In vitro study of endogenous CEST agents at 3 T and 7 T

Abstract

Chemical exchange saturation transfer (CEST) has been an intensive research area in MRI, providing contrast mechanisms for the amplified detection and monitoring of biomarkers and physiologically active molecules. In biological tissues and organs, many endogenous CEST agents coexist, and their CEST effects may overlap. The interpretation of such overlapped CEST effects can be addressed when the individual CEST effects originating from various metabolites are characterized. In this work, we present the in vitro measurements of the CEST effects from endogenous CEST agents that are commonly found in biological tissues and organs, at the external magnetic fields of 3 T and 7 T and under various pH conditions. Together with the proton NMR spectra measured at 11.7 T, these CEST effects have been evaluated in consideration of the chemical exchange rates, chemical shifts, and acidities of the labile protons. Amine protons of small metabolites might not be visible at 3 T, but some of them can be probed at 7 T, wherein their CEST effects may overlap with those from coexisting amide and hydroxyl protons.

Graphical abstract

Between 3 T and 7 T, CEST contrast may carry different information regarding metabolites content. Amine protons may not be visible at 3 T by the CEST methodology, but some of them with lower acidity can be probed at 7 T.

1. INTRODUCTION

Over the past decade, chemical exchange saturation transfer (CEST) MRI has been intensively studied as a promising molecular imaging tool (1–4). In CEST, chemical exchange between the labile protons of solutes and water protons transfers the saturation of the former to the latter, which enables the amplified detection of such solutes. Combined with MRI, CEST is expected to provide contrast mechanisms for monitoring biomarkers and metabolites in tissues and organs.

Several endogenous CEST agents and their CEST activities have been identified and studied, including amide proton transfer (APT) between the amide protons of proteins and peptides and water protons (5), glycoCEST (6), gagCEST (7–9), GluCEST (10–12), CrCEST (13–17), MICEST (18,19), and glucoCEST (20), respectively ascribed to glycogen, glycosaminoglycan (GAG), glutamate (Glu), creatine (Cr), myo-inositol (MI), and glucose. In addition to the name of a molecule, each endogenous CEST application implies targeted tissues and/or organs and distinct experimental parameters (2). For example, gagCEST would be used for synovial joints and intervertebral discs while glycoCEST would be applied to the liver and skeletal muscle, although the exchangeable protons of GAG and glycogen may have similar chemical shifts and exchange rates.

In biological tissues and organs, however, several endogenous CEST agents may coexist, such as Glu, Cr, and MI in the brain (21). In GluCEST (10) and MICEST (18), the experimental parameters were chosen to pick up mostly the interesting CEST effect, which were supported by in vitro phantom and simulation studies. For example, Glu was expected to contribute to about 70 – 75 % of GluCEST, but the measured in vivo CEST effects did not show the characteristics of the Glu’s CEST effect (10).

On the other hand, the in vivo CEST measurement usually requires a significant subset of the whole Z-spectrum in order to reliably compensate the B₀-field inhomogeneity (22), and the scanning of several Z-spectra, each of which is optimized for a single CEST agent (e.g. hydroxyl, guanidine, or amine), would be time-consuming and redundant. Instead, all the coexisting CEST effects may be included in a single Z-spectrum if measured with suitable experimental parameters, and their separation may be pursued based on the experimentally measured features of the individual CEST effects.

In this work, we present the in vitro CEST effects arising from metabolites that are common in most biological tissues and organs. The Z-spectra and magnetization transfer ratio asymmetry (MTR_asym) curves were measured in clinical scanners at 3 T and 7 T for various pH values of the solutions. In addition, ¹H NMR spectra were obtained in a high-resolution NMR spectrometer at 11.7 T to seek the resonance peaks of labile protons, where larger chemical shifts may help to bring the systems into the slow exchange regime. Once the resonance peaks are observed, the chemical shifts and exchange rates of labile protons can be respectively measured by the peak positions and full widths at half maximum (FWHM’s), which give the upper limits of chemical exchange rates. With a help of the chemical shifts and exchange rates measured at 11.7 T, the CEST effects from the selected metabolites were evaluated in consideration of the acidities of their labile protons. The results may provide information useful for interpreting in vivo data and designing new applications.

2. RESULTS AND DISCUSSION

2.1. Glutamic acid

Glu is an amino acid and the most abundant excitatory neurotransmitter in the human brain (21,23). Its molecular structure at physiological pH is shown in Fig. 1a. The pK_a’s of α-carboxylic, α-amino, and the side chain protons are respectively known to be 2.19, 9.67, and 4.25 (24).

