Gadolinium-modulated ¹⁹F signals from Perfluorocarbon Nanoparticles as a New Strategy for Molecular Imaging

Abstract

Recent advances in the design of fluorinated nanoparticles for magnetic resonance molecular imaging have enabled specific detection of ¹⁹F nuclei, providing unique and quantifiable spectral signatures. However, a pressing need for signal enhancement exists because the total ¹⁹F in imaging voxels is often limited. By directly incorporating a relaxation agent (gadolinium) into the lipid monolayer that surrounds the perfluorocarbon, a marked augmentation of the ¹⁹F signal from 200nm nanoparticles was achieved. This design increases the magnetic relaxation rate of the ¹⁹F nuclei 4-fold at 1.5 T and effects a 125% increase in signal, an effect which is maintained when they are targeted to human plasma clots. By varying the surface concentration of gadolinium, the relaxation effect can be quantitatively modulated to tailor particle properties. This novel strategy dramatically improves the sensitivity and range of ¹⁹F MRI/MRS and forms the basis for designing contrast agents capable of sensing their surface chemistry.

Introduction

Developments in the field of molecular imaging promise precise elucidation of structural and functional alterations that occur in vivo during disease development, progression, and treatment. Magnetic resonance imaging (MRI) provides many advantages for molecular imaging including superb resolution, excellent soft tissue contrast, and lack of ionizing radiation. However, sensitivity remains an issue for traditional paramagnetic contrast agents that are used to detect low density molecular epitopes because millimolar intravoxel concentrations of gadolinium-based agents are required to achieve diagnostic contrast-to-noise levels (1,2). The use of superparamagnetic agents may enhance sensitivity, but multiple susceptibility artifacts in the image field can challenge specificity, especially at higher field strengths. Furthermore, neither of these agents provides a unique signature as they are only detected indirectly through their effect on the signal from protons (water). The high concentration (~88 M) of protons distributed throughout the body, while providing a high intrinsic anatomical signal, nevertheless can also present as a potentially confounding and ubiquitous background signal for the detection of targeted contrast agents with MRI.

We have proposed and demonstrated a perfluorocarbon (PFC)-based nanoprobe for MR molecular imaging and spectroscopy that exhibits a fluorine (¹⁹F) signal which can be detected independently of the anatomical proton signal, while also allowing detection of multiple unique signatures with reasonable sensitivity for epitope detection with the use of slightly modified clinical imaging equipment (2–4). Recently, this approach has been extended to cellular labeling for MRI-based tracking (5,6). This nanoparticle comprises a liquid perfluorocarbon core of selected composition that is stabilized by a surrounding lipid-surfactant monolayer, creating a stable water-soluble emulsion. Molecular targeting ligands and gadolinium chelates can be incorporated into the lipid monolayer to provide localized signal amplification upon binding, which has been demonstrated both in vitro and in vivo (7,8). However, as compared to the proton signal measured in the case of the gadolinium T₁ contrast effect, the ¹⁹F signal from the perfluorocarbon core exhibits the added benefit of no confounding background tissue signal (9). Moreover, the ¹⁹F signal is quantifiable and can be used to differentiate nanoparticles with different perfluorocarbons based upon their unique NMR spectrum (4).

In this work, we sought to amplify the utility of ¹⁹F MRI/MRS by mimicking an approach used in ¹H imaging: namely, change the relaxation times of the ¹⁹F spins by adding an agent capable of altering local magnetic fields. To conceptualize this novel approach, we posed the question: “Would the proximity of a large number of gadolinium chelates (50,000–100,000) coupled tightly to the lipid monolayer of the perfluorocarbon nanoparticle affect the MRI/MRS signal from the ¹⁹F core material?” If this were the case, then it could provide the basis for designing a T₁ modulated nanoparticle ¹⁹F signal based on the proximate interactions between the gadolinium and perfluorocarbon. Previous studies have suggested that mixing perfluorocarbon emulsions with gadolinium chelates in solution results in almost no effect on the ¹⁹F relaxation (10). However, one study illustrated a measurable relaxation effect of gadolinium on aqueous ¹⁹F, suggesting that proximity of the gadolinium atoms to the ¹⁹F nuclei may be the critical factor for such a hypothetical interaction (11). A more recent study utilizing various fluorine-lanthanide complexes has verified that placing the lanthanide close to the fluorine nucleus results in shortened T₁ times and increases the sensitivity for NMR studies (12).

