Design and Synthesis of Luminescent Lanthanide-Based Bimodal Nanoprobes for Dual Magnetic Resonance (MR) and Optical Imaging

Current biomedical imaging techniques are crucial for the diagnosis of various diseases. Each imaging technique uses specific probes that, although each one has its own merits, do not encompass all the functionalities required for comprehensive imaging (sensitivity, non-invasiveness, etc.). Bimodal imaging methods are therefore rapidly becoming an important topic in advanced healthcare. This bimodality can be achieved by successive image acquisitions involving different and independent probes, one for each mode, with the risk of artifacts. It can be also achieved simultaneously by using a single probe combining a complete set of physical and chemical characteristics, in order to record complementary views of the same biological object at the same time. In this scenario, and focusing on bimodal magnetic resonance imaging (MRI) and optical imaging (OI), probes can be engineered by the attachment, more or less covalently, of a contrast agent (CA) to an organic or inorganic dye, or by designing single objects containing both the optical emitter and MRI-active dipole. If in the first type of system, there is frequent concern that at some point the dye may dissociate from the magnetic dipole, it may not in the second type. This review aims to present a summary of current activity relating to this kind of dual probes, with a special emphasis on lanthanide-based luminescent nano-objects.


Introduction
Current biomedical imaging techniques are vital for the diagnosis of various diseases. Each imaging mode has its own merits and disadvantages and uses specific probes exhibiting particular physical and chemical properties. However, a single technique does not encompass all the functionalities required for comprehensive imaging. For instance, magnetic resonance imaging (MRI) has the advantage of being a non-invasive technique for in vivo 3D-tomography, but it is limited by low target sensitivity. Also, optical imaging (OI) is non-invasive and has the advantage of a high sensitivity and a high specificity, but it is limited by the poor light tissue penetration. Most of the time, it requires appropriate and expensive endoscopic optical fibers for local photoexcitation and/or detection. Therefore, multimodal imaging methods, with enhanced signal sensitivity, better spatial resolution, and the ability to relay information about biological systems at the molecular and cellular levels are becoming important tools with an absolute necessity of designing multimodal probes combining ideally, in single objects, all the desired properties.
Nanoparticles (NPs) as platforms bringing together several functionalities offer such an opportunity. They can be easily produced by colloidal chemistry with different sizes allowing thus a strict control of their size-dependent physico-chemical properties (surface plasmon, quantum confinement, superparamagnetism) and then a strict control of their MR imaging is based on the magnetic nuclear resonance of water protons in the body. With MRI, three-dimensional images of entire live specimens can be obtained with high resolution and without the use of ionizing radiation. The MR contrast is essentially derived from the environment variation of water protons leading to different signal intensities.
In practice, when a living body is placed in a large static magnetic field B 0 , many of the free hydrogen nuclear magnetic moments align themselves with the direction of the magnetic field. When a magnetic radio-frequency pulse B RF , with an appropriate frequency (the gyromagnetic resonance frequency) is applied perpendicular to B 0 , the nuclear magnetic moments of the hydrogen atoms tilt away from B 0 . At the end of the pulse, the magnetic moments are in the plane perpendicular to B 0 . They then relax to realign themselves parallel to B 0 , by reducing their B RF induced transverse magnetization component (M xy ) to zero and increasing their longitudinal magnetization component (M z ) to its initial value, with characteristic T 2 and T 1 relaxation times, respectively.
The Bloch equations for the relaxation of the longitudinal (M z ) and transverse (M xy ) components of the nuclear magnetization of water protons to their equilibrium state indicate that the MRI signal is proportional to the proton density of the targeted tissues and to the product (1−e −T R /T 1 ) −T E /T 2 , where T R is the time of the B RF pulse and T E the time between the application of the B RF pulse and the measurement of the response signal.
In the presence of CAs, the T 1 and T 2 relaxation times are significantly shortened, affecting thus the MRI signal, inducing the desired water proton response contrasting.
The efficiency of CAs is usually expressed by the enhancement of the relaxation rate, also called relaxivity, r i , defined according to the following equation: (1/T i ) = (1/T i ) 0 + r i [CA] (1) where (1/T i ) is the inverse of T i(=1 or 2) in the presence of CA, at a [CA] concentration, while (1/T i ) 0 is the inverse of T i(=1 or 2) in its absence. It is generally stated in MRI textbooks that the ratio r 2 /r 1 determines whether a given CA is more suitable as a T 1 (positive) or as a T 2 (negative) MRI agent. When CAs affect longitudinal and transverse relaxivities to a similar degree they are better suited for T 1 -weighted imaging and are commonly called positive CAs, whereas when they preferentially increase transverse relaxivity, typically r 2 /r 1 >>1, they are better suited for T 2 -weighted imaging and are called negative CAs. The interaction between CAs and water protons is of two types, depending on whether the water molecule is directly attached to the CA or not. It is of inner-sphere type if it proceeds through a direct water attachment and it is of outer-sphere type if it does not (Figure 1), the former being more effective for positive CAs while the latter for the negative ones.
ween the application of the BRF pulse and the measurement of the respons presence of CAs, the T1 and T2 relaxation times are significantly shortened s the MRI signal, inducing the desired water proton response contrasting.
The efficiency of CAs is usually expressed by the enhancement of the rela called relaxivity, ri, defined according to the following equation: (1/Ti) = (1/Ti)0 + ri [CA] ere (1/Ti) is the inverse of Ti(=1 or 2) in the presence of CA, at a [CA] concentra i)0 is the inverse of Ti(=1 or 2) in its absence. It is generally stated in MRI tex ratio r2/r1 determines whether a given CA is more suitable as a T1 (positive gative) MRI agent. When CAs affect longitudinal and transverse relaxi ilar degree they are better suited for T1-weighted imaging and are comm itive CAs, whereas when they preferentially increase transverse relaxivity 1 >>1, they are better suited for T2-weighted imaging and are called negativ The interaction between CAs and water protons is of two types, dep ether the water molecule is directly attached to the CA or not. It is of inner-s t proceeds through a direct water attachment and it is of outer-sphere typ (Figure 1), the former being more effective for positive CAs while the la ative ones. In practice, the relaxivity is more often limited by slow exchange (due to a long m) when only a limited number of water molecules can be relaxed.
In the outer-sphere model introduced by Ayant et al. [23] and Freed et al. [24] for paramagnetic agents, and adapted by Gillis et al. to superparamagnetic ones [25], the increase in the relaxation rate 1/T2 compared to pure water originates from fluctuating dipolar interactions between nuclear spins of water protons and the electronic magnetic moment of the CAs. For a limited range of diameters called the 'motional averaging regime', the CA can be considered immobile during TE compared to the random trajectories of water molecules diffusing around the CA sphere. In this case, Vuong et al. have shown that r2 follows a universal scaling law that is quadratic both with the magnetization and with a, the radius of the outer sphere also called the relaxometric size, defined as the minimum approach distance between H2O molecules and the CA center [26]. The diffusional correlation time of outer-sphere water molecules (D) can also contribute to the CA relaxivity, even if the relationship between D and r2 is not evident.
Finally, CAs must be non-toxic and they must be engineered accordingly. An agent with a high relaxivity value allows reducing its dose and, thus, its potential toxicological effect. These two criteria have to be absolutely respected to allow CA clinical use and consequently to drive their design and synthesis [2].

