Tumor acidosis is a consequence of altered metabolism, which can lead to chemoresistance and can be a target of alkalinizing therapies. Noninvasive measurements of the extracellular pH (pHe) of the tumor microenvironment can improve diagnoses and treatment decisions. A variety of noninvasive imaging methods have been developed for measuring tumor pHe. This review provides a detailed description of the advantages and limitations of each method, providing many examples from previous research reports. A substantial emphasis is placed on methods that use MR spectroscopy and MR imaging, including recently developed methods that use chemical exchange saturation transfer MRI that combines some advantages of MR spectroscopy and imaging. Together, this review provides a comprehensive overview of methods for measuring tumor pHe, which may facilitate additional creative approaches in this research field.

1. Introduction

An emerging hallmark of cancer is the dysregulation of cellular energetics, which involves reprograming cellular energy metabolism to most effectively support neoplastic proliferation [1]. Under aerobic conditions, normal cells convert glucose to pyruvate via glycolysis in the cytoplasm and thereafter dispatch the pyruvate to the oxygen-consuming mitochondria to produce carbon dioxide and ATP (Figure 1) [2]. Under anaerobic conditions, pyruvate is converted to lactic acid in the cytosol. This anaerobic metabolism is inefficient as it produces ~18-fold fewer ATP molecules relative to mitochondrial oxidative phosphorylation. Warburg first reported that cancer cells limit their energy metabolism largely to glycolysis, even in the presence of oxygen [3]. Despite lower energy production efficiency, increased glycolysis allows the diversion of glycolytic intermediates into various biosynthetic pathways, including those generating nucleosides and amino acids [4]. This in turn facilitates the biosynthesis of the macromolecules and organelles required for assembling new cells, supporting the active cell proliferation in neoplastic disease.

The consequence of increased intracellular production of lactic acid is extracellular tumor acidosis. To maintain an intracellular pH (pHi) that is slightly alkaline (~pH 7.4), tumor cells upregulate several proton extrusion mechanisms such as the Na+/H+ exchanger (NHE), transporter, carbonic anhydrase IX, vacuolar-ATPase, and the H+/K+ ATPase [5, 6]. Excess protons are excreted into the extracellular matrix, causing the extracellular pH (pHe) of the tumor microenvironment to become acidic. In certain tumor types such as human MCF-7 mammary carcinoma, the pHe has been measured to be as low as pH 6.44 [6]. Chronic exposure to acidic pHe has been reported to promote invasiveness and metastatic behavior in several tumor types [79].

Most chemotherapeutic agents enter cancer cells via passive diffusion across the cell membrane, which requires the agent to be in a nonionized form. Thus, the cell toxicities of weak-base drugs such as daunorubicin, doxorubicin, and mitoxantrone are greatly reduced under acidic conditions [1012] (Figure 2). When the pH drops below the pKa of these weak bases, they become predominantly protonated and positively charged and are therefore less permeable to cell membranes, resulting in cellular drug resistance. Conversely, weak acid chemotherapeutic drugs such as chlorambucil, cyclophosphamide, and 5-fluorouracil have higher cytotoxicity at lower pH [12]. Hence, knowing the pHe of the tumor microenvironment can enable physicians to select the best chemotherapy based on tumor pHe and provide personalized chemotherapy for each individual patient.

Research has been conducted to investigate alkalinizing therapy as an option to increase tumor pHe that can enhance chemotherapy and counteract acid-mediated invasion and metastasis. In vitro results have shown that the cytotoxicity of doxorubicin has 2.25-fold enhancement when the pH of the cell culture media was raised from 6.8 to 7.4 [13]. In the same study, 200 mM bicarbonate in drinking water was introduced ad libitum to a mouse model of MCF-7 mammary carcinoma. In vivo results showed that the tumor volume of the MCF-7 tumors in mice treated with both bicarbonate and doxorubicin were significantly smaller than mice treated with only doxorubicin (). In another study, bicarbonate treatment was also shown to reduce the formation of spontaneous metastases, with fewer metastatic lung lesions and longer survival times in mouse models of MDA-MB-231 metastatic breast cancer [14] (Figure 3). Computer simulations showed that the addition of a moderate amount of bicarbonate in blood that is ~40% higher than normal serum concentration can reduce the amount of intratumoral and peritumoral acidosis and can almost completely eliminate tumor invasion [15]. Therefore, being able to accurately monitor pHe in tumors and normal tissues will greatly aid in evaluating the utility of alkalinizing treatment for cancer care.

