Revealing a Third Dissolved-Phase Xenon-129 Resonance in Blood Caused by Hemoglobin Glycation

Hyperpolarized (HP) xenon-129 (129Xe), when dissolved in blood, has two NMR resonances: one in red blood cells (RBC) and one in plasma. The impact of numerous blood components on these resonances, however, has not yet been investigated. This study evaluates the effects of elevated glucose levels on the chemical shift (CS) and T2* relaxation times of HP 129Xe dissolved in sterile citrated sheep blood for the first time. HP 129Xe was mixed with sheep blood samples premixed with a stock glucose solution using a liquid–gas exchange module. Magnetic resonance spectroscopy was performed on a 3T clinical MRI scanner using a custom-built quadrature dual-tuned 129Xe/1H coil. We observed an additional resonance for the RBCs (129Xe-RBC1) for the increased glucose levels. The CS of 129Xe-RBC1 and 129Xe-plasma peaks did not change with glucose levels, while the CS of 129Xe-RBC2 (original RBC resonance) increased linearly at a rate of 0.015 ± 0.002 ppm/mM with glucose level. 129Xe-RBC1 T2* values increased nonlinearly from 1.58 ± 0.24 ms to 2.67 ± 0.40 ms. As a result of the increased glucose levels in blood samples, the novel additional HP 129Xe dissolved phase resonance was observed in blood and attributed to the 129Xe bound to glycated hemoglobin (HbA1c).

A high glucose concentration in blood leads to excessive non-enzymatic chemical interactions between glucose and proteins in the blood [27]. Hemoglobin is the protein most affected by glycation. Structurally, hemoglobin has four subunits, consisting of two α and two β chains, each of which carry an iron-containing heme group responsible for oxygen binding [28]. During the glycation process, a non-enzymatic reaction occurs between glucose and the α-amino groups of valine residues at the N-terminus of the β-chains [29]. The most abundant form of glycated hemoglobin, which is widely used in clinical practice for the evaluation of glycemia in diabetes mellitus, is hemoglobin A1c (HbA1c) [30]. Glycated hemoglobin is naturally present in healthy individuals (~4%), whereas glycation levels up to 20% have been reported in patients diagnosed with diabetes [31]. Elevated HbA1c levels (>6.5%) are routinely used for the diagnosis of diabetes mellitus [32]. The standard HbA1c test is used for the diagnosis of type 2 diabetes and prediabetes. It shows the amount of glycated hemoglobin and reflects the average blood glucose level over the past three months [33].
Hemoglobin glycation induces the formation of oxygen-derived free radicals, which are responsible for causing oxidative stress in erythrocytes [34], leading to an increase in membrane lipid peroxidation [35] and membrane damage in diverse cell types. The glycation of iron-containing heme proteins can cause the degradation of heme and further reactions with H 2 O 2 , leading to increased iron release, as well as ferryl myoglobin formation [27,36]. In addition, iron overload caused by high concentrations of glycated hemoglobin has been shown to be associated with changes in the structure of the red blood cells (RBCs) and increased thrombotic events [37]. Numerous studies have been conducted evaluating the effect of elevated levels of glycated hemoglobin on red blood cell distribution width (RDW); however, the data published has been inconclusive [38]. Some studies have also suggested that hyperglycemia may have some effect on erythropoiesis and RBC survival at concentrations of glucose higher than 40 mM [35,38].
It is known that the HP 129 Xe signal from blood originates from Xe in plasma and the RBCs [1]. 129 Xe binds to hemoglobin in hydrophobic cavities close to the external surface in both α and β chains [39]. Therefore, alternations in the structure of hemoglobin caused by glycation are expected to have a direct effect on the HP 129 Xe magnetic environment and its dipole-dipole interaction with hemoglobin, potentially resulting in CS changes. Furthermore, the overall increase in free radical levels can also be anticipated to cause an effect on HP 129 Xe CS. The destruction of heme groups accompanied by the release of iron would be expected to have potential effects on HP 129 Xe-RBC CS, as well as on the T * 2 relaxation process.
The present lung studies with HP 129 Xe exploit the measuring of the signal intensity of 129 Xe bound to tissue and RBC [2][3][4]7,8]. However, the elevated glucose level in blood may have a considerable impact on the 129 Xe spectroscopic properties and should be taken into account for patients with diabetes. This work, for the first time, investigates the effect of glucose concentration in blood on the physical properties of dissolved HP 129 Xe, such as CS and on the effective spin-spin relaxation time T * 2 . Moreover, using non-linear curve fitting, we demonstrated that the spectrum of HP 129 Xe in blood contains not two (as was conventionally thought) but three dissolved-phase resonances. We report the resonance frequencies, T * 2 relaxation times, and the CS dependance on glucose level of the three observed spectral components in sterile citrated sheep blood. In addition, we propose a potential mechanism for glycation-related changes in the physical parameters of the dissolved HP 129 Xe. This research suggests that for a comprehensive understanding of dissolved-phase imaging, glucose levels in the blood should also be taken into account along with blood oxygenation, since both play distinct roles.