Positive Contrast Agents
Positive CAs are mainly ( Figure 2) paramagnetic metallic complexes. Their average size is in the nanometer range and they can be easily distributed over the whole body by intravenous administration. Since they are small, they can diffuse into the extravascular media and may be eliminated by glomerular filtration.
The most used paramagnetic cation for this purpose is Gd 3+ . This cation has a high electronic magnetic moment thanks to its seven non-paired 4f electrons. When submitted to a strong static magnetic field, B0, its moment is aligned parallel to the direction of the field. Its effective magnetic moment is 7.95 µ B. Free Gd 3+ are nephrotoxic and neurotoxic [27]. For these reasons, they are never used without strong chelation to polydentate hydrophilic ligands ( Figure 2) to make them as stable as possible in physiological media. ligands (the commercial names are given in parenthesis). Reproduced from [2], with permission from ACS, 1999 Other paramagnetic cations, chelated by the same kind of ligands, have also been investigated as potential positive CAs. These cations are mainly from the first series of d-block elements, like Mn 2+ and Fe 3+ , which have an effective magnetic moment of 5.92 µ B, Figure 2. The semi-developed formula of Gd-chelates involving linear (top) or cyclic (down) ligands (the commercial names are given in parenthesis). Reproduced from [2], with permission from ACS, 1999 The most used paramagnetic cation for this purpose is Gd 3+ . This cation has a high electronic magnetic moment thanks to its seven non-paired 4f electrons. When submitted to Nanomaterials 2021, 11, 354 5 of 34 a strong static magnetic field, B 0 , its moment is aligned parallel to the direction of the field. Its effective magnetic moment is 7.95 µ B . Free Gd 3+ are nephrotoxic and neurotoxic [27]. For these reasons, they are never used without strong chelation to polydentate hydrophilic ligands ( Figure 2) to make them as stable as possible in physiological media.
Other paramagnetic cations, chelated by the same kind of ligands, have also been investigated as potential positive CAs. These cations are mainly from the first series of d-block elements, like Mn 2+ and Fe 3+ , which have an effective magnetic moment of 5.92 µ B , smaller than the 7.92 µ B of Gd 3+ . This magnetization decrease is expected to decrease total relaxivity. Typically, chelating Mn 2+ by N,N'-dipyridoxylethylenediamine-N,N'-diacetate-5,5'-bis(phosphate) (DPDP) forms a Mn-DPDP complex with a longitudinal relaxivity r 1 of 1.6 mM −1 .s −1 (T = 25 • C, B 0 = 0.47 T), smaller than the usual 4 mM −1 .s −1 of monomeric Gd complexes within the same operating conditions [28,29]. Besides Gd 3+ , other paramagnetic lanthanide cations were tested. The most commonly studied is Dy 3+ , which has the largest effective magnetic moment value (10.65 µ B ) and which has been expected to allow reaching higher longitudinal relaxivity values. In fact, its highest effective magnetic moment and its shortest electronic relaxation time (~10 −13 s), make it much more investigated for T 2 contrast enhancement at ultrahigh field [30,31].
Compartmentalization of free or chelated paramagnetic cations, mainly Gd 3+ , was also explored as chemical strategy to enhance r 1 (Figure 3). Liposomes [38], carbon nanotubes (CNTs) [39][40][41], or the cavities of certain proteins, like apoferritin [42,43] were used in order to significantly increase the number of Gd-coordinated water molecules and then to enhance the longitudinal relaxivity of the resulting architectures. Unfortunately, despite the very promising physical properties measured on this new generation of probes, their weak stability in biological media compromised their clinical use. Finally, with always the same goal of increasing the number of interacting water molecules, biocompatible inorganic particles, like silica or gold, were decorated by monomeric Gd-chelates ( Figure 3), with very encouraging results [44]. , x FOR PEER REVIEW 5 of 34 relaxivity. Typically, chelating Mn 2+ by N,N'-dipyridoxylethylenediamine-N,N'-diacetate-5,5'-bis(phosphate) (DPDP) forms a Mn-DPDP complex with a longitudinal relaxivity r1 of 1.6 mM -1 .s -1 (T = 25 °C, B0 = 0.47 T), smaller than the usual 4 mM -1 .s -1 of monomeric Gd complexes within the same operating conditions [28,29]. Besides Gd 3+ , other paramagnetic lanthanide cations were tested. The most commonly studied is Dy 3+ , which has the largest effective magnetic moment value (10.65 µ B) and which has been expected to allow reaching higher longitudinal relaxivity values. In fact, its highest effective magnetic moment and its shortest electronic relaxation time (~10 −13 s), make it much more investigated for T2 contrast enhancement at ultrahigh field [30,31].
Compartmentalization of free or chelated paramagnetic cations, mainly Gd 3+ , was also explored as chemical strategy to enhance r1 ( Figure 3). Liposomes [38], carbon nanotubes (CNTs) [39][40][41], or the cavities of certain proteins, like apoferritin [42,43] were used in order to significantly increase the number of Gd-coordinated water molecules and then to enhance the longitudinal relaxivity of the resulting architectures. Unfortunately, despite the very promising physical properties measured on this new generation of probes, their weak stability in biological media compromised their clinical use. Finally, with always the same goal of increasing the number of interacting water molecules, biocompatible inorganic particles, like silica or gold, were decorated by monomeric Gd-chelates ( Figure 3), with very encouraging results [44].     The last family of positive CAs is that based on heavy paramagnetic inorganic NPs, which have intrinsically high τ R values. These particles are functionalized by hydrophilic molecules or biomolecules to make them biocompatible. Most of them are 3d-and 4f-block metal oxides like Mn 3 O 4 [48], MnO [49], and Gd 2 O 3 [50][51][52]. Their longitudinal relaxivity depends mainly on their size, the spin of the metallic cation and the organic coating, which should allow water to interact with the paramagnetic surface cations. Experimentally, their r 1 values were found to be of the same order of magnitude as those of molecular CAs (Table 2). Recently, ultra-ultrasmall superparamagnetic iron oxide NPs (UUSPIOs) with a very small size (less than 5 nm) and a very small magnetization (≤10 Am 2 .kg −1 ), have been included in this last class of CAs [53,54]. Their measured r 2 /r 1 values close to 1 explain their classification. They are not yet commercially available, but they would be an excellent alternative to the more expensive gadolinium-based CAs, currently used and still suspected of toxicity [55].

Negative Contrast Agents
Negative CAs are exclusively superparamagnetic NPs coated with hydrophilic ligands or polymers to improve their ability to form stable aqueous colloids. Their introduction in a solution or a tissue induces local magnetic field gradients, which accelerate the loss of phase coherence of the water proton magnetic nuclear moments, improving MR contrasting at their proximity [26,57].
Finally, d-block metal NPs, like Fe or Fe-Co, protected from air and water oxidati have been tested. These CAs have higher magnetization (typically 185 Am 2 .kg -1 for Fe-ba NPs at room temperature) than all SPIO-and ferrite-based particles (typically 60-90 Am 2 · at room temperature) [71]. As a consequence, despite their very reduced size, their r2 an /r1 values are expected to be very high (Table 4).    [68], and Zn 1-x Mn x Fe 2 O 4 [69] among others. Exchange-coupled magnetic NPs, commonly called enhanced ferrite nanoparticles (EFNPs), consisting of a metallic core, mainly iron, coated by a ferrite shell, mainly iron oxide, form another class of negative CAs. Their structure allows increasing the total magnetization of the engineered particles and prevents the oxidation of their highly magnetized metallic cores [70]. For comparison, such EFNPs exhibit r 1 and r 2 relaxivities of 7.19 and 9.96 mM −1 ·s −1 (expressed per particle), respectively, at room temperature for an applied B 0 of 2.4 T. These values are higher than those for commercial iron oxide-based CAs commonly used in human MR examinations if expressed per particles.
Finally, d-block metal NPs, like Fe or Fe-Co, protected from air and water oxidation, have been tested. These CAs have higher magnetization (typically 185 Am 2 .kg -1 for Fe-based NPs at room temperature) than all SPIO-and ferrite-based particles (typically 60-90 Am 2 .kg −1 at room temperature) [71]. As a consequence, despite their very reduced size, their r 2 and r 2 /r 1 values are expected to be very high (Table 4). Table 3. Main characteristics of iron oxide-based negative CAs. Note, the reported r 2 and r 2 /r 1 values were expressed per mM of iron atoms.

SPIO
An important issue of negative CA design is their surface modification to make them biocompatible for in vitro and in vivo uses. This modification usually involves: (i) the synthesis of hydrophilic polymer brushes from the CA particle surface ('grafting-from' reactions), using different coupling ligands (silane, carboxylate, phosphonate . . . ) [75,76]; (ii) the attachment of preformed polymer brushes at the surface of the particles by different surface reactions (click chemistry, diazonium chemistry . . . ) [77]; (iii) the self-assembly of polymer chains with the magnetic particles, based mainly on electrostatic interactions [78], leading to core@shell [79] or embedded [80,81] morphologies; (iv) the grafting of specific molecules [82,83] or biomolecules [6,84] with targeting ability. This list is, of course, not exhaustive but it gives an idea of the further research directions in the field of negative CAs. The polymers most used are polysaccharides (Dextran, Alginate, Chitosan . . . ), polyethyleneglycols (PEG), polyvinylpyrrolidone (PVP), and polyvinyl alcohol (PVA), polycaprolactone (PCL), polyacrylic acid (PAA), certain polypeptides, and fatty acids, due to their ability to increase the aqueous colloidal stability of the particles [72]. The targeting biomolecules can be specific proteins like transferrin [84] and TRAIL [6], short peptides like RGD [85,86] and Tat [87,88], or just small molecules like folic acid [83] and dopamine [82].
Surface modification may also involve an inorganic coating or embedding with inert and hydrophilic silica [89][90][91] or hydroxyapatite [92], which may also serve as matrix for drugs, or radioisotopes, transforming the initial CAs into theranostic agents.

Optical Imaging Probes
OI applied to living organs or cells has become a fundamental tool for imaging functional lesions in vivo and in vitro. It offers the possibility of real-time spatio-temporal monitoring of biological processes in a non-invasive way. This technique takes advantage of the phenomenon of intrinsic luminescence induced by endogenous biological matter or by exogenous species introduced near the biological structure to be imaged. Interestingly, certain exogenous dyes, in addition, to be able to illuminate a given area of their biological environment, they are able to express its physicochemical state, acting as metabolism marker. Their optical signal may be thus correlated to the variation of biological parameters like the pH and the calcium concentration [93][94][95][96].
Nowadays, OI is routinely used for in vitro observations, thanks mainly to fluorescence microscopy. However, its use for in vivo diagnostics is still in progress and to date, only endoscopic light exciting or detecting systems are used. Indeed, the incident as well the emitted photon may be: (i) reflected by the surface of the tissue; (ii) absorbed by the tissue, losing its energy; (iii) finally, scattered by the tissues. As a consequence, a loss of coherence of the light occurs and it becomes impossible to know where the photons are going out or coming from limiting the OI operating spectral range to the body transparency window namely between 600 and 1000 nm ( Figure 5) [97]. Anyway, technological advances in the field of electronics, with the development of very high-resolution, ultra-sensitive and very small CCD cameras as well as optical fibers allowed the fabrication and the commercialization of endoscopic imaging systems, coming very close to the target tissue and detecting directly, inside the body, the emitted photons. Systems adapted to a conventional bronchoscope, in which the conventional beam is replaced by blue laser light (442 nm), have been thus successfully used to image the lungs and bronchi. The emitted light has been collected by means of an intensified CCD camera, connected to the image beam of an endoscope. Optical imaging devices using endoscopic pathways and allowing the measurement of the autofluorescence of some endogenous porphyrins have been also proposed for urology, gynecology, and otorhinolaryngology [98][99][100]. More recently, two low energy photon excitation technology has been developed, in replacement of the conventional single high energy photon excitation one, making the dye photoexcitation less-harmful and deep-penetrating, even carried out outside the body [101].  Exogenous dyes usually produce light locally after a photonic excitation. Each dye can be defined by its excitation and emission spectra, its lifetime (the time taken by the fluorophore to emit after the cessation of the excitation), and its quantum yield. In practice, the excitation wavelength chosen to trigger the reaction of a specific dye may lead to a 'parasite' luminescence of endogenous dyes, which must be discriminated to avoid OI artifacts. For such a purpose, light information is often processed as a function of time. Indeed, as soon as the excitation ceases, the fluorescence gradually fades, according to a variable time constant which depends on the nature of the fluorophore. If the exogenous dyes exhibit long relaxation times, longer than those of endogenous ones, one has to wait for the extinction of the endogenous fluorescence before opening the camera shutter [102,103]. Reversely, if their relaxation times are shorter, the camera Figure 5. Absorbance of various tissue and blood components from 200 nm to 10 µm. The optical imaging window ranging from 650 to 1450 nm represents the range where tissue penetration is greatest, HbO 2 and Hb referring to oxygenated and deoxygenated hemoglobin. Reproduced from [97] with permission from ACS, 2012.
Exogenous dyes usually produce light locally after a photonic excitation. Each dye can be defined by its excitation and emission spectra, its lifetime (the time taken by the fluorophore to emit after the cessation of the excitation), and its quantum yield. In practice, the excitation wavelength chosen to trigger the reaction of a specific dye may lead to a 'parasite' luminescence of endogenous dyes, which must be discriminated to avoid OI artifacts. For such a purpose, light information is often processed as a function of time. Indeed, as soon as the excitation ceases, the fluorescence gradually fades, according to a variable time constant which depends on the nature of the fluorophore. If the exogenous dyes exhibit long relaxation times, longer than those of endogenous ones, one has to wait for the extinction of the endogenous fluorescence before opening the camera shutter [102,103]. Reversely, if their relaxation times are shorter, the camera shutter must be closed quickly, and the endogenous fluorescence contribution must be subtracted from the recorded signal.
Finally, the quantum yield Φ of an exogenous dye, measures the ratio of the number of photons emitted to the number of photons absorbed during the lifetime of its excited state. It is usually defined by the equation where k r and k nr are the rate constants of its radiative and non-radiative de-excitation after absorption of the incident photon. In practice, it is determined by comparing the absorption and emission spectra of a reference compound, excitable at the same wavelength as the dye and whose emission covers the same range as that of the compound of interest. The higher the Φ, the more efficient the exogenous dye is. In contact with physiological media, Φ may decrease drastically, reaching luminescence extinction. In the case of molecular dyes, this extinction is mainly due to the chemical reactivity of the fluorophores with free oxygen-based radicals. In the case of solid ones, it is much more due to the interaction of their organic coatings with the available bio-organics, leading to photon reabsorption. These differences justify the classification of all the optical imaging probes into two classes: molecular dyes, including organics and lanthanide chelates, and inorganic, mainly solids including PPs, QDs, CDs, DCNPs and UCNPs. However, for our purposes, we will classify all these agents into two main families, Ln-containing and Ln-free dyes.

