1 Introduction

The need for methods to assess renal function that are both robust and noninvasive is increasingly becoming recognized. One approach is to use MRI in combination with contrast agents that are responsive to physiological changes [1]. The most common type of MR contrast agents are paramagnetic agents that enhance contrast on MR images. These contrast agents act by modulating proton (1H) relaxation, either during recovery (T1 component) or decay (T2 component) of the MR signal. In order for a contrast agent to be useful to identify changes in a specific physiological phenomenon, its sensitivity to detect and quantify physiological changes on MRI and its pharmacokinetic properties must be balanced [2].

From a technical point of view, paramagnetic contrast agents have a concomitant effect on all components of MR relaxation; the interplay between T1 and T2 relaxation depends on the dose and location of the contrast agent as well as the MR imaging sequence used [3]. This complexity is exacerbated by intrinsic factors such as deoxygenated blood or air cavities that may confound the signal being investigated [4]. A case in point is when imaging cells labeled with contrast agent: changes in contrast created by such intrinsic tissue factors could render the labeled cells indistinguishable from the surrounding tissue. This drawback is heightened by the signal sensitivity constraints due to the low concentrations of contrast agent (<μM range). As a result, any quantification of the physiological observation is hampered, and in some cases, no definitive conclusions can be made.

The use of fluorinated agents in combination with fluorine (19F) MRI overcomes the hurdle of signal ambiguity because the 19F MR signal is specific to the administered 19F agent. While the 19F nucleus has similar MR sensitivity and resonance frequency as 1H, unlike 1H, it is practically absent in vivo, that is, there are no endogenous 19F organic forms in vivo. Therefore, any exogenous fluorinated compounds introduced in vivo can be detected with high selectivity and absolute specificity. This has been exploited to noninvasively track cells in vivo using 19F MRI after labeling them with 19F compounds, typically in the form of nanoparticles (NP). Interestingly, the chemical shift range for 19F is much larger than that of 1H. Organofluorine compounds span a chemical shift range of over 350 ppm [5]. Thus multiple 19F nuclei resonances for organofluorine compounds can be easily separated, even under rather inhomogeneous conditions [6]. The unique properties of 19F nuclei as well as 19F compounds have been the driving force behind the increasing interest in 19F MRI. However, the application of 19F MRI to a wide range of biomedical research fields has been restricted by the low abundance of 19F nuclei in vivo following exogenous administration. This constraint is compounded by the fact that the signal sensitivity of current MR hardware remains limited, making the detection of 19F compounds present at low concentrations extremely challenging. Therefore, one strategy in 19F MRI to circumvent this is to boost the signal-to-noise ratio (SNR) as a function of time, that is, the SNR efficiency to enhance signal sensitivity and improve detection limits.

From a biological perspective, several key components must be considered when using contrast agents to study renal function. Among other physicochemical properties, size will determine the pharmacokinetic properties of the contrast agent as well as its uptake by cells in vivo. For example, low-molecular-weight-gadolinium-based contrast agents (GBCAs) are filtered through the glomerulus and are not reabsorbed by the renal tubule. Conversely, larger contrast agents such as paramagnetic iron oxide particles or 19F nanoparticles commonly follow a one-compartment pharmacokinetic model within the blood [7]. These nanoparticles are too large to be filtered by the glomerulus and are eventually taken up by phagocytic blood cells. Such nanoparticles including the iron replacement therapy ferumoxytol have a long intravascular half-life (>14 h) and are therefore useful for vascular and perfusion-weighted MRI [8]. Unless inflammatory processes are ongoing, these particles will localize primarily in the liver and will eventually undergo elimination via hepatic clearance. The clearance of nanoparticles via liver or kidney is dependent on their size and surface properties [9] but also on the health status of these organs, such that clearance should be critically studied [10] especially prior to patient studies.

This introduction chapter is complemented by two separate chapters describing the experimental procedure and data analysis, which are part of this book.

This chapter is part of the book Pohlmann A, Niendorf T (eds) (2020) Preclinical MRI of the Kidney—Methods and Protocols. Springer, New York.