CS Analysis Using a Conventional Three-Peak Model (3PM)
The measurements of the CS of the 129 Xe dissolved in plasma ( 129 Xe-plasma) and 129 Xe dissolved in RBC ( 129 Xe-RBC) resonances were conducted for the glucose concentration range of 0-55 mM. The acquired MRS spectra were first fitted using the conventional three-peak model (3PM): gas peak, plasma peak, and RBC peak. Figure 1A,B show MRS acquired for 10 mM and 45 mM glucose concentrations in the blood, respectively. The increase in glucose concentration did not cause a significant change in the 129 Xe-plasma CS. Conversely, the 129 Xe-RBC resonance shifted from 217.86 ppm (10 mM of glucose) to 219.07 ppm (45 mM). Figure 1C shows three representative cumulative Lorentzian 3PM fits for 10 mM, 20 mM, and 45 mM glucose concentrations. The downfield shift of the HP 129 Xe-RBC resonance with respect to the gaseous 129 Xe resonance can be clearly observed with increasing glucose concentration. The observed CS increased linearly with the increased glucose concentration ( Figure 1D). The fit of the CS dependence of the 129 Xe-RBC peak on glucose concentration revealed a (0.025 ± 0.004) ppm/mM 129 Xe-RBC change rate. A strong positive correlation (r = 0.91) was observed between the 129 Xe-RBC peak CS and the glucose concentration in the measured blood samples. It can be clearly seen that the 129 Xe-plasma peak was completely unaffected by an increase in glucose concentration (r = 0.1 with a slope of the line equal to (0.001 ± 0.005) ppm/mM. The observed downfield shift of the HP 129 Xe-RBC resonance is similar to the HP 129 Xe-RBC resonance frequency change due to the increase in blood oxygenation, albeit with a total span of only~1.25 ppm.

CS Analysis Using a Conventional Three-Peak Model (3PM)
The measurements of the CS of the 129 Xe dissolved in plasma ( 129 Xe-plasma) an dissolved in RBC ( 129 Xe-RBC) resonances were conducted for the glucose concen range of 0-55 mM. The acquired MRS spectra were first fitted using the conven three-peak model (3PM): gas peak, plasma peak, and RBC peak. Figure 1A,B show acquired for 10 mM and 45 mM glucose concentrations in the blood, respectivel increase in glucose concentration did not cause a significant change in the 129 Xe-p CS. Conversely, the 129 Xe-RBC resonance shifted from 217.86 ppm (10 mM of gluc 219.07 ppm (45 mM). Figure 1C shows three representative cumulative Lorentzia fits for 10 mM, 20 mM, and 45 mM glucose concentrations. The downfield shift of 129 Xe-RBC resonance with respect to the gaseous 129 Xe resonance can be clearly ob with increasing glucose concentration. The observed CS increased linearly with creased glucose concentration ( Figure 1D). The fit of the CS dependence of the 129 X peak on glucose concentration revealed a (0.025 ± 0.004) ppm/mM 129 Xe-RBC chang A strong positive correlation (r = 0.91) was observed between the 129 Xe-RBC peak C the glucose concentration in the measured blood samples. It can be clearly seen th 129 Xe-plasma peak was completely unaffected by an increase in glucose concentrati 0.1 with a slope of the line equal to (0.001 ± 0.005) ppm/mM. The observed downfie of the HP 129 Xe-RBC resonance is similar to the HP 129 Xe-RBC resonance frequency c due to the increase in blood oxygenation, albeit with a total span of only ~1.25 ppm

T * 2 Relaxation Measurements
Following the measurements of CS, T * 2 relaxation was assessed based on Lorentzian fits of the measured spectra to 3PM. No significant change was observed in T * 2 values for HP 129 Xe dissolved in both RBC and plasma pools with an increase in glucose content ( Figure 2). The mean T * 2 values were equal to (1.48 ± 0.09) ms and (0.87 ± 0.07) ms for HP 129 Xe dissolved in plasma and RBC, respectively.

* Relaxation Measurements
Following the measurements of CS, T * relaxation was assessed based on Lorentzian fits of the measured spectra to 3PM. No significant change was observed in T * values for HP 129 Xe dissolved in both RBC and plasma pools with an increase in glucose content ( Figure 2). The mean T * values were equal to (1.48 ± 0.09) ms and (0.87 ± 0.07) ms for HP 129 Xe dissolved in plasma and RBC, respectively.