Ln-Free Dyes
This luminescent agent family is rich. Their major representatives are organic chromophores. Briefly, they consist of natural or synthetic conjugated organic molecules. Their excitation is very fast (10 -17 s) and their excited lifetime is of few nanoseconds [8]. Their radiative relaxation leads to the emission of a photon of energy lower than that of the one absorbed (Stoke displacement). Most of these molecular dyes are commercialized: coumarins, fluoresceins, rhodamines, and cyanines. Their absorption and emission cover the entire near-UV to near-IR spectral range ( Figure 6). Their quantum yields are relatively low, about 10-15% in visible light and 2-4% in IR [104,105].
11, x FOR PEER REVIEW 10 of 34

Ln-Free Dyes
This luminescent agent family is rich. Their major representatives are organic chromophores. Briefly, they consist of natural or synthetic conjugated organic molecules. Their excitation is very fast (10 -17 s) and their excited lifetime is of few nanoseconds [8]. Their radiative relaxation leads to the emission of a photon of energy lower than that of the one absorbed (Stoke displacement). Most of these molecular dyes are commercialized: coumarins, fluoresceins, rhodamines, and cyanines. Their absorption and emission cover the entire near-UV to near-IR spectral range ( Figure 6). Their quantum yields are relatively low, about 10-15% in visible light and 2-4% in IR [104,105]. These last years, dipyromethene borane derivatives, commonly known as bodipy, have emerged. Thanks to their higher quantum yields, approaching 100% for their aza-based compounds (Figure 7), and their emission wavelengths range between 520 and 670 nm [106] with a life-time of some nanoseconds (from 3.9 ns in water to 5.7 ns in methanol), they have attracted a lot of interest. They suffer nevertheless from certain drawbacks: (1) their Stoke displacement is low, making it difficult to filter the scattered incident light (at the excitation wavelength of the system) and the emission signal; and (2) their absorption and emission bands are broad, making it awkward to use different bodipy markers simultaneously [106]. These last years, dipyromethene borane derivatives, commonly known as bodipy, have emerged. Thanks to their higher quantum yields, approaching 100% for their aza-based compounds (Figure 7), and their emission wavelengths range between 520 and 670 nm [106] with a life-time of some nanoseconds (from 3.9 ns in water to 5.7 ns in methanol), they have attracted a lot of interest. They suffer nevertheless from certain drawbacks: (1) their Stoke displacement is low, making it difficult to filter the scattered incident light (at the excitation wavelength of the system) and the emission signal; and (2) their absorption and emission bands are broad, making it awkward to use different bodipy markers simultaneously [106]. methanol), they have attracted a lot of interest. They suffer nevertheless from certain drawbacks: (1) their Stoke displacement is low, making it difficult to filter the scattered incident light (at the excitation wavelength of the system) and the emission signal; and (2) their absorption and emission bands are broad, making it awkward to use different bodipy markers simultaneously [106]. Ln-free dyes class includes nanosolids, PPs and QDs. PPs, mainly gold NPs, exhibit tunable optical properties including Mie scattering, surface plasmonic resonance (SPR), surface-enhanced luminescence and surface Raman scattering. They are used for high-sensitivity and high-resolution optical imaging [107][108][109]. Their surface plasmon resonance which corresponds to the interaction with the light of the free conductive electrons on their surface, is the major feature. It causes enhanced absorption and scattering intensities at the SPR wavelength. The intensity and position of the SPR can be controlled by the size and the shape of the PPs but also by the dielectric constant of their surrounding medium. The SPR wavelength is thus red-shifted with increasing the size of PPs, changing their shape from an isotropic to an anisotropic morphology (nanorods [110], nanostars [110,111]… Figure 8), or coating them with dielectric materials of high refractive index (iron oxide [112,113], polyvinyl pyrrolidone (PVP) [114,115] Ln-free dyes class includes nanosolids, PPs and QDs. PPs, mainly gold NPs, exhibit tunable optical properties including Mie scattering, surface plasmonic resonance (SPR), surface-enhanced luminescence and surface Raman scattering. They are used for highsensitivity and high-resolution optical imaging [107][108][109]. Their surface plasmon resonance which corresponds to the interaction with the light of the free conductive electrons on their surface, is the major feature. It causes enhanced absorption and scattering intensities at the SPR wavelength. The intensity and position of the SPR can be controlled by the size and the shape of the PPs but also by the dielectric constant of their surrounding medium. The SPR wavelength is thus red-shifted with increasing the size of PPs, changing their shape from an isotropic to an anisotropic morphology (nanorods [110], nanostars [110,111] . . . Figure 8), or coating them with dielectric materials of high refractive index (iron oxide [112,113], polyvinyl pyrrolidone (PVP) [114,115] . . . ). Besides, PPs show superior photostability, which means that they can serve as multi-colored optical probes and sensors for in vivo real-time imaging. Additionally, the non-toxic and chemically stable ones, like gold NPs, can be delivered into living organisms non-invasively. They can be easily functionalized by various biotargets thanks to metal-SH bio-conjugation, making them, gold PPs in particular, of primary importance for in vivo and in vitro OI. The last Ln-free dye type is that of QDs. QDs exhibit a strong emission resulting usually from the excitation by photons of energy hex higher than their band gap Eg. This emission proceeds as radiative recombination of the photogenerated hole (in the valence band) and the electron (in the conduction band) pair (exciton). The radiative recombination of excitons can be direct (band-to-band) or indirect. In the latter case, lattice defects or impurities (doping), associated with levels of energy located in the gap, act as recombination centers, and an intermediate step is involved in the return of the excited electron from the conduction band to the valence band of the semiconductor. The surface defects of the nanocrystalline semiconductor can be a preferred site for recombination. One can very well imagine that this surface-related mechanism is important, maybe the most important, to the point of greatly modulating the luminescence properties of this kind of dye. Additionally, due to quantum confinement effect, ultrafine QD bands may split into discrete levels and their band gap increases when the radius of the nanocrystals r decreases, leading to Eg values higher than those of bulk counterparts [116]. The wavelength of the band-to-band radiative de-excitation decreases, making any QD-based biolabeling size-dependent (Table 5, Figure 9). Besides, PPs show superior photostability, which means that they can serve as multicolored optical probes and sensors for in vivo real-time imaging. Additionally, the nontoxic and chemically stable ones, like gold NPs, can be delivered into living organisms non-invasively. They can be easily functionalized by various biotargets thanks to metal-SH bio-conjugation, making them, gold PPs in particular, of primary importance for in vivo and in vitro OI.
The last Ln-free dye type is that of QDs. QDs exhibit a strong emission resulting usually from the excitation by photons of energy hυ ex higher than their band gap E g . This emission proceeds as radiative recombination of the photogenerated hole (in the valence band) and the electron (in the conduction band) pair (exciton). The radiative recombination of excitons can be direct (band-to-band) or indirect. In the latter case, lattice defects or impurities (doping), associated with levels of energy located in the gap, act as recombination centers, and an intermediate step is involved in the return of the excited electron from the conduction band to the valence band of the semiconductor. The surface defects of the nanocrystalline semiconductor can be a preferred site for recombination. One can very well imagine that this surface-related mechanism is important, maybe the most important, to the point of greatly modulating the luminescence properties of this kind of dye. Additionally, due to quantum confinement effect, ultrafine QD bands may split into discrete levels and their band gap increases when the radius of the nanocrystals r decreases, leading to E g values higher than those of bulk counterparts [116]. The wavelength of the band-to-band radiative de-excitation decreases, making any QD-based biolabeling size-dependent (Table 5, Figure 9). Table 5. Main spectroscopic characteristics of luminescent semiconducting nanocrystals for optical fluorescence imaging. ε is the molar extinction coefficient associated with the main absorption, FWHM the width at half height of the emission band and Φ f the quantum yield.
Ref.    In practice, QDs are passivated by hydrophilic organic ligands, to improve the dispersion of the particles in aqueous media and minimize the non-radiative recombination of the photogenerated exciton by interaction with the external environment. It should be noted that organic 'passivation' is not always enough to avoid fluorescence extinction. As an alternative, each QD can be replaced by a semiconductor-semiconductor core-shell counterpart, whose quantum efficiency may exceed 50% [126,[129][130][131]. Of course, the engineered semiconducting hetero-junctions have to be finally surrounded by a hydrophilic organic [132][133][134] or inorganic (mainly silica) [135,136] thin layer to make them soluble in physiological media.
Despite all these very interesting optical properties, QDs have been seldom employed in OI. Their toxic heavy metals composition with the risk of release of these elements makes their use still controversial. Studies are underway to determine the dose effects for each QD family, based on their size, chemical nature, and surface state [137][138][139][140]. Only efficient clearance [141] may promote their application in clinic.
Recently a new class of metal free semiconductive NPs has emerged. It is that of surface-passivated carbon dots (CDs). They have successfully been used as in vitro [16] and in vivo [142] biolabels. To date, CDs are considered as safe materials, non-toxic towards different cell lines even at high concentrations [143].

