Organic Solvent-Free, One-Step Engineering of Graphene-Based Magnetic-Responsive Hybrids Using Design of Experiment-Driven Mechanochemistry.

In this study, we propose an organic solvent-free, one-step mechanochemistry approach to engineer water-dispersible graphene oxide/superparamagnetic iron oxide (GO/SPIOs) hybrids, for biomedical applications. Although mechanochemistry has been proposed in the graphene field for applications such as drug loading, exfoliation or polymer-composite formation, this is the first study to report mechanochemistry for preparation of GO/SPIOs hybrids. The statistical design of experiment (DoE) was employed to control the process parameters. DoE has been used to control formulation processes of other types of nanomaterials. The implementation of DoE for controlling the formulation processes of graphene-based nanomaterials is, however, novel. DoE approach could be of advantage as one can tailor GO-based hybrids of predicted yields and compositions. Hybrids were characterized by TEM, AFM FT-IR, Raman spectroscopy, and TGA. The dose-response magnetic resonance (MR) properties were confirmed by MR imaging of phantoms. The biocompatibility of the hybrids with A549 and J774 cell lines was confirmed by the modified LDH assay.

Graphene is a single layer carbon sheet (∼1 nm/layer) with sp2 hybridization.[1] When oxidized to graphene oxide (GO), the material becomes water dispersible, which is a favorable property particularly in biomedicine. The nanometer-range dimension and large surface areas of GO provide great potential for targeted drug delivery and tissue engineering.[2,3] GO can also be covalently or noncovalently functionalized to provide additional features.[4] For example, GO hybridized with superparamagnetic iron oxide nanoparticles (SPIOs) has been reported as a theranostic platform (a single therapeutic and diagnostic system),[5] as a magnetic resonance imaging (MRI)/photoacoustic (PA) multimodal imaging system,[6] and as magnetically targeted nanocarriers capable of localized hyperthermia.[7]

Traditionally, the hybridization of GO with iron oxide nanoparticles has been reported using in situ synthesis,[8] direct graphitization of oleic acid coated iron oxide nanoparticles,[9] precipitation methods,[10−12] solvent-thermal method by autoclaving at 180 °C for 4–16 h,[13] direct mixing with heating at 98 °C,[14] and microwave-heated synthesis.[15] However, all of these methods involve the use of multistep processing,[8] organic solvents,[8,12−15] or external-heating.[8−10,13−15] An organic solvent-free, one-step process for formation of GO/SPIOs is still to be found. Mechanochemistry, however, has been proposed in the graphene field to exfoliate graphite with[16] or without drug loading,[17−21] to synthesize GO from graphite via oxidation reactions,[22,23] to functionalize graphene,[24−28] or to synthesize various graphene/graphite-polymer composites.[29−33] Mechanical force provides an alternative source of energy to control chemical reactivity that is usually achieved by thermal and chemical treatment.[34] Herein, we propose a one-step mechanochemistry approach to produce water-dispersible GO/SPIO hybrids, using a high-speed vibrational ball-mill. The process is deemed superior to previously reported methods, as it is simple and is organic solvent-free. Furthermore, no heating step is required. It is hypothesized in this study that the mechanical force, an alternative source to heat, can enhance the interaction between oleic acid coated SPIOs and GO, forming water-dispersible hybrids. In brief, freeze-dried GO and oleic acid coated hydrophobic SPIOs were loaded into the milling jar, as dry powders, and dry milled for 1–2 h at 10–20 Hz. The control over the mechanical force and process parameters was achieved by implementation of the statistical design of experiment (DoE) (Figure A), as will be described later. GO or SPIOs were milled alone as controls. Oleic acid coated SPIOs, milled in absence of GO remained hydrophobic and sedimented, from aqueous dispersions, upon standing. GO remained water-dispersible before and after ball milling and did not respond to the magnet. The water-dispersible hybrids were separated from hydrophobic SPIOs and GO, by leaving the dispersion to stand for 2 min, followed by magnetic separation, respectively, as described in the methods section. Figure B shows the black dispersion of ball-milled GO/SPIO hybrids, which turned transparent within 30 s of exposure to the magnet, indicating the magnetic responsiveness of the hybrid.

Figure 1

Engineering graphene-based magnetic responsive hybrids using mechanochemistry. (A) Schematic representation of GO/SPIO hybrid formulation. (B) Images of the water-dispersible GO/SPIO hybrid before and after exposure to a magnet (rated 0.42 T on its surface). (C) Representative thermogravimetric curves of GO, SPIOs, and GO/SPIO hybrid. The 3D response surface at center points (1.5 h) for (D) hybrid yield (MTotal) and (E) GO/SPIOs wt/wt % (RPFinal).