2. Imaging Methods to Measure In Vivo pH

A variety of biomedical imaging methods have been developed in an attempt to measure tumor pHe (Table 1). Each method has shown a remarkable ability to produce quantitative measurements during in vivo studies, although the accuracies of these quantitative measurements have been difficult to confirm, due to a lack of a “gold standard”. Perhaps more importantly, most of these methods have disadvantages which complicate or eliminate the possibility that the method can be translated to the radiology clinic to diagnose cancer patients. The following descriptions provide a summary of each method, including advantages and disadvantages.

3. pH Measurements with Methods other than MRI

3.1. pH Electrode

The traditional ion-selective glass electrode is one of the first types of instrumentation that was used to measure pH in a human tumor. Although the use of an electrode is not an imaging method, pH measurements with an electrode are useful to compare to measurements made with noninvasive imaging methods. The pH of surface tumors such as malignant melanomatosis can be measured using a pH electrode because these tumors are easily accessible to the percutaneous technique [16]. At the completion of each procedure, the nodule has to be excised to avoid the possibility of tumor fungating. More recently, a needle-shaped pH electrode has been developed to minimize the invasiveness and improve the speed of this technique (Figure 4). The pH measurement is assumed to be accurate because this microelectrode can be accurately calibrated with external buffer solutions. The spatial resolution depends on the number of insertions and the spacing between each insertion into the solid tumor, which has ranged between 4 and 20 insertions at 0.5–1.0 cm intervals in tumor studies with canines [17]. When the tumor is not on the surface, an additional localizer step such as MRI is required to locate the tumor and assist the placement of the pH electrode into the tumor and to direct the electrodes away from necrotic areas in the tumor. Due to inevitable tissue damage during electrode insertion into tissue, the measured pH must be a weighted average of pHi and pHe and may also be affected by potential damage to microvasculature [18].

3.2. Optical Imaging with Fluorescence

Optical imaging with a fluorescence dye is a low cost imaging tool with great versatility. A fluorescence agent can be engineered to couple with a peptide, antibody, or other biomolecule, which can bind to a tumor biomarker and enhance selectivity for tumor tissue [25]. The pH can be measured independent of concentration by assessing the ratio of fluorescence signals at different emission wavelengths (Figure 5) [1921, 26] or at different fluorescence lifetimes [27]. Near-infrared (NIR) light at 700–900 nm wavelength is preferred for in vivo imaging because NIR light can propagate through tissue to a depth of approximately one centimeter and still be adequately detected relative to background signals, due to lower tissue absorption and autofluorescence in this wavelength regime [25]. Optical imaging has high sensitivity and can detect nanomolar concentrations of agents. It also has good temporal resolution on the order of seconds [14, 26, 28, 29]. Surface-accessible tumors can be imaged at excellent micron-scale spatial resolution. Based on these merits, preclinical optical imaging has been applied to surface-accessible tumors such as the rabbit ear chamber [29] and window chamber models of mammary carcinoma [14], and optical imaging has been used during image-guided surgery that provides access to the surface of tumors [30]. However, optical imaging cannot interrogate tumors that are located in deep tissues, which severely limits the utility of this imaging method for measuring tumor pHe. Furthermore, lower optical signals generated from deeper regions of surface-accessible tumors can weigh the measurement to the characteristics of the tumor surface. For these reasons, pHe measurements with optical imaging have not been translated to routine clinical use.

3.3. Positron Emission Tomography

Positron emission tomography (PET) is a widely used molecular imaging modality in both clinical and research settings. Whole-body PET imaging of mice can be performed in less than 10 minutes. To image tumor pHe with PET, a 64Cu radioactive nuclide has been conjugated to the pH low insertion peptide (pHLIP), which folds to form an α-helix and inserts itself into cell membranes when the tumor pHe is acidic (Figure 6) [2224, 31]. This method has high sensitivity and therefore can produce images with an administration of agent as low as ~0.01 ng/kg of mouse weight. Higher retention of the agent within the tumor has been shown to correlate with lower acidity [22]. However, this method depends on an equilibrium between the peptide conformations that can and cannot insert into a cell membrane, and therefore the fraction of membrane-inserted peptide has a sigmoidal relationship with pH. This causes the pH measurement to be semiqualitative instead of quantitative. Furthermore, PET has spatial resolution of ~2 mm [32] and may not be as applicable to measure pH in small tumors such as tumors in lymph nodes and may not be able to adequately assess tumor heterogeneity.

