Characterization of amine proton exchange for analyzing the specificity and intensity of the CEST effect: from humans to fish

Chemical exchange saturation transfer (CEST) at about 2.8 ppm downfield from water is characterized besides other compounds by exchanging amine protons of relatively high concentration amino acids and is determined by several physiological (pH, T) and experimental (B0, B1, tsat) parameters. Although the weighting of the CEST effect observed in vivo can be attributed mainly to one compound depending on the organism and organ, there are still several other amino acids, proteins and molecules that also contribute. These contributions in turn exhibit dependences and thus can lead to possible misinterpretation of the measured changes in the CEST effect. With this in mind, this work aimed to determine the exchange rates of six important amino acids as a function of pH and temperature, and thus to create multi‐pool models that allow the accurate analysis of the CEST effect concerning different physiological and experimental parameters for a wide variety of organisms. The results show that small changes in the above parameters have a significant impact on the CEST effect at about 2.8 ppm for the chosen organisms, i.e. the human brain (37 °C) and the brain of polar cod (1.5 °C), furthermore, the specificity of the CEST effect observed in vivo can be significantly affected. Based on the exchange rates ksw(pH, T) determined for six metabolites in this study, it is possible to optimize the intensity and the specificity for the CEST effect of amino acids at about 2.8 ppm for different organisms with their specific physiological characteristics. By adjusting experimental parameters accordingly, this optimization will help to avoid possible misinterpretations of CEST measurements. Furthermore, the multi‐pool models can be utilized to further optimize the saturation.


| INTRODUCTION
Chemical exchange saturation transfer (CEST) imaging enables the indirect detection of endogenous and exogenous compounds such as amino acids, proteins and other molecules with exchangeable protons exhibiting, e.g., amide (-NH), amine (-NH 2 ) or hydroxyl (-OH) groups. The principle of CEST is the change of the NMR signal of water after the selective saturation of magnetization of the exchangeable protons. The CEST effect is mainly determined by the concentration of compounds and the exchange rate (k sw ) between the labile bound protons from these and water, whereby k sw in turn is determined by the chemical structure and physiological parameters. Furthermore, the ratio of k sw and the chemical shift difference (Δω) between the resonances of water and the exchangeable protons of the compound determines the intensity of the expected CEST effect. The different exchange regimes are defined as slow if ksw Δω ( 1, intermediate if ksw Δω1 , and fast if ksw Δω ) 1: 1 CEST imaging is mostly sensitive to the slow-to intermediate-exchange regime defined by k sw s À1 Â Ã ≤ Δω rad s Â Ã and for this reason is dependent on the field strength used. Therefore, a trend in medical research is to focus on higher magnetic fields (B 0 ) for CEST imaging. 2 The CEST effect, being observed after saturation at about 2.8 ppm downfield from water is, among other things, largely determined by exchangeable amine protons of relatively highly concentrated amino acids, which play an important role in the metabolism of the central nervous system of vertebrates and fulfil a variety of specific functions. 3 For example, they form the basis for the synthesis of proteins and nucleic acids or are involved in energy metabolism and neurotransmission. Their non-specific functions include, e.g., their involvement in osmoregulation. Besides glutamate (Glu), whose CEST effect has already been established in preclinical research, [4][5][6][7][8][9] several amino acids play important roles in the metabolism of the central nervous system. Due to their function and concentration, the amino acids alanine (Ala), γ-aminobutyric acid (GABA), aspartate (Asp), glutamine (Gln) and taurine (Tau) are also particularly relevant for the consideration of the total CEST effect at about 2.8 ppm. Lee et al.
published the examination of CEST effects from endogenous agents that are typically found in biological tissues and organs, e.g. Glu, Gln, GABA and Tau, at different pH conditions and B 0 values. 10 In combination with 1 H NMR spectra, these CEST effects have been analyzed in consideration of the k sw , chemical shifts and acidities of the labile protons. 10 Khlebnikov et al. have also studied the most important metabolites that contribute to the CEST effect in the human brain. 11 Their exchange rates were investigated at physiological pH and 37 C. Additionally, the contributions of these metabolites to the total CEST effect were determined at different field strengths. However, the k sw as well as their pH and temperature (T) dependences for the metabolites were only investigated selectively and not for a wider range of values, 4,10-13 which does not allow the exchange rate to be determined as a function of pH and temperature (k sw (pH, T)).