Ln-Based Dyes
The optical properties of Ln cations derive from the electronic transitions of the excited state, populated under illumination, towards the ground state (f-f transitions), either directly or indirectly by transiting towards excited levels of lower energies ( Figure 10). As a consequence, their emission bands are narrow and of specific energies [144]. Moreover, their excitation life-times are longer, from a few microseconds to several milliseconds, than those of organic chromophores. This property makes it easy to discriminate between their luminescence signal and that of endogenous dyes [145,146]. Despite all these very interesting optical properties, QDs have been seldom employed in OI. Their toxic heavy metals composition with the risk of release of these elements makes their use still controversial. Studies are underway to determine the dose effects for each QD family, based on their size, chemical nature, and surface state [137][138][139][140]. Only efficient clearance [141] may promote their application in clinic.
Recently a new class of metal free semiconductive NPs has emerged. It is that of surface-passivated carbon dots (CDs). They have successfully been used as in vitro [16] and in vivo [142] biolabels. To date, CDs are considered as safe materials, non-toxic towards different cell lines even at high concentrations [143].

Ln-Based Dyes
The optical properties of Ln cations derive from the electronic transitions of the excited state, populated under illumination, towards the ground state (f-f transitions), either directly or indirectly by transiting towards excited levels of lower energies ( Figure  10). As a consequence, their emission bands are narrow and of specific energies [144]. Moreover, their excitation life-times are longer, from a few microseconds to several milliseconds, than those of organic chromophores. This property makes it easy to discriminate between their luminescence signal and that of endogenous dyes [145,146]. As biolabels, they are used as hydrophilic complexes (molecular dyes) or as luminescent ceramic NPs.

Molecular Dyes
Ln cations are chelated to hydrophilic polydentate ligands to form stable complexes, for which the energy of the involved f-f-transitions are weakly affected by either the cation environment or the experimental measurement conditions [144,148]. Timely, to achieve their clinical use, these molecular edifices are upgraded by the covalent grafting of labeling functions [149].
The design of this family of molecular dyes involves also two main improvements, in relation to some intrinsic limitations. Indeed, the photoluminescence intensity of Ln As biolabels, they are used as hydrophilic complexes (molecular dyes) or as luminescent ceramic NPs.

Molecular Dyes
Ln cations are chelated to hydrophilic polydentate ligands to form stable complexes, for which the energy of the involved f-f-transitions are weakly affected by either the cation environment or the experimental measurement conditions [144,148]. Timely, to achieve their clinical use, these molecular edifices are upgraded by the covalent grafting of labeling functions [149].
The design of this family of molecular dyes involves also two main improvements, in relation to some intrinsic limitations. Indeed, the photoluminescence intensity of Ln cations is weak [147]. The electric dipolar f-f transitions, at the origin of the luminescence spectrum, are forbidden by the Laporte and sometimes spin multiplicity rules [149]. It results that Ln 3+ have extremely low extinction coefficients (few units M −1 .cm −1 in the best cases) and their direct excitation through these transitions is difficult and consequently their photoluminescence is weak. This can be overcome by sensitizing the lanthanide cations through a covalent modification of their ligand with some aromatic molecules, called antenna, (Figure 11) having an appropriate electronic structure for fluorescence resonance energy transfer (FRET) [150]. Secondly, the risk of their photoluminescence quenching by the OH, NH, and CH oscillators in their inner coordination sphere is not negligible at all [149,151]. To minimize this effect while obtaining extremely stable water-soluble edifices, Ln coordination must be balanced between the strength of the electrostatic interactions between Ln 3+ and the ligands and steric or electrostatic repulsion interactions between the ligands around the cations, while filling as much as possible the coordination sphere, avoiding water bonding. There are essentially three strategies for such a purpose. The first is to use highly pre-organized ligands featuring macrocycles such as triazacyclononane or 1,4,7,10-tetraazacyclododecane (Cyclen) [152][153][154]. The second strategy is to provide numerous negatively charged functions such as carboxylates or phosphonates [154,155] ( Figure 11).
x FOR PEER REVIEW 14 of 34 cases) and their direct excitation through these transitions is difficult and consequently their photoluminescence is weak. This can be overcome by sensitizing the lanthanide cations through a covalent modification of their ligand with some aromatic molecules, called antenna, (Figure 11) having an appropriate electronic structure for fluorescence resonance energy transfer (FRET) [150]. Secondly, the risk of their photoluminescence quenching by the OH, NH, and CH oscillators in their inner coordination sphere is not negligible at all [149,151]. To minimize this effect while obtaining extremely stable water-soluble edifices, Ln coordination must be balanced between the strength of the electrostatic interactions between Ln 3+ and the ligands and steric or electrostatic repulsion interactions between the ligands around the cations, while filling as much as possible the coordination sphere, avoiding water bonding. There are essentially three strategies for such a purpose. The first is to use highly pre-organized ligands featuring macrocycles such as triazacyclononane or 1,4,7,10-tetraazacyclododecane (Cyclen) [152][153][154]. The second strategy is to provide numerous negatively charged functions such as carboxylates or phosphonates [154,155] (Figure 11). Figure 11. Schematic representation of a luminescent lanthanide label. Reproduced from [149], with permission from RSC Publisher, 2016.

Solid Inorganic Dyes
Ln-doped up-conversion and down-conversion luminescent ceramic nanocrystals (UCNPs and DCNPs) have been seriously considered for OI application. With an appropriate surface modification, they may form an interesting class of inorganic dyes, particularly UCNPs. In practice, they consist of a transparent crystalline host lattice that accommodates the dopants. For high-yield light emission, the host lattice must closely match that of the dopant ions, have low phonon vibration energies, and good chemical stability [156,157]. Based on these criteria, the most commonly used host lattices for the synthesis of UCNPs are fluorides [158][159][160], oxides [161][162][163], and sometimes phosphates [164][165][166] (Table 6). Fluoride matrixes are the most efficient UCNPs [167][168][169]. These are commonly doped with two luminescent cations: the activator, emitting visible light and the sensitizer, absorbing the photoexcitation, and transferring the required energy to the activator (Table 6). To minimize cross-relaxation energy loss, the concentration of the sensitizer is relatively high (~20 mol. %), while that of the activator is low (below 2 mol. %) [169][170][171][172]. In this mechanism, higher-energy photons are emitted by sequential absorption of lower-energy photons ( Figure 12) [17,173], leading to very strong emission and increasing the optical detection sensitivity.

Solid Inorganic Dyes
Ln-doped up-conversion and down-conversion luminescent ceramic nanocrystals (UCNPs and DCNPs) have been seriously considered for OI application. With an appropriate surface modification, they may form an interesting class of inorganic dyes, particularly UCNPs. In practice, they consist of a transparent crystalline host lattice that accommodates the dopants. For high-yield light emission, the host lattice must closely match that of the dopant ions, have low phonon vibration energies, and good chemical stability [156,157]. Based on these criteria, the most commonly used host lattices for the synthesis of UCNPs are fluorides [158][159][160], oxides [161][162][163], and sometimes phosphates [164][165][166] (Table 6). Fluoride matrixes are the most efficient UCNPs [167][168][169]. These are commonly doped with two luminescent cations: the activator, emitting visible light and the sensitizer, absorbing the photoexcitation, and transferring the required energy to the activator (Table 6). To minimize cross-relaxation energy loss, the concentration of the sensitizer is relatively high (~20 mol. %), while that of the activator is low (below 2 mol. %) [169][170][171][172]. In this mechanism, higher-energy photons are emitted by sequential absorption of lower-energy photons ( Figure 12) [17,173], leading to very strong emission and increasing the optical detection sensitivity.   UCPNs exhibit great advantages over other types of fluorescent materials. Thanks to their NIR excitation, they allow enhanced tissue penetration depths. Excitation by the light of such low energy avoids DNA or RNA photo-damage. Moreover, the chemical composition of the host and the nature of the dopant improve their stability against ) ) UCPNs exhibit great advantages over other types of fluorescent materials. Thanks to their NIR excitation, they allow enhanced tissue penetration depths. Excitation by the light of such low energy avoids DNA or RNA photo-damage. Moreover, the chemical composition of the host and the nature of the dopant improve their stability against photobleaching, photoblinking, and photochemical degradation, and significantly reduce their cytotoxicity [17]. Finally, their optical properties do not depend on their size and shape. The only requirement for their in vivo use is that they must be small enough and almost uniform in size. As the particles are small, they are able to diffuse inside the body and to reach the target organs.
Finally, the judicious surface functionalization of UCNPs allows in vitro and in vivo, targeting and molecular events detection, making these objects not only useful for the selective detection of cells, but also for the elucidation of biological processes. For instance, polyethyleneimine-coated NaYF 4 :Yb/Er particles conjugated with folic acid were employed for in vitro imaging of HT29 adenocarcinoma cells and OVCAR3 ovarian carcinoma cells [180]. NaGdF 4 :Er/Yb particles functionalized with heparin and basic fibroblast growth factor (bFGF) molecules were also employed for in vitro imaging of HeLa cells. Heparin molecules not only provided water dispersibility, but their interaction with the growth factor resulted in the required conformation of bFGF to interact with receptors on the cell membrane of epithelial cancer cells, optimizing the targeting abilities of the luminescent nanoprobes [181].