In our nanoparticle formulation, bisoleate DTPA derivatives are employed as chelates for the gadolinium ions, which constrains the final complex to the nanoparticle surface such that the separation from the perfluorocarbon core is no more than the length of the lipid monolayer components (~15 Å) but prevents direct contact between the gadolinium and fluorine nuclei. We postulated that this proximity condition could alter ¹⁹F relaxation and therefore serve as a mechanism for tailoring the ¹⁹F MR signal based on the presence or absence of the gadolinium. Such an agent could then potentially be used as a different type of “smart” contrast agent probe that could report on the biochemical status of its local environment using ¹⁹F signals, as others have shown for ¹H MRI probes (13–15).

Methods

Nanoparticle formulation and characterization

Nanoparticle emulsions were formulated as described previously(16,17). Briefly, they were composed of 20% (v/v) of perfluoro-15-crown-5 ether (C₁₀F₂₀O₅; Exfluor Research Corp., Round Rock, TX), 1.5% (w/v) of a surfactant co-mixture, and 1.7% (w/v) glycerin, with water comprising the balance. Perfluoro-15-crown-5 ether is a cyclic perfluorocarbon with 20 chemically equivalent fluorine atoms per molecule which produces a single resonance on the NMR spectrum. The surfactant co-mixture was comprised of lipophilic gadolinium-diethylene-triamine-pentaacetic acid-bisoleate (Gd-DTPA-BOA; Gateway Chemical Technologies, St. Louis, MO), lecithin (Avanti Polar Lipids, Inc., Alabaster, AL), and dipalmitoyl-phosphatidylethanolamine (Avanti Polar Lipids, Inc., Alabaster, AL) in varying molar ratios of 0:98:2 (no gadolinium), 10:88:2, 20:78:2, or 30:68:2 (30 mol% gadolinium). For targeting to human fibrin clots, the surfactant was modified to also contain 1 mol% biotinylated dipalmitoyl-phosphatidylethanolamine (Avanti Polar Lipids, Inc., Alabaster, AL). The mixture of surfactant components, perfluorocarbon, and water was emulsified at 20,000 PSI for 4 min in an ice bath (S110 Microfluidics emulsifier, Microfluidics, Newton, MA). Particle sizes were determined at 25° C with a laser light scattering submicron particle sizer (Malvern Instruments, Malvern, Worcestershire, UK). Particle size ranged from 171–232 nm between the different formulations (Table 1). The number of gadolinium molecules per nanoparticle was calculated based upon the nominal particle radius according to the following equation:

$$\frac{\#Gd}{NP}=\frac{[Gd]}{[NP]}=[Gd]·\frac{\mathit{PFC}}{4/3·\pi ·r^3},$$

where [Gd] and [NP] are the concentration of gadolinium and nanoparticles, respectively, in the final emulsion volume. The latter is calculated based upon the volume of PFC used in to make the emulsion and the volume that each nanoparticle occupies of that volume.

Table 1.

Selected properties of the nanoparticles used in the study

Amount of Gd- DTPA-BOA in lipid monolayer	[Gd] in total volume (mM)	[Gd] in PFC volume (mM)	Nominal particle diameter (nm)	Estimated # Gd per nanoparticle
10 mol%	1.44	6.56	232	2.6x104
20 mol%	4.50	20.46	191	4.5x104
30 mol%	6.52	29.63	171	4.7x104

Precise measurement of the gadolinium content of the different nanoparticle constructs was analyzed by standard neutron activation techniques conducted at the Research Reactor facility at the University of Missouri (MURR) (18). After lyophilization, the mass of gadolinium was calculated from the beta decay of ¹⁶¹Gd produced through neutron capture on ¹⁶⁰Gd. Individual samples and standards were irradiated in a thermal neutron flux of about 5 x 10¹³ n·cm⁻²·s⁻¹ for 7 seconds, allowed to decay for 30 seconds, and counted on a high-resolution gamma-ray spectrometer for 300 seconds. Table 1 lists the relevant characteristics of these different nanoparticle formulations.

