“Magnetic resonance imaging contrast enhancement in vitro and in vivo by octanuclear iron-oxo cluster-based agents”

Abstract

A water-soluble octanuclear cluster, [Fe₈], was studied with regard to its properties as a potential contrast enhancing agent in magnetic resonance imaging (MRI) in magnetic fields of 1.3, 7.2 and 11.9 T and was shown to have transverse relaxivities r₂ = 4.01, 10.09 and 15.83 mM s⁻¹, respectively. A related hydrophobic [Fe₈] cluster conjugated with 5 kDa hyaluronic acid (HA) was characterized by ⁵⁷Fe-Mössbauer and MALDI-TOF mass spectroscopy, and was evaluated in aqueous solutions in vitro with regard to its contrast enhancing properties [r₂ = 3.65 mM s⁻¹ (1.3 T), 26.20 mM s⁻¹ (7.2 T) and 52.18 mM s⁻¹ (11.9 T)], its in vitro cellular cytotoxicity towards A-549 cells and COS-7 cells and its in vivo enhancement of T₂-weighted images (4.7 T) of a human breast cancer xenografted on a nude mouse. The physiologically compatible [Fe₈]-HA conjugate was i.v. injected to the tumor-bearing mouse, resulting in observable, heterogeneous signal change within the tumor, evident 15 minutes after injection and persisting for approximately 30 minutes. Both molecular [Fe₈] and its HA-conjugate show a strong magnetic field dependence on r₂, rendering them promising platforms for the further development of T₂ MRI contrast agents in high and ultrahigh magnetic fields.

1. Introduction

Over the past three decades, research in magnetic resonance imaging (MRI) has grown into a multidisciplinary field, spanning fundamental research to clinical diagnosis [1–3]. The major reasons for this growth are its non-invasive, ionizing radiation-free nature and its ability to perform 3D visualization [4].

Voxel brightness in MRI is governed by four parameters: local bulk magnetic susceptibility, local proton density, proton longitudinal (spin-lattice) relaxation time, T₁, and proton transverse (spin-spin) relaxation time, T₂. Pulse sequences that emphasize natural differences between the T₁ and/or T₂ times are available, often providing sufficient contrast to visualize the desired objects [5]. However, when these differences are not sufficiently differentiated, image contrast can be further improved by the artificial manipulation of T₁ and T₂ in the tissues to be imaged. This is achieved by the administration of physiologically acceptable, paramagnetic inorganic materials, which act as contrast enhancing agents (CAs). CAs shorten the T₁ and T₂ times of protons – paramagnetic resonance enhancement (PRE) [6] -- in tissues accessible to the CA, thus improving the contrast between types of tissue, radically emphasizing anatomic and pathologic features of concern [7].

Targeted CAs offer another dimension of molecular specificity to the abundant anatomical and functional information that MRI already provides [8, 9]. Target binding provides the pharmacodynamic effect of increasing the relaxivity of the CA, and therefore the MR signal. For example, using a targeted CA at a field strength of 1.5 T yielded images of contrast quality previously achieved only at 3.5–7.0 T with the non-targeted counterpart [10]. Other examples of the advantages of targeted CA in cancer diagnosis and detection have been shown with folate receptor (FR)-targeted compounds, octeotride, Herceptin, and others [11–13]. Also, CAs which contain an aptamer targeting moiety provide a single system where optimization of magnetic properties, pharmacokinetics and biodistribution can all be addressed at the same time [14]. The benefits and disadvantages of targeted CAs have been reviewed, and even though they yet remain to be introduced to the clinic, the increased relaxivity achieved by target binding outweigh the high cost and the synthetic and regulatory complications [15].

Gadolinium (Gd) chelates, with seven unpaired electrons and slow electron-spin relaxation, have been the most widely used positive CAs (brightening the voxels in their proximity) having a significant T₁-shortening effect [5, 7, 16–17]. For negative contrast (darkening of the voxels), dextran-coated superparamagnetic iron oxide nanoparticles (SPIONs), which induce large local field inhomogeneity, thus shortening the T₂ of protons in their vicinity, are generally used [18–23]. Despite the success of Gd³⁺-chelates as T₁-CAs, the discovery of nephrogenic systemic fibrosis (NSF), associated with agents involving acyclic chelates [24], has challenged the widespread use of these MRI CAs [25–29]. Besides NSF, recent in-depth studies have demonstrated Gd phosphate precipitation in mouse embryonic 3T3-L1 fibroblast cells grown in media containing Gd-based CAs [30] and more seriously, Gd accumulation in the brain of cadavers of patients with a history of Gd-based MRI-CA injections [31].