Figure 1.

Proton NMR spectra, Z-spectra, and MTR_asym curves of Glu and Cr solutions. (a) Proton NMR spectrum of a Glu solution at pH = 5.6. (b,c) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (b) 3 T and (c) 7 T of 100 mM Glu solutions titrated at different pH values. (d) Proton NMR spectrum of a Cr solution at pH = 7.4. (e,f) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (e) 3 T and (f) 7 T of 100 mM Cr solutions titrated at different pH values.

The ¹H NMR spectrum of Glu at pH 5.6 and 11.7 T is shown in Fig. 1a. One peak with the FWHM of ~700 Hz was observed at ~3 ppm downfield of the water signal, which may be assigned to the α-amino protons. No peaks were observed for the higher pH values at 11.7 T, suggesting that the chemical exchange rate may exceed the inverse of the chemical shift difference between the α-amino protons and water protons at 11.7 T.

The measured FWHM of the α-amino protons at pH 5.6 suggests that their chemical exchange would make a transition from the slow exchange regime to the fast exchange regime when the external magnetic field is lowered through ~5.5 T, which is evidenced by the Z-spectra measured at 3 T and 7 T (the black solid lines in Figs. 1b and 1c): The Z-spectrum at 7 T clearly reveals a peak at ~2.9 ppm, which becomes a plateau on the Z-spectrum at 3 T.

For pH 6.2 and 6.8, Glu produced a substantial amount of MTR_asym (the red and green dashed lines in Figs. 1b and 1c), although there were no peaks observed at ~3 ppm on the ¹H NMR spectrum at 11.7 T. For pH 6.2, the Z-spectrum and MTR_asym curve at 7 T (the green solid and dashed lines in Fig. 1c) display rather a clear peak, which suggests that the α-amino protons are still in the slow exchange regime. For pH 7.4, the Z-spectrum at 7 T (the blue solid line in Fig. 1c) shows the asymmetry around the zero frequency offset, and the corresponding MTR_asym curve (the blue dashed line in Fig. 1c) peaks at ~1.6 ppm. However, the Z-spectrum at 3 T (the blue solid line in Fig. 1b) looks symmetric around the zero frequency offset. For pH 8.0, the Z-spectra at 3 T and 7 T (the purple solid lines in Figs. 1b and 1c) are symmetric around the zero frequency offset, suggesting that the α-amino protons should be in the fast exchange regime.

2.2 Creatine

Cr is a nitrogenous organic acid and participates in the storage and transport mechanisms for high-energy phosphate bonds (21). Its molecular structure at physiological pH is shown in Fig. 1d. The pK_a’s of the carboxylic proton and protons in the guanidine group are respectively known to be 2.67 and 11.02 (24).

The ¹H NMR spectrum of Cr at pH 7.4 and 11.7 T is shown in Fig. 1d. One peak with the FWHM of ~70 Hz was observed at ~1.9 ppm downfield of the water signal, which may be assigned to the protons in the guanidine group. At the downfield tail of this peak, one small peak was located between 2.4 ppm and 2.5 ppm. The peak at ~1.9 ppm was observed through the whole range of pH investigated in this study, and its FWHM was 55 Hz, 64 Hz, 74 Hz, 69 Hz, and 77 Hz for pH 5.6, 6.2, 6.8, 7.4, and 8.0, respectively. This rather constant FWHM seems consistent with the high pK_a value of the protons in the guanidine group.

At 3 T and 7 T, the chemical exchange of the protons in the guanidine group is probably in the slow exchange regime because their chemical shift (~1.9 ppm) may be larger than the chemical exchange rate, the upper limit of which would be ~ π × 70 Hz ≈ 220 s⁻¹, based on the FWHM observed at 11.7 T. Indeed, the MTR_asym curves at 3 T and 7 T displayed a clear peak at 2 ppm for all five pH values (Figs. 1e and 1f). The peak grows as pH increases, which is consistent with the high pK_a value of those exchangeable protons.

Note that Haris et al. reported a low pK_a value (= 6.6) for the exchangeable protons in Cr (13). They estimated the pK_a value based on the integrated intensity of the proton peak as a function of pH at 9.4 T, which might not be the conventional way of measuring pK_a. Following the Henderson–Hasselbalch equation (25), pK_a is usually given as the pH of a solution when the concentrations of an acid and its conjugate base are same. The NMR peak of an exchangeable proton would disappear from the NMR spectrum, regardless of its pK_a, when the chemical exchange rate exceeds the chemical shift difference.