Four-Peak Spectroscopic Model (4PM) Analysis
With the blood glucose level increase, the RBC resonance in the 129 Xe spectrum became more asymmetrical and its linewidth increased slightly ( Figure 3). In addition, a small splitting of the RBC peak was noted for higher glucose concentrations (red arrow on Figure 3). Considering the glycation process in RBCs, the observed asymmetry and splitting in the RBC peak can be interpreted by a four-peak spectroscopic model (4PM). In general, structural and functional changes to hemoglobin occur as a result of glycation [27]. Therefore, we can subdivide the RBC pool into HP 129 Xe bound to non-glycated HbA0 and HP 129 Xe bound to glycated HbA1c. These two pools can be characterized by their own HP 129 Xe resonances. The suggested 4PM includes one gas phase resonance and three dissolved-phase resonances: 129 Xe-plasma, 129 Xe-RBC1, and 129 Xe-RBC2.

Four-Peak Spectroscopic Model (4PM) Analysis
With the blood glucose level increase, the RBC resonance in the 129 Xe spectrum became more asymmetrical and its linewidth increased slightly ( Figure 3). In addition, a small splitting of the RBC peak was noted for higher glucose concentrations (red arrow on Figure 3). Considering the glycation process in RBCs, the observed asymmetry and splitting in the RBC peak can be interpreted by a four-peak spectroscopic model (4PM). In general, structural and functional changes to hemoglobin occur as a result of glycation [27]. Therefore, we can subdivide the RBC pool into HP 129 Xe bound to non-glycated HbA 0 and HP 129 Xe bound to glycated HbA 1c . These two pools can be characterized by their own HP 129 Xe resonances. The suggested 4PM includes one gas phase resonance and three dissolved-phase resonances: 129 Xe-plasma, 129 Xe-RBC1, and 129 Xe-RBC2.
The suggested 4PM was used to fit the experimental data ( Figure 4A,B). The application of the 4PM did not affect the plasma peak position. The position of the HP 129 Xe-RBC1 peak was barely affected by glucose: 216.08 ppm for the 10 mM solution and 215.92 ppm for the 45 mM solution. The 129 Xe-RBC2 peak, however, shifted downfield with increasing glucose concentrations: 220.32 ppm for the 10 mM solution and 220.78 ppm for the 45 mM. The residual error for the 3PM, once plotted, showed distinct peaks at 216, 218, and 222 ppm for the 10 mM sample and at 216 ppm and 219 ppm for the 45 mM sample ( Figure 4C,D). Once the spectrum is fitted to the 4PM, however, the residual error becomes flat at the RBC resonance position (between 210 ppm and 240 ppm). This indicates that 4PM fits the acquired HP 129 Xe blood spectra more accurately compared to the conventional 3PM. The suggested 4PM was used to fit the experimental data ( Figure 4A,B). The application of the 4PM did not affect the plasma peak position. The position of the HP 129 Xe-RBC1 peak was barely affected by glucose: 216.08 ppm for the 10 mM solution and 215.92 ppm for the 45 mM solution. The 129 Xe-RBC2 peak, however, shifted downfield with increasing glucose concentrations: 220.32 ppm for the 10 mM solution and 220.78 ppm for the 45 mM. The residual error for the 3PM, once plotted, showed distinct peaks at 216, 218, and 222 ppm for the 10 mM sample and at 216 ppm and 219 ppm for the 45 mM sample ( Figure  4C,D). Once the spectrum is fitted to the 4PM, however, the residual error becomes flat at the RBC resonance position (between 210 ppm and 240 ppm). This indicates that 4PM fits the acquired HP 129 Xe blood spectra more accurately compared to the conventional 3PM.   . HP 129 Xe MRS spectra acquired for a pure blood sample (black line) and for a 25 mM blood glucose concentration (purple line). The RBC peak became asymmetrical and broader with the addition of glucose. A small splitting of the RBC peak can be seen (red arrow).