3.4. Electron Paramagnetic Resonance

Electron paramagnetic resonance (EPR) spectroscopy is a technique for studying material with unpaired electrons. Because most stable molecules have all of their electrons paired, only a few molecules are sensitive to EPR. This limitation also means that EPR offers great specificity and there is no competing background signal [35]. The use of pH-sensitive nitroxides offers a unique opportunity for noninvasive assessment of pH values from 0 to 14 in living animals with sensitivity in the μM regime [36]. The pH can be determined by measuring the frequency separation of the spectral peaks or by quenching the EPR signal within a pH-dependent polymer (Figure 7) [33]. Unfortunately, tissues must be irradiated with high power to perform EPR studies, which limits this technique to the study of small animal models and peripheral human tissues.

4. Magnetic Resonance Based Methods other than CEST MRI

Magnetic resonance imaging (MRI) and MR spectroscopy (MRS) are excellent whole-body imaging tools that provide excellent soft tissue contrast with little radiation exposure. Beyond anatomical information, recent MRI and MRS developments have focused on providing environmental biomarker evaluations (pH, temperature, and oxygen) and molecular information (proteins, enzyme activity, gene expression, metabolites, and metal ions) [37]. In recent years, innovations in MR instrumentation have drastically improved spatial and temporal resolution, and it is now possible to image in vivo tissues with ~0.1 mm spatial resolution.

4.1. Magnetic Resonance Spectroscopy

Magnetic resonance spectroscopy (MRS) has been used for more than three decades to examine metabolite distributions among living cells and tissues. The most common applications of in vivo MRS are used to detect endogenous signals from 1H, 31P, or 23Na [38]. 1H MRS is the most favorable spectroscopic method because 1H has the highest inherent sensitivity with 99.98% natural abundance and the highest gyromagnetic ratio of stable isotopes [39], and most clinical MR instruments have the capability to detect 1H MR signals. Furthermore, the most atoms in the human body are hydrogen atoms, 67% by atomic percent. Tumor pHe can be measured by detecting endogenous metabolites that have chemical shifts that are sensitive to pH. For example, pH is correlated with the chemical shift difference between the two protons on histidine, both of which are pH-sensitive. Such analyses do not require known concentrations of histidine [40]. Another example is (±)2-imidazole-1-yl-3-ethoxycarbonylpropionic acid (IEPA), which can measure pH by comparing the pH-dependent chemical shift of one proton with the pH-independent chemical shift of a second proton in the same molecule (Figure 8) [34]. Thus a ratio of the two chemical shifts correlates with pH in a concentration independent manner. However, the use of IEPA to measure pHe is questionable, because IEPA is a pH buffer and can alter the pHe by interfering with the clearance of extracellular proton to the blood stream.

31P MRS has been used to measure the chemical shift of endogenous inorganic phosphate . However, this chemical shift is weighted to reporting the pHi rather than the pHe, because ~2-3 mM of is in the intracellular compartment and only ~1 mM is in the extracellular compartment [39]. The exogenous contrast agent 3-aminopropylphosphonate (3-APP) can be used to measure pHe via 31P MRS (Figure 9) [41]. The chemical shift of 3-APP is referenced to the chemical shift of the alpha peak of APT or GPC and therefore information about the concentration of 3-APP is not required. Furthermore, 3-APP has low toxicity and can measure pHe within a physiological range of pH 6 to 8. This method of analysis allows simultaneous pHe and pHi measurements with 3-APP and [43, 44] and has been used to study alkalosis and acidification of tumor models [45, 46]. This method has been a popular choice as a cross-reference when developing new methods to measure pHe [22, 47, 48].

