Compartmentalized agents: A powerful strategy for enhancing the detection sensitivity of chemical exchange saturation transfer contrast

Since the very beginnings of the chemical exchange saturation transfer (CEST) technique, poor overall sensitivity has appeared to be one of its strongest limitations for future applications. Research has therefore focused on designing systems, such as supramolecular and nanosized agents, that contain a high number of magnetically equivalent mobile spins. However, the number of mobile spins offered by these systems is still limited by their composition and surface/volume ratio. The design of compartmentalized agents, that is, systems where an aqueous inner core is separated from the MRI‐detected bulk pool via a semipermeable barrier/membrane, is very much a step forward for the technique. These vesicular systems can (i) act as biocompatible and versatile carriers for dia‐, para‐, and hetero‐nuclear CEST probes, thus offering new application options; and (ii) act as CEST probes themselves via the encapsulation of a suitable agent (e.g., a paramagnetic shift reagent) that can change the resonance frequency of the spin pool in the inner compartment only. LipoCEST agents were the pioneers in the latter category, as they are able to grant picomolar sensitivity (in terms of nanoparticle concentration), and paved the way for new applications for CEST agents, especially in the theranostic research area. The use of larger, natural vesicular systems, such as yeasts and cells, in which the huge number of intravesicular spins lowers the detection threshold to a femtomolar limit, is a further step forward in the development of compartmentalized CEST agents. Finally, interesting combinations of nanovesicular and cellular compartmentalized systems have been proposed, thus highlighting how the approach has the potential to drive CEST agents towards completing their journey to mature clinical translation.


| INTRODUCTION
Chemical exchange saturation transfer (CEST) contrast is one of the most relevant achievements in the field of molecular MRI in recent decades. [1][2][3][4] Although clinical applications are currently limited to CEST-active endogenous biomolecules and endogenous-like exogenous diamagnetic CEST (diaCEST) agents, research has been very dynamic in exploring novel routes for the design of clinically translatable CEST contrast agents. It was soon realized, even in the early days of the technique, that the future of such agents would be driven by solutions to a number of issues: mainly the safety of the procedure, in terms of energy use (especially for paramagnetic agents 5,6 ), and the poor overall sensitivity of contrast detection. [7][8][9] The latter point has received a great deal of attention and many strategies have been adopted for the design of highly sensitive CEST probes. [9][10][11] However, although ST efficiency can be maximized by increasing the exchange rate of the saturable spins, k ex , this approach is limited by two conditions: (i) the necessity to meet the slow-to-intermediate exchange regime (Δω > k ex ); and (ii) large k ex values require highenergy saturation pulses that may exceed the limits imposed by specific absorption rate (SAR) constraints. In typical 1 H-water MRI detection, where the molar fraction of the free water pool is very high, the current concentration threshold per single saturable spin corresponds to a few mM.
For this reason, the problem of sensitivity has been primarily faced by designing systems that contain a high number of magnetically equivalent (or pseudo-equivalent) mobile spins, and in which the threshold molecular concentration for contrast detection scales downwards proportionally to that number. [9][10][11][12] This has led to a move from molecular to supramolecular and finally to nanometric probes that contain thousands of mobile spins and possess sensitivity on the μM scale.
Despite the unquestionable relevance of this achievement, further sensitivity improvements would allow epitopes that are present in very low concentrations to be detected, meaning that CEST contrast could be used in molecular imaging procedures.
The number of saturable mobile protons that are presented by a nanosystem is affected by the surface area and composition of the nanoparticle, which must be designed to ensure high biotolerability.
These considerations have paved the way for other strategies that aim to increase the number of CEST-active spins, including the use of "compartmentalized" agents. In this case, the saturable spins typically belong to a pool of water molecules, whose exchange with the bulk water pool occurs via a water-permeable barrier/membrane. 10,13,14 Besides the great advantages in sensitivity granted by the huge number of magnetically equivalent and saturable mobile protons, such agents have the valuable feature of being highly biocompatible, as compartmentalization is typically generated by natural or natural-like membranes, such as phospholipid-based bilayers, meaning that both artificial (e.g., liposomes) and natural (e.g., cells) vesicular systems are very suitable candidates.
This report aims to review this research field, and thus provide the readers with a description of the state-of-the-art and a critical view of the future perspectives of these probes. A general classification of compartmentalized CEST agents is provided in Figure 1.

| NANOVESICULAR SYSTEMS
Nanovesicular-based CEST systems are classified herein into two groups: (i) compartmentalized CEST agents, where "stand-alone" CEST agents are loaded into the inner core of a nanovesicular carrier; and (ii) nanovesicular CEST agents, where the saturated spins, which can be detected by CEST via the loading of proper molecules, are provided by the nanovesicle itself.