Ln-Based Nanoprobes for Dual Magnetic and Optical Imaging
There are several ways to introduce bimodality in an imaging probe. They depend on whether the magnetic and the luminescent species are distinct, just assembled in a common compartment or if they form a single object. On this basis, there is a first classification of the bimodal probes: probes in which the magnetic and luminescence properties come from two distinct species, brought together only for this purpose, and probes in which the magnetic and luminescence properties come from the same chemical object. In vivo, the former dual probes may be dissociated and impacted by microenvironmental conditions (pH, enzymatic activities), the optical label and the magnetic dipole, becoming tracked separately, making the later cases much more suitable for the desired application. Among these architectures, there are those constructed using micellar chemistry. In practice, luminescent and paramagnetic Ln chelates are assembled within the same micellar structure [182,183] as illustrated in Figure 13. There is also the possibility to interact simultaneously with molecular chromophores and paramagnetic chelates with selected nanocontainers like zeolites [184] or CNTs [185].

Ln-Based Nanoprobes for Dual Magnetic and Optical Imaging
There are several ways to introduce bimodality in an imaging probe. They depend on whether the magnetic and the luminescent species are distinct, just assembled in a common compartment or if they form a single object. On this basis, there is a first classification of the bimodal probes: probes in which the magnetic and luminescence properties come from two distinct species, brought together only for this purpose, and probes in which the magnetic and luminescence properties come from the same chemical object. In vivo, the former dual probes may be dissociated and impacted by microenvironmental conditions (pH, enzymatic activities), the optical label and the magnetic dipole, becoming tracked separately, making the later cases much more suitable for the desired application. Among these architectures, there are those constructed using micellar chemistry. In practice, luminescent and paramagnetic Ln chelates are assembled within the same micellar structure [182,183] as illustrated in Figure 13. There is also the possibility to interact simultaneously with molecular chromophores and paramagnetic chelates with selected nanocontainers like zeolites [184] or CNTs [185]. Let us focus on the 'two-in-one' bifunctional probes. There are several types: all-molecular, hybrids, and all-inorganics. The first sub-family consists of polynuclear complexes, involving both magnetic and luminescent lanthanide cations. They can be used only as positive MRI contrast agents, and their optical properties are mainly of the down-conversion type. The second sub-family is a kind of core-shell structure made from a luminescent core, typically lanthanide-doped UCNPs, and a shell containing paramagnetic Gd 3+ complexes and reversely a magnetic one, mainly iron oxide NPs, and a shell of luminescent Ln 3+ complexes covalently bonded to the former. Depending on their architecture, they can proceed with up-or down-conversion luminescence while leading to positive or negative MRI contrasting. They are again small and offer more versatility for both in vivo and in vitro imaging. Within these sub-classes, there are also dual-Lanthanide-chelated silica particles [186][187][188][189]. The chelates can be embedded into the silica core or just attached to its outer surface. The former configuration is much more appropriate for dual imaging since it allows direct contact between water molecules and paramagnetic cations as it limits light scattering with the silica matrix. Finally, the third sub-family includes two types: (i) doped single crystals like Er-Tr or Ho doped superparamagnetic iron oxide Fe3O4 NPs [190,191], Eu doped paramagnetic gadolinium oxide Gd2O3 NPs [192,193], or Gd co-doped UCNPs and (ii) crystalline hetero-nanostructures like Fe3O4@LaF3:Ce,Tb [194,195], NaYF4:Yb 3+ -Er 3+ @Fe3O4 [196] or Let us focus on the 'two-in-one' bifunctional probes. There are several types: allmolecular, hybrids, and all-inorganics. The first sub-family consists of polynuclear complexes, involving both magnetic and luminescent lanthanide cations. They can be used only as positive MRI contrast agents, and their optical properties are mainly of the downconversion type. The second sub-family is a kind of core-shell structure made from a luminescent core, typically lanthanide-doped UCNPs, and a shell containing paramagnetic Gd 3+ complexes and reversely a magnetic one, mainly iron oxide NPs, and a shell of luminescent Ln 3+ complexes covalently bonded to the former. Depending on their architecture, they can proceed with up-or down-conversion luminescence while leading to positive or negative MRI contrasting. They are again small and offer more versatility for both in vivo and in vitro imaging. Within these sub-classes, there are also dual-Lanthanide-chelated silica particles [186][187][188][189]. The chelates can be embedded into the silica core or just attached to its outer surface. The former configuration is much more appropriate for dual imaging since it allows direct contact between water molecules and paramagnetic cations as it limits light scattering with the silica matrix. Finally, the third sub-family includes two types: (i) doped single crystals like Er-Tr or Ho doped superparamagnetic iron oxide Fe 3 O 4 NPs [190,191], Eu doped paramagnetic gadolinium oxide Gd 2 O 3 NPs [192,193], or Gd co-doped UCNPs and (ii) crystalline hetero-nanostructures like Fe 3 O 4 @LaF 3 :Ce,Tb [194,195], NaYF 4 :Yb 3+ -Er 3+ @Fe 3 O 4 [196] or NaYF 4 :Yb 3+ -Er 3+ @NaGdF 4 [197,198] core-shell NPs. Even if both kinds of probes combine high sensitivity of time-resolved fluorescence and high spatial resolution of MRI, there are some reports pointing out the possibility of a host lattice induced optical quenching in the former [199], making the latter much more studied. There is a vast range of possibilities for this sub-family of probes thanks to the thousands of combinations that can be made between the core and the shell, while remaining the resulting architectures always small in size for improved in-body diffusion.

Molecular Dual Probes
Molecular dual probes consist of polynuclear complexes, in which both luminescent and paramagnetic lanthanides are simultaneously chelated. The design of such architectures is a real challenge for coordination chemists. Whereas the presence of at least one inner-sphere water molecule is required for good MRI efficiency, it is not for non-quenched luminescence imaging [200]. Some synthetic strategies were nevertheless proposed to overcome this drawback. A versatile pyridine-based scaffold for Ln 3+ complexation was for instance constructed (Figure 14a,b) [201]. In this structure, the paramagnetic Ln cations, particularly Gd 3+ , are bishydrated, giving good MRI efficacy (a relaxivity r 1 of 6.21 mM −1 ·s −1 at 500 MHz and 25 • C per Gd 3+ ) while the luminescent ones, particularly Nd 3+ , are sensitized by the aromatic pyridine moieties, tacking advantage from the resulting 'antenna effect' to improve their photoconversion rate, compensating the quenching caused by the two inner-sphere waters.
particularly Gd 3+ , are bishydrated, giving good MRI efficacy (a relaxivity r1 of 6.21 mM −1 ·s −1 at 500 MHz and 25 °C per Gd 3+ ) while the luminescent ones, particularly Nd 3+ , are sensitized by the aromatic pyridine moieties, tacking advantage from the resulting 'antenna effect' to improve their photoconversion rate, compensating the quenching caused by the two inner-sphere waters.
Other derivatives of these pyridine-based ligands were prepared for the same purpose. They were obtained by extending the pyridine group with a triazole ring (Figure 14c,d) [151] or by replacing it by an isoquinoline (Figure 14e,f) [202]. The former compounds gave quantum yields of 0.01% and 0.02% for the NIR-emitting Nd(III) and Yb(III) complexes, respectively, while the latter gave slightly higher yields of 0.013-0.016% and 0.028-0.040% for the same metal cations. In all cases the relaxivity of Gd(III) complexes was of the same magnitude, ranging from 6 to 8 mM −1 .s −1 at 20 MHz and 37 °C. NIR emission could also be observed after complexation of the tripodal hydroxyquinolinate ligand (Figure 14g) to Nd(III) or Yb(III) or that based on triazacyclononane and 8-hydroxyquinolinate/phenolate binding units (Figure 14h), which are considered to be good sensitizers for luminescent lanthanide cations, leading to quantum yields around 0.02%, as for the pyridine-based complexes. Interestingly, these two ligands are not equivalent in their MRI applications. Indeed, while the latter has a large r1 value for its bishydrate Gd(IIII) complexes of about 9.1 mM -1 .s -1 at 20 MHz, as a result of a long rotational correlation time, fast water exchange and slow electronic relaxation, the former has not, due to a slow water exchange [203].  Other derivatives of these pyridine-based ligands were prepared for the same purpose. They were obtained by extending the pyridine group with a triazole ring (Figure 14c,d) [151] or by replacing it by an isoquinoline (Figure 14e,f) [202]. The former compounds gave quantum yields of 0.01% and 0.02% for the NIR-emitting Nd(III) and Yb(III) complexes, respectively, while the latter gave slightly higher yields of 0.013-0.016% and 0.028-0.040% for the same metal cations. In all cases the relaxivity of Gd(III) complexes was of the same magnitude, ranging from 6 to 8 mM −1 .s −1 at 20 MHz and 37 • C. NIR emission could also be observed after complexation of the tripodal hydroxyquinolinate ligand (Figure 14g) to Nd(III) or Yb(III) or that based on triazacyclononane and 8-hydroxyquinolinate/phenolate binding units (Figure 14h), which are considered to be good sensitizers for luminescent lanthanide cations, leading to quantum yields around 0.02%, as for the pyridine-based complexes. Interestingly, these two ligands are not equivalent in their MRI applications. Indeed, while the latter has a large r 1 value for its bishydrate Gd(IIII) complexes of about 9.1 mM −1 .s −1 at 20 MHz, as a result of a long rotational correlation time, fast water exchange and slow electronic relaxation, the former has not, due to a slow water exchange [203].
Encouraging results were also obtained by replacing linear polydentate ligands with cyclic ones, involving aromatic moieties as bridging ligands (Figure 15) [204][205][206][207][208][209]. The resulting macromolecule exhibits a characteristic slow tumbling with high longitudinal relaxivity, due to the formation of nanosized aggregates. Their luminescence quantum yields remain relevant taking advantage of the antenna effects of their aromatic moieties.

