Перенос дисфункции химического обмена MRI для нейродегенеративных заболеваний: Обновленная информация о клинических и предклинических исследованиях.

Introduction

The prevalence of neurodegenerative diseases is increasing as human longevity extends, posing significant challenges to public health and socioeconomic systems. These diseases are characterized by the progressive loss of selected groups of neurons and associated cognitive and/or movement deficiencies. Common pathological features of these conditions are the misfolding and aggregation of abnormal proteins, altered metabolism, and neuroinflammation in the central nervous system (Dugger and Dickson, 2017). Screening and early diagnosis of neurodegenerative diseases allow for early intervention, improving prognosis and reducing financial burdens on individuals and the public health sector (Barnett et al., 2014; Afentou et al., 2019). There is an emerging need for developing imaging biomarkers for neurodegenerative diseases; one available option is structural imaging that detects brain morphological abnormalities, despite little noticeable change can be detected especially in the early stages before the onset of brain atrophy (Frisoni et al., 2010). Molecular imaging provides insights into the spatial distribution and molecular specificity of pathological changes, especially in early stages with subtle changes in metabolism and abnormal protein aggregation (Nordberg, 2011). While many current techniques depend on radioactive isotopes, recent advances in magnetic resonance imaging (MRI) have enabled the detection of brain pathology in vivo without ionizing radiation. In particular, chemical exchange saturation transfer (CEST) MRI is emerging as a powerful tool for detecting specific chemical groups, such as amide, amine, and aliphatic protons in exchange with the free water pool, offering high sensitivity and improved resolution compared with isotope-based tomography or magnetic resonance spectroscopy (MRS) (van Zijl and Yadav, 2011; van Zijl et al., 2018).

This review begins with an introduction to the fundamentals of CEST MRI, covering saturation pulses, readout, post-processing, and contrast types. It then focuses on the applications, findings, and perspectives of CEST MRI in clinical and preclinical studies of neurodegenerative diseases, alongside advanced in vivo CEST MRI techniques involved in these studies.

Search Strategy and Selection Criteria

In the present review, we searched papers from Google Scholar ( https://scholar.google.com/ ) and PubMed ( https://pubmed.ncbi.nlm.nih.gov/ ) and included full-text papers published in English between 2009 and 2024. To ensure relevant results were retrieved, we focused on finding combinations of neurodegenerative diseases such as “neurodegenerative disease,” “Alzheimer’s disease,” “Parkinson’s disease,” “multiple sclerosis,” “Huntington’s disease,” “amyotrophic lateral sclerosis,” and imaging modality keywords such as “CEST,” “chemical exchange saturation transfer,” “GluCEST,” “amide proton transfer,” “glucose,” “GlucoCEST,” “creatine,” “CrCEST,” “Myo-inositol,” “MRI” in the title, abstract, or keywords. Further screening of titles and abstracts was performed to ensure relevance to our research objectives, and if deemed pertinent, the entire article was accessed. We included research papers that were aligned with the following inclusion criteria for a given neurodegenerative disease: studies that mainly focused on the detection of in vivo biomarkers using CEST techniques, either in human or animal models. A full list of publications with essential details is provided in Additional Table 1 , and a visualization of all the preclinical and clinical publications can be seen in Figure 1 .

Timeline of CEST MRI studies for neurodegenerative diseases.

CEST MRI studies with authors represented by colored dots on the timeline. Shaded plot is an annual number of publications of the same disease type. Starred “*” publications of the same year with the same name are the same publications but reported two CEST contrasts. AD: Alzheimer’s disease; APT: amide proton transfer; CEST: chemical exchange saturation transfer; CrCEST: creatine-weighted CEST; GluCEST: glutamate-weighted CEST; glucoCEST: glucose-weighted CEST; GlyCEST: glycine-weighted CEST; HD: Huntington’s disease; MICEST: myo-inositol-weighted CEST; MRI: magnetic resonance imaging; MS: multiple sclerosis; PD: Parkinson’s disease; rNOE: relayed Nuclear Overhauser Effect.

Chemical Exchange Saturation Transfer Magnetic Resonance Imaging Techniques

CEST MRI is a family of molecular imaging tools that detects the concentration and chemical environment of exchangeable protons in certain chemical groups, such as amide groups on mobile proteins and peptides, and amine groups in glutamate. The abundance of the chemical groups in vivo and huge signal amplification by the CEST mechanism enables endogenous contrast agents to detect metabolism change by CEST MRI. This was initially demonstrated by Ward et al. (2000) through the selective radiofrequency (RF) saturation of the protons in a group of chemicals, including amino acids, sugars, nucleotides, and heterocyclic ring chemicals. In CEST signal acquisition, protons capable of exchange on solute molecules, such as -OH, -NH, and -NH 2 groups, are magnetically labeled (saturated) by RF irradiation. The migration of this magnetic label (saturation) through proton exchange is then monitored by observing the signal reduction in the bulk water proton pool (van Zijl and Yadav, 2011; Zhou et al., 2023). In a state of chemical equilibrium, the proton exchange process between the solute and water is characterized or quantified by a classical “dual-pool” exchange model ( Figure 2 A ). For successful CEST imaging, the exchange rate k ex should be smaller than or equal to the chemical shift difference (Δω) between the solvent and the solute (van Zijl and Yadav, 2011). If the (Δω) is insufficiently large, saturation of solute protons results in partial saturation of the water pool, a phenomenon termed the direct (water) saturation (DS) effect. To maintain a discernible spectral separation between water and solute protons, particularly at lower field strengths, k ex must reside within the slow to intermediate exchange rate regime (Ward et al., 2000). Additionally, k ex must surpass the longitudinal relaxation rate of the solute protons, thereby allowing ample time for the chemical exchange to occur before complete relaxation.

Fundamentals and quantification of CEST MRI.

(A, B) CEST principles illustration using a dual-pool model. (C) Z-spectrum (or CEST spectrum) generation. (D) MTR asym analysis. Reprinted with permission from van Zijl and Yadav (2011). CEST: Chemical exchange saturation transfer; k sw / k ws : proton exchange rate of solute-to-water or water-to-solute; MTR: magnetization transfer ratio; ppm: parts per million; RF: radiofrequency pulse; S sat / S 0 : signal intensity from saturated/unsaturated tissue; t sat : saturation time.

The concentration of the abundant water proton pool is approximately 111 M, significantly exceeding that of the saturated solute proton pool. Due to the inherently low concentration of solute protons, merely transferring their saturation effect to water protons is insufficient to elicit a discernible impact (van Zijl and Yadav, 2011). Upon each exchange of a saturated solute proton with an unsaturated water proton, the latter becomes the new unsaturated solute proton that can be saturated again. This process is repeated with each exchange. If the saturation time ( t sat ) extends into the second range, continuous RF pulses can significantly amplify the saturation effect. Over time, this amplified effect becomes detectable in the water signal. This phenomenon allows indirect imaging of solutes at low concentrations ( Figure 2 B ).

For CEST MRI, magnetic field strength impacts not only the sensitivity but also the specific absorption rate (SAR) and the overall feasibility of the CEST imaging protocol (Gao et al., 2021). High-field MRI scanners are generally favored for CEST imaging because they offer better sensitivity and higher signal-to-noise ratio, which enhance the detection of exchangeable protons with small frequency offsets (Dula et al., 2013). However, in a clinical setting, lower field strengths such as 3T are more commonly available and still provide sufficient sensitivity for certain metabolites with inherently larger frequency shifts, opening up possibilities for clinical integration.

The contrast observed in CEST imaging is influenced by the saturation pulses, readout techniques, and the analysis methods employed. An overview of the fundamental principles and process of CEST MRI will be presented, followed by introductions to available contrasts in the CEST family.

Saturation pulse

The selection of a saturation pulse is important for optimizing contrast sensitivity and specificity. For the saturation part, the conventional methods for creating CEST involve the use of continuous wave saturation pulses or pulse trains ( Figure 3 A ). The selection of saturation pulse parameters, including shape, power, duration, delay time, and frequency offset, is essential for the effective saturation of exchangeable protons and the minimization of SAR for RF safety (Liu et al., 2013). Using lower saturation power can help minimize the direct saturation of water when targeting MR frequencies closer to the water resonance. However, the clinical translation of CEST imaging with long continuous wave pulses faces challenges due to hardware and SAR limitations. The majority of CEST applications on human MRI scanners employ a series of block or shaped RF pulses for CEST saturation, which remains the most commonly used approach for human studies to date (Sun et al., 2008; Zhu et al., 2010; Schmitt et al., 2011; Zu et al., 2011). CEST contrast can be further enhanced by optimizing pulsed CEST parameters, including pulse shapes, sub-pulse duration, number of pulses, and inter-pulse delay, offering an alternative approach to improve CEST sensitivity ( Figure 3 B–D ; Schmitt et al., 2011; Zu et al., 2011; Xiao et al., 2015; Kim et al., 2019). The use of shaped pulses, such as Gaussian shapes and Block Pulse Trains ( Figure 3 C ), has been shown to produce comparable saturation to square-shaped continuous wave pulses while minimizing off-resonance effects, particularly for protons with slow exchange rates. Despite the potential benefits of optimized saturation parameters, there is currently no consensus on the ideal saturation time or power for clinical CEST MRI studies. Hardware limitations, including RF amplifier and transmitter coil capabilities, can also impact pulse selection and generate significant B1 inhomogeneity, underscoring the need for standardization to ensure consistent results across radiology clinics (Jones et al., 2018).

Examples of a saturation pulse sequence for CEST.

(A) CW RF saturation. (B1–B3) Three pulse-train RF saturation examples with high DC sat (≥ 90%), respectively, Single-lobe sinc-Gaussian or any other saturation pulses may be used in (B1) and (B2), and Fermi pulses in (B3). (C) Shorter pulse-train RF saturation with high DC sat = 95%. (D) Pulse-train RF saturation with standard DC sat = 50%. Single-lobe sinc-Gaussian saturation pulses are used as an example. Reprinted with permission from Zhou et al. (2022). CW: Continuous wave; DC sat : Saturation duty cycle (= t p /[t p + t d ]); t d : Interpulse delay; t p : Individual pulse element duration in a pulse train or a pulsed steady state; T sat : saturation time.

Readout

Data acquisition is a crucial component of the CEST sequence, regardless of the saturation scheme employed. To optimize readout efficiency while maintaining image quality, various fast readout sequences have been utilized. These include echo planar imaging (EPI) (Akbey et al., 2019; Mueller et al., 2020), turbo spin echo (Togao et al., 2017; Liu et al., 2019a), gradient and spin echo (Zhou et al., 2013), gradient echo (Zaiss et al., 2018; Deshmane et al., 2019), optimized turbo spin echo with variable flip angle evolutions (Zhang et al., 2020a), periodically rotated overlapping parallel lines enhanced reconstruction (Guo et al., 2020; Sui et al., 2021), and stack-of-spirals gradient echo readout (Wang et al., 2024). The deployment of these sequences has significantly enhanced the efficiency of CEST imaging, enabling faster data acquisition and improved image quality.