| Compartmentalized CEST agents
Nanovesicular particles, such as liposomes, have been widely used as carrier systems for drug-delivery purposes for some time, meaning that exploring their ability to transport CEST agents, both diamagnetic and paramagnetic, and assessing their MRI detection, was an obvious step to make. 15 The compartmentalization of a CEST probe into a vesicular system introduces two exchange dynamics that together define the overall CEST contrast (Figure 2A): the first is the exchange between the saturated pool of the probe's mobile spins and the intravesicular protons ( Figure A, step 2); and the second is the exchange between the latter and the bulk pool, which is mediated by the water permeability of the nanoparticle membrane (Figure 2A, step 3). High concentrations of the encapsulated CEST probe mean that the saturation of the intravesicular pool is typically rather high, thus allowing the CEST contrast on the bulk pool to be detected. Demetriou et al. have developed an analytical model to describe saturation transfer contrast in liposomes that encapsulate diaCEST agents, using D-glucose and 2-deoxy-D-glucose (2-DG) as probes. 16 Figure 2B reports representative MTR asym spectra of liposome-encapsulated D-glucose (left) and free D-glucose in solution (right), showing the effect of membrane water exchange on the CEST signal. The model considers several parameters including pH, temperature, irradiation amplitude, and intraliposomal water content. Interestingly, it was found that the compartmentalization of the CEST probes inside the liposomes induced a slowing of the exchange rate of the mobile spins of the sugar molecules, which enhanced the detection threshold of the encapsulated probes compared with the free molecules. This finding was particularly relevant because the relatively fast exchange regime of sugar mobile protons is one of the challenges for glucoCEST detection in the clinical field.
One of the first examples of the liposome encapsulation of CEST probes was reported by Opina et al., who designed a pH-sensitive liposomal CEST probe in which a 150 mM solution of the pH-responsive paramagnetic CEST (paraCEST) agent Tm-DOTA(gly) 4 was encapsulated in the aqueous core of the nanocarrier. 17 Interestingly, the encapsulation of the complex did not prevent the manifestation of its pH dependence, although the dynamic range of the responsiveness was rather attenuated because of the additional exchange at the liposome bilayer. However, this drawback was compensated for by a consistent amplification in the detection sensitivity (liposome vs. metal complex concentration) of approximately four orders of magnitude.
Another representative example has been reported by Liu et al., who formulated three liposomes that encapsulated diaCEST probes with different Δω values (glycogen, L-arginine, and poly-L-Lys), and who were able to detect them separately in the lymph nodes of mice after the nanocarriers were injected into the foot pad. 18 F I G U R E 1 General classification of the compartmentalized chemical exchange saturation transfer (CEST) agents. SR, shift reagent The well-consolidated use of liposomes as drug-delivery systems has stimulated research into using MRI-CEST contrast to predict nanocarrier tumor accumulation and to monitor the therapeutic outcome of cancer treatments that exploit the synergy between a tumor vasculature permeability enhancer (e.g., TNF-α) and a liposome-based chemotherapeutic such as Doxil. Chan et al. coencapsulated barbituric acid (BA), in the role of a diaCEST probe, into the core of liposomal doxorubicin, and have demonstrated the potential of this system to report the TNF-α-enhanced tumor accumulation of the nanomedicine in vivo. 19 Subsequently, the same research team demonstrated the potential of this approach in monitoring the vaginal distribution and retention of BA-loaded mucus-penetrating liposomes. 20 DiaCEST contrast allows drugs to be directly imaged (the so-called label-free approach). Li et al. have screened the CEST performance of 22 chemotherapeutics that contain pyrimidine, purine, and folate structural motifs, and reported the feasibility of detecting the tumor uptake of liposomes loaded with one of them, gemcitabine, after pretreatment with TNF-α to increase the permeability of the tumor endothelium. 21 With the aim of designing CEST probes with high clinical translatability, Chen et al. have encapsulated the clinically approved CT agent, iodixanol, into liposomes, and were thus able to visualize the tumor delivery of the nanocarrier via both CT and MRI-CEST imaging. 22 Another F I G U R E 2 (A) General scheme of the two exchange dynamics that define overall chemical exchange saturation transfer (CEST) contrast; (B) Example of MTR asym for (left) D-glucose-containing liposomes and (right) free D-glucose at variable temperatures (B 1 = 1.5 μT) (adapted from 16 ) original use of compartmentalized agents has been proposed by Zhang et al., who tested the capability of arginine-modified C-dots (AC-dots), which bear a very large number of exchangeable protons on their surface, to act as highly sensitive CEST probes. 23 The aim of this work was to use the AC-dots for cell-tracking purposes, with the nanoparticles (diameter 4.7 nm) being encapsulated into liposomes in order to promote their uptake into human glioma cells while maintaining the system's ability to generate CEST contrast. After transplantation into the striatum of a mouse brain, the labeled cells were more clearly visible in MRI-CEST experiments than the contralateral brain region, which was inoculated with cells labeled with blank liposomes.