Even though 31P MRI has 100% natural abundance, its signal sensitivity relative to 1H MRI signals is only 6.6%. For comparison, 19F is 100% naturally abundant and has a MR sensitivity of 83% relative to 1H [39]. Furthermore 19F MRS has a relatively large chemical shift difference and an almost lack of endogenous MR signal in normal tissues. The fluorinated compound ZK-150471 emits two 19F MR signals that have a frequency difference which is dependent on pH (Figure 10) [42]. This 19F agent has been used to detect changes in tumor pHe in response to treatment with hydralazine and/or heating. In a study that compared 19F and 31P MRS at a 1 cm3 spatial resolution, 31P MRS required ~40 min of acquisition time, while its counterpart 19F MRS only required ~5 min [47]. However, pH-sensitive 19F contrast agents have practical problems that limit in vivo use, such as the instability of some fluorinated compounds and nonspecific accumulation in normal tissues, which can result in low sensitivity in the tumor tissue. To improve stability and specificity, 19F contrast agents can be encapsulated into nanogels that specifically target a tumor. The diameter of the nanogel is pH-sensitive, and indirectly measuring a change in diameter via 19F MRS provides a pH measurement [49]. In addition, the accuracy of pHe measurements with 19F agents may be questionable, partly due to many potential biomolecular interactions with the agent that can change the chemical shift of the 19F nucleus.

Despite the promise of 31P and 19F MRS for measuring tumor pHe, these MRS methods have many detriments for clinical use. Most clinical MRI instruments do not have the capability to measure isotopes other than 1H. A typical 1H MRS result requires a long acquisition time greater than 30 min and provides only coarse resolution of approximately 1 cm3. Due to the low signal to noise ratio (SNR), MRS requires a high degree of magnetic field homogeneity and minimal movement [39]. Over the years, strategies have been developed to reduce acquisition time and/or to increase spatial resolution. For example, point resolved spectroscopy (PRESS) [69] and stimulated echo acquisition mode (STEAM) [70] have been developed to enable simultaneous acquisition of spectra from multiple volumes at 1 mm3 spatial resolution. However, the relationship between SNR, spatial resolution, and acquisition time are dictated by the physics of the nuclear spins and electronics. Hence, improved resolution or reduced acquisition time is always associated with reductions in SNR. For example, a 40 mL voxel interrogated with a clinical 1.5 T MRI instrument requires 30 min acquisition time, while a 105 mL voxel requires only 6 min of acquisition time with a 1.5 T MRI instrument [40].

4.2. Hyperpolarized 13C Spectroscopy

The 13C isotope has 1.1% natural abundance and a MRS sensitivity of 0.016% relative to 1H MRS at 37°C. Dynamic nuclear polarization (DNP) can be used to increase the sensitivity of 13C MRS [71]. This technique involves cooling a 13C labeled molecule to ~1 K and then transferring polarization from electron spins to 13C nuclei with microwave irradiation. The sample is then warmed rapidly to body temperature while retaining a high level of nuclear spin polarization. A ratio of and 13CO2 from injected 13C labeled bicarbonate is correlated with pH (Figure 11) [48]. Each acquisition is ~5 sec and yields a resolution of 0.375 mm3 per voxel. Despite the ultrafast acquisition, hyperpolarized 13C has a rapid decay of ~20 sec and requires a special transceiver coil and hyperpolarizer. Furthermore, 13C labeled bicarbonate can only measure a weighted average of pHi and pHe.

4.3. Relaxivity MRI

The use of a paramagnetic metal complex as a MRI contrast agent is now widely accepted in diagnostic radiology. Agents currently approved for clinical use are based on low-molecular weight chelates of gadolinium that partition throughout all extracellular space and enhance the MR relaxation of nearby water protons. The relaxivities of some MRI contrast agents are pH-dependent, such as Gd(III) chelates that have a pH-sensitive ligand, which can block water from accessing the Gd(III) ion only under certain pH conditions [37, 7280]. Other agents currently approved for clinical use are based on superparamagnetic iron oxide nanoparticles that enhance the MR relaxation of nearby water protons. For example, iron oxide nanoparticles encapsulated within pH-responsive nanocapsules [81] and pH-sensitive hydrogels [82] exhibit different relaxation properties under different pH conditions.

