Engineering of inorganic nanoparticles as magnetic resonance imaging contrast agents

Abstract

Magnetic resonance imaging (MRI) is a highly valuable non-invasive imaging tool owing to its exquisite soft tissue contrast, high spatial resolution, lack of ionizing radiation, and wide clinical applicability. Contrast agents (CAs) can be used to further enhance the sensitivity of MRI to obtain information-rich images. Recently, extensive research efforts have been focused on the design and synthesis of high-performance inorganic nanoparticle-based CAs to improve the quality and specificity of MRI. Herein, the basic rules, including the choice of metal ions, effect of electron motion on water relaxation, and involved mechanisms, of CAs for MRI have been elucidated in detail. In particular, various design principles, including size control, surface modification (e.g. organic ligand, silica shell, and inorganic nanolayers), and shape regulation, to impact relaxation of water molecules have been discussed in detail. Comprehensive understanding of how these factors work can guide the engineering of future inorganic nanoparticles with high relaxivity. Finally, we have summarized the currently available strategies and their mechanism for obtaining high-performance CAs and discussed the challenges and future developments of nanoparticulate CAs for clinical translation in MRI.

1. Introduction

Diagnostic imaging techniques to detect lesion information (e.g., type, location, and stage) are currently playing diverse roles in the combat against serious diseases such as cancer.^1-3 Among all existing imaging techniques, magnetic resonance imaging (MRI) is a popular choice because of its exquisite soft tissue contrast, high spatial resolution, lack of ionizing radiation, unlimited signal penetration depth, and wide clinical applicability.⁴ Contrast agents (CAs) can be used to further improve the sensitivity and detection capability of MRI. The use of CAs in MRI has greatly expanded the possible conditions that may be explored, offering physicians unprecedented noninvasive visualization of many ailments.^5,6 Most clinically available CAs are chemical agents, traditionally gadolinium (Gd³⁺) chelates. Generally, these Gd³⁺ chelates have very short circulation lifetimes in the body, work in a nonspecific manner, and have certain limitations as molecular CAs for long-time tracking. Another disadvantage of clinically used Gd³⁺ chelates is their relatively low proton relaxation efficiency, which means that high doses are necessary for imaging (e.g., mM concentrations of Gd³⁺ chelates at target areas).⁷ However, high-dose injection of Gd³⁺ chelates may be harmful to the body as it has been reported that the released free Gd³⁺ ions may cause severe nephrogenic system fibrosis syndrome in vivo.⁸ Therefore, the imaging performances of clinically available CAs to date have been far from optimal.

Nanoparticles naturally lend themselves to be used as CAs in MRI, where magnetically responsive atoms are essential.⁹ T₁ or spin–lattice contrast may be enhanced through the use of paramagnetic centers such as gadolinium^10,11 or manganese ions,^12,13 whereas iron oxide-based nanosystems are often used for T₂-weighted^14-16 or spin–spin imaging. Unlike other molecular imaging agents, these nanoparticles are not directly imaged in the resulting MRI image; in fact, the metal ions perturb the magnetic properties of protons around them, which are the actual target of a traditional MRI scan.⁶ Among various types of nanoscale imaging agents, inorganic nanoparticle (NP)-based CAs (hereafter, “NPs” refers to inorganic nanoparticles) have received great attention and have been regarded as promising imaging probes. This is due to their unique electronic, magnetic and optical properties at the nanoscale, facile intracellular uptake and accumulation, and excellent in vivo stability.¹⁷ The high surface areas of these NPs also greatly enhance their chemical reactivity, allowing for easy surface modification with biological and bioactive materials. These characteristics have caused NPs to emerge as promising candidates for next-generation imaging CAs.¹⁸

The contrast behavior of NPs strongly depends on their perturbation of magnetic properties of nearby protons, which can be tuned through a series of factors (e.g., size, surface, and shape) and various innovative approaches or strategies. Great efforts have been made for the rational design and manufacture of NPs with high relaxivity to elevate imaging sensitivity of CAs and even impart MRI with molecular imaging capabilities for better monitoring of biological processes (e.g., tumor microenvironment, cell tracking, and cellular interactions). This has resulted in the wide expansion of this research area, leading to the formation of sub-branches such as ultra-high field MRI,^19-21 multi-modal imaging,^22-24 and incorporation of theranostic techniques into MRI.²⁵

Both the rational design and engineering of nanoparticulate CAs play important roles in this field, whichmay, inthemeantime, broaden biomedical applications of MRI. However, the complexity of this process is much greater than initially expected, calling for broad expertise in a multidisciplinary manner through chemistry, physics, biology, and engineering. The goal of this review is not to comprehensively summarize nanoparticulate CAs for MRI that have been reported so far,²⁶ as there are many reviews available that list the major types of NPs as CAs (Gd³⁺-based T₁ CAs,^11,27 magnetic T₂ CAs,^14,15,28-30 and dual- T₁/T₂ CAs^9,31) or summarize a special application of CAs (e.g., bioresponsive CAs,^32-34 and multimodal CAs²⁴). Instead, this review intends to comprehensively discuss the main rules and mechanisms of nanoparticulate CAs, key factors influencing the relaxation of CAs, and innovative strategies to engineer CAs with high relaxivity. As the Chinese aphorism goes, “It is better to teach a man to fish than to give him a fish,” we hope that this timely and comprehensive review can familiarize readers with the general principles of designing high-performance CAs, stimulate extensive further studies, and attract extensive attention from researchers worldwide.

2. Basic rules for MRI contrast agents

Since the mechanisms of MRI have been thoroughly described elsewhere,^11,29 herein, we will focus on the basic rules of CAs for MRI, an in-depth understanding of which is extremely important for formulating design principles of CAs. The contrast enhancement of MRI originates from the interaction between CAs and surrounding water protons, which shorten the longitudinal (T₁) or transverse (T₂) relaxation time of nearby water molecules. Usually, T₁ CAs are applied to increase signal intensity, resulting in a positive contrast enhancement (brighter image) in T₁-weighted MRI, whereas T₂ CAs can decrease signal intensity, providing a negative contrast enhancement (darker image) in T₂-weighted MRI. Relaxivity (r_i, i = 1, 2) is therefore used to evaluate the efficacy of CAs, which is defined as the slope of a plot of relaxation rate (1/T_i, i = 1, 2) versus concentration of CAs:

$$1/T_i=1/T_{i0}+r_{i}C$$ (1)

where T_i0 is the relaxation time without CAs and C is the concentration of CAs (mM). The values of r₁, r₂, and r₂/r₁ are typically used to evaluate MRI efficiency of nanoparticulate CAs. The transition metal ions (Mn²⁺, Fe²⁺/Fe³⁺, Co²⁺, and Cu²⁺) and lanthanide metal ions (Eu³⁺, Gd³⁺, Ho³⁺, and Dy³⁺) are the major candidate relaxation centers that can be utilized to construct NPs with proper chemical compositions for enhanced contrast in MRI. The choice of a metal ion requires consideration of its electron motion, nanostructure-dependent effects, and involved relaxation mechanism to obtain high-relaxivity CAs.

2.1 Choice of ions

The selection of appropriate ions to construct nanoparticulate agents should obey certain rules. Inherently, one metal ion includes both the electron orbital motion and electron spin motion. The fast electron orbital motion is significantly higher than the speed of water proton relaxation, while the electron spin motion is slower and can be closely in tune with the slow water proton relaxation. Only the electron spinmagneticmoment of ions can efficiently influence the longitudinal water proton relaxation (r₁). Therefore, r₁ will be higher if the total electron angular momentum (J) contains a larger contribution from electron spin angular momentum (S) of adjacent ions. On the other hand, it will be small if J consists only of electron orbital angular momentum (L).³⁵ Table 1 summarizes the outer orbital, quantum parameters (S, L, and J) and effectivemagnetic moment (μ_eff) of candidate metal ions for MRI, from which we can determine suitable ions for chemical composition of CAs. For example, since Gd³⁺ has seven unpaired electrons and its J (J = 7/2) value is completely derived from S (S = 7/2, L = 0), inorganic Gd³⁺-based NPs (e.g., Gd₂O₃, NaGdF₄, GdF₃) have been widely explored as T₁ CAs.^36-49 Notably, Mn²⁺ plays a highly important role in biological processes as a cofactor of enzymes and through neurotransmitter release control; moreover, Mn²⁺ has five unpaired electrons (J = 5/2, S = 5/2, L = 0), showing prominent contrast effects that have been shown to enable visualization of brain anatomic structure and its neuronal activity.⁵⁰ Therefore, Mn²⁺-based NPs constitute a new imaging category and are promising candidates as T₁ CAs.^51-69

Table 1.

Outer orbital and quantum parameters of metal ions

Ion	Orbital	S	L	J	(μeff μB)
Mn2+	3d5	5/2	0	5/2	5.92
Mn4+	3d3	3/2	3	9/2	3.87
Fe2+	3d6	2	2	4	4.90
Fe3+	3d5	5/2	0	5/2	5.92
Co2+	3d7	3/2	3	9/2	3.87
Co3+	3d6	2	2	4	4.90
Cu2+	3d9	1/2	2	5/2	1.73
Eu2+	4f7	7/2	0	7/2	7.94
Eu3+	4f6	3	3	0	—
Gd3+	4f7	7/2	0	7/2	7.94
Dy3+	4f9	5/2	5	15/2	10.65
Ho3+	4f10	2	6	8	10.61

The r₂ of a CA is largely proportional to the square of μ_eff of the ion, as T₂ CAs show reduced signal dependence on induced magnetic moment of these CAs.⁷⁰ Among all lanthanide ions, Dy³⁺ (together with Ho³⁺) has the highest effective magnetic moment (μ_eff = 10.6 μ_B) and much attention has been paid to synthesizing Dy³⁺- or Ho³⁺-based NPs as T₂ CAs,^70-72 especially for ultra-high field imaging.^19-21,73 Studies of d-glucuronic acid-coated ultra-small Ln₂O₃ NPs (Ln = Eu, Gd, Dy, Ho, or Er) show that Gd₂O₃ NPs exhibit the highest r₁ value, whereas Dy₂O₃ NPs possess the highest r₂ value among all studied NPs.⁷⁴ Unlike the lanthanides, iron is one of the most abundant metal elements in living organisms and deeply involved in various biological processes including cellular respiration by redox enzymes and oxygen transport by hemoglobin.⁷⁵ Magnetic iron oxide NPs (e.g., Fe₃O₄, γ-Fe₂O₃, Fe@Fe₃O₄, MnFe₂O₄, etc.) are therefore very popular T₂ CAs that can produce negative imaging contrast.^29,30,76-84 The magnetization of paramagnetic NPs, such as Gd-based NPs and Mn-based NPs, directly depends on the total number of metal ions, and no magnetization is present in the absence of an external magnetic field. However, magnetic iron oxide NPs exhibit strong intrinsic magnetization, which may cause microscopic field inhomogeneity and accelerate the dephasing of protons for MRI contrast enhancement when an external magnetic field is applied.

In addition to relaxation impact from ion species, the ion valency also affects relaxation interaction between ions and water protons. Variable valencies (e.g., Mn⁴⁺/Mn²⁺ ions,⁵⁴ Co³⁺/Co²⁺ ions,⁸⁵ Eu³⁺/Eu²⁺ ions⁸⁶) can endow these ions with tunable MRI performance under certain reducing or oxidative microenvironments, thus creating responsive MRI CAs for molecular imaging.³² Mn⁴⁺-based NPs are widely used as pH or redoxresponsive MRI CAs because Mn⁴⁺ (J = 9/2, S = 3/2, L = 3) with negligible relaxation interactions can be easily reduced to Mn²⁺ (J = 5/2, S = 5/2, L = 0), which shows high MRI contrast capability according to the rules mentioned above. Overall, understanding the basic principles of NPs such as ion species, ion valency, and magnetic properties will facilitate the engineering of nanoparticulate CAs with excellent MRI performance for various applications.

2.2 Mechanisms of contrast enhancement

The relaxation enhancement of water protons near a paramagnetic center generally follows the inner sphere, secondary sphere, and outer sphere mechanism (Fig. 1).⁸⁷ The water molecules directly coordinated to the paramagnetic center are responsible for inner sphere relaxation contribution, while bulk water molecules in the nearby environment affect outer sphere relaxation. In some cases, water molecules that are hydrogen bonded to chelators are regarded as the secondary sphere. The contrast enhancement of individual CAs closely depends on their chemical composition, structure, and surfacemodification.

Fig. 1.

Inner sphere, secondary sphere, and outer sphere can influence relaxation rates of MRI CAs.