2 Overview of Applications

Quantitative measurements of renal function are necessary to detect changes already during the early stages of kidney disease, to support disease staging and to screen emerging therapeutic strategies. Noninvasive and rapid MR methods are needed to quantify disease components such as inflammation, hypoxia and tissue damage. While commonly available MR contrast agents act by modulating 1H relaxation, contrast agents containing 19F have also been developed to investigate disease components such as inflammation and to quantify pathological changes in the tissue microenvironment (e.g., changes in tissue oxygenation (pO2) or pH) using 19F MRI.

2.1 19F MRI to Study Environmental Changes in the Kidney In Vivo

There is an abundant repertoire of exogenous MRI contrast agents that act by modulating 1H relaxation and which are triggered by changes in pO2, pH or temperature as well as changes in protein, metabolite or metal ion concentrations (reviewed elsewhere [1]). Here we focus on the application of 19F probes that elicit a desired MR signal following changes in pO2 and pH and hold promise as tools for screening kidney disease.

2.1.1 Renal Tissue Oxygenation

Healthy kidneys operate under relatively low O2 conditions, requiring extensive extraction of O2 from flowing blood. While several regulatory mechanisms act to maintain oxygenation within homeostatic limits, the kidneys are nonetheless particularly susceptible to hypoxic injury [11]. While O2 consumption determines perfusion in most nonrenal tissue such as the brain, perfusion determines O2 consumption in the kidney [12]. Therefore, a sudden temporary impairment of the blood flow to the kidney results in acute kidney injury (AKI) as a result of impaired O2 delivery. Approaches for identifying and treating AKI as early as possible should aim at detecting and restoring imbalances the homeostasis between renal oxygenation and factors such as nitric oxide that control the renal microcirculation [13].

Noninvasive methods to measure kidney function in vivo, specifically tissue pO2 and blood perfusion, are crucial. Arterial spin labeling MRI determines intrarenal perfusion without the need of exogenous contrast agents, and will be reviewed in a separate chapter of this book. MRI techniques that utilize the native 1H contrast from blood—blood-oxygenation-level-dependent (BOLD) MRI—have been proposed [14]. The level of renal oxygenation within each voxel is quantified by mapping the shortening of 1H T2* induced by deoxyhemoglobin—deoxyHb [15]. However, the relation of T2* to tissue pO2 may be confounded by factors such as hematocrit, oxyHb dissociation curve, and the vascular volume fraction. This underscores the need for careful scrutiny of this relation in order to make quantitative interpretations [16].

Perfluorocarbon (PFC) emulsions offer an alternative to quantify changes in tissue pO2 during vascular pathology [17, 18]. More recently PFC NPs were also employed as a theranostic strategy for studying perfusion and treating AKI [19]. Primarily PFC NPs have been used to assess perfusion in the renal microvasculature following AKI [20]. 19F MRI sensitively detected decreased RBV and pO2 in the cortical–medullary (CM) junction 24 h following unilateral (left) renal ischemia–reperfusion [20], where increased T2* was also observed by 1H BOLD MRI (Fig. 1a). The T2*, 19F signal, and pO2 in the renal cortex of injured kidneys were all comparable to the contralateral control kidney, suggesting recovery of perfusion and oxygenation in this region. In the inner medulla, however, vascular leakage and hemorrhage-induced extravascular retention of PFC NPs (Fig. 1b, c) resulted in a false readout of reduced 1H T2*, increased 19F signal intensity, and unchanged pO2.

Fig. 1
figure 1

Assessing perfusion in the renal microvasculature following acute kidney injury using PFCs. (a) Composite Representative 1H T2*-weighted image (top row), 19F spin density weighted image (middle row), and pO2 map (bottom row) in the left control and right injured kidneys of the same mouse. (b) Fluorescence images showing FITC-lectin labeled perfused blood vessels in control and injured kidneys (top row) and zoom-in view of FITC-lectin (green) labeled perfused vessels in the CM junction (bottom row). (c) Fluorescence images showing rhodamine labeled PFC NPs in injured and control kidneys (top row) and zoom-in view showing PFC NPs (red) in the renal medulla (bottom row). Adapted from Hu L. et al. (2013) [20] with permission from John Wiley & Sons Inc.

