Synthesis of Carbon Onion and Its Application as a Porous Carrier for Amorphous Drug Delivery

Given the great potential of porous carrier-based drug delivery for stabilising the amorphous form of drugs and enhancing dissolution profiles, this work is focussed on the synthesis and application of carbon onion or onion-like carbon (OLC) as a porous carrier for oral amorphous drug delivery, using paracetamol (PA) and ibuprofen (IBU) as model drugs. Annealing of nanodiamonds at 1100 ◦C produced OLC with a diamond core that exhibited low cytotoxicity on Caco-2 cells. Solution adsorption followed by centrifugation was used for drug loading and results indicated that the initial concentration of drug in the loading solution needs to be kept below 11.5% PA and 20.7% IBU to achieve complete amorphous loading. Also, no chemical interactions between the drug and OLC could be detected, indicating the safety of loading into OLC without changing the chemical nature of the drug. Drug release was complete in the presence of sodium dodecyl sulphate (SDS) and was faster compared to the pure crystalline drug, indicating the potential of OLC as an amorphous drug carrier.


Introduction
Poor solubility and dissolution of drugs have necessitated the formulation of drugs with more than the required dose, leading to economic wastage and an increased risk of toxicity [1,2]. Hence, it is very important for pharmaceutical scientists to develop effective techniques to improve the solubility and dissolution behaviour of poorly water-soluble drugs.
Several studies have been performed to enhance solubility of drugs and, in recent years, adsorption of drugs onto porous materials/adsorbents to produce stabilised amorphous drugs has attracted increasing interest [3][4][5]. Unlike crystalline forms that have a long range molecular order, amorphous forms of drugs lack three-dimensional molecular order, resulting in higher solubility,. However, amorphous forms have poor stability and often tend to convert back to crystalline forms, which is a major obstacle to formulation [6]. Loading of drugs into the pores of a carrier via adsorption restricts the crystallisation of the drug, resulting in a stable amorphous form. In addition to stabilisation, adsorption of drugs onto porous carrier increases the effective surface area of the drug that is in contact with the dissolution media, resulting in higher solubility [7,8]. Various inorganic and organic nanocarriers with suitable porosity were investigated. From organic nanocarriers, cubosomes, hexosomes and spongosomes are interesting and offer high internal channels and porosity to load with amorphous drugs [9].

Characterisation of Drug Loaded Carbon Onion Complex
To investigate the solid state characteristics of the drug in drug/OLC complex, XRD patterns were recorded on a D8 ADVANCE diffractometer (Bruker, Billerica, MA, USA) in the angular range of 10 • to 50 • (2θ) with a step size of 0.02 • , using a Cu-Kα source operated at 30 kV and 30 mA.
The presence of crystalline drug can be detected from the melting peak of DSC curves. DSC analysis was performed using a Q100 DSC (TA Instruments, Leatherhead, UK). Sample (2-3 mg) was transferred to a Tzero aluminium pan and temperature scan was performed at a heating rate of 10 • C/min under nitrogen gas.
FTIR was used to investigate the surface chemistry of OLC and to identify possible interactions between the drug and OLC. Studies were performed using a Nicolet iS5 spectrometer (Fisher Scientific, Loughborough, UK) equipped with an iD5 ATR accessory with a laminated diamond crystal at an angle of incidence of 42 • . The spectra were obtained in the range of 500-4000 cm −1 (wavenumber) at a spatial resolution of 4 cm −1 and were an average of 16 scans.
The porosity of the carrier before and after drug loading was investigated by nitrogen adsorption technique using a Micrometrics Accelerated Surface Area and Porosimetry (ASAP) 2420 system (Micrometrics, Norcross, GA, USA). Adsorption-desorption isotherms were obtained at 77 K in the relative pressure range between 0.01 and 0.9. Prior to analysis, all samples were degassed at 40 • C for 24 h. Pore size distribution was computed with the ASAP 2420 software using non-local density functional theory (NLDFT) model assuming slit-shaped pores

Drug Release Studies
Drug release studies were performed at 37 • C using a USP Type II dissolution apparatus (Erweka GmbH, Langen, Germany). Sample powders were filled in hard gelatin capsules (size 000) prior to analysis and 900 mL of 0.1M sodium phosphate buffer of pH 5.8 and pH 7.2 was used as dissolution media for paracetamol and ibuprofen, respectively. Dissolution was performed at paddle stirring speed of 100 rpm with and without 1% w/v sodium dodecyl sulfate (SDS) added to the medium. 5 mL samples were withdrawn at specific time intervals and replaced with fresh dissolution medium. The withdrawn samples were filtered using 0.2 µm syringe filters and analysed using UV spectrophotometry (Jenway) at wavelengths of 257 nm and 264 nm for paracetamol and ibuprofen, respectively.