The total mass MTotal (SPIOs + GO) (mg) and the composition of the hybrid i.e. the final relative weight percentage of GO/SPIOs (RPFinal) were measured by thermogravimetric analysis (TGA). As illustrated in Figure C, GO decomposes completely at 800 °C, in compressed air. Oleic acid coated SPIOs demonstrated a weight loss of 31.48 ± 0.01% at 800 °C (n = 3), due to oleic acid combustion. Ball-milled GO/SPIO hybrids, however, showed a weight loss of 34–69% at 800 °C, indicating the presence of GO and SPIOs in the sample.

DoE constitutes an efficient procedure for planning experiments.[35] It offers the advantage of being able to investigate the effect of two or more factors, on a system, while keeping the number of runs minimal. As a result, DoE offers advantages of obtaining the same amount of information while keeping the number of experimental conditions minimal, without losing precision. A full 2-level factorial DoE can investigate the effect of three factors, i.e., variables, 2-factor interactions (2FI), and 3-factor interactions (3FI) on the responses, from only 8 experiments. Implementation of DoE for controlling the formulation processes of other types of nanocarriers such as liposomes,[36−38] nanosuspension,[39] and solid lipid nanoparticle[40] have been reported. There are no studies, however, which have employed this approach in processing graphene-based nanomaterials. To better understand the impact of formulation and process parameters, on GO/SPIO hybrid system formulation, DoE with a 23 (2-level 3 factors) full factorial design with replicated center points was introduced.[41] Three factors were included in the design: milling frequency in Hz (A), milling time in h (B), and the initial relative weight percentage of GO/SPIOs (RPInitial), being loaded into the milling jar (C). Actual values of each factor were represented in coded pattern, i.e., +1 (high level) and −1 (low level) (Table S1 in the Supporting Information). Responses include the total weight of the hybrids in mg (MTotal) and RPFinal. MTotal represents the yield whereas RPFinal reflects the content of SPIOs within the hybrid. The design layout and response values are summarized in Table .

Table 1

Full Two-Level, Three-Factor Factorial Design

	factor			response
std	Aa (Hz)	Bb (h)	Cc (%)	MTotald (mg)	RPFinale (%)
1	10	1	25	107.10	9.77
2	20	1	25	77.53	10.41
3	10	2	25	125.71	8.63
4	20	2	25	95.83	14.04
5	10	1	75	44.50	32.60
6	20	1	75	20.46	61.42
7	10	2	75	50.15	62.64
8	20	2	75	26.82	81.74
9	15	1.5	50	66.98	24.16
10	15	1.5	50	64.13	38.82
11	15	1.5	50	77.19	25.94

Milling frequency in Hz.

Milling time in h.

Initial GO/SPIOs wt/wt %.

Total hybrid weight in mg.

Final GO/SPIOs wt/wt % in the hybrid.

The effect of factors, 2FI and 3FI on MTotal and RPFinal were calculated using eq S1 (Figure S1 and Table S2 in the Supporting Information). Detailed information on the effect evaluations and the establishment of predictive equations are described in the Supporting Information, methods and Figures S2–S5.

The final fitted model for the MTotal with the coded factors of milling frequency (A), milling time (B), and RPInitial (C) is shown in eq . A good correlation was obtained between the observed and predicted values as indicated by the R2 value of 0.9795, adjusted R2 value of 0.9707 and predicted R2 value of 0.9578. In brief, milling frequency, milling time and the RPInitial significantly influenced the response MTotal (p-value <0.05). RPInitial (C) was the dominating factor toward MTotal, as it exhibited the largest coefficient compared to A and B. Higher milling frequency (A) and higher RPInitial (C) resulted in a lower MTotal, as indicated by their negative coefficients. Longer milling time (B), however, increased MTotal, as indicated by its positive coefficient. The predictive equation of coded and actual factors is represented in eq . A comparison of the predicted and actual (experimental) values for MTotal is shown in Figure S3 in the Supporting Information. A 3D factorial response surface was produced for MTotal according to the predictive eqs and 2 and is shown in Figure D.The final fitted model for the RPFinal with the coded factors of milling frequency (A), milling time (B), and RPInitial (C) is shown in eq . A good correlation was obtained between the observed and predicted values as indicated by the R2 value of 0.9438, adjusted R2 value of 0.9197 and predicted R2 value of 0.8662. In brief, higher milling frequency, higher RPInitial and longer milling time all resulted in a higher RPFinal as reflected by their positive coefficient. The predictive equation of coded and actual factors is represented in eq . A comparison of the predicted and actual (experimental) values for RPFinal is shown in Figure S5 in the Supporting Information. A 3D factorial response surface was produced for RPFinal according to the predictive eqs and 4 and is shown in Figure E.Using DoE, it was established that a higher initial SPIOs loading, i.e., lower RPInitial value, and longer milling time resulted in hybrids of higher SPIOs content, i.e., lower RPFinal value (Figure E). Higher milling frequency adversely affected both MTotal and SPIOs content in the hybrids (Figure D, E). From these studies, it was concluded that the DoE approach could be reliably implemented in the process of formulating GO-SPIOs hybrids.

