Lanthanide(III) Complexes of Rhodamine–DO3A Conjugates as Agents for Dual-Modal Imaging

Two novel dual-modal MRI/optical probes based on a rhodamine–DO3A conjugate have been prepared. The bis(aqua)gadolinium(III) complex Gd.L1 and mono(aqua)gadolinium(III) complex Gd.L2 behave as dual-modal imaging probes (r1 = 8.5 and 3.8 mM–1 s–1 for Gd.L1 and Gd.L2, respectively; λex = 560 nm and λem = 580 nm for both complexes). The rhodamine fragment is pH-sensitive, and upon lowering of the pH, an increase in fluorescence intensity is observed as the spirolactam ring opens to give the highly fluorescent form of the molecule. The ligands are bimodal when coordinated to Tb(III) ions, inducing fluorescence from both the lanthanide center and the rhodamine fluorophore, on two independent time frames. Confocal imaging experiments were carried out to establish the localization of Gd.L2 in HEK293 cells and primary mouse islet cells (∼70% insulin-containing β cells). Colocalization with MitoTracker Green demonstrated Gd.L2’s ability to distinguish between tumor and healthy cells, with compartmentalization believed to be in the mitochondria. Gd.L2 was also evaluated as an MRI probe for imaging of tumors in BALB/c nude mice bearing M21 xenografts. A 36.5% decrease in T1 within the tumor was observed 30 min post injection, showing that Gd.L2 is preferentially up taken in the tumor. Gd.L2 is the first small-molecule MR/fluorescent dual-modal imaging agent to display an off–on pH switch upon its preferential uptake within the more acidic microenvironment of tumor cells.


■ INTRODUCTION
Molecular imaging is a rapidly growing area of chemistry that aims to visualize cellular function and structure in a noninvasive manner with the in vivo use of specially designed imaging agents. The combination of magnetic resonance imaging (MRI) and optical imaging offers synergistic advantages over either modality alone. MRI has high spatial resolution and good softtissue contrast 1,2 while optical imaging displays high sensitivity and can give information on the local chemical environment. 3 Lanthanide-based optical probes have superior fluorescence properties in comparison to organic fluorophores, with luminescence lifetimes stretching to the millisecond region, and time-gated techniques can be employed in order to eliminate interfering background autofluorescence. 4−6 With the constant advances in cancer treatment, it is becoming increasingly important to detect the early signs of the disease and to establish the efficacy of the concurrent treatment in a noninvasive manner. This can be achieved by the in vivo use of a dual-modal MRI/fluorescent probe that responds to a characteristic trait of tumor cells. 7−9 It is known that the mitochondrial potential in cancer cells is greater than that of healthy cells, 10 and the design of a probe that can accumulate in these energized mitochondria will lead to a tumor-targeting agent. The combination of MRI and fluorescence will allow for improved diagnostic accuracy by tumor localization via MRI imaging and guided surgery via fluorescence imaging.