EPI has been extensively developed and modified due to its superior acquisition efficiency, with variants such as snapshot EPI (Mueller et al., 2020) and spatiotemporal encoding (Huang et al., 2017). However, EPI-based fast imaging is susceptible to image distortion caused by magnetic susceptibility and ghosting artifacts from lipid signals, which can compromise CEST image quality (Sun et al., 2005; Villano et al., 2021). Promising methods like gradient and spin echo CEST can still face challenges with susceptibility and lipid artifacts associated with their EPI component. In contrast, turbo spin echo is less prone to these artifacts and has become a commonly used readout approach for CEST on clinical scanners. To tackle issues of limited covering regions in a CEST scan, multislice imaging is an approach to achieve volumetric CEST imaging, with techniques such as simultaneous multislice gradient echo and phase-offset multiplanar CEST-FISP sequences (Randtke et al., 2017). However, conventional multislice methods are limited by their time-consuming nature and the decrease in CEST effect over time and variations among different slices (Zhang et al., 2023).

Optimized readout techniques, such as those utilizing variable flip angle evolutions, have shown promise in improving CEST imaging efficiency while minimizing artifacts. These approaches can offer advantages such as reduced SAR and faster acquisition times, making them suitable for clinical applications. By leveraging advances in readout technology, CEST imaging can be further optimized for improved efficiency and image quality.

Chemical Exchange Saturation Transfer Magnetic Resonance Imaging Post Processing

Z-spectra and B 0 /B 1 correction

Post-processing steps are crucial in CEST MRI measurement to separate contributions from different exchange pools and interpret the saturation effect in CEST data. The term “Z-spectra” used in CEST arises because the saturation effect primarily affects the acquired signal indirectly via the longitudinal magnetization vector along the Z-axis. Around the central water frequency of 0 ppm, the Z-spectrum exhibits a symmetric dipolar saturation distribution ( Figure 2 C ). Correcting the static magnetic ( B 0 ) field inhomogeneities is necessary for CEST post processing (Paech and Loi, 2023). A commonly used method for B0 correction is water saturation shift referencing (Kim et al., 2009). At ultra-high field strengths, further correction for RF transmit ( B 1 ) field inhomogeneities is desired for accurate interpretation of CEST data, as the strength of the CEST effect is highly dependent on the applied saturation power (Windschuh et al., 2015).

After calibrating the B 0 and B 1 inhomogeneities, Z-spectrum analysis methods are applied. Z-spectrum analyses typically involve computing the difference between the acquired Z-spectrum (Z) and a reference spectrum ( Z ref ).

Z-spectra analysis

MTR asym analysis

Magnetization transfer ratio asymmetry ( MTR asym ) analysis eliminates magnetization transfer (MT) and DS effects from the acquired signal by assuming water resonance frequency symmetry (Zhou et al., 2003b). Based on the assumption that the DS effect is symmetric around the water resonance frequency, CEST effects can be isolated by setting the Z ref as the Z value at the opposite frequency offset. MTR asym is especially effective on CEST data without CEST/MT on the negative side of Z-spectra ( Figure 2 D ). However, cares must be taken when using MTR asym to analyze in vivo data that includes multiple molecular contributions, including relayed nuclear Overhauser effect (rNOE) and magnetization transfer contrast (MTC).

Lorentzian difference analysis

Lorentzian difference analysis (LDA) employs a Lorentzian line shape to model DS effects as the Z ref (Jones et al., 2013). The CEST contrast is then calculated using the difference between Z and Z ref . If the rNOE peak is of interest and the MT contrast needs to be excluded, an additional Lorentzian term to account for the MT effect can be introduced when constructing the Z ref (Zhang et al., 2017a).

Multi-pool Lorentzian fitting

Multi-pool Lorentzian fitting assumes that all exchange pools follow a Lorentzian line shape and approximates the acquired Z-spectrum as a linear combination of multiple Lorentzian terms (Zaiß et al., 2011). By fitting the summation of Lorentzian terms, the characteristics of each pool can be determined.

Other traditional methods

Apart from the commonly employed post-processing analyses aforementioned, several alternative approaches are also available for specific CEST studies. Polynomial and Lorentzian line-shape fitting (PLOF) combines polynomial and Lorentzian line shapes to distinguish the creatine (Cr) signal from other CEST signals, such as amine and aromatic protons at 2 ppm (Chen et al., 2019a). Currently, PLOF is primarily used for analyzing creatine signals (Wang et al., 2024) and amide CEST (Sui et al., 2021). The three-offset method separates the CEST peak by assuming that the MT effect and the DS effect within a certain CEST peak boundary can be approximated as linear (Xu et al., 2014). This method is more suitable for high-field applications, as it relies on well-defined CEST peak boundaries to determine the acquisition region. The double saturation powers (DSP) method is used to isolate the amide proton transfer (APT) and rNOE signals while suppressing MT and DS signals (Zhao et al., 2023). DSP constructs the Zref based on the Z-spectrum acquired with a high B1 amplitude ( Z H ). Then, the CEST contrast is calculated by subtracting the Z-spectrum acquired with a low B1 amplitude ( Z L ) from the Z ref . The DSP method can quantify the APT and rNOE effects with considerably increased specificities and reduced confounding signals from DS, MT, and fast-exchange CEST effects.

To compensate for T 1 contributions, the apparent exchange-dependent relaxation (AREX) method is proposed based on an approximate solution of the Bloch-McConnell equations under several assumptions (Zaiss et al., 2015). AREX excludes T 1 contributions by analyzing the Z-spectrum inversely, followed by multiplication by the longitudinal relaxation rate (R 1 ), and can be used for all Z-spectrum analysis methods mentioned above to obtain an exchange rate-dependent parameter that is insensitive to spillover, MT, and R 1 .

Artificial intelligence-based analysis methods

Recent years have seen a rise in the use of artificial intelligence for CEST processing, significantly enhancing the efficiency and accuracy of data analysis and enabling better extraction of biological information (Cohen et al., 2018; Zaiss et al., 2019a; Glang et al., 2020; Kim et al., 2020; Perlman et al., 2021; Huang et al., 2022). Methods such as deepCEST (Zaiss et al., 2019a; Glang et al., 2020) have demonstrated the potential of neural networks in analyzing the Z-spectrum. A subsequent study demonstrated that AREX contrasts could also be predicted using deep neural network (deepAREX) (Huang et al., 2022). Some other studies have shown that deep learning-based analysis methods with sparse frequency offsets can effectively accelerate CEST imaging (Li et al., 2020; Cheema et al., 2024; Chen et al., 2024; Xiao et al., 2024). Deep learning has been applied for tumor classification (Bie et al., 2022) and water saturation shift referencing fitting (Mohammed Ali et al., 2023), demonstrating improved stability and efficiency. Additionally, deep learning-based methods have also been proposed for improving CEST MRI in signal-to-noise ratio (Chen et al., 2023a; Radke et al., 2023; Kurmi et al., 2024) and spatial resolution (Wu et al., 2023; Zu et al., 2023; Liu et al., 2024a, b; Pemmasani Prabakaran et al., 2024; Xu et al., 2024; Yang et al., 2024), thereby improving the quality of CEST contrast maps.

CEST MR fingerprinting (MRF) is an innovative technique for the simultaneous quantification of multiple proton exchange parameters (Geades et al., 2017; Cohen et al., 2018; Zhou et al., 2018; Heo et al., 2019). By leveraging deep learning which addresses the limitations of traditional dictionary-based methods, CEST MRF can offer faster quantitative parameter estimation (Kim et al., 2020; Perlman et al., 2020, 2022a, b, 2023; Kang et al., 2021; PCohen and Otazo, 2023; Cohen et al., 2023; Singh et al., 2023). These advancements contribute to a deeper understanding of molecular processes and pathology.

Chemical Exchange Saturation Transfer Magnetic Resonance Imaging Contrasts

CEST MRI is a molecular imaging technique that visualizes metabolic dynamics in living tissues ( Figure 4 ). A comprehensive list of these CEST contrasts along with their exchangeable proton groups and typical chemical shift is presented in Table 1 .

Commonly used endogenous CEST contrasts.

Chemical exchange of protons for Amide (3.5 ppm), amine (3 ppm), amine/guanidium (2 ppm), hydroxyl (0.8 – 1.2 ppm), Nuclear Overhauser effect (–1 to –5 ppm) showing decreased water signal on corresponding frequency shifts in a Z-spectrum, generated by simulation for typical tissue environment. CEST: Chemical exchange saturation transfer; DS: direct (water) saturation; MTC: magnetization transfer contrast.

Table 1 Other chemical exchange saturation transfer (CEST) contrasts and their properties Exchangeable group Referred name Chemical shift (ppm) Contrast source –NH2 Amine CEST 1.8–3.0 Endogenous GluCEST (Glutamate) 1.8–3.0 –NH3 + CrCEST (Creatine) 1.8–2.0 –OH Hydroxyl CEST 0.5–1.5 Endogenous & Exogenous Exogenous: GlucoCEST (Glucose) MICEST (Myoinositol) 0.3–0.6

Other chemical exchange saturation transfer (CEST) contrasts and their properties

Amide proton transfer

Neurodegenerative diseases, such as Alzheimer’s disease (AD) and Parkinson’s disease (PD), are often characterized by proteinopathies. APT imaging, which primarily detects amide protons on proteins, has naturally been one of the first techniques applied to image these pathologies in vivo . APT-weighted (APTw) MRI offers a promising alternative for detecting amyloid and tau pathologies in AD, as well as Lewy bodies composed of aggregated α-synuclein in PD. Compared with positron emission tomography (PET) and single-photon emission computed tomography, APTw MRI is more cost-effective, widely available, and avoids the use of ionizing radiation, making it well-suited for integration into routine clinical MRI workflows. Endogenous mobile proteins can be detected in vivo using APTw MRI, which has already been adapted for research and clinical applications, particularly in imaging tumors and strokes, on standard 3T MRI scanners.

APT imaging is the most studied branch of CEST imaging for its feasibility and robustness at 3T and sensitivity to mobile peptides and proteins in tissues (Zhou et al., 2003b). The APT effect in MRI serves as a non-invasive pH meter in biological tissues by exploiting the pH-sensitive chemical shift variations of amide protons found in proteins and peptides. This phenomenon manifests as a downfield shift in the NMR spectrum under acidic conditions due to increased electron shielding, and an upfield shift in alkaline environments due to reduced shielding (Zhou et al., 2003b). The assessment of the APT effect typically employs the MTR asym , as mentioned in the previous chapter, with Δω located at 3.5 ppm (Zhou et al., 2019).