In addition to proton MRI-CEST detection, liposomes have also been used as carriers of heteronuclear CEST probes, as reported by Schnurr et al., who formulated nanovesicles that were loaded with cryptophane-A, which is a host ligand for 129 Xe-CEST detection and was decorated with a targeting moiety to favor uptake by human brain microvascular endothelial cells. 24 This was a nice proof-of-concept for the contrast performance of the developed probe in vitro, and its specificity using human aortic endothelial cells as a control.

| Nanovesicular CEST agents
Liposomes were the first vesicular carriers to be identified as the most suitable nanosystems for designing compartmentalized CEST probes ( Figure 3). They are characterized by an aqueous core that may contain from 10 7 small unilamellar vesicles (SUVs) to 10 11 giant unilamellar  10,13,25,26 ). CEST, chemical exchange saturation transfer; GUVs, giant unilamellar vesicles; LUVs, large unilamellar vesicles; MLVs, large multilamellar vesicles; OLVs, oligolamellar (multivesicular) vesicles; SR, shift reagent; SUVs, small unilamellar vesicles vesicles (GUVs), water protons that exchange with the external bulk pool via the liposomes' bilayered phospholipid-based membrane ( Figure 3A). Different types of liposome have been defined in accordance with their size and number of bilayers (i.e., SUVs, large unilamellar vesicles, large multilamellar vesicles, oligolamellar [multivesicular] vesicles, and GUVs) ( Figure 3B). The usual Δω > k ex requirement must be fulfilled if their potential as CEST agents is to be fully exploited. As the resonance frequency of the intraliposomal water protons does not differ from that of the bulk pool, paramagnetic shift reagents (SRs) for water protons have been encapsulated into liposomes. 10,11 The best candidates were found to be macrocyclic lanthanide (III) complexes containing one metal-coordinated and fast-exchanging water molecule, whose protons experience the largest chemical shift effects. The lanthanide-induced shift (LIS) effect is linearly correlated to the Ln-complex concentration, and the best performance ($ 31 ppm/M) was measured at 25 C for the [Tm-DOTMA] À complex 27 ( Figure 3C).
However, the maximum amount of SR that can be encapsulated is mainly limited by osmotic constraints, and the chemical shift separation in the first generation of liposomal-based CEST probes (called lipoCEST) was 3.1 ppm ($ 5650 rad/s at 7 T and 39 C). Fortunately, the shift was large enough to exceed the exchange rate of the intraliposomal water protons ($ 585 s À1 ), thus allowing saturation transfer that was sufficiently efficient to detect sub-nM amounts of liposomes. Another important advantage of lipoCEST is the relatively slow exchange regime, which means that high-intensity pulses are not required for saturation.
Many factors are involved in generating efficient lipoCEST systems, with membrane permeability to water and the number of equivalent exchangeable protons both having a considerable influence ( Figure 3D). It is important to remember that the exchange rate of the intravesicular water protons is directly proportional to water permeability and is inversely correlated with particle size. 28 However, these variables are both rather rigid due to the constraints imposed by the clinical application of the nanovesicles. In fact, the typical hydrodynamic diameter of clinically approved liposomes ranges from approximately 80 to 120 nm, and the required stability of the nanoparticles, which guarantees the delivery of the therapeutic cargo, is typically achieved by including saturated phospholipids (e.g., DPPC and DSPC) and cholesterol in the vesicle membrane, but these components drastically reduce water permeability, thus making the CEST detection of the agents more difficult. 29 The in vivo potential of lipoCEST in molecular imaging was demonstrated by Flament et al., who formulated a Tm (III)-based lipoCEST that is decorated with the RGD tripeptide and that successfully targeted α v β 3 integrin receptors (marker of neo-angiogenesis) in an orthotopic mouse model of a human brain tumor ( Figure 4). 30 Although a shift between the two exchanging proton pools of a few ppm may be sufficient for in vivo detection, increasing Δω is certainly beneficial for in vivo applications because of the reduced dynamic range of CEST contrast as caused by endogenous MTC effects.