Morphological examination of GO and GO/SPIO hybrids was performed using transmission electron microscopy (TEM) and atomic force microscopy (AFM). The TEM and AFM images of GO sheets are shown in Figures S6 and S7 in the Supportng Information, respectively. The height of the GO sheets, as shown with AFM, is ∼1 nm confirming the formation of single-layer GO sheets. The size (surface area) of GO was generally small, with a medium area of 181.35 nm2 and an average of 441 nm2 (n = 438) (Figure S7 in the Supporting Information). TEM images of the GO/SPIO hybrids showed electron-dense structures, typical of SPIOs, associated with GO sheets confirming the formation of GO/SPIO hybrids (Figure A). AFM images showed height between 10 and 20 nm, which agreed with the theoretical size of the SPIOs (Figure B and Figure S8 in the Supporting Information). Some larger SPIOs aggregates could be also found (Figure S9 in the Supporting Information). This was expected as the milling processing occurred under dry conditions. The breaking of the SPIOs aggregates is dependent on the process-capability of the ball-mill, which is one of the limitations of this approach.

Figure 2

Characterization of GO/SPIOs hybrid. (A 1–3) TEM images of GO/SPIO hybrids on lacey carbon supports are shown. SPIOs were found as scattered black spots due to their higher electron density. (A 4–6) The 4× magnification of selected areas from A 1–3. (B) 2D and 3D AFM height images are shown. The height of the SPIOs is around 10–20 nm. (C) IR and Raman spectra of starting materials and the hybrid.

In our study, stabilization of GO-SPIOs hybrids, upon ball-milling, could be a result of enhanced hydrophobic or π–π interactions between the oleic acid coated SPIOs and aromatic rings of GO. Alternatively, GO’s oxygen-containing functional groups, i.e., COOH could interact with SPIOs, mimicking the ligand exchange process. Mechanical force has been reported to exchange the ligand on oleic acid coated iron oxide nanoparticles with poly(acrylic acid), polyethylenimine, and glutathione.[42] A slight to moderate increase in temperatures inside the milling jar may have occurred. Temperatures, however, remained lower (<40 °C) than those reported in previous studies (>95 °C), used to formulate GO/SPIOs hybrids.[8−10,13,14] Avoiding high-temperature is of critical importance when conjugating biomolecules i.e., targeting peptides or antibodies.

IR and Raman spectra of GO, SPIOs and the hybrids are shown in Figure C. The C–H stretching vibrations, at 2920 and 2851 cm–1, corresponding to the oleic acid coating can be seen in both SPIOs and GO/SPIOs. The presence of oleic acid in GO/SPIO hybrids may have facilitated flocculation overtime (Figure B, left panel). Future work could involve PEGylation of GO/SPIO hybrid to improve physical stability. Characteristic D and G bands can be seen in GO and GO/SPIOs hybrids with ID/IG peak-height ratios of 1.30 ± 0.04 and 1.27 ± 0.04, respectively (p > 0.05, n = 3). Both IR and Raman spectra confirmed the formation of the hybrid. In addition to mechanical forces, it is anticipated that the mild milling may result in local heat generation, at the nanointerface, which is required to enhance hydrophobic or π–π interactions forming the hybrid. The weak but characteristic Raman peaks of γ-Fe2O3 (maghemite) can be seen at 327, 383, 505, and 680–720 cm–1 (Figure C Raman spectra). Under the same laser power (1%), these peaks were difficult to be visualized in GO/SPIOs hybrids, as shown in Figure S10 in the Supporting Information.