MR imaging and spectroscopy

The majority of imaging and spectroscopy was performed on a 1.5 T clinical MR system (NT Intera, Philips Medical Systems, Andover, MA) with peak gradients of 30 mT/m (150 mT/m/ms) which was outfitted with a secondary radio frequency (RF) transmit/receive system tuned for fluorine imaging at 61 MHz. An inductively coupled, 10 cm long by 5 cm diameter, single turn solenoid coil was built for both transmission and reception of the fluorine MR signal. High-voltage variable capacitors made of Teflon for MR compatibility (Johanson, Boonton, NJ, and Voltronics, Denville, NJ) were used for tuning to different loads. The measurements at higher fields were completed on 4.7 T and 11.7 T Varian INOVA small bore imaging systems using custom built T/R volume coils.

Fluorine relaxation rate (R₁ and R₂) measurements were performed using inversion recovery and variable echo time spin-echo spectroscopy with a minimum of 10 data points per measurement and at least three consecutive measurements (coefficient of variation was typically 0.5–3%). The actual inversion and echo times used were adjusted based upon the expected relaxation rate. The data was analyzed using NUTS data processing software (Acorn NMR, Inc., Livermore, CA). Fluorine imaging of gadolinium (30 mol%) and non-gadolinium containing nanoparticle samples at 1.5 T was performed with the following common scan parameters: 2D single slice acquisitions with 256x256 mm field of view and 1 mm x 1 mm x 5 mm resolution. The other parameters, which varied amongst the different sequences, were: TR = 5.07 ms, TE = 2 ms, 45 NSA, 58 second duration for the T₁-weighted gradient echo scan; TR = 200 ms, TE = 2 ms, 1 NSA, 52 second duration for the traditional gradient echo scan; and TR = 5.17 ms, TE = 2.2 ms, 40 NSA, 53 second duration for the balanced steady-state free precession sequence. Contrast-to- noise ratios were calculated as the absolute value of the difference between the signals from two regions of interest divided by the standard deviation of the signal in a region of air. All measurements were performed at room temperature and ambient oxygenation levels unless otherwise specified.

The effect of aqueous solution versus particle-complexed gadolinium

The ¹⁹F longitudinal and transverse relaxation times of the nanoparticles formulated with different amounts of gadolinium in their lipid layer (10, 20, or 30mol%) were measured using an undiluted emulsion sample (20% v/v PFC). To compare the effect of a water soluble gadolinium agent on the ¹⁹F relaxation, 0.5 mL of a nanoparticle emulsion containing no gadolinium was mixed with varying amounts (5, 50, or 500 μL) of a standard clinically approved MR contrast agent (gadodiamide, Omniscan™, GE Healthcare). The total sample volume was increased to 1 mL by the addition of saline solution, bringing the final gadolinium concentration to 2.5, 25, or 250 mM in the total sample volume or 0.55, 5.5, or 55 mM in the “PFC volume”, the latter of which was used for calculating an effective relaxivity for comparison to the effect of gadolinium when contained on the nanoparticle itself.

Fibrin clot formation and targeting

Fresh-frozen human plasma anticoagulated with sodium citrate was used to form fibrin clots by combining plasma, 500 mM calcium chloride, and 1 U/μL thrombin (Sigma) in a volume ratio of 375:22:3 (17). These reagents were quickly mixed and added to an approximately 1.5 cm long template containing a string down the center while the clot solidified. It was then teased from the template and immersed in saline in a 5 mL snap-cap tube. Fibrin-targeting of the nanoparticles was induced by serially incubating the clots with saline containing 125 rg biotinylated anti-fibrin monoclonal antibody (19) (National Institute for Biological Standards and Control, NIB 1H10) overnight at 4 °C, followed by 125 rg avidin for 1 hour at 37 °C, and then 100 rL of the biotinylated nanoparticles containing 30 mol% gadolinium for 1 hour at 37°C to complete the binding. All samples were rinsed three times with sterile saline after each incubation step to remove any unbound reagents.

Transmetallation reactions

Two types of transmetallation reactions were performed: steady-state and kinetic. The steady-state reactions utilized gadolinium-containing nanoparticles in solution (675 μL), to which were added various amounts of 250 mM ZnCl solution to obtain ratios of Zn:Gd of 0, 0.1, 0.5, 0.75, 1, or 10. A stock solution of 10x phosphate buffer containing 0.26 M KH₂PO₄ and 0.41 M Na₂PO₄ and water were added to obtain a final volume of 920 μL for all samples. Previous studies indicate that steady state is reached at ~50 hrs with a 1:1 ratio of Zn:Gd at 37°C (20), but room temperature was chosen for this experiment to avoid potential bacterial growth and subsequent nanoparticle degradation in the samples. Thus, these reactions were allowed to react at room temperature and were measured at both three and nine days, with comparable results. To verify that these transmetallation reactions do not result in nanoparticle degradation, the particle sizes were measured and compared (n=4 each) between the native gadolinium emulsion, a “cracked” sample of the same emulsion (1:1 mixture of gadolinium emulsion and 2-propanol, which destroys the lipid monolayer), the 1:1 Zn:Gd transmetallation sample, and nanoparticles incubated in serum.