MRI is a noninvasive diagnostic technique that relies on the enhancement of local water proton relaxation. 11,12 Different tissues have different relaxation times, resulting in endogenous magnetic resonance contrast. Exogenous agents can be used to enhance this contrast by selectively shortening the T 1 (longitudinal) or T 2 (transverse) relaxation time; these agents include Gd(III) chelates, manganese complexes, and superparamagnetic nanoparticles. 13,14 The most successful and commonly used of these exogenous agents are complexes containing the Gd(III) ion because of its high magnetic moment (7.9μ B ) given by its seven unpaired electrons and its totally symmetric electronic ground state ( 8 S 7/2 ). 15 According to Solomon−Bloembergen−Morgan theory, the relaxivity of Gd(III) complexes is governed by a number of factors, including the number of bound water molecules (q), the rotational tumbling time (rotational correlation time) of the complex (τ R ), and the mean residence lifetime of Gd(III)bound water molecules (τ M ). 2 An effective way to increase the relaxivity of a Gd(III) complex is to increase the number of bound water molecules. Herein we report two novel probes possessing different hydration states. We envisioned that the relaxivity could be improved by decreasing the number of coordination sites on the Gd(III) chelate from eight to seven, allowing for a change from one bound water molecule to two. It is reported however that heptadentate ligands, when bound to Gd(III), have unfavorable properties for use as MRI probes in terms of thermodynamic and kinetic stability because of the coordinatively unsaturated nature of the complex. 16 Eu(III) and Tb(III) chelates based on a DO3A core have received a lot of interest as luminescent lanthanide systems. 17−20 These lanthanides are characterized by their long-lived luminescent excited states (ms time scale) and linelike emission spectra, and their complexes have been shown to be useful in biomedicine as luminescent probes. 21 Lanthanide metals or ions however, have low absorption coefficients due to Laporte-forbidden f−f transitions. 22 This can be overcome by attaching an aromatic chromophore to the macrocyclic core to act as an "antenna" by transferring absorbed excitation energy from its triplet excited state to the excited state of the coordinated lanthanide ion. 23 Rhodamine derivatives have received significant attention as fluorescent labeling agents because of their long absorption and emission wavelengths, large absorption coefficients, and high quantum yields. 24 They have been used in a wide range of applications, having great success as chemosensors (both in vitro and in vivo), 25,26 protein labels, 27 and dual-modal imaging agents. 28,29 Rhodamine derivatives have had particular success as mitochondria targets, 30−32 as their delocalized organic cationic forms tend to accumulate in the mitochondria of tumor cells as a result of the increased negative mitochondrial potential. 33−35 The spiro ring-opening mechanism of rhodamine has also led to its extensive use as a chemosensor for metal detection as well as a pH-sensitive probe. 36−38 Under acidic conditions, the rhodamine moiety exists in its highly fluorescent pink-colored open form. Upon addition of OH − , the structure exists in its spirocyclic form, which is colorless and nonfluorescent.
In recent years there have been an increasing number of reports of multimodal imaging probes wherein two or more imaging modalities are combined. 39,40 The combination of modalities leads to probes that give enhanced visualization and improved reliability of data by synergistically combining imaging techniques to overcome their inherent disadvantages. MRI gives high anatomical resolution and deep tissue penetration but lacks sensitivity. Optical imaging, however, boasts high sensitivity but has limited tissue penetration. As discussed earlier, the combination of the two techniques yields a probe that is able to provide a more complete picture of the biological area of interest. There are few reports of dual-modal MRI/optical imaging agents in the literature that are not of nanoparticle nature. These include Gou et al.'s description of the binding of a cyanine−DTPA conjugate to BSA to afford relaxivity rate constants of 15 mM −1 s −1 . 41 The aliphatic nature of the chelate is unfavorable, however, as release of Gd(III) ions into the body from these types of complexes has been related to nephrogenic systemic fibrosis (NSF). 42 In 1998, it was reported that a rhodamine−Gd(III) complex conjugate showed no MRI contrast enhancement when introduced to Xenopus laevis embryos. 43 The same complex conjugate, Gd(Rhoda-DOTA), was studied alongside a series of hydrophobic fluorescent dye− Gd(III) complex conjugates in 2011 by Yamane et al., 44 who found that incorporation of the hydrophobic fluorescent moiety  improves cell permeability compared to Gd(DOTA). They reported that Gd(Rhoda-DOTA) showed a large increase in r 1 relaxivity when bound to albumin and efficient cell permeation when incubated with HeLa cells. Again, however, no significant change in the MR signal of HeLa cells was observed in the presence of the complex, even though its fluorescence was observed by microscopy. Similarly, a rhodamine moiety has previously been conjugated to DO3A and consequently labeled with 64 Cu to arrive at a dual-modal PET/optical probe. 29 It was found that 64 Cu(DOTA-LRB) had a surprisingly high tumor uptake compared with other 64 Cu-labeled organic cations. The probe was able to selectively localize to the tumor mitochondria with long tumor retention times. 45 The probe and its derivatives have been the subjects of extensive in vitro and in vivo studies showing that there is promise for a Gd(III) MRI/ optical analogue to be as successful in vivo.