The multiple contributions evident in MTR asym(3.5 ppm) have led to the designation of APT images defined by MTR asym(3.5 ppm) as APT-weighted (APTw) images (Li et al., 2023). This offset has been strategically chosen for its proximity to the water proton chemical shift, which facilitates efficient water suppression and increases the signal-to-noise ratio, thereby improving the detection of pH-dependent changes. Furthermore, it minimizes spectral overlap with other tissue components, such as lipids and metabolites, ensuring clearer discrimination of the APT effect. APTw imaging has gained broad recognition as a reliable indicator for brain tumor characterization (Zhou et al., 2003a). This consensus has served as the foundation for the development and subsequent commercial release of the APTw imaging protocol for routine use in 3T clinical MRI systems.

For the clinical implementation of APTw MRI, rapid volumetric imaging is essential. Three-dimensional imaging is favored as it ensures uniform saturation effects across the volume, thereby reducing variability in saturation losses due to T 1 relaxation that may occur between slices. In particular, a study (Zhu et al., 2010) has established a 3D APTw MRI method ( Figure 3 , which permits rapid acquisition on 3T clinical MRI equipment (Togao et al., 2016).

In addition to semi-quantitative MTR asym analysis, a comprehensive suite of advanced data processing techniques has been developed for the precise quantification of CEST. These include model-based Z-spectrum methods, such as multi-pool Lorentzian fitting (Zaiß et al., 2011) introduced in Post Processing section, and Bloch-McConnell equation fitting (Tee et al., 2012), which facilitate the quantification of various tissue parameters. The choice of the optimal APT/CEST analysis technique is contingent upon the RF saturation amplitude settings. Moreover, MR fingerprinting has made significant progress in achieving absolute quantification of CEST parameters, including exchange rates and concentrations (Cohen et al., 2018, 2023; Zhou et al., 2018; Perlman et al., 2023). This approach holds promise for enhancing detection accuracy, but further clinical verification is still required.

The Food and Drug Administration (FDA) has granted clinical authorization to APT imaging in the detection, staging, and diagnosis of brain tumors and the identification of tumor genetic markers (Jiang et al., 2017; Zhou et al., 2022). However, the clinical application of APT/CEST in NDDs is currently limited due to some technical issues, including scanner RF amplifier limitations, SAR requirements, long scan time, and the possibility of B 0 / B 1 inhomogeneity artifacts (Zhou et al., 2019).

Glutamate-weighted chemical exchange saturation transfer

Glutamate, the most abundant excitatory neurotransmitter in the human brain, serves as a cornerstone in synaptic transmission, plasticity, and the intricate network of neural communications (Wang and Reddy, 2017). It is synthesized in the tricarboxylic acid cycle and serves as the chemical precursor for gamma-aminobutyric acid (GABA), which is the main inhibitory neurotransmitter in the mammalian central nervous system. Glutamate is released by glutamatergic neurons upon action potential and binds to receptors on the postsynaptic neuron. Subsequently, it is cleared from the synapse by glutamine synthetase on the astroglia, transported back to the glutamatergic neuron, and converted back to glutamate by the enzyme glutaminase. This process is known as the glutamate–glutamine cycle. Impaired glutamate transport causes delayed synaptic clearance and accumulation, leading to excitotoxicity and neuronal death, and dysregulated glutamate metabolism can be a promoting factor to amyloid-beta (Aβ) aggregation, synaptic dysfunction, and neuroinflammation (Andersen et al., 2022). Aberrations in glutamate metabolism have been implicated as a pivotal factor in the pathogenesis of AD, and glutamate clearance impairment has been observed in the cortex and hippocampus region of AD patients (Scott et al., 2011).

Magnetic resonance-based detection, pioneered by single-voxel spectroscopy, has been the leading method for glutamate quantification, with its drawback being limited spatial resolution. Spectroscopic imaging and CEST method provide high spatial resolution imaging of glutamate (Cember et al., 2023). Amine protons on glutamate can be labeled and exchanged with a water pool to produce glutamate-weighted CEST (GluCEST) contrast. A promising method for imaging glutamate involves targeting its rapidly exchanging amine protons, which resonate at approximately 3 ppm relative to the water signal. Studies have shown that, at physiological concentrations and pH, these glutamate protons contribute significantly more to CEST signals at 3 ppm under high irradiation power ( B 1 ) compared with other brain metabolites (Cai et al., 2012). This finding suggests that the “GluCEST” technique offers high specificity for imaging glutamate. A potential confounding factor in GluCEST imaging is the contribution of protein lysine amine protons, which also resonate near 3 ppm and undergo rapid exchange with water. This overlap can make it challenging to attribute the CEST signal exclusively to glutamate, particularly at lower magnetic field strengths where glutamate signals are visually inseparable (De Graaf, 2019). As a result, GluCEST methods have primarily been applied on high-field scanners (7T or higher) to achieve better specificity. However, this approach comes with challenges, including increased SAR limits and B 1 inhomogeneities, which can complicate data acquisition at these high field strengths.

Glucose-weighted chemical exchange saturation transfer

The hydroxyl protons of cyclic hexoses, such as D-glucose (D-Glc) and its analogs (e.g., 2-DG, 3-OMG, D-GlcN, and GlcNAc), can be detected under low pH and low temperature in various peaks of NMR from 0.6 ppm to 2.9 ppm. However, at physiological pH, these hydroxyl protons exchange rapidly with water, causing their resonances to broaden and merge with the water signal in the Z-spectrum. Typically, under such conditions, the hydroxyl proton signals are observed within the 0–2 ppm range in the Z-spectrum using glucose-weighted CEST (GlucoCEST) MRI. Based on the number of protons involved, the maximum signal appears around 1.2 ppm for most glucose analogs which has become a popular offset frequency to be used in GlucoCEST studies (Van Zijl et al., 2007; Chan et al., 2012; Walker-Samuel et al., 2013; Zaiss et al., 2019b). When sugars are used as contrast agents, the dynamic difference curve of the Z-spectrum is analyzed. This allows for the partial subtraction of interfering resonances from other molecules and the acquisition of glucose kinetics information, a technique known as dynamic glucose-enhanced (DGE) MRI (Xu et al., 2015; Paech et al., 2017b; Huang et al., 2020; Paech and Radbruch, 2021; Knutsson et al., 2023).

A key distinction between DGE approaches and studies aimed at detecting endogenous exchangeable species lies in the administration of the contrast agent: glucose (or its analogs) is administered intravenously during the MRI scan. Consequently, changes in the observed signal can be directly attributed to the injected contrast agent. Most studies quantify the DGE signal by calculating the difference between post-injection scans at each time point and the average of pre-injection scans, followed by analyzing the resulting dynamic difference curve (Knutsson et al., 2023).

Despite significant research efforts, the exact origin of the DGE contrast remains unclear. While glucose hydroxyl protons are believed to be the primary contributors to the DGE signal, other physiological effects may also influence the observed signal changes. For instance, a sudden increase in arterial osmotic pressure could induce large signal alterations. Additional factors, such as changes in blood flow, oxygenation levels, and even volume shifts in blood or cerebrospinal fluid, may also play a role (Kim et al., 2022b).

Developing new MRI techniques with enhanced sensitivity to non-radioactive glucose analogs holds promise as a potential alternative to the widely used FDG-PET. Although stronger CEST signals have been observed in the human brain at 7 Tesla (Xu et al., 2015; Paech et al., 2017a; Schuenke et al., 2017a, b; Knutsson et al., 2018), the results at the more widely used clinical field strength of 3 Tesla have been less encouraging (Herz et al., 2019; Xu et al., 2020). As such, the development of DGE MRI methodology for clinical applications remains incomplete, presenting many opportunities for further advancement.

Relayed nuclear overhauser effect

The rNOE, a dipole-dipole coupling phenomenon initially discerned by Albert Overhauser in 1951, manifests as a discernible variation in the NMR signal intensity when the proximity between two nuclei falls below a critical distance threshold (Overhauser, 1953). In CEST studies, a signal attenuation centered at –3.5 ppm on the Z-spectrum is attributed to rNOE interactions (van Zijl et al., 2018; Huang et al., 2021a; Zhou et al., 2023). These rNOE signals, originating from the MT between free water and bound water associated with aliphatic chains (–CH2–n) on lipids, can induce a substantial drop in the water signal at a clinical field strength of 3 Tesla. This results in a clear “dip” in the CEST spectrum for the human brain. This suggests that the technique can be used to image lipid changes, such as demyelination and cell membrane damage in neurodegenerative diseases.

The transfer of protons between tissue molecules and water can occur via a single direct step, such as the proton or water exchange. In this process, the saturation of the original proton pool is continuously refreshed, leading to an accumulated effect in a coupled large proton pool. Alternatively, it may involve multiple MT pathways. One such instance is the rNOEs that originate on aliphatic protons, which subsequently undergo a relay mechanism facilitated by proton or water exchange between a specific molecule and the aqueous environment. This process culminates in the aggregation of magnetization perturbations, a phenomenon known as the rNOE (van Zijl et al., 2018). Thus, in vivo , Z-spectrum comprises signals from both CEST effects and non-exchangeable rNOE proton pools, predominantly observed in mobile macromolecules ( Figure 5 ). Recent advancements have clarified that MTC effects ( Figure 5 ) are predominantly attributed to rNOEs, although they may also encompass some direct CEST effects from exchangeable protons or water bound in the solid (Zhou et al., 2023). In general, the rNOE effect is determined by several parameters, including proton concentrations, exchange and cross-relaxation rates, temperature, and pH (Zhou et al., 2023).

Examples of CEST/rNOE/MT MRI contrast mechanisms.

The targeted irradiation of specific molecular components using an RF pulse elicits a reduction in water signal intensity through a multitude of intricate mechanisms, including: (A) CEST, (B) rNOE, (C) MTC, (D) ihMT, and (E) rNOE after binding. Reprinted with permission from Zhou et al. (2023). CEST: Chemical exchange saturation transfer; ihMT: inhomogeneous magnetization transfer; MRI: magnetic resonance imaging; MT: magnetization transfer; MTC: magnetization transfer contrast; RF: radiofrequency pulse; rNOE: relayed nuclear Overhauser effect.

Studies have indicated that the detection of rNOEs in MRI, without significant MTC interference, can be optimized by adjusting the tsat and B 1 (Huang et al., 2021a). This can be attained through the execution of a low-power steady-state pulsed CEST experiment and the subsequent removal of the DS contribution utilizing the LD analysis (Jones et al., 2013).