For T₁ CAs, on the basis of the classical Solomon–Bloembergen–Morgan (SBM) theory,^35,87,88 common strategies for enhancing relaxivity mainly include increasing the number of bound water molecules (q), enhancing rotational correlation time (τ_R), and optimizing water residence time (τ_M). Higher τ_R values imply slower molecular tumbling and lead to increased proton relaxivity. τ_M is the inverse of water exchange rate in the inner sphere (k_ex = 1/τ_M), which is favorable for proton relaxivity in the case of fast water exchange (high k_ex). The relationship between τ_R and τ_M can be described by the following equation:⁸⁹

$$1/{\tau }_{ci}=1/T_{i\mathrm{e}}+1/{\tau }_{\mathrm{R}}+1/{\tau }_{\mathrm{M}}i=1,2$$ (2)

where τ_c is the correlation time parameter and T_ie (i = 1, 2) is the electronic relaxation time. Since it is challenging to predict and experimentally alter T_ie, approaches to obtain highly efficient CAs commonly focus on the regulation of τ_R and τ_M.³⁷ By immobilizing Gd³⁺ on the surface or lattice of NPs, prolonging τ_R will result in significantly enhanced r₁, which is a commonly used strategy because these nanoparticulate CAs can couple the tumbling rate of Gd³⁺ chelates to that of much larger species (e.g., titanium, gold, and silica NPs).^90-98 Another widely used approach to enhance r₁ is the optimization of τ_M,^99,100 which is readily accessible because nanoparticulate CAs inherently have relatively slow tumbling rates.

Different from T₁ CAs that influence the relaxation of water proton through direct energy exchange between water and paramagnetic ions,¹⁰¹ magnetized NPs can induce increased T₂ relaxivity, which reduces phase coherence of water molecules in an inhomogeneous magnetic field generated by magnetized NPs.^102-104 Based on the quantum mechanical outer sphere theory, T₂ relaxivity is given by eqn (3)¹⁰⁵

$$1/T_2=(256{\pi }^2{\gamma }^2/405)V^{\ast }{M}_{\mathrm{s}}^2{r}^2/D(1+L/r)$$ (3)

where r and M_s are the effective radius of magnetic NPs and saturation magnetization, respectively, γ is the proton gyromagnetic ratio, V* is the volume fraction, D is the diffusivity of water molecules, and L is the thickness of an impermeable surface shell on the magnetic core. It should be noted that M_s value of a magnetic NP can be tuned by its high surface area, where the interior spins of magnetic NPs are usually aligned, while the spins of surface atoms are magnetically disordered. This is known as the spin canting effect, which is non-negligible because NPs have a large surface-to-volume ratio.²⁹ According to eqn (3), T₂ relaxivity of magnetic NPs is highly dependent on effective radius of the magnetic core, M_s, and thickness of an impermeable surface.

3. Factors influencing relaxivity of MRI contrast agents

Recently, great advances in nanotechnology have stimulated the quick development of nanoparticulate CAs, which have two main traits of interest: the dominant quantum confinement effect and the large surface area. Size-dependent magnetic, electronic, and optical properties originating from dominant quantum confinement effects can be utilized in the rational design of high-performance MRI CAs at the nanoscale. Moreover, the large surface area of NPs offers increased chemical reactivity for surface modification with additional functional species, enabling enhanced magnetic relaxivity and higher-quality MRI with reasonable detection limits. Typically, three techniques are applied for obtaining high-relaxivity CAs at the nanoscale: size control, surface modification, and shape regulation (Fig. 2).

Fig. 2.

Factors influencing relaxation of MRI CAs: size, surface, and shape effects at the nanoscale.

3.1 Size effects

Inorganic NPs present a unique size-dependent effect on physical properties including surface plasmon resonance of metal nanoparticles,^106,107 fluorescence emission of semiconductor NPs,¹⁰⁸ and superparamagnetism or heating of magnetic NPs.^109,110 By controlling nucleation and growth, the size of NPs can be well controlled. The dependence of cellular uptake, biodistribution, and clearance rate on NP size has been widely investigated by many groups.^111-115 To date, several types of CAs have been developed for various applications in MRI, which can be classified according to their chemical composition (Gd-, Mn-, Fe-, Dy- or Ho-based CAs; Table 2). A deep understanding of the involved mechanisms between size effect and proton relaxation will undoubtedly benefit the rational design of T₁ or T₂ CAs for high-performance MRI. For T₁ CAs, smaller-sized NPs usually generate greater contrast enhancement because their surface-to-volume ratio is the dominant contributing factor, which increases as the size of NPs is decreases. However, τ_R will become much lower for large NPs and plays a dominant role in affecting relaxivity, causing larger NPs to possess higher r₁ values. For T₂ CAs, larger-sized magnetic NPs usually have higher magnetization than smaller-sized NPs, and therefore, r₂ values increase as size increases. However, this trend breaks down when sizes of magnetic NPs become too large.

Table 2.

Representative inorganic nanoparticle size effect on relaxivity

NPs	Shape	Approximate size (nm)	Surface coating	Relaxivity (mM−1 s−1)	Field (T)	Ref.
NaGdF4	Sphere	2.5/4.0/6.5/8.0	PVP	r1 = 7.2/4.5/3.3/3.0	1.5	49
NaGdF4	Sphere	5/15/20	PEG	r1 = 6.2/5.7/8.78	3.0	41
NaGdF4	Sphere	2.1/3.2/8.6	PAA	r1 = 47.3/27.5/15.3	0.5	36
NaGdF4	Sphere	3/4/5	PEG	r1 = 58.8/39.4/28.6	1.4	39
Gd3O2	Sphere	2.2/3.8/4.6	SiPEG	r1 = 8.8/8.8/4.4	7.0	45
Gd3O2	Sphere	4.4/5.4/10.1	Albumin	r1 = 18.49/16.22/12.26	1.5	129
MnO	Irregular	7/15/20/25	PEG	r1 = 0.37/0.18/0.13/0.12	3.0	69
Mn2O3	Sphere	3/5/11/13	PEG	r1 = 2.38/1.39/0.99/0.41	1.5	132
Mn2O3	Tetrahedron	6/8/20	PEG	r1 = 1.08/0.63/0.25	1.5	132
Fe2O3	Sphere	2.2/3/12	PEG	r1 = 4.78/4.77/2.37	1.5	135
MnFe2O4	Sphere	2/3/3.9	PEG	r1 = 8.43/8.23/6.98	3.0	136
Fe3O4	Sphere	4/6/9/12	DMSA	r2 = 78/106/130/218	1.5	79
Fe3O4	Octapod	20/30	HDA-G2	r2 = 679/209	7.0	137
Fe3O4	Sphere	8/23/37/65	PVP	r2 = 173/204/240/249	7.0	138
Fe3O4	Rod	30a/40/50/60/70	PEI	r2 = 312/381/427/545/608	3.0	80
MnFe2O4	Sphere	6/9/12	DMSA	r2 = 208/265/358	1.5	84
NaDyF4	Sphere	5.4/9.8/20.3	PEG	r2 = 32/51/101	9.4	19
NaDyF4	Rod	19a/20/25/35	PEG	r2 = 75/65.2/91.4/204.4	9.4	20
NaHoF4	Sphere	10.6/14.1/16.4	PEG	r2 = 31.9/51.5/70.4	9.4	20
NaHoF4	Sphere	3.2/7.4/13.2	PEG	r2 = 83.4/45.4/131.7	7.0	21

^aLength of nanorods.

3.1.1 Gd-based contrast agents

The recently developed Gd³⁺-based NPs such as Gd₂O₃-,^43-45,116 NaGdF₄-,^36,38-41 GdF₃-,^46,117,118 and Gd³⁺-doped NPs (e.g., quantum dots,^119-123 upconversion NPs,^124-126 and carbon dots^127,128) have been regarded as promising T₁ CAs because of their controllable size and morphology. By optimizing the reaction parameters during both nucleation and growth, Johnson et al. synthesized size-tunable ultra-small NaGdF₄ NPs with uniform sizes of around 2.5 nm, 4 nm, 6.5 nm, and 8 nm (Fig. 3a) and found that the r₁ value increased dramatically with decreasing size from 3.0 mM⁻¹ s⁻¹ for 8.0 nm NPs to 7.2 mM⁻¹ s⁻¹ for the smallest 2.5 nm ones (Fig. 3b).⁴⁹ Correlations of NPs’ sizes and their r₁ values showed that contributions from the surface-tovolume ratio increased with decreasing NP size, becoming the dominant contributing factor. Several other research groups also reported similar size effects of NaGdF₄ NPs on relaxivity.^36,39 However, τ_R became much lower for larger NaGdF₄ NPs (>15 nm) and played a dominant role in affecting r₁ value, as demonstrated by Gao and co-workers, who found r₁ values of 5.7 mM⁻¹ s⁻¹ for 15 nm NaGdF₄ NPs, which increased to 8.78 mM⁻¹ s⁻¹ for 20 nm ones (Fig. 3c).⁴¹ Inspired by this, we recently synthesized ultra-small sized and renal-clearable NaGdF⁴ nanodots (~2 nm) with high r₁ values of 8.93mM⁻¹ s⁻¹.⁴⁰ These nanodots exhibited excellent magnetic resonance angiography (MRA) performance, as the vascular signal was much stronger than that achieved using clinical MRI CAs (Fig. 3d). With the new CAs, even capillary vessels and atherosclerotic plaques could be clearly delineated (Fig. 3e).

Fig. 3.

(a) TEM images of oleate-stabilized NaGdF₄ NPs of varied particle sizes (scale bar is the same for all images). (b) Plots of 1/T₁ versus Gd³⁺ concentration for varied NaGdF₄ NPs in water (1.5 T). Reproduced with permission from ref. 49. Copyright 2011, American Chemical Society. (c) R₁ relaxivity of aqueous solutions containing NaGdF₄ NPs of different sizes or Gd–DTPA with different concentrations of Gd³⁺. Reproduced with permission from ref. 41. Copyright 2013, American Chemical Society. (d) Magnetic resonance angiography of rabbits within 3 min after injection of ultra-small NaGdF₄ nanodots or Magnevist. AA: abdominal aorta; IVC: inferior vena cava. (e) Transverse cross-sectional images of rabbit atherosclerotic plaques before and after injection of ultra-small NaGdF₄ nanodots or Magnevist at the same dosage (13 mg Gd per kg). Reproduced with permission from ref. 40. Copyright 2014, Wiley-VCH.

The same trends were also observed for ultra-small Gd₂O₃ NPs,^45,129 such that r₁ values increased from 4.4 mM⁻¹ s⁻¹ (4.6 nm Gd₂O₃) to 8.8 mM⁻¹ s⁻¹ (2.2 nm Gd₂O₃).⁴⁵ Based on the fitting of Gaussian, Lorentzian, and log-normal functions, Lee et al. concluded that the optimal range of diameters for high-relaxivity Gd₂O₃ NPs should be 1–2.5 nm with a theoretical maximum r₁ value of 10.1–10.6 mM⁻¹ s⁻¹.⁴⁸ In that study, they reported a large r₁ value of 9.9 mM⁻¹ s⁻¹ of ultra-small Gd₂O₃ NPs coated with d-glucuronic acid. Consistent with ultra-small NaGdF₄ NPs, such a large r₁ value of ultra-small Gd₂O₃ NPs was attributed to the large surface-to-volume ratio, as well as the cooperative induction of surface Gd³⁺ ions for longitudinal relaxation. It is generally accepted that the r₁ value should be as large as possible, and for high-performance T₁ MRI CAs, the r₂/r₁ ratio should be close to unity, while a greater value of r₂/r₁ will generally result in higher contrast enhancement efficiency for T₂ MRI CAs.⁹ The r₂/r₁ ratio of ultra-small Gd₂O₃ NPs was estimated to be 1.06 (r₁ = 9.9 mM⁻¹ s⁻¹, r₂ = 10.5 mM⁻¹ s⁻¹), which seemed to satisfy this condition.

Biologically, these ultra-small sized agents can escape from recognition and can be capture by the liver and then be rapidly cleared through the kidneys, in which the basal lamina has pores of approximately 10 nm to filter metabolites (e.g., ultrasmall sized NPs) from the bloodstream.^38,48,130 Such rapid renal clearance diminishes in vivo toxicity of NPs owing to the suppressed tissue accumulation and certainly facilitates their future clinical translation and application. In addition, ultrasmall Gd³⁺-based NPs can be integrated with additional components such as multifunctional nanoplatforms for non-targeted blood pool imaging,⁴⁰ targeted molecular imaging,⁴¹ pH/H₂O₂- responsive MRI,¹³¹ and brain tumor diagnosis and image-guided therapy.¹³⁰

3.1.2 Mn-based contrast agents

As another major class of T₁ CAs, Mn-based NPs (e.g., MnO NPs, Mn₃O₄ NPs, and MnO₂ nanosheets) have gained extensive interest as GSH- or pHresponsive MRI CAs due to their readily changeable valency and easy release of manganese under acidic conditions.^54,55,63,133 The size-dependent magnetic properties of MnO and Mn₃O₄ NPs with monodispersed sizes in the range of 6–22 nm were initially reported by Seo et al.¹³⁴ Subsequently, the size-dependent T₁ relaxation properties of MnO NPs were investigated by Hyeon and co-workers,⁶⁹ demonstrating higher T₁ signal enhancement for smaller MnO NPs than larger ones (Fig. 4a and b). The interaction between proton spins and surface Mn²⁺ ions on NPs is regarded as the main dipolar interaction responsible for their relaxation properties. Clear T₁-weighted MR images of the brain, liver, kidney, and spinal cord with a fine anatomical structure were obtained after administration of MnO NPs. Since MnO NPs spontaneously get oxidized to Mn₃O₄ NPs in air,⁶⁸ An et al. reported measurements of T₁ and T₂ relaxation of spherical and tetrahedral Mn₃O₄ NPs with varied uniform sizes.¹³² In agreement with the previous results, the r₁ value increased as the size of Mn₃O₄ NPs decreased for both spherical (Fig. 4c) and tetrahedral NPs (Fig. 4d), and 3 nm-sized spherical NPs exhibited the highest r₁ value of 2.38 mM⁻¹ s⁻¹, again indicating that higher surface-to-volume ratios lead to higher relaxivity of ultrasmall NPs.