In this paper, we report two novel methods for conjugating fluorescent rhodamine derivatives to DO3A. We evaluated the two rhodamine−DO3A Gd(III) derivatives as multimodal MRI/optical imaging agents by measuring their relaxivity and fluorescence properties as well as their in vitro tumor cellular localization. The Tb(III) analogues were synthesized to assess their dual-luminescence properties, and it was found that the organic and metal-based luminescence can be separated on different time scales.
■ RESULTS AND DISCUSSION Synthesis. Two new synthetic routes were developed to obtain L1 and L2 (outlined in Schemes 1 and 2, respectively), both commencing from commercially available rhodamine B and involving five straightforward steps. Reaction of the carboxylic acid moiety on rhodamine B with bromoethylamine resulted in amide bond formation, giving product 2. The ligand precursor 3 was prepared by N-alkylation of the tri-tert-butyl ester derivative of cyclen (tri-t-Bu-DO3A). Removal of the tertbutyl ester groups was then carried out using trifluoroacetic acid (TFA) to obtain L1. The successful isolation of the product was confirmed through 1 H and 13 C NMR spectroscopy along with high-resolution mass spectrometry. L1 was coordinated to a series of lanthanide metal chloride salts [Ln = Gd(III), Tb(III)] in water in a mildly acidic environment (pH ∼5.5). After purification, the absence of free lanthanide ions was verified by the use of xylenol orange indicator solution. 46 For each complex, a peak corresponding to the molecular ion with the correct isotopic pattern could be observed by highresolution electrospray ionization mass spectrometry. The Ln.L1 complexes were found to be only partially water-soluble and therefore unsuitable for in vivo applications. Thus, the second ligand L2 was developed to overcome the solubility issues of L1 in water. The synthesis of L2 was very similar to that of L1. Two subsequent amide bond formation reactions on rhodamine B were carried out, first with ethylenediamine to form 4 and then with chloroacetyl chloride resulting in product 5. This was then conjugated to tri-t-Bu-DO3A to give 6 in 39% yield, and subsequent deprotection with TFA gave L2 in the same manner as for L1. The second amide bond was introduced to increase the hydrogen-bond donor and acceptor ability and thereby to increase the water solubility. L2 was complexed with Gd(III) and Tb(III) as above, and the complexes Ln.L2 were found to be water-soluble.
Photophysical Properties. Complexes Ln.L1 and Ln.L2 ( Figure 1) behave as dual-modal MRI/optical imaging probes with Gd(III) and dual-luminescent probes with Tb(III). In this section, we discuss the photophysical properties of both systems. The photophysics of the two ligands with Gd(III) were found to be very similar, as they exhibited fluorescence emission at 580 nm upon excitation at 560 nm. This short Stokes shift and orange fluorescence was expected, as the fluorescence properties of rhodamine and its derivatives have been extensively studied. 47 Modification of the carboxylic acid moiety of rhodamine B does not lead to significant changes in the photophysical properties of the chromophore, as no   electronic or structural changes occur on the xanthene core, so a very small bathochromic shift is observed (rhodamine B: λ ex = 553 nm, λ em = 572 nm in ethanol).
The photophysical properties of the Tb(III) analogues were investigated ( Figure 2). Under excitation at λ ex = 310 nm, complex Tb.L1 displays exclusively green emission arising from 5 D 4 − 7 F J transitions of terbium. The rhodamine-centered emission is of a very low intensity in comparison with the Tb(III) emission, indicating efficient ligand-to-lanthanide energy transfer. At λ ex = 560 nm, emission is seen at 580 nm, featuring fluorescence from the xanthene core. Therefore, two fluorescence pathways are available to Tb.L1: emission via the lanthanide with energy transfer from rhodamine and direct excitation of the xanthene core. Conversely, under excitation at λ ex = 310 nm, the emission spectrum of complex Tb.L2 is dominated by rhodamine emission at 580 nm. Applying a 0.1 ms delay to the fluorescence measurement allowed for isolation of the short-lived organic fluorescence of the rhodamine moiety from the long-lived lanthanide luminescence in both complexes. The extra amide bond in L2 renders the complex water-soluble but also introduces a quenching pathway. The additional N−H bond is close to the lanthanide center, and fluorescence deactivation may occur via the N−H vibrational energy level oscillator. 48 Therefore, when complexed with a luminescent lanthanide ion, both ligands behave as dual-luminescent probes. In both Tb.L1 and Tb.L2, excitation at 560 nm leads to emission from the organic chromophore at 580 nm, with the fluorescence originating from the xanthene core. Using time-gating techniques, Tb(III) emission is induced in both complexes upon excitation of the isolated phenyl ring of the rhodamine moiety at 310 nm by energy transfer to the Tb(III) metal center (Figures 2 and 3). In Tb.L1 only, Tb(III) emission can be seen upon excititation at 310 nm without the use of a time delay. The quantum yields of both terbium complexes were determined, and the results were in agreement with the observed spectra, confirming the efficient energy transfer to the terbium metal center in Tb.L1 (Φ = 18.6%) and the inefficient energy transfer in Tb.L2 (Φ = 0.8%).
