Fe₅C₂ Nanoparticles with High MRI Contrast Enhancement for Tumor Imaging

Stable colloidal nanoparticles with strong magnetization have attracted much attention for their great potential in biomedical applications.^[1] In particular, they can be used as contrast probes in magnetic resonance imaging (MRI) by inducing hypo-intensities on T₂ and T₂*-weighted MRI maps. ^[2] So far, the most commonly used MRI contrast agents are iron oxide nanoparticles. However, due to moderate magnetization (∼8 0 emu/g for Fe₃O₄ and ∼70 emu/g for Fe₂O₃), ^[1a,3] iron oxide nanoparticles as MR probes suffer from suboptimal contrast enhancement. r₂ relaxivity, the measure of contrast ability, is ∼100 mM⁻¹s⁻¹ for Feridex^[1a], a clinically used iron oxide formula, and is ∼200 mM⁻¹s⁻¹ for iron oxide nanoparticles made from more advanced synthetic approaches. ^[4] To exceed this limit, efforts have been made to fabricate nanoscale particles with even higher magnetization. For instance, Fe,^[5] CoFe,^[6] Co,^[7] FePt,^[8] and SmCo₅^[9] nano-particles have been recently prepared by us and others via wet chemistry routes. Translation of these nanoparticles into biomedical practices, however, has encountered difficulties of various sorts. For instance, Fe, Co, and CoFe are very reactive at ambient conditions, and are susceptible to rapid oxidization and magnetism loss. ^[5–7] FePt and SmCo ₅ nanoparticles need to be annealed at high temperature to be granted with high magnetism. However, the annealing process is accompanied with severe particle aggregation. ^{[9a, 10]} More critically, most of these nanoparticles are associated with high toxicity that limits their wide use in biological settings. ^{[8b, 11]}

Iron carbides (Fe₃C, Fe₅C₂, and Fe₇C₃), used primarily in metallic alloys and hard coatings, are one of the most ancient materials in human history. Iron carbides are also known to possess appealing magnetic properties. Fe₅C₂ for instance, has a magnetic moment of ∼ 140 emu/g, ^[12] a value that is comparable to that of Co (∼1 60 emu/g) and two times higher than that of Fe₃O₄. Unlike Co or Fe, iron carbides boast good chemical stability, and hence are largely immune to oxidization-induced magnetization drop. Such strong and stable magnetism suggests their potential applications in biomedicine. The related exploits, however, have seldom been attempted. The main challenge comes from the difficulty of preparing nanoparticulate iron carbides. Conventionally, iron carbides are made at high-temperatures and in reducing conditions, ^[13] which are hard to replicate in a laboratory setting. There have been reports on utilizing laser pyrolysis ^[14] and high temperature sol-gel methods to prepare iron carbide nanoparticles. ^[13,15] The products, however, do not afford the size and/or colloidal stability that are desired for bio-medical applications.

A breakthrough by us recently allowed one to prepare Fe5C2 nanoparticles by a facile and mild wet-chemistry route.^[12] The products possess close-to-bulk magnetization, small and tunable particle size (<25 nm), narrow size distribution, and good colloidal stability. Following this success, we herein evaluated the potential of Fe₅C₂ nanoparticles as a MR contrast probe, which, to the best of our knowledge, is the first time. We found that Fe₅C₂ nanoparticles, after surface modification with phospholipids, could be stably dispersed in aqueous solutions (Scheme S1). The particles displayed an r₂ relaxivity of 464.02 mM⁻¹s⁻¹, which is among the highest of all MR probes reported. When coupled to a tumor targeting ligand, Fe₅C₂ nanoparticles were able to home to tumors and induce more prominent signal change than Fe₃O₄ nanoparticles. These observations strongly suggest the great promise of Fe₅C₂ in MR imaging as well as in other biomedical fields.