The suggested 4PM was used to fit the experimental data ( Figure 4A,B). The application of the 4PM did not affect the plasma peak position. The position of the HP 129 Xe-RBC1 peak was barely affected by glucose: 216.08 ppm for the 10 mM solution and 215.92 ppm for the 45 mM solution. The 129 Xe-RBC2 peak, however, shifted downfield with increasing glucose concentrations: 220.32 ppm for the 10 mM solution and 220.78 ppm for the 45 mM. The residual error for the 3PM, once plotted, showed distinct peaks at 216, 218, and 222 ppm for the 10 mM sample and at 216 ppm and 219 ppm for the 45 mM sample ( Figure  4C,D). Once the spectrum is fitted to the 4PM, however, the residual error becomes flat at the RBC resonance position (between 210 ppm and 240 ppm). This indicates that 4PM fits the acquired HP 129 Xe blood spectra more accurately compared to the conventional 3PM.  . It can be clearly seen that the conventional 3PM shows some residual signal present (green arrows). The proposed 4PM, however, results in a flat baseline with almost no residual signal at the RBC resonance position (between 210 ppm and 240 ppm) indicating a better fit of the RBC signal. Therefore, the proposed 4PM better suits the acquired HP 129 Xe MRS spectra analysis compared to the conventional 3PM.
It should be noted that the 4PM fit worked better for glucose concentrations above 5 mM. This was indicated by slightly higher R 2 values. In addition, small alterations in the peak position for the RBC resonances did not affect R 2 significantly. Therefore, a lower accuracy of the recalculated spectral parameters can be anticipated for the pure blood and Int. J. Mol. Sci. 2023, 24, 11311 6 of 13 5 mM samples compared to the samples with higher glucose levels. This is in accordance with our hypothesis that the second HP 129 Xe RBC peak originates from HbA1c. There is no naturally occurring HbA1c in the pure blood and the concentration is expected to be low in a 5 mM glucose sample.
The proposed 4PM did not affect the CS of the HP 129 Xe gas peak and 129 Xe-plasma peak. The HP 129 Xe-RBC1 CS change was not observed with a glucose level increase ( Figure 5A). The HP 129 Xe-RBC2 resonance frequency, however, increased with a rate of (0.015 ± 0.002) ppm/mM. A strong Pearson correlation was observed between the HP 129 Xe-RBC2 peak position and the blood glucose level (r = 0.95). The T * 2 relaxation time for 129 Xe-plasma did not depend on the glucose concentration in the sample ( Figure 5B). Conversely, the 129 Xe-RBC1 T * 2 time increased non-linearly from (1.58 ± 0.24) ms up to (2.67 ± 0.40) ms over the studied range of glucose concentrations. The 129 Xe-RBC2 T * 2 relaxation time increased from (0.66 ± 0.10) ms to (1.23 ± 0.19) ms over a 0-10 mM glucose concentration range and leveled out at approximately (0.91 ± 0.03) ms at higher glucose levels.
the acquired HP 129 Xe MRS spectra analysis compared to the conventional 3PM.
It should be noted that the 4PM fit worked better for glucose concentrations above 5 mM. This was indicated by slightly higher R 2 values. In addition, small alterations in the peak position for the RBC resonances did not affect R 2 significantly. Therefore, a lower accuracy of the recalculated spectral parameters can be anticipated for the pure blood and 5 mM samples compared to the samples with higher glucose levels. This is in accordance with our hypothesis that the second HP 129 Xe RBC peak originates from HbA1c. There is no naturally occurring HbA1c in the pure blood and the concentration is expected to be low in a 5 mM glucose sample.
The proposed 4PM did not affect the CS of the HP 129 Xe gas peak and 129 Xe-plasma peak. The HP 129 Xe-RBC1 CS change was not observed with a glucose level increase (Figure 5A). The HP 129 Xe-RBC2 resonance frequency, however, increased with a rate of (0.015 ± 0.002) ppm/mM. A strong Pearson correlation was observed between the HP 129 Xe-RBC2 peak position and the blood glucose level (r = 0.95). The T * relaxation time for 129 Xeplasma did not depend on the glucose concentration in the sample ( Figure 5B). Conversely, the 129 Xe-RBC1 T * time increased non-linearly from (1.58 ± 0.24) ms up to (2.67 ± 0.40) ms over the studied range of glucose concentrations. The 129 Xe-RBC2 T * relaxation time increased from (0.66 ± 0.10) ms to (1.23 ± 0.19) ms over a 0-10 mM glucose concentration range and leveled out at approximately (0.91 ± 0.03) ms at higher glucose levels.