Fig. 4.

(a) TEM images of MnO NPs of varied sizes (7, 15, 20, and 25 nm) in water. (b) T₁-weighted MR image of MnO NPs in a 3.0 T clinical MRI system. Reproduced with permission from ref. 69. Copyright 2007, Wiley-VCH. Size-dependent T₁-weighted MR images and relaxivities of water-dispersible spherical (c) and tetrahedral (d) Mn₃O₄ NPs at varied Mn²⁺ concentrations. (e) The r₁ and r₂ values and r₂/r₁ ratios of Mn₃O₄ MPs from c and d. Reproduced with permission from ref. 132. Copyright 2012, Wiley-VCH.

To further increase surface-to-volume ratio, Mn-based NPs were etched into hollow structures by selectively removing the core MnO phase from the Mn₃O₄ shell phase, thus exposing more active Mn²⁺ ions to water protons.⁶⁸ As a result, the hollow nanostructure with a larger water-accessible surface area possessed a much higher r₁ value of 1.42 mM⁻¹ s⁻¹ compared to the solid ones (r₁ = 0.21 mM⁻¹ s⁻¹) at 3.0 T. Such hollow structures could be used as a drug delivery system with marked MRI contrast enhancement for the monitoring of drug release.^139,140

3.1.3 Fe-based contrast agents

Iron oxide NPs are generally regarded to be safer than Gd- or Mn-based CAs because iron is essential in the human body, mostly stored as ferritin in the blood. Fe³⁺ has 5 unpaired electrons (J = 5/2, S = 5/2, L = 0), the same as Mn²⁺, making Fe³⁺ a promising candidate for T₁ CAs.^136,141-143 Small-sized iron oxide NPs show enhanced T₁ contrast effects due to the existence of a large number of Fe³⁺ ions on the surface, suppressing T₂ relaxation by their small magnetic moment. The magnetic properties and r₁ values of various small-sized iron oxide NPs from 1.5 to 12 nm have been investigated (Fig. 5a).¹³⁵ By canting the most surface spins (around 93.6%), 2.2 nm-sized iron oxide NPs with low magnetic moments exhibited a relatively high r₁ value of 4.78 mM⁻¹ s⁻¹, while the r₁ values of 12 nm NPs (nearly 38.6% of spins were canted, Fig. 5b) decreased to 2.37 mM⁻¹ s⁻¹ (Fig. 5c). This phenomenon was further verified in cellular studies, in which significant T₁ signal enhancement for 3 nm-sized NPs and attenuated T₁ signal for cells incubated with 12 nm-sized iron oxide NPs were found (Fig. 5d). Another study from Zhang et al. also indicated similar size-dependences of r₁ values based on ultra-small MnFe₂O₄ NPs.¹³⁶ Recently, exceedingly small iron oxide NPs with a hydrodynamic size of 4.7 nm, consisting of approximately 3 nm inorganic cores and about 1 nm organic hydrophilic shell, were reported to have high T₁ relaxivity (r₁ = 5.2 mM⁻¹ s⁻¹ at 1.5 T).¹⁴² Excreted mostly within the first 2.5 h through the urine, the exceedingly small iron oxide NPs have been successfully applied for high-performance magnetic resonance angiography (MRA) imaging (Fig. 5e and f). Strong positive contrast from both the heart and blood vessels has been found, which decreased over time, while the signal from the bladder increased, indicating renal clearance.

Fig. 5.

(a) Field-dependent magnetization curves at 300 K for iron oxide NPs of varied sizes. (b) Description of spin canting effect (canting layer = 0.9 nm) in iron oxide NPs of varied sizes. Red and black colors represent magnetic cores and magnetically disordered shells. (c) Plots of 1/T₁ over concentration of iron oxide NPs of 3 nm and 12 nm diameter. (d) T₁-weighted MR images of MCF-7 cell pellets after 24 h incubation with iron oxide NPs of 3 nm and 12 nm in diameters. Reproduced with permission from ref. 135. Copyright 2011, American Chemical Society. (e) T₁-weighted magnetic resonance angiography (MRA) of a mouse injected with ultra-small iron oxide NPs at 7 T. (f) Five different perspectives of MRA images, which were extracted from a 3D scan, at 4 minutes post-injection. Reproduced with permission from ref. 142. Copyright 2017, National Academy of Sciences.

Importantly, iron oxide NPs are also widely used as T₂ CAs, where the magnetic moment of these NPs decreases rapidly as the size decreases due to reduced magnetic anisotropy and spin disordering on the surface of NPs.^30,144 The size-dependence of T₂ contrast has been extensively investigated both theoretically and experimentally. Theoretically, two different regimes are predicted. The first is called the “motional averaging regime” (MAR) that describes T₂ relaxation for relatively small iron oxide NPs,¹⁰⁴ where water diffusion near NPs occurs on a much faster timescale than the resonance frequency shift, resulting in increased r₂ values with increasing NP size. However, this theory breaks down for larger NPs because the induced surrounding perturbing field around larger NPs is much stronger, and proton diffusion becomes a nondominant factor for signal decay. Therefore, the relaxation rate is independent of diffusion and does not significantly increase with further increase in size of NPs. This process is termed the “static dephasing regime” (SDR), first introduced by Yablonskiy and Haacke.¹⁴⁵ Importantly, this tendency has been validated on NPs of varied sizes synthesized by many different groups.

Generally, small iron oxide NPs synthesized by various research groups are in the MAR, where larger NPs possess higher magnetization and therefore larger r₂ values. For example, such size effects on T₂ contrast have been reported by Cheon and co-workers,⁷⁹ using small and size-controlled Fe₃O₄ NPs (Fig. 6a), where the variation of NPs size from 4 to 6, 9, and 12 nm resulted in r₂ values of 78, 106, 130, and 218 mM⁻¹ s⁻¹, respectively (Fig. 6b–d). Similar size effect on relaxation rates has also been reported by many other research groups.^{80,83,105,146-149} Importantly, this size effect is a universal phenomenon, which is not limited to spherical NPs. For example, octapod iron oxide NPs with average edge lengths of 30 nm exhibited a large r₂ value of 679.25 mM⁻¹ s⁻¹, almost three times higher than that of the NPs of 20 nm in edge lengths.¹³⁷ However, according to the SDR theory, the r₂ value reaches a plateau as the size continues to increase. It has been reported by Chen and co-workers that the size effect became less significant for single iron oxide NPs of 50 nm (Fig. 6e).¹³⁸ They found that r₂ values increased rapidly from 173.37 to 203.86 and 239.98 mM⁻¹ s⁻¹ at 7 T, when the particle size increased from 8 to 23 and 37 nm, respectively (Fig. 6f), but r₂ value was only slightly higher at 248.89 mM⁻¹ s⁻¹ for 65 nm-sized iron oxide NPs (Fig. 6g). This SDR theory has also been experimentally verified in cells loaded with iron oxide NPs,¹⁵⁰ as well as in an induced clustering process.¹⁵¹

Fig. 6.

(a) TEM images of iron oxide NPs of 4, 6, 9, and 12 nm. (b) Upper panel: Size-dependent T₂-weighted MR images of iron oxide NPs in aqueous solution at 1.5 T; lower panel: size-dependent changes from red to blue in color-coded MR images based on T₂ values. (c) The magnetization of iron oxide NPs measured by a SQUID magnetometer. (d) Size-dependent r₂ values of iron oxide NPs. Reproduced with permission from ref. 79. Copyright 2005, American Chemical Society. (e) Size-dependent T₂-weighted MR images of PVP-coated iron oxide NPs in aqueous solution with various concentrations at 7 T. (f) Plots of 1/T₂ against Fe concentration of PVP-coated iron oxide NPs. (g) Size-dependent r₂ values of iron oxide NPs. Reproduced with permission from ref. 138. Copyright 2010, American Chemical Society.

3.1.4 Dy- or Ho-based contrast agents

Ultra-high field MRI, the future of next-generation MRI scanners, brings not only new opportunities for biomedical research, but also great challenges in developing exogenous CAs, since most currently used CAs are only suitable at relatively lowmagnetic field strengths (e.g., 1.5 T or 3.0 T) but are inefficient at higher magnetic field strengths (e.g., 7.0 T). Dysprosium (Dy³⁺) and holmium (Ho³⁺) are promising candidates for imaging in an ultra-high magnetic field due to their relatively short electronic relaxation time (~10⁻¹³ s) and large magnetic moment (~10.6 μ_B).⁷¹ The size-dependent properties of NaDyF₄ NPs under ultra-high magnetic fields were first reported by van Veggle and co-workers,¹⁹ demonstrating that T₂ contrast was enhanced with increasing NP size and magnetic field strength (Fig. 7b and c), while r₁ value remained almost unchanged (Fig. 7a).

Fig. 7.

(a) r₁ and (b) r₂ values of NaDyF₄ NPs (three different sizes) at 3 and 9.4 T. (c) (i) TEM images of 5.4, 9.8, and 20.3 nm NaDyF₄ NPs, (ii) phantom MR images of NPs at 1.0 mM Dy³⁺ concentration at 9.4 T. Reproduced with permission from ref. 19. Copyright 2012, American Chemical Society. (d) Schematic of the effect of size and a coating layer on the relaxation of NaDyF₄ and NaHoF₄ NPs. Reproduced with permission from ref. 20. Copyright 2016, American Chemical Society. (e) T₂-weighted MR images of NaHoF₄ NPs of varied sizes in aqueous solutions at 7.0 T. (f) r₂ values obtained for varied sized NaHoF₄ NPs at 1.5 T, 3.0 T, and 7.0 T. (g) Simulated curves of B²φ from Curie contribution versus diffusion correlation time (τ_D), which is related to NP size. Reproduced with permission from ref. 21. Copyright 2016, Elsevier Ltd.

In a follow-up study, van Veggle and co-workers investigated in detail the involved relaxation mechanisms of NaDyF₄ and NaHoF₄ NPs in the range of 9–40 nm based on the outer sphere theory (Fig. 7d).²⁰ The larger particle had a larger volume fraction, and a three-fold increase in core size resulted in r₂ values that were 30 times larger. Further theoretical simulation results showed that Curie contribution became predominant under an increasing magnetic field. Recently, we have reported the size effect of NaHoF₄ NPs on magnetic relaxation (Fig. 7e).²¹ Both experimental and theoretical findings revealed that Curie contribution to contrast enhancement was the principal mechanism when particle size was smaller than 7 nm (Fig. 7f and g), as the φ(τ_D) value from Curie relaxation remarkably decreased with increasing diffusion correlation time (τ_D) from 1 ns (3 nm NaHoF₄) to 5 ns (7 nm NaHoF₄). The φ(τ_D) value then slightly decreased as τ_D became larger, explaining the opposite change of r₂ values that was observed when NP size was less than 7.4 nm. As the size of NPs increased to 13 nm and 29 nm, the contribution of dipolar mechanism could not be neglected, and all relevant mechanisms mentioned above contributed to MRI contrast enhancement.

3.2 Surface effects

Apart from the size effect on MRI relaxation, surface properties of NPs also play an important role in MRI contrast enhancement based on interaction between NPs and water molecules, which happens primarily on the surface of NPs. Individual contributions from the inner sphere, secondary sphere, and outer sphere depends on the CA’s structure and composition, which can be regulated by surface modifications of tuned composition and thicknesses. Most NPs developed for in vivo MRI applications could be stabilized and functionalized with coating materials such as organic surface ligands, dense or mesoporous silica shells, or an inorganic nanolayer (e.g., Fe_xO_y or NaGdF₄ layers). These modifications, here termed surface effects, can in turn impact the relaxation of water protons in various configurations, such as hydration, hydrogen binding, and diffusion. For T₁ CAs, surface effects can enhance proximity of water to center paramagnetic ions by tuning the distance of water protons to the NP surface, generating greater contrast than small-molecule CAs alone. Such a process depends on the species, chain length, density, and functional group of organic ligands, as well as the thickness of silica shells or inorganic nanolayers. For T₂ CAs, proper surface modification of magnetic NPs will impact intrinsic surface spin disordering, increase inhomogeneity in the local magnetic field, or induce much stronger magnetization, leading to improvement in T₂ contrast effects.

3.2.1 Organic surface ligands

Organic surface ligands are critical for the synthesis of NPs, not only for controllable growth during synthesis but also for ensuring their biocompatibility for in vivo application. An early study by Lee et al. suggested that different types of organic surface moieties greatly impacted T₁ contrast enhancement.⁶⁰ Organic ligands including mercaptosuccinic acid (MSA), poly(maleic anhydride-alt-1-octadecene) (PMAO), PMAO–PEG, and Pluronic PF127 were capped on MnO NPs. The MnO@MSA exhibited the largest r₁ values of 2.55 mM⁻¹ s⁻¹ at pH 7.4 and 6.94mM⁻¹ s⁻¹ at pH 5.0 among all the coatings, indicating that r₁ relaxivity of the NPs depended on the surface ligands.

