More specific biochemical information can be obtained from magnetic resonance spectroscopy (MRS) or chemical shift imaging (CSI) of small molecules which reports directly on the imaging agent and its concentration. However this technique suffers from the inherent low sensitivity of NMR and millimolar (mM) concentrations of agent are required. Higher field strength scanners provide improved sensitivity and this has led to an increased interest in chemical shift agents. Numerous ¹⁹F-MRI agents have been proposed on the basis of the relatively high sensitivity of ¹⁹Fand the lack of endogenous ¹⁹F signal.(1) More sensitive still is to use the ¹H nucleus, but here a challenge is the narrow ¹H chemical shift window with its high background of water, fat, and endogenous metabolites.

Some notable ¹H CSI probes include YbDOTMA (Figure 1) or its thulium analogue,(2,3) hexamethyldisiloxane (HMDS), and (±)imidazol-1-yl)succinic acid (ISUCA) that have been proposed as responsive agents for temperature, partial oxygen pressure (pO₂) or pH, respectively.(2,4,5) YbDOTMA is a highly soluble, inert complex that has 12 equivalent methyl protons for higher sensitivity. The paramagnetic Yb(III) shifts these protons outside the typical diamagnetic proton chemical shift range, with its high endogenous background signal, without causing significant line broadening. However on its own, YbDOTMA has no specific biological targeting and is rapidly excreted. HMDS has 18 equivalent protons at 0 ppm. Injection of neat, water insoluble HMDS directly into muscle or tumours in rats enabled quantitative studies of tissue oxygenation, but the invasive route of administration and the very slow clearance of the agent limit the applicability of this agent to animal studies.(5) The water soluble ISUCA has a very short blood half-life and contains only one proton that is shifted from the natural background. Therefore, a continuous infusion is required to maintain a concentration sufficiently high for pH determination.(4) In general, the sensitivity of these contrast agents is low and concentrations in the millimolar range are required for imaging.

Here, we describe approaches to increase the sensitivity of ¹H-CSI probes by encapsulation of the chemical shift reporters inside liposomal nano-carriers. We demonstrate the potential for multiplexed imaging using three different probes with chemical shift signatures that are outside the biological proton shift window. We further show that incorporating a relaxation agent within the nano-carrier results in improved detection sensitivity.

The trimethylsilyl (TMS) group containing nine magnetically equivalent protons with a chemical shift of 0 ppm is used in NMR experiments as a standard since it is shifted from most common resonances. Therefore, it is not surprising that compounds containing this moiety were already used as sensors in MRS.(5,6) We reasoned that sodium 3-(trimethylsilyl)-1-propanesulfonate (DSS, Figure 1) would offer high water solubility for nano-encapsulation. Furthermore, small hydrophilic anions are often well tolerated and we performed a cell toxicity assay on human umbilical vein endothelial cells (HUVEC, see supporting information) that confirmed the low toxicity of this compound (EC₅₀ = 5 mM).

Protons exhibiting chemical shifts downfield of natural background are usually acidic and exchange with bulk water. This leads to line broadening and loss of signal. However, azolium salts bearing aliphatic substituents on the heteroatom(s) undergo slow proton exchange of the C₂-proton in protic solvents and, therefore, can be considered as CSI probes as exemplified by ISUCA.(7) 3-(2-Hydroxyethyl)thiazolium bromide (HETB, Figure 1) was chosen as a model compound for this family of probes because the C₂-proton of thiazolium salts have larger chemical shifts than their imidazolium analogues (9.8 ppm in this case).

Paramagnetic lanthanide ions shift protons in their proximity as a function of distance and orientation of the lanthanide-proton vector. Here, we investigate the chemically inert ytterbium complex of the symmetric ligand 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (YbDOTA, Figure 1) since it has a resonance at 19.9 ppm corresponding to four magnetically equivalent protons.(8)

To demonstrate the possibility for multiplex detection of these probes, phantoms were placed in an agar gel and imaged on a 9.4T small bore animal scanner with a STEAM sequence and VAPOR water suppression.(9,10) The five phantoms contained equimolar proton concentrations of a) DSS, b) HETB, c) HETB, DSS, YbDOTA, d) blank, and e) YbDOTA. To compensate for the bulk magnetic susceptibility effect of YbDOTA on the chemical shifts of DSS and HETB, equimolar YbCl₃ was added to the samples that did not contain YbDOTA. Color overlays represent the chemical shift images generated at the characteristic resonance frequencies of each of the probes and show that simultaneous detection of all investigated compounds is possible.