Discussion
In normal physiological conditions, 95-98% of the hemoglobin in RBCs is present in the HbA0 state [31]. Once glucose levels elevate in the blood, HbA1c is generated as part of the glycation process [40]. In fact, the concentration of HbA1c hemoglobin increases linearly with the glucose concentration [30]. Watala et al. demonstrated that there are three major sites where structural changes occur in HbA1c: at the amino N-terminus of the β chain (~1/3 of the total amount of glucose binds to this site), α amino sites (~1/3 of the total

Discussion
In normal physiological conditions, 95-98% of the hemoglobin in RBCs is present in the HbA 0 state [31]. Once glucose levels elevate in the blood, HbA 1c is generated as part of the glycation process [40]. In fact, the concentration of HbA 1c hemoglobin increases linearly with the glucose concentration [30]. Watala et al. demonstrated that there are three major sites where structural changes occur in HbA 1c : at the amino N-terminus of the β chain (~1/3 of the total amount of glucose binds to this site), α amino sites (~1/3 of the total glycation), and lysine residues (~40% of all glycation) [41]. Even if glucose is mostly linked to N-terminal valine, the glycation affects the spatial structure of the whole hemoglobin molecule. Ye at al. demonstrated that glycation transforms alphahelices into beta-sheets, which results in conformation changes or the unfolding or even aggregation of hemoglobin [42]. Moreover, Sen at al. observed that the conformation caused by glycation increased the exposure of the hydrophobic tryptophan residues, like Trp14 in alpha-chains, which created one of the xenon-binding sites [27]. These structural changes result in significant alterations in hemoglobin conformation [27,41]. Glycationinduced conformational changes in hemoglobin weaken the heme-globing linkage in HbA 1c and make it more thermolabile compared to HbA 0 [27]. Thus, the global changes in spatial hemoglobin's structure (tertiary and quaternary structure) may affect the shape of hydrophobic cavities or the binding sites of xenon, potentially resulting in a different chemical shift.
Savino et al. identified a total of twelve 129 Xe binding sites per Hb 4 tetramer: the α 1 chain contains four binding sites; the α 2 chain contains three; three more are located in the β 2 chain; and two can be found in the β 1 chain [43]. Considering the glycation-induced structural changes in HbA 1c , it is possible and fairly likely that the number, size, and/or spatial location of 129 Xe binding sites are different in HbA 1c compared with HbA 0 . In addition, the glycation of hemoglobin results in the accumulation of free OH· radicals due to an iron-mediated Fenton's reaction [27]. Increasing OH· concentrations are anticipated to result in the deshielding of the HP 129 Xe nuclei. This hypothesis is supported by our experimental data, indicating the linear shift of the RBC HP 129 Xe resonance downfield with respect to the resonance frequency of the gaseous HP 129 Xe due to the increase in glucose concentration. The linear change in the HP 129 Xe CS as a function of glucose level is in agreement with a known linear increase in HbA 1c concentration caused by an increase in glucose concentration.
Considering our hypothesis that HP 129 Xe binding sites are affected by glycation and are different in HbA 1c compared to HbA 0 , it is reasonable to assume the existence of two HP 129 Xe resonances-one from 129 Xe bound to HbA 0 and another originating from 129 Xe bound to HbA 1c . Therefore, instead of utilizing a conventional three peak model (gas phase resonance, plasma resonance, and RBC resonance) for spectroscopic data analysis, a four-peak model (gas phase resonance, plasma resonance, and two RBC resonances) can be used, according to our hypothesis. Spectral data were analyzed using both models and the residual analyses better supported the hypothesis of the 4PM. After 4PM Lorentzian fits, the residual error was smooth and almost completely flat in the region between 213 and 235 ppm. On the contrary, several residual peaks were observed in the same region after 3PM Lorentzian fits, indicating the presence of certain spectral component that remained unaccounted for by the 3PM. The residual analysis findings were further supported by the visual observation of asymmetry of the RBC peak and the presence of a small peak splitting for higher glucose concentrations.
It should also be mentioned that the 4PM Lorentzian fits of the spectroscopy data acquired for higher glucose concentrations were much better compared to the 4PM Lorentzian fits for the low glucose levels. For the pure blood samples as well as for the 5 mM glucose concentration, perturbations in 129 Xe-RBC1 and 129 Xe-RBC2 peak positions did not affect R 2 significantly, if at all. Therefore, it can be concluded that the 4PM functions better for blood glucose levels above 5 mM. For concentration ranges between 0 mM and 5 mM, the 4PM can potentially result in some level of inaccuracy due to the four-peak fitting process. This can plausibly be explained by a low glycation level of the 0 mM and 5 mM samples.
The utilization of the 4PM demonstrated that the CS of one of HP 129 Xe-hemoglobin peaks (RBC1) is independent of glucose level, whereas the resonances of the RBC2 peak experience a downfield shift linearly with an increase in the glucose level, and thereby, an increase in HbA 1c concentration. Although the 4PM preserves the overall linear trend of the 129 Xe CS evolution, the net span of CS changes was reduced from~1.25 ppm (for 3PM) down to~0.9 ppm. The CS change in HP 129 Xe resonances due to the glycation of hemoglobin is much smaller than the previously reported changes due to the blood oxygenation by Norquay et al. [24]. It should be mentioned that the actual physiological range of glucose levels in the blood has an upper limit of~35 mM. Therefore, the net expected physiologically relevant change in HP 129 Xe RBC resonances due to HbA 1c formation is limited to~(0.53 ± 0.07) ppm, which is one order of magnitude smaller compared to the blood oxygenation CS changes. The plasma resonance position was unaffected by the elevation of the blood glucose level.
The analysis of the T * 2 relaxation times indicated that the HP 129 Xe-plasma resonance was unaffected by glucose level. Interestingly, the T * 2 values for the RBC1 peak increased non-linearly over the studied range of glucose concentrations. The T * 2 dynamics of the RBC2 peak were similar for glucose levels below 10 mM, albeit leveling out at higher concentrations.