Furthermore, the addition of polyethylene glycol (PEG) to the surface (i.e., PEGylation) of NPs has been widely applied to improve biocompatibility of NPs, prolong in vivo circulation, and reduce the capture by the reticuloendothelial system (RES).^152-155 Recently, Johnson et al. reported a compact micellization strategy that could achieve ultra-high r₁ values (around 80 mM⁻¹ s⁻¹ at 1.4 T), about 25 times higher than that of clinically used Gd³⁺ chelates, by regulating the surface ligands (Fig. 8a–c).³⁹ The compactly packed PEGylated lipids would bring the water protons closer to the surface of NaGdF₄ NPs and thus lead to enhanced relaxivity. They also found that the relaxivity was independent of outer PEG chain length (Fig. 8d and e), whereas increasing the inner core NaGdF⁴ size resulted in lowered relaxivity values (Fig. 8f and g). Two important criteria to achieve ultra-high relaxivity with Gd³⁺-based NPs were proposed: (i) extremely small core NPs (<5 nm) and (ii) surface coating strategies that can render compact and individually stabilized NPs in water.

Fig. 8.

(a) A TEM image of ultra-small oleate-stabilized NaGdF₄ NPs of 3 nm diameter. (b) A representative negative-stained TEM image of DSPE– PEG-coated NP micelles. Inset: Schematic of a NaGdF₄ NP confined within a DSPE–PEG micelle. (c) r₁ values of a clinical MRI contrast agent (Dotarem) and the compact NP micelles at low (1.41 T) and high (7 T) fields. (d) Schematic and (e) r₁ values of NaGdF₄ NPs (3 nm) confined within DSPE–PEG micelles with varied PEG chain length. (f) Schematic and (g) r₁ values of NaGdF₄ NPs confined within DSPE–PEG micelles with variable core NP size (3–5 nm). Reproduced with permission from ref. 39. Copyright 2016, American Chemical Society. (h) r₁ and r₂ values of NaGdF₄ NPs coated with different ligands measured at 0.5 T. (i) Schematic of the strong hydrogen-bonding capacity of PAA to water molecules to improve r₁. Reproduced with permission from ref. 36. Copyright 2017, American Chemical Society.

After surface PEGylation of ultra-small NaGdF₄ nanodots, our group further conjugated DTPA on the surface of NPs to capture potentially released Gd³⁺ ions, effectively preventing their liberation.⁴⁰ The Gd³⁺ ions captured by DTPA could directly interact with water protons to further enhance relaxivity. Recently, Yan and co-workers presented another simple strategy to improve r₁ values by taking advantage of surface ligands’ strong interaction with water protons.³⁶ NaGdF₄ NPs coated with poly(acrylic acid) (PAA) exhibited largely enhanced r₁ values compared to those coated by polyethylenimine (PEI) or PEG (Fig. 8h), which resulted from the strong hydrogen-bonding capacity of PAA to water molecules according to their theoretical calculation (Fig. 8i).

By rationally designing the functional group of surface ligands, the transverse r₂ value can also be regulated for iron oxide NPs. Zeng et al. reported the surface modification of PEG2000 ligands with three different functional groups: diphosphate (DP-PEG), hydroxamate (HX-PEG), and catechol (CC-PEG) on 3.6 nm and 10.9 nm Fe₃O₄ NPs (Fig. 9a).¹⁴⁹ The r₂ values increased with increasing sizes for all three ligand-modified Fe₃O₄ NPs, and the HX- or CC-PEG modified NPs showed much higher r₂ values than the unconjugated functional group of DP-PEG for both sized Fe₃O₄ NPs (Fig. 9b), while the r₁ value only changed slightly with different anchored groups and NP sizes (Fig. 9c). The π–π conjugation and p–π conjugation from the CC and HX groups were believed to create increased inhomogeneity in the local magnetic field around the underlying Fe₃O₄ NPs when compared to the unconjugated DP group (which has no π–π or p–π conjugation), which effectively accelerated transverse relaxation of the water protons nearby.

Fig. 9.

(a) Left panel: Chemical structures of PEGs used for exchanging the hydrophobic ligands of 3.6 nm(S: small) and 10.9 nm(L: large) iron oxide NPs; right panel: TEM images of resulting PEGylated NPs (scale bars: 50 nm); insets: photographs of aqueous solutions of PEGylated NPs at an equal Fe concentration of 20 mM. Comparison of (b) r₂ and (c) r₁ values of different kinds of PEGylated NPs. Reproduced with permission from ref. 149. Copyright 2014, Wiley-VCH. (d) Left panel: Schematic of three transfer approaches: ligand exchange of oleic acid with a water-soluble polymer (top), coating of individual NPs with the amphiphilic polymer (middle), and embedding into lipid micelles (bottom). Reproduced with permission from ref. 147. Copyright 2007, American Chemical Society. (e) Schematic of a nanoparticle with 4.8 nm iron oxide core and DSPE-PEG1000 coating. (f) T₂ relaxivity of iron oxide NPs at a constant iron concentration. (g) T₂ relaxivity of NPs on a per-particle basis. Iron oxide NPs with two core sizes, 5 and 14 nm, and five PEG sizes, 550, 750, 1000, 2000, and 5000 Da, were evaluated. Reproduced with permission from ref. 105. Copyright 2010, American Chemical Society.

Besides the surface groups, the ligand species and the methods of modification also affect T₂ contrast enhancement. The species effects of surface ligands on magnetic properties were verified by Duan et al. by coating ultra-small Fe₃O₄ NPs with different levels of hydrophilic ligands including hyperbranched PEI, PEG grafted PEI (PEG-g-PEI), and poly(maleic acid)-octadecene (PMO).¹⁴⁸ The hydrophilicity of surface ligands was found to greatly impact the intrinsic surface spin disordering of NPs for water proton relaxation, where the most hydrophilic PEI coated Fe₃O₄ NPs showed the highest r₂ value, nearly three times higher than that of the least hydrophilic PMO-coated NPs. Among different methods for surface modification of MnFe₂O₄ NPs such as ligand exchange, encapsulation with polymer shells, and embedding them into micelles (Fig. 9d), embedding NPs into lipid micelles was found to be the most efficient method for greatly enhancing r₂ relaxation.¹⁴⁷ The larger-sized core inside lipid micelles and the contribution from dephasing of proton spins of the lipid molecules due to high viscosity of these molecules were considered to be responsible for this phenomenon.

The accessibility of water molecules to the magnetic core of NPs depends on the thickness of surface ligands, which can influence T₂ relaxation of NPs. Tromsdorf et al. found that increasing PEG chain length from PEG350 to PEG550 on the surface of Fe₃O₄ NPs led to decreased r₂ and almost unchanged r₁ values due to aggregation of NPs.¹⁴¹ However, slight changes were found for both r₁ and r₂ values when PEG chain length was increased from 550 to 2000. Another report by Tong et al. indicated that Fe₃O₄ NPs had a critical PEG ligand size, at which the r₂ value changed dramatically.¹⁰⁵ The coating thickness could be well controlled by varying the size of DSPE–PEG from 500 Da to 5000 Da (Fig. 9e). Interestingly, the r₂ value of 14 nm Fe₃O₄ NPs increased by 2.5-fold when PEG molecular weight decreased from5000 to 1000 Da (Fig. 9f). It was suggested that 14 nm iron oxide NPs and PEG1000 surface modification was the optimal combination of core size and coating thickness according to T₂ relaxivity on a per-particle basis (Fig. 9g). However, r₂ values did not increase further when PEG size decreased to 750 and 550 Da. Such effects of PEG chain length on T₂ relaxivity can be partly attributed to changes in the shell/core ratio, i.e., L/r in eqn (3), under the assumption that the coating was water impermeable.

3.2.2 Surface silica coatings

The growth of a dense or mesoporous silica shell on the surface of MRI CAs has been a commonly used strategy since silica exhibits excellent biocompatibility, and the active surface or pore volume of the shells can be used for conjugating or loading biological molecules, functional particles, and therapeutic agents.^156-158 Our group extensively investigated the effect of both dense (dSiO₂) and mesoporous (mSiO₂) silica shells and their thickness on the MRI relaxation of Gd³⁺-based upconversion nanoparticles (UCNPs) (Fig. 10a).¹⁵⁹ The thicknesses of dSiO₂ and mSiO₂ shell could be well controlled and uniformly coated on each particle (Fig. 10b and c). It was found that both r₁ and r₂ values were independent of thickness of mSiO₂ shell, the thickness of which could be readily controlled from 11.1 nm to 23.4 nm. The dominant outer sphere mechanism determined by the translational diffusion of water molecules was thought to be responsible for this phenomenon, and the diffusion of water molecules in the mesoporous channels was not affected by the increased thickness of the mSiO₂ shell (Fig. 10d). However, the diffusion of water molecules in the dSiO₂ shell, in which large pore channels did not exist, became much more restricted by thicker dSiO₂ shells, leading to longer translational diffusion time and thus larger r₁ and r₂ values. Therefore, the r₁ value of UCNPs@dSiO₂ increased from4.03mM⁻¹ s⁻¹ to 9.7mM⁻¹ s⁻¹ as the thickness of dSiO₂ shell increased from 2.6 nm to 16.9 nm, while r₂ values almost doubled (from 57.58 mM⁻¹ s⁻¹ to 102.08 mM⁻¹ s⁻¹, Fig. 10e).

Fig. 10.

(a) Schematic illustration of coating UCNPs with dSiO₂ and mSiO₂ shells. (b) TEM images of (b) UCNP@mSiO₂ and (c) UCNP@dSiO₂ with varied shell thicknesses. Scale bar: 20 nm. (d) The plot of r₁ and r₂ versus mSiO₂ shell thickness. (e) The plot of r₁ and r₂ versus dSiO₂ shell thickness. Reproduced with permission from ref. 159. Copyright 2014, Wiley-VCH. (f) Schematic of the synthesis of HMnO@mSiO₂ NPs and labeling of mesenchymal stem cells. HMnO denotes hollow structure manganese oxide NPs. (g) TEM images and r₁ value of MnO–lipid PEG, MnO@dSiO₂, HMnO@mSiO₂ NPs. (h) No MRI contrast enhancement (red arrow) was detected in mice transplanted with unlabeled mesenchymal stem cells into the brain, whereas hyperintense signals (green arrows) were detected in mice transplanted with HMnO@mSiO₂-labeled mesenchymal stem cells, which were still visible 14 days after injection in the brain. The scheme and figures (with minor modifications) are licensed under the ACS AuthorChoice license. Reproduced with permission from ref. 62. Copyright 2011, American Chemical Society.

In another study, both mSiO₂ and dSiO₂ shells were used to coat the surface of MnO NPs.¹⁶⁰ With these modifications, the r₁ value decreased substantially from 0.17 mM⁻¹ s⁻¹ for MnO NPs to 0.07 mM⁻¹ s⁻¹ for MnO@dSiO₂. For MnO@mSiO₂ nanostructures, the r₁ value was similar to MnO NPs (0.16 mM⁻¹ s⁻¹), which was attributed to the existence of worm-like nanochannels in these NPs, where increased water exchange rates could be achieved. Similar results were also reported by Kim et al., who found that mSiO₂-coated MnO NPs showed a much higher r₁ value of 0.65 mM⁻¹ s⁻¹ than those of dSiO₂-coated NPs (r₁ = 0.08mM⁻¹ s⁻¹) and PEGlyated MnO NPs (r₁ = 0.11mM⁻¹ s⁻¹). The r₁ value could be further increased to 0.99 mM⁻¹ s⁻¹ after acid etching of MnO NPs to generate a hollow nanostructure (i.e., HMnO@mSiO₂) (Fig. 10f and g).⁶² Such a hollow structure allowed more interior Mn²⁺ ions to interact with water molecules on the inner surface, resulting in a significantly higher r₁ value than that of MnO NPs encapsulated with PEG phospholipids. ⁶⁹ Therefore, intracranial transplantation of HMnO@ mSiO₂-labeled mesenchymal stem cells could be monitored by serial MRI over 14 days (Fig. 10h). Among a number of surface modifications strategies of NPs with organic ligands and silica, Lee et al. found that silica-coated MnO exhibited the highest stability at neutral pH and the slowest dissolution rate when dispersed in acetate buffer solutions and showed continuous contrast enhancement in the rat brain.⁶⁰

3.2.3 Inorganic surface nanolayers

Coating an inorganic nanolayer (e.g., Fe₂ O₃, Fe₃ O₄, NaGdF₄, and so forth) on the surface of NPs to induce stronger magnetization and ensure adequate exposure to water molecules nearby is an attractive strategy to obtain high-relaxivity T₂ or T₁ CAs. MR relaxation enhancement effects of γ-Fe₂O₃ nanolayers were first reported by Xu and co-workers based on core/shell, yolk/shell, or hollow structures (Fig. 11a–d).⁷⁷ Field-dependent magnetization measurements showed that all these NPs were superparamagnetic (Fig. 11e), and the saturated magnetization (M_s) of the γ-Fe₂O₃ hollow nanostructure was the highest among these nanostructures (about 50.8 emu g⁻¹). Additionally, both the growth of a 5 nmFe_xO_y layer and self-assembly of ultra-small Fe₃O₄ on the surface of UCNPs generated much higher r₂ values than conventional iron oxide NPs.^161,162 Cheong et al. synthesized iron/iron oxide core/shell (Fe@Fe₃O₄) NPs by oxidizing the surface of iron nanocrystals in the air.¹⁶³ These core/shell Fe@Fe₃O₄ NPs had a r₂ value of 324 mM⁻¹ s⁻¹, more than twice that of Fe₃O₄ NPs (145 mM⁻¹ s⁻¹, Fig. 11h and i). Such an increase in r₂ value was attributed to the presence of a single-crystal iron core, which can induce much stronger magnetization than Fe₃O₄ NPs.