Although the in vivo observed CEST effect is often dominated by one compound, others contribute as well. Cai et al. could show that 70-75% of the CEST effect observed in vivo in the human brain is determined by Glu, while 25-30% remained from other products. 4 However, these percentages may change depending on physiological (i.e. concentration, temperature and pH) and experimental parameters (i.e. B 0 , saturation amplitude (B 1 ) and saturation time (t sat )). Therefore, a determination of k sw (pH, T) will simplify both the analysis of CEST effects and the optimization of the pulse sequences used. 14 There is a growing interest in alternative and non-endothermic animal models for in vivo studies (e.g. birds 15 or fish [16][17][18][19][20][21], which considerably extends the number of compounds of interest and the range of temperature and pH in which this methodology is applied. Recently, the application of the TauCEST effect allowed us to determine relative changes in intracellular pH (pH i ) in the brain of the polar cod (Boreogadus saida) due to changes in CO 2 concentrations under polar conditions in vivo. 22 This is based on the exceptionally high Tau concentration in saltwater fish (about 20 mM in the brain of polar cod (B. saida) 23 and up to 65.7 mM in the brain of stingray (Dasyatis sabina) 24 ) compared with adult rodents (around 2 mM to 6-8 mM), 25-28 due to their osmoregulatory function. This aspect, in combination with the body temperature of 1 C, results in a contribution of Tau of about 65% to the observed total CEST effect in the physiological pH range in vivo. 22 Even in conventional preclinical research, other physiological conditions can occur during the measurements, for example, due to temperature changes. Pyrexia and anesthesia without an external control system can cause changes between 32 and 40 C even at a basal body temperature of 37 C. 29 Furthermore, the specificity of the in vivo observed CEST effect changes due to changes in pH (e.g. hypercapnia, tumor tissue 30,31  direct saturation to retain nearly the maximal amide proton transfer (APT) effect. 32 This methodical approach utilized the fact that direct saturation width and MTC magnitude increase strongly with B 1 , while the effect of compounds with relative slowly exchanging protons as in APT only depends slightly on B 1 .
However, the amino acids have similar physical properties, including nearly identical resonance frequencies, which makes experimental separation considerably more difficult. This requires a more detailed analysis of these metabolites and the effect of physiological and experimental parameters. This work aimed to simulate and analyze as accurately as possible the intensity and specificity of the expected in vivo CEST effect of amino acids for different organisms and their corresponding body temperatures, based on a determination of k sw (pH, T) for six important brain metabolites. For the metabolites investigated in more detail, z-spectra and their corresponding asymmetry curves were determined experimentally for 36 combinations of pH values and temperatures. For the amino acids, which are particularly relevant for analyzing the total CEST effect at about 2.8 ppm, the exchange rates were determined by fitting the measured data with the Bloch-McConnell equation. Subsequently, the maximum CEST effect was investigated as a function of the experimental parameters B 0 , B 1 and t sat for the human brain as a representative of the clinical and preclinical research (37 C body temperature) and for polar cod as an example for a new field of application (1.5 C).
Finally, the intensity and specificity of the CEST effect are analyzed as a function of temperature and pH value using optimal experimental parameters for both organisms. in vitro measurements. To minimize the variables in the determination of exchange rates, the longitudinal relaxation times (T 1 ) of water were determined for each temperature and each series of measurements using a RARE (rapid acquisition with relaxation enhancement) sequence (T R = 5000 ms, 3000 ms, 1500 ms, 800 ms, 400 ms, 227 ms). was mixed with filtered and deionized water, heated to boiling and then immersed in a water bath at 46 C. After 15 min the agarose solution was filled into a 50 mL Falcon tube, into which the NMR tubes were inserted. 13,34 The phantoms were wrapped with heating tubings connected to a circulation thermostat (Lauda Eco RE 630S, Lauda-Brinkmann, Delran, NJ) for measurements at defined temperatures (1-37 C), while temperature measurements inside the magnet were performed with a two-point calibrated fiber-optical thermometer (Luxtron 504, Polytec, Waldheim, Germany).