Kinetic measurements of the change in ¹⁹F relaxation during transmetallation were acquired by targeting nanoparticles formulated with 30 mol% gadolinium to the surface of a human fibrin clot formed on a string. Fluid containing the transmetallation reagents (26 mM KH₂PO₄, 41 mM Na₂PO₄, and ZnCl in 5 mL of deionized water) was added to the clots for a specified length of time and then removed to quench the reaction long enough to perform ¹⁹F relaxation rate measurements at 1.5 T. After the measurement was made, the process was resumed by re-adding the transmetallation fluid. The rate of the reaction was varied by altering the concentration of zinc in the surrounding fluid (either 40 mM or 40 μM, n=3 each).

Results

As an initial demonstration, the ¹⁹F relaxation rates (R₁ and R₂) of nanoparticles formulated with or without 30 mol% Gd-DTPA-BOA in the lipid layer were compared at three different field strengths (Figure 1). The effect of the gadolinium was quite pronounced, resulting in a substantial influence on both the R₁ and R₂ relaxation rates. The effectiveness of gadolinium as a relaxation agent for ¹⁹F nuclei as a function of increasing field strength is consistent with previously published data showing the same trend for ¹H R₁ relaxation using a similar nanoparticle construct (21). Importantly, in view of the desirability of a pronounced increase in R₁ combined with a relatively limited R₂ effect, the strongest benefit of including the gadolinium for ¹⁹F imaging of perfluorocarbon nanoparticles is observed at 1.5 T, the most prevalent clinical imaging field strength.

Figure 1.

Incorporation of gadolinium (30 mol%) into the nanoparticle outer lipid membrane induces a dramatic and field-dependent effect on core ¹⁹F relaxation rates (a: R₁, b: R₂).

To further delineate the relationship of concentration and proximity of the gadolinium to its effect on the fluorine relaxation rates, a series of gadolinium dilution experiments was conducted. Dilution of nanoparticle emulsions formulated to contain the same amount of gadolinium on their surface with additional water (no dilution, 1:10, or 1:100) resulted in no change in either the fluorine R₁ or R₂ at 1.5 T (data not shown), indicating that the effect on the fluorine nuclei is not mediated by the total gadolinium concentration but rather by the local concentration on the particle surface. We then created nanoparticles with different concentrations of gadolinium on the particle surface by altering the amount added during formulation (see Table 1), which serves to change the local concentration of gadolinium on the particle surface. This results in a linear change in the fluorine relaxation rates (Figure 2), which allows calculation of effective relaxivity values for both gadolinium on the nanoparticle surface and gadolinium in solution, respectively: 1.02x10⁻¹ vs. 1.29x10⁻⁴ (s·mM)⁻¹ for r₁ and 4.15x10⁻¹ vs. 1.11x10⁻¹ (s·mM)⁻¹ for r₂.

Figure 2.

Gadolinium must be in very close proximity to the perfluorocarbon core of the nanoparticle to affect ¹⁹F relaxation at 1.5 T. ¹⁹F relaxation rates (R₁ and R₂) as a function of gadolinium concentration (millimoles Gd per liter of perfluorocarbon) for nanoparticles either formulated with different amounts of gadolinium on their surface (black squares) or mixed with an aqueous chelate of gadolinium (gray diamonds). A linear regression line through these points gives an r²>0.96 for R₁ and r²>0.97 for R₂ and was used to calculate an effective relaxivity on ¹⁹F nuclei.