Considering that the HP 129 Xe-plasma resonance was completely unaffected by alternations in glucose concentration, the detected changes of HP 129 Xe RBC relaxation should originate from internal alternations in RBCs' magnetic susceptibilities and/or from changes in HP 129 Xe spin-spin interactions with hemoglobin. There are two plausible mechanisms accounting for HP 129 Xe transverse relaxation changes within the RBC. It was previously suggested that glycation induces iron release from heme pockets of hemoglobin [27] and, therefore, should reduce the spin-spin interaction between HP 129 Xe enclosed in hemoglobin and the heme-iron atoms. Alternatively, conformational changes of HbA 1c may occur in a way that the heme pockets become spatially distanced from HP 129 Xe binding sites, thereby reducing the HP 129 Xe-namely iron spin-spin interactions. Considering that iron release and conformational changes occur during the glycation process itself, it is possible that both mechanisms contribute to the observed changes in the T * 2 relaxation of HP 129 Xe within the RBC.
Although a careful study of the position of the 129 Xe binding sites in HbA 1c is required for a proper explanation of the observed results, it is plausible that conformational changes in hemoglobin due to glycation occur in such a way that the dissociation constant of the HP 129 Xe-HbA 1c becomes substantially lower compared to the dissociation constant of the HP 129 Xe-HbA 0 complexes. Unfortunately, an accurate assessment of the relative changes in dipole-dipole interaction strength is not possible without knowledge of the location of 129 Xe atoms in HbA 1c .
Despite the fact that the knowledge of the 129 Xe-HbA 1c binding sites is of critical importance before attempting a quantitative description of the relaxation mechanisms responsible for the changes in HP 129 Xe relaxation and the alternations of the CS, it should be possible to hypothesize the origin of the 129 Xe-RBC1 and 129 Xe-RBC2 peaks by combining our experimental results with previous knowledge on the glycation process and HbA 1c . Due to the glycation process, iron is released from the heme pocket of hemoglobin [27]. In addition, hydrogen peroxide is produced by autoxidizing glucose, and its concentration builds up with time [44]. Hydrogen peroxide further initiates iron release from hemoglobin [45], affecting HbA 1c much faster compared to HbA 0 [27]. The iron released from heme pockets participates in an iron-dependent Fenton's reaction that becomes a source of free OH· radicals within RBCs [27]. The charge of free radicals deshields the HP 129 Xe nuclei, resulting in a downfield chemical shift detected via MRS. On the other hand, the release of iron may result in a decrease in the spin-spin interaction between the HP 129 Xe bound to the hemoglobin. Since the iron release is much higher for HbA 1c , it is plausible that the reduction in spin-spin interaction would be substantial for HP 129 Xe enclosed within HbA 1c . The T * 2 of HP 129 Xe bond to HbA 0 would be less affected due to the lower amount of iron released. Therefore, based on our results, we hypothesize that the RBC1 peak originates from HP 129 Xe-HbA 1c , while the RBC2 peak corresponds to the HP 129 Xe-HbA 0 signal. It is further plausible that the CS of the HP 129 Xe-HbA1c signal is unaffected by the free OH· radical levels due to the conformational changes of the HbA1c and potential enclosure of HP 129 Xe within the protein structure. On the other hand, the HP 129 Xe-HbA 0 signal experiences a linear downfield CS change due to the interaction with free radicals. It is also possible that the small change in T * 2 of HP 129 Xe-HbA 0 is a result of substantially slower iron extraction from HbA 0 .
Our study has several limitations. First, the initial incubation time of the blood samples with glucose solution was relatively short (only~1 h). The glycation process is relatively slow, consisting of several steps and lasting a span of several days [44]. In the first step, the non-enzymatic Maillard early phase reaction takes place. This occurs after the binding of glucose transferred from the blood to the hemoglobin in RBCs and dictated by the interaction between the aldehyde group of the reducing glucose with the N-terminal amino base and ε-amino base of the lysine residue with the consequent formation of the Schiff base (almidine). This reaction is reversible-it is followed by an irreversible Amadori rearrangement, which results in the production of a stable form of HbA 1c (ketoamine) [46]. The Amadori rearrangement is considered to be the limiting reaction in the glycation process [47,48]. The amount of HbA 1c and synthetized H 2 O 2 , and therefore, the final concentration of free radicals, increases gradually over the glycation time course. Koga et al. used a 1 h incubation time in their study and confirmed the formation of the labile-glycated hemoglobin (aldimine); however, the formation of stable glycated hemoglobin (ketoamine) was not observed [46]. Thus, it is likely that we observed only the initial changes in the HP 129 Xe CS and T * 2 relaxation time. A longer incubation time period will result in more pronounced changes in CS and in T * 2 relaxation time for HP 129 Xe dissolved in RBCs. It should be noted that longer incubation times must be moderated by the fact that a small percentage of hemoglobin will be converted into methemoglobin in standing blood over time [49]. The formation of methemoglobin from hemoglobin within the red blood cells is an ongoing oxidative process that can result from the exposure of blood to the atmosphere in long standing blood samples. Methemoglobin forms when hemoglobin is oxidized to contain iron in the ferric [Fe 3+ ] state rather than the normal ferrous [Fe 2+ ] state [49]. Since methemoglobin is paramagnetic, this itself will cause additional changes to the CS and T * 2 relaxation time [23]. Therefore, longer incubation times are a tradeoff between allowing the full glycation of the blood samples versus limiting the formation of methemoglobin. In addition, all experiments were conducted on sterile citrated sheep blood. While it is not clear how the presence of citric acid affects protein glycation, the reproduction of these experiments may be performed in other types of animal blood.
Our work is a pioneering proof-of-concept study that should be further expanded by performing experiments with purified human hemoglobin, which is very low in the glycated form and separated RBCs from healthy and hyperglycemic subjects. This will allow us to study the effects of glucose on human RBCs. Consequently, the present study can be translated to the utilization of human blood drawn from healthy participants and hyperglycemic participants to observe how glycose-related changes are affected by the presence of other biological moieties from human blood. Furthermore, glycation level measurements are vital for obtaining reliable conclusions regarding the glucose effect on the 129 Xe spectroscopic properties in blood. Additionally, no studies have yet been performed that investigate HP 129 Xe binding to the glycated form of hemoglobin, which can be assessed in future X-ray diffraction (XRD) studies.