PFCs and oxygen exhibit similarly low cohesive energy densities, such that both require little energy for mutual solubility [21]. The high oxygen-dissolving capacity of PFC emulsions makes them ideal as O2 carriers. PFCs were, in fact, considered as artificial blood substitutes [22]. Fluosol-DA—a mixture of perfluorodecalin (PFD) and perfluorotributylamine—was the first PFC emulsion studied in humans in a trial of Japanese patients [23]. It was soon recalled due to its rather low oxygen transport capacity compared to that of red blood cells [24]. PFCs ultimately failed as blood substitutes (even as next generation products appeared in the clinic). Although O2-deprived tissues easily extract O2 in PFCs, the oxygen dissociation curve of PFC emulsions is linear and monotonic, in contrast to the sigmoidal hemoglobin dissociation curve. Therefore, most of the O2 dissolved in PFCs is released at high O2 partial pressure in the arteries, with little O2 being available for the capillaries where O2 partial pressure is low but needed the most [22]. Interestingly Fluosol-DA (the first PFC that was used in patients [23]) was also used to acquire the first 19F MR images in vivo, showing its accumulation in the rat liver [25]. The PFC in this emulsion gives complex MR spectra and has since been replaced by other PFCs with better imaging properties [26].

The longitudinal relaxation rate (1/T1) of 19F in PFCs was shown to depend linearly on O2 partial pressure [27]. This led to the first noninvasive in vivo tissue O2 assessments in tumor tissue and liver [28] and later in myocardial tissue [29]. The latter study provided the first proof of concept study for rapid, noninvasive measurements of O2 oxygen tension changes in response to ischemia and reperfusion [29]. It is claimed that the size of PFC nanoparticles (>100 nm) confers upon them a good safety profile without renal toxicity, since they are not expected to be cleared through glomerular filtration. Therefore, unlike other imaging agents that undergo renal clearance (e.g., iodinated CT compounds, gadolinium MRI chelates), PFC emulsions do not increase the workload of the kidney and are not immediately considered as nephrotoxic [10].

2.1.2 Acid–Base Homeostasis in the Kidney and pH Sensors

The kidney plays a major role in acid–base homeostasis; this balance depends on its capacity to reabsorb filtered bicarbonate, mainly from the proximal tubule, and to excrete an equivalent amount of protons at the distal tubule [30]. CEST MRI was originally developed to measure pH with exogenous pH-responsive MRI contrast agents [31]. The pH in the kidney following AKI gradually increased from 6.7 to 7.2 as measured with CEST-MRI pH-responsive probes [32]. A major drawback of paramagnetic probes is that signal intensity is not only proportional to the pH but also to contrast agent concentration. Particularly in tumors, in vivo pH mapping assumes a constant contrast agent concentration across tissue notwithstanding poor perfusion where acidolysis occurs [33].

Approaches such as ratiometric 19F/1H probes have been developed to overcome the issue of quantification; the signal is normalized to the concentration of the contrast agent by means of a concentration reporting (19F) moiety [34]. Another advantage of 19F MR probes is the larger range in chemical shifts in 19F (compared to 1H). These chemical shifts are highly sensitive to the immediate surroundings, such that small chemical changes can lead to dramatic changes in chemical shift [6]. The first 19F MR probe identified as a candidate to probe cellular pH was the vitamin B6 analog 6-fluoropyridoxol (6-FPOL) reported to have a pKa in the physiological range and large chemical shift response [35]. Subsequently the feasibility of using 6-FPOL to probe cellular pH was demonstrated in whole blood [36]. The concept of employing 19F MR probes for in vivo imaging was introduced when a series of fluorinated vitamin B6 analogs with a chemical shift sensitivity in the range 7.4–12 ppm were synthesized as pH indicators [37]. The sensitivity of 19F MRS probes can be further improved by incorporating paramagnetic metal ions that further amplify the chemical shift between protonated and deprotonated forms by the presence of covalent and dipolar contributions [38, 39]. Iron (II) complexes were also explored to exploit the pH-dependent spin-state population change; the 19F NMR chemical shift was in a sensitivity range of 13.9 ppm per pH unit at 37 °C [40]. More recently fluorinated nickel (II) complexes with diamagnetic and paramagnetic properties were designed for ratiometric pH mapping by 19F MRI [41]. Supramolecular nanostructures formed by fluorinated peptide amphiphiles transitioning from cylindrical to ribbon-like shape as pH increases from 4.5 to 8.0 resulting in enhanced MR signals. These nanostructures represent a good strategy to design stimulus-responsive 19F MRI contrast agents for responses to pH changes in vivo [42].