The luminescence decays of the Tb(III) complexes in both H 2 O and D 2 O were also measured (see the Supporting Information). The luminescence lifetimes in D 2 O were increased in comparison to H 2 O because less nonradiative deactivation was induced by O−D vibrations than by O−H vibrations. 49 From these differences in the rates of energy transfer to H 2 O and D 2 O oscillators, the hydration state can be calculated using eq 1, in which the term 0.06 reflects the quenching effect of unbound water molecules and n is the number of coordinated secondary amide groups, reflecting the quenching effect of the proximate NH oscillators. 50 The emission decays were fitted to monoexponential decays, and the corresponding lifetime measurements of both terbium complexes confirmed the expected coordination numbers of seven for Tb.L1 and eight for Tb.L2 (Table 1).
Relaxometric Studies. T 1 measurements of the Gd(III) complexes were performed at 400 MHz (9.4 T, 25°C), and their r 1 values were determined. The r 1 values of 8.5 and 3.8 mM −1 s −1 obtained for complexes Gd.L1 and Gd.L2, respectively, were within the range for Gd(III) chelates with hydration states of two and one respectively. These hydration states were confirmed by the lifetime measurements of the terbium analogues.
It is well-known that the presence of hydrophobic moieties within a probe is one of the basic structural requirements for ligand binding to human serum albumin (HSA). 51 HSA is the most abundant protein in blood, and binding to it results in an increase in intravascular retention and relaxivity (due to a decrease in τ R of the complex). 52 In view of the fact that the rhodamine moiety has extended aromatic rings and hence hydrophobicity, a study was carried out to investigate possible interactions between Gd.L2 and HSA. The relaxation rates of a series of solutions of Gd.L2 with increasing concentrations of HSA were measured. The relaxation rate was found to remain constant across the series (Figure 4), therefore indicating that there is no significant binding of the probe to the protein. The binding mechanisms of rhodamine B with bovine serum albumin (BSA) were previously investigated, and site-selective binding was observed. However, these results were obtained using high molar ratios of rhodamine B to BSA (>50). 53 Such high concentrations are not practical for use of MRI probes in vivo, and indeed, large changes in relaxivity have been observed in cases where the molar ratio of probe to serum albumin have been much lower. 54,55 It is hypothesized that the steric bulk of the complex is unsuitable for selective binding to HSA.
pH Titration. As previously discussed (vide supra), the lactam ring of rhodamine can be intramolecularly opened and closed via changes in pH. 36 The pH sensitivity of Gd.L2 was investigated, and to show the sensitivity of the proton-triggered ring opening, a pH titration was carried out wherein the emission intensity was monitored at various pH values ( Figure  5). The pH titration showed that Gd.L2, which was nonfluorescent at pH 6.5 or above, rearranged into its fluorescent form at acidic pH. The fluorescence of Gd.L2 at pH 4.3 was 50-fold brighter than that at pH 6.5 and 15-fold brighter than that at pH 6.0, demonstrating that Gd.L2 is a sensitive, acid-responsive probe capable of sensing small changes in pH in the range 4−7.