2.3. myo-Inositol

Inositol is a six-fold alcohol of cyclohexane. There are nine possible stereoisomers, and MI is the most widely occurring one in nature. MI serves as the structural basis for the various inositol phosphate second messengers and as an osmolyte, and is often considered as a glial marker (21).

The ¹H NMR spectrum of MI at pH 7.4 and 11.7 T is shown in Fig. 2a. Three peaks at ~0.8, ~0.9, and ~1.1 ppm were identified. Their FWHMs were 10 – 15 Hz, estimated by a fitting with multiple Lorentzian lines. The peak at ~0.8 ppm became broader for pH 5.6 and 7.4, while the other peaks were broader for pH 5.6 and 6.8.

Figure 2.

Proton NMR spectra, Z-spectra, and MTR_asym curves of MI and Cho solutions. (a) Proton NMR spectrum of a MI solution at pH = 7.4. (b,c) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (b) 3 T and (c) 7 T of 100 mM MI solutions titrated at different pH values. (d) Proton NMR spectrum of a Cho solution at pH = 7.4. (e,f) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (e) 3 T and (f) 7 T of 100 mM Cho solutions titrated at different pH values.

The Z-spectra at 3 T and 7 T (Figs. 2b and 2c) showed the CEST effects centered at 0.7 to 0.8 ppm. Interestingly, MI at pH 6.2 revealed the largest CEST effect at 3 T but the smallest at 7 T, which could be understood in relation to the broadness of the peaks observed at 11.7 T. The protons in MI may have the slowest chemical exchange rate at pH 6.2, and the chemical exchange rates at 7 T fall into the slow to intermediate exchange regimes, which cross into the intermediate to fast exchange regimes when the external magnetic field is lowered to 3 T.

2.4. Choline

Choline (Cho) is an essential nutrient and usually exists as quaternary ammonium salt. The molecular structure of the cation is shown in Fig. 2d. The ¹H NMR spectrum of Cho at pH 7.4 and 11.7 T is shown Fig. 2d. One peak with the FWHM of ~130 Hz was observed at ~1 ppm downfield of the water signal, which may be assigned to the hydroxyl proton. The peak was not observed for the other pH values. The Z-spectra at 7 T displayed CEST effects centered at ~0.8 ppm except for pH 8.0. At 3 T, those CEST effects became broader, suggesting that they transition into the intermediate exchange regime.

2.5. Chondroitin sulfate

Chondroitin sulfate (CS) is a sulfated GAG, an unbranched carbohydrate with a repeating disaccharide unit composed of N-acetylgalactosamine and glucuronic acid. CS is prevalent in cartilage and is usually found attached to a core protein to constitute proteoglycan aggregates. The repeating disaccharide unit of CS-A sodium salt is shown in Fig. 3a. Each disaccharide unit has three hydroxyl groups and one amide group, each of which includes one labile proton (7,26).

Figure 3.

Proton NMR spectra, Z-spectra, and MTR_asym curves of CS solutions. (a) Proton NMR spectra of CS solution at different pH values. The vertical lines at 0.93, 2.2, and 3.2 ppm indicate the labile proton species. (b,c) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (b) 3 T and (c) 7 T of 70 mM CS solutions titrated at different pH values.

The ¹H NMR spectra of CS at 11.7 T and various pH values are shown in Fig. 3a. At pH 7.4, two peaks appeared at the downfield side of the water signal. The peak at ~0.9 ppm is broader with the FWHM of ~200 Hz and may be assigned to the hydroxyl protons. The other peak at ~3.2 ppm is narrower with the FWHM of ~30 Hz and can be assigned to the amide protons. With decreasing pH, the intensities of the two peaks became smaller and disappeared at pH 5.6, while several narrow resonances appeared at 1.1 to 1.2 ppm and 3.3 to 3.7 ppm for pH 5.6 and 6.2 and at 2.2 ppm for pH 5.6. The peaks between 1.1 ppm and 1.2 ppm and between 3.3 ppm and 3.7 ppm should be respectively the protonated hydroxyl and amine protons, while the peak at 2.2 ppm was assigned to hydroxyl protons (26), which may be different from those hydroxyl protons observed at pH 7.4.

According to the FWHMs measured at 11.7, the hydroxyl protons at ~0.9 ppm may be in the slow to intermediate exchange regime at 7 T and in the intermediate to fast exchange regime at 3 T. For pH 5.6 and 8.0, the MTR_asym was observed to be reduced at 3 T as seen in Fig. 3b. Otherwise, the hydroxyl protons seem to produce observable MTR_asym at both of 3 T and 7 T, as seen in Figs. 3b and 3c.