While Gd-chelates remain the principal MRI-CAs in clinical use, their potential toxicity issues have launched a search for CAs with alternative metal-ions [32–34]. Early efforts to develop Mn²⁺-based MRI-CAs showed accumulation of metal ions in the brain [35, 36]; nevertheless, several Mn-based CAs have been the subjects of recent pre-clinical investigations [37–39]. Natural products with high paramagnetic ion content have also been investigated [40]. Unlike Gd, iron is a natural cellular constituent and often a cofactor for enzymes and receptors, having the advantage of excretion through the normal biochemical pathways of iron metabolism [41], although an excess of iron can still be toxic [42, 43]. Iron is also inexpensive and the most abundant transition metal on Earth. Consequently, mononuclear Fe³⁺ complexes have been studied as hepatobilliary MRI-CAs for a long time [44–46]. Dextran coated SPIONs of sizes between 5 and 50 nm [47] have been used for MR imaging of liver cancer [21–23]. A variety of methods have been developed that allow control of surface properties and size of iron oxide nanoparticles for use as MRI-CAs [48]. It should be noted that in spite of the growing number of new magnetic clusters in the literature, the application of polynuclear transition-metal complexes (to be differentiated from superparamagnetic nanoparticles, mentioned above) in MRI contrast enhancement has been less explored, possibly due to their water-insolubility and ligand lability.

Herein we present a preliminary evaluation of the MRI-CA properties of the novel [Fe₈(μ_4-O₄(μ−4-R-pz)₁₂Cl₄] ([Fe₈]; pz = pyrazolato anion, C₃H₃N₂^-) family of complexes and we propose them as the basis for further development of improved CAs (Fig. 1). We discuss the in vitro water-proton relaxation properties induced by these paramagnetic clusters in a directly water-soluble form, R = CH₂CH₂OH (1), and compare these results with the corresponding values of gadoversetamid (2-[bis[2-[carboxylatomethyl-[2-(2-methoxyethylamino)-2-oxoethyl]amino]ethyl]amino]acetate;gadolinium(3+), brand name Optimark™), a clinically approved Gd-based CA (in spite of its known instability in vivo [24]). We also describe the in vitro and in vivo evaluation of a prototypical targeted MRI-CA consisting of a conjugate of [Fe₈] and a tyramine-functionalized hyaluronic acid ([Fe₈]-Tyr-HA, 2) in which [Fe₈] is the hydrophobic variant with R = H. In compound 2, HA is a vector targeting the CD44 proteoglycan, which is overexpressed in ovarian, breast and other cancer cells (vide infra).

Figure 1.

Ball-and-stick diagram of [Fe₈(μ_4-O₄(μ−4-R-pz)₁₂Cl₄] (R = H, [Fe₈]; R = CH₂CH₂OH (1), not shown). Color coding: Yellow, Fe; red, O; blue, N; green, Cl; and black, C. H-atoms not shown for clarity

2. Experimental

2.1. Materials and methods

The following reagents were obtained from commercial sources and used as received: phosphate buffered saline (PBS), anhydrous ferric chloride, triethylamine, pyrazole, ferrozine, hydroxylamine, trichloroacetic acid (TCA), sodium acetate, ammonium citrate, and tris(hydroxymethyl)aminomethane (Tris). Tetrahydrofuran (THF) was distilled over CaCl₂, while anhydrous CH₂Cl₂ was used as received. Complexes [Fe₈] and 1 were prepared as previously described [49–51]. The tyramine (Tyr)-modified HA (Tyr-HA, MW = 5 kDa; average 13.4 disaccharide units), Tyr-HA, was also prepared by a procedure described previously [52]. Approximately 12% of the pendant carboxylic groups of the HA chain were condensed with Tyr, forming Tyr-HA via an amide bond (Scheme 1); e.g., on average, 1.6 Tyr units per 5 kDa HA chain. This level of substitution is sufficiently low to still preserve strong CD44 binding. Additional experimental details, along with an instrumentation list, are provided in Supplementary Information.

Scheme 1.

Tyramine functionalization of hyaluronic acid.

2.2. Proton relaxation times

Water proton relaxation times were determined by a Bruker Avance 500 MHz (11.9 T), a Bruker Avance 300 MHz (7.2 T) and a SpinMaster 55 MHz (1.3 T) instruments at 298 K. Longitudinal and transverse relaxation times (T_i, i = 1 and 2) were determined by spin-inversion recovery experiments (180º-τ−90º) and the Carr-Purcell-Meiboom-Gill (CPMG, 90º-τ/2–180º-τ/2) pulse sequences, respectively. Relaxivity values, r₁ and r₂, were determined by best-fit plots of T_i^–1 vs. concentration of compound 1, according to T_i^–1 = T_i,0^–1 - r_i[Fe₈]. Fresh solutions were prepared prior to use for T₁ and T₂ measurements. To determine the relaxivity of the paramagnetic cluster in aqueous solution, 5.0 mM stock solution of 1 in deionized water was prepared and was successively diluted to concentrations in the range of 0.7 – 5.0 mM. For 2, a 1.1 mM stock solution was to prepare diluted aliquots in the 0.1 – 1.1 mM range. For each concentration, T₁ and T₂ experiments were run in triplicate, using three independently prepared sets of fresh solutions, in order to obtain nine values. Relaxivities, r₁ and r₂, were determined from the slope of plots generated from the mean of the nine values of 1/T₁ or 1/T₂ vs. CA concentration.