A titration curve was produced by plotting the emission intensity at λ em = 580 nm versus pH, which yielded a pK a value of 5.11. It was hypothesized that the change in the charge of Gd.L2 on going from pH ∼ 7.0 to 4.0 would aid the probe to    localize in the more acidic microenvironment of tumor cells. 56 It has been shown that the microenvironment of tumor cells is more acidic than that of healthy tissues (Warburg effect). Upon entering an acidic environment, the probe can undergo rearrangement to its ring-opened form, which is fluorescent. This cationic acid form of the probe would then be expected to accumulate in the mitochondria. 57 In Vitro/in Vivo Studies. To test the viability of Gd.L2 as a dual-modal agent for cancer imaging, in vivo and in vitro studies were carried out. Because of a lack of water solubility, no in vivo studies could be carried out on L1 complexes. The addition of an extra amide functionality within L2 introduced water solubility, and this allowed for the investigation of its biological properties via in vitro and in vivo studies. Cell penetration is crucial for the use of imaging probes in the monitoring of biological mechanisms. To this end, preliminary in vitro cell studies were carried out using Gd.L2. Rhodamine has been widely used to determine the mitochondrial potential in tumor cells. 9 The negative mitochondrial potential is reported to be greater in tumor cells than in normal cells, and organic cations such as rhodamine have been shown to accumulate in the energized mitochondria. 57,58 With this in mind, colocalization experiments were undertaken wherein a known mitochondrial stain (MitoTracker Green FM) was tested against Gd.L2. Confocal microscopy images of human embryonic kidney (HEK293) cells and primary mouse islet cells (∼70% insulin-containing β cells) were recorded after a 30 min incubation with Gd.L2 and MitoTracker Green ( Figure 6). Figure 6 show clear accumulation of Gd.L2 within each cell type. However, a clear colocalization of Gd.L2 and MitoTracker was not evident ( Figure 6C, box). Moreover, we noted that in cells where significant uptake of Gd.L2 was apparent (as indicated in the primary cells in D−F), MitoTracker accumulation was lower and more diffuse than in neighboring cells, and little, if any, colocalization of red and green staining was evident ( Figure 6F). These findings are probably best explained by the fact that the Gd.L2 fluorescence, which is strongly suppressed at pH >6.5 ( Figure 5B), is only readily detectable in cells in which mitochondria are relatively depolarized, such that the mitochondrial matrix pH falls to levels permissive of Gd.L2 fluorescence. Such mitochondria, however, are expected to be poorly able to retain MitoTracker, whose accumulation depends on the inner mitochondrial membrane potential and the pH gradient (i.e., high intramitochondrial pH >7). These results are thus consistent with the known properties of both MitoTracker and Gd.L2 but are unable to definitively establish a mitochondrial localization of the latter.

Images B, C and E, F in
BALB/c nude mice with neck M21 (human melanoma cell) xenograft implantations were injected intravenously with 0.1 mmol kg −1 Gd.L2. The mice were imaged in a 4.7 T MRI instrument at different times (preinjection, 30 min post injection, and 60 min post injection) to evaluate the effect of the contrast agent. Thirty minutes after the injection of Gd.L2, a 36.5% decrease in tumor average T 1 was observed compared with the baseline of the preinjection image (Figure 7a,b). This decrease in longitudinal relaxation time was retained at least up to 60 min (31.3% decrease; Figure 7c). These preliminary images demonstrate that Gd.L2 functions successfully in vivo by shortening the T 1 of protons in the tumor sufficiently so as to give a more detailed image. The observed accumulation of Gd.L2 in the tumor area is consistent with the predicted effect of a lower pH environment on the probe; the rearrangement of the lactam ring at acidic pH introduces a positive charge in the molecule that can promote the intracellular uptake of the molecule responsible of the average T 1 decrease observed in the tumor.