For comparison, we also mixed an aqueous gadolinium chelate (Omniscan™, gadodiamide) with a non-gadolinium containing nanoparticle emulsion after formulation and measured the effect on the fluorine relaxation rates. While gadodiamide and Gd-DTPA-BOA clearly differ chemically and exhibit different ¹H relaxivities, this comparison highlights critical differences in relaxation behavior under these conditions. In both cases, gadolinium concentrations were calculated as amount of gadolinium compared to a “perfluorocarbon volume” (not total volume) so that a relevant comparison can be made between the two experiments. As shown in Figure 2b, both the r₁ and r₂ relaxivities are increased by including the gadolinium on the particle surface, although the difference in r₁ is much more striking. The aqueous chelate did not substantially alter the ¹⁹F r₁ relaxation and demonstrated a lesser effect on ¹⁹F r₂ than did gadolinium on the particle surface.

The alteration in ¹⁹F relaxation time translates into considerable contrast enhancement on ¹⁹F MR images, as demonstrated in Figure 3. Samples of nanoparticles formulated with or without 30 mol% gadolinium on their surface were imaged at 1.5 T using three variations of a 2D gradient echo pulse sequence (short TR with crusher gradients, longer TR with crusher gradients, or balanced steady-state free precession) over a range of flip angles to characterize the ability to differentiate these nanoparticles based upon the influence of gadolinium. To produce a reasonable comparison between these different scan sequences, resolution (1x1x5mm) and total scan duration (~52 seconds) were kept approximately constant, necessitating varied numbers of signal averages. By defining a contrast to noise ratio (CNR) of 5 as the minimum amount of detectable contrast, it is apparent that these two different nanoparticle formulations were visibly differentiable at almost all flip angles with all three sequences (Figure 3a). The individual images shown for each sequence (Figure 3b) demonstrate that the contrast can be modulated depending on the sequence chosen. For both T₁-weighted gradient echo sequences with crusher gradients (T1wFFE), the sample containing gadolinium produced the brightest signal. In the case of the steady-state free precession sequence, where the signal generated is weighted by T₁/T₂ instead of simply T₁ (22), the contrast is reversed so that the nanoparticles without gadolinium give the higher signal.

Figure 3.

Nanoparticles containing gadolinium generate substantial contrast effects in ¹⁹F images acquired at 1.5 T. a) Contrast to noise ratio generated when simultaneously imaging nanoparticles with and without gadolinium using three different gradient echo sequences over a range of flip angles. b) Image with maximal contrast to noise for each of the three imaging sequences.

As a practical validation for modulation of the ¹⁹F signal, a series of transmetallation experiments were performed to deplete the gadolinium from the particle surface. Transmetallation of gadolinium with zinc, a diamagnetic metal ion that has a minimal effect on relaxation rates, has been used in the literature to study the stability of a variety of chelating agents (20,23). The DTPA ligand is known to be susceptible to this process, providing a mechanism whereby the gadolinium can be controllably removed from the surface of the nanoparticles, and the effect on the fluorine nuclei can be measured. By including an excess of phosphates in solution, the displaced gadolinium is bound and forms an insoluble precipitate, thereby driving the reaction to completion.

In the first set of reactions, the amount of zinc added to nanoparticles in solution was varied, and the reaction was allowed to proceed for 9 days at room temperature to ensure equilibrium conditions. When the amount of zinc added compared to gadolinium is kept below 1:1 and the reaction is allowed to proceed to completion, the essential effect is partial replacement of gadolinium on the particle surface with zinc, the degree to which is dictated by the ratio (i.e., a ratio of 0.5 would indicate that half of the gadolinium has been replaced). Measurement of the fluorine relaxation rate showed a linear dependence on the ratio of zinc to gadolinium added to the reaction when it was less than or equal to 1 (Figure 4a), presenting further evidence that the relationship between local surface gadolinium concentration and fluorine relaxation is predictable and quantifiable. When zinc is added in vast excess (ratio Zn:Gd = 10), almost all of the gadolinium is replaced, and the ¹⁹F R₁ value becomes very similar to that of control nanoparticles formulated without gadolinium, confirming the predicted endpoint of the transformation. The results obtained by varying the gadolinium on the nanoparticle surface by changing the amount used in the formulation compared with those obtained from the transmetallation reaction experiments indicate that the change in R₁ as a function of molar percentage of gadolinium present is very similar (slopes of 0.0278 ± 0.011 vs. 0.0192 ± 0.003, respectively). Because the sizes of the native, 1:1 Zn:Gd transmetallation-reacted, and serum-incubated emulsions are similar to each other (163.2 ± 3.3 nm, 196.4 ± 9.4 nm, and 195.4 ± 23.4, respectively, p>0.6) and statistically different from a fully degraded (or “cracked”) nanoparticle sample (493.4 ± 221.6 nm, p<0.01 from all other samples), the measured effect on ¹⁹F relaxation rates is most likely not attributable to nanoparticle degradation, but instead to actual gadolinium removal.