Creatine-weighted chemical exchange saturation transfer

Creatine is a metabolite derived from the amino acids glycine, arginine, and methionine, with exchangeable protons from its guanidium group. It is predominantly found in muscle fibers, with smaller concentrations present in the brain. Initially, creatine is detected using MRS. Guanidinium CEST was initially studied in Cr molecules (Sun et al., 2007, 2008). In early creatine-weighted CEST (CrCEST) studies, phantom pH was adjusted to slow down the chemical exchange rate for optimizing the detectability of creatine signals. In 2012, a high-spatial-resolution imaging technique for creatine was developed using CEST, known as CrCEST, allowing for improved spatial resolution and eliminating signal contamination from phosphocreatine (Haris et al., 2012). CrCEST provided a novel approach for measuring creatine changes in vivo under various pathological conditions. The first application of CrCEST focused on muscle, mapping energetic changes during exercise with high spatial resolution (Kogan et al., 2014).

More recently, researchers discovered that guanidinium CEST signals are also present in a variety of proteins and brain Z-spectra at 2 ppm. However, these signals overlap with CrCEST and phosphocreatine (PCr) CEST signals (Jin et al., 2017; Zhang et al., 2017b; Chen et al., 2019a). To distinguish between GuanCEST from proteins and creatine, the CEST signal at 2 ppm associated with creatine is referred to as CrCEST, while the protein-derived GuanCEST is termed ArgCEST because only arginine (Arg) residues in proteins contain guanidine groups. Although free arginine amino acids also contribute to GuanCEST, their effect is negligible due to the extremely low concentration of free arginine in tissues (< 1 mM) compared with the much higher concentrations of arginine residues in mobile proteins. ArgCEST and CrCEST, along with PCrCEST, are among the few CEST contrasts that can be reliably extracted from the crowded in vivo Z-spectrum (Xu et al., 2023a). It should be noted, however, that at 2 ppm, other CEST and rNOE contrasts may also be present, such as aromaticNOE, amineCEST, and even hydroxylCEST. These overlapping signals underscore the complexity of interpreting Z-spectra in vivo .

Myo-inositol-weighted chemical exchange saturation transfer

Among endogenous molecules, myo-inositol also exhibits a detectable CEST signal, referred to as myo-inositol-weighted CEST (MICEST). Myo-inositol is a carbocyclic sugar abundantly found in the brain and is one of the metabolites commonly observed using MRS (Jansen et al., 2006). It plays a crucial role in cellular processes as a precursor for secondary messengers in eukaryotic cells, such as inositol phosphate, phosphatidylinositol, and phosphatidylinositol phosphate (Kim et al., 2024). These molecules are essential components of lipid membranes, including the myelin sheath in the central nervous system. Myo-inositol is considered to be a glial marker, and an increase in its content is believed to represent glial proliferation. Altered myo-inositol concentrations have been associated with various brain disorders, including AD, multiple sclerosis (MS), and brain tumors (Albrecht et al., 2016; Bizzarri et al., 2016).

Myo-inositol was first studied using 1 H-MRS, usually assigned to the peak present at 3.56 ppm from water. In 2011, CEST imaging was introduced to visualize myo-inositol distribution with improved spatial resolution. A brain mapping study in healthy human volunteers demonstrated significantly higher MICEST signals in white matter compared with gray matter at 7 Tesla (Haris et al., 2011). High-resolution mapping of myo-inositol using MICEST has proven to be a valuable tool for monitoring regional changes in myo-inositol concentrations across different disease states. Unlike glucose-based CEST imaging, MICEST does not require the administration of exogenous contrast agents. Its non-invasive nature makes myo-inositol mapping a promising biomarker for disease diagnosis and progression monitoring.

Glycine-weighted chemical exchange saturation transfer

Glycine is an important inhibitory neurotransmitter, with its concentration in the brain known to fluctuate in neuropsychiatric disorders. Mapping regional glycine concentrations and their distribution in the brain can offer valuable insights into the pathogenesis of neurodegenerative diseases and assist in disease staging. Glycine resonates at 2.85 ppm relative to the water peak. A previous GluCEST study has also measured the exchange rates of various metabolites, including glycine (Wermter et al., 2015). Glycine exhibits a high chemical exchange rate with free water, thus, high saturation power in a short pulse time is required to image glycine-weighted CEST (GlyCEST). Imaging GlyCEST requires high saturation power delivered in short pulse durations. However, these high-power saturation pulses can increase the SAR, imposing significant demands on the MRI system and potentially leading to system limitations. The GlyCEST effect is pH-dependent and becomes more pronounced under acidic conditions. Additionally, contributions from other brain metabolites, magnetization transfer effects, and rNOE may influence and modulate GlyCEST signals, adding complexity to the interpretation of results (Ohno et al., 2021).

Chemical Exchange Saturation Transfer Magnetic Resonance Imaging in Neurodegenerative Diseases

Applications in Alzheimer’s disease

AD is a progressive neurological disorder that gradually impairs memory, thinking, and behavior. It is the most common cause of dementia, accounting for 60% to 70% of the cases, affecting approximately 55 million populations around the globe, according to the World Health Organization (World Health Organization, 2021, 2022). The burden on caregivers and the public health sector is enormous, with the expenses related to dementia projected to more than double, increasing from $1.3 trillion annually in 2019 to $2.8 trillion by 2030 (Long et al., 2023). More than 95% of Alzheimer’s cases develop after the age of 65, known as late-onset AD, supporting that the rising onset of AD is correlated with increased human longevity (Long and Holtzman, 2019).

Early diagnosis of AD remains challenging, yet successful identification and timely intervention improve patient outcomes. A preclinical stage as verified by various longitudinal studies, occurring 10 to 20 years before clinical symptoms, is marked by the initial buildup of amyloid-beta in the precuneus and other cortical regions. The following cognitive impairment presents subtle symptoms that do not interfere significantly with daily activities (Dunne et al., 2020), and it is recognized as a risk state that may eventually lead to dementia (Rasmussen and Langerman, 2019). Advanced Alzheimer’s dementia can eventually lead to complete dependency on caregivers, placing a substantial burden on public resources and affecting the patient’s quality of life and life expectancy (Kim et al., 2021). Several pathological criteria have been adopted for diagnosis, including imaging biomarkers, such as MRI and PET scans (Kim et al., 2022a; Veitch et al., 2022), and fluid biomarkers from cerebrospinal fluid and blood (Zetterberg and Blennow, 2021). Early identification of Alzheimer’s disease at the preclinical and mild cognitive impairment (MCI) stages allows for improved therapeutic outcomes by interventions to mitigate or halt further progression (Guest et al., 2020; Zhang et al., 2021).

The pathological cascade in AD starts years before clinical symptoms appear (Knopman et al., 2021), and knowledge of these early altered metabolism guides CEST biomarker design. The primary neuropathological criteria for AD diagnosis include the deposition of extracellular β-amyloid, or Aβ, as neuritic plaques and the intracellular accumulation of hyperphosphorylated tau proteins as neurofibrillary tangles (NFTs) (Self and Holtzman, 2023), and is the primary target for APT/rNOE imaging in CEST MRI. These formations are accompanied by neuroinflammation, synaptic homeostasis disruption and dysfunction, and neuronal apoptosis, reflected by alterations in neurotransmitter concentration and metabolites such as glutamate, creatine, glycine, and myo-inositol, detectable by GluCEST, CrCEST, glyCEST, and MICEST, respectively. In addition to the pathological features mentioned above, AD is also characterized by a decreased number of glucose transporters at the blood-brain barrier (Kyrtata et al., 2021). Reduced glucose uptake in the hippocampus and cortical regions, as will be discussed later in glucoCEST studies, has been noted in individuals genetically predisposed to AD or those with mild or no cognitive impairment, who eventually progress to the disease, revealing valuable supporting evidence for brain atrophy, neuronal dysfunction and early diagnosis of AD (Ossenkoppele et al., 2013; Protas et al., 2013).

Amide proton transfer

As AD is characterized by protein accumulation and loss of neurons, APT has intuitively been one of the first to be used for imaging these pathologies in vivo . Amyloid and tau pathology detection using APTw MRI provides an alternative imaging solution that can integrated into clinical MRI routine well, with lower cost, wider availability, and no ionizing radiation compared with PET and single-photon emission computed tomography. Endogenous mobile proteins can be detected using APTw MRI in vivo and have been adapted in research and clinical practice for tumors and strokes on clinical 3T MRI scanners. However, for AD, the detection specificity of APTw MRI is inherently lower compared with a molecular-specific PET scan, due to its non-specific detection of amide protons in various protein and peptide species. APTw MRI studies so far attempt to detect altered proteostasis in general, assuming major contribution from Aβ pathology and tauopathy, as a supplementary to other clinical diagnosis criteria.

In animal studies, AD rodent models were engineered to mimic part of the pathology, such as rTg4510 mouse for tauopathy (Santacruz et al., 2005) and APPswe/PS1ΔE9 mouse for amyloid deposition (Borchelt et al., 1997). Declined APTw signals were detected on rTg4510 mice, in the cortex and hippocampus region but not the thalamus (Wells et al., 2015); recovery of the signals was observed after treatment with anti-tau drugs in a follow-up study from the same group (Holmes et al., 2016). The signal decrease in AD animal models can be partly explained by the fact that the exchange rate between amide protons and water protons was diminished when aggregation of soluble amyloid and tau proteins into amyloid plaques and NFTs occurred. Lowered pH values caused by AD brain acidosis can also attenuate the chemical exchange of amide protons.

Early diagnosis of AD for humans focuses on the detection of pathological characteristics at early AD stages such as mild cognitive impairment. However, APTw signal intensity change observed in human AD patients differs from those observed in preclinical rodent models. A human study has shown enhanced APTw signal intensities in the hippocampus and amygdala regions for amnestic MCI (aMCI) patients (Guo et al., 2021), and APTw value increases are also observed in various brain regions such as the gray and white matter in the temporal and occipital lobes, white matter in the putamen, thalamus, frontal lobe and pons (Zhang et al., 2020b). For patients with symptomatic dementia due to AD, APTw signal increases were also observed in the bilateral hippocampus (Wang et al., 2015), anterior cingulate, and putamen (Oh et al., 2021).