2.3. MRI of phantoms

MRI phantoms were obtained using 3 mL solutions in 5 mL glass vials supported on a Teflon base. The MRI experiments were performed with phantoms containing varying concentrations of Fe₈-based CAs, using gadoversetamide as a standard comparison for gadolinium-based CAs, and one phantom with deionized water (blank). All images and scans were acquired using standard T₁-weighted and T₂-weighted multi-echo spin-echo imaging protocols at room temperature in a scanning suite at 15–18 ºC. For each sample, five 5mm slices were obtained with 1 mm space in between slices at the smallest field-of-view available for the sequence using 512 × 512 imaging matrices. Measurements of the image integrated signal intensity of equal circular Regions-of-Interest (ROIs) within the center slices image of each sample, the image of deionized water (blank) and the background were obtained. For T₁ weighted images the TE = 9–20 ms, TR was varied from 100 ms to 3,000 ms, and the excitation angle was 90 degrees. For T2 weighted images TR = 3,000 ms, TE was varied from 30 ms to 120 ms in the multi-echo sequence, and the excitation angle was 90 degrees. T₁ and T₂ were determined by measuring the MR image integrated intensity changes as a function of the repetition time TR and the echo time TE, respectively. These relaxation times were calculated using least squares fitting of the image intensities versus time curves obtained from the phantom images.

2.4. Synthesis of [Fe₈]-Tyr-HA conjugate 2

Tyramine-functionalized HA (Tyr-HA) was coupled to [Fe₈] via its phenolic groups, forming the new Fe₈-Tyr-HA conjugates (2). A solution of Tyr-HA (0.050 g, ~10^–2 mmol) in 10 mL of deionized water was prepared. To this solution, [Fe₈] (0.018 g, ~1.2 × 10^–2 mmol) in the minimum amount of THF was added drop wise with stirring at room temperature. The solution was stirred for approximately 10 min and the solvent was removed using a rotary evaporator. The dark brown solid product was rinsed with CH₂Cl₂ to remove any excess [Fe₈] and vacuum dried in a dessicator.

2.5. Solubility and stability study of 2

The stability of 2 was evaluated under physiologically relevant conditions. It was prepared as a stock solution (130 μM) in water and then diluted to 3.3 μM in water and also in pH 7.4, 25 mM Tris, 0.1 M NaCl buffer containing 10% fetal bovine serum. Four separate sets of the final solutions were prepared. The solutions were then monitored by UV-Vis spectroscopy every 15 minutes for 2 hours and then every hour for 24 hours at 25 ºC temperature. See Supplementary Information for experimental details.

2.6. Cell culture

A-549, COS-7 and MDA-MB-468 cells were maintained in accordance with ATCC protocols. The cells were cultured in 75 cm² flasks with phenol red DMEM (Sigma, D6429) containing 1% of glutamine and 4.5 g/mL of glucose and sodium pyruvate, and supplemented with 10% FBS (HyClone), 1% of antibiotic solution prepared with 11 mg/mL streptomycin and 7 mg/mL penicillin (Calbiochem®, EMD Biosciences Inc.) in a humidified incubator with 5% CO₂ for A-549 and COS-7 and without CO₂ for MDA-MB-468 at 37 °C. All experiments were conducted after 3 to 10 cell passages.

2.7. Cell viability assay

A stock solution of 90 µM of 2 was prepared in nanopure-quality water. A-549 and COS-7 cells at >80% confluence were seeded (5.0 × 10⁴ cells/mL) into 96-well plates with 100 μL of DMEM per well. After 24 h, cells were incubated with 30 μM of of 2 in a total volume of 200 μL of DMEM for 48 h. Next the viability of the cells was measured using the CellTiter 96 aqueous non-radioactive cell proliferation assay (Promega). 1 h before completing the incubation time, 20 µL of 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt (MTS), and phenazine methosulfate (PMS) solution was added to each well (333 µg/ml MTS + 25 µM PMS). After 1 h, the absorbance of each well was measured at 492 nm in a Multiskan FC, Thermoscientific plate reader spectrophotometer. Cells treated with 150 µM linoleic acid were used as positive control (cytotoxic agent, low viability), and 1 mg/mL α-lactalbumin and untreated cells were negative controls (no cytotoxicity, 100% viability).