Fig. 11.

HRTEM images of (a) FePt@Fe₂O₃ yolk–shell NPs, (b) Pt@Fe₂O₃ yolk–shell NPs, (c) FePt@Fe₃O₄ core–shell NPs obtained by the seed-growth method, and (d) γ-Fe₂O₃ hollow NPs. (e) Room-temperature field-dependent magnetization measurements of different NPs. Reproduced with permission from ref. 77. Copyright 2008, American Chemical Society. (f) TEM image of 16 nm iron/iron oxide core/shell NPs. (g) HRTEM image showing a core of single-crystal a-Fe and a shell consisting of multiple domains of iron oxide. (h) T₂-weighted MR images at 9.4 T of iron/iron oxide core/shell NPs and iron oxide NPs. (i) r₂ values of the core/shell and oxide NPs determined from the same samples as in (h). Reproduced with permission from ref. 163. Copyright 2011, Wiley-VCH.

The coating of a NaGdF₄ layer is an intriguing approach not only to suppress surface quenching of UCNPs for enhanced luminescence emission but also to endow NPs with MRI capabilities, creating a dual-modality imaging system.^125,165,166 Results from different groups have suggested that the presence of Gd³⁺ ions on the surface of NPs might play a major role in accelerating relaxation, whereas the contribution from inner Gd³⁺ ions could be neglected; however, no direct evidence has been found to support this hypothesis.^46,48,49 To investigate this issue, our group designed Gd³⁺-doped core NPs with multilayer coating of NaYF₄ and/or NaGdF₄.¹⁶⁴ The r₁ value was calculated to be 1.24 mM⁻¹ s⁻¹ for the core NPs, which decreased to nearly zero (r₁ = 0.037 mM⁻¹ s⁻¹) after coating a ~4 nmNaYF₄ layer on the core NPs, and recovered to 1.18 mM⁻¹ s⁻¹ after further coating with a NaGdF₄ layer, clearly indicating the different effects between the bulk and surface Gd³⁺ ions in shortening T₁ relaxation time.

To further validate the hypothesis that Gd³⁺ ions in the inner crystal lattice contributed little to longitudinal relaxivity, Gd³⁺-free core NPs were coated with a NaGdF₄ layer of varied thicknesses (Fig. 12). With an increase in NaGdF₄ layer thickness from approximately 0.2 nm to 3.7 nm, r₁ values decreased gradually from 6.18 mM⁻¹ s⁻¹ to 2.18 mM⁻¹ s⁻¹, confirming that the main contribution to shortening T₁ values was by surfacebound Gd³⁺ ions on the NPs. Subsequently, these NaGdF₄ coated UCNPs were applied as dual-targeted nanoprobes by conjugating angiopep-2 peptide to cross the blood–brain barrier and target brain tumors for MRI-aided preoperative diagnosis and intraoperative identification of brain tumors by upconversion luminescence in an animal model.¹⁶⁶

Fig. 12.

TEM images and schematic of Gd-free core (a, f and k) and core@NaGdF₄ with varied NaGdF₄ shell thicknesses of 0.2 nm (b, g, and l), 0.7 nm (c, h and m), 2.0 nm (d, i, and n), and 3.7 nm (e, j, and o). Yellow dots in l, m, n, and o represent Gd³⁺ ions. The r₁ values of silica-coated water-soluble core@NaGdF₄ NPs of various sizes are shown in the middle of the schematic illustrations. Reproduced with permission from ref. 164. Copyright 2011, Wiley-VCH.

3.3 Shape effects

The shape effects of NPs can be as important as the size and surface effects in determining the properties of nanomaterials. The unusual shapes of these NPs could potentially dictate their interactions with single cells or whole organisms.¹⁶⁷ Shape-dependent relaxivity of NP-based MRI CAs has been revealed by many different groups, where shape anisotropy of magnetic NPs or shape confinement of Gd³⁺-based CAs can improve the relaxation of nearby water molecules. Employing the highly controllable shape of spherical or cubic magnetic CoFe₂O₄ NPs, Song et al. found that shape was an important factor in determining coercivity of magnetic NPs due to the varying effects of surface anisotropy.⁸² Very strong localized magnetic field inhomogeneity was produced by shape anisotropy, and the generated magnetic field strength was found to be dependent on the shape of magnetic NPs due to demagnetization effect.¹⁶⁸ Shape-dependent enhancement of r₂ values was also observed with superparamagnetic NPs.^80,81,137 For example, faceted irregular cobalt ferrite nanostructures showed increased r₂ values when compared to spherical NPs of similar volume.⁸¹ Zhao et al. synthesized branched octapod-shaped iron oxide NPs (edge length of 30 nm, Fig. 13a) with an ultra-high r₂ relaxivity of ~679.3 mM⁻¹ s⁻¹,¹³⁷ which was a much more effective T₂ CA for small tumor detection than conventional spherical iron oxide NPs (Fig. 13d and e) because of their unique morphology and larger effective boundary radius (Fig. 13c). The significant enhancement was attributed to an increase in the effective radius of the magnetic core, a parameter that is impacted by shape anisotropy and directly affects r₂ values (Fig. 13b).^80,137

Fig. 13.

Shape effects on T₂ contrast. (a) TEM image (scale bar, 100 nm) and (b) higher magnification TEM image (scale bar, 20 nm) of octapod-30 NPs of uniform four-armed iron oxide NPs. (c) A cartoon showing octapod and spherical iron oxide NPs. With the same geometric core volumes, the octapod NPs have much larger effective volumes (radius, R) than the spherical NPs (radius, r). (d) Comparison of r₂ values of different iron oxide NPs. (e) In vivo sagittal MR images and (f) quantification of signal-to-background ratio (*P = 0.01) of orthotopic liver tumor models at 0, 0.5, 1, 2 and 4 h after intravenous injection of octapod-30 and spherical-16 iron oxide NPs. Reproduced with permission from ref. 137. Copyright 2013, Nature Publishing Group. (g and h) Top: TEM images of 16 nm iron oxide NPs (g) and 50 nm NRs (h). Bottom: Local magnetic field generated by the NRs and spherical Fe₃O₄ NPs of equivalent material volumes under an applied magnetic field of 3 T. (i) Schematic of the quantum mechanical outer sphere model of Fe₃O₄ NPs and NRs of the same material volume. Reproduced with permission from ref. 80. Copyright 2015, Royal Society of Chemistry.

Recently, Mohapatra et al. suggested that although spherical iron oxide NPs exhibited higher magnetization than iron oxide nanorods (NRs) of the same volume,⁸⁰ high relaxivity (r₂ = 608mM⁻¹ s⁻¹) was obtained for 50 nm NRs (Fig. 13g and h), while the relaxivity of the same volume of spherical NPs (the volume of a 50 nm NR was equivalent to that of a 16 nm NP) was only 297 mM⁻¹ s⁻¹. This phenomenon was believed to be a result of the larger surface area of the anisotropic morphology for NRs. Using the software Comsol Multiphysics, the induced local field inhomogeneity surrounding the 50 nmNRs was calculated to be significantly stronger than that induced by 16 nmNPs (Fig. 13g and h). The higher outer sphere diameter of rod-shaped NPs resulted in a stronger local magnetic inhomogeneity with a larger number of water protons experiencing rapid dephasing, which gave a higher r₂ value. On the other hand, spherical NPs exhibited a smaller volume of magnetic field variation. Therefore, water protons diffused slowly and thus dephased slowly, resulting in a much lower r₂ value than NRs of equivalent volume (Fig. 13i).⁸⁰

Importantly, the shape effects of NPs on r₂ values cannot be translated to r₁ values because mechanisms for enhancements in r₂ and r₁ are different. An early study from Hyeon and co-workers demonstrated the significant effects of shape on r₁ values using spherical and tetrahedral Mn₃O₄ NPs.¹³² Recently, Rotz et al. described the surface conjugation of DNA–Gd onto gold nanostars and nanospheres⁹⁵ and found that DNA–Gd@stars exhibited remarkable r₁ values up to 98 mM⁻¹ s⁻¹, much higher than those of DNA–Gd@spheres, where r₁ values were 25 times larger than FDA-approved Gd chelates at 25 °C (Fig. 14a and b). Nuclear magnetic relaxation dispersion (NMRD) analysis indicated that secondary-sphere relaxivity contributed significantly to the production of such an exceptionally high r₁ value of DNA–Gd@stars, while secondary-sphere enhancements were not observed for DNA–Gd@spheres (Fig. 14c). The shape of gold nanostars with hydrophilic microenvironment played a meaningful role in the sequestration of water molecules around the Gd³⁺ complexes, which facilitated longer access to secondarysphere water molecules and thereby generated greater contrast enhancement. Further studies from the same group reported that r₁ depended on the number of branches on the nanostars, as r₁ increased with increasing average number of branches (Fig. 14d).⁹⁴ Researches from many other groups have also demonstrated that the shape of NPs can greatly impact their cellular uptake, biodistribution, clearance, and biocompatibility in vivo.^169-173 Thus, an in-depth understanding of the effects of shape on relaxation can enable the rational design of high-relaxivity CAs with outstanding performance for in vivo applications.

Fig. 14.

Effects of NP shape on T₁ relaxivity. (a) Schematic showing r₁ values of different kinds of MRI CAs including Gd³⁺ chelates, DNA–Gd@spheres, and DNA–Gd@stars. (b) NMRD profiles for water solutions of DNA–Gd@stars and DNA–Gd@spheres. (c) Simulated deconvolution of DNA–Gd@stars NMRD profiles into their inner, secondary, and outer sphere contributions. Reproduced with permission from ref. 95. Copyright 2015, American Chemical Society. (d) Schematic showing the dependence of relaxation on branch numbers of DNA–Gd@stars. Reproduced with permission from ref.94. Copyright 2016, American Chemical Society.

4. Strategies to achieve high relaxivity for MRI contrast enhancement

Size control, surface modifications (e.g., organic surface ligands, silica shells, and inorganic nanolayers), and shape regulation at the nanoscale have been widely applied to regulate the behavior of CAs, resulting in largely accelerated relaxation for efficient T₁ or T₂ MRI contrast enhancement. A deep understanding of the involved nucleation, growth, and surface chemistry of NPs enables intelligent engineering of the key factors that affect relaxation, either independently or in combination. The development of such high-relaxivity CAs is crucial for molecular MRI applications such as early-stage disease detection,^174-176 gene expression monitoring,^177,178 drug release,^140,179-181 and cell tracking.^182-185 In addition to the traditionally investigated factors influencing the performance of CAs at the nanoscale, several novel strategies have been exploited to obtain high-relaxivity CAs, such as confining CAs in mesopores, doping with metal ions, self-assembly or disassembly, and so forth (Fig. 15).

Fig. 15.

Schematic of engineering strategies to obtain high-relaxivity CAs for efficient MRI contrast enhancement.

4.1 Confinement in mesopores

The accessibility of water molecules to the paramagnetic center is highly important for the design of high-relaxivity CAs. Mesoporous silica nanoparticles (MSNs) provide an ideal platform for ready access of water molecules to the center due to their tunable pore structures and high surface areas. Dispersing CAs such as Gd³⁺ chelates,^90-92 MnO NPs,^61,64,139 and iron oxide NPs¹⁴³ into the mesopores of MSNs can efficiently prolong τ_R and thus greatly increase relaxivity. For example, Lin et al. loaded Gd–Si–DTPA complexes into the hexagon-ordered nanochannels (~2.4 nm diameter) of MSNs, obtaining extremely high r₁ values of 10.2 mM⁻¹ s⁻¹ at 9.4 T and 28.8 mM⁻¹ s⁻¹ at 3.0 T.⁹¹ The geometrical confinements of Gd³⁺ chelates, gadofullerenes, and gadonanotubes within the porous structure of silicon microparticles (SiMPs) (Fig. 16a) were all reported by Ananta et al.,¹⁸⁶ demonstrating that such a strategy can successfully influence the paramagnetic behavior of Gd³⁺ ions, resulting in increased r₁ values. By prolonging τ_R and reducing the mobility of outer sphere water molecules, r₁ values about 4–50 times larger than their original values without SiMPs were obtained (Fig. 16b).

Fig. 16.