Various follow-up studies improve the APTw imaging for AD detection, in terms of acquisition time, sensitivity, and advancement of post-processing. Accelerated acquisition by using a saturation transfer difference ΔST(±3.6) map, together with radial sampling and 2D-UTE acquisition module was proposed by Chen et al. (2019b). The low B1 feature of the proposed scheme minimized the effects of direct water saturation and the semi-solid MTC contamination to the APTw signal. Significantly decreased −3.6 ppm but not +3.6 ppm signal on cortex, corpus callosum and thalamus was observed on APPswe/PS1ΔE9 mouse; the author credited the decrease to the effect of protein aggregation involved in AD based on comparison with phantom experiments. Wang et al. (2019a) demonstrated APTw imaging using saturation with frequency alternating RF irradiation (APT SAFARI ) with improved image uniformity and contrast, and also a better correlation between APTw signal and behavioral or pathology scores in AD model rats and normal control. Huang et al. (2022) applied deep learning to enhance CEST MRI (deepCEST) and apparent exchange-dependent relaxation (deepAREX) for imaging AD (5XFAD) in mice on a 3T system. This study differentiated significantly lower amide-weighted signals (3.5 ppm) using accelerated deep learning Z-spectrum processing associated with amyloid β-peptide deposits in both central and anterior brain slices of AD mice, as validated by immunohistochemistry. The optimized deepCEST/deepAREX techniques show promise for differentiating AD pathologies at 3T MRI using low B 1 power with fast processing time without compromising accuracy, expecting clinical translation on humans.

Conclusive evidence is yet to be verified for both limited animal and human studies so far, and whether these animal models faithfully mimic pathological conditions in human AD patients is still under debate. Multiple factors could be the candidate for inconsistency of APTw signal alterations between human AD patients and AD rodent models. Zhuang et al. (2023) summarized that besides Aβ pathology and tauopathy expressed or injected in mouse models, unexplored AD pathophysiological pathways in human patients involving various protein species and altered pH environment could contribute to APT signals inconsistency with animal results via MT, rNOE and other spillover mechanisms (Yu et al., 2018). Other possible explanations include MT, rNOE and spillover effects interfering with MTR asym methods in APTw and T 1 w contamination.

Exogenous CEST agents for APTw detection of AD pathology such as Angiopep-2 provided an alternative approach to track amyloid pathology. Angiopep-2 is an artificial peptide that binds to Aβ with high affinity in both human AD brain and rodent models as evidenced by post-mortem brain staining (Wang et al., 2017), and its in vivo visibility under APTw images was verified in APP/PS1 mice after intravenous injection of angiopep-2 which can penetrate through the blood-brain barrier, and increased signal was observed in the cortex and hippocampus region in the mice, suggesting potential for clinical AD diagnosis and Aβ-targeted treatment evaluation (Wang et al., 2019b). A following longitudinal study demonstrated its dynamic monitoring ability of amyloid deposition on APPswe/PSEN1dE9 mice, showing elevated APTw signal as the mice age (Xu et al., 2023b).

Glutamate-weighted chemical exchange saturation transfer

GluCEST imaging has demonstrated the ability to detect early changes in glutamate metabolism disrupted in AD rodent models. GluCEST detection in APP/PS1 transgenic mice revealed diminished contrast throughout the brain, including the hippocampus ( Figure 6 ), a region particularly susceptible to glutamate-related pathology (Haris et al., 2013b). This reduction of glutamate correlated well with glutamate and the glutamate-total creatine (Glu/tCr) ratio measured by proton magnetic resonance spectroscopy ( 1 H-MRS). In the P301S transgenic mouse PS19 line that overexpresses the human P301S mutant tau protein, a reduction in GluCEST contrast was detected within the Cornu Ammonis sub-region of the hippocampus (with synaptic loss) but not dentate gyrus subregion of the hippocampus (without synaptic loss) (Crescenzi et al., 2017). It is worth noting that GluCEST signal reduction and synaptic loss are detected before tau tangles occur, suggesting early diagnosis capability of GluCEST for AD (Yoshiyama et al., 2007; Crescenzi et al., 2014). Igarashi et al. (2020) further demonstrated that the GluCEST signal diminished before the onset of hypoperfusion in specific areas of the cortex and hippocampus in a 5×FAD mouse model of AD by comparing the GluCEST results with regional cerebral blood flow (CBF) imaging results, a standard diagnostic biomarker for AD and mild cognitive impairment (Dai et al., 2009).

Brain GluCEST mapping across the hippocampus region.

(A) MRI brain image of a wild-type mouse. (B) The corresponding GluCEST map. (C, D) MRI image and corresponding GluCEST map from the age-matched AD mouse. Reprinted with permission from Haris et al. (2013b). AD: Alzheimer’s disease; GluCEST: glutamate-weighted CEST; MRI: magnetic resonance imaging.

Further human research is required to validate the efficacy of GluCEST in detecting AD-related glutamatergic alterations, as promising animal model outcomes require clinical confirmation. Notably, in the pathological progression of AD, the degeneration of glutamatergic neurons is highly associated with compromised glucose metabolism, given that the synthesis pathway of glutamate is dependent on glucose (Matthews et al., 2021). This suggests that glucoCEST and GluCEST could work in a complementary fashion, serving as a valuable toolkit in the diagnosis of AD (Orzyłowska and Oakden, 2022).

Glucose-weighted chemical exchange saturation transfer

Altered cerebral glucose uptake has been considered a hallmark of AD (Niwa et al., 2002), and glucose consumption is heavily linked with synaptic function (Stoessl, 2017). Studies have observed glucose hypermetabolism or hypometabolism in various AD mouse models at different stages (Luo et al., 2012; Poisnel et al., 2012; Tolomeo et al., 2018). These occur prior to noticeable neuronal loss and cerebral atrophy, which suggests detecting glucose metabolism alteration as an early diagnosis tool for AD patients. Currently, glucose metabolism detection is achieved by 18 F-fluorodeoxyglucose (FDG) PET, and reduced glucose uptake in the posterior cingulate and temporoparietal regions has been established as a common diagnostic criterion for AD (Wang et al., 2016). Again, repeated PET imaging for AD patients is not desired due to ionizing radiation exposure. With higher resolution and lower cost, GlucoCEST is a promising candidate to evaluate metabolism by revealing the effects of glucose uptake and metabolism (Chan et al., 2012).

GlucoCEST for assessing AD brain is still in its infancy; no human study has been conducted so far. Available preclinical studies on AD mouse models deliver D-glucose by intravenous or intraperitoneal injection, followed by continuous GlucoCEST monitoring to capture glucose accumulation and removal. In the APP/PS1 mouse model, GlucoCEST imaging revealed a slower rate of D-glucose uptake in the whole-brain parenchyma, specifically in the thalamus and entorhinal cortex of 6-month-old AD rodents compared with age-matched controls (Huang et al., 2020). This indicates impaired glucose transport and metabolism occurred even before significant amyloid plaque formation, suggesting the potential for early AD diagnosis. Furthermore, the reduced clearance of D-glucose in cerebrospinal fluid was observed, implying dysfunction of the glymphatic system, which is known to precede amyloid angiopathy and Aβ accumulation in the brain parenchyma. In the later stage (16 months) of AD, GlucoCEST signal intensities were diminished significantly, as with similar results obtained in 7–8-month Tau4RDK mice with Tauopathy (Chen et al., 2021b) and 20-month APP23 mice (Tolomeo et al., 2018). Chen et al. (2021c) investigated correlations between glucoCEST signal intensity and myoinositol, using rat models with an intracerebroventricular injection of Aβ 25–35 . GlucoCEST enhancement correlates with increased myoinositol levels in the hippocampus, implying that glucoCEST is responsive to neuroinflammatory processes. An earlier study using 9.5-month rTg4510 AD mice with tauopathy and glucoCEST imaging every 8 minutes within 100 minutes after glucose injection showed higher contrast only in the cortex (Wells et al., 2015), partly explained by elevated retention of glucose in the cortex due to suppressed glucose metabolism alongside an incrementing delivery by increased cerebral blood flow to the cortical regions.

Relayed nuclear Overhauser effect

The relayed nuclear Overhauser effect (rNOE) signal is mainly derived from aliphatic protons. Mobile macromolecules, including proteins, carbohydrates, and membrane lipids, contribute significantly to rNOE in the Z-spectrum, suggesting opportunities for probing in vivo macromolecules such as Aβ peptides, tau proteins, glycosaminoglycans (GAGs), and lipids in the AD brain (Zhou et al., 2023). As rNOE is sensitive to macromolecular content in the brain, these metrics can serve to assess disrupted cell membrane lipid integrity in the early stages of AD.

Early explorations of rNOE on APP swe ;PS1ΔE9 mice detected strong rNOE peaks in the full z-spectrum on an 11.7T scanner (Chen et al., 2019b). Although variations between APP mice and control groups were insignificant, it revealed the potential of protein aggregates of the AD brain to alter the background signal due to changes in rNOE contributions.

Swain et al. (2023) explored rNOE for early-stage AD pathology in APP NL-F knock-in mouse models. They employed a standard spoiled gradient-echo readout with saturation parameters sensitive to rNOE contrast, generating a comprehensive Z-spectrum from offsets ranging from –100 to +100 ppm, allowing for detailed analysis of multiple metabolites and macromolecules despite long scan time. Apart from significant rNOE MTR and rNOE changes in common AD detection targets in grey matter such as the hippocampus, the author observed a specific signal alteration in the entorhinal cortex, suggesting it as a potential biomarker for AD detection.

Creatine-weighted chemical exchange saturation transfer

The disruption of creatine metabolism, specifically the creatine kinase reaction (CKR), is a key aspect of AD pathology, measured by the phosphocreatine/creatine (PCr/Cr) ratio. Phosphorus-31 magnetic resonance spectroscopy (31P MRS) has been the established method for assessing CKR function; however, low sensitivity, low resolution, and the need for specialized equipment limit its application. CrCEST MRI offers possibilities for high-resolution creatine mapping. Due to the pH-dependent exchange rate of creatine guanidinium protons, it is sensitive to both creatine concentration and pH. This sensitivity makes CrCEST a valuable tool for understanding cerebral pH regulation, which is vital for cell viability and neuronal function in AD.

Chen et al. (2021a) evaluated the feasibility of CrCEST MRI for early AD detection using the amyloid precursor protein/presenilin-1 (APP) and tau (Tau4RΔK) mouse models. The research examined CrCEST signal variations influenced by creatine concentration and pH, validated by MRS-based methods and hypercapnia experiments. Results indicated that reduced CrCEST contrast effectively differentiated between AD and wild-type mice at an early age (6–7 months), primarily due to changes in cerebral pH rather than creatine concentration.

Shahid et al. (2024) combined CrCEST with GluCEST on the ARTE10 mouse model of amyloid pathology. Z-spectra were acquired using a 9.4 Tesla small animal MRI system with two radiofrequency saturation amplitudes (1.47 μT for creatine and 5.9 μT for glutamate) to achieve distinct CEST contrasts. MRS results showed a significant decrease in glutamate and the glutamate-total creatine (Glu/tCr) ratio, along with increased levels of total creatine, phosphocreatine, and glutathione in ARTE10 mice compared with wild-type. In the same regions of interest, Glu-AREX and Cr-AREX were positively associated with the Glu/tCr ratio, indicating the involvement of neurotransmitter metabolites and energy metabolism in Aβ-mediated synaptic degradation in the hippocampus. This multi-contrast study demonstrates the potential of CEST and MRS techniques to track the heterogeneous pathophysiology of AD in vivo .