2.8. In vivo Studies

Imaging studies were conducted in the Small Animal Imaging Facility (SAIF) of MD Anderson, under a protocol approved by the Institutional Animal Care and Use Committee (IACUC). Female nude mice (Harlan) were implanted in their mammary fat pads with human MDA-MB-468 breast tumor cells, according to our previous protocols [53, 54]. Depending on the maturity of the mice and the extent of development of the mammary tree, usually the fat pads immediately inferior to those most superior were injected, either on one side of the tree or on both. Once the orthotopic tumors had grown to at least 5 mm in their longest diameter (~ 3 weeks), these xenografted mice were MR imaged on a 4.7 T scanner. To both confirm intravenous (i.v.) injection of the Fe₈-Tyr-HA and to probe the plasma kinetics of its clearance, mice were scanned dynamically using a fast spoiled gradient echo sequence (TE = 1.6 ms, TR = 76 ms, 30° excitation angle). To assess tumor localization of this CA, anatomic imaging sequences were utilized: T₂-weighted (TE_eff = 70 ms, TR=4150 ms), and T₂*-weighted (TE_min = 3.06 ms, TE = 3.64 ms, TR = 1200 ms, 30° excitation angle). MRI scans were obtained immediately prior to and after administration of Fe₈-Tyr-HA_, either by direct intra-tumoral or i.v. injection. Application of ultrasound gel on these superficial tumors minimized air-induced artifacts on the MR images.

3. Results and Discussion

We have previously studied the physical chemistry of octanuclear ferric complexes of the general formula [Fe₈(μ₄-O)₄(μ−4-R-pz)₁₂X₄] -- pz = pyrazolato anion; X = Cl, Br, NCS -- with various R-substituents at the pyrazole 4-position: R = H, Cl, Br, I, CH₃, (CH₂)_2,3Cl and (CH₂)_2,3OH (Fig. 1) [49, 50, 51]. The parent compound, with R = H and X = Cl, has an effective magnetic moment of 7.0 B.M. at ambient temperature, resulting from the antiferromagnetic coupling of eight high-spin Fe³⁺ centers [49]. The almost spherical shape (no dipole moment) and surface defined by carbon, hydrogen, and halogen atoms of these molecules render them highly hydrophobic. More recently, however, we have also reported the synthesis of two new variants of these octanuclear complexes with R = -CH₂CH₂OH (1) and -CH₂CH₂CH₂OH, which, because of their hydrophilicity, render the octanuclear complex directly soluble in water [50]. In aqueous solutions, 1 forms two types of H-bonded super-clusters consisting of ~13 and ~90 [Fe₈] units [50]. The fact that a well-resolved ¹H-NMR spectrum can be recorded for paramagnetic 1 implies that its electron spin relaxation time, T_ie, is shorter than the NMR timescale of 10^–11- 10^–13 s. This is also consistent with typical T_ie values of 10⁻¹⁰ s reported for high spin Fe(III) compounds [7]. In aqueous media, hydrolysis of their terminal Fe-Cl bonds leads to the formation of H-bonded aggregates with dimensions of 5–7 nm (~13 Fe₈ molecules) and 40–60 nm (~90 Fe₈ molecules), as determined by DLS [50].

3.1. Relaxivity of complex 1

The proton longitudinal relaxation rate (R₁) and transverse relaxation rate (R₂) of aqueous solutions of 1 were measured in three different magnetic fields in the 0.7 – 5.0 mM range, limited by the solubility of 1. The plots (Fig. 2 and S1) show the expected linear relationship with the concentration of the paramagnetic cluster. For 1, the slope of the least-square fit lines indicates unremarkable longitudinal relaxivity values (r₁, Table 1) and marginal field dependence within experimental error. In contrast, the transverse relaxation rates (R₂) are faster -- significantly higher transverse relaxivities (r₂, Table 1) -- and show a dependence on the magnetic field. The R₂ value of the 5.0 mM solution at 11.9 T was treated as an outlier, as no measurements could be conducted at higher concentration to examine whether it represents an inflexion point of the line.

Figure 2.

Plot of 1/T₂ versus concentration of [Fe₈O₄(4-OHCH₂CH₂-pz)₁₂Cl₄], 1, in water, 298 K, at 1.3, 7.2 and 11.9 T.

Table 1.

Longitudinal and transverse relaxivities^a in aqueous solutions at 25 ºC.

		r1 (mM−1s−1)			r2 (mM−1s−1)	[r2/r1]
	1.3 T	7.2 T	11.9 T	1.3 T	7.2 T	11.9 T
[Fe8O4(4-OHCH2CH2-pz)12Cl4] (1)	0.68	0.75	0.76	4.01 [5.9]	10.09 [13.5]	15.83 [20.8]
Phantom imaging of 1 (at 1.5 T)	0.86	-	-	4.84 [5.6]	-	-
Fe8-Tyr-HA (2)	0.09	0.12	0.12	3.65 [40.6]	26.20 [218.3]	52.18 [434.8]
Gadoversetamide (Optimark™)	3.8b	3.8c	-	4.2b	4.7c	-

^a For comparison to mononuclear Gd CAs, the relaxivity is reported in mM^–1 of the cluster (as opposed to mM^–1 of metal);

^b At 1.5 T and 37 ºC (from ref. 63);

^c At 4.7 T and 37 ºC (from ref. 63).