Among these relaxation-based MRI contrast agents, Gd-DOTA-4Amp has been applied to measure in vivo pHe with the greatest success [85]. Knowledge on the concentration of Gd-DOTA-4Amp is required to convert the -weighted MRI signal to a pHe measurement, because the MRI signal is dependent on both concentration and the relaxation effect of the agent. A pH-unresponsive contrast agent with an analogous chemical structure can be used to account for the concentration of pH-responsive agent [37]. For example, two serial boluses of Gd-DOTA-4AmP and Gd-DOTP have been used to measure tumor pHe in C6 glioma and renal carcinoma [5052]. The -weighted MRI signal from Gd-DOTP was used to determine the temporally dynamic concentration of this agent in the tumor, which was assumed to be identical to the temporally dynamic concentration of Gd-DOTA-4AmP. Based on this concentration, the -weighted MRI signal of Gd-DOTA-4AmP was used to determine the pHe. However, this method with serial injections makes the risky assumption that the biodistributions are identical for two MRI contrast agents that are administered at different times. To avoid this assumption, a single cocktail of Gd-DOTA-4AmP and Dy-DOTP has been administered to a mouse model of C6 glioma (Figure 12) [53]. The Gd-DOTA-4Amp dominated the relaxation process generated by the agents, whereas Dy-DOTP dominated the relaxation process of the agents. The relaxation effect was then used to determine the pixelwise concentration of Dy(III), which was assumed to be identical to Gd(III). This value was used to convert the pixelwise -weighted MRI signals into maps. In a similar in vitro study, the ratio of and relaxation processes of a single agent [(GdDOTAam)33 − Orn205] was shown to be correlated with pH [86]. However, this large-molecule contrast agent has not been demonstrated with in vivo studies and may not be clinically translatable because large-molecule Gd(III) chelates are potentially toxic.

5. Chemical Exchange Saturation Transfer MRI

A chemical exchange saturation transfer (CEST) MRI experiment can selectively detect multiple proton exchanging sites within the same contrast agent. This provides a great advantage relative to relaxivity-based MRI methods, because a ratio of CEST effects from the same CEST agent is independent of concentration. If one or both of the CEST effects are dependent on pH, then this ratio can be used to accurately measure pH. CEST MRI has advantages relative to other imaging methods that measure pH. CEST MRI does not require special instrumentation such as hyperpolarizer for 13C MRS or a detector coil for 19F, 31P, or 13C MRS. CEST MRI allows full body imaging and it is not limited by depth of view as with fluorescence imaging. CEST MRI uses relatively low energy radio frequency and does not raise safety concerns like EPR or PET imaging. Most importantly, CEST MRI is noninvasive, which is a great improvement relative to invasive electrodes and microsensors.

5.1. The CEST Mechanism

CEST MRI employs an endogenous or exogenous chemical agent that has a proton which can exchange with water at a slow-to-moderate rate of 100–5000 Hz (Figure 13(a)). The protocol for CEST MRI starts with saturating a specific resonance frequency of an exchangeable proton on the CEST agent (Figure 13(b)). Saturation is a condition wherein the number of nuclear spin magnetic moments (“spins”) aligned against the field is increased at the expense of spins aligned with an external magnetic field. This soft irradiation pulse temporary disrupts the Boltzmann distribution of spins and leads to a decrease in MRI signal amplitude of the exchangeable proton in NMR spectra, which depends on the net difference of spins aligned with and against the magnetic field [87]. Due to the natural process of chemical exchange of labile protons between molecules, this proton on the CEST agent is transferred to a nearby water molecule (Figure 13(c)). The spin aligned against the field from the exchangeable proton is transferred to the population of spins for water, in exchange for a spin aligned with the field that is transferred to the population of spins for the agent. This transfer of saturation from the exchangeable proton to the water proton reduces the MRI signal amplitude of water. In practice, the CEST MRI protocol starts with a series of selective saturation RF pulses (Figure 13(d)), followed by a standard MRI acquisition sequence that can measure the amplitude of water signal throughout the image (Figures 13(d) and 13(e)). A CEST spectrum is obtained by iterating a series of saturation frequencies and recoding the normalized water signal amplitudes (Figures 13(e) and 13(f)).