(a) Schematics showing (i) Magnevist (MAG), (ii) gadofullerenes (GF), and (iii) gadonanotubes (GNT) and cartoons showing Magnevist, GFs and GNTs entrapped within the porous structure of SiMPs. (b) r₁ values of six MRI CA nanostructures in comparison with corresponding Gd-based CAs at 1.41 T and 37 °C. Reproduced with permission from ref. 186. Copyright 2010, Nature Publishing Group. (c) Schematic of Gd₂O₃ confined in MSNs. Reproduced with permission from ref. 187. Copyright 2016, Royal Society of Chemistry. (d) TEM image of hybrid mesoporous composite nanocapsules (HMCNs). Inset: STEM image with scale bar = 100 nm. (e) Relaxivity of an aqueous suspension of HMCNs after 4 h of soaking in buffer solutions at pH 5.0 and 7.4 at 37 °C. Reproduced with permission from ref. 61. Copyright 2012, Elsevier Ltd.

Recently, the use of such geometrical confinement for enhanced relaxivity was further confirmed by Gao and co-workers.¹⁸⁷ Ultrasmall Gd₂O₃ NPs with numerous coordinating water molecules were dispersed into worm-like mesoporous channels of MSNs, which, by increasing the tumbling time and creating a geometrically confined space, remarkably increased r₁ value from 14.93 mM⁻¹ s⁻¹ to 45.08 mM⁻¹ s⁻¹ at 0.5 T. (Fig. 16c). Thus, confinement of CAs in a porous structure offers a new and widely applicable strategy for strong T₁ contrast enhancement, where mesopores also hold the potential for drug loading and delivery for therapeutic applications.

As described above, loading small MnO or iron oxide NPs within the mesopores of MSNs was demonstrated to be more beneficial for relaxation enhancement than traditional silicacoated structures because such coating layers shielded the surface paramagnetic centers from interaction with the water molecules.^139,143 To further enhance the r₁ value of MnO NPs to exceed that of clinical Gd³⁺ chelates (e.g. Magnevist, r₁ = ~3.4 mM⁻¹ s⁻¹), we loaded small MnO NPs into hollow mesoporous silica nanoparticles (HMSNs) via an in situ redox reaction to further facilitate the interaction between manganese paramagnetic centers and water protons (Fig. 16d).⁶¹ Using this strategy, we were able to obtain a maximum r₁ value of 8.81 mM⁻¹ s⁻¹, which is more than twice that of commercial Gd³⁺ chelates (Fig. 16e). In addition, the well-defined mesoporosity of HMSNs can enable efficient encapsulation of photosensitizers (i.e., protoporphyrin) for therapeutic purposes, as well as further confinement of paramagnetic Mn ions based on metalloporphyrin chelating chemistry.¹⁸⁸ This mesoporous structure endowed these intriguing sensitizers with remarkably enhanced MRI performance (r₁ = 9.43 mM⁻¹ s⁻¹).

By a facile but efficient physical-vapor-infiltration method, ultra-small iron oxide NPs have also been confined to the mesoporous networks of MSNs.¹⁴³ In particular, the highly dispersed iron species in the mesoporous channels can be used as anchoring sites for anti-cancer drugs, such that gradual exposure of these iron paramagnetic centers along with controllable drug release from the mesopores enables monitoring of drug release in vivo through synchronously intensified MRI contrast. In another recent report, Zhang et al. incorporated both GdPO₄ and g-Fe₂O₃ NPs in ordered mesoporous carbon composites, which exhibited satisfactory r₁ and r₂ values of 2.7 and 183.7 mM⁻¹ s⁻¹, respectively.¹⁸⁹

4.2 Metal alloying and metal ion doping

The MRI performance of magnetic NPs is strongly dependent on their magnetic characteristics. Metal alloy-based NPs such as FeCo and FePt NPs are another class of CAs,^{77,146,190-194} which possess magnetic spins that align in parallel to the external magnetic field, resulting in typically higher magnetic moments than traditional ferromagnetic NPs. For example, the magnetic moment of FeCo alloys is approximately 2.4 μ_B per magnetic atom, while that of Fe₃O₄ NPs is approximately 1.3 μ_B.¹⁴⁶ These FeCo alloys have a comparably high M_s of 215 emu per gram of metal when coated with a graphite shell, leading to an exceptionally high r₂ value of 644 mM⁻¹ s⁻¹, much larger than that of conventional Fe₃O₄ NPs (usually about 100 mM⁻¹ s⁻¹). High X-ray absorption of Pt (6.95 cm² g⁻¹ at 50 keV) makes FePt alloy NPs promising candidates for MRI/CT dual-modality CAs.¹⁹⁴ By conjugating FePt NPs with an antibody targeting Her2/neu, in vivo dual-modality imaging was achieved in cancer lesions in an animal model. Recently, it was reported that FePt NPs could also be used as an infrared-active nanoplatform for four imaging modalities (optical/photoacoustic/CT/MRI), which may allow for multiscale imaging from the cellular level to wholebody imaging.¹⁹²

According to eqn (3), the r₂ value of magnetic NPs is proportional to their M_s. Thus, metal ion doping has been verified to be an effective method to engineer the magnetism of iron oxide NPs and obtain extremely high r₂ values, as investigated in detail by Cheon and co-workers.⁸⁴ They synthesized various types of MFe₂O₄ NPs in which Fe²⁺ ions can be readily replaced by other transition-metal dopants M²⁺ where M = Mn, Ni, or Co. The magnetic moments per unit of MnFe₂O₄, CoFe₂O₄, and NiFe₂O₄ were estimated to be 5 μ_B, 3 μ_B, and 2 μ_B, respectively, leading to gradually decreasing M_s values from 110 emu per gram of metal for MnFe₂O₄ NPs to 85 emu per gram of metal for NiFe₂O₄ NPs (Fig. 17a). As a result, MnFe₂O₄ NPs exhibited the highest r₂ value of 358 mM⁻¹ s⁻¹, two-fold higher than that of conventional iron oxide NPs. Such an enhancement is significant for clinical applications as the dosages of nanoparticulate CAs can potentially be decreased. By further co-doping Zn²⁺ and Mn²⁺ ions in (Zn_xMn_1−x)Fe₂O₄ NPs, the same group obtained NPs that exhibited an extremely high magnetization value (175 emu g⁻¹) and provided the largest r₂ value (860 mM⁻¹ s⁻¹) among the CAs reported to date (Fig. 17c).⁸³ In this study, it was also found that the magnetism was dependent on amount of metal ion doping and x = 0.4 doping of Zn²⁺ demonstrated the best contrast (Fig. 17b).

Fig. 17.

(a) TEM images, mass magnetization values, magnetic spin structures, magnetic moments, T₂-weighted images and colormaps of MnFe₂O₄, FeFe₂O₄, CoFe₂O₄, and NiFe₂O₄ NPs. Reproduced with permission from ref. 84. Copyright 2007, Nature Publishing Group. (b) Plots of r₂ versus Zn²⁺ doping level in (Zn_xM_1−x)Fe₂O₄ (M = Mn²⁺, Fe²⁺) NPs at 4.5 T. (c) Comparison of r₂ values of NPs, showing that Zn²⁺ doped NPs have significantly enhanced MRI contrast when compared to conventional iron oxide NPs. Reproduced with permission from ref. 83. Copyright 2009, Wiley-VCH.

4.3 Self-assembly and disassembly

“Activatable” or “smart” nanoprobes with tunableMRI relaxivity through self-assembly or disassembly in the presence of certain pathophysiological parameters can serve as high-relaxivity CAs, resulting in enhanced detection sensitivity of MRI for specific molecular or biological events. For example, such a self-assembly strategy allows for small NPs to circulate in the blood pool over a prolonged time duration, which can then be triggered in the tumor to self-assemble into a large structure for enhanced MRI. The trigger may include many physiological characteristics, such as molecular interactions,¹⁹⁵ enzymes,¹⁹⁶ receptors,¹⁹⁷ and redox status.¹⁹⁸ On the other hand, the disassembly strategy utilizes the different tumor microenvironments to release CAs preconfined in NPs, also achieving amplified MRI contrast. Both strategies are highly versatile and useful as they could be potentially used to monitor in vivo biological processes of the tumor microenvironment and improve the accuracy of cancer diagnosis or monitoring the therapeutic efficacy.

4.3.1 Self-assembly

Liang and co-workers reported the controlled self-assembly of Gd³⁺-based NPs through the condensation reaction between 1,2-aminothiol and 2-cyanobenzothiazole. ¹⁹⁹ The probe was activated through disulfide reduction by tris(2-carboxyethyl)phosphine or glutathione (GSH) in cells and generated intermediate 1,2-aminothiol groups to react with the cyano groups on 2-cyanobenzothiazole, which results in selfassembly of large Gd³⁺ NPs with enhanced relaxivity. Presumably due to an increased rotational correlation time, a more than 2-fold increase in r₁ value was observed for the larger assembly when compared to its small molecule form of Gd³⁺ chelates. Through further optimization of self-assembly strategy by intermolecular macrocyclization chemistry, the same group obtained biothiol-specific Gd³⁺-based CAs with amplified relaxivity after self-assembly, which were capable of detecting GSH concentrations with high sensitivity using MRI.²⁰⁰ In addition to redoxactivatable self-assembly, Ye et al. reported novel caspase-3/7 activated Gd-based CAs, which could self-assemble into Gd³⁺ NPs upon reduction and activation by caspase-3/7, an important early biomarker for apoptosis.²⁰¹ Higher r₁ relaxivity and longer retention time in apoptotic tumors were found with such self-assembling CAs, suggesting the potential for MRI of in vivo enzyme activity.

Self-assembly of small magnetic NPs into nanoclusters to increase effective magnetic size is also a widely used strategy for enhancing their magnetic properties and thus obtaining a higher r₂ value. Using charge compensation and electrostatic adsorption between oppositely charged moieties, Berret et al. assembled 6.3 nm iron oxide NPs into larger nanoclusters, resulting in remarkably increased r₂ values from 39 mM⁻¹ s⁻¹ for single NPs to 71 mM⁻¹ s⁻¹ for 70 nm nanoclusters and 162 mM⁻¹ s⁻¹ for 170 nm nanoclusters.²⁰² The detection sensitivity of MRI using these NPs was thus substantially improved. A similar effect was also reported by Ai et al. in their first proofof- concept study regarding clustering of monodispersed iron oxide NPs inside the core of polymeric micelles.²⁰³ When compared to single core nanoparticles, multi-core nanoclusters with an average core size of 50 nm (self-assembled from 13 nm iron oxide NPs) provided much higher M_s values and subsequently exhibited higher performance in MRI contrast enhancement.²⁰⁴

Although larger sized iron oxide NPs present enhanced magnetic properties for MRI contrast enhancement, an ideal scenario is the self-assembly of small iron oxide NPs into larger nanostructures at the target sites (e.g., tumor) for substantially amplified relaxation, since smaller NPs (below renal clearance threshold of approximately 10 nm) exhibit longer blood circulation half-life and hence are cleared more slowly from the blood pool than their larger counterparts. Mao and co-workers reported that ultrafine iron oxide NPs (~3.5 nm) with high T₁ relaxivity exhibited prolonged in vivo circulation life time and could be self-assembled or clustered into larger ones in the tumor, thereby switching from bright T₁ contrast to dark T₂ contrast in MRI.²⁰⁵ In another report, two complementary iron oxide NPs that were linked to a peptide substrate for enzymatic cleavage and a functional group (azide or alkyne) for click reaction¹⁹⁶ were able to self-assemble into a larger nanocluster network in the presence of matrix metalloproteinase enzymes, leading to approximately 160% enhancement of T₂ signal intensity.

Recently, Liang and co-workers reported a caspase-3/7- instructed biorthogonal cyclization to self-assemble Fe₃O₄@1 NPs in apoptotic cells and tumors for greatly enhanced T₂ MRI (Fig. 18a).¹⁹⁷ Caspase-3/7-mediated self-assembly of Fe₃O₄@1 NPs (r₂ = 185 mM⁻¹ s⁻¹) induced an approximately 65% increase of r₂ value when compared to the control group (r₂ = 112 mM⁻¹ s⁻¹) (Fig. 18b). More importantly, in vivo MRI revealed that Fe₃O₄@1 NPs provided specific T₂ enhanced contrast in apoptotic tumors, whereas the control Fe₃O₄@1-Scr NPs showed no contrast enhancement (Fig. 18c). In another recent study, Gao and co-workers synthesized Fe₃O₄ NPs modified with a responsive peptide sequence,¹⁹⁸ which enabled GSH-triggered self-assembly of Fe₃O₄ NPs within the tumor microenvironment and the formation of aggregates through inter-particle cross-linking reactions (Fig. 18d and e). Both in vitro and in vivo experiments demonstrated that such self-assembly strategies substantially enhanced MRI performance (Fig. 18f).

Fig. 18.