| Methods of data analysis
The CEST images were analyzed by determining the mean signal intensities from different regions of interest in the tubes of the phantoms using ImageJ (Version 1.48v, US National Institutes of Health, Bethesda, MD). Further calculations and data fitting were done using programs written in Scilab (Version 6.0.1, Scilab Enterprises, Versailles, France). The signal minimum in the z-spectrum at the water signal was fitted to a Lorentzian line shape to correct for B 0 inhomogeneities. 35 The exchange rates and the transverse relaxation times (T 2 ) of water were determined by fitting the Bloch-McConnell equations, 36 modified for a two-pool chemical exchange, to the experimental data. 13,37 A saturation time of 12 s allowed us to postulate steady-state conditions. The set of coupled linear first-order homogenous differential equations for the two-pool case without algebraic simplifications was numerically solved using a Nelder-Mead algorithm. 38 Since the relaxation times of the metabolites are inconsequential to the fitting routine, these values were fixed to T 1m = 1000 ms and T 2m = 100 ms. 36,39 For the amino acids, the chemical shift difference between the amine protons and water resonance was assumed to be 2.8 ppm for Ala, Asp, Gln and Tau, 10,11,40 3.0 ppm for Glu 4,41 and 2.75 ppm for GABA. 12 Despite small shifts in peak position due to changes in pH and temperature, all chemical shifts reported here were assumed to be constant to maintain consistency. 4 All phantom experiments were repeated five times.

| Data modelling
The pH dependence of the proton exchange rate in aqueous solutions can be defined as 14,42,43 where k a , k b and k buffer denote the exchange rate of acid-catalyzed exchange, base-catalyzed exchange and other contributions, e.g. from the buffer, respectively.
The Arrhenius equation describes the temperature dependence of exchange rates 44 : with A being a characteristic constant for the chemical process, E A the activation energy, R = 8.314 J (mol K) À1 the general gas constant and T the absolute temperature.
Assuming that the amine proton exchanges of amino acids are predominantly base catalyzed and buffer catalyzed, it can be assumed for each temperature that 14,45 The temperature dependence of the pK of water (pK w (T)) is given by the solution of the Van't Hoff equation 46 : with the standard reaction enthalpy for the self-dissociation of water ΔH 0 R = 55.84 kJ mol À1 and a pK w = 14 at T 0 = 25 C = 298.15 K. 14 Thus k sw (pH, T) is given by For the modeling of k sw (pH, T), only exchange rates below 22 000 s À1 and with a standard deviation of less than 10% were included, to ensure the validity of the data and the modeling.

| Simulations
Simulations were performed by numerically solving the Bloch-McConnell equations using the two-pool model. Additionally, for a multi-pool simulation, the model was extended according to Sun. 37 We solved the Bloch equations and calculated the spectra by using the 'ode' (ordinary differential equation) solver in Scilab.