Figure 4.

¹⁹F relaxation times at 1.5 T directly reflect the concentration of gadolinium on the nanoparticle surface. a) ¹⁹F relaxation rate (R₁) measurements on nanoparticles in solution after the addition of different amounts of zinc and an excess of phosphate buffer after steady-state has been reached. The line of best fit to the first four data points is R₁ = −1.7·ratio+ 2.87 (s⁻¹) with an r²>0.97. b) Measurements of transmetallation kinetics using ¹⁹F relaxation rate (n=3, mean ± SD) of nanoparticles targeted to human fibrin clots in vitro where the reaction rate was varied by changing the concentration of zinc (40 μM vs. 40 mM). Dotted lines on both plots represent the R₁ relaxation rate measured on control samples that did not contain gadolinium.

To show that this particle can be targeted to human molecular epitopes while still retaining this property, kinetic measurements of the change in ¹⁹F relaxation were acquired for nanoparticles formulated with 30 mol% gadolinium. After targeting the nanoparticles to the surface of a human fibrin clot formed on a string with the use of a biotinylated anti-fibrin antibody (17), the ¹⁹F relaxation rate of the nanoparticles remained the same as when in solution (data not shown). Fluid containing the transmetallation reagents was added to the clots for a specified length of time and then removed to quench the reaction long enough to perform ¹⁹F relaxation rate measurements at 1.5 T. After the measurement was made, the process was resumed by adding the transmetallation fluid. The rate of the reaction was varied by altering the concentration of zinc in the surrounding fluid (either 40 mM or 40 μM). As shown in Figure 4b, these two different conditions were easily differentiated in targeted fibrin clots using the fluorine relaxation rate as the indicator.

One potential confounding factor for use of this proposed tailored relaxation mechanism is the well-characterized ability of perfluorocarbons to dissolve a large amount of oxygen (24). Because oxygen can readily diffuse through lipid membranes and is paramagnetic, it induces a change in the magnetic relaxation properties of perfluorocarbons, a characteristic that has been used by others to non-invasively measure oxygenation levels in vivo (25–29). Inclusion of gadolinium on the nanoparticle surface is hypothesized to abrogate the oxygen sensing ability of the perfluorocarbons. To test this, the ¹⁹F relaxation rate of the nanoparticles was measured at 1.5 T after bubbling the samples with 0, 21, 50, or 100% oxygen in nitrogen for 30 minutes (Fig. 5). The slope of the line (or “relaxivity” of oxygen on perfluoro-crown-ether) is virtually identical in the case of nanoparticles formulated with or without 30 mol% gadolinium, indicating that these two paramagnetic agents induce an additive effect on ¹⁹F relaxation at these concentrations. However, within the range of relevant oxygen concentrations in vivo (50–100 mmHg), the magnitude of fluorine relaxation change induced by gadolinium removal (up to 320% change) far outweighs that of altered oxygen levels (13% change between 0–150 mmHg). The ability to use a given ¹⁹F agent as an oxygen sensor depends on both a predictable relationship between R₁ and pO₂, as well as a wide dynamic range of measurable R₁ values, as indicated by the slope to intercept ratio (30). Because the gadolinium-containing nanoparticles have a ¹⁹F relaxation rate that is approximately 10% that of the particles without gadolinium at 1.5 T, the sensitivity of this agent to serve as an oxygen sensor and the potential confounding effects of oxygen changes on detecting gadolinium removal is greatly reduced.

Figure 5.

Gadolinium and oxygen are synergistic ¹⁹F relaxation agents. Effect of gadolinium on the oxygenation-dependent relaxation rate (R₁) of PFC nanoparticles at 1.5 T. The linear regression lines show that the “relaxivities” are statistically similar for both nanoparticles (0.0036 s⁻¹mmHg⁻¹ with Gd and 0.0039 s⁻¹mmHg⁻¹ without Gd, r²>0.99).