A significant Δω increase was achieved by exploiting the downfield bulk magnetic susceptibility (BMS) shift contribution that occurs when the paramagnetic SR is confined in nonspherical compartments that can orient themselves with the external magnetic field 31,32 ( Figure 5A). The semipermeable nature of liposome membranes mean that they can modify their shape upon osmotic stress 33 ( Figure 5B-D). Hence, when the intraliposomal core is hypotonic, liposomes shrink and adopt a discoidal shape. The BMS contribution has led to a three-fold enhancement of the absolute shift of the intraliposomal water protons, as reported in Figure 5C-D, in the case of lipoCEST that were loaded with Ln-HPDO3A complexes (Ln = Dy or Tm). 34 Interestingly, the Δω value is also dependent on the orientation of the nonspherical nanovesicles along the magnetic field, which depends on the sign of the magnetic anisotropy of the particle membrane. It has been demonstrated that the shift of the intraliposomal water protons can be further increased by incorporating amphiphilic paramagnetic Ln-complexes into the lipoCEST bilayer 25 ( Figure 5D). Moreover, the direction of the A further extension of the range of accessible Δω values has been achieved by encapsulating, within liposomes, neutral multimeric SRs, which can lead to an increase in the concentration of paramagnetic centers without affecting the osmolarity of the inner nanocarriercore, 26 and by encapsulating a neutral SR with two metal-coordinated water molecules. 35 More recently, Abozeid et al have designed a new version of a lipo-CEST agent in which the paramagnetic SRs that are loaded into the nanovesicles are Co (II) ion-based d-metal complexes, with promising Δω values being obtained. 36 Very interestingly, the shape of a lipoCEST agent, and consequently its Δω value, can also be influenced by attaching large chemical moieties onto the liposome surface. It has been observed that the conjugation of streptavidin, in stoichiometric deficit, to a biotinylated lipoCEST agent (1:4 streptavidin:biotin), induced a three-fold downfield shift (from 4 to 12 ppm). However, the effect was completely lost when the biotin moieties that were exposed on the liposome surface were saturated with streptavidin. 37 It has been hypothesized that high shift is consequent to the F I G U R E 5 (A) Change of liposome shape from spherical to osmotically shrunken. (B) TEM image of osmotically shrunken lipoCEST agents (cigar-like structure). (C) Variation of intraliposomal water resonance upon changing external osmolarity. (D) Z-spectra of osmotically shrunken lipoCEST probes that encapsulate [Tm-HPDO3A] as the shift reagent (SR) and incorporate an amphiphilic (Tm-or Dy-based) SR in the membrane that changes orientation with the main magnetic field. (E) Multiple visualization of lipoCEST agents injected into the muscles of a mouse. Green = CEST contrast at 3.5 ppm (spherical agent that encapsulated [Tm-DOTMA] À ) and red = CEST contrast at À17 ppm (aspherical vesicles that entrapped [Dy-DOTMA] À ). (F) Range of chemical shifts that can be exploited using lipoCEST probes (adapted from 31,32 ). CEST, chemical exchange saturation transfer loss of the spherical shape of the nanocarrier resulting from the spatially anisotropic binding of streptavidin when it is present in lower amounts.
As more streptavidin is added, the distribution of the protein on the liposome surface becomes isotropic, and the particles readopt the original spherical shape. This switch has been harnessed to design a smart lipoCEST probe in which the Δω value is responsive to the enzymatic activity of MMP-2. The system is based on the incorporation, into the liposome membrane, of a chemical, to which biotin is bound via a peptidic linker that can be cleaved by the enzyme. 37 In principle, any vesicular system that possesses a water-permeable coating may act as a highly sensitive CEST probe. However, liposomes have by far attracted the most attention because of their high clinical translatability, and only a single example of a nonliposomal nanovesicular CEST probe, which is based on polymeric membranes, named polymerCEST, has been reported to date. 38 More recently, the achievements obtained by extending the range of Δω values have been applied to the use of giant microvesicles (size 1-2 μm) to further improve the sensitivity of CEST detection 39 ( Figure 6A, confocal microscopy). The bigger vesicles (named giantCEST) are formulated with saturated phospholipids, have displayed asymmetry in their Z-spectra ( Figure 6B,C), and possess a sensitivity of approximately 1 pM (particle concentration), while the corresponding value for SUV (120 nm), with similar formulation and saturation conditions, was approximately 9.5 nM ( Figure 6D). Quite surprisingly, the intravesicular shift of the water protons in giantCEST was lower than that of the analogous SUV probes (5.2 vs. 14 ppm, Figure 6E), despite them sharing the same membrane composition and inner core payload (and consequently experiencing the same osmotic stress). It has been found that this observation reflected the very limited sensitivity of giant particles when responding to osmotic gradients. This then attenuates the contribution of BMS to the Δω value.
The micron size of GUV allowed single particles, which were loaded with different fluorophores in the aqueous core and bilayer, to be observed using fluorescence confocal microscopy, and these experiments highlighted that GUVs mainly reacted to osmotic stress by producing invaginations in their bilayers and thus transforming into anisotropic multilamellar particles without dramatic changes in their overall shape ( Figure 6G). A combination of liposomes that encapsulate a paraCEST agent and conventional lipoCEST probes has been explored as a means of expanding the range of imaging tools that can visualize payload release from liposomal nanocarriers. 40 The strategy consists of mixing two liposomes that are formulated to release their payloads after being triggered by ultrasound (US) and acidic pH, and then report drug release via a modulation in CEST contrast. The US-sensitive formulation was a conventional lipoCEST agent whose contrast reports on nanocarrier integrity, whereas the pH-dependent formulation was designed to encapsulate a lanthanide-based SR within a liposome with very low water permeability in order for the agent to be "invisible" when the nanocarrier is intact and for the CEST contrast to activate at acidic pH when the paraCEST probe is released.