For the amide protons, on the other hand, it has been reported that their CEST effects may be compensated in the MTR_asym analysis by the so-called nuclear Overhauser effects (NOE) between the protons in the N-acetyl groups of CS and water protons (7). While the NOE were not clearly observed in this study, the CEST effects from the amide protons were observed to be small compared with those from the hydroxyl protons, due to their slow chemical exchange with the water protons. The CEST effects were most pronounced for pH 6.8 at 7 T as seen in Fig. 3c.

2.6. Bovine Serum Albumin

Serum albumin is a globular protein and the most abundant blood protein in mammals. In the CEST field, bovine serum albumin (BSA) has been used as a system providing CEST effects and NOE without confounding magnetization transfer (MT) effects (27).

The ¹H NMR spectra measured at 11.7 T showed broad and overlapped resonances at 2 to 4 ppm downfield of the water signal, as seen in Fig. 4a. There exists a peak at ~2.8 ppm, the intensity of which varies with pH. While the Z-spectra measured at 3 T showed a little bump centered at ~3 ppm for pH 6.5, 6.8, and 7.1, as seen in Fig. 4b, the Z-spectra measured at 7 T displayed a CEST peak at ~2.7 ppm except for pH 8.0, as seen in Fig. 4c. Besides, the MTR_asym curves show a broad asymmetry along the frequency offsets, probably due to fast exchanging species. It seems that the experimental conditions used in this study did not produce noticeable NOE on the upfield side of the water signal.

Figure 4.

Proton NMR spectra, Z-spectra, and MTR_asym curves of BSA solutions. (a) Proton NMR spectra of BSA solution at different pH values. The vertical lines at 0.93, 2.2, and 3.2 ppm indicate the labile proton species. (b,c) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (b) 3 T and (c) 7 T of 12 wt% BSA solutions titrated at different pH values.

2.7. Glycogen & Glucose

Glycogen is a multibranched polysaccharide of glucose, which is used in many organisms as an energy source and metabolic intermediate. The schematic of glycogen structure and the molecular structure of glucose are shown respectively in Figs. 5a and 5d, together with the ¹H NMR spectra at 11.7 T and pH 7.4.

Figure 5.

Proton NMR spectra, Z-spectra, and MTR_asym curves of glycogen and glucose solutions. (a) Proton NMR spectrum of a glycogen solution at pH = 7.4. (b,c) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (b) 3 T and (c) 7 T of 100 mM glycogen solutions titrated at different pH values. (d) Proton NMR spectrum of a glucose solution at pH = 7.4. (e,f) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (e) 3 T and (f) 7 T of 100 mM glucose solutions titrated at different pH values.

The ¹H NMR spectra of glycogen measured at different pH values revealed a resonance with the FWHM of ~200 Hz at ~1.2 ppm downfield of the water signal except for pH 6.8, at which the peak disappeared. In addition, two tiny peaks were found at ~2.2 ppm and ~3.0 ppm downfield of the water signal. While no peaks were observed on the ¹H NMR spectra of glucose, the resonances of the exchanging hydroxyl protons were reported for low pH and low temperature (pH 3.5 and 4°C (1)). The reported chemical shifts of glucose (1.2 ppm, 2.2 ppm, and 2.8 ppm downfield of the water signal (28)) seem to match those observed with glycogen in this study.

As expected from the FWHM measured at 11.7 T, the Z-spectra of glycogen indicated that the protons at ~1.2 ppm are in the intermediate to fast exchange regime at 3 T (Fig. 5b) and in the slow to intermediate exchange regime at 7 T (Fig. 5c). The MTR_asym curves manifest the asymmetry beyond the frequency offset of 2 ppm, which may originate from the protons at ~2.2 ppm and ~3.0 ppm downfield of the water signal. For the same pH, glucose generally displayed bigger and broader CEST effects than glycogen (Figs. 5e and 5f), which is consistent with the literature (28). At 7 T, the Z-spectra for pH 6.2, 7.4 and 8.0 even manifested a dip at ~2.8 ppm (Fig. 5f).

2.8 Glutamine

Glutamine (Gln) is the amide of Glu, in which the hydroxyl group of the side chain of glutamic acid is replaced by an amine functional group. Its molecular structure at the physiological pH is shown in Fig. 6a. The pK_a’s of α-carboxylic and α-amino protons are respectively known to be 2.17 and 9.13 (24), while it seems that the pK_a of the amide protons is not available.