The CEST agent typically has a low concentration in the 1 to 10 mM range. The MR spectroscopic resonances of the agent are not directly observable during a standard MRS acquisition due to the low concentration of agent relative to the concentration of water. However, with continuous RF irradiation (Figure 13(d)), the saturation from the agent is continuously transferred to the bulk water (Figure 13(c)), resulting in a reduction of water intensity that is observable by MRI (Figures 13(e) and 13(f)). Hence, this continuous transfer of saturation serves as amplification for the agent and allows for indirect observation of the CEST agent at low mM concentrations [88].

5.2. Endogenous CEST MRI

An ingenious application of CEST is to detect exchanging groups that are already present in the tissue, such as hydroxyl, amide and amine groups in proteins and peptides. Endogenous CEST avoids the need of an exogenous contrast agent, resulting in high impact in both in vivo applications and clinical translatability. For example, glycoCEST detects hydroxyls (0.5–1.5 ppm) from glycogen in liver and muscle [89], gagCEST detects hydroxyls (0.9–1.5 ppm) from glycosaminoglycan in cartilage [90], gluCEST detects amines (3.0 ppm) from glutamate in brain [91], and amide proton transfer (APT) detects amides (3.5 ppm) from proteins and peptides in brain [59].

Among the established endogenous CEST methods studied, only APT can measure differences in pH. The chemical exchange between the amide protons and the bulk water is base-catalyzed, so that the exchange rate decreases with decreasing pH. The pH-weighted APT imaging method has been widely used to study acute ischemic stroke in an animal model (Figure 14) [64, 83, 84, 9294].

When using APT to evaluate pH, the generated contrast depends on many parameters which compromises the measurement of pH. These other parameters include water proton concentration, amide proton concentration, spin-lattice relaxation rate, and saturation time when measuring CEST via at 3.5 ppm (Equation (1)) [59]. For this reason, the pH-weighted APT image can measure relative changes in pH, but no absolute pH value can be calculated. In the past, pH-weighted APT images have been correlated with apparent diffusion coefficient (ADC) images [64, 84, 94], isotropic diffusion-weighted images [59, 95], and lactic acid content using nonimaging methods [83, 94]. Alternative quantification methods have been applied, such as calibrating with various pH solutions of creatine [60], cross-reference with 31P MRS [59], and employing QUEST/QUESP to establish a relationship between the amide proton transfer ratio (APTR) and pH [96]. ConsiderDue to the high concentration of mobile proteins and peptides in the cytoplasm, APT mainly measures intracellular pH [59]. Furthermore, as shown in (1), an increase in APTR can potentially be caused by an increase in cytosolic protein and peptide content rather than intracellular pH. In fact, the increased protein and peptide content in the tumor is the most likely explanation for changes in APT contrast in tumors, because the intracellular pH is highly regulated by active proton exporting systems [5] and there is usually only a small difference (<0.1 pH unit) between a malignant tumor and normal tissue. Thus, APT imaging would be a more appropriate tool to provide visual information about the presence and grade of tumor based on increased content of mobile proteins and peptides, as shown in past clinical studies [60, 97].

5.3. Exogenous PARACEST Agents

Paramagnetic Gd(III) is the most commonly used contrast agent for MRI because it is the most efficient at relaxing bulk water protons, providing excellent contrast in a weighted image. However, having such a rapid water relaxation does not allow for sufficient time to generate CEST between an agent and water, which makes it unsuitable for CEST detection. On the contrary, other lanthanide ions have slower exchange rates and are more suitable as PARACEST agents [99].

PARACEST agents contain a lanthanide ion that greatly shifts the MR resonance frequency of the exchangeable amide protons from the MR frequency of water, which expands the range of MR frequencies that can generate a CEST effect, away from the direct saturation of water experienced in the APT experiment. This expanded frequency range facilitates the development of an agent with two CEST effects that have different MR frequencies. Due to the variety of the chelation chemistry, PARACEST agents can be designed to possess both pH-responsive and pH-unresponsive CEST effects within a single agent. A ratio of the two CEST effects can then be used to measure pHe without complications of the concentration term (Equation (2)) [100], unlike approaches with Gd(III) pH-responsive relaxivity agent or APT. ConsiderAime and colleagues were the first to design a series of pH-responsive ratiometric PARACEST agents (Ln-DOTAM-Gly, Ln = Pr, Nd, Eu) [5456]. The CEST effect from an amide is pH-responsive while CEST from the metal-bound water is pH-unresponsive. The ratio of CEST effects from Pr-DOTAM-Gly gave the most sensitive pH response over a range of pH 5.5–7.5. Unfortunately, owing to fast exchange rate of metal-bound water at physiologic temperature, high saturation of 87.6 μT power was required [56]. Such high saturation power exceeds the specific absorption rate (SAR) and limits the agent’s applicability for in vivo tumor pHe measurements.