Myo-inositol-weighted chemical exchange saturation transfer

Myo-inositol is a marker of glial cell proliferation, often observed in neuroinflammation responses, and has been shown to increase in early AD pathology (Haris et al., 2011; Murray et al., 2014). Haris et al. (2013a) used MICEST to detect Alzheimer’s pathology in 20-month-old APP-PS1 mice with a 9.4T scanner. The APP-PS1 mice exhibited higher MICEST contrast compared with the wild-type, which aligns with increased myo-inositol/creatine ratios observed in 1 H MRS. Immunohistochemistry confirmed elevated glial cell proliferation in APP-PS1 mice.

A more recent study (Yanez Lopez et al., 2019) explored MICEST to assess microglial activation induced by lipopolysaccharide (LPS) in young wild-type and AD mice using an Agilent 9.4T scanner. Results suggest that CEST biomarkers could differentiate inflammation activation levels in wild-type (WT) and AD mice validated by MRS. However, LPS-induced responses were highly variable and did not consistently align with genotype, and the sensitivity of CEST could not be explained by changes in myoinositol levels measured by MR spectroscopy. Further research is required to validate the feasibility of MICEST for detecting neuroinflammatory responses in AD mouse models.

Glycine-weighted chemical exchange saturation transfer

Glycine is an important inhibitory neurotransmitter, and its concentration in the brain is known to change in neuropsychiatric disorders. In AD, both blood and brain glycine levels differ from those in healthy individuals. Measuring regional glycine concentrations and mapping their distribution in the brain can provide insights into the pathogenesis of AD and aid in disease staging.

A pilot study by Ohno et al. (2021) explored the feasibility of GlyCEST imaging to map glycine concentration distributions in the murine brain. The study demonstrated that GlyCEST can quantitatively measure glycine levels, showing good agreement with biochemical measurements. In C57BL/6 mice, glycine concentrations were higher in the thalamus than in the cortex. In the 5×FAD AD mouse model, GlyCEST indicated lower glycine levels in the cerebral cortex and thalamus compared with wild-type mice, but not in the hippocampus. Although the field of GlyCEST is in its infancy, GlyCEST shows potential for further research in AD.

Applications in Parkinson’s disease

PD is a progressive neurodegenerative disease primarily characterized by the degeneration of dopaminergic neurons in the substantia nigra pars compacta, a midbrain structure that is pivotal in dopamine production (Poewe et al., 2017). The neuronal degeneration in this region disrupts the basal ganglia circuitry, which is vital for motor control, and it may lead to a significant decrease in dopamine, a neurotransmitter essential for movement regulation (Alexander, 1994). Two hallmarks of PD are the accumulation of α-synuclein and the loss of dopaminergic neurons in the substantia nigra. Intracellular Lewy bodies formed by α-synuclein aggregates can spread in a prion-like manner from the brainstem, limbic system and temporal paralimbic cortex to the prefrontal and sensory-association cortex, and are central to the pathology of PD (Braak et al., 2004; Barba et al., 2022). These possibly neurotoxic α-synuclein aggregates and oligomers in neural cells are closely associated with synaptic dysfunction. Along with complex interactions among mitochondrial dysfunction, lysosomal and vesicle transport issues, synaptic transport problems, and neuroinflammation, these mechanisms primarily accelerate dopaminergic neuron apoptosis, though various other motor and non-motor circuits are affected.

PD is a prevalent condition affecting over 10 million globally (Parkinson’s Foundation, 2024). It is presumably the fastest-growing neurological condition worldwide, with both its incidence and prevalence surging over the past two decades (Dorsey et al., 2018). Several factors contribute to this rise, including an aging population, improved diagnostic strategies, prolonged survival rates, and environmental pollution (De Miranda et al., 2022). Typical symptoms of PD include motor features such as tremor, rigidity, and slowness, and non-motor ones like cognitive decline, depression, and pain (Sveinbjornsdottir, 2016). Apart from significant personal impacts with accumulating disability and a slow progression that can extend for decades, it also imposes considerable stress on caregivers and represents a growing socioeconomic burden on society (2022), similar to other neurodegenerative diseases.

Diagnosis of PD is a challenging task for clinicians (Tolosa et al., 2021). It exhibits marked heterogeneity in its presentation and causes. Moreover, alternative underlying pathologies can produce similar appearing clinical manifestations, referred to as Parkinsonian syndromes (Srivanitchapoom et al., 2018). Except for genetic testing in selected cases, a definitive diagnosis can only be established through post-mortem identification of hallmark neuropathological changes in the brain (Bloem et al., 2021). The early stages of PD are often challenging for clinical practitioners to identify, which is evident from the average delay of about 10 years between the first noticeable symptoms before a diagnosis (Gaenslen et al., 2011). Detecting PD early allows for timely disease-modifying intervention, potentially enhancing patient outcomes (Noyce et al., 2016).

There has been a need in the radiology field to develop reliable imaging biomarkers for diagnosing and evaluating the severity of PD. Existing molecular imaging methods such as single-photon emission computed tomography, and PET can detect objective alterations in the neurochemical machinery of the brain, primarily by radioactively labeled molecules involved in the pathophysiology pathway. Some examples that already passed FDA approval and are used in diagnosis are [ 18 F]F-DOPA, precursors in dopamine synthesis, and [ 123 I]FP-β-CIT (DatScan®) as dopamine transporters (Bidesi et al., 2021). Again, these radiotracer-based imaging methods face substantial hurdles for routine checkups and repetitive scans for the cost and ionizing radiation exposure.

Despite various MR-based imaging techniques developed for PD research, such as MR spectroscopy functional MR imaging, diffusion tensor imaging, and susceptibility-weighted imaging (for excessive accumulation of iron), primarily focused on substantia nigra (Bae et al., 2021), clinically reliable in vivo MR biomarkers for PD diagnosis and staging have yet to be investigated (Ryman and Poston, 2020). Over the past decade, research on PD assessment using CEST MRI has advanced in both animal models and in patient studies, using APTw imaging to reflect protein and cellularity changes in the substantia nigra, and extending to GluCEST attempts to reveal glucose metabolism.

Amide proton transfer

Li et al. (2014) were pioneers in demonstrating the potential of APTw/CEST imaging for detecting PD at a clinical 3T scanner on human patients. Their findings revealed that signal intensities in the substantia nigra of both brain hemispheres were lower in PD patients compared with healthy controls. Subsequent studies by the same group confirmed that when PD patients were grouped according to the Hoehn and Yahr scale or disease duration ( Figure 7 ), signal intensities in the substantia nigra notably declined as PD progress (Li et al., 2015, 2017). For PD patients exhibiting symptoms on one side, the APTw or total CEST intensities in the substantia nigra on the affected side were notably lower compared with healthy controls. The reduced MTR asym spectrum in the substantia nigra across a broad offset range suggests these changes might result from dopaminergic neuron loss, affecting all water-exchanging chemicals (Li et al., 2023). For other regions showing signal alterations, the globus pallidus, putamen, and caudate exhibited rising APTw and total CEST signal intensities in PD patients, with peak signal changes in early disease stages, which then slightly declined in advanced stages. The elevated APTw signals in these regions could be linked to increased cytosolic proteins and peptides, such as misfolded alpha-synuclein and neuroinflammation-related tau proteins (Pan et al., 2021). These findings suggest that APTw/CEST imaging could be valuable for diagnosing and monitoring PD progression.

APTw images of PD patients with different stages.

(A) A normal control male (yellow appearance in SN). (B) An early-stage PD appearance (green). (C) A late-stage PD appearance (light blue). Reprinted with permission from Li et al. (2017). APTw: Amide proton transfer-weighted; PD: Parkinson’s disease; SN: substantia nigra.

Recent studies use multipool CEST MRI for integrated analysis using APTw and other pools. In a study by Mennecke et al. (2023) conducted at 7T on human participants and PD patients, researchers proposed an optimized post-processing pipeline for CEST multipool evaluation to enhance reproducibility. Using a five-pool Lorentzian fit model, the researchers aimed to improve the reproducibility of the amide pool, Lorentzian fitting at 3.5 ppm, and aliphatic rNOE amplitude maps by refining individual post-processing steps. These advancements allowed for the detection of subtle changes in CEST MRI contrast, revealing age-related effects in healthy subjects and alterations due to neurodegeneration in early-stage PD patients. The method successfully identified subtle amide pool changes associated with aging and detected decreased aliphatic rNOE in early-stage PD, providing valuable insights into the disease’s progression. Chen et al. (2024) accelerated multipool CEST MRI for PD by employing a deep learning and compressed sensing approach. A deep learning network based on a modified U-Net was proposed to recover dense Z-spectra from sparse ones. In vivo experiments using a rat PD model induced by 6-hydroxydopamine demonstrated that this method could achieve superior fidelity in retrieving dense Z-spectra compared with existing techniques (Simola et al., 2007). Significant differences were observed in APT, CEST at 2 ppm, nuclear Overhauser enhancement, and direct saturation between the striatum regions of wild-type and PD models.

Tian et al. (2023) explored APT imaging in PD patients using a 3T clinical scanner and focused on cortical changes rather than substantia nigra. The findings suggest that cortical changes in APT-related parameters could serve as imaging biomarkers for non-invasive PD diagnosis. Although no correlations were found between these signals and clinical assessments, 3D-APT imaging identified significant changes in the temporal cortex, a known site of early Lewy pathology. In particular, MTR Rex values showed notable differences in the temporal grey matter between PD patients and controls.

Glutamate-weighted chemical exchange saturation transfer

Glutamate metabolism plays a crucial role in the pathophysiology of PD. Excessive glutamate and its induced excitotoxicity are one of the key candidates for the degeneration of dopaminergic neurons and concomitant motor dysfunction in PD patients (Wang et al., 2020). For the dopamine-deficient Parkinsonian mouse model induced by MPTP, elevated levels of glutamate in the striatum and motor cortex are also reported (Bagga et al., 2013; Chassain et al., 2013).