The CEST asymmetry was determined as where M sat refers to the magnetization with saturation at a positive (Δω) or a negative (ÀΔω) offset from the water resonance. 13 The metabolites that were used for the multi-pool simulation of the human brain were Ala (0. For all metabolites, the experimentally determined k sw values show an exponential behavior as a function of pH and temperature. Using Equations 1-5 base-and buffer-catalyzed exchange rate constants and activation energies were obtained for the different amino acids, as shown in Table 1. Also, the pK values of the amine protons of the amino acids were included in Table 1 as a basis for the following discussion. 47 Note that the effective base-catalyzed rate constant for Tau is a factor of 10 higher compared with Ala, GABA, Asp and Glu, while their effective activation energy is the lowest. 3.2 | Dependence of the CEST effect on B 0 , B 1 amplitude and t sat T A B L E 2 Experimental parameters yielding the largest CEST effects for Ala, Asp, GABA and Glu at different field strengths at 37 C (pH 7.0, 10 mM), based on the measurements described above F I G U R E 2 Simulated CEST effects as a function of B 1 and t sat for Glu at 37 C (A) and 1.5 C (B) and for Tau at 37 C (C) and 1.5 C (D) (10 mM, pH 7, B 0 = 7 T) ranges for CEST imaging. On average, for a maximum CEST effect at a body temperature of 37 C, a higher B 0 requires a higher B 1 and a longer t sat . Table 3 summarizes the results for the experimental parameters that achieve the maximum CEST effect in the brain of the polar cod at 1.5 C. In contrast to an application at 37 C, on average, a lower B 1 with higher B 0 but longer t sat is required to achieve the maximum CEST effect at 1.5 C. Note that compared with the other amino acids significantly higher B 1 and shorter t sat are required for Gln and Tau. Figure 2 presents the different CEST effects and their optimal experimental parameters B 1 and t sat at 37 C and at 1.5 C for the amino acids Glu (A, B) and Tau (C, D). While for Glu at 37 C a high B 1 amplitude and relatively short t sat would be optimal (Figure 2A), the maximum GluCEST effect at 1.5 C can be achieved with a much lower B 1 amplitude and a longer t sat ( Figure 2B). For Tau, however, no significant CEST effect can be detected at 37 C and pH 7 ( Figure 2C), but this changes when the temperature becomes 1.5 C, whereby a significant effect can be seen with a high B 1 amplitude and a short t sat ( Figure 2D). The in vivo expected CEST effects for both organisms require a relatively large B 1 amplitude and a short t sat for maximum signal intensity. For the simulation of the specificity of the GluCEST effect for the human brain, the contour plots clearly show that independent of the B 1 amplitude and t sat the proportion of Glu in the total CEST effect is always between 68% and 71%. In contrast, the specificity of TauCEST in the fish model shows a strong dependence on the B 1 amplitude. However, the average percentage of Tau is above 50%, and it can be up to 78%.  Figure 4C and 4D shows the specificity for F I G U R E 3 Simulated in vivo expected CEST effects and the contributions for the dominant metabolite as a function of B 1 and t sat : total CEST effect for the human brain (37 C) at 3 ppm (A) and the contribution of Glu (C) (7 T), and total CEST effect for the polar cod (1.5 C) at 2.8 ppm (B) and the contribution of Tau (D) (9.4 T). The contour plots overlaid on the maps represent the regions with steps of 0.5% for the plots of the total CEST effects and 2% for the maps of the corresponding contributions of Glu and Tau Glu (human brain) and Tau (polar cod brain). For the two simulations, an increase of the total CEST effect with decreasing pH value occurs. While in the human brain a rise in temperature results in a lower CEST effect, the same is true for the effect in the brain of the polar cod for basic pH values, but it turns into an increase in the effect for the more acidic pH range. The specificity for GluCEST in the human brain shows almost no change depending on pH and temperature and is approximately constant at 70%. The dependence of the specificity of the TauCEST effect in the brain of polar cod is different, with a strong increase in the acidic pH range and for lower temperature.

| Intensity and specificity of the CEST effects of amino acids
The contributions of the different amino acids to the in vivo expected CEST effect in the human brain and the brain of polar cod as a function of the saturation frequency are shown in Figure 5. The main percentage contribution of Glu to the total CEST effect in the human brain is almost independent of the saturation frequency, while the remaining percentage is mainly due to Asp and GABA (see Figure 5A). Although for all saturation frequencies the main contribution to the CEST effect in the brain of polar cod is due to Tau, it is noteworthy that its percentage contribution is lowest at the maximum intensity of the total CEST effect (see Figure 5B). The remaining contributions are mainly due to Glu and Gln.

| DISCUSSION
This work aimed to characterize the amine proton exchange of important amino acids. Therefore, the exchange rates of six amino acids were determined experimentally as a function of pH and temperature and the influence of experimental parameters was simulated. The intensity and specificity of the CEST effect of amino acids at about 2.8 ppm as expected in the human brain were analyzed by simulations. Furthermore, this was also done for an ectothermic animal model, the polar cod, to demonstrate the strong temperature dependence and the broad applicability of the CEST effect. only Gln has another group of labile bound protons at 2.2 ppm, which is characterized by a very slow exchange rate of less than 50 s À1 , that, following the Arrhenius equation (see Equation (2)), slows down further with decreasing temperatures; therefore, at a B 1 amplitude of 5.87 μT such slow exchanging protons can be neglected.