■ CONCLUSIONS
We have reported the preparation and full characterization of two rhodamine−DO3A conjugate derivatives. The DO3A units have been coordinated to paramagnetic Gd(III) and luminescent Tb(III) centers. Complexation with Tb(III) gives dual-luminescent probes with two different time scales, and photophysical data collected also show the impact of distance of the sensitizer and linker design from the lanthanide metal center on the efficiency of energy transfer. When complexed with Gd(III), they form dual-modal MRI/fluorescent probes. Ln.L2 is water-soluble and suitable for in vitro/in vivo studies, and Gd.L2 has r 1 = 3.84 mM −1 s −1 (9.4 T, 25°C) and has been shown to be viable both in vitro and in vivo. The pH sensitivity of the probe means that it functions as an "off−on" luminescent probe that is sensitive to the small pH changes within cells. By coexpression with a known mitochondrial stain, the probe has been hypothesized to localize in the mitochondria of HEK cells. These studies demonstrate that the probe is cell-penetrating, has a pH sensitivity to acidic microenvironments, and is a tumor specific, dual-modal MRI/fluorescent contrast agent.

■ EXPERIMENTAL SECTION
Materials and Conditions. No special precautions were taken to exclude air or moisture during reactions or workups, unless otherwise stated. Products 2 59 and 4 60 were prepared via literature methods from  commercially available starting materials. All other materials were purchased from commercial suppliers and used without further purification.
Instrumentation. 1 H and 13 C{ 1 H} NMR spectra were recorded at ambient temperature on a Bruker 400 MHz spectrometer and internally referenced to the residual solvent peaks of CDCl 3 at 7.26 ppm ( 1 H) and 77.16 ppm ( 13 C{ 1 H}) or MeOD at 3.31 ppm ( 1 H) and 49.15 ppm ( 13 C{ 1 H}). 13 C{ 1 H} spectra were fully assigned where possible using 2D correlation spectroscopy. Mass spectrometry analyses were conducted by the Mass Spectrometry Service, Imperial College London. Microanalyses were carried out by Stephen Boyer of the Science Centre, London Metropolitan University. ICP-MS analysis was carried out by Stanislav Strekopytov of the Natural History Museum.
Fluorescence Measurements. Absorption and fluorescence spectra of all complexes in either aqueous or methanolic solutions depending on solubility were obtained at room temperature on a PerkinElmer Lambda 25 spectrometer and a Cary Varian luminescence spectrometer (using SCAN for Windows), respectively. Samples were held in 10 mm × 4 mm quartz Hellma cuvettes. Lifetimes were measured by direct excitation of the sample at 325 nm with a 40 ms pulse of light (50 pulses per point) followed by monitoring of the integrated intensity of light emitted at 545 nm during a fixed gate time of 0.1 ms at a delay time later. Delay times were set at 0.1 ms, covering four or more lifetimes. Excitation and emission slits were set to 10:10 nm bandpass respectively. The obtained decay curves were fitted to a simple monoexponential first-order decay curve using Microsoft Excel. Fluoresence quantum yields of terbium complexes were determined from the integrated fluorescence intensities of the complexes following a previously reported method 61 using the reference compound 62 Na 3 [Tb(dpa) 3 ] in water or methanol at 279 nm (quantum yield = 26.5% or 18.2%, respectively).
pH Titration. A 1.49 mM solution of Gd.L2 in 0.1 M NaCl (1 mL) was prepared, from which 15 × 67 μL aliquots (microlitre pipet, Eppendorf) were taken. Each aliquot was adjusted to the required pH using a digital pH meter (Jenway 3510) equipped with a glass electrode (Jenway 924005). The pH was monitored and adjusted to acidic or basic conditions using small aliquots of a 1 M HCl or 1 M NaOH solution, respectively. The pH was allowed to stabilize. The solution was then added to a quartz cuvette for fluorescence analysis. The fluorimeter was operated using an excitation slit width of 10 nm and an emission slit width of 2.5 nm.