Evidence from GluCEST imaging by Bagga et al. (2016) on a dopamine-deficient Parkinsonian mouse model induced by MPTP aligned with the abovementioned observations. The study revealed elevated glutamate levels in the striatum and motor cortex, supported by 1 H-MRs and immunohistochemical analysis. A follow-up study from the same group confirmed the reliable repeatability and specificity of GluCEST in detecting striatal glutamate changes in the MPTP model (Bagga et al., 2018). It demonstrated low variability and a strong association with 1 H-MRS findings. Notably, it highlighted that a portion of the GluCEST signal comes from sources other than glutamate which were undetectable by MRS, aligning with theoretical estimates from the 2016 study.

Applications in multiple sclerosis

MS is a complex chronic autoimmune disease that leads to damage to the central nervous system, and it is also the primary cause of non-traumatic disability among young adults (Browne et al., 2014). The likelihood of women developing MS is more than three times that of men, which may be attributed to the combined effects of female hormones and other factors (Hunter, 2016). The specific etiology of MS remains unclear; however, it is established that genetic and environmental factors play a significant role in the disease’s pathogenesis (Chandran et al., 2008).

Based on disease progression and the severity of symptoms, MS is primarily classified into four types: (i) Relapsing-remitting MS, the most common subtype, characterized by intermittent relapses and the emergence of new or more severe symptoms. (ii) Secondary progressive MS, in which most relapsing-remitting MS patients eventually develop over time, marked by a gradual worsening of symptoms without acute relapses. (iii) Primary progressive MS is an uncommon subtype in which symptoms gradually worsen from the onset of the disease without periodic relapses. (iv) Progressive-relapsing MS, the rarest type with steady disease progression from the beginning and occasional relapses during the disease (Inglese and Bester, 2010; Loma and Heyman, 2011).

Due to the relapsing-remitting characteristic of MS, there may be overlap between different stages of the disease, making classification challenging; but it can generally be made based on the presence of inflammation or demyelination (Van der Valk and De Groot, 2000).

The pathological features of MS include demyelination, axonal damage, inflammation, and varying degrees of remyelination, which collectively impair communication between the brain and the body (Chandran et al., 2008). In the early stages of MS, lymphocytes and macrophages in the body actively express adhesion molecules, which facilitate the destruction of the blood–brain barrier (Balasa et al., 2021). Auto-reactive lymphocytes that cross the blood-brain barrier also increase significantly, infiltrating the central nervous system and causing damage to normal-appearing white matter (NAWM) (Van Langelaar et al., 2020). The dehydrated myelin is composed of lipids (75%–80%) and proteins (20%–25%) (Min et al., 2009), and it will be attacked by lymphocytes, leading to the loss of oligodendrocytes, which play an essential role in the formation and maintenance of myelin. Although the process of myelin regeneration still occurs in the early stages of MS, it becomes increasingly difficult as the disease progresses, resulting in permanent lesions or scars in the brain and spinal cord that cannot be reversed (Wu and Alvarez, 2011).

MS can lead to a variety of symptoms, including motor impairments, cognitive impairment, and neurological dysfunction, which can occur in all subtypes of MS (Goverover et al., 2016). For example, the most common initial symptom of relapsing-remitting MS is unilateral optic neuritis, which can result in changes in color perception and vision loss; inflammatory lesions in the spinal cord may lead to myelitis (Ford, 2020); and problems with memory, attention, or other abilities. Currently, various assessment methods are available to quantify the severity of MS symptoms, such as the expanded disability status scale, which is the most commonly used tool for evaluating MS and was proposed in 1983 by neurologist John Kurtzke (Kurtzke, 1983; Barkhof, 1999). However, the expanded disability status scale lacks sensitivity in assessing cognitive function. To monitor it, the brief international cognitive assessment in MS is also widely used in clinical practice (Langdon et al., 2012). Additionally, there are various diagnostic methods, including clinical observation, cerebrospinal fluid analysis, and MRI examination, but there are currently no effective treatments available. The main purpose of clinical intervention is to slow down the disease progression. Notably, previous studies have confirmed the presence of axonal damage in the early stages of MS, indicating that the nervous system may already have sustained permanent damage before the onset of visible symptoms, which further underscores the importance of timely diagnosis and intervention (De Stefano et al., 2001; Filippi et al., 2003).

MRI is one of the effective tools for the diagnosis of MS. Traditional MRI sequences, including T 1 -weighted, T 2 -weighted, and contrast-enhanced imaging, are capable of revealing the morphology and activity of MS lesions (Oghabian et al., 2022). However, these diagnostic sequences typically rely on morphological changes in normal-appearing gray matter or NAWM, which often appear in the later stages of the disease. Fortunately, CEST MRI as an advanced molecular imaging technique offers a promising approach for early diagnosis and intervention by identifying pathological changes at the molecular level.

Amide proton transfer

Preliminary studies have been conducted on the application of APTw MRI technology in patients with MS. A human study analyzed the APT asymmetry (APT asym ) in four MS patients and ten healthy controls (Dula et al., 2011). The results revealed an increase in APT asym observed in certain MS lesions compared with healthy tissue, and significant differences in APT signals between white matter and gray matter in the human brain were also observed. In addition to the brain, APTw can also be applied to diagnose the spinal cord of MS patients. According to By et al. (2018), significant differences in APTw signals were noted in NAWM after respiratory correction between MS patients and the healthy control group.

The study by Lee et al. (2021) evaluates the use of APTw 7T MRI for monitoring myelination changes in a cuprizone-induced rat model of demyelination and remyelination, focusing on the corpus callosum. It demonstrates that APTw MRI effectively detects variations in amide proton concentrations during demyelination and remyelination. The demyelination group showed significantly higher APTw signals than controls, while the remyelination group had signals similar to controls but lower than the demyelination group. This indicates APTw MRI’s potential as a non-invasive biomarker for early detection and monitoring of myelination changes in conditions like MS.

Relayed nuclear Overhauser effect

In a study published by Huang et al. (2021b), relayed nuclear Overhauser effect weighted (rNOEw) imaging was used on the whole brains of MS patients, including those with neuromyelitis optica spectrum disorders (NMOSD), a disease that presents with clinically similar neurologic symptoms to MS. Compared with the normal control group and the NMOSD group, MS patients exhibited a significantly decreased rNOEw contrast ( Figure 8 ), with notable differences observed in both white matter and gray matter. Additionally, compared with NMOSD lesions, MS lesions were more numerous and had larger size but exhibited a lower rNOEw contrast, indicating that rNOEw imaging can effectively differentiate between MS, NMOSD, and normal control. This technique may be useful in the detection of early myelin and cellular damage, aiding in the development of accurate treatment plans.

Comparison of rNOEw imaging contrast in three groups of experimental subjects.

A notable attenuation of rNOEw contrast was discerned in MS patients, exhibiting a distinct contrast from both NMOSD cohorts and NC. Reprinted with permission from Huang et al. (2021b). MS: Multiple sclerosis; NC: normal control; NMOSD: neuromyelitis optica spectrum disorder; rNOEw: relayed nuclear Overhauser effect weighted.

A study by Chen et al. (2023b) investigates the application of CEST MRI to detect myelin changes in a cuprizone mouse model at 3T, focusing on rNOE and amide contrasts. The research reveals that rNOE contrast at –3.5 ppm significantly decreases during demyelination and recovers during remyelination, while amide contrast at 3.5 ppm decreases during demyelination but remains low during remyelination. These findings are validated by immunohistochemistry, showing a strong correlation with myelin content. The results highlight the potential of CEST MRI as a sensitive tool for monitoring myelin pathology, with implications for diagnosing and understanding MS.

Glutamate-weighted chemical exchange saturation transfer

Glutamate is the primary free amino acid in the central nervous system and is essential for normal brain function (Gasiorowska et al., 2021). In MS, excitotoxicity induced by glutamate leads to apoptosis of oligodendrocytes and neurons (Vercellino et al., 2007); pathology of MS can be partly mimicked in animal models. In a study using a rat model of demyelinated MS, GluCEST imaging was applied with a pre-clinical 7T MRI system to investigate glutamate signal changes (Lee et al., 2020). The experiment used lysophosphatidylcholine-induced MS rats and control groups. Lysophosphatidylcholine exposure, which targets myelinating oligodendrocytes while preserving neuronal axons, was used to induce focal demyelination in brain regions such as the striatum, optic chiasm, spinal cord, corpus callosum, and hippocampus (Burrows et al., 2019). GluCEST-weighted signals were analyzed using magnetization transfer ratio asymmetry at 3.0 ppm in the corpus callosum and hippocampus. Results indicated increased GluCEST-weighted signals in the lysophosphatidylcholine-induced demyelination rat brains compared with controls, suggesting alterations in glutamate levels associated with demyelination.

A research team conducted two clinical studies using GluCEST on MS patients. The first study utilized a conventional asymmetry analysis of GluCEST, exhibiting an elevated GluCEST contrast on 3.0 ppm in the prefrontal cortex of MS patients compared with healthy counterparts (O’Grady et al., 2019). Notably, this study also documented a negative correlation between GluCEST in cortical areas and cognitive performance indices, suggesting a potential link between glutamate metabolism and cognitive decline. However, this study employed the conventional GluCEST quantification method without applying MT removal or T 1 correction. Refined methodology incorporating rigorous corrections for confounding factors such as MT, spillover effects, and T 1 relaxation, achieved through a Lorentzian fitting algorithm in conjunction with the AREX approach were integrated in the following study (O’Grady et al., 2022). A notable increase in the corrected AREX (AREX Corr ) within white matter lesions was observed, compared with both NAWM and healthy white matter controls. These findings suggest the transformative potential of glutamate-weighted CEST at high field strengths as a metabolic imaging modality.

Other chemical exchange saturation transfer contrasts

CEST MRI signals at offsets of 1 and 2 ppm were investigated as potential biomarkers for early pathological changes, in a study using the experimental autoimmune encephalomyelitis mouse model of MS (Liu et al., 2019b). These signals, analyzed through a histogram-guided method, revealed a more heterogeneous distribution with lower peak heights in the hindbrain of experimental autoimmune encephalomyelitis mice compared with the control group. The author claims that the CEST contrast at 1 ppm is sensitive to exchangeable hydroxyls in myo-inositol, choline, glucose, and cell-surface glycoproteins, as well as pH changes. Meanwhile, the 2 ppm offset is typically attributed to creatine and guanidinium amines in proteins and peptides, with signals decreasing at acidic pH, which may occur in inflammatory environments like the experimental autoimmune encephalomyelitis mouse brain. This suggests that CEST MRI at these offsets can serve as novel imaging biomarkers of MS, complementing existing functional and physiological MRI measures.