2.2 19F MRI to Study Renal Inflammation In Vivo

Together with renal tissue hypoxia and tubular injury, inflammation is one underlying component in the pathology of AKI [43]. Studies suggest that systemic inflammation causes ischemic injury in one organ and that repercussions follow in distant organs downstream of the ischemic event [44,45,46] (Fig. 2). Therefore, methods that can non-invasively quantify the level of inflammation in different organs would be crucial to identify the causality and significance of inflammation during both systemic and kidney-specific disease.

Fig. 2
figure 2

Inflammation as one underlying factor in the pathology of AKI. AKI can be triggered by a variety of clinical disorders ranging from renal ischemic events to sepsis and exposure to nephrotoxic substances. The kidney is tightly connected with the interior milieu of the body such that even the slightest of changes will be rapidly detected and communicated between the kidney and the rest of the body

19F MRI is commonly used in association with an intravenous administration of PFC emulsions to study the distribution of inflammatory cells in vivo [26, 47,48,49,50,51]. 19F MRI is performed after an in vivo labeling of inflammatory cells with a 19F compound—typically in nanoparticle (NP) form. The 19F-NPs are administered intravenously in order for them to be taken up by inflammatory cells that are migrating through the systemic circulation into inflamed organs. Typically, PFCs such as perfluoro-15-crown-5 ether (PFCE) are used to prepare NPs for cell labeling. PFCE is biologically stable and provides a high payload of 19F nuclei since it contains 20 chemically equivalent 19F spins.

In a rat model of orthotopic kidney transplantation, 19F MRI was used to study inflammation in the kidney undergoing rejection [48]. Three days following transplantation, PFCE NPs were administered intravenously and 1H/19F images acquired 2 and 24 h following NP infusion (Fig. 3a). At the earlier time point, the 19F signal was mainly found in the circulation but also in small but roughly equal amounts in both the native and transplanted kidneys (Fig. 3a). By day 4 posttransplantation (i.e., 24 h following PFCE NP), the transplanted kidney underwent moderate (Grade 2) rejection and showed an intense signal in the allograft kidney and in liver and spleen, but the 19F signal was virtually undetectable above the level of image noise in the native kidney (Fig. 3a). T2*-weighted imaging reveals hemorrhaging in the medulla on day 4 post-kidney transplantation, which can be detected both in animals administered ultrasmall superparamagnetic iron oxide (USPIO) NPs but also in transplants without USPIO (saline control) (Fig. 3b). This can confound the detection of USPIO-labeled macrophages and exhibits one of the advantages of 19F MRI over iron oxide-based NPs.

Fig. 3
figure 3

Inflammation in the kidney in a rat model of orthotopic kidney transplantation experiencing rejection. (a) Composite 1H/19F images of an orthotopic kidney transplant recipient ∼2 h post–PCE injection on POD 3 with 19F signal found in the vasculature and equally in both the native (upper) and graft (lower) kidneys (left) and ~24 h post–PCE injection with 19F signal in the allograft kidney, as well as other organs (right). (b) Allograft kidney on POD 4 following an injection (left) of 6 mg/kg BW USPIO (upper row) or following an injection of saline (lower row) and 24 h later (right). Adapted from Hitchens T.K. et al. (2011) [48] with permission from John Wiley & Sons Inc.

Apart from labeling inflammatory cells in vivo, NPs have been used to label specific immune cell types such as dendritic cells (DCs) ex vivo, prior to their administration in vivo [52,53,54]. This kind of labeling is particularly useful to label cellular therapies that will be administered in patients as part of a treatment strategy [55]. In addition to cancer patients, kidney transplantation patients would also benefit from cell-based therapies, specifically the types of cells inducing tolerance [56]. Monocyte-derived suppressor cell products are currently being investigated alongside immunosuppressive therapies in kidney transplant recipients with promising results [57]. These candidate therapies would benefit from a labeling strategy that will enable them to be tracked in vivo over time with 19F MRI. The possibility of quantifying the 19F MR signal would allow an objective assessment of the spatial and temporal cellular distribution within the same individual. This kind of labeling and 19F MRI would result in absolute cell quantification, since the 19F MR signal is generated from the cells alone, by direct detection of the 19F spins in the cells labeled with 19F NPs, without any confounding background MR signal. However, several biological and technical considerations need to be made, in order to achieve a truly absolute MR signal quantification: not all cell types can be labeled with the 19F NPs currently available (Subheading 3.1.1).