Due to the dominant base catalysis (see Equation (3)) and the Arrhenius equation (see Equation (2)), the exchange rates experimentally determined for the different metabolites show the expected mono-exponential course as a function of pH and temperature. Lee et al. assigned the resonances of exchangeable protons of brain metabolites and the positions and full widths at half maximum of the signals using 1 H NMR spectra measured on a high-resolution 11.7 T spectrometer. 10 The exchange rates determined on this basis are in some cases lower than the values found in the present study. One reason for this is that the method used by Lee et al. to determine exchange rates using the half-width is limited in its validity to slow exchange regimes. 51 Furthermore, the solutions were prepared without the use of a corresponding phosphate buffer. However, amine and hydroxyl protons of amino acids show a strong dependence on buffer properties, as they occur in the natural environment. 54 The osmolarity and concentration of phosphate, therefore, has a catalytic effect for the exchange of the amine protons and must be included in the considerations for an analysis of the in vivo CEST effect expected at about 2.8 ppm. Khlebnikov et al. measured the k sw for the most important metabolites for the CEST effect in the human brain in the physiological pH range at 37 C. 11 Although the exchange rates in both these and the present work were determined by fitting z-spectra with the Bloch-McConnell equation, the exchange rates determined by Khlebnikov et al. are slightly but systematically lower. This difference may be attributed to the 2 mM higher concentrated PBS solution used in the present study and its influence on the exchange rates as described above.
The acid constants (pK value) of the amino groups of the metabolites investigated vary within a range of 9.0-10.6 units (see Table 1).
According to the Brønsted law of acid/base catalysis, the logarithm of k sw is proportional to the pK value of the amine protons, 55,56 which is reflected in the data obtained. A comparison of the pK values and the determined exchange rates for 25 C and pH 7.0 shows a similar order, ie a lower pK value corresponds to a faster k sw . In addition, the exchange rates of Glu and Asp show a comparable behavior due to similar pK values. 10 Furthermore, the remarkably fast exchange rates of Gln and Tau match their low pK values.
A determination of k sw (pH, T) supports both the analysis of CEST effects and the optimization of the pulse sequences used. 14 For this reason, the experimentally determined exchange rates were interpolated based on the Equations 1-5 and the base-and buffer-catalyzed exchange rates and activation energies for the different metabolites were determined (see Figure 1 and Table 1).
The determined effective base-catalyzed exchange rates (k b,eff ) for amine protons (T = 25 C) are higher by a power of 10 than those for the base-catalyzed exchange of amide protons of PLL k b (37 C) = 1. 92 Â 10 9 s À1 L mol À151 and of guanidinium protons of creatine k b,eff (25 C) = (3.01 ± 0.16) Â 10 9 s À1 L mol À1 . 14 The base-catalyzed exchange rates for the amine protons of GABA, Gln, Glu and Tau in the work of Khlebnikov et al. are in the same order of magnitude as those determined in the present work. 11 Differences can be explained by the slight deviations in the determination of the exchange rates described above and by the different temperatures on which the analysis is based. The base-catalyzed exchange constants are an indicator of the sensitivity of the exchange rates as a function of the pH value. Thus, pH changes have a much greater influence on the exchange of Tau and Gln than that, for example, of Asp.
The activation energies of the base-catalyzed exchange, 20.32 kJ mol À1 on average, are lower than those determined for amide protons of peptide groups (71.12 kJ mol À1 ) 44 and the guanidinium protons of creatine (32.27 kJ mol À1 ). 14 This aspect fits the high exchange rates of the amine protons compared to the amide and guanidinium protons, since the lower the activation energy the faster the reaction proceeds. Liepinsh and Otting were able to show that amino acids whose amino groups have a low pK value are more sensitive to buffer properties than those with higher pK values. 54 This is reflected in the data obtained for buffer-catalyzed exchange rates (see Table 1).