To avoid detecting the CEST effect from a metal-bound water, the same group measured the CEST effects of hydroxyl groups in Yb-HPDO3A, an analogue to FDA approved ProHance (Gd-HPDO3A) (Figure 15) [61, 101]. The ratio of hydroxyl CEST effects arises from the two isomeric forms and was linearly correlated within the pH range of 5.2–6.7. Using this agent, the pHe of early and late stage murine melanoma flank tumors were measured to be pH 6.1 and 5.8, respectively. Although this approach is promising, these pHe values are far below the typical range of acidic tumors, which ranges from pH 6.4 to 7.2. Therefore, this result may have a systematic error that produces pH measurements that are too low. Another limitation is that Yb-HPDO3A can only measure pHe below pH 6.7, which is less than the physiologic pH range for tumors and normal tissues. An interesting variation of pH measurements with Yb-HPDO3A used a PAMAM dendrimer to carry multiple Yb-HPDO3A monomers [57].

Our research group has previously developed Yb-DO3A-oAA that has an aryl amine and amide (Figure 16) [65, 100]. The chemical exchange rate of an aryl amine is sufficiently slower due to the hydrogen bonding to proximal carboxylates. This allows the use of lower saturation powers (20 μT [102]) that are safer for in vivo preclinical measurements of tumor pHe. In addition, the base-catalyzed chemical exchange rates of an amide and aryl amine are different, and a ratio of these CEST effects is linearly correlated with pH within the range of 6.35–7.57. Using this agent, the pHe measured in a mammary carcinoma flank tumor was pH , which is well within the physiologic pH range [65]. Nonetheless, the use of the lanthanide ion and 20 μT saturation power is not considered safe for clinical translatability.

An alternative to measuring the magnitude of two CEST peaks is measuring a ratio of CEST at two frequencies. Wu and colleagues took a ratio of CEST signals at 55 and 49 ppm of Eu-DO3A-tris(amide), which only has one CEST effect that varies with pH between these ppm values [103]. A ratio of CEST at 55 and 49 ppm is linearly correlated with pH within the range of 6.0 and 7.6 and is independent of agent concentration.

An alternative to ratiometric PARACEST is the use of linewidth analysis of a single CEST peak (Figure 17) [62, 63]. The linewidth of the CEST effect of Tm-DOTAM-Gly-Lys is independent of agent concentration for given saturation pulse conditions. Within the physiological range of pH 6.0–7.5, linewidth analysis is less sensitive to temperature. The agent was directly injected into mouse’s leg muscle and a pH value of was measured. This example also provides the possibility of conjugating DOTAM with peptides to improve pharmacokinetic properties such as high cellular uptake and longer intracellular retention [62, 63, 104]. However, this method of analysis is not suitable for overlapping CEST effects.

5.4. Exogenous DIACEST Agents

The first documented CEST MRI contrast agent was urea, which decreased the water proton signal in ex vivo kidney tissue [105]. As the chemical exchange of an amide proton with water is base-catalyzed, the CEST effect of urea was shown to be dependent on pH [106]. This CEST effect is characterized as diamagnetic CEST (DIACEST) as opposed to PARACEST agent with metal ions.

DIACEST agents typically have exchangeable protons with a chemical shift that is less than 10 ppm from the water resonance. For example, a hydroxyl proton has a chemical shift of ~1 ppm, an amine proton resonates at ~2-3 ppm, and an amide proton typically has a MR frequency greater than 2.5 ppm, with the basic and aromatic proton shifted more down field [107]. The first reported DIACEST agent, 5,6-dihydrouracil, has a ratio of CEST effects from the two amides that follows a sigmoidal relationship with pH [105].