Sample Preparation
A 500 mM solution of glucose was prepared by dissolving 3.6 g of D-glucose (Sigma-Aldrich, St. Louis, MO, USA) in 40 mL of 1× phosphate-buffered saline (PBS) at pH 7.4 at room temperature. The mixture was stirred for complete dissolution and used as a stock solution for the preparation of the blood samples. Various volumes of the glucose solution in PBS were mixed with 20 mL of fresh citrated sheep blood (Cedarlane, Burlington, CA, USA) to create the following set of concentrations: 5,10,15,20,25,35,45, and 55 mM. The volume of glucose solution added to the blood was at least one order of magnitude smaller than the volume of the blood, in order to minimize blood dilution effects. A total of 20 mL of the pure citrated sheep blood without any additives served as a control sample. All blood samples were allowed to equilibrate to room temperature for approximately 1 h. The reported incubation times for glycation in the literature vary between 30 min up to hundreds of hours, depending on the incubation protocol [50]. Moreover, the vast majority of studies were able to confirm hemoglobin glycation for all of the incubation time periods. Therefore, successful hemoglobin glycation is expected within our experimental time frame.

Magnetic Resonance Spectroscopy (MRS)
129 Xe gas was polarized up to 56% using a XeBox-10E polarizer (Xemed, Durham, NH, USA) and disposed into 1 L Tedlar bags. The experimental setup ( Figure 6) used for mixing HP 129 Xe with blood was similar to that used by Norquay et al. [25]. Mixing was performed using an exchange module (Superphobic MicroModule 0.5 × 1 G680 Contactor; Membrana, Charlotte, NC, USA). A steady flow of HP 129 Xe was set through the exchange module with the help of a pressure chamber pressurized with a continuous flow of N 2 gas. The flow rate of N 2 into the pressure chamber was controlled by a ventilator. The solution of sheep blood in the 10 mL syringe was connected to an exchange module and placed in a custom-built dual 1 H/ 129 Xe quadrature MRI birdcage coil, while the other empty syringe was connected to the other side of the exchange module. The blood was pumped back and forth manually through the exchange module perpendicular to the 129 Xe flow for~6 s. This technique allowed a sufficient amount of HP 129 Xe to dissolve in the sheep blood and avoid the formation of gas bubbles in the blood. of the pure citrated sheep blood without any additives served as a control sample. All blood samples were allowed to equilibrate to room temperature for approximately 1 h. The reported incubation times for glycation in the literature vary between 30 min up to hundreds of hours, depending on the incubation protocol [50]. Moreover, the vast majority of studies were able to confirm hemoglobin glycation for all of the incubation time periods. Therefore, successful hemoglobin glycation is expected within our experimental time frame.