When samples were preheated with higher power laser (50%) during Raman measurement, the appearance of 5 peaks, characteristic of α-Fe2O3 at 226, 294, 407, 497, 613 cm–1, could be seen in GO/SPIO hybrids (Figure C and the Raman spectra in FIgure S10 in the Supporting Information). It is known that Fe3O4 and γ-Fe2O3 (maghemite) can undergo phase-transformation to α-Fe2O3 (hematite), which exhibits different Raman spectrum. This transformation can be artificially triggered when heated by laser.[43] Raman spectroscopy did not only confirm the formation of the hybrid but also suggested that if heat was generated during the milling process, then the amount produced was not sufficient to induce transformation of γ-Fe2O3 to α-Fe2O3. The Raman spectra of SPIOs and GO/SPIO hybrids measured using different laser power are shown in Figure S-10.

In vitro toxicity of the GO/SPIO hybrids was also evaluated using A549 cells (adenocarcinoma human alveolar basal epithelial cells) and J774 cells (murine macrophages). Optical microscopy images of A549 and J774 cells, incubated with GO/SPIO hybrids, were captured after 24 and 72 h. GO/SPIO hybrids are expressed based on SPIOs concentration. Cells incubated with GO/SPIO hybrids showed normal morphology at all concentrations tested while cells treated with dimethyl sulfoxide (DMSO), used as a positive control, appeared unhealthy and detached from the plate. The extent of GO/SPIO hybrids taken up by cells was concentration- and time-dependent (Figure S11 in the Supporting Information).

The modified lactate dehydrogenase (mLDH) assay was used to assess the cytotoxicity of GO/SPIO hybrids on A549 cells and J774 cells using the method described in the methods in the Supporting Information.[44] Cells were incubated with the hybrids at 10, 25, 50, and 100 μg/mL for 24 and 72 h. DMSO resulted in viability of 2.2% and 1.80% for A549 and J774 cells, respectively. The cytotoxicity result from the mLDH assay is shown in Figure . No statistically significant effect on cell viability was observed in all treated groups up to 100 μg/mL for 72 h. Reported studies in the literature showed no significant toxicity for SPIOs or GO up to 100 μg/mL.[45,46]

Figure 3

Modified LDH assay of A549 and J774 cells incubated with GO/SPIO hybrids. The A549 and J774 cells were incubated with GO/SPIO hybrids at SPIOs concentrations of 10, 25, 50, and 100 μg/mL for 24 and 72 h. The toxicity was assessed using the modified LDH assay. No significant effect on cell viability was observed.

To demonstrate the imaging potential of the GO-SPIOs hybrids, their MR contrast properties were determined by an in vitro MR phantom study using a 7T MR scanner. The GO/SPIOs hybrid samples were suspended in 1% agarose to mimic the tissue-like environment in vivo.[47] Their transverse (T2) relaxation times were measured at increasing Fe concentrations and the R2 (1/T2) values were calculated. The GO/SPIOs hybrids exhibited a concentration-dependent darkening effect with an r2 relaxivity of 144 mMs–1 (Figure A, B). The obtained values are higher than some of those GO-based MRI imaging nanocomposites reported in literature.[7,48] The obtained r2 relaxivity was comparable to values reported for the commercially approved MRI contrast agent Feridex (127.47 and 108 mM s–1)[49,50] Regional signal reductions were seen with the GO/SPIO hybrids and became more obvious at higher Fe concentrations. The heterogeneous hypointense signals are likely to be due to the inhomogeneous presence of iron oxide in the hybrids as observed in TEM and AFM images. The T2-weighted contrast enhancement, combined with water-dispersibility, made the GO-SPIOs as a potential MR imaging agent in the future.

Figure 4

Phantom MR imaging of GO/SPIO hybrids. (A) T2-weighted phantom MR imaging of GO/SPIO hybrids at increasing Fe concentrations. (B) R2 relaxation rate analysis as a function of Fe concentration.

In conclusion, engineering GO/SPIO hybrids using organic solvent-free mechanochemistry is reported for the first time. The mechanochemistry approach replaced the traditional solvent-based and/or external heat-based treatment for GO/SPIO hybridization. The formulation process was rendered eco-friendly, i.e., organic solvent-free without significant heating being produced, as confirmed by Raman spectroscopy. The preparation time was dramatically reduced to a maximum of 1–2 h. The purification step is rapid, simple and only requires a magnet. The resulting GO/SPIO hybrids showed no significant toxicity on A549 and J774 cells at the SPIOs concentrations up to 100 μg/mL and 72 h. Hybrids produced were water-dispersible and showed dose-dependent MR properties. The mechanochemical approach was combined with the DoE, which has been used for the first time, to design GO-based hybrids of predicted yields and compositions. Simple preparative methods combined with predicted design of hybrid nanomaterial provides the foundation for future applications with quality by design (QbD), which is important for clinically relevant materials.[51] This approach could be extended to prepare other types of nanohybrids after careful consideration.