Another "smart" lipoCEST agent has been designed, by Langereis et al., to report on the heat-triggered release of liposomal content. 41 Another example has recently been published by Han et al., who designed a hydrogel that was embedded with liposomes that were, in turn, loaded with BA, with the aim of developing a therapeutic wafer for the treatment of glioblastoma multiforme and to monitor drug release by CEST-MRI. 42 The CEST contrast at 5 ppm from the bulk water, corresponding to the mobile protons of BA (acting as drug surrogate), reported that the drug was in the hydrogel, whereas saturation at À3.4 ppm, corresponding to nuclear overhauser effect transfer from the fatty acid chains of the liposome bilayer, provided information on the number of nanocarriers embedded within the wafer. The system was successfully tested in vivo by transplanting the gel into a mouse brain, thus allowing the two CEST responses to be monitored over time. 42 3 | CELL-BASED SYSTEMS

| Moving from lipoCEST to cell-based CEST agents
The development of lipoCEST agents clearly demonstrated the significant advantages of exploiting the water molecules entrapped inside biological compartments as CEST mobile protons. In fact, the very high abundance of water molecules in biological systems surely makes this pool of equivalent protons suitable for increasing the sensitivity of CEST-MRI agents. The data from SUV and GUV vesicles showed that larger vesicular sizes provide a considerable increase in water content and, consequently, a significant enhancement in CEST efficiency. Moreover, studies on lipoCEST agents have demonstrated that a pivotal role in producing efficient compartmentalized CEST agents was played by the shape of the container, the water permeability of the membranes, and the payload of the paramagnetic SR. To summarize, optimal compartmentalized CEST systems must display the following properties: (i) a large amount of equivalent water protons (with larger vesicles being able to entrap more protons); (ii) high SR payload (i.e., a good capability to entrap large amounts of paramagnetic SRs); (iii) proper membrane formulation (for a good exchange  39 ). CEST, chemical exchange saturation transfer; GUVs, giant unilamellar vesicles; SUVs, small unilamellar vesicles rate); (iv) anisotropic shape (so that BMS can contribute to the shift of the intracompartmental water NMR signal); and (v) good biocompatibility.
It is therefore obvious that cells, which are natural, micrometric water-filled compartments, should be considered as carriers for the development of highly sensitive CEST probes.
Cells are natural systems derived from living organisms that generally display high biocompatibility and sizes in the micrometric range (eukaryotic cell radii range from 5 to 50 μm). Eukaryotic cell membranes are composed of a mixture of saturated and unsaturated phospholipids, with the presence of cholesterol modulating water permeability. This, coupled with the presence of transmembrane transporter and channel proteins, ensures the separation of the inner and outer cell compartments (needed for the CEST effect), and appropriate values of the water exchange rate ( Figure 7A). Cells are normally not spherical, and some can change and modulate their shape. Eukaryotic cells normally display a wide variety of different morphologies, including spheroid, ovoid, cuboidal, cylindrical, flat, lenticular, fusiform, discoidal, crescent, ring stellate, and polygonal.
Finally, several methods with which to load cells with exogenous small molecules (such as hydrophilic paramagnetic SRs) and nanoparticles have been reported, [44][45][46][47] and these methods can be specific either for entrapment in the cytosol (electroporation, 48 and hypotonic swelling 49 ) or in endosomes (macropinocytosis, receptor-mediated endocytosis, etc. 50,51 ). The loading of the proper number of SRs into the inner cavity is therefore not challenging, meaning that it is possible to shift the resonance of the intracellular water protons compared with the extracellular pool. In 2014, our group demonstrated that lanthanide-loaded erythrocytes have the potential to act as highly sensitive CEST agents, 43

| ErythroCEST: red blood cells as CEST agents
Red blood cells (RBCs) were the first cell type to be considered for use as a potential CEST system. They were chosen because they possess some positive peculiarities: (i) they are the most abundant natural cells ($ 5 Â 10 9 RBCs/ml of human blood); (ii) they have a well-known anisotropic shape (biconcave disk, anucleated); and (iii) most of their biological/biochemical properties are well known (e.g., their membrane composition in terms of both phospholipids and the water transporters that regulate the water exchange across the membrane). 52,53 The intracellular half-life of water molecules in RBCs has been the object of many studies over the last 50 years (most of them based on NMR approaches), and this value has recently been remeasured as 19.1 ± 0.65 ms, 54 which is the shortest residence time that has been reported so far for eukaryotic cells, and is similar to the values measured for giant liposomes. RBCs can simply be considered as micrometric containers that are mainly filled with hemoglobin (Hb), with an intracellular volume, as routinely measured by automatic hemochromocytometry (of $ 90 fl). This corresponds to an intracellular content of approximately 6 x 10 12 equivalent water protons, which can be saturated to generate a CEST effect. Due to the intracellular residence time, the chemical shift separation Δω must be higher than approximately 300 rad/s, and paramagnetic SRs (lanthanides and d-transition metals) can be internalized within RBCs to efficiently separate the resonance of the inner water protons from that of the extracellular water protons, and to generate dipolar and BMS shift contributions. Shifting internal water using a dipolar-only contribution is not feasible because this would entail the internalization of a concentration of SR that cannot be attained using the available cell-labeling strategies (of the order of hundreds of millimoles of Ln-complexes).