11, x FOR PEER REVIEW 18 of 34
Encouraging results were also obtained by replacing linear polydentate ligands with cyclic ones, involving aromatic moieties as bridging ligands (Figure 15) [204][205][206][207][208][209]. The resulting macromolecule exhibits a characteristic slow tumbling with high longitudinal relaxivity, due to the formation of nanosized aggregates. Their luminescence quantum yields remain relevant taking advantage of the antenna effects of their aromatic moieties. To date, the given above polydentate ligands have been principally bonded to only one type of lanthanide cations, only luminescents or only paramagnetics. The combination of the two types of cations would provide cocktail of paramagnetic/luminescent Ln 3+ complexes, assumed to be good candidates for the development of efficient bimodal agents. Very promising results have been obtained by using these ligands for f-d heteropolymetallic complexes, like Gd 3+ and Ru 2+ or Gd 3+ and Ti 4+ ones [210,211].
Respecting this general chemical approach, more specific architectures started to be proposed these last 5-6 years, based on the spatial separation of the luminescent and the paramagnetic cation compartments. The f-f molecular architecture built by Debroye et al. and abbreviated as (GdL)3Eu, is a good example (Figure 16). In this structure, three DTPA have been chosen as Gd 3+ chelating units, and have been linked to a central Eu 3+ chelate, consisting of a para-substituted pyridine-2,6-dicarboxylate derivative, via an amide bond, achieving bright-red luminescence with a quantum yield of 10% and an enhanced longitudinal relaxation rate of 31 mM −1 .s −1 per molecule at 40 MHz in water at 310 K [212]. To date, the given above polydentate ligands have been principally bonded to only one type of lanthanide cations, only luminescents or only paramagnetics. The combination of the two types of cations would provide cocktail of paramagnetic/luminescent Ln 3+ complexes, assumed to be good candidates for the development of efficient bimodal agents. Very promising results have been obtained by using these ligands for f-d heteropolymetallic complexes, like Gd 3+ and Ru 2+ or Gd 3+ and Ti 4+ ones [210,211].
Respecting this general chemical approach, more specific architectures started to be proposed these last 5-6 years, based on the spatial separation of the luminescent and the paramagnetic cation compartments. The f-f molecular architecture built by Debroye et al. and abbreviated as (GdL) 3 Eu, is a good example (Figure 16). In this structure, three DTPA have been chosen as Gd 3+ chelating units, and have been linked to a central Eu 3+ chelate, consisting of a para-substituted pyridine-2,6-dicarboxylate derivative, via an amide bond, achieving bright-red luminescence with a quantum yield of 10% and an enhanced longitudinal relaxation rate of 31 mM −1 .s −1 per molecule at 40 MHz in water at 310 K [212]. This spatial separation strategy was also successfully employed to build a new class of dual probes that of lanthanide complex dendrimer conjugates. The paramagnetic and luminescence properties of Dy(III) and Yb(III) were thus exploited by chelating them with the all-oxygen-donor-hexadentate ligand TREN-bis(1-Methyl)-3,2-HOPO-TAM-NX (X = 1, 2, or 3), in which TREN, HOPO, and TAM correspond to tris(2-aminoethyl)amine, hydroxypyridinonate, and tris(hydroxyethyl)aminomethane, respectively. The resulting complexes were subsequently conjugated to the esteramide dendrimer ( Figure 17) to improve bioavailability, solubility, and relaxivity [213]. In these structures, all the lanthanide ions are peripheral, accessible to free water. Since they are also bishydrated their MRI efficacy is good. The large scaffolding and mass of the dendrimer, to which up to eight complexes may be covalently conjugated, increases the tumbling time of the complex and contributes also to the improvement of the MRI contrast. The r1 value was 7.60 mM −1 .s −1 at 27 °C for an applied static field of 1.41 T. The presence of water in the inner sphere of the Yb 3+ complexes, of course, reduces their NIR luminescence quantum yield, but does not extinguish it. In mouse serum, the value was 0.17% [213]. Despite their ~40 kDa molecular weight, these probes have been excreted by glomerular filtration over several days making the engineered probes quite safe [213]. Figure 17. Structures of (a) Ln(III) complexes for MRI and NIR imaging agents and that of (b) the esteramide dendrimer scaffold. Reproduced from [213], with permission from ACS Publisher, 2012. This spatial separation strategy was also successfully employed to build a new class of dual probes that of lanthanide complex dendrimer conjugates. The paramagnetic and luminescence properties of Dy(III) and Yb(III) were thus exploited by chelating them with the all-oxygen-donor-hexadentate ligand TREN-bis(1-Methyl)-3,2-HOPO-TAM-NX (X = 1, 2, or 3), in which TREN, HOPO, and TAM correspond to tris(2-aminoethyl)amine, hydroxypyridinonate, and tris(hydroxyethyl)aminomethane, respectively. The resulting complexes were subsequently conjugated to the esteramide dendrimer ( Figure 17) to improve bioavailability, solubility, and relaxivity [213]. In these structures, all the lanthanide ions are peripheral, accessible to free water. Since they are also bishydrated their MRI efficacy is good. The large scaffolding and mass of the dendrimer, to which up to eight complexes may be covalently conjugated, increases the tumbling time of the complex and contributes also to the improvement of the MRI contrast. The r 1 value was 7.60 mM −1 .s −1 at 27 • C for an applied static field of 1.41 T. The presence of water in the inner sphere of the Yb 3+ complexes, of course, reduces their NIR luminescence quantum yield, but does not extinguish it. In mouse serum, the value was 0.17% [213]. Despite their~40 kDa molecular weight, these probes have been excreted by glomerular filtration over several days making the engineered probes quite safe [213]. This spatial separation strategy was also successfully employed to build a new class of dual probes that of lanthanide complex dendrimer conjugates. The paramagnetic and luminescence properties of Dy(III) and Yb(III) were thus exploited by chelating them with the all-oxygen-donor-hexadentate ligand TREN-bis(1-Methyl)-3,2-HOPO-TAM-NX (X = 1, 2, or 3), in which TREN, HOPO, and TAM correspond to tris(2-aminoethyl)amine, hydroxypyridinonate, and tris(hydroxyethyl)aminomethane, respectively. The resulting complexes were subsequently conjugated to the esteramide dendrimer ( Figure 17) to improve bioavailability, solubility, and relaxivity [213]. In these structures, all the lanthanide ions are peripheral, accessible to free water. Since they are also bishydrated their MRI efficacy is good. The large scaffolding and mass of the dendrimer, to which up to eight complexes may be covalently conjugated, increases the tumbling time of the complex and contributes also to the improvement of the MRI contrast. The r1 value was 7.60 mM −1 .s −1 at 27 °C for an applied static field of 1.41 T. The presence of water in the inner sphere of the Yb 3+ complexes, of course, reduces their NIR luminescence quantum yield, but does not extinguish it. In mouse serum, the value was 0.17% [213]. Despite their ~40 kDa molecular weight, these probes have been excreted by glomerular filtration over several days making the engineered probes quite safe [213]. Figure 17. Structures of (a) Ln(III) complexes for MRI and NIR imaging agents and that of (b) the esteramide dendrimer scaffold. Reproduced from [213], with permission from ACS Publisher, 2012. Figure 17. Structures of (a) Ln(III) complexes for MRI and NIR imaging agents and that of (b) the esteramide dendrimer scaffold. Reproduced from [213], with permission from ACS Publisher, 2012.

Ln(III) complex a) b)
The last strategy for engineering all-molecular bimodal probes is based on organic light-emitting diode (OLED) and lighting device technology. It uses oligomers in which specific lanthanide binding sites can be attached to a peripheral structure (Figure 18a) or connected in series through bridging ligands (Figure 18b) along the oligomer chains. The general idea of this strategy consists in exploiting well identified organic polymer scaffolds for the selective sequestration of luminescent and paramagnetic lanthanides, separating them from each other, and keeping the water molecules bonded to the paramagnetic ones away from the complexation sphere of the luminescent ones [214]. Oligomeric ligands, L, were obtained by coupling two tridentate 2,6-bis(benzimidazole-2-yl)pyridine binding units at the 1 and 4 positions of a rigid phenyl spacer ( Figure 19). Because of the small number of torsional degrees of freedom imposed by the polyaromatic scaffold, L was reacted with Ln(hfac) 3 salts, where hfac-is the bidentate hexafluoroacetylacetonate anion, leading to a linear saturated single-stranded complex [Ln 2 (L)(hfac) 6 ]. In these architectures, there are no water molecules directly attached to the lanthanide centers. The hydrophilic character of the ligands involved, particularly hfac, allowed their indirect attachment through hydrogen bonding, conserving a certain magnetic effect of the chelated paramagnetic cations on the surrounding water molecules.
The last strategy for engineering all-molecular bimodal probes is based on organic light-emitting diode (OLED) and lighting device technology. It uses oligomers in which specific lanthanide binding sites can be attached to a peripheral structure (Figure 18a) or connected in series through bridging ligands (Figure 18b) along the oligomer chains. The general idea of this strategy consists in exploiting well identified organic polymer scaffolds for the selective sequestration of luminescent and paramagnetic lanthanides, separating them from each other, and keeping the water molecules bonded to the paramagnetic ones away from the complexation sphere of the luminescent ones [214]. Oligomeric ligands, L, were obtained by coupling two tridentate 2,6-bis(benzimidazole-2-yl)pyridine binding units at the 1 and 4 positions of a rigid phenyl spacer (Figure 19). Because of the small number of torsional degrees of freedom imposed by the polyaromatic scaffold, L was reacted with Ln(hfac)3 salts, where hfac-is the bidentate hexafluoroacetylacetonate anion, leading to a linear saturated single-stranded complex [Ln2(L)(hfac)6]. In these architectures, there are no water molecules directly attached to the lanthanide centers. The hydrophilic character of the ligands involved, particularly hfac, allowed their indirect attachment through hydrogen bonding, conserving a certain magnetic effect of the chelated paramagnetic cations on the surrounding water molecules.