Figure 6.

Proton NMR spectra, Z-spectra, and MTR_asym curves of Gln solutions. (a) Proton NMR spectrum of a Gln solution at pH = 7.4. (b,c) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (b) 3 T and (c) 7 T of 100 mM Gln solutions titrated at different pH values.

The ¹H NMR spectra of Gln at 11.7 T are shown in Fig. 7a. Except for pH 8.0, three peaks were observed at ~2.1 ppm, ~2.8 ppm, and ~2.9 ppm downfield of the water signal. For pH 8.0, there was only one peak at ~2.9 ppm with the FWHM of ~80 Hz. The FWHMs of the peaks for the other pH values were between ~30 Hz and ~55 Hz.

Figure 7.

Z-spectra and MTR_asym curves of Aps, GABA, and taurine. (a,b) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (a) 3 T and (b) 7 T of 100 mM aspartic acid titrated at different pH values. (c,d) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (c) 3 T and (d) 7 T of 100 mM GABA titrated at different pH values. (e,f) Z-spectra (solid lines) and MTR_asym curves (dashed lines) at (e) 3 T and (f) 7 T of 100 mM taurine solutions titrated at different pH values.

In consideration of the pK_a of its α-amino protons and the existence of the amide protons, the CEST effects from Gln (Figs. 7b and 7c) may be contrasted to those from Glu (Figs. 1b and 1c). The MTR_asym of Gln decreases as the pH increases from 5.6 to 7.4, like the MTR_asym of Glu, but increases back when the pH becomes 8.0, which may be due to the amide protons of Gln that Glu does not have.

2.9 Aspartic acid, γ-Aminobutyric acid, and taurine

Aspartic acid (Asp) is an α-amino acid, the molecular structure of which is shown in Fig. 7a at the physiological pH. The pK_a’s of the α-carboxylic, α-amino, and side-chain protons are 2.09, 9.82, and 3.86, respectively (24). The Z-spectra and MTR_asym curves of Asp measured at 3 T and 7 T are shown in Figs. 7a and 7b, which are similar to those of Glu, as expected from their similar pK_a values. However, we could not observe any resonances from the α-amino protons at 11.7 T for any pH values used in this study. When one of the α-amino protons is replaced by an acetyl group, Asp turns into N-acetylaspartic acid (NAA) and its amine becomes an amide. The ¹H NMR spectrum of NAA at 11.7 T revealed a peak at 3.2 ppm downfield of the water signal, which may be assigned to the amide proton. NAA solutions did not manifest any CEST effects at 3 T and 7 T for any pH values used in this study, probably because the amide proton may have very high pK_a compared to the amine protons in Asp.

γ-Aminobutyric acid (GABA) is the most abundant inhibitory neurotransmitter in the brain (29). Its molecular structure at the physiological pH is shown in Fig. 7c. The pK_a’s of the carboxylic and amine protons are 4.23 and 10.43, respectively (24). The ¹H NMR spectrum at 11.7 T revealed the resonance of the amine protons at ~2.8 ppm downfield of the water signal for pH 5.6 only, with the FWHM of ~260 Hz. The Z-spectra and MTR_asym curves measured at 3 T and 7 T shown in Figs. 7c and 7d display the dependence of its CEST effects on the pH values consistent with its amine protons’ pK_a value, which is higher than those of Glu and Asp.

Taurine is a sulfonic acid, the molecular structure of which is shown in Fig. 7e at the physiological pH. The pK_a’s of the carboxylic and amine protons are 1.5 and 8.74, respectively (24). The ¹H NMR spectra at 11.7 T revealed a single peak with the FWHM of ~100 Hz at ~3 ppm downfield of the water signal for pH 5.6 and triplets centered at the same frequency position for pH 6.2 and 6.8. For pH 7.4 and 8.0, no peaks were observed. The Z-spectra and MTR_asym curves measured at 3 T and 7 T are shown in Figs. 7e and 7f, which seem consistent with the amine protons’ pK_a value.

2.10. 3 T vs. 7 T

Additional CEST scans were performed in a high-throughput setting with a set of selected phantom solutions at pH 7.4. The MTR_asym curves obtained at 3 T and 7 T are presented in Fig. 8. Notice that the positive and negative peaks appearing at the smallest frequency offset (40 Hz at 3 T and 100 Hz at 7 T) may be due to the imperfect B₀ correction and that the MTR_asym curves shown in Fig. 8 may look different from those individually presented in Figs. 1, 2, 6, and 7, probably because of the dependence of the B₁ field and signal sensitivity on the position of a phantom inside the coils.