On the other hand, the BMS contribution can allow the necessary shift separation to be easily reached (requiring < 10 mM concentration of Ln-complexes, considering the more effective Ln ions), but requires magnetically oriented, not spherical, compartments. It has been reported that nonspherical cells, such as RBCs and spermatozoa, can orient with an external strong magnetic field as a result of the anisotropy of magnetic susceptibility that is intrinsic to their cellular membranes. [55][56][57] RBCs naturally tend to align their disk plane with the magnetic field (B 0 ), and there is a higher percentage of oriented cells with higher magnetic field strength; it has been estimated that approximately 50% of RBCs are oriented parallel to the field at B 0 = 2 T, and this percentage increases to 100% at 4 T. 58 This observation suggests that the internalization of paramagnetic SRs into RBCs is expected to generate a BMS contribution.
RBCs are suitable for loading with small hydrophilic metal complexes, with Ln-HPDO3A complexes (Ln = Eu, Gd, Dy, Tm, and Yb) being preliminary tested. Several internalization routes for entrapping exogenous molecules inside RBCs have been investigated over the last 2 decades, with the aim of (i) loading drugs to generate long-circulating drug-delivery systems 59 ; and (ii) loading labeling agents to make RBCs detectable by imaging techniques (e.g., MRI, OI, and SPECT). 60,61 Of the different routes that are normally available for the loading of small molecules inside cells (micropinocytosis, electroporation, hypotonic swelling, endocytosis, sonoporation, etc.), hypotonic swelling has been shown to be one of the most efficient and safest because it does not affect the functionality of RBCs, which retain their hematological properties (i.e., MHC, MCHC, MCV, and RDW), mechanical properties (response to osmotic stress, deformability), and the capability to carry oxygen. 62 In fact, an intracellular Ln-HPDO3A concentration of approximately 4.5 mM (corresponding to $ 3-4 x 10 8 complexes/RBC) has been obtained using this method. This amount led to a Δω of 6.5 ppm for the Dy-complex, 4.9 ppm for Gd (III), 4.5 for Tm (III), 2.7 ppm for Yb (III), and 2.4 ppm for Eu (III), according to the order of the effective magnetic moments (μ eff ) of the Ln ions ( Figure 7B,C). As the molar dipolar shift reported for Dy-HPDO3A is approximately À19 ppm/M and considering an intracellular concentration of 4.5 mM, the dipolar contribution to the observed shift of 6.5 ppm accounted for less than 1.5%, and it can therefore be said that the chemical shift values for the intracellular water protons were dominated by the BMS contribution. Interestingly, when the paramagnetically loaded RBCs were exposed to a hypotonic environment, the chemical shift separation disappeared in accordance with the switch from the biconcave to the spherical cell shape and the loss of their anisotropic orientation in the magnetic field. The CEST contrast detection of SR-loaded RBCs (called erythroCEST) was the highest displayed so far by CEST agents (less than 1 pM in terms of cell concentration).
The excellent biocompatibility of this system and its high sensitivity makes it very attractive for in vivo applications, especially as a reporter of blood volume. As proof of concept, erythroCESTs were used to assess vascular volume in the tumor region in subcutaneous murine models of cancer, where CEST contrast can report on intratumor blood perfusion ( Figure 7D,E). Research on this topic is still ongoing with the aim of testing and validating this approach in preclinical biomedical research.