(a) Schematic of intracellular caspase-3/7-triggered aggregation of Fe₃O₄@1 NPs. (b) Plots of 1/T₂ versus metal concentration in the presence or absence of caspase-3/7. (c) In vivo T₂-weighted coronal images of Fe₃O₄@1 NPs or Fe₃O₄@1-Scr NPs in saline or DOX-treated (i.e., apoptotic) mice in 0 h (top) or 3 h (bottom) post injection. To make tumor apoptotic, about 8 mg kg⁻¹ of doxorubicin (DOX) was injected intravenously, once every 4 days three times. Reproduced with permission from ref. 197. Copyright 2016, American Chemical Society. (d) Schematic of ^99mTc-labeled Fe₃O₄ NPs and their responsiveness to GSH-triggering within the tumor microenvironment to form aggregates through inter-particle crosslinking reactions. (e) TEM images of the nonresponsive probe (left panel) and responsive probe (right panel) after being treated with GSH. (f) Temporal evolution of transverse relaxation rate R₂ for both the responsive probe and nonresponsive control recorded on a 3.0 T MRI scanner during incubation with GSH (inset: T₂-weighted images of probe solutions acquired at different incubation time points). Reproduced with permission from ref. 198. Copyright 2017, Wiley-VCH.

4.3.2 Disassembly

In general, disassembly strategies switch the relaxivity of CAs from “off” to “on” after responding to pathophysiological parameters such as low pH value and high GSH concentration within the tumor, thereby maximizing the signal from region-of-interest and minimizing the signal from background tissues to improve specificity and sensitivity of CAs. For example, Viger et al. encapsulated a large number of ultra-small Gd₂O₃ NPs in biodegradable polymer capsules, which showed extremely weak MRI contrast due to effective silencing of both inner and outer sphere relaxation.¹³¹ The r₁ of the ultra-small Gd₂O₃ NPs could be recovered after disassembling the large capsules under acidic conditions and/or in the presence of H₂O₂, which are both characteristic of many diseases such as cancer (Fig. 19a). In this study, encapsulation of ultra-small Gd₂O₃ NPs within pH-responsive polymer particles resulted in deactivation of their contrast enhancement at pH 7.4. However, burst degradation was observed within 1 min at pH 6.5 (approximately the extracellular pH of diseased tissue), leading to an instantaneous 4.2-fold increase of r₁ value, which remained constant thereafter (Fig. 19b). This concept was also verified using a peroxide-degradable polymer for encapsulation of ultra-small Gd₂O₃ NPs, indicating a general strategy of disassembly (Fig. 19c).¹³¹ A redox-modulated disassembly of Gd³⁺-based nanosystem was developed by Martinelli et al.²⁰⁶ The disassembly was achieved via the cleavage of S–S bridges in response to reducing environments, leading to the release of Gd³⁺ complexes and a subsequent significant change in relaxivity.

Fig. 19.

(a) Schematic of how a degradable polymermatrix is able to control the interaction of water molecules with Gd₂O₃ NPs (purple spheres). (b) Magnetic relaxivity of Gd₂O₃ NPs encapsulated in pH-responsive materials demonstrates a jump from neutral pH to mild acidity. (c) Increasing concentrations of H₂O₂ results in corresponding increases in the T₁ relaxation rates of Gd₂O₃ NPs encapsulated in an H₂O₂-responsive polymer. Reproduced with permission from ref. 131. Copyright 2013, American Chemical Society.

Similar to the above-mentioned strategies used for Gd-based NPs, the release of Mn²⁺ from the disassembled matrix at lowpH sites was also employed to obtain high-relaxivity CAs for pH-responsive MRI.^61,63 For example, Kim et al. reported core–shell Fe₃O₄@Mn₃O₄ NPs, where Mn²⁺ ions released from the Mn₃O₄ shell under reducing conditions allowed Fe₃O₄ NPs to interact with water protons, thereby enhancing contrast of both T₁ and T₂-weighted MRI.²⁰⁷ By confining Mn²⁺ within pH-sensitive calcium phosphate NPs (PEGMnCaP NPs, Fig. 20a), Mi et al. found a remarkable increase of r₁ from 4.73 mM⁻¹ s⁻¹ to 19.96 mM⁻¹ s⁻¹ after exposure to acidic environments simulating intratumoral pH, wherein the pH-sensitive CaP was disassembled (Fig. 20b).⁵² The enhanced MRI signal by the dissembled CaP matrix and released Mn²⁺ was attributed to the increase in relaxivity after binding of Mn²⁺ with the protein. Such Mn²⁺-triggered signal amplification could be further used for the precise imaging of tumors. After intravenous injection of PEGMnCaP NPs, enhanced MRI signal in tumor areas could be clearly visualized within 6 h after injection. Such a release of Mn²⁺ from the nanocomposites allowed these NPs to sensitively detect millimeter-sized metastatic tumors in the liver of mice, which were otherwise invisible (Fig. 20c). The signal of the gallbladder gradually increased due to hepatic clearance of PEGMnCaP. As hypoxic regions in tumors usually exhibit a lower local pH than surrounding tissues, the contrast in hypoxic regions was 20–30% higher than that in normally oxygenated regions within the tumors after intravenous injection of PEGMnCaP NPs (Fig. 20d and e).

Fig. 20.

(a) Schematic of the hybrid structure of PEGMnCaP. PEGMnCaP consists of a CaP-based core and a PEG shell. The Mn²⁺ ions are trapped in the CaP core. (b) r₁ value of PEGMnCaP in physiological environments at different pH levels, with and without proteins (for example, HSA). (c) MR images of liver metastasis using 1 T MRI scanner after i.v. injection of PEGMnCaP NPs. Scale bar, 1 cm. (d) MR images of a hypoxic region within a C26 tumor at 1 T, 4 h after the i.v. injection of PEGMnCaP NPs. (e) Staining of tumor tissues with pimonidazole confirmed that the hypoxic regions (brown) are at the same location as the tumor regions with higher MRI contrast enhancement. Scale bar, 1 mm. Reproduced with permission from ref. 52. Copyright 2016, Nature Publishing Group. (f) Schematic of the disassembly of PEG/Mn–HMSNs through “manganese extraction” and release of Mn²⁺ component intracellularly. TEM images showing structural evolution of Mn–HMSNs after biodegradation at a GSH concentration of (g) 5.0 and (h) 10 mM at pH 5.0 for 48 h. (i) Plot of 1/T₁ versus Mn concentration for PEG/Mn–HMSN solution at varied GSH concentrations and (j) at varied pH. Reproduced with permission from ref. 208. Copyright 2016, American Chemical Society.

By doping Mn²⁺ ions into HMSN (Mn–HMSNs), we have recently reported a strategy called “manganese extraction” to promote the disassembly of Mn–HMSNs and enhance MRI contrast (Fig. 20f).²⁰⁸ Such a Mn²⁺ doping approach enabled accelerated disassembly of Mn–HMSNs in either mildly acidic or reducing microenvironments by breaking the Mn–O bond and subsequently triggering “manganese extraction,” which further accelerated breakage of Si–O–Si bonds in the framework and biodegradation of HMSNs (Fig. 20g and h). Such a fast disassembly of Mn–HMSNs enhanced the interaction probability of Mn paramagnetic centers with water molecules, allowing pH- and GSH-responsive MRI to be realized (Fig. 20i and j). The disassembly of Mn–HMSNs also accelerated the release of anticancer drugs to obtain tumor-sensitive theranostic functions.

4.4 Other strategies

Valency change

In principle, any physical or chemicalmethod to change the interaction between CAs and water molecules such as the number of bound water molecules (q), rotational correlation time (τ_R), and water residence time (τ_M) of T₁ CAs or to enhancemagnetic properties of T₂ CAs will undoubtedly result in enhanced MRI relaxivity of CAs. One major strategy for MRI contrast enhancement focuses on valency changes of ions, which has been widely applied through using redox-activated chemical exchange saturation transfer (CEST) agents that can be detected by MRI.^85,86,209 As previously discussed, the valency of ions affects electron spin that in turn influences relaxation of water protons. Owing to its pH- and redox-responsive properties, MnO₂ can be reduced to Mn₂₊ by GSH or H₂O₂/H^+.51,53,210 For example, Chu et al. reported that MnO₂ nanosheets could serve as highly effective generators to trigger self-assembly of photosensitizers (sinoporphyrin sodium, DVDMS) after reduction by GSH and H₂O₂ (Fig. 21a).²¹¹ Such processes could be monitored by in vitro and in vivo MRI (Fig. 21b and c), which was validated by TEM imaging of ex vivo tumor tissues (Fig. 21d).

Fig. 21.

(a) Schematic illustration showing the fabrication process of MnO₂/DVDMS and the reaction in GSH solution or with H₂O₂ (pH 5.5). (b) Plot of 1/T₁ versus Mn concentration for MnO₂/DVDMS (black line), MnO₂/DVDMS + GSH (red line), and MnO₂/DVDMS + H₂O₂/H⁺ (blue line) solutions. (c) T₁-weighted MR images before and after injection of MnO₂/DVDMS. (d) TEM images of MCF-7 tumor sections at 24 h after injection of MnO₂/DVDMS. Reproduced with permission from ref. 211. Copyright 2017, Wiley-VCH. (e) UV-vis-NIR absorption spectra and (f) plot of 1/T₁ versus Gd³⁺ concentration of PEG–Na_xGdWO₃ nanorods after oxidization with H₂O₂ for varied time periods: 1#: 0 h; 2#: 0.5 h; 3#: 1 h; 4#: 3 h. (g) Schematic of the affinity of oxygen vacancies for oxygen atoms and its impact on the interaction between water molecules and Gd³⁺ ions. Reproduced with permission from ref. 37. Copyright 2017, American Chemical Society.

Research from our group also suggested that MnO₂ nanosheets could generate oxygen to overcome the hypoxic environment of tumors by reacting with H₂O^{2 53} and the released Mn²⁺ could be used for pH-responsive MRI.⁵⁵ Zhao et al. designed a redox-responsive MnO₂ nanosheet, in which manganese atoms were coordinated in an octahedral geometry to six oxygen atoms that prevented direct contact with the aqueous environment, therefore making no contributions to the alteration of relaxation of water protons.⁵⁴ Intracellular GSH could then reduce endocytosed MnO₂ nanosheets, generating numerous Mn²⁺ ions for MRI with greatly enhanced r₁ and r₂ values.

Oxygen vacancy

To date, the strategy of increasing the number of bound water molecules has not been employed in inorganic Gd³⁺-based NPs. Recently, we introduced oxygen vacancy into Gd³⁺-based NPs by designing a novel Gd³⁺-doped tungsten bronze (termed as PEG-Na_xGdWO₃ nanorods).³⁷ For the first time, controllable oxygen vacancies in inorganic nanosized CAs were generated for accelerated proton relaxation, favorably resulting in a substantially enhanced r₁ value due to the oxygen vacancies’ natural affinity to oxygen atoms in water molecules.²¹² Such a strategy was validated with oxygen-deficient PEG-Na_xGdWO₃ nanorods, which exhibited extraordinarily high r₁ values up to 80 mM⁻¹ s⁻¹ at 0.7 T and 32.1 mM⁻¹ s⁻¹ on a clinical 3.0 T scanner (Fig. 21e and f), offering excellent blood pool imaging and tumor diagnosis capabilities at a low CA dosage. NMRD profiles indicated that r₁ value was apparently increased by changes in the distribution of the number of bound water molecules (q) either through secondary- or outer-sphere effects (Fig. 21g).

Distance regulation

Recently, Choi et al. have introduced a distance-dependent magnetic resonance tuning systembased on the interaction between a superparamagnetic ‘quencher’ and a paramagnetic ‘enhancer’, where T₁ signal was tuned “on” or “off” depending on separation distance between the enhancer and quencher.²¹³ As the enhancer moved far away from the quencher (d₁ > d_c, where d_c is the critical separation distance), the electron spin fluctuation of the enhancer accelerated water proton relaxation, resulting in a stronger T₁ signal (“on” state); however, when the enhancer and quencher were in close proximity to one another (d₂ < d_c), the resulting spin fluctuation became too slow to induce effective water proton relaxation, resulting in a low T₁ signal “off” state (Fig. 22a). The feasibility of this strategy was demonstrated through the control of thickness of SiO₂ shell, thereby separating the enhancer and quencher from 18 to 2 nm (Fig. 22b). The d_c was set to 7 nm, where the T₁ relaxivity quenching efficiency was 50% (Fig. 22c). Measurements from X-band electron paramagnetic resonance and NMRD profiles revealed distance-dependent values of correlation time (τ_c) and T_1e (for details, please see eqn (2)), indicating that the enhancer electron spin fluctuation became slower as separation distance decreased. Furthermore, such a distance-dependent tuning effect of MRI signal was verified using three representative models of molecular interactions (i.e., cleavage, binding, and folding/unfolding) via modular combinations of enhancers and quenchers though interactive linkers (Fig. 22d).

Fig. 22.

(a) Schematic representation of distance-dependent magnetic resonance tuning. (b) Schematic of a nanoscale ruler that shows a variable T₁ MRI signal, dependent on the separation distance between the paramagnetic enhancer (Gd-DOTA) and superparamagnetic quencher (i.e., 12 nm Zn_0.4Fe_2.6O₄ particle). Lower panel: T₁-weighted MR image and color map image of a solution containing the enhancer and quencher with varied separation distances. (c) A plot of r₁ values versus separation distance. (d) Schematic of modular combinations for the preparation of distance-dependent magnetic resonance sensors, operated using three different modes of interactions (cleavage, binding, and conformational changes) and subsequent T₁ MRI signal outputs. Reproduced with permission from ref. 213. Copyright 2017, Nature Publishing Group.