Figure 8.

MTR_asym curves of MI, Cr, Glu, Cho, Asp, GABA, taurine, and Gln at the physiological pH. (a) MTR_asym curves at 3 T. (b) MTR_asym curves at 7 T.

Between 3 T and 7 T, the amine protons of Glu, Asp, and GABA show the most noticeable changes in the MTR_asym curves, which showed almost no CEST effects at 3 T but manifested ones spreading over the frequency offsets at 7 T. At both of 3 T and 7 T, the CEST effects from Cr, MI, and Cho were focused at the chemical shifts of their exchangeable protons, in contrast to very little CEST effects from Gln and taurine. These observations suggest that the CEST effects from Cr and MI could be better identified at 3 T without the overlap by the CEST effects from Glu, Asp, and GABA. Higher external magnetic fields are beneficial to the CEST measurement: The frequency separation between different CEST agents will increase, which also helps to reduce the direct saturation of the water signal during the CEST measurement, and the proton species with higher exchange rates can be probed. As seen in Fig. 8, however, the contribution from more proton species at the higher magnetic fields could complicate the analysis and interpretation if some of them are in the intermediate exchange regime.

The behavior of amine protons seems consistent with the simple proton exchange model (30,31) in consideration of the pH of solutions and the pK_a values and chemical shifts of the amine protons. The amine protons of Glu, Gln, Asp, GABA, and taurine have similar chemical shifts (2.7 ppm – 3 ppm) but span a range of pK_a values (8.7 – 10.4). Therefore, the difference between their CEST effects may be ascribed to the difference in the pK_a value. Indeed, the pK_a values of the amine protons of Glu, Asp, and GABA are higher than those of Gln and taurine, so their chemical exchange rates are low enough to enter into the intermediate exchange regime at 7 T.

2.11. Miscellaneous

The exchange regimes at 3 T and 7 T for the metabolites studied in this work are summarized in Table 1, together with the chemical shifts and π times FWHMs measured at 11.7 T. While these results may give some insights into the applications of CEST methodology on biological tissues and organs, such extrapolation should proceed with caution. Chemical exchange phenomena depend on temperature as well as exchange catalysts other than H⁺ and OH⁻ (30–32), and our in vitro CEST measurements were performed at room temperature and without any other exchange catalysts such as inorganic phosphate ions HPO₄²⁻ and H₂PO₄⁻. In addition, the concentrations of CEST agents used in this work were one or two orders of magnitude higher than the in vivo concentrations (33), which were intended to intensify CEST effects.

Table 1.

Functional groups carrying exchangeable protons for the metabolites studied in this work, with exchange regimes at 3 T and 7 T for pH 7.4, chemical shifts measured at 11.7 T.

	Exchange regime		Chemical shift (ppm)	π × FWHM (s−1) at 11.7 T and 25 °C
	3 T	7 T
Hydroxyl	intermediate
MI			~0.8, ~0.9, ~1.1	90 – 200 (pH = 7.4) a
Cho		intermediate	~1	~ 400 (pH = 7.4)
CS			~0.9, ~2.2	~ 600 (pH = 7.4)
Glucose			n/a b	n/a
Glycogen			1.2, 2.2, 3.0	~ 600 (pH = 7.4)
Guanidine
Cr	slow	slow	~2	~ 220 (pH = 7.4) c
Amine	fast
GABA			~2.8	~ 800 (pH = 5.6)
Asp		intermediate	n/a	n/a
Glu			~3	~ 2000 (pH = 5.6) d
Taurine		fast	~3	~ 300 (pH = 5.6)
Amide
CS	slow	slow	3.2	~ 90 (pH = 7.4)

^aIn (18), the exchange rate was reported to be 600 s⁻¹ at pH 7.4 and 9.4 T.

^bIn (28), the chemical shifts of glucose were reported to be 1.2, 2.2, and 2.8 ppm.

^cIn (13), the exchange rate was reported to be 950 s⁻¹ in phosphate buffered saline (PBS) at 9.4 T and 37°C.

^dIn (10), the exchange rate was reported to be 5500 s⁻¹ in PBS at pH 7.0, 7 T, and 37°C.