Discussion

The implications of these results are manifold. This is the first demonstration of relaxation enhancement of bulk perfluorocarbons obtained through the use of gadolinium for ¹⁹F MR imaging. The hydrophobic nature of perfluorocarbons makes it extremely difficult to force an interaction of the gadolinium lanthanide with the ¹⁹F nuclei, except through a lipid encapsulation approach. Using a similar lipid-gadolinium interaction, Ellena et al. have demonstrated a substantial relaxation change of lipid bilayer-localized perfluorooctylbromide (PFOB, a linear perfluorocarbon) in response to Gd⁺³ bound to phospholipid headgroups at 11.7 T. However, they did not find any effect on the “micellar PFOB” fraction (31), which agrees with our findings at 11.7 T with perfluoro-crown ether containing nanoparticles. While PFOB and perfluoro-crown ether are chemically distinct and may be affected differently by gadolinium, combining these results may provide some insight into the physical mechanism underlying this effect. We hypothesize that those ¹⁹F nuclei in very close proximity to the gadolinium on the lipid layer are primarily affected and then serve to induce relaxation in the bulk perfluorocarbon (lattice) inside the nanoparticle through diffusional and rotational interactions in the fluid core. The most efficient R₁ relaxation will occur when the frequency of this molecular motion is closely matched to the Larmor frequency of the nucleus (32), which our data indicate occurs at lower field strengths for this agent.

As with standard ¹H imaging, the main benefit of this effect is the ability to generate a higher signal from the ¹⁹F nuclei in a shorter imaging time with traditional imaging sequences. For quantitative spectroscopy, this feature might also serve to improve sensitivity by allowing reduced repetition times and therefore, more signal averages in a given amount of available scan time. Furthermore, the precision of this nanoparticle design serves to reduce the total amount of gadolinium needed to induce a measurable effect, which militates against the dominance of r₂ relaxivity (reduction in signal) that is demonstrated when the gadolinium is mixed in solution at high concentrations (Figure 2). It has also been previously demonstrated in a variety of animal models that these gadolinium containing perfluorocarbon nanoparticles can be utilized for sensitive ¹H imaging of molecular targets due to their long circulation half-life and stability in vivo (7,8,33), which presents the opportunity to use combined ¹⁹F and ¹H imaging to simultaneously detect and/or quantify nanoparticles targeted to a particular epitope.

These experiments also demonstrate that the fluorine relaxation rate is a very sensitive and quantitative indicator of the presence and amount of gadolinium on the nanoparticle surface. This proximity requirement adds a high degree of specificity and provides the basis for novel MRI applications. If, for example, a short cleavable linker were included between the chelated gadolinium and the lipid anchor, it might be possible to monitor gadolinium-chelate release in vitro or in vivo using these nanoparticles. The aim would be to confer release specificity with selected enzymatic substrates, as has been reported for other “smart” contrast agents (13,34), yet with a signal readout that is unique and quantifiable, and which does not have a confounding background. Alternatively, it should be possible to formulate perfluorocarbon nanoparticles with variable but known gadolinium concentrations to achieve sets of particles with T₁ relaxation times that can be discriminated with traditional T₁ mapping methods. This approach might enable simultaneous detection of the multiple T₁ modulated particles targeted to different molecular epitopes that can be registered against essentially no confounding background. These possibilities represent aims for the future following the initial description of a novel gadolinium-fluorine contrast effect that creates tailored perfluorocarbon nanoparticles which exhibit unique spectral signatures for MRI/MRS.

The specific gadolinium chelate used in this study (DTPA) may permit some dissociation of Gd⁺³ ions in vivo, which may contribute in part to untoward side effects such as nephrogenic fibrosing dermopathy (35). Transmetallation reactions such as those used in this study have been used as a way to test the stability of gadolinium chelates to quantify their potential for releasing free Gd⁺³ in vivo, and the results in this study are consistent with those found in the literature (20). In anticipation of eventual clinical application, an alternative chelate with stronger binding affinity would be more useful and easily substituted, such as a macrocyclic compound (DOTA analogue). The actual application would not be transmetallation per se, but rather something like detection of tissue enzyme activity that would cleave the Gd-chelate complex at some specific linker site from the PFC particle, resulting in a change in ¹⁹F signal (T₁). The stable Gd-chelate complex, used in a dose twenty times lower than that typically administered in an MRI exam (4.6x10⁻³ mmol Gd/kg body weight vs. 0.1 mmol/kg, respectively),.then would dissociate from the nanoparticle and be cleared rapidly by the kidney.