The ratio of r₂/r₁ is an indicator of relaxometric properties of an imaging agent, and it is used to classify a given type of MRI CA. In general, for mononuclear paramagnetic chelates, r₂/r₁ ratios range between 1 and 2, whereas for SPIONs, typical values are > 3 and can be as large as 50 in the case of superparamagnetic colloids [55, 56]. Gd complexes, with r₁ values ranging from 4–16, are used primarily as T₁-agents [57], whereas the SPIONs are T₂-agents with a r₁ ~30 mM⁻¹ s⁻¹ and r₂ ~100 mM⁻¹ s⁻¹ [58]. The relaxivity r₂/r₁ ratios of compound 1 range from 5.9 (at 1.3 T) to 20.8 (at 11.0 T), suggesting the Fe₈ system as a T₂-contrast agent.

The magnetic field strength is one of the factors influencing relaxation properties. Relaxation rates are described by the Solomon–Bloembergen–Morgan theory, which determines that the correlation time, τ_Ci, is dominated by the most rapid one among three contributing factors: T_ie (electron-spin relaxation times of the metal ion), τ_R (re-orientational correlation time of the metal-proton vector), and τ_m (rate of water-proton exchange) [59]; the former is magnetic field-dependent, influencing the overall τ_Ci, and hence the relaxation rates. Consequently, the increase of r₂ values of 1 is tentatively attributed to a slower electron spin relaxation of the [Fe₈] cluster with increasing magnetic field. In the case of Gd-chelates, T_ie may be the limiting factor below 0.5 T, but not at 1.5 T. Instead, 1/τ_Ci is dominated by τ_R above 1.5 T. In the present case, 1 has a shorter T_ie (10^–11- 10^–13 s, vide supra) than mononuclear Fe(III), Mn(II) and Gd(III) complexes (T_ie = 10⁻⁸ - 10⁻¹⁰ s), resulting in a different influence on τ_Ci and ultimately on the relaxation rates. The longitudinal relaxivity of 1 is practically invariable, whereas its transverse relaxivity increases with increasing magnetic field strength. The low r₁ values are attributed to the inter-cluster H-bonded aggregation of [Fe₈] units (previously reported [50]), which render the Fe-ions in the interior of the aggregate inaccessible to H₂O molecules. On the other hand, the magnetic field dependence of r₂ values are attributed to the increased population of the closely-spaced [49] higher spin-states at increasing magnetic fields. These results indicate that contrast enhancement by compound 1 will improve in T₂-weighted images acquired in stronger magnetic field scanners, while T₁-weighted ones will not benefit from a stronger field. Three other octanuclear Fe(III) complexes have been studied with regard to their relaxivities, alas in narrower concentration ranges and in only one or two magnetic field strengths [60–62]. The r₁ values of these three complexes (1.1 – 5.4 mM⁻¹s⁻¹) are higher than those of 1, however, their r₂ values (3.1 – 8.4 mM⁻¹s⁻¹) are all lower than those of 1.

3.2. Phantoms

Phantom imaging allows the visualization of the CA effect. Imaging of 1 at four concentrations (5 mM, 2.5 mM, 1 mM and 0.5 mM), and of deionized water as control, allowed estimation of the r₁ and r₂. For comparison, solutions of the same four concentrations of gadoveretsamide were also imaged. Both T₁- and T₂-weighted images were obtained. For the T₁-weighted sequence, scans were run using five repetition times, TR = 100, 200, 400, 600, 800 and 2000 ms (Fig. 3). For the T₂-weighted multi-echo sequence, scans were run with four different echo times, TE = 30, 60, 90 and 120 ms (Fig. 4). For 1, the calculated r₁ and r₂ (Table 1) from these phantom images at 1.5 T are in good agreement with those determined by relaxometry at 1.3 T (see previous section). Visual inspection of the phantom images confirms that at 1.5 T, gadoveretsamide is superior to 1 as a positive (T₁) contrast agent, but of approximately of the same efficiency as a negative (T₂) contrast agent, in accordance with expectation based on the respective r₁ and r₂ values (Table 1).

Figure 3.

T₁-weighted imaging, scanned with varying TR; TE = 9.6 s, water blank and gadoversetamide reference (upper row), 1 (lower row), 18 ºC, 1.5 T.

Figure 4.

T₂-weighted imaging, scanned with varying TE; TR= 3000 s, water blank and gadoversetamide reference (upper row), 1 (lower row), 18 ºC, 1.5 T.

3.3. Characterization of Fe₈-Tyr-HA (2)

The four labile chloride terminal ligands of [Fe₈] allow its coupling to other moieties containing functional groups that can be metathetically exchanged for chloride, such as phenols. Hyaluronic acid (HA) is an unbranched polysaccharide distributed extensively throughout connective, epithelial and neural tissues, whose extra-cellular binding receptors include the cell surface glycoprotein CD44, a main receptor for HA [64–68]. Taking a cue from the well-known affinity of iron for phenolate ligands (e.g., transferrins), we prepared tyramine-functionalized HA, Tyr-HA.