Hybrid Probes
Hybrid probes are produced by combining a luminescent inorganic particle with molecular paramagnetic chelates or a paramagnetic or superparamagnetic inorganic particle with molecular luminescent chelates, in a core-shell type structure.
In the first case, Ln-doped UCNP cores have been conjugated to Gd complexes, as reported by Carron et al. Citrate-capped 20 nm Yb 3+ -and Tm 3+ -doped NaGdF4 particles were reacted with a mono-amino derivative of Gd-DOTA, leading to multiple paramagnetic surface centers, with improved tumbling time and longitudinal relaxivity. The authors reported a r1 value of 25 mM −1 .s −1 per Gd 3+ ion at 60 MHz and 310 K, seven times larger than that of the Gd-DOTA precursor, which is 3.23 mM −1 .s −1 under the same conditions [217]. Moreover, the excitation of the nanoconstructs in the water at 980 nm L1 L2 L3 Figure 18. Two different strategies for the introduction of lanthanide binding sites into molecular multifunctional probes. The chelating units are (a) connected to the periphery (type I) or (b) in series through bridging ligands (type II). Reproduced from [215] and [216], with permission from ACS, 2008, and RCS, 2011, respectively.
The last strategy for engineering all-molecular bimodal probes is based on organic light-emitting diode (OLED) and lighting device technology. It uses oligomers in which specific lanthanide binding sites can be attached to a peripheral structure (Figure 18a) or connected in series through bridging ligands (Figure 18b) along the oligomer chains. The general idea of this strategy consists in exploiting well identified organic polymer scaffolds for the selective sequestration of luminescent and paramagnetic lanthanides, separating them from each other, and keeping the water molecules bonded to the paramagnetic ones away from the complexation sphere of the luminescent ones [214]. Oligomeric ligands, L, were obtained by coupling two tridentate 2,6-bis(benzimidazole-2-yl)pyridine binding units at the 1 and 4 positions of a rigid phenyl spacer (Figure 19). Because of the small number of torsional degrees of freedom imposed by the polyaromatic scaffold, L was reacted with Ln(hfac)3 salts, where hfac-is the bidentate hexafluoroacetylacetonate anion, leading to a linear saturated single-stranded complex [Ln2(L)(hfac)6]. In these architectures, there are no water molecules directly attached to the lanthanide centers. The hydrophilic character of the ligands involved, particularly hfac, allowed their indirect attachment through hydrogen bonding, conserving a certain magnetic effect of the chelated paramagnetic cations on the surrounding water molecules.

Hybrid Probes
Hybrid probes are produced by combining a luminescent inorganic particle with molecular paramagnetic chelates or a paramagnetic or superparamagnetic inorganic particle with molecular luminescent chelates, in a core-shell type structure.
In the first case, Ln-doped UCNP cores have been conjugated to Gd complexes, as reported by Carron et al. Citrate-capped 20 nm Yb 3+ -and Tm 3+ -doped NaGdF4 particles were reacted with a mono-amino derivative of Gd-DOTA, leading to multiple paramagnetic surface centers, with improved tumbling time and longitudinal relaxivity. The authors reported a r1 value of 25 mM −1 .s −1 per Gd 3+ ion at 60 MHz and 310 K, seven times larger than that of the Gd-DOTA precursor, which is 3.23 mM −1 .s −1 under the same conditions [217]. Moreover, the excitation of the nanoconstructs in the water at 980 nm resulted in an intense up-converted emission of Tm(III) at 800 nm. To increase the L1 L2 L3 Figure 19. Structure of ligands L resulting from Suzuki-Miyaura coupling of two tridentate 2,6bis(benzimidazole-2-yl)pyridine binding units at the 1 and 4 positions of a rigid phenyl spacer. Reproduced from [214], with permission from ACS, 2011.

Hybrid Probes
Hybrid probes are produced by combining a luminescent inorganic particle with molecular paramagnetic chelates or a paramagnetic or superparamagnetic inorganic particle with molecular luminescent chelates, in a core-shell type structure.
In the first case, Ln-doped UCNP cores have been conjugated to Gd complexes, as reported by Carron et al. Citrate-capped 20 nm Yb 3+ -and Tm 3+ -doped NaGdF 4 particles were reacted with a mono-amino derivative of Gd-DOTA, leading to multiple paramagnetic surface centers, with improved tumbling time and longitudinal relaxivity. The authors reported a r 1 value of 25 mM −1 .s −1 per Gd 3+ ion at 60 MHz and 310 K, seven times larger than that of the Gd-DOTA precursor, which is 3.23 mM −1 .s −1 under the same conditions [217]. Moreover, the excitation of the nanoconstructs in the water at 980 nm resulted in an intense up-converted emission of Tm(III) at 800 nm. To increase the chemical stability and the biocompatibility of this type of dual probes silica coating was used. Typically, UCNPs were first prepared and then coated by a silica thin layer embedding Gd-DTTA complexes as illustrated in Figure 20 [19]. chemical stability and the biocompatibility of this type of dual probes silica coating was used. Typically, UCNPs were first prepared and then coated by a silica thin layer embedding Gd-DTTA complexes as illustrated in Figure 20 [19]. Conversely, about 10 nm-sized USPIOs particles were decorated with Eu(III) ions encapsulated in a DO3A organic scaffold. The organic ligand was modified to bear an ethoxysilane group which allows its attachment by sol-gel chemistry to the surface of an iron oxide crystal [218]. The red luminescence emission of Eu(III) was found at 614 nm. Relaxometric studies showed a r2 and a r2/r1 value of 114.8 mM -1 .s -1 and 18.9 (per iron atom), respectively, at 60 MHz and 27 °C in water, making them particularly valuable for negative MRI contrasting.
To improve the biocompatibility of this type of bimodal probes, mesoporous silica embedding of the iron oxide NPs was achieved, the silica matrix containing the luminescent lanthanide complexes. Two morphologies were largely explored: magnetic multicore [219] and single core [220] morphologies (Figure 21), the latter being much more suitable than the former for the desired application due to the strong silica photo-scattering. By confining the lanthanide-based molecular dyes at the extreme surface of the silica shell, photo-scattering may be avoided or at least minimized improving thus the OI capability of the resulting dual probes. Moreover, the size of the latter is significantly reduced compared to that of the former, meaning better in-body diffusion after intravenous administration. Of course, within these embedded multifunctional nanostructures, superparamagnetic iron oxide NPs can be replaced by perovskite magnetite or spinel ferrite oxide NPs. Conversely, about 10 nm-sized USPIOs particles were decorated with Eu(III) ions encapsulated in a DO3A organic scaffold. The organic ligand was modified to bear an ethoxysilane group which allows its attachment by sol-gel chemistry to the surface of an iron oxide crystal [218]. The red luminescence emission of Eu(III) was found at 614 nm. Relaxometric studies showed a r 2 and a r 2 /r 1 value of 114.8 mM −1 .s −1 and 18.9 (per iron atom), respectively, at 60 MHz and 27 • C in water, making them particularly valuable for negative MRI contrasting.
To improve the biocompatibility of this type of bimodal probes, mesoporous silica embedding of the iron oxide NPs was achieved, the silica matrix containing the luminescent lanthanide complexes. Two morphologies were largely explored: magnetic multicore [219] and single core [220] morphologies (Figure 21), the latter being much more suitable than the former for the desired application due to the strong silica photo-scattering. By confining the lanthanide-based molecular dyes at the extreme surface of the silica shell, photo-scattering may be avoided or at least minimized improving thus the OI capability of the resulting dual probes. Moreover, the size of the latter is significantly reduced compared to that of the former, meaning better in-body diffusion after intravenous administration. Of course, within these embedded multifunctional nanostructures, superparamagnetic iron oxide NPs can be replaced by perovskite magnetite or spinel ferrite oxide NPs. chemical stability and the biocompatibility of this type of dual probes silica coating was used. Typically, UCNPs were first prepared and then coated by a silica thin layer embedding Gd-DTTA complexes as illustrated in Figure 20 [19]. Conversely, about 10 nm-sized USPIOs particles were decorated with Eu(III) ions encapsulated in a DO3A organic scaffold. The organic ligand was modified to bear an ethoxysilane group which allows its attachment by sol-gel chemistry to the surface of an iron oxide crystal [218]. The red luminescence emission of Eu(III) was found at 614 nm. Relaxometric studies showed a r2 and a r2/r1 value of 114.8 mM -1 .s -1 and 18.9 (per iron atom), respectively, at 60 MHz and 27 °C in water, making them particularly valuable for negative MRI contrasting.
To improve the biocompatibility of this type of bimodal probes, mesoporous silica embedding of the iron oxide NPs was achieved, the silica matrix containing the luminescent lanthanide complexes. Two morphologies were largely explored: magnetic multicore [219] and single core [220] morphologies (Figure 21), the latter being much more suitable than the former for the desired application due to the strong silica photo-scattering. By confining the lanthanide-based molecular dyes at the extreme surface of the silica shell, photo-scattering may be avoided or at least minimized improving thus the OI capability of the resulting dual probes. Moreover, the size of the latter is significantly reduced compared to that of the former, meaning better in-body diffusion after intravenous administration. Of course, within these embedded multifunctional nanostructures, superparamagnetic iron oxide NPs can be replaced by perovskite magnetite or spinel ferrite oxide NPs.

All-Inorganic Dual Probes
The last category of single objects for dual MRI and OI applications consists of all-inorganic particles. These are of two types, single crystals, combining both optical and

All-Inorganic Dual Probes
The last category of single objects for dual MRI and OI applications consists of allinorganic particles. These are of two types, single crystals, combining both optical and magnetic properties, or core-shell crystalline structures concentrating each property in one compartment, the core or the shell.