Applications in Huntington’s disease

Huntington’s disease (HD) is an inherited neurodegenerative disorder characterized by progressive motor dysfunction, cognitive decline, and psychiatric disturbances. It is caused by an abnormal CAG trinucleotide repeat expansion in the Huntingtin gene, leading to the production of a toxic mutant Huntingtin protein with an expanded polyglutamine tract. This results in protein aggregation and altered interaction with other proteins. HD is dominantly inherited and typically affects about 1 in 10,000 individuals in Europe and the US, with approximately 40,000 patients in each region, which is considered to be a rare disease. The disease is fatal within 10–15 years after symptom onset, which usually occurs in adulthood, although juvenile cases with larger repeat expansions progress more rapidly.

Clinically, HD is marked by involuntary movements such as chorea and dystonia, cognitive deficits including impaired strategy and planning, and psychiatric symptoms such as depression and irritability. Neuroimaging has been instrumental in understanding HD’s progression; striatal atrophy observed via MRI is currently a key biomarker, while PET scans have shown reduced glucose metabolism in affected brain regions. Despite these insights, existing methods provide limited information on the underlying biological mechanisms of the disease.

Advancements in MRS have revealed changes in brain metabolites, such as decreased glutamate and N-acetyl-aspartate in HD patients and models. However, MRS is typically restricted to single-voxel measurements. CEST imaging, particularly GluCEST, offers a promising non-invasive technique to map glutamate distribution with high anatomical resolution. Studies have demonstrated the potential of GluCEST to detect local alterations in glutamate levels in both rodent and human brains, which could indicate regional metabolic defects and provide further insights into HD pathophysiology.

Glutamate-weighted chemical exchange saturation transfer

Pépin et al. (2016) have employed GluCEST imaging to investigate HD in 12-month-old Ki140CAG mice, specifically using a high-field 11.7T scanner. The Ki140CAG mouse model is a knock-in model of HD containing 140 CAG repeats inserted in the murine huntingtin gene. It is characterized by the slowly progressive appearance of the symptoms and is considered to mimic more closely the clinical form of HD than severe and rapid mouse models (Menalled et al., 2003). Reduced glutamate levels across various brain regions were observed, with the most significant decrease observed in the corpus callosum. This reduction was more pronounced in homozygous mice, indicating faster disease progression. The study highlighted that the corpus callosum exhibited the largest GluCEST variations, suggesting early metabolic changes and potential compensatory mechanisms in brain energy metabolism. The author noted in the following communication that GluCEST may miss disease processes when glutamate levels are preserved (Pépin et al., 2020).

A subsequent longitudinal study from the same group assessed Ki140CAG mice over time using multiple imaging techniques including GluCEST (Pérot et al., 2022). It revealed early defects in the anterior corpus callosum at 5 months on diffusion MRI, preceding GluCEST changes at 8 and 12 months, which then spread to the frontal and piriform cortices and pallidum. Additionally, magnetization transfer imaging indicated reduced signals in the septum at 12 months.

Despite these promising findings, the requirement for high-field scanners limits clinical translation. Further research is needed to explore GluCEST’s potential in clinical settings and its ability to reveal early metabolic alterations in HD.

Discussion

This review has summarized the recent advancements in the application of CEST MRI for neurodegenerative diseases, including AD, PD, MS, and HD. CEST MRI is an emerging molecular imaging modality that provides high-resolution imaging of specific metabolites and molecular changes, offering the potential for early detection of neurodegenerative diseases. Its sensitivity allows the identification of biomarkers at early stages, which is crucial for diagnosing and monitoring disease progression.

Discussion of in vivo results

Despite various CEST contrasts for detecting neurodegenerative diseases, differences may vary from preclinical rodent models to human trials. Conclusive evidence is yet to be verified for both limited animal and human studies so far, and whether these animal models faithfully mimic pathological conditions in human AD patients is still under debate. As mentioned previously, there were inconsistencies in APTw signal changes between human AD patients and AD rodent models. In addition to Aβ pathology and tauopathy observed or induced in mouse models, unexplored AD pathophysiological pathways in human patients, such as those involving diverse protein species and altered pH environments, may contribute to inconsistencies in APT signals between animal studies and human results due to MT, rNOE, and other spillover effects.

The underlying pathological changes detected by CEST MRI are still not fully understood, particularly in the context of neurodegenerative diseases. For example, APTw and creatine signals are intuitively correlated with protein concentrations and guanidium concentration respectively, but other factors such as pH and temperature may also contribute to signal changes, by changing the exchange rates of the protons. In vivo conditions are inherently complex, and further work involving histopathology and postmortem analyses will be necessary to clarify the contributions in CEST signals and the underlying pathological indications.

CEST MRI has shown promising results in clinical studies for detecting neurodegenerative diseases and differentiating disease states. Studies on AD, PD, MS, and HD have demonstrated the potential of CEST MRI not only to distinguish patients from healthy controls but also to track disease progression proved by longitudinal studies on rodents. APTw imaging, the earliest developed and most extensively explored contrast, has been particularly useful in detecting proteinopathies, hallmarks of many neurodegenerative diseases. Given that these diseases often involve protein aggregation (e.g., amyloid plaques and tau fibrils in AD or alpha-synuclein in PD), APTw appears well-suited to detect these pathological changes. However, proteinopathies are not always the earliest markers of neurodegeneration. The capability of CEST MRI to detect impaired neurotransmitter metabolism, demyelination, neuroinflammation, and pH changes that often occur before overt protein aggregation, positions it as a valuable tool for early diagnosis. These early-stage markers are more challenging to detect due to the lower concentration, weaker separability in Z-spectrum, and complexity of the underlying biological processes, but they represent an important area for future research. Translation to clinical practice remains a challenge, and larger studies with broader patient populations are needed to validate CEST biomarkers for clinical application.

Chemical exchange saturation transfer magnetic resonance imaging contrasts in detecting neurodegenerative diseases

CEST MRI offers several advantages as a non-invasive, radiation-free technique, making it suitable for repetitive scans and longitudinal studies. Unlike other contrast agent-dependent MRI techniques and radiotracer-dependent PET, CEST MRI leverages endogenous contrast mechanisms, avoiding the potential toxicity of exogenous substances. Additionally, CEST MRI could be a cost-effective alternative by eliminating the need for radiotracers and contrast agents, which can be expensive and are not as widely available. However, one challenge is the preference for high-field MRI scanners, particularly for certain CEST contrasts such as GluCEST and CrCEST, which require ultra-high-field scanners (≥ 7T) to resolve peaks in the Z-spectrum that are closely spaced. Sensitivity and spectral resolution are significantly enhanced at 7T and higher fields, making them uniquely suited for certain CEST-based applications (Geethanath and Vaughan Jr, 2019). For APTw imaging and rNOE contrasts, more widely available 3T MRI scanners are often sufficient and have been increasingly used in studies aimed at clinical adaptations. MRI scanners with low field strengths, such as 1.5T, are currently more clinically available. CEST MRI at 1.5T shows insufficient Z-spectrum separability and near-zero asymmetry for in vivo CEST applications, despite optimization attempts on saturation power and pulse schemes, and limited in vivo studies have been reported so far in general. There are some publications demonstrated that the feasibility of in vivo 1.5T CEST MRI for tumor diagnosis (Chan et al., 2019, 2021a, b). To the best of our knowledge, there are no published studies on 1.5T CEST MRI for neurodegenerative diseases as of the time when we conducted the review.

Challenges in clinical translation

While AD and PD have dominated CEST MRI research, other neurodegenerative diseases such as HD, amyotrophic lateral sclerosis, and rarer conditions remain mostly underexplored. For instance, amyotrophic lateral sclerosis has only been investigated in a small pilot study involving five patients, showing reduced CEST signal intensity in the motor cortex, with no extensive in vivo studies available (Dai et al., 2019). These diseases, such as AD and PD, involve metabolic changes and proteinopathies that CEST MRI could potentially detect, yet they have received far less attention. Additionally, variations in CEST MRI signals across different ethnic groups have been noted. For example, higher APTw values were observed in the hippocampus and amygdala of healthy Europeans compared with Chinese subjects, highlighting the need for multi-ethnicity and multi-centered studies to analyze these differences (Guo et al., 2021).

CEST MRI has generated significant enthusiasm in neurodegenerative disease imaging, yet challenges remain for its routine clinical implementation. The major hurdle is the variability in MRI scanner types, magnetic field strengths, imaging parameters of saturation and acquisition, and post-processing algorithms across institutions complicates result comparisons and limits generalizability. Establishing standardized protocols and imaging parameters is essential for multi-site clinical trials and large-scale validation studies. CEST studies usually require the optimization of the B 1 field to achieve high labeling efficiency for exchanging protons, while managing the SAR. The CEST imaging is time-consuming due to the Z-spectrum acquisition across multiple frequency offsets. Solutions such as compressed sensing and deep learning approaches are being explored to accelerate acquisition and reconstruction processes (Cheema et al., 2024; Chen et al., 2024; Xu et al., 2024). Motion correction is crucial for human applications, as small movements can introduce errors in CEST quantification. Another key challenge is the absence of standardized methods for CEST data analysis. Additionally, interpreting CEST contrast is complex, as it depends on molecular alterations, including concentration and exchange rates, and other factors such as water accessibility. Recent developments in pulse sequences and post-processing methods aim to address these problems. Combined with artificial intelligence-based methods, these advancements are expected to facilitate fast and accurate CEST MRI, opening new avenues for clinical applications.

Conclusion

In this review, we provided an updated overview of essential CEST MRI techniques, as well as preclinical and clinical studies of CEST MRI in various neurodegenerative diseases including AD, PD, MS, and HD, focusing on connections between pathology and CEST MRI manifestations and feasibilities in clinical practice. Notably, proteinopathies commonly presented in neurodegenerative diseases, as well as various metabolic alterations presented starting from early stages of diseases including glutamate, glucose, and creatine are key targets for CEST MRI. Signal intensity alterations are observed in patient groups as well as animal models of modest sizes, providing valuable insights into the clinical value of CEST MRI for neurodegenerative diseases. Future explorations are suggested on larger scale validation and clinical translation of CEST MRI for early diagnosis of neurodegenerative diseases.

Additional file:

Additional Table 1 :
Summary of clinical and preclinical studies of CEST MRI for neurodegenerative diseases in the past 15 years.

Summary of clinical and preclinical studies of CEST MRI for neurodegenerative diseases in the past 15 years

Funding Statement

Funding:
This work was supported by The University of Hong Kong, China (109000487, 109001694, 204610401, and 204610519) and National Natural Science Foundation of China (82402225) (to JH).

Footnotes

Conflicts of interest:
The authors declare no conflicts of interest.

C-Editors: Zhao M, Sun Y, Qiu Y; T-Editor: Jia Y

Data availability statement:

All relevant data are within the paper and its Additional file .

References

Associated Data

Supplementary Materials

Summary of clinical and preclinical studies of CEST MRI for neurodegenerative diseases in the past 15 years

Data Availability Statement

All relevant data are within the paper and its Additional file .