Tuning inhomogeneity

A novel strategy to greatly enhance r₂ relaxivity by tuning magnetic field inhomogeneity of iron oxide nanoclusters has been recently reported by Chen and co-workers.²¹⁴ Using both experimental validations and theoretical simulations, the group found that the clustering of iron oxide NPs could enhance local field inhomogeneity due to reduced field symmetry. Clusters of mixed 5 nm plus 15 nm iron oxide NPs exhibited a much higher r₂ value of 533.4mM⁻¹ s⁻¹, when compared to clusters containing 5 nm NPs only (r₂ = 231.6 mM⁻¹ s⁻¹) or 15 nm NPs only (r₂ = 358.3 mM⁻¹ s⁻¹) (Fig. 23a), indicating the important roles of particle distribution and fraction in generating local field inhomogeneity. By artificially involving iron oxide NPs with heterogeneous geometries (e.g., iron oxide cubes and plates to assemble clusters) (Fig. 23b and c), the levels of field inhomogeneity were further elevated when compared to iron oxide sphere clusters. The iron oxide cube clusters showed the highest r₂ value (589.3 mM⁻¹ s⁻¹) due to the presence of spinpolarized cubes in nanoclusters (Fig. 23d). Landau–Lifshitz–Gilbert simulations confirmed their hypothesis (Fig. 23e and f), suggesting that regulating magnetic field inhomogeneity is a powerful way to establish correlations between magnetization and T₂ relaxivity of iron oxide NPs.

Fig. 23.

(a) Columns showing r₂ values of iron oxide (IO) clusters, as well as the single IO-5 and IO-15 NPs. C1: clusters of 5 nm NPs only; C2: clusters of 15 nm NPs only; C3: clusters of mixed 5 nm NPs and 15 nm NPs. (b and c) TEM and high-resolution TEM images, as well as cartoons of clusters C6 (IO cubes) and C7 (IO nanoplates), respectively. (d) Columns showing r₂ values of IO clusters C6 and C7, as well as the single IO cubes and plates. (e and f) Simulation models and calculated stray fields for clusters C6 and C7, respectively. Color bars represent log₁₀(H_d), where H_d is the calculated stray field. Reproduced with permission from ref. 214. Copyright 2017, Nature Publishing Group.

5. Summary and outlook

The design and fabrication of high-relaxivity and ultra-sensitive CAs for MRI provides not only a great challenge to overcome, but also provides great opportunities to make significant contributions in the biomedical field. The imaging performances of commercially available MRI CAs have been far from optimal, which necessitates the development of novel nanosized CAs with large enough contrast effects for highly sensitive detection of lesions. Undoubtedly, controllable synthesis and facile engineering of inorganic nanoparticulate CAs and their excellent in vivo behavior as imaging agents will greatly deepen our understanding and subsequent effective diagnosis and treatment of diseases.

This review article provides a comprehensive summary of methods that have been investigated to construct highperformance CAs, including size control of Gd-, Mn-, Fe-, Dyand Ho-based NPs, surface modification with different ligands, coating of silica shells (dense or mesoporous shells) or inorganic nanolayers on NPs, and regulation of shape of NPs for accelerated relaxation of water protons. Some widely applied strategies to achieve high relaxivity forMRI contrast enhancement such as confinement of CAs in mesopores, doping with metal ions, and self-assembly or disassembly have been discussed. In addition, other recently developed novel strategies including valency change, oxygen vacancy, and distance-dependent tuning have been briefly described. While there is no simple answer to the question of “how do these factors impact relaxivity of NPs,” several principles that can guide the engineering of future NPs as high-performance MRI CAs are as follows:

1. Size effect of NPs will determine whether surface-tovolume ratio or τ_R is the dominant contributing factor for generating T₁ contrast with any given NP. Meanwhile, size-dependent magnetization of magnetic NPs can be utilized in the design of T₂ CAs with high r₂ values.

2. Surface modification of NPs is indispensable not only for controlling NP growth but also for increasing compatibility and functionalization of NPs. The surface of NPs can be further adjusted to enhance or reduce MRI signal after size, which depends on coating species, hydrophilicity, density, and thickness of organic ligands, silica shells, and inorganic layers.

3. Shape confinement of Gd³⁺-based CAs has recently been utilized to impact relaxation of water protons by engineering Gd³⁺ chelates into dense and hydrophilic microenvironments that are closely related to the shape of NPs. In the meantime, the magnetic field inhomogeneity enhanced by shape anisotropy of magnetic NPs can be artificially regulated to enhance T₂ relaxivity by reducing symmetry of the induced field.

4. Biodistribution, pharmacokinetics, and metabolism of NPs are also mediated by factors such as particle size, surface, and shape. The successful engineering of NPs as high-performance imaging agents must consider their behavior in vivo. Ultrasmall sized NPs with higher relaxivity are promising imaging agents because they can be excreted by the kidneys on a short time scale.

5. Strategies to achieve high relaxivity for MRI contrast enhancement are mainly focused on tuning the interaction between NPs and nearby water protons, either enhancing or reducing the contrast of MRI. By further exploiting tumor microenvironment characteristics such as acidic pH, unevenness of blood flow, and hypoxia, engineering of NPs for stimuliresponsive MRI can be realized for the precise imaging of tumors and detection of their pathological conditions with MRI.

Despite the recent significant progress and promising results in preclinical animal models, the investigation, evaluation, and optimization of NP-based MRI CAs for clinical use are still in their infancy. To facilitate and enable the successful translation of these CAs from laboratories to clinics, we outline several considerations for future research on the interdisciplinary cooperation among engineers, material scientists, biologists, clinicians, and others toward future (molecular) MRI (Fig. 24).

1. Although some factors/strategies have been explored by regulating the key parameters that influence the perturbation of protons by the dipolar field of NPs, we believe that more novel factors/strategies for the design of high-relaxivity CAs are on the horizon. In addition, because the biodistribution, excretion, and pharmacokinetics of NPs are dependent on their size, shape, and surface status, rigorous and extensive attention should be paid to deep understanding the co-relations among nanoparticle characteristics (e.g., size, shape, and surface status), in vivo metabolism, and MRI contrast enhancement efficiency. Through a basic understanding of how these factors work together, future clinical applications may be realized. Synergistic integration of these factors may offer promising approaches in fabricating low-toxicity and high-performance CAs that are amenable to clinical translation which will eventually benefit patients.¹⁸

2. Efficient systemic delivery of CAs to the targeted tissues is of utmost importance for accurate and sensitive detection by MRI, which can include passive accumulation of elaborately synthesized NPs to maximize enhanced permeability and retention (EPR) effect, as well as active targeting of NPs via various ligands such as antibodies, antibody fragments, peptides, small molecules, and proteins. The impacts of nanoparticle size, surface modification, and shape on proton relaxation have been well studied, while their relationship with the EPR effect remains to be further elucidated. Generally, ultra-small sized renal clearable NPs exhibit poor EPR effect.²¹⁵ Future studies should focus on the specific interaction between MRI contrast performance of CAs and EPR effect to develop a clinically relevant and useful strategy for obtaining high-relaxivity NPs with an optimal balance between renal clearance and strong EPR effects over a suitable time scale (e.g., less than a few hours).²¹⁶

3. The hypoxic microenvironment in various tumors has received tremendous attention from researchers because it is closely related to the tumor’s aggressiveness and resistance to therapy,^217,218 and a number of NPs have been used to effectively eradicate hypoxic tumors in preclinical animal models.^219-222 Although in this review article we have discussed various strategies based on NPs for efficient MRI, there has been no literature report on hypoxia-responsive NPs as MRI CAs except on several chemical complexes for hypoxia-related MRI such as T₁ or T₂ CAs,^{223 19}F-based probes,²²⁴ and CEST probes.⁸⁶ We believe that research on this subject is likely already in progress, and the NPs that respond to hypoxia via, for example, valency change, size/morphology variation, assembly, and disassembly for enhanced MRI, will likely play multiple roles in accurate and noninvasive delineation of hypoxic areas, more precise and effective treatment, and real-time and noninvasive monitoring of therapeutic efficacy.

4. The next generation of MRI CAs will be developed not only based on design principles of NPs described in this review article, but also on the concurrent development of advanced nanotechnology, novel MRI sequences and scanners, and new biological applications. The synthesis of these NPs needs further optimization, and NPs that are synthesized in an easy, low-cost, and large-scale way is prerequisite to future clinical translation, a challenge which may be overcome by seeking more facile, reproducible, and economic one-pot syntheses without harmful reagents. With an in-depth understanding of the involved processes, we firmly believe that good biocompatibility and biodegradability, high relaxivity, and novelMRI contrast mechanisms can all be achieved with the future generation of novel inorganic NPs to usher in the new era of 21st century precision medicine (Table 3).

Fig. 24.

Summary and outlooks of designing inorganic NPs for MRI.

Table 3.

Full names and their corresponding abbreviations

Full name	Abbreviation	Full name	Abbreviation
5-Aminolevulinic acid	5-ALA	Mesoporous silica nanoparticles	MSNs
Catechol	CC	Mercaptosuccinic acid	MSA
Chemical exchange saturation transfer	CEST	Motional averaging regime	MAR
Computed X-ray tomography	CT	Nanoparticles	NPs
Dense silica	dSiO2	Nanorods	NRs
Diethylenetriaminepentaacetic acid	DTPA	Nuclear magnetic relaxation dispersion	NMRD
Diffusion correlation time	τD	Number of bound water molecules	q
Diphosphate	DP	Poly(acrylic acid)	PAA
Contrast agents	CAs	Photoacoustic	PA
Distearyl phosphatidylethanolamine	DSPE	Polyethylene glycol	PEG
2,3-Dimercaptosuccinic acid	DMSA	Polyethylenimine	PEI
Effective magnetic moment	μeff	Polyvinylpyrrolidone	PVP
Electron spin angular momentum	S	Poly(maleic anhydride-alt-1-octadecene)	PMAO
Electron orbital angular momentum	L	Poly(maleic acid)-octadecene	PMO
Enhanced permeation and retention	EPR	Reticuloendothelial system	RES
Glutathione	GSH	Silicon microparticles	SiMPs
1-Hexadecylamine	HDA	Rotational correlation time	τR
High-resolution	TEM	HRTEM Saturation magnetization	Ms
Hollow mesoporous silica nanoparticles	HMSNs	Sinoporphyrin sodium	DVDMS
Hydroxamate	HX	Solomon–Bloembergen–Morgan	SBM
Lanthanide	Ln	Static dephasing regime	SDR
Magnetic resonance angiography	MRA	Transmission electron microscopy	TEM
Magnetic resonance imaging	MRI	Total electron angular momentum	J
Magnevist	Gd–DTPA	Upconversion nanoparticles	UCNPs
Mesoporous silica	mSiO2	Water residence time	τM

Biographies

Dalong Ni

Dalong Ni received his PhD degree in 2016 from the Shanghai Institute of Ceramics, Chinese Academy of Sciences (SICCAS), under the supervision of Prof. Wenbo Bu. He then joined the Department of Radiology at the University of Wisconsin-Madison as a postdoctoral fellow under the supervision of Prof. Weibo Cai. His research interests focus on design, synthesis, and biomedical applications of multifunctional nanoplatforms for imaging and therapy applications, such as MRI and MRI-guided cancer therapy.

Wenbo Bu

Wenbo Bu received his PhD degree from Nanjing Tech University (China) in 2002. He has been a full-time professor in SICCAS since 2008. In 2016, he moved from SICCAS to East China Normal University. His research mainly focuses on design and synthesis of multifunctional rareearth- based nanomaterials for cancer imaging and therapeutic applications.

Emily B. Ehlerding

Emily B. Ehlerding received her MS degree in Medical Physics from the University of Wisconsin- Madison in 2016 and her BS in Chemistry and Physics from Manchester University in 2014. She is currently a PhD candidate in Medical Physics at UW-Madison, working with antibody-based platforms for multimodality imaging of immunotherapy-related targets.

Weibo Cai

Weibo Cai received his PhD degree in Chemistry from University of California, San Diego, in 2004 and is now an associate professor at the University of Wisconsin- Madison (http://mi.wisc.edu). He has published more than 230 peerreviewed articles (H-index 4 60) and has received many awards such as the Society of Nuclear Medicine and Molecular Imaging (SNMMI) Young Professionals Committee Best Basic Science Award (2007), the European Association of Nuclear Medicine Springer Prize (2011 & 2013), American Cancer Society Research Scholar (2013–2017), and UW-Madison Vilas Distinguished Achievement Professor Award (2017). His trainees have received more than 80 awards to date, such as the 2012 Berson-Yalow Award from SNMMI.

Jianlin Shi

Jianlin Shi received his PhD degree from SICCAS. He has been a full-time professor at SICCAS since 1994. His research interests include the synthesis of mesoporous silica nanoparticles and mesoporous silica-based nanocomposites for biomedical applications. He has published over 400 scientific papers, which have been cited more than 22 000 times by other scientists, and has an H-index of 79.