3 Basic Concepts of 19F MRI

3.1 19F imaging Probes

The design and synthesis of novel fluorine-rich molecular labels and imaging probes is an important domain for 19F MRI. For several years, a wide range of fluorine probes has been available for experimental and even clinical applications [58, 59].

3.1.1 19F Probes for Cellular Uptake

Current procedures used to label cells with nanoparticles take advantage of the intrinsic nature of cells to engulf foreign bodies, a cellular property known as phagocytosis. Dendritic cells are intrinsically phagocytic and therefore ideal candidates for tracking with 19F MRI [52, 53, 55, 60,61,62]. Cells labeled with clinical grade 19F nanoparticles have been administered as immunotherapies in cancer patients [55]. This tracking method could support several other cell therapy studies, including those dealing with kidney transplantation [63].

Enhancing cellular uptake is desirable, even for cells that are intrinsically phagocytic. Several studies have investigated the incorporation of phosphatidylserine (PS) in nanoparticle-based therapeutics to enhance their uptake [64]. After renal ischemia–reperfusion, the signal from PS-containing microbubbles was two-fold higher than that for standard lipid microbubbles [65]. Phospholipids such as PS and phosphatidylethanolamine (PE) are important components of biological cell membranes and are important components for endocytosis in phagocytic cells such as DCs [66]. We previously reported that enrichment of 19F nanoparticles with 1,2-dipalmitoyl-sn-glycero-3-PE (DPPE) resulted in a striking increase (at least one order of magnitude) in cytoplasmic uptake in DCs [62]. Interestingly, membrane lipid-raft processes have been proposed as trafficking mechanisms for transporting therapeutic cargo such as siRNA [67] or targeting ligands [68] into the cytoplasm, circumventing the usual endosomal uptake pathway.

Labeling cells with low intrinsic phagocytic properties (such as T cells) will prove more difficult using the above strategy, even if the nanoparticle size is lower than 180 nm [69]. Some research groups claim to have labeled T cells using the same above principles [70]. T cells are not able to internalize nanoparticles , even small iron oxide nanoparticles (core diameter = 12.5 nm), as has been clearly shown by transmission electron microscopy (TEM), considered to be a definitive method, and also by confocal microscopy [71]. For the latter method nanoparticles with fluorescent tags were used to corroborate the in vivo 19F signals with ex vivo fluorescence microscopy signals, but the authors also included a cell membrane stain [71]. This is a necessary control to distinguish membrane association from intracellular distribution, in light of the point spread function contribution [72]. Another key point is that the fluorescent tag dissociates from the fluorine label over time, both in vitro and in vivo [73]. This has a critical impact on the interpretation of long-term experiments validated by standard fluorescent techniques.

Before taking the next steps toward administering cells in vivo, it is essential to understand how nanoparticles are recognized, internalized, trafficked, and distributed within the specific host cell [74]. This is critical when contemplating labeling of cells that may ultimately be applied in vivo to patients in a therapeutic context. Cells that are not intrinsically phagocytic will not readily internalize NPs but might still exhibit a 19F MR signal ex vivo, due to reasons such as transitional binding of NPs to the cell membrane or incomplete clearing of noninternalized NPs from the cell culture medium. The 19F MR signals following administration of ineffectively labeled cells with contaminating NPs would be misrepresentative of the actual distribution of the cells in vivo. Therefore, we need to ask the right questions regarding the specific cell type to be labeled, and the methods used prior to transfer of the cells into living organisms [75].