| CellCEST: eukaryotic cells as CEST agents
More recently, our group has deepened the study of using Ln-loaded cells as CEST agents by investigating the effect of entrapping Ln-complexes inside cell types other than RBCs, which are characterized by a well-defined size and shape (unlike all other eukaryotic cells) and the absence of intracellular vesicles (e.g., endosomes), thus allowing the internalized SR to be confined in the cytosol. 50 Murine macrophages (J774.A1) were the first to be considered for the generation of cellCEST agents, and Ln-HPDO3A complexes (Ln = Gd, Eu, Dy, and Tm) were loaded inside cells via hypotonic swelling (allowing direct loading into the cytosol) or macropinocytosis (loading in endosome-like vesicles). About 1.5 x 10 10 complex molecules were internalized per cell, which corresponds to an intracellular Ln-HPDO3A concentration of approximately 10 mM. A CEST effect (downfield to the extracellular water protons) was observed for all the tested SRs, regardless of whether they were loaded using hypotonic swelling or macropinocytosis. The chemical shift separation was found to be influenced by three properties of the entrapped paramagnetic metal ions: (i) the intracellular concentration; (ii) the μ eff value of the Ln ion (Dy > Gd > Tm > Eu); and (iii) the intracellular compartmentalization, which is on the internalization route. It was observed that the localization of the SR in the intracellular vesicles enhanced the BMS effect and larger Δω values were consequently found than for cytosolic confinement. For instance, the internalization of [Dy-HPDO3A] in J774A.1 cells generated a chemical shift separation of 8 ppm in the case of micropinocytosis, whereas 3.5 ppm was observed in the case of hypotonic swelling ( Figure 8A). As the CEST contrast is proportional to the T 1 of the bulk pool that is not saturated, the internalization of [Gd-HPDO3A] quenched the effect (especially for cytoplasmatic loading), whereas the strong T 2 effect caused by the Dy (III) ion influenced the frequency bandwidth of the CEST effect ( Figure 8A,B).
Analogously to erythroCEST, cellCEST shows very high sensitivity. When mixing Dy-labeled cells with unlabeled ones, the CEST contrast was proportional to the percentage of labeled cells, and the detection threshold appeared to be lower than 10% of the labeled cells (corresponding to One potential and relevant application for cellCEST is in the assessment of cell proliferation rate. In fact, upon cellular division, a mother cell dilutes its SR content between the two newborn cells. In such a way, there is a dilution of intracellular concentration of the paramagnetic SRs, resulting in a Δω reduction ( Figure 8C). CellCEST contrast has been successfully used to assess the proliferation rate both in vitro and in vivo, showing that murine TS/A breast cancer cells proliferate faster than murine J774A.1 macrophages in vitro, and that the in vivo proliferation rate of TS/A is lower than it is in vitro (Figure 8C,D).
As a proof of concept, cellCEST detection was tested in vivo in a murine model by labeling murine TS/A breast cancer cells with , and the results obtained demonstrated the feasibility of this approach for imaging cells and their proliferation rate in vivo ( Figure 8D).
In addition, CellCEST contrast can also be used to gain information on cell viability as dead cells cannot generate a CEST effect due to the loss of SR and increase in membrane water permeability.

| YeastCEST: yeast cells as CEST agents
The cellCEST approach has more recently been further developed by Patel et al., who investigated the effect of high-spin macrocyclic Co (II) complexes as paramagnetic SRs upon entrapment inside Saccharomyces cerevisiae cells. 63 The complexes were based on the 1,4,7triazacyclononane ligand ( Figure 9A), and the complexes have been tested as pH paraCEST agents, as fluorescent agents, and to generate cellCEST agents. 63,64 Yeast cells were loaded with the Co (II)-complexes using a heat treatment transformation protocol 65 at a complex concentration of about 50 mM. It was reported that Co (II)-labeled yeast cells produced broadened CEST spectra with slight asymmetry, and the typical signal at 1-2 ppm (already observed for RBCs and other eukaryotic cells) was observed for both the control and Co (II)-labeled cells ( Figure 9B). The labeled yeast cells displayed broad saturation transfer in the range of 1 to 4 ppm. As discussed above, the signal broadening was a consequence of Co (II)induced T 2 * effects, while the decrease in the ST effect was due to T 1 shortening. ICP-MS measurements reported that a very high amount of Co (II) was entrapped inside the yeast cells ($ 10 10 -10 11 ions/cell, corresponding to intracellular concentrations of 1.4-17.6 and 0.3-4 M, depending on the Co (II) complex used and assuming the yeast cells were spherical with a diameter in the range of 5 ± 2 μm). Even when considering the rather low μ eff value of Co (II) (2.8-3.2), 66 the very high intracellular concentration of the SR meant that a larger chemical shift separation for the intracellular water protons was expected than was actually observed (< 2 ppm). As hypothesized by the authors, this may be the result of SR sequestering by yeast organelles, as was demonstrated in fluorescence microscopy experiments. However, one would expect that the F I G U R E 9 (A) Chemical structures of Co-based shift reagents used to generate yeastCEST. (B) Z-and ST-spectra of yeast cells labeled with co-complexes (adapted from 63 ). CEST, chemical exchange saturation transfer confinement of the SR inside the organelles would influence CEST contrast rather than the Δω value. A more sound hypothesis, again proposed by the authors, suggests that the small shift separation relies on the partial loss of the BMS contribution due to the oxidation, promoted by the oxidative intracellular environment of Co (II) to diamagnetic Co (III).