|
Dependence of the CEST effect on B 0 , B 1 amplitude and t sat On the basis of the experimentally determined model functions k sw (pH, T) for the different amino acids, the influence of the experimental parameters B 0 , B 1 and t sat was investigated at two different temperatures (10 mM, 1.5 C and 37 C) (see Tables 2,3 and Supplementary Material). These findings can be used as a starting point for the optimization of existing pulse sequences, but also as a basis for the development of new CEST sequences and processing models.
As reflected in our results the intensity of the CEST effect increases with higher B 0 (see Supplementary Material), since Δω increases proportionally to the magnetic field strength and thus the influence of direct saturation of water is reduced. As a result, even metabolites with faster exchange rates meet the requirements for the CEST effect, i.e. a slow to medium exchange regime k sw ≤ Δω. Also, the T 1 relaxation time increases with a higher B 0 field, which delays the time until thermodynamic equilibrium is reached and can lead to a higher CEST effect. Additionally, the CEST effect increases with B 1 but decreases again when direct saturation becomes important (see also Figures 2 and 3). 57 When a very high B 1 amplitude is selected, the CEST effect is partially or even completely cancelled out (see Figures 2 and 3). 35,57 The influence of the B 1 amplitude on the CEST effect can be explained most simply in terms of the saturation efficiency α. This parameter, introduced by Sun et al., is a measure of the intensity of the maximum CEST effect and depends on both k sw and the applied B 1 amplitude. 58 In general, a faster k sw requires a higher B 1 amplitude to achieve the same CEST effect and similar saturation efficiency (see Figure 2). The CEST effect builds up over a defined t sat , whereby after a defined saturation period of about 5 T 1w the thermodynamic equilibrium state is reached. However, the influence of the direct saturation shows a biexponential behavior as a function of t sat , and since interferences with direct water saturation increase with t sat , an optimal t sat can exist for faster-exchanging systems where the spillover is not yet fully effective (see Figure 2). 57 The temperature dependence of exchange rates is characterized by the Arrhenius equation. 44 Accordingly, even small temperature changes can have a large effect on the exchange rate, given the exponential relationship. As a consequence, metabolites with a fast exchange at high temperatures can be assigned to the slow to medium exchange regime at lower temperatures and become of interest for CEST imaging (e.g. Tau).
Due to the relationships between the experimental parameters and the exchange rate discussed in the previous section, a temperature change thus also results in different optimal experimental parameters for CEST imaging (see Figure 2). This aspect is still strongly influenced by different composition of metabolite concentrations in different organisms or at different (patho-) physiological conditions (see Figure 3).
At present, CEST measured in vivo is often described as a metabolite weighted contrast emphasizing the main contribution of one metabolite. 11 However, it is important to consider the contributions of other compounds and the dependence of the specificity on temperature and pH to avoid physiological misinterpretations. In a recent study by Chen et al., the authors conclude that tCr accounts for 80% of the guanidinium peak of the CEST effect in the brain of a mouse at optimal experimental parameters. Cai et al. showed that the CEST effect observed in vivo in the human brain is mainly determined by Glu, with a contribution of 70-75% (3 ppm, 37 C, 7 T). 4 This result is well reflected in our simulations, where Glu dominates the in vivo expected CEST effect in the human brain independent of the experimental parameters, with a share of 68-71%. The situation is different for the CEST effect expected in vivo for a model organism such as the polar cod, where the specificity of the Tau dominated in vivo CEST effect increases with increasing B 1 amplitude. An optimal choice of experimental parameters and a consistent analysis of the CEST effect as a multi-pool model together with adequate knowledge of the metabolite composition is therefore essential.

| Intensity and specificity of the CEST effects of amino acids
The choice of optimal experimental parameters is a trade-off between the intensity of the CEST effect and its specificity (see Figure 3). Additionally, the choice of saturation parameters is constrained by the specific absorption rate (SAR). As SAR increases with B 1 and B 0 , the allowed B 1 amplitude at 7 T is limited for human application. 4 For this reason, a B 1 amplitude of 3.6 μT, as already used by Cai et al., was chosen for the simulations in the present study. 4 These guidelines do not apply to in vivo applications on animal models such as polar cod. However, any uncontrolled body temperature due e.g. to heating of the surrounding seawater during the measurement should be avoided at all costs, so in the present work a B 1 of 4.4 μT at 9.4 T was used. This is consistent with previous in vivo applications, where no warming could be detected during the temperature-controlled measurements. 22 However, an optimization of the saturation scheme and RF pulse shapes can lead to an increase in the intensity of the CEST effect.