Statistical Analysis
Statistical analysis was carried out on the data using GraphPad Prism software (Version 6.0 for Windows, GraphPad Software, San Diego, CA, USA). Statistically significant differences were denoted for p values of less than 0.05.

Surface Morphology
SEM pictures ( Figure S1) revealed a significant difference in the morphology of OLC compared to pristine nanodiamonds. OLC showed tightly bound aggregates reaching several micrometers in size, compared to NDs showing loosely bound agglomerates. The high agglomerating character of NDs (due to the presence of surface functional groups) results in the formation of OLC aggregates, thereby leading to aggregate induced porosity [13].

Surface Characterisation
The Raman spectrum of ND ( Figure S2 and Table S1) showed two dominant peaks: one at 1325 cm −1 , corresponding to the sp 3 diamond bonds (F 2g ) with a small background of sp 2 disorder peak (D band), and the second at 1595 cm −1 , corresponding to sp 2 dimer bonds (G band), indicating the presence of graphitic carbon fragments, characteristic of ND produced from detonation process [35,36]. Additionally, the relative broadening of the diamond peak is typical of nanocrystalline diamonds [31].
Post annealing, the Raman spectrum consists of a broad D band at 1345 cm −1 , attributed to an increase in the sp 2 disorder with a background of sp 3 diamond, indicating the transformation of diamond to graphite [37]. This position is also very near the one recorded in previous studies for high purity OLC [31]. Moreover, the G band situated at 1587cm −1 corresponds to the optical phonon mode of the E 2g symmetry in graphite and is known as the tangential mode for the OLC. The FWHM of G band decreased after annealing, suggesting the formation of graphitic shells [38][39][40].
HRTEM images (Figure 1) of ND show a diamond core with a typical lattice spacing of~2 Å {111} [41], surrounded by amorphous carbon. Images of OLC show the presence of a diamond core surrounded by 3-5 graphitic shells, with a typical lattice spacing of~3 Å {002} [42], indicating the transformation of diamond to graphite, further supporting the data from Raman studies. The Raman spectrum of ND ( Figure S2 and Table S1) showed two dominant peaks: one at 1325 cm −1 , corresponding to the sp 3 diamond bonds (F 2g ) with a small background of sp 2 disorder peak (D band), and the second at 1595 cm −1 , corresponding to sp 2 dimer bonds (G band), indicating the presence of graphitic carbon fragments, characteristic of ND produced from detonation process [35,36]. Additionally, the relative broadening of the diamond peak is typical of nanocrystalline diamonds [31].
Post annealing, the Raman spectrum consists of a broad D band at 1345 cm −1 , attributed to an increase in the sp 2 disorder with a background of sp 3 diamond, indicating the transformation of diamond to graphite [37]. This position is also very near the one recorded in previous studies for high purity OLC [31]. Moreover, the G band situated at 1587cm −1 corresponds to the optical phonon mode of the E 2g symmetry in graphite and is known as the tangential mode for the OLC. The FWHM of G band decreased after annealing, suggesting the formation of graphitic shells [38][39][40].
HRTEM images (Figure 1) of ND show a diamond core with a typical lattice spacing of ~ 2 Å {111} [41], surrounded by amorphous carbon. Images of OLC show the presence of a diamond core surrounded by 3-5 graphitic shells, with a typical lattice spacing of ~ 3 Å {002} [42], indicating the transformation of diamond to graphite, further supporting the data from Raman studies.
Furthermore, the electronic microscopic observations of annealed ND do not show any evidence of any large graphitic structures, such as graphitic ribbons, which corroborates the findings of the Raman spectrum and is in accordance with the selected annealing conditions [21,31,32].

OLC-1100
OLC-1100 Furthermore, the electronic microscopic observations of annealed ND do not show any evidence of any large graphitic structures, such as graphitic ribbons, which corroborates the findings of the Raman spectrum and is in accordance with the selected annealing conditions [21,31,32].