An EDS spectrum of the solid Fe₈-Tyr-HA conjugate (Fig. S2) showed the presence of iron and traces of chlorine in the conjugate, which was confirmed by an elemental analysis showing a 7.21% Fe-content. TGA analyses were performed for [Fe₈], Tyr-HA and Fe₈-Tyr-HA (2): A sample of [Fe₈] was decomposed to FeO after heating beyond 400 ºC under air, while Tyr-HA was decomposed, leaving only 4% w/w of residue upon heating beyond 600 ºC. Finally, a sample of Fe₈-Tyr-HA was decomposed leaving a 17% w/w residue, attributed to 4% from the Tyr-HA plus 13% from the Fe₈ components, respectively. The latter matched satisfactorily with the 11.2% FeO residue calculated based on the elemental analysis of the same sample (Fig. S3). The combined elemental analysis and TGA results suggest that approximately 84% of Tyr-HA functionalized oligosaccharides are attached to an [Fe₈] unit, e.g., the composition of 2 is approximately 16% Tyr-HA and 84% Fe₈-Tyr-HA. The IR spectrum showed the diagnostic Fe-O stretch of Fe₈-cluster at 477 cm^–1 in an area where the spectrum of iron-free Tyr-HA is featureless (Fig. S4).

Deconvolution of the 80 K ⁵⁷Fe-Mössbauer spectrum (Fig. 5) revealed three components: Site A (blue trace, δ = 0.44 mm s^–1, ΔE_Q = 0.43 mm s^–1) attributed to the four iron centers of the Fe₄O₄-cubane core; site B (red trace, δ = 0.36 mm s^–1, ΔE_Q = 0.83 mm s^–1) attributed to outer Fe centers of Fe₈ coordinated to the phenolic oxygen atoms of tyramines; and site C (green trace, δ = 0.37 mm s^–1, ΔE_Q = 1.89 mm s^–1) attributed to outer Fe centers coordinated to μ-O-atoms, resulting from the hydrolysis of Fe-Cl bonds and forming bridges between adjacent Fe₈-units. The above spectral deconvolution assumed that 50% of total iron occupies A sites, resulting in occupation of sites B and C of 20% and 30%, respectively. Based on the Mössbauer spectrum interpretation, the following average structural model (Scheme 2) is proposed for the Fe₈ content of 2: Three Fe₈-units are linked by three μ-O atoms allowing the remaining six outer iron sites to be coordinated by phenolic O-atoms of tyramines. The assignment of the deconvoluted spectral features to sites A, B and C is based on previous experimental and theoretical studies of [Fe₈] and related complexes [69]. Scheme 2 depicts the types of bonding leading to clustering of Fe₈-Tyr-HA units.

Figure 5.

⁵⁷Fe-Mössbauer spectrum of 2 at 80 K. Site A, blue; site B, red; site C, green.

Scheme 2.

Tentative modes of bonding leading to the aggregate formation

In the solid state, Fe₈-Tyr-HA formed spheres ranging in diameter from 100 to 200 nm, as determined by TEM images (Fig. S5). In aqueous solution (7.5 mg Fe₈-Tyr-HA in 1.5 mL) the average size of particles was 143.7(4) nm (with a polydispersity index, PDI = 0.19), as determined by DLS. The particle size distribution remained invariant in the 4.5 to 9.5 pH range, with a sharp increase observed at pH = 10 (Fig. S6). In contrast, the ζ-potential of particles varied from −29 to −45 over the same pH range, consistent with gradual deprotonation of pendant carboxylic groups of HA at increasing pH (Fig. S7). Gel filtration chromatography estimated a 15.9 kDa mass for 2 in buffered aqueous solution, suggesting the presence of a three-ligand aggregate (Supplementary Information and Fig. S8). MALDI TOF spectroscopic results are also consistent with the formation of aggregates (Supplementary Information and Fig. S9).

3.4. Solubility and stability study of conjugate 2

The ferrozine assay showed that 2 was reasonably water soluble and yielded relatively high iron concentrations. Conjugate 2 reached a maximal soluble iron concentration of 5.51±0.43 mM or 0.17±0.015 mM compound concentration in water. The compound is extremely stable in water. At 3.3 μM it does not demonstrate any UV-Vis spectral changes even after 24 hours in solution (S10). The stability of the compound at the same concentration was monitored under physiologically relevant conditions (pH 7.4, 25 mM Tris buffer, 100 mM NaCl, 10% fetal bovine serum (FBS)) every fifteen minutes for two hours and then every hour up to 24 hours [Fig. 6]. The shoulder at 280 nm, attributable to the Tyr-HA, remains virtually constant for 12 hours. Between the 12 and 24 h time points the absorbance increases either due to slow binding to biomolecular species in the FBS or a gradual speciation change because of partial dissociation of Tyr-HA and compound aggregation as discussed in the analysis of the gel filtration chromatographic data (Supplementary Data). This FBS-induced aggregation will obviously affect the relaxivity, therefore contrast enhancement, achieved in vivo compared to those determined in vitro. No new absorbance shoulders or maxima appear during the spectral changes.