3.1.2 19F Probes with Increased Signal Sensitivity

Fluorinated probes are commonly designed with the goal of improving signal strength by increasing the number of fluorine atoms per molecule [59]. Linear 19F polymers such as perfluoropolyethers (PFPEs) are the simplest choice to achieve increased fluorine content but the NPs produced with these polymers are polydisperse and the presence of multiple chemical shifts presents several peaks in the 19F spectra [76]. Superfluorinated branched 19F probes such as 19FIT [77] and PERFECTA [78] promise several advantages, specifically a very high fluorine nuclei density with a single 19F spectral peak. These probes therefore afford enhanced sensitivity and better imaging performance than linear perfluoropolymers. Unlike PERFECTA [78], the bispherical fluorocarbon 19FIT is water soluble [77] and therefore possesses low organ retention, but it cannot be emulsified into nanoparticles for cell labeling strategies. Increasing the density of 19F nuclei can only go so far in terms of increased sensitivity of 19F probes. Ultimately, a limit to this density will be reached.

Another strategy to promote 19F MR signal is to manipulate the MR properties of the 19F probes in order to favor ideal conditions for MR signal acquisition. One approach is to decrease the T1 of the 19F nuclei, which is typically in the range of 0.5–3 s for diamagnetic compounds. Paramagnetic metal ions such as lanthanides(III) or iron(II) reduce the T1 of 19F compounds [38]. Paramagnetic centers affect the relaxation of 19F nuclei (paramagnetic relaxation enhancement or PRE effect) via a number of mechanisms, most notably electron–nucleus dipole–dipole interaction and Curie relaxation [39], both of which are influenced by the effective magnetic moment (μeff) of the paramagnetic metal [79]. Other than the μeff of the metal ion, the PRE effect is also influenced by the distance separating the 19F nuclei and the paramagnetic metal ion [39]. While long distances will severely diminish PRE, short distances will substantially shorten the spin–spin relaxation time T2. Thus T1 reduction by molecular design should be dealt with carefully since it could also lead to a reduction in the T2 and thereby line broadening of the resonance, which ultimately affects sensitivity [80]. The most sensitive metal-based probes are, in fact, not those whose metal ions have the highest μeff, but rather those who have the ideal distance separating 19F from the paramagnetic metal ion (4.5–7.5 Å) thus resulting in higher T2/T1 ratios [81]. Nonetheless even at larger 19F–M distances (>10 Å), the PRE effect can still be notable as is the case for a DOTA-based 19F chelator [82]. In paramagnetic PFC nanoparticles, the 19F MR relaxation enhancement has also been explained by diffusion mechanisms attributed to the changes in Brownian motion of PFC molecules inside the NP in conjunction with their ability to permeate into the lipid surfactant coating [83].

Introduction of paramagnetic metals in the design of 19F probes not only results in a decrease T1 of 19F nuclei but also leads to paramagnetic induced chemical shifts. The CEST signal generated by paramagnetic involvement (PARACEST) will also be modulated by pH (Subheading 2.1.2). A macrocyclic Fe(II) complex bearing both exchangeable 1H and 19F was shown to behave as a ratiometric pH-responsive MRI contrast agent, whereby the CEST, but not the 19F MR signal was pH-dependent [84]. This enables accurate pH mapping without the use of multiple probes. More recently a DOTA macrocyclic structure complexed with Ni(II) metal was shown to exhibit larger separations between the 1H CEST signal and water [85]. Though the in vivo application of 19F PARACEST probes remains to be investigated, the available complexes will likely provide valuable information in the near future when probing pH changes in vivo.

3.2 MR Data Acquisition

3.2.1 Pulse Sequence Optimization

One way of improving signal sensitivity is to avoid signal losses such as those occurring during spin relaxation. Most fluorinated compounds generally have long T1 relaxation times, which effectively increases the acquisition time. Together with the fact that there is a large range of relaxation times for 19F MR, this motivates more reliable optimization of pulse sequences that is informed by the MR properties of the 19F compounds under investigation. Sequence performance is largely dependent on the relaxation times of the addressed spin system but also on the conditions the nuclei are in, such as tissue oxygenation level [17].

The choice of the best pulse sequence and acquisition method is dictated by the MR properties, particularly the relaxation times, of the nuclei being measured. The acquisition protocol for each sequence should be optimized by tailoring detection schemes to the properties of the nucleus—also in relation to its environment—using relationships empirically drawn from simulations [86]. Relaxation times determine the time frames of signal losses, so the time orchestration for data acquisition should be assigned correctly. Other parameters, such as acquisition bandwidths, should also be properly planned to avoid recording unnecessary noise and to achieve the highest possible SNR [87].