We followed our reported protocol to prepare Fe₅C₂ nanoparticles.^[12] Briefly, a mixture of octadecylamine and cetyltrimethyl ammonium bromide (CTAB) was rigorously stirred and degassed under argon. The solution was heated to 393 K, at which time Fe(CO) ₅ was injected. The resulting mixture was further heated to 453 K and kept at the temperature for 10 min. Finally, the temperature was increased to 623 K and kept for 10 min before cooling to room temperature. The nanoparticles were purified with ethanol and redispersed in hexane. Transmission electron microscope (TEM) analysis showed that the nanoparticles are comprised of a Fe₅C₂ core (∼19 nm) and a Fe₃O₄ shell of ∼2 nm^[12], and the overall particle size was around 23 nm ( Figure 1a). The magnetism of the particles was studied by a SQUID magnetometer. A saturation magnetization (Ms) of ∼125.4 emu/g was observed at room temperature (Figure 1b, black curve), which is close to the bulk materials’. This value is also among the highest reported with colloidally stable nanoparticles, compared to that of ∼7 0 emu/g for iron oxide nanoparticles,^[16] ∼100 emu/g for Fe nanoparticles,^[5] and ∼115 emu/g for Co nanoparticles.^[7] Notably, the particles showed no remnant magnetization or coercivity at room temperature, i.e. they are superparamagnetic. This ensures good particle dispersibility that is critical to their uses in a biological context. Another important feature is that the particles are very stable in the air. After one week aging in air the Fe₅C₂ nanoparticles showed an almost unchanged hysteresis loop (Figure 1b, red curve).

Figure 1.

a) TEM image of as-synthesized Fe₅C₂ nanoparticles in hexane. (b) Hysteresis loops of Fe₅C₂ nanoparticles that were as-synthesized (black curve) and after exposure to the air for one week (red curve). The two curves were almost identical, suggesting good stability of the nanoparticles. No remnant magnetization and coercivity were observed, indicating that they were superparamagnetic at room temperature. The Ms values were determined to be 125.4 and 122.6 emu/g, respectively, for as-synthesized Fe₅C₂ nanoparticles and those after one week aging. c) TEM image of DSPE-PEG-COOH coated Fe₅C₂ nanoparticles in water. The red arrows point to the DSPE-PEG-COOH coating that is manifested as dim halos surrounding the particles. d) The hydrodynamic sizes of Fe₅C ₂ nanoparticles are ∼22 and ∼35 nm, respectively, before and after the DSPE-PEG-COOH coating. The results correlate well with the TEM observations.

The as-synthesized Fe₅C₂ nanoparticles are coated with a layer of octadecylamine and are hydrophobic. To render them water soluble, we imparted a phospholipid derivative, 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N[carboxy(polyethylene glycol)-2000] (DSPE-PEG-COOH), onto the particle surface. DSPE-PEG-COOH is an amphiphilic compound. It immobilizes on the particle surface through hydrophobic-hydrophobic interactions between its alkyl chains and the octadecylamine coating. The PEGylated, hydrophilic section faces outward, interacting with the aqueous surroundings to afford the particles with good dispersibility. The particles are stable in water, PBS, and serum (Figure S1). Figure 1c is a TEM image of DSPE-PEG-COOH coated nanoparticles in water. There is a dim halo (pointed by red arrows) surrounding each nanoparticle, which is attributed to the phospholipid coating. The dynamic light scattering (DLS) results were well correlated with the TEM observations, finding average sizes of 22 and 35 nm (Figure 1d) for particles before and after the surface coating, respectively.

We next assessed the r₂ relaxivity of these phospholipid-coated Fe₅C₂ nanoparticles. Briefly, we suspended Fe₅C₂ nanoparticles at a concentration gradient (0.05, 0.1, 0.25, 0.5, and 1 μg Fe/mL) in 1% agarose gel. We then subjected these gel samples to MR scans on a 7.0 T magnet using T₂ -weighted fast spin-echo sequences of different TEs. As a comparison, Fe₃O₄ nanoparticles (from Ocean Nanotech, Springdale, AR) samples at the same Fe concentrations were also prepared and scanned. While both sets of samples showed concentration dependent signal drop (Figure 2a), Fe₅C₂ nanoparticles induced greater hypo-intensities at the same concentrations. The linear fitting shown in Figure 2b indicates that the r₂ relaxivity of the Fe₅C₂ nanoparticles is 464.02 mM⁻¹s⁻¹, which is ∼2.6 times higher than that of Fe₃O₄ nanoparticles (178.30 mM⁻¹s⁻¹). This value is among the highest of all the MR probes reported.

Figure 2.

a) Phantom studies with Fe₅C₂ and Fe₃O₄ nanoparticles dispersed in 1% agarose gel at different concentrations on a 7T magnet. Fe₅C₂ nanoparticles induced more significant signal drop than Fe₃O₄ nanoparticles at the same Fe concentrations. b) r₂ relaxivities derived from the imaging results in a). Fe₅C₂ nanoparticles have an r₂ of 464.02 mM⁻¹s⁻¹ ( R² = 0.9995) compared to that of 178.30 mM⁻¹s⁻¹ (R² = 0.9758) for Fe₃O₄ nanoparticles.