Figure 6.

The stability of 3.3 μM conjugate 2 at pH 7.4 in 10% FBS monitored by changes in its UV-vis spectrum. Inset: Absorbance change at 280 nm over time.

3.5. Relaxivity of complex 2

The longitudinal and transverse relaxivities, r₁ and r₂, of 2, determined in a similar manner as those of 1 (Fig. S11, S12) are listed in Table 1. Its r₁ values are low compared to 1 significantly and much lower compared to Gd-based agents [57]. This is tentatively attributed to the trapping of Fe₈ units in the interior of the Fe₈-Tyr-HA conjugate, limiting the exchange of water molecules in its solvation sphere with bulk water. In contrast, the r₂ values of 2 are higher than both 1 and Gd-agents, showing a stronger magnetic field dependence than 1. In fact, the r₂ values reached at higher field place 2 in the regime of nanoparticulate SPIOs [58]. The spectacular difference between the r₂ values of 1 and 2 is attributed to the slower tumbling of the latter due to its polymeric nature (high molecular weight).

3.6. In vitro cytotoxicity evaluation of Fe₈-Tyr-HA (2)

The cellular cytotoxicity of conjugate 2 was examined by evaluating its ability to induce cell death in A-549 cells, which overexpress CD44, and COS-7 cells, which do not. The determination of in vitro cytotoxicity was designed to be a first-order assessment of potential toxicity anticipated in vivo when used as a CA. Tumors may have an enhanced permeability and retention (EPR) effect and are thereby able to uptake significantly larger solutes, and retain them due to dysfunctional lymphatic drainage, than most normal tissues in vivo [70, 71]. The HA moiety on conjugate 2 would also facilitate higher CD44-mediated interaction with the conjugate. Nonetheless, using the MTS assay, conjugate 2 even at as high a concentration as 30 μM demonstrated no cytotoxicity against either cell line (Fig. 7), an encouragement for in vivo applications of the compound because it was not expected to be significantly toxic.

Figure 7.

The cell viability of A-549 and COS-7 was examined in the presence of 30 μM conjugate 2, 150 µM linoleic acid (LinOA) as a positive control, and 1 mg/mL α-lactalbumin (LA) as a negative control.

3.7. In vivo evaluation Fe₈-Tyr-HA (2)

In vivo pilot experiments in four mice demonstrated that the lead HA-Tyr-Fe₈ conjugate was physiologically compatible in the dose ranges administered (consistent with the cell viability studies), although these doses were largely limited by solubility and considerations of optimal injection volumes. A nude mouse bearing orthotopic human MDA-MB-468 breast tumor xenografts [53, 54], which overexpress CD44, were scanned using T₂-weighted and T₂*-weighted anatomic imaging sequences before and after administration of conjugate 2. Figure 8 illustrates clear visibility of CA following direct (intratumoral) injection.

Figure 8.

Pre- (top panels) and post-injection (bottom panels) MR scans (4.7 T) of a mouse bearing MDA-MB-468 human breast tumors implanted bilaterally in mammary fatpads; the tumor on the left was injected intra-tumorally with ~70 μg of Fe₈-Tyr-HA. Left panels-T₂; right panels-T₂*. The arrow in the upper left panel indicates the presence of overlaying ultrasound gel; the arrow in the upper right panel indicates one of the bilateral tumors; arrow in lower right panel indicates the injection site of the CA.

Mice were also scanned dynamically using a fast spoiled gradient echo sequence following i.v. injection of the Fe₈-Tyr-HA to both confirm injection and to initially assess plasma clearance pharmacokinetics. Clear negative contrast in circulating blood viewed in the inferior vena cava (Fig. S13), appeared rapidly (~1 min) post-injection, with strong signal attenuation still persisting at ~50% of maximum attenuation for more than 10 min following injection.

Next, having established the feasibility of detecting signal attenuation following systemic administration of the Fe₈-Tyr-HA, a mouse also bearing an MDA-MB-468 breast xenograft was injected i.v. with the CA via a tail vein catheter. A comparison of T₂-weighted scans (Fig. 9), acquired before and ~ 45 min after injection of 7 mg of compound 2 in 200 μL PBS, revealed heterogeneous signal change within the tumor that was already evident by 15 minutes after injection (not shown) and that was increasingly conspicuous with time over this window of observations. The observation that the CA was largely cleared from the blood stream while the tumor contrast enhancement was maximized, corroborated the conclusion that the Fe₈-Tyr-HA targeted the tumor.