On their own, small molecules like YbDOTA have limited applicability for in vivo CSI because of their non-specific distribution and their rapid blood clearance. However, it is known that large payloads of small molecules can be encapsulated inside liposomes or emulsions.(11) Moreover stealth liposome formulations have very long circulation times resulting in persistent MR signal.(12) Nanoparticle platforms have been widely used in molecular imaging to enable target-specific imaging of inflammation, angiogenesis, apoptosis, and atherosclerosis, demonstrating the broad applicability of such systems.(13–16)

We encapsulated 135 mM DSS (isotonic) inside DSPC/cholesterol/PEG2000-PE liposomes resulting in a 1.2 M concentration of equivalent protons at 0 ppm in the liposome core. The vesicles showed a narrow size distribution (polydispersity index<0.11) and were stable over at least two weeks if stored at 4 °C. Figure 3A shows a ¹H NMR spectrum of a 2.05 nM liposome suspension (6.3 μM DSS) in citrated blood plasma. We recorded spectra and calculated signal to noise ratios (SNR) for the TMS resonance for a series of dilutions (Figure 3C). The spectra show that the TMS resonance of DSS and the plasma resonances are well separated allowing for highly sensitive detection of liposomes; at the lowest dilution, a 57 pM liposome concentration was still readily apparent with SNR=7.1, Figure 3B. In vivo CSI will be less sensitive than ex vivo spectroscopy, but assuming a detection limit of 1 mM in protons for clinical CSI at 3T, this would represent a 40 nM detection limit for the DSS containing liposomes. This is well within the range of many vascular molecular imaging targets such as activated platelets, fibrin, the α_vβ₃ integrin, and the selectins. (17)

An additional problem in CSI experiments is the relatively long longitudinal relaxation times (T₁) of nuclei under investigation. Such long T₁ values require long repetition times (TR) to allow for recovery of the magnetization, and this in turn increases acquisition times. Reduction of T₁ would reduce scan times or allow more scans to be performed in the same time period leading to even higher sensitivity.(1,18,19) A benefit of the liposome approach is that relaxation agents can also be encapsulated along with the chemical shift reporter. To demonstrate this, a dilute solution of the gadolinium(III) complex of 10-(2-hydroxypropyl)-1,4,7,10-tetraazacyclododecane-1,4,7-triacetic acid (Gd-HPDO3A), a clinically approved MRI contrast agent, was co-encapsulated into liposomes together with DSS. CSI imaging with varying TR was performed on two phantoms containing liposome-encapsulated DSS with and without 2 mM Gd-HPDO3A inside the liposomes. As shown in Figure 4A, there is a significant signal loss for diamagnetic liposomes when TR is shortened whereas the relaxation agent counteracts this loss. This short TR allows for >3-fold reduction in acquisition time.

Additionally, the ability of Gd(III) to shorten T₁ is a function of the distance between the Gd(III) ion and the nucleus. Since Gd is entrapped within liposomes in the proximity of DSS, only the DSS protons are affected; endogenous nuclei will not experience significant changes in relaxation and thus loose signal faster with shortened TR than the imaging probe. This is illustrated in Figure 4B where the SNR of DSS co-encapsulated with Gd-HPDO3A is compared with the SNR of the most intense peak of plasma in the same voxel. The plasma signal relaxes slower than the DSS signal due to the local separation of Gd from the plasma components. Thus, the addition of small amounts of relaxation agent not only allows shorter imaging times, it also enables more selective detection of the imaging agent at a favorably shorter TR by suppressing the background signal.

Supplementary Material