Magnetic Resonance Spectroscopy (MRS)
129 Xe gas was polarized up to 56% using a XeBox-10E polarizer (Xemed, Durham, NH, USA) and disposed into 1 L Tedlar bags. The experimental setup ( Figure 6) used for mixing HP 129 Xe with blood was similar to that used by Norquay et al. [25]. Mixing was performed using an exchange module (Superphobic MicroModule 0.5 × 1 G680 Contactor; Membrana, Charlotte, NC, USA). A steady flow of HP 129 Xe was set through the exchange module with the help of a pressure chamber pressurized with a continuous flow of N2 gas. The flow rate of N2 into the pressure chamber was controlled by a ventilator. The solution of sheep blood in the 10 mL syringe was connected to an exchange module and placed in a custom-built dual 1 H/ 129 Xe quadrature MRI birdcage coil, while the other empty syringe was connected to the other side of the exchange module. The blood was pumped back and forth manually through the exchange module perpendicular to the 129 Xe flow for ~6 s. This technique allowed a sufficient amount of HP 129 Xe to dissolve in the sheep blood and avoid the formation of gas bubbles in the blood. A clinical Philips Achieva 3.0 T MRI scanner (Philips, Andover, MA, USA) equipped with a custom-built dual 1 H/ 129 Xe quadrature MRI birdcage coil was used for all 129 Xeblood spectroscopy acquisitions. The coil was tuned to 35.33 MHz, and the scanner was shimmed on the 1 H signal from blood to correct for B1 inhomogeneities. Prior to taking any measurements, the coil was calibrated for the blood samples by applying a series of ten RF pulses with a s10° flip angle centered on the gas peak position of HP 129 Xe resonance (for this, 1 mL syringes filled with HP 129 Xe were placed into the coil near the blood sample). A clinical Philips Achieva 3.0 T MRI scanner (Philips, Andover, MA, USA) equipped with a custom-built dual 1 H/ 129 Xe quadrature MRI birdcage coil was used for all 129 Xeblood spectroscopy acquisitions. The coil was tuned to 35.33 MHz, and the scanner was shimmed on the 1 H signal from blood to correct for B 1 inhomogeneities. Prior to taking any measurements, the coil was calibrated for the blood samples by applying a series of ten RF pulses with a s10 • flip angle centered on the gas peak position of HP 129 Xe resonance (for this, 1 mL syringes filled with HP 129 Xe were placed into the coil near the blood sample).
To measure the effect of glucose concentration on the CS of the 129 Xe dissolved in plasma ( 129 Xe-plasma) and in RBC ( 129 Xe-RBC) resonances, high-resolution single voxel spectroscopy was acquired for all samples. The receiver bandwidth was equal to 22 kHz and the number of samples was set to 4096, yielding a 0.15 ppm spectral resolution. A total of 0 ppm was set in between the peaks of 129 Xe dissolved in plasma and the RBCs. A 90 • rectangular excitation pulse was utilized. An FID spectrum was acquired with a TR of 189.6 ms and a TE of 0.25 ms. Measurements were repeated five times for each sample.

Data Reconstruction and Statistical Analysis
All data were initially analyzed using custom MatLab scripts in Matlab 2020b (Mathworks, Natick, MA, USA). The MRS spectra were postprocessed in Origin2021b (OriginLab, Northampton, MA, USA) for CS and T * 2 relaxation time assessment. HP 129 Xe MRS spectra were fitted to either three (gas resonance, 129 Xe-plasma resonance, and 129 Xe-RBC resonance) or four (gas resonance, 129 Xe-plasma resonance, and two 129 Xe-RBC resonances) Lorentzian curves. The quality of fit was accessed via R 2 value and through the residual fit error, which was automatically calculated at the end of the fitting algorithm. The comparison between the three and four peak model was performed based on the R 2 and the residuals.
The full width half maximum (FWHM) of Lorentzian curves was utilized for calculations of the T * 2 of 129 Xe dissolved in RBCs and plasma using the following equation [51]: To evaluate statistical significance of the acquired results, a paired two-tailed t-test was used with a statistical significance level of 0.05. Pearson's correlation coefficient was calculated between the CS changes of each HP 129 Xe dissolved phase resonance and the glucose concentration. All statistical analysis was conducted using Origin2021b v.9.8.5.212 software.

Conclusions
In this work, for the first time, we demonstrated glucose-induced changes in CS and T * 2 relaxation of HP 129 Xe dissolved in blood. By using a residual analysis, we observed the evidence for a third dissolved-phase resonance, which was attributed to HP 129 Xe bound to HbA 1c produced during HbA 0 glycation. Our four-peak model suggests that one dissolved 129 Xe resonance originates from plasma, whereas two other peaks originate from 129 Xe encapsulated by HbA 0 and HbA 1c (the fourth peak stems from the gas phase resonance). We observed a linear dependence of the 129 Xe-HbA 0 resonance CS on the blood glucose levels. The 129 Xe-HbA 1c T * 2 changed non-linearly with an increase in glucose level. The HP 129 Xe-plasma resonance was not affected by alternations in the glucose level.

Data Availability Statement:
The data that support the findings of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest:
Author I.C.R. was employed by the company Xemed LLC, Durham, NH, USA. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.