Figure 9.

T₂-weighted scans (4.7 T) of MDA-MB-468-bearing nude mouse before (left) and ~45 min after (right) i.v. injection of 7 mg of Fe₈-Tyr-HA. Setup was essentially identical to that for the experiment shown in Figure 8; blue arrow in left panel indicates ultrasound gel; red arrow in right panel indicates visual evidence of tumor accumulation of the CA.

4. Conclusions

Mononuclear paramagnetic metal-ions with no magnetic anisotropy and large energy separation between spin-states, such as Gd³⁺, Mn²⁺ and Fe³⁺, have long electron-spin relaxation times (T_1e) and are therefore ideal proton relaxation enhancers. On the other hand, paramagnetic exchange-coupled polynuclear complexes, with their closer-spaced spin-states, show relatively faster electron-spin relaxation and are not optimal promoters of proton-spin relaxation. Not surprisingly, the former ions have attracted most of the attention with regard to their potential in MRI contrast enhancement. Complex 1, which gave well-defined ¹H-NMR spectra and therefore is believed to have a T_1e < 10^–8 s, falls into the latter category. However, polynuclear complexes may have other advantages over mononuclear ones, such as higher spin-states populated at ambient temperature and chemical stability under physiological conditions, while still able to promote proton nuclear relaxation with an observable MRI contrast.

The clinically approved Gd-complexes show a decrease in relaxivity with increasing magnetic field strength in the 1.5 – 7.0 T range, which is a potential drawback for their application in even larger fields, as the newer 3 T and 4.5 T clinical MRI scanners are becoming more common in practice [72, 73]. The scenario is even worse in case of small animal scanners that usually operate at ultrahigh 4.7 T - 7 T, or even higher magnetic fields [74, 75]. On the contrary, an opposite trend is observed in the Mn(II)-oxo clusters [76]. The [Fe₈]-based CAs described here show the same trend of improving contrast enhancement with increasing magnetic field, which, considering the current trend towards higher-field clinical MRI scanners, places them in an advantageous position for future development. The low r₁ values of 1 and 2 have been tentatively attributed to clustering of [Fe₈] units, preventing the close approach of H₂O molecules to interior Fe-ions. In future development, this can be remedied by increasing the steric bulk of new [Fe₈] variants that will prevent clustering. Because the Fe₈ complexes are amenable to chemical manipulation, which can modify their chemical, magnetic, and pharmacokinetic properties, we anticipate that improved variants of 1 will be available soon. It should be emphasized that the octanuclear complex 1 is well-defined molecule (e.g., not an iron oxide nanoparticle), and therefore not comparable with SPIOs in a straightforward manner. Nevertheless, the r₂ values of both 1 and 2 are in the same range as SPIOs.

With these thoughts, we have proceeded with the development of physiologically compatible conjugates of [Fe₈] clusters with targeting ligands. The HA conjugates described here represent the first effort in that direction. The 5 kDa HA of 2 is at the low end of the HA molecular weight range of 5 – 150 kDa. Because the longer HA chains are known to have higher affinity for CD44, we project that analogues of 2 employing larger HAs will allow contrast enhancement of tumors with lower doses of the administered CA. This may be further improved by optimization of the [Fe₈]/HA ratio in the conjugate. The small extent of Fe₈ dissociation from Tyr-HA, currently a drawback, can be remedied by increasing the percentage of Fe-coordinating Tyr groups in the Tyr-HA vector.

In summary, we have prepared an HA conjugate of the paramagnetic [Fe₈] MRI module and shown that the T₂ relaxation properties of this conjugate are greatly improved compared to those of the unconjugated module. It is expected that further optimization of the conjugate composition -- e.g., by varying the length of the HA chain, degree of functionalization, and steric bulk of the [Fe₈] units – will lead to CAs with superior contrast enhancing properties. Work in that direction, along with detailed biological evaluation, is in progress.

5. Abbreviations

CA

contrast agent

CPMG

Carr-Purcell-Meiboom-Gill

DLS

dynamic light scattering

DMEM

dulbecco’s modified eagle medium

EDS

energy dispersive spectroscopy

EPR

enhanced permeability and retention

FBS

fetal bovine serum

HA

hyaluronic acid

IACUC

Institutional Animal Care and Use Committee

MALDI-TOF

matrix assisted laser desorption ionization-time-of-flight

MRI

magnetic resonance imaging

MTS

3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium

NSF

nephrogenic systemic fibrosis

PBS

phosphate buffered saline

PDI

polydispersity index

PMS

phenazine methosulfate

PRE

paramagnetic resonance enhancement

ROI

region of interest

SAIF

small animal imaging facility

SPION

superparamagnetic iron oxide nanoparticle

TCA

trifluoroacetic acid

TE

echo time

TEM

transmission electron microscopy

TR

repetition time

Tyr

tyramine