The phantom was imaged with a ¹³C 2D CSI sequence (10 × 10 matrix, 10° hard pulses, 7.5 mm isotropic in-plane resolution) on a clinical 3 T MRI scanner. After imaging, dynamic ¹³C NMR spectroscopy was performed for 3 T T₁ measurement (10° hard pulses, TR = 3 s, n = 2).

Results & Discussion

We investigated several dicarboxylic acids using ¹³C MRS to identify nuclei that demonstrated a pH-dependent chemical shift (Figure 1). All compounds tested had two carboxylic acid groups separated by either one carbon (derivatives of malonic acid) or two carbons. All molecules also had a known or predicted pK_a close to the physiologic range (i.e. near 7 - 7.4) and contained at least one carbon nucleus without directly bonded protons, making them likely to have long T₁ relaxation time constants amenable to use with hyperpolarized imaging.²⁵ Strikingly, the intermediate carbons of all malonic acid derivatives in this study demonstrated larger pH-dependent chemical shifts than did the carboxylic acid carbons themselves. This finding was somewhat surprising, considering that the carbonyl carbons are closer in proximity to the acidic protons in each molecule. Two of the malonic acid derivatives, highlighted in yellow in Figure 1, demonstrated large chemical shifts over the tested pH range: diethylmalonic acid (DEMA) and cyclopropane-1,1-dicarboxylic acid (CPDA). Of the compounds with two carbons separating the dicarboxylic acid moieties, the cis enantiomers demonstrated larger pH-dependent chemical shifts than the trans. However, these molecules exhibited smaller pH-dependent carbonyl chemical shifts than the quaternary carbons in the malonates.

Following the dicarboxylate investigation, two-step synthetic routes were developed for the isotopically-enriched, deuterated versions of DEMA and CPDA (Figure 2). These syntheses were based on a previously reported method applied to valproic acid.²³ In addition to ¹³C labeling the pH-sensitive quaternary carbon, the functional groups were deuterated for each molecule in order to lengthen the ¹³C T₁.²⁵ The overall reaction yields were 64% for DEMA and 45% for CPDA. The reaction products were confirmed to be the target molecules by both NMR (¹H, ¹³C) and high-resolution mass spectroscopy (see ESI). Based upon a preliminary T₁ comparison between the two synthesized compounds (Figure 2), we chose DEMA for further development as a hyperpolarized pH probe.

The pH-dependent chemical shift behavior of the DEMA quaternary carbon was characterized via NMR spectroscopy (Figure 3a). The CS difference between DEMA and tert-butanol (tBuOH) was plotted against pH and fitted to a sigmoidal curve. The pK_a was determined to be 7.39 under these conditions, similar to the reported value of 7.29.²⁶ The slight difference may be attributable to temperature and/or isotopic enrichment. We demonstrated that the NMR titration curve could be used to measure the solution pH from HP spectra of co-polarized DEMA and tBuOH (Figure 3b). The pH measured from HP spectra was within 0.1 pH unit of the pH measured with a conventional pH electrode (Figure 3c, n = 5). The solution-state polarization, back-calculated to the time of dissolution, was 13.7 ± 0.6% (n = 3). The T₁ values for the HP signal at 3 T and 11.7 T were 84.3 ± 1.4 s (n = 2) and 105.6 ± 5.2 s (n = 3), respectively. The T₁ was longer at the higher field strength, as might be expected for a quaternary carbon nucleus dominated by dipole-dipole relaxation.²⁷ Minimal variation in T₁ was observed over the physiologic pH range (Figure S1). The HP DEMA linewidth broadened due to chemical exchange as pH increased from 6.8 (13.1 Hz) to 7.5 (18.7 Hz), as expected based on the exchange mechanism, which is both acid- and base-catalyzed.^28,29

Figure 3.

(a) MR pH titration curve for [2–¹³C,D₁₀]DEMA. CS difference between DEMA and tBuOH is plotted against pH, and best-fit equation to the data is displayed. Inset: representative ¹³C MR spectrum of DEMA (upfield) and tBuOH (downfield). (b) HP DEMA peak at two pH values (circled points in (c)), demonstrating pH-dependent chemical shift. Spectra are referenced to tBuOH peak. (c) Plot of pH calculated from spectra using equation in (a) vs. pH electrode measurements (n = 5). pH values agree within 0.1 pH unit.

In order to demonstrate that HP DEMA could be used with spectroscopic imaging techniques, we performed an imaging phantom experiment on a clinical 3 T MRI scanner. This allowed us to simultaneously measure the pH in several solutions (Figure 4a). As before, the pH in three of five tubes was measured by using the CS difference between the HP DEMA and tBuOH peaks (Figure 4b), and these pH values agreed with electrode measurements within 0.1 pH unit (Figure 4c). Two tubes had pH values at the high and low ends of the measurable pH range. However, the extreme high and extreme low pH tubes demonstrated CS differences of 6.9 and 10.3 ppm, respectively, which agree with the minimum and maximum ppm values determined for the titration curve in Figure 3a.