The simulations for the CEST effect for the human brain and the brain of the polar cod show considerable differences due to different metabolite concentrations and especially due to the different temperatures and thus different k sw values (see Figure 4). The specificity of the metabolite dominating the CEST effect in the brain of polar cod at 1.5 C is strongly dependent on pH and T. For high temperature and an acidic pH the k sw of Glu is in the optimal slow-to intermediate-exchange regime (k sw ≤ Δω), whereas this optimal regime is reached for Tau at low temperature and basic pH.
Considerations of the in vivo expected CEST effect of amino acids for the human brain for various typical pathological conditions show that, e.g., for acute cerebral ischemia with acidification of pH by 0.3 pH units, an increase in the CEST effect can be expected, but the specificity of Glu remains at 70%. For the clinical picture of a glioblastoma, an alkalinization of 0.5 pH units can be expected, but even in this particular case a slight decrease in specificity to 63% can be assumed.
The specificity of Tau behaves differently in the example of the brain of a polar cod. Polar cod can be easily exposed to temperature changes of 4 C to À2 C within a very short time due to, e.g., vertical movements in the water column; simulating this in an in vivo application would result in a change of the specificity of Tau of almost 20%, which in turn would result in a specificity below 50% for Tau. In contrast, acidification of pH i due to acute incubation of polar cod with CO 2 to simulate ocean acidification scenarios (À0.2 pH units) would result in a 10% increase.
In a previous in vivo study on polar cod (see Supplementary Material), we could show that a simulated ocean acidification scenario (i.e. CO 2enriched sea water (4900 μatm)) resulted in an increase by 3.2% of the CEST effect in brain of polar cod. 22 According to our simulations, this corresponds to a decrease in pH i by 0.4 pH units, as expected from the effect of elevated CO 2 concentrations in the blood.
These examples show that careful considerations of the specificity of the CEST effect are needed to minimize the risk of misinterpretations for different pH and temperature dependences of metabolites involved in the CEST effect. One approach can be the shift of the saturation frequency at which the analysis is performed ( Figure 5). In contrast to the Glu weighted CEST effect in the human brain, whose specificity is nearly constant over the saturation frequencies studied, the specificity of Tau in the expected CEST effect in the brain of polar cod shows a strong dependence on the saturation frequency. An evaluation of the CEST effect at about 2 ppm would thus result in an approximately 20% increase in specificity. This is facilitated by a shift of the asymmetry curve for Tau towards lower ppm values due to intermediate-to fast-exchange chemical shift averaging. 4 The rather small intensity loss of the CEST effect at this offset frequency can be neglected or compensated by extending the measurement time, i.e. averaging. The gain in specificity from the significantly reduced influence of other compounds on the total CEST effect, however, is much more desirable to avoid physiological misinterpretation.
The in vivo expected CEST effects presented in this study focus on the CEST effect of highly concentrated amino acids with exchangeable protons at about 2.8 ppm downfield from water and its behavior in terms of intensity and specificity for different experimental and physiological parameters. We have shown this for two very different organisms to demonstrate the wide range of CEST applications. However, it should be emphasized that other effects have been neglected in this study, such as MTC and upfield nuclear Overhauser enhancement, as there are several approaches to remove these effects from the in vivo CEST data, as mentioned in the introduction. Furthermore, the APT effect of amide protons of mobile proteins and peptides was not included in the modeling of this study. Amide protons are by far the most highly concentrated group of exchanging protons of mobile proteins, even if other groups occur at much lower concentrations. Furthermore, protein conformational changes, which can occur in vivo, further complicate their inclusion in simulations. 11

| CONCLUSION
In the present work, the exchange rates k sw (pH, T) have been determined for six amino acids that are important for brain metabolism. These functions can be used for the simulation of multi-pool models of a wide variety of organisms and their physiological properties. Based on these multi-pool models, it is possible to optimize saturation parameters and improve the analysis of CEST effects of amino acids measured in vivo.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Data will be made available in a public repository after approval.