| Combined liposome/cell CEST systems
The properties of lipoCEST and erythroCEST agents have been combined to generate a new CEST system in the form of the supramolecular aggregation of these two compartmentalized systems. 67 Spherical lipoCEST agents that encapsulate Dy-HPDO3A and were properly formulated to carry residual positive charges on their outer surface were electrostatically bound to unlabeled RBCs that are known to possess a negatively charged membrane. The lipoCEST agent displayed the typical upfield resonance (Δω = À4.2 ppm) that is caused by the dipolar shift of the intraliposomal water protons, whereas unlabeled RBCs only displayed the typical CEST peak with a Δω of 1-2 ppm, caused by the saturation transfer from endogenous mobile protons (sugars, amines, etc.).
When the LipoCEST agents were bound to the RBCs, a new downfield CEST signal (Δω = 4.8 ppm) appeared, in addition to the upfield Lipo-CEST resonance, and this was attributed to the resonance of the intracellular water protons that had been shifted by BMS effects induced by the lipoCEST agents anchored on the RBC membranes. As expected for a BMS contribution, it was observed that the shift was directly proportional to: (i) the number and size of the anchored lipoCEST vesicles; (ii) the concentration of the encapsulated SR; and (iii) the effective magnetic moment of the lanthanide ion. The Dy-based lipoCEST/RBC construct was intravenously administered into tumor-bearing mice. The CEST contrast that was measured in the tumor region provided data on both the vascular volume (via the CEST signal from the RBCs) and liposome release (via the CEST signal from lipoCEST agents).
Another smart example of a CEST-detectable liposome/cell system was reported by Chan et al., who designed microcapsules that incorporate liposomes and hepatocytes, which were used as a cell model, and that were decorated with a layer of arginine-rich protamine sulfate. This system demonstrated the ability to report the pH decrease that occurs as a consequence of cell apoptosis via the well-known pH-dependent modulation of CEST contrast. 68 Although the system displayed some limitations, such as partial overlapping with endogenous CEST contrast, it displays a very high level of clinical translatability.

| CONCLUSIONS
The intrinsic low sensitivity of the CEST MRI technology has driven the development of new agents that bear multiple, magnetically equivalent, exchangeable protons. In this scenario, the use of compartmentalized vesicular systems, in which an aqueous inner core is separated from the bulk compartment via a water-permeable membrane, is an important and successful strategy that is able to provide compartmentalized systems, whose diagnostic/therapeutic activity is often investigated, with particular imaging capabilities.
Liposomes are the most representative example. In addition to acting as carriers for diaCEST and paraCEST probes, the nanovesicular nature of this carrier means that it can encapsulate paramagnetic SRs within its inner cavity to generate a new class of highly sensitive agents whose imaging response can report on the integrity of the nanosystem, thus granting it potentially interesting applications in the theranostic field.
Larger, natural vesicular systems, such as yeast and cells, have been successfully investigated with the aim of improving contrast detection sensitivity and thus further extending the landscape of applications for CEST-based agents, while also improving biocompatibility.
Compartmentalized CEST systems represent an innovative step forward in the field, with multiple advantages in respect to small molecular diaCEST and paraCEST. Firstly, they allow increasing the sensitivity, reaching an efficiency suitable for molecular imaging and targeting procedures. Moreover, nanosystems include a highly diverse group of systems whose features can be modulated at will (biodistribution in tissues, size, formulation, etc.), making possible the optimization based on the necessary application. More generally, the separation of an inner compartment from the external microenvironment opens up the possibility of properly modulating the CEST features.
However, several issues still need to be addressed to achieve the full clinical translation of these probes. Except for the diaCEST-active molecules that can already be injected into humans (e.g., glucose and iodinated CT agents), all the systems described in this review make use of chemicals (in most cases, metal complexes) that have never been tested in clinical trials. Furthermore, in nanoparticle-based systems, the compound that conveys the CEST effect, whether directly or indirectly, will display a pharmacokinetic profile that is driven by the nanocarrier, and, for this reason, may bring with it toxicity issues due to the intracellular fate of nanosystems. Hence, the future of these probes will strongly depend on the possibility of designing CEST-based imaging procedures that can respond to a specific medical demand that is not yet met by other available methodologies. Meanwhile, these systems will continue to make a valuable contribution to basic and preclinical research.

ACKNOWLEDGMENTS
Financial support from the European Union's Horizon 2020 research and innovation programme ("GLINT -GlucoCEST Imaging of Neoplastic Tumours" project), the Italian Ministry of Research ("Rationally designed nanogels embedding paramagnetic ions as MRI probes" -PRIN project) and FOE support for the Multi-Modal Molecular Imaging Italian Node (MMMI) within Euro-Bioimaging ERIC, is gratefully acknowledged. Open Access Funding provided by Universita degli Studi di Torino within the CRUI-CARE Agreement.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in IRIS AperTO at https://iris.unito.it.