Figure 4.

HP phantom imaging with [2–¹³C,D₁₀]DEMA: (a) T₂-weighted ¹H image of tubes containing ~5 mM co-polarized DEMA and tBuOH at varying pH values. Electrode pH measurements are displayed near each tube. (b) Overlaid ¹³C spectra from color-coded voxels, highlighting pH-dependent DEMA chemical shift observed via imaging. Spectra are referenced to tBuOH peak. (c) Plot of pH values calculated from spectra in (b) vs. electrode measurements, demonstrating agreement within 0.1 pH unit. The highest and lowest pH values are not plotted but demonstrated chemical shifts very close to the minimum and maximum CS differences, respectively, seen in the MR titration curve in Figure 3a.

The HP agents developed in this work, in addition to others reported previously,^19–21 represent a departure from previous techniques in HP pH imaging using ¹³C-bicarbonate. Important similarities exist between ¹³C pH agents that are “ratiometric” (eg. ¹³C-bicarbonate¹⁷), which quantify pH using the intensities of two separate ¹³C NMR resonances, and “chemical-shift” (eg. ACES,²¹ DEMA), which quantify pH based upon a change in observed ¹³C NMR frequency. In both cases, the pH-sensing molecule exists in both a protonated state and a deprotonated state, and the molecule exchanges between the two states with an overall first-order rate constant, k, representing both forward and reverse reaction rates. Ratiometric and chemical-shift sensors differ in the magnitude of the exchange rate constant, k, relative to the CS dispersion, Δf.³⁰ For ratiometric pH sensors, the exchange is much slower relative to the CS dispersion (k << Δf), leading to the observation of two distinct resonances via MR spectroscopy. In the case of ¹³C-bicarbonate, the resonances for bicarbonate and CO₂ are separated by a large CS difference of 35.5 ppm. Furthermore, the chemical exchange between the two states is rate-limited by CO₂ hydration to form bicarbonate.³¹ Conversely, simple protonation-deprotonation of ACES or DEMA is fast relative to the total CS dispersion over all pH values (k >> Δf), as is generally the case for these reactions.²⁸ Therefore, these molecules exhibit one MR resonance, with a chemical shift that is a weighted average of the chemical shifts of the protonated and deprotonated molecular states.

MR chemical-shift sensors of pH possess certain advantages and disadvantages relative to ratiometric sensors. The presence of a single peak is a significant boon concerning high spatial resolution imaging, since all HP molecules contribute to the magnitude of the single peak, and because imaging resolution is not limited by the signal of the lower of two peaks. However, these sensors also possess significant challenges. The resonant frequency, which gives a readout of pH, is also sensitive to main magnetic field inhomogeneity and changes in susceptibility throughout the imaging volume. These effects can be accounted for by co-injecting a pH-insensitive HP molecule, in our case tBuOH, that is used as a chemical shift reference. Our experimental results in phantoms demonstrate that we can use this approach for highly accurate pH imaging. The ability to resolve different pH values in vivo will depend upon image acquisition parameters, voxel size, and B₀ inhomogeneity within each voxel. High-resolution pH imaging, which may be achievable using DEMA, should provide relevant data about pH gradients within tissue. As is the case with other magnetic resonance-based pH imaging approaches,^21,32 the buffering capacity of DEMA could potentially alter the tissue pH. However, the signal gains resulting from hyperpolarization, and from the chemical shift imaging based approach compared against a ratiometric approach, have the potential to minimize these effects.

DEMA exhibits some striking properties that make it amenable to high spatial resolution imaging. Firstly, the T₁ relaxation time constant is one of the longest measured for HP ¹³C compounds.²⁵ Interestingly, the T₁ increases with field strength, as opposed to the vast majority of HP compounds ¹³C-enriched at carbonyls, whose relaxation is dominated by chemical shift anisotropy. However, the T₁ is still exceptionally long at a clinical field strength of 3 T. Combined with the high polarization obtainable for this compound, the long T₁ offers significant flexibility in terms of spatial resolution and timing of HP imaging.

Conclusions

We report a novel compound for pH measurement via ¹³C MRSI, [2–¹³C,D₁₀]diethylmalonic acid (DEMA). The pH is measured via changes in NMR chemical shift, potentially circumventing SNR limitations found with HP bicarbonate. The HP imaging pH accuracy and long T₁ values make DEMA a strong potential candidate for high spatial resolution in vivo pH mapping.