Chemical Characterisation
The surface chemistry of ND and OLC was examined using XPS. The spectrum ( Figure S3) of ND contained an intense carbon (C 1s) peak and two weak peaks, corresponding to nitrogen (N 1s) and oxygen (O 1s), marked at 284, 400, and 532 eV, respectively. Also, a series of CKLL Auger peaks appear between 1200 and 1270 eV that can provide information about the chemical bonding [43,44]. The atomic percentage of carbon increased, while the atomic percentage of nitrogen and oxygen decreased post annealing (Table S2), which could be due to the reduction in the surface dangling bonds, such as -C-H and -C-O-bonds, during annealing [36].
The relative contents of sp 3 and sp 2 carbon in the samples were determined by analysing the C-KLL X-ray excited Auger spectra ( Figure S4 and Table S2) with the method used by Jones et al. [45]. The KLL spectrum was smoothened by Savitzy-Golay with 15 points and differentiated with 9 points [45]. The distance (energy 'D') between the maximum of positive excursion and the minimum of negative excursion of the first differential was determined. Figure S5 shows XRD patterns of PA, indicating typical diffraction peaks of crystalline PA located at 15.
, respectively. Figure S6 shows XRD patterns of ibuprofen, indicating typical diffraction peaks of crystalline ibuprofen located at 16.4 • , 17.4 • , 20 • and 22 • , respectively. As the typical XRD patterns for both drugs studied are located at the 2θ ranging from 10 to 40 • , the XRD diffraction presented in Section 3.4 is also shown in the same 2θ range for comparison purpose.
Post-annealing, the D value was found to increase, indicating an increase in the content of sp 2 carbon [46]. Relative contents of sp 3 and sp 2 carbon were determined from the linear interpolation of the D value of 13 eV for diamond (100%-sp 3 ) and 22 eV for graphite (100%-sp 2 ). The D value of ND was found to be 13 eV, despite the presence of sp 2 carbon, and the exact reason for this is not clear. Post annealing, the sp 2 content was found to increase to 68%, indicating the conversion of diamond to graphite, consistent with the results from Raman spectroscopy. OLC also contains 32% of sp 3 carbon, indicating the presence of diamond core, consistent with the results from HRTEM analysis.

Particle Size Analysis
OLC produced from thermal annealing of NDs, showed an aggregate size in the range of 0.45-21.5 µm ( Figure 2 and Table 1).

Chemical Characterisation
The surface chemistry of ND and OLC was examined using XPS. The spectrum ( Figure S3) of ND contained an intense carbon (C 1s) peak and two weak peaks, corresponding to nitrogen (N 1s) and oxygen (O 1s), marked at 284, 400, and 532 eV, respectively. Also, a series of CKLL Auger peaks appear between 1200 and 1270 eV that can provide information about the chemical bonding [43,44]. The atomic percentage of carbon increased, while the atomic percentage of nitrogen and oxygen decreased post annealing (Table S2), which could be due to the reduction in the surface dangling bonds, such as -C-H and -C-O-bonds, during annealing [36].
The relative contents of sp 3 and sp 2 carbon in the samples were determined by analysing the C-KLL X-ray excited Auger spectra ( Figure S4 and Table S2) with the method used by Jones et al. [45]. The KLL spectrum was smoothened by Savitzy-Golay with 15 points and differentiated with 9 points [45]. The distance (energy 'D') between the maximum of positive excursion and the minimum of negative excursion of the first differential was determined. Figure S5 shows XRD patterns of PA, indicating typical diffraction peaks of crystalline PA located at 15.4°, 18.1°, 20.3°, 23.3°, 24.2°, and 26.4°, respectively. Figure S6 shows XRD patterns of ibuprofen, indicating typical diffraction peaks of crystalline ibuprofen located at 16.4°, 17.4°, 20° and 22°, respectively. As the typical XRD patterns for both drugs studied are located at the 2 ranging from 10 to 40°, the XRD diffraction presented in Section 3.4 is also shown in the same 2 range for comparison purpose.
Post-annealing, the D value was found to increase, indicating an increase in the content of sp 2 carbon [46]. Relative contents of sp 3 and sp 2 carbon were determined from the linear interpolation of the D value of 13 eV for diamond (100%-sp 3 ) and 22 eV for graphite (100%-sp 2 ). The D value of ND was found to be 13 eV, despite the presence of sp 2 carbon, and the exact reason for this is not clear. Post annealing, the sp 2 content was found to increase to 68%, indicating the conversion of diamond to graphite, consistent with the results from Raman spectroscopy. OLC also contains 32% of sp 3 carbon, indicating the presence of diamond core, consistent with the results from HRTEM analysis.