In general, the optimization of the parameters for the off-resonance pre-saturation irradiation has been suggested for individual CEST applications (10–19). Such optimizations are illustrated in Fig. 9 with simulated Z-spectra and MTR_asym curves. The simulations were performed by numerically solving the Bloch-McConnell equations for a two-pool exchange model (38), based on the parameters measured from the ¹H NMR spectra shown in Figs. 2d, 1d, and 1a. The duration and RF amplitude of the off-resonance pre-saturation irradiation were searched to maximize the peak height of MTR_asym curves. The CEST effects from the amine protons interfere with the other CEST effects, even when the duration and RF amplitude are optimized for the guanidine and hydroxyl protons (Figs. 9a and 9b). From the optimal duration and RF amplitude for the guanidine protons (Fig. 9b) to those for the amine protons (Fig. 9c), the increases of the MTR_asym at 1500 Hz are ~5 times and ~2.4 times for the guanidine and amine protons, respectively. The optimal RF amplitude for the amine protons is large due to the higher exchange rate (k_ex,amine = 2000 s⁻¹), so the pre-saturation irradiation may have the excitation range or bandwidth large enough to simultaneously induce multiple CEST effects. Notice that the optimal conditions for the hydroxyl and guanidine protons happen to be very close (Figs. 9a and 9b).

Figure 9.

Simulated CEST effects at 11.7 T with the duration (t_sat) and RF amplitude (ω_1,sat) of the off-resonance pre-saturation irradiation optimized for individual CEST agents. Z-spectra and MTR_asym curves with the parameters optimized for (a) the hydroxyl proton of Cho, (b) the guanidine protons of Cr, and (c) the amine protons of Glu. In each graph, black, red, and blue lines represent amine, guanidine, and hydroxyl protons, respectively.

For in vivo applications, the scan time and specific absorption rate (SAR) as well as hardware limitations are additional factors to determine the duration and RF amplitude of the off-resonance pre-saturation irradiation. As shown in Fig. 9, the optimal durations come much sooner than the time for a steady state, indicating that the T₁ relaxation of the abundant pool prevents CEST effects from being further accumulated. Smaller RF amplitude would be favorable in order to reduce SAR and to decrease overlaps between CEST effects at different frequency offsets and the so-called spillover effect (35) caused by the direct saturation of the abundant pool, as demonstrated in Fig. 10 by simulated Z-spectra and MTR_asym curves for three RF amplitudes under the same RF energy deposition by the off-resonance pre-saturation irradiation.

Figure 10.

Simulated CEST effects at 11.7 T under three different durations (t_sat) and RF amplitudes (ω_1,sat) of the off-resonance pre-saturation irradiation with the same RF energy (ω_1,sat² × t_sat). (a) t_sat = 2.0 sec, ω_1,sat/2π = 50 Hz. (b) t_sat = 1.0 sec, ω_1,sat/2π = 100 Hz. (c) t_sat = 0.5 sec, ω_1,sat/2π = 200 Hz. In each graph, black, red, and blue lines represent amine, guanidine, and hydroxyl protons, respectively.

However, there would exist a lower limit for the RF amplitude of the off-resonance pre-saturation irradiation. In biological tissues and organs, for example, semisolid macromolecules may cause MT effects and shorten T₂ of bulk water. Unless such MT effects happen to be symmetric with respect to the water signal, they may make additional contributions to MTR_asym, which interferes with the measurement of CEST effects. Such asymmetric MT effects may be removed from MTR_asym through the so-called uniform magnetization transfer (uMT) phenomenon (35,36). The uMT phenomena assume the efficient and uniform saturation of the semisolid macromolecules by the off-resonance pre-saturation irradiation simultaneously at two frequency offsets (37), which would be facilitated by stronger amplitudes. In our previous works, we could achieve the uMT phenomena with B_1,rms of 1.4 – 1.7μT for each frequency component (35,36).

3. CONCLUSION

The CEST effects from endogenous CEST agents were measured in vitro at 3 T and 7 T. These CEST effects were analyzed in consideration of the chemical shifts and pK_a’s of the exchangeable protons and the pH of solutions. For small metabolites, amine protons might not be visible at 3 T, but some of them with higher pK_a values can be probed at 7 T. These amine protons, however, fall into the intermediate exchange regime at 7 T, and their CEST effects may hence overlap with those from coexisting amide and hydroxyl protons. Conversely, the dependence of CEST effects on the external magnetic field can be exploited to unravel individual CEST effects.