Many pH dependent DIACEST agents have been investigated, including poly-L-lysine (single amide, 3.75 ppm) and polyamidoamine dendrimer (three amides, 3.4–3.6 ppm) [96], poly-L-arginine (guanidyl amide and amide, 1.8 and 3.6 ppm resp.), and poly-L-threonine (hydroxyl and amide, 0.6 and 3.5 ppm, resp.) [58]. The DIACEST agents glycogen, L-arginine, and poly-L-lysine have also been encapsulated in liposomes for lymph node mapping in vivo. No pH dependence on the CEST effect was reported for these liposome-encapsulated agents, but these pioneering studies offer new approaches for studying the spatial and temporal dynamics of complex biological systems. As another intriguing approach, the hydroxyl peak of poly-L-threonine has been shown to drop as a function of pH, because the exchange rate becomes too fast compared to the chemical shift difference between water and the hydroxyl group at high pH [58].

Iopamidol (Isovue, Bracco Imaging S.p.A.) is a FDA approved contrast agent for CT/X-ray imaging that has been shown to be a DIACEST agent (Figure 18) [98, 108]. This agent has five hydroxyl protons, two amide protons that share the same MR frequency, and an additional amide proton with a different MR frequency. Due to the iodinated aryl ring, the amide protons have MR frequencies that are significantly shifted from the water frequency, which facilitates the detection of the CEST effects from these amide protons. The ratio of the CEST effects from these amide protons is correlated with pH in the range of 5.5–7.4 via a third-order polynomial function [66]. In vivo kidney pH mapping was performed with healthy mice that were acidified and alkalinized with ammonium chloride and bicarbonate and with mice that had acute kidney injury [67]. No pH mapping of a tumor was performed, presumably because the agent could not reach a sufficient concentration in the tumor for statistically significant CEST detection.

We have developed a similar CEST MRI method that uses a similar FDA approved CT/X-ray contrast agent, iopromide (Ultravist, Bayer Healthcare, Inc.) (Figure 19) [68]. The measurement of tumor acidosis with this method is known as “acidoCEST MRI.” This approach uses a rapid FISP MRI acquisition protocol to compensate for the potentially rapid in vivo pharmacokinetics of the agent and processes the results using signal averaging, Gaussian filtering, cubic spline smoothing, and Lorentzian line shape fitting to detect both CEST effects from this agent. A log10 ratio of these CEST signals is linearly correlated with pHe, without influence from the agent’s concentration or other environmental effects such as the endogenous relaxation time of the tissue. This approach can be used to measure the average pHe of the tumor and can also be used to construct pixelwise maps of tumor pHe. To date, we have used acidoCEST MRI to measure the tumor pHe in MCF-7 and MDA-MB-231 breast cancer models; Raji, Ramos, and Granta 519 lymphoma models; and orthotopic and xenograft models of SKOV3 ovarian cancer, which demonstrates the broad applicability of this method. In addition, iopromide is clinically approved for CT diagnoses, saturation power, and durations needed for acidoCEST MRI have already been implemented in the clinic, and most clinical MRI instruments can perform FISP MRI studies. Therefore, acidoCEST MRI is poised for immediate translation to the clinic (Table 2).

6. Conclusions

The noninvasive measurement of pH has evolved from optical imaging, PET, EPR imaging, MR spectroscopy, and -weighted MRI, towards the newer technique of CEST MRI. Improvements in CEST MRI pulse sequences and image analysis methods have provided the ability to perform relatively routine research studies. Methods using endogenous proteins to generate CEST, PARACEST agents, and DIACEST agents are each under continued development. The approaches that measure endogenous CEST signals or signals from DIACEST agents that are already clinically approved have great potential for clinical translation. These inevitable clinical studies may provide the foundation for providing improved diagnoses and rapid evaluations of early responses to anticancer therapies and treatments.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.


The authors would like to acknowledge the support from the Phoenix Friends of the Arizona Cancer Center, the Community Foundation of Southern Arizona, the Better than Ever Program, R01CA167183-01 and P50 CA95060. Liu Qi Chen was supported through the Anne Rita Monahan Foundation.