Single Nanocrystals
It is possible to have light sources and magnetic dipoles in a same matrix material by inserting the desired components into a same nanocrystal. These probes are expected to offer high photostability, a narrow emission band, and a broad absorption band, combining the high sensitivity of time-resolved fluorescence with the high spatial resolution of MRI. However, to date, most of the bifunctional nanocrystals synthesized to consist of downconversion emitters and iron oxide NPs. In addition to the fluorescence quenching by iron oxide NPs, down-conversion fluorescence has some intrinsic limitations, such as autofluorescence, low light penetration into the biological tissue, and photon damage risks to the biological specimen. Tb-doped γ-Fe 2 O 3 nanocrystals, combining superparamagnetism and luminescence, were prepared and evaluated [221]. Functionalized by amino groups they were dispersed in water as a stable colloid. They exhibited green photoluminescence but only under UV excitation (235 nm). This feature prevents the use of such probes for magnetic-fluorescent bimodal in vivo imaging and restricts them to in vitro imaging. The same conclusions can be made concerning Dextran-coated Eu-doped (5 mol%) ultrasmall iron oxide nanocrystals (hydrodynamic diameter between 20 and 40 nm) [191]. Indeed, despite interesting superparamagnetic behavior with relatively high r 1 and r 2 relaxivity values (15.4 and 33.9 mM −1 .s −1 in the water at 0.47 T and 37 • C, respectively), their red photoluminescence can not be used for in vivo and in vitro optical imaging, since it can not be activated without a UV excitation (254 nm).
As an alternative, (Gd, Yb, Tb)PO 4 nanocrystals were investigated as up-conversion systems. They exhibit ultraviolet, blue, and green up-conversion emissions upon excitation with a 980 nm continuous wave laser diode [227]. They are also efficient T 2 -weighted contrast agents with a r 2 /r 1 relaxivity ratio (per Gd atom) larger than 2, between 11 and 12 at 20 MHz and 300 K in water for free particles, and around 22 when the particles are coated with Dextran. These results suggest that the lack of water molecules into the Gd inner coordination sphere, capable of exchanging efficiently with the bulk water, leads to inefficient T 1 relaxation. In other words, the T 2 -relaxation process, which has a strong outer-sphere contribution from field inhomogeneities created by the magnetized particles, that the water protons experience as they diffuse nearby, appears to be more efficient, particularly for the Dextran-coated particles. These results are very promising and open real opportunities for efficient dual MRI and OI applications.
Other paramagnetic UCNPs were successfully employed as dual probes. Typically, ultrasmall paramagnetic Gd 2 O 3 oxide nanocrystals co-doped with luminescent Ln cations like Gd 2 O 3 :Yb 3+ -Er 3+ [222,228], Gd 2 O 3 :Yb 3+ -Ho 3+ [228], and Gd 2 O 3 :Yb 3+ -Tm 3+ [228] proved to be suitable for the desired application. All exhibit strong visible light emission after NIR excitation ( Figure 22) and relatively high longitudinal relaxivity, ranging from 13 to 16 mM −1 .s −1 (per Gd atom), in water at 3.0 T and 310 K. The r 1 value of these contrast agents is proportional to the number of hydration water molecules, which corresponds directly to the number of surface unpaired electrons of Gd 3+ ions. Compared to the standard Gd-DTPA agent, which has only one Gd 3+ ion coordinated to only one water molecule, the Gd 2 O 3 :Yb 3+ -Ln 3+ particles have all their surface Gd 3+ ions available to bind several water molecules. Within the same strategy, co-doped diamagnetic nanocrystals like LaF3:Yb 3+ -Ho 3+ were also used. They offer both up-conversion properties and MRI contrast capabilities. Their measured in water r1 and r2 reached a value of 0.12 and 28.18 mM −1 .s −1 per Ho, at 11.0 T (500 MHz) and 27 °C [229].
Focusing on UCNP cores, hetero-nanostructures have been also built by growing epitaxially a paramagnetic shell of the same crystallographic type. The first reported system is that of NaYF4:Yb 3+ -Er 3+ @NaGdF4 nanoconstructs [197,198], which have the advantage of less surface optical quenching thanks to the spatial isolation of the luminescent core from its environments. Similar architectures were prepared based on singly Er-doped paramagnetic NaGdF4:Er 3+ @NaGdF4:Er 3+ core-shell particles [232]. By adjusting the Er concentration in the shell and in the core, for example, 10% in the core and 12 at % in the shell, the active shell may play the role of a sensitizer for the luminescence of the Ln 3+ cations in the core (Figure 25). Green and red emissions were thus obtained under a 1540 nm excitation thanks to this process [232]. Of course, the availability of Gd 3+ ions at the probe surface makes the engineered core-shell particles valuable for inner-sphere magnetic interaction with water molecules and, therefore, efficient positive MRI contrasting. Figure 25. Energy transfers and de-excitations involved in NaGdF4:12%Er 3+ @NaGdF4:10%Er 3+ active-core/active-shell NPs under 1540 nm excitation. Reproduced from [232], with permission from Beilstein-Institut Publisher, 2017.

Conclusions
After introducing the basic principles of MRI and OI and describing the various probes used to enhance image contrast in one case and provide a light source in the other . TEM images of (a) NaYF 4 :Yb 3+ -Er 3+ @Fe 3 O 4 and (b) NaYF 4 :Eu 3+ @Fe 2 O 3 particles. Reproduced from [196] and [231], with permissions from John Wiley and Sons, 2014, and MDPI, 2020, respectively.
Focusing on UCNP cores, hetero-nanostructures have been also built by growing epitaxially a paramagnetic shell of the same crystallographic type. The first reported system is that of NaYF 4 :Yb 3+ -Er 3+ @NaGdF 4 nanoconstructs [197,198], which have the advantage of less surface optical quenching thanks to the spatial isolation of the luminescent core from its environments. Similar architectures were prepared based on singly Er-doped paramagnetic NaGdF 4 :Er 3+ @NaGdF 4 :Er 3+ core-shell particles [232]. By adjusting the Er concentration in the shell and in the core, for example, 10% in the core and 12 at % in the shell, the active shell may play the role of a sensitizer for the luminescence of the Ln 3+ cations in the core ( Figure 25). Green and red emissions were thus obtained under a 1540 nm excitation thanks to this process [232]. Of course, the availability of Gd 3+ ions at the probe surface makes the engineered core-shell particles valuable for inner-sphere magnetic interaction with water molecules and, therefore, efficient positive MRI contrasting. Figure 24. TEM images of (a) NaYF4:Yb 3+ -Er 3+ @Fe3O4 and (b) NaYF4:Eu 3+ @Fe2O3 particles. Reproduced from [196] and [231], with permissions from John Wiley and Sons, 2014, and MDPI, 2020, respectively.
Focusing on UCNP cores, hetero-nanostructures have been also built by growing epitaxially a paramagnetic shell of the same crystallographic type. The first reported system is that of NaYF4:Yb 3+ -Er 3+ @NaGdF4 nanoconstructs [197,198], which have the advantage of less surface optical quenching thanks to the spatial isolation of the luminescent core from its environments. Similar architectures were prepared based on singly Er-doped paramagnetic NaGdF4:Er 3+ @NaGdF4:Er 3+ core-shell particles [232]. By adjusting the Er concentration in the shell and in the core, for example, 10% in the core and 12 at % in the shell, the active shell may play the role of a sensitizer for the luminescence of the Ln 3+ cations in the core (Figure 25). Green and red emissions were thus obtained under a 1540 nm excitation thanks to this process [232]. Of course, the availability of Gd 3+ ions at the probe surface makes the engineered core-shell particles valuable for inner-sphere magnetic interaction with water molecules and, therefore, efficient positive MRI contrasting. Figure 25. Energy transfers and de-excitations involved in NaGdF4:12%Er 3+ @NaGdF4:10%Er 3+ active-core/active-shell NPs under 1540 nm excitation. Reproduced from [232], with permission from Beilstein-Institut Publisher, 2017.

Conclusions
After introducing the basic principles of MRI and OI and describing the various probes used to enhance image contrast in one case and provide a light source in the other case, we have reviewed the different nanometric magnetic and luminescent architectures that can be used simultaneously for bimodal imaging, focusing on the lanthanide luminescence phenomenon within single objects. We have classified the corresponding bimodal probes into three families, the all-molecular ones, the hybrids, and the . Energy transfers and de-excitations involved in NaGdF 4 :12%Er 3+ @NaGdF 4 :10%Er 3+ active-core/active-shell NPs under 1540 nm excitation. Reproduced from [232], with permission from Beilstein-Institut Publisher, 2017.

Conclusions
After introducing the basic principles of MRI and OI and describing the various probes used to enhance image contrast in one case and provide a light source in the other case, we have reviewed the different nanometric magnetic and luminescent architectures that can be used simultaneously for bimodal imaging, focusing on the lanthanide luminescence phenomenon within single objects. We have classified the corresponding bimodal probes into three families, the all-molecular ones, the hybrids, and the all-inorganic ones. Each of them has advantages and drawbacks for in vivo dual imaging but, in all the cases, their optical properties remain excellent, particularly when they involve the up-conversion feature. Their fluorescence intensity, brightness, photostability, the width of excitation wavelength window and narrow emission window, etc. are often good, making them very efficient biomarkers. This is particularly true when they can be excited by NIR light to avoid tissue scattering and limit luminescence quenching in an aqueous environment. The engineered probes may also be efficient MRI contrast agents, taking advantage of high molecular weight and then a long tumbling time, a high exposed surface and therefore a large number of interacting water molecules, a high concentration of paramagnetic cations in a small volume which gives high field inhomogeneity, and other factors reinforcing the contrast. Finally, dual imaging contrast agents could provide diagnostic information at the early stages of certain diseases and avoid invasive procedures, since they overcome the low sensitivity of MRI and the low detection limit of OI.