Particle Size Analysis
OLC produced from thermal annealing of NDs, showed an aggregate size in the range of 0.45-21.5 µ m ( Figure 2 and Table 1).  All samples have a broad particle size distribution in the range of 0.5 to 21.5 µm. This distribution is slightly bimodal, with two peaks between 1.6 microns and 2 microns and at 3.8 microns and shows a majority of particles with a size between 1 and 5 µm.
The primary particle size of OLC depends on the primary particle size of ND used; however, the aggregate size depends on various factors, such as annealing temperature and aggregate size of ND used for the synthesis [47]. Detonation ND usually has a primary particle size of less than 10 nm; however, the presence of surface functional groups facilitates the formation of tight aggregates of size 20-30 nm, and these primary aggregates form secondary aggregates of 100 nm to a few micrometres in size. In addition, the presence of van der Waals interactions between these aggregates often leads to the formation of agglomerates. Therefore, application of these agglomerates of ND as a starting material in the synthesis of OLC resulted in the production of aggregates of OLC bonded by graphitic layers [14].

Drug Loading Efficiency
UV results indicated that, with an increase in the drug concentration, a statistically significant difference was found in the drug loading (p < 0.05 for PA and p < 0.01 for IBU; one-way ANOVA). Drug loading reached a maximum when the drug concentration in the loading solution reached the saturation point ( Figure 3).
The primary particle size of OLC depends on the primary particle size of ND used; however, the aggregate size depends on various factors, such as annealing temperature and aggregate size of ND used for the synthesis [47]. Detonation ND usually has a primary particle size of less than 10 nm; however, the presence of surface functional groups facilitates the formation of tight aggregates of size 20-30 nm, and these primary aggregates form secondary aggregates of 100 nm to a few micrometres in size. In addition, the presence of van der Waals interactions between these aggregates often leads to the formation of agglomerates. Therefore, application of these agglomerates of ND as a starting material in the synthesis of OLC resulted in the production of aggregates of OLC bonded by graphitic layers [14].

Drug Loading Efficiency
UV results indicated that, with an increase in the drug concentration, a statistically significant difference was found in the drug loading (p < 0.05 for PA and p < 0.01 for IBU; one-way ANOVA). Drug loading reached a maximum when the drug concentration in the loading solution reached the saturation point ( Figure 3).
The higher loading efficiency for ibuprofen (IBU) compared to paracetamol (PA) could be attributed to the higher solubility of IBU in ethanol. Although the loading efficiency was higher, the fraction of adsorbed to the unadsorbed drug remaining in the initial solution used for OLC loading was lower for IBU compared to that of PA, which could be due to stronger interactions between IBU and ethanol or due to saturation of OLC.  The higher loading efficiency for ibuprofen (IBU) compared to paracetamol (PA) could be attributed to the higher solubility of IBU in ethanol. Although the loading efficiency was higher, the fraction of adsorbed to the unadsorbed drug remaining in the initial solution used for OLC loading was lower for IBU compared to that of PA, which could be due to stronger interactions between IBU and ethanol or due to saturation of OLC.

Solid State Analysis of Drug
Thermal analysis results of PA/OLC complex with different drug loadings were compared ( Figure 4A). In the case of PA/OLC complex, the complex with a drug loading of 15.5% exhibited a melting peak at 169 • C, characteristic of monoclinic Form I paracetamol [48], suggesting the presence of crystallinity; however, the complex with drug loading ≤11.5% did not show any crystallinity, which could indicate that saturation has been reached and any loading higher than 11.5% results in crystallisation of PA.