A number of 19F MR studies using different pulse sequences have aimed to optimize the acquisition method to promote SNR efficiency and signal sensitivity [17, 87,88,89,90]. A large proportion of 19F MR studies, especially those applying perfluorocarbons (PFCs) in vivo, have employed the turbo spin echo (SE) rapid acquisition using relaxation enhancement (RARE) sequence [26, 47, 50, 51, 91]. This method reduces acquisition time by accumulating multiple echoes (dictated by the turbo factor) within a single repetition time [92]. Optimal settings for echo train lengths and repetition time (TR) according to measured relaxation times (T1/T2 combinations) were calculated for RARE to improve sensitivity thresholds [90].

Although RARE is commonly been for studying PFCs in vivo, other sequences will be better suited for 19F compounds that unlike typical PFCs have short T2 relaxation times. In such cases signal detection might be technically challenging and will demand acquisition strategies properly adapted for the compound and its environment. Gradient echo (GE) sequences were used to maximize SNR efficiency for imaging paramagnetic macrocyclic 19F molecules complexed to lanthanides [87]. These molecules exhibit T1 values in the order of 1–15 ms and T2* values in the order of 0.4–12 ms [88]. In the latter study a radial zero echo time (ZTE) sequence yielded a sensitivity gain of 27-fold when compared to nonparamagnetic molecules, while the GE sequence yielded a gain of 11-fold [88]. Additionally, the highest SNR efficiency gain was achieved when T2/T1 were close to 1, particularly for the ZTE pulse sequence , and when T1 was in the range of 1–5 ms [88]. Even 19F molecules with unique spectral properties will require a more individualized adjustment of the MR parameters; such molecules might benefit from sequences such as ultrashort echo time (UTE) that acquire signal before the spins dephase, taking into consideration destructive phase interference [93]. An efficient k-space sampling strategy will be necessary to achieve maximal SNR and artifact-free MR images [94]. To image perfluorooctyl bromide (PFOB) that has a rich spectrum with large chemical shifts, a radial UTE with a balanced steady-state free precession (bSSFP) pulse sequence showed superiority above other sequences, such as SE, GE, or bSSFP with Cartesian sampling [89]. However, radial sampling might not always be viable since k space sampling pattern effects need to be considered, which might altogether result in reduced SNR efficiency [86]. Another way to speed up data acquisition is synchronous multinuclear imaging. For this, the MR system needs to have the architecture suited to acquire images from more than one nucleus in a single pulse sequence [95]. Simultaneous proton and 19F MRI was performed in a preclinical model using a clinical MR scanner at 3 T [95].

3.3 Hardware Considerations

The similarity between 19F and 1H MR properties with respect to sensitivity and resonance frequency is an advantage, from an engineering perspective since similar pulse power settings can be used, and dual-tuned radio frequency (RF) coils for 19F and 1H MRI are more convenient to design [50, 96]. One way of improving the 19F MR signal in vivo is to increase SNR by improving the sensitivity of the measuring instrument. Two hardware strategies to improve sensitivity for 19F MR are to increase the strength of the static magnetic field (B0), and to increase the sensitivity of the RF coil by cryogenic cooling.

3.3.1 Cryogenic Cooling

The sensitivity of the RF coil is an important factor limiting SNR and thereby the level of spatial and temporal resolution that is targeted for 19F MRI [97]. Even when employing an RF coil with a homogeneous transmit field (B1+) distribution and dedicated for brain 19F and 1H MRI, the spatial resolution of in vivo 19F MRI was limited to approximately 600 μm [50]. This level of spatial resolution prohibits a detailed understanding of inflammation dynamics. To overcome the sensitivity constraints in 19F MR and improve detail of inflammatory cell location, we applied the concept of cryogenically cooling RF coil hardware to improve SNR by reducing thermal noise [98,99,100]. The cryogenically cooled RF coil (CryoProbe™) technology is commercially available for small animal MRI of 1H and X-nuclei such as carbon (13C) and phosphorous (31P). For these X-nuclei this technology is expected to be particularly useful for increasing SNR. A gain in SNR of 3.0- to 3.5-fold was reported for the 13C CryoProbe [101]. We recently compared a 19F CryoProbe with a room temperature 19F RF coil of similar size and reported on a maximal SNR gain of 15-fold and were able to study inflammation at an isotropic resolution of 150 μm [100].