As a novel nanomaterial for bio-related applications, a primary concern is toxicity. This was studied by (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assays with U87MG (human glioblastoma) cells. Within the tested range (0–25 μg Fe/mL), U87MG cells retained over 80% viability over 24 h, indicating good tolerance to the Fe₅C₂ nanoparticles ( Figure 3a). The cytotoxicity was also studied by live/dead assays (Invitrogen), where green (calcein AM) and red (ethidium homodimer-1) fluorescence mark live cells and dead cells, respectively. The outcomes were consistent with the MTT results, finding no substantial cell death when the particle concentration was below 25 μg Fe/mL (Figure 2b). Notably, though a small amount of CTAB was used as a precursor in the synthesis, none remained on yielded Fe₅C₂ nanoparticles, as confirmed in our previous studies.^[12]

Figure 3.

a) MTT assays with Fe₅C₂ nanoparticles on U87MG cells. The cells retained over 80% viability in the tested concentration range (0 – 25 μg Fe/mL). b) Live/dead assays with Fe5C2 nanoparticles on U87MG cells. No red fluorescence (ethidium homodimer-1) was observed in the tested concentration range (0 – 25 μg Fe/mL). c) Cellular uptake studies with U87MG cells. The cells were incubated with either RGD-Fe₅C₂ or Fe₅C₂ nanoparticles (5 μg Fe/mL) for 1h, and then subjected to Prussian blue staining. Positive staining was found with almost all the cells when RGD-Fe₅C₂ nanoparticles were used. On the contrary, Fe₅C₂ nanoparticles showed little cellular uptake.

The multiple carboxyl groups on the particle surface allow for easy coupling with biomolecules. For a proof-of-concept study, we coupled a peptide, c(RGDyK), onto the Fe₅C₂ nanoparticles (Scheme S1). c(RGDyK) is an RGD derivative with high affinity towards integrin α_vβ ₃ that is overexpressed on tumor vasculature and tumor cells of various types.^[17] The coupling was mediated by 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) and sulfo-(N-hydroxysulfosuccinimide) (NHS), and the particles were collected and purified using a microcon centrifugal filter unit (YM-100, Millipore). The final products, c(RGDyK)-conjugated Fe₅C₂ (hereafter referred to as RGD-Fe₅C₂), were dispersed in PBS. No substantial aggregation was observed during the coupling and purification processes.

The targeting specificity was first assessed in vitro with U87MG cells, which are integrin α_vβ₃ positive. Briefly, RGD-Fe₅C₂ nanoparticles (5 αg Fe/mL) were incubated with fixed U87MG cells for 1 h. After removing the solution, the cells were washed with PBS and subjected to Prussian blue staining to reveal the Fe distribution. Almost all of the cells were positively stained, suggesting good particle uptake (Figure 3c). In contrast, when non-targeting Fe₅C₂ nano-particles (not conjugated with the c(RGDyK)) were used, there was a much lower level of cellular uptake. The results confirm that the particle targeting and uptake were mainly mediated by RGD-integrin interactions.

With the encouraging phantom and in vitro data, we next evaluated the particles in vivo in U87MG tumor bearing mice (n = 3). The animal model was established by subcutaneously injecting approximately 5 × 10⁶ U87MG cells into the right hind limb of each mouse. In vivo MR imaging was conducted when the tumor size reached about 200–500 mm³, at which time RGD-Fe₅C₂ or Fe₅C₂ nanoparticles were intravenously administered at a dose of 10 mg Fe/kg. A similar dose was used by us and others in previous investigations with Fe₃O₄ nanoparticles.^[18] For comparison, Fe₃O₄ nanoparticle (20 nm, from Ocean Nanotech) and c(RGDyK) conjugates (RGD-Fe₃O₄ were injected at the same dose.) T₂-weighted MR images were acquired before, and 4 and 24 hours after the particle injection ( Figure 4a). In the RGD-Fe₅C₂ group, significant signal drop in the tumor areas was observed at both 4 and 24 h time points, manifested as unevenly distributed black spots (pointed by red arrows). This distribution pattern is a reflection of the tumor heterogeneity, and is consistent with the observations made by us^[18b] and others^[19] previously. On the contrary, little signal change was visualized in the control group, where Fe₅C₂ nanoparticles were injected. RGD-Fe₃O₄ also induced hypointensities in tumors but the signal change was not as dramatic as that with RGD-Fe₅C₂. By choosing signals in the leg muscle as a reference, we compared the relative intensity change. Compared to RGD-Fe₃O₄, RGD-Fe₅C₂ induced 42.6% and 60.7% greater signal drop at 4 and 24 h time points, respectively. After the 24 h imaging, we sacrificed the animals and performed Prussian blue staining with the tumor sections. In parallel to the in vivo observations, we found many areas of positive staining in the RGD-Fe₅C₂ group, but few were present in the Fe₅C₂ group (Figure 4b). To further elucidate the targeting mechanism, we also performed Prussian blue and integrin β₃ double staining with tumors from the RGD-Fe₅C₂ group (Figure 4b). We observed overall good overlap between the two stainings, confirming that the tumor targeting was mediated by RGD-integrin interactions. The distribution of RGD-Fe₅C₂ nanoparticles among major organs was also studied by Positive Prussian blue staining, which found many iron deposits in the liver and spleen. This is mainly attributed to the uptake by the Kupffer cells and macrophages as part of the reticuloendothelial system (RES) intervention (Figure 4c), though body iron stores may have contributed to a certain level.^[18a] As expected, few particles were found in other major organs such as the lung, kidneys, and brain.