Relaxivity Measurements. Gd.L1 was dissolved in a methanol/ water (50:50) solvent mixture to give a solution of concentration 5 mM. Gd.L2 was dissolved in water to give a solution concentration of 5 mM. The resulting solutions were placed in 1.7 mm diameter capillaries, which were sealed. The 1/T 1 measurements were performed on a Bruker Avance 400 spectrometer (400 MHz). These experiments were carried out a minimum of three times, and the relaxivity results given are averages of all experiments. The concentration of Gd(III) was checked by the chemical shift measurement of HOD induced by magnetic susceptibility. 63 HSA Relaxivity Titration. In Vivo Tumor Models. All of the animal experiments were performed by licensed investigators in accordance with the United Kingdom Home Office Guidance on the Operation of the Animal (Scientific Procedures) Act 1986 and in keeping with the newly published guidelines for the welfare and use of animals in cancer research. 64 Female BALB/c nude mice (aged 6−8 weeks; Harlan) were used. M21 cells were injected subcutaneously on the back of mice (5 × 10 6 cells in 100 μL of sterile PBS). Animals were used when the xenografts reached ∼100 mm 3 . Tumor dimensions were measured continuously using calipers, and tumor volumes (V) were calculated using the equation V = πabc/6, where a, b, and c represent the lengths of the three orthogonal axes of the tumor.
In Vivo MRI. In vivo MRI was performed in a 4.7 T horizontal-bore DirectDrive MRI system (Varian, Palo Alto, CA, USA) equipped with 40 G/cm actively shielded gradients (VnmrJ 3.1). For imaging of the tumors and generation of the T 1 parametric maps, a 45 mm diameter saddle coil was used in transmit mode and a separate 20 mm diameter stripline resonator was used as a surface coil to accommodate the tumors. 65 Imaging was performed with an inversion recovery (IR) sequence and the following parameters: axial images; FOV = 35 × 35 mm; No slices = 14; slice thickness = 1 mm (zero slice gap); image bandwidth = 156 kHz; T R = 4500 ms; T I = 6, 14.6, 35.4, 86.1, 209, 508, 1230, or 3000 ms; NA = 2; image matrix = 192 × 192; TA = 12 min. During imaging, animals were anesthetized with 2% isofluorane and respiration and body temperature were monitored via SA physiological monitoring systems (SA Instruments, Stony Brook, NY, USA). Body temperature was monitored by a rectal probe and maintained at 35 ± 1°C. T 1 maps were generated in ImageJ 66 by fitting the acquired data to the equation S = S 0 [1 − 2 exp(−T I /T 1 )], where T I is the inversion time and T 1 is the longitudinal relaxation time. In order to assign the signal with the appropriate sign for the fitting process, the χ 2 value of the fit was evaluated during the fitting. ROIs were then drawn on the T 1 maps to delineate the whole tumor with consideration to avoid the edges of the tumor and thus contamination from neighboring tissues and partial volume effects.
In Vivo Tissue Isolation and Imaging. Islet isolation was carried out as shown in ref 67. In short, a female CD1 mouse (8−12 weeks old) was sacrificed by cervical dislocation. The pancreas was injected with 1 mg/mL collagenase solution through the bile duct and excised. The pancreas was digested by heating with collagenase solution, and islets were isolated through the use of ficoll gradient centrifugation. Islets were hand-picked and kept at 37°C in Islet medium (RPMI supplemented with 10% FBS and 1% penicillin/streptomycin). Islets were dispersed into clusters and single cells according to the procedure in ref 67. Briefly, islets were incubated with Hank's based enzyme-free dissociation buffer (Invitrogen, Paisley, UK) and centrifuged. All but 1 mL of the supernatant was removed, and the pellet was resuspended vigorously via pipetting. The cells were washed and resuspended with RPMI 1640 and then loaded onto a poly-L-lysine-treated coverslip. On the day of visualization, cells were washed twice with KREBS buffer (140 mM NaCl, 3.6 mM KCl, 0.5 mM NaH 2 PO 4 , 0.2 mM MgSO 4 , 1.5 mM CaCl 2 , 10 mM HEPES, pH 7.4, 2 mM NaHCO 3 ) pre-equilibrated with 95:5 O 2 /CO 2 and containing 25 mM glucose. Cells were then incubated with KREBS buffer containing 215 mM Gd.L2 and 100 nM MitoTracker Green for 30 min and washed twice with KREBS buffer before visualization. Images were captured using a Zeiss Axiovert confocal microscope coupled to a Nipkow spinning-disk head (Yokogawa CSU-10) using a 63× oil objective. A solid-state laser