Surface RF coils provide a very high SNR close to the region where they are applied. However, a decrease in signal amplitude with increasing distance from their surface is inherent to the geometry of their design. Inhomogeneities in the transmit (B1+) field cause a spatially varying flip angle (FA) and as a result signal quantification is hampered. Likewise, the receive (B1) field of surface RF coils is position-dependent [102]. In the particular case of the 19F CryoProbe, we observed a strong FA decrease with increasing distance from the surface of the coil [100]. Using a receive-only surface RF coil in combination with a volume resonator used for transmission would overcome B1+ inhomogeneities. With this configuration the transmitted FAs will be homogeneous, thus enabling quantification [103]. This option is not available when using CryoProbes, since they are typically built in the form of transceive (transmit–receive) RF coils with no active decoupling. Particularly for quantitative 19F MRI, a correction of the B1+ inhomogeneity will be necessary when comparing concentrations of PFC in vivo with an external standard of known 19F concentration. Differences in the B1+ field between the region of interest in vivo and the external standard will result in errors when quantifying in vivo 19F signals. Several approaches have been implemented to correct B1+ inhomogeneities retrospectively, even for 19F MRI [104, 105]. These approaches typically include RF pulse sequences for which an analytical knowledge of the signal intensity dependency on the FA exists.

3.3.2 Magnetic Field Strengths

Increasing the strength of the static magnetic field (B0) will increase SNR of the measured signal [106], a strategy which is actively being pursued for clinical application [107]. Intrinsic SNR is expected to grow at least linearly with increasing B0 strengths [97, 108, 109]. The principle of reciprocity can be used at low field strengths to calculate the receive field (B1) sensitivity of a single channel RF coil in terms of the transmit field (B1+) that can be easily measured [110]. In studies performed at frequency (f) <1 MHz the maximum sensitivity was calculated to be proportional to the f: sensitivity ∝ f1.75 [97, 108]. When going to higher magnetic field strengths, the increasing f and the influence of wave propagation need to be considered [106]. The homogeneity of the B1+ field is expected to decline with increasing B0, thereby influencing the overall SNR gain. An experimental study investigating SNR dependency on B0 in the human brain revealed that SNR ∝ B01.65 in the range of 3.0–9.4 T [109]. A recent simulation study also showed that SNR grows superlinearly with frequency, particularly in deeper regions of the sample; in less deep regions the SNR versus B0 trend approached linearity [111].

An increase in magnetic field strength is particularly beneficial for X-nuclei imaging. An increase in B0 from 9.4 to 21.1 T results in an SNR gain of ~3 compared to a gain of ~2 for 1H MR [112]. We recently studied the influence of increasing B0 on improving 19F MR signal sensitivity [113]. Prior to the practical experiments, we estimated the expected SNR gain when moving from 9.4 to 21.1 T to be 2.7 [113]. We estimated SNR gain using the principle of reciprocity [110] and ignoring all flip angle, sequence and relaxation-dependent effects. The estimated SNR gain represents a baseline expectation, based on the information available for the specific hardware of each system. When introducing the relationship of the reference power for both MR systems, we estimated an SNR gain of ~3.8. In our practical experiments the ratio in SNR efficiency (SNR/√t ratio) was estimated to be 7.29, when comparing the best possible conditions for both 21.1 and 9.4 T, and when including T1 relaxation effects [113]. Differences between the actual SNR gains determined experimentally and those expected from simulations are conceivable, due to minor inaccuracies in the assumptions made for the EMF simulations.

All research efforts ranging from new 19F probe design, through optimization of 19F MR data acquisition methods up to hardware developments hinge on the need for promoting 19F MR signal in vivo, in order to answer crucial questions during pathology including kidney disease. Most of the efforts devoted to discovery and proof-of-principle research will benefit from an integration of all available tools and technologies, in order to reach the maximum potential for practical applications for patient health care.