4. EXPERIMENTAL

4.1. Solutions

The following solutes (all purchased from Sigma-Aldrich, St. Louis) were dissolved into deionized water: L-glutamine (CAS # 56-85-9), creatine (CAS # 57-00-1), myo-inositol (CAS # 87-89-8), choline chloride (CAS # 67-48-1), chondroitin sulfate A sodium salt (from bovine trachea, CAS # 39455-18-0), albumin from bovine serum (CAS # 9048-46-8), glycogen (type II, from oyster, CAS # 9005-79-2), D-(+)-glucose (CAS # 50-99-7), L-glutamic acid monosodium salt monohydrate (CAS # 6106-04-3), L-aspartic acid (CAS # 56-84-8), γ-aminobutyric acid (CAS # 56-12-2), taurine (CAS # 107-35-7), and N-acetylL-aspartic acid (CAS # 997-55-7). From each solution, five phantoms with different pH values (5.6, 6.2, 6.8, 7.4, and 8.0) were prepared.

4.2. CEST experiments at 3 T and 7 T

At 3 T and 7 T, the CEST measurements were performed in a high-throughput setting on whole-body Siemens scanners (Siemens, Erlangen, Germany) respectively with a Tx/Rx 15-channel knee coil (Siemens, Erlangen, Germany) and a volume-transmit, 24-element receive head coil array (Nova Medical, Boston, MA). For the signal acquisition, a segmented GRE acquisition with centric phase encoding order was used. The following parameters were used for the GRE sequence, flip angle = 15°, TR = 24 ms, TE = 3.5 ms, dwell time = 15 μs, slice thickness = 5 mm, matrix size = 96 × 96. The field of view (FOV) was 170 mm × 170 mm. For the off-resonance pre-saturation irradiation, a train of 20 Gaussian pulses was used. Each Gaussian pulse was 100 ms long, followed by a delay of 500 μs, and the nominal flip angle was 1440° (B_1,rms = 1.4 μT). The frequency offsets for the pre-saturation irradiation were varied from-1000 Hz to 1000 Hz with a step size of 40 Hz at 3 T and from-2500 Hz to 2500 Hz with a step size of 100 Hz at 7 T, respectively. The Gaussian pulse can perturb a spin within a frequency range between-50 Hz and +50 Hz from its frequency offset. For the WASSR acquisition (22), a train of two 100 ms-long 180° Gaussian pulses with an inter-pulse delay of 500 μs, were used as the off-resonance pre-saturation irradiation, and the frequency offset was varied from-250 Hz to 250 Hz with a step size of 10 Hz at 3 T and from-500 Hz to 500 Hz with a step size of 20 Hz at 7 T, respectively. For each phantom solution, a Z-spectrum was obtained by averaging the signal intensity, spline-interpolated, and B₀-corrected, from which a MTR_asym curve was evaluated.

4.3. Proton NMR spectra at 11.7 T

All NMR experiments were performed using a Bruker Avance 500 MHz NMR spectrometer equipped with a broadband observe (BBO) probe. Each solution was prepared in a 5-mm NMR tube, and the temperature was set to be 298 K. Since the solutions did not contain deuterium, the deuterium lock was not used. Proton NMR spectra were obtained by using a 90° pulse with its duration being about 10 to 12 μs and averaging 64 free induction decay signals acquired for 0.5 sec with the spectral width of 100 kHz. The relaxation delay from the end of the acquisition and the beginning of the next 90° pulse was 5 sec.

4.4 Numerical simulations

Z-spectra for hydroxyl, guanidine, and amine protons were numerically obtained by solving the Bloch-McConnell equations (38). The parameters used for hydroxyl, guanidine, and amine protons were from the ¹H NMR spectra at 11.7 T shown in Figs. 2d, 1d, and 1a, respectively. The T₁ (T₂) relaxation times for the abundant and solute pools were respectively 4 sec (30 ms) and 4 sec (60 ms). The T₁ relaxation times of the abundant and solute pools were assumed to be same, and the value was from the saturation recovery experiment on water protons. The T₂ relaxation times were determined from the FWHMs of water signal and methyl and methylene signals, which were measured about 10 Hz and 5 Hz, respectively. The sizes of solute pools were determined from the ratios of the integrated peak intensities between the bulk water and exchangeable protons, and 0.1%, 0.4%, and 0.3% were used respectively for hydroxyl, guanidine, and amine protons, corresponding to the solute concentration of 110 mM. The RF amplitude and duration of the off-resonance pre-saturation irradiation were respectively varied from 10% to 1000% of exchange rates and T₁ relaxation time with the step size of 10%, to find the combination to maximize MTR_asym.