Figure 4.

a) MR imaging results. U87MG tumor bearing mice were intravenously injected with RGD-Fe₅C₂, RGD-Fe₃O₄, or Fe₅C₂ nanoparticles (10 mg Fe/kg). Scans were performed before, and 4 and 24 h after the administration. In both the RGD-Fe₅C₂ and RGD-Fe₃O₄ groups, black areas (in the RGD-Fe₅C₂ group highlighted by red arrows) were found distributed in the tumors (circled by yellow dotted lines). RGD-Fe₅C₂ induced more significant signal drop than RGD-Fe₃O₄ at both time points. b) Left column: Prussian blue staining with tumor tissues. Positive staining was found in the RGD-Fe₅C₂ group but not in the Fe₅C₂ group. Scale bars, 50 μm. Right column: Prussian blue and integrin β₃ double staining with the tumor tissues. Good overlap was observed in the RGD-Fe₅C₂ group, indicating that the tumor accumulation was mainly caused by RGD-integrin interactions. Scale bars, 10 μm. c) Prussian blue staining with tissue samples from the liver, spleen, lung, kidneys, and brain. Particles were accumulated in the liver and spleen but not in the other major organs. Scale bars, 10 μm.

Though an ancient material, the uses of Fe₅C₂ in a biological setting have seldom been exploited. The current study offers the first investigation in this regard. The in vitro, in vivo, and pathological studies suggest the great potential of Fe₅C₂ nanoparticles for bio-applications. Despite the promise, however, more studies are needed to fully understand and optimize this new type of biomaterial. First, systematic toxicity studies are needed. Though relative low cytotoxicity was observed in the current study, the metabolism and long-term impact of Fe₅C₂ nanoparticles within a biological system is largely unknown and needs thorough investigations. In one pilot study, we incubated Fe₅C₂ nanoparticles in a pH 5.0 buffer solution. After 72 h, it was found that most of the nanoparticles were partially or completely degraded (Figure S2). It is therefore believed that, similar to iron oxides, Fe₅C₂ can be degraded in an acidic lysosome environment, releasing Fe to join the body iron stores. Second, it is worthwhile to explore the impact of particle size on the magnetic properties, cellular uptake, in vivo circulation half-lives, bio-distribution, and targeting selectivity of Fe₅C₂ nanoparticles. An increased particle size may lead to higher magnetization and r₂ relaxivity, but may also cause a shorter circulation time and a higher level of RES uptake. Balance has to be struck to achieve the best performance. Third, the surface coating may affect the pharmacokinetics of nanoparticles and is worthy of optimization. In the current study, a PEGylated phospholipid was used as the coating. Other coating materials such as silica,^[20] dopamine derivatives,^[5,21] and tri-block co-polymers^[18b] are also expected to be able to modify Fe₅C₂ nanoparticles. Fourth, as a proof-of-concept investigation, this study chose RGD as a tumor targeting ligand. In future investigations, it is important to evaluate performance of Fe₅C₂ nanoparticles when conjugated with different targeting motifs. Ultimately, there are many other potential applications with Fe₅C₂ nanoparticles. These include liver and lymph node imaging for which iron oxide nanoparticles are dominantly used. Moreover, with a high magnetism, Fe₅C₂ nanoparticles also hold promise in magnetic separation/purification, drug delivery, gene transfection, cell tracking, and magnetic hyperthermia. ^[1b,22] The related investigations are underway.