Solid State Analysis of Drug
Thermal analysis results of PA/OLC complex with different drug loadings were compared ( Figure 4A). In the case of PA/OLC complex, the complex with a drug loading of 15.5% exhibited a melting peak at 169 °C, characteristic of monoclinic Form I paracetamol [48], suggesting the presence of crystallinity; however, the complex with drug loading ≤11.5% did not show any crystallinity, which could indicate that saturation has been reached and any loading higher than 11.5% results in crystallisation of PA.
Solid state characteristics of IBU/OLC complex with different drug loadings were also compared and the complex with a drug loading of 31.6% and 36.6% exhibited a melting peak at 74.8 °C [49,50], suggesting the presence of crystallinity; however, the complex with loading 20.7% did not show any crystallinity ( Figure 4B), which could indicate that saturation has been reached and any loading higher than 20.7% results in crystallisation of ibuprofen.   Solid state characteristics of IBU/OLC complex with different drug loadings were also compared and the complex with a drug loading of 31.6% and 36.6% exhibited a melting peak at 74.8 • C [49,50], suggesting the presence of crystallinity; however, the complex with loading 20.7% did not show any crystallinity ( Figure 4B), which could indicate that saturation has been reached and any loading higher than 20.7% results in crystallisation of ibuprofen.
The results from DSC were confirmed by X-ray diffraction studies ( Figure 5). As mentioned previously in Section 3.1.3, the XRD diffraction graphs presented in Figure 5 only demonstrate the data within the 2θ range from 10 • to 40 • for comparison purpose. For OLC, there is a negligible broad graphitic (002) peak at round 2θ = 25.2 • , which comes from the onion-like nanographite. Diamomd (111) and (220) planes characteristic peaks are located at around 2θ = 43.7 • and 75.1 • , which are not shown in the 2θ range selected in Figure 5.
The results from DSC were confirmed by X-ray diffraction studies ( Figure 5). As mentioned previously in Section 3.1.3, the XRD diffraction graphs presented in Figure 5 only demonstrate the data within the 2 range from 10 o to 40 o for comparison purpose. For OLC, there is a negligible broad graphitic (002) peak at round 2=25.2 o , which comes from the onion-like nanographite. Diamomd (111) and (220) planes characteristic peaks are located at around 2=43.7 o and 75.1 o , which are not shown in the 2 range selected in Figure 5.
XRD patterns of PA/OLC complex with a loading of 15.5% showed peaks at 15.4°, 18.1°, 20.3°, 23.3°, 24.2° and 26.4°(2θ), corresponding to crystalline PA [51]. However, PA/OLC complex with loadings of 11.5% ( Figure 5) and 7.1% (data not shown) did not show any peaks corresponding to crystalline PA, suggesting that the drug was present completely in an amorphous form.
In the case of IBU loading, the complex with loadings of 31.6% ( Figure 5) and 36.6% (data not shown) showed peaks at 16.4°, 17.4°, 20° and 22° (2θ), corresponding to crystalline IBU [52,53]. However, IBU/OLC complex with a loading of 20.7% did not show any diffraction pattern corresponding to crystalline IBU. Therefore, all subsequent studies were performed on PA/OLC complex-11.5% and IBU/OLC complex-20.7% only, which showed no crystallinity in the sample.

Drug-Carrier Interactions
It is important to consider possible chemical interactions between drugs and OLC, since these interactions may affect the chemical nature and stability of drugs. FTIR spectra of PA/OLC complex 11.5% and PA/OLC physical mixture containing an equivalent amount of drug as that of the complex showed peaks corresponding to an NH amide band stretch at 3320 cm −1 and a broad phenolic OH stretch at 3129 cm −1 , similar to that of Pure PA ( Figure 6A). Spectra of PA/OLC complex did not show any significant shift in the existing peaks or new peaks, confirming physical adsorption of PA. Also, spectra of IBU/OLC complex-20.7% and IBU/OLC physical mixture containing an equivalent amount of drug as complex showed peaks corresponding to carbonyl CO stretch at 1694 cm −1 , similar to pure IBU, and no new peaks were found, indicating the absence of any chemical interactions ( Figure 6B) [54]. Thus, the interaction of drug/OLC is inferred to be through physical adsorption.  [51]. However, PA/OLC complex with loadings of 11.5% ( Figure 5) and 7.1% (data not shown) did not show any peaks corresponding to crystalline PA, suggesting that the drug was present completely in an amorphous form.
In the case of IBU loading, the complex with loadings of 31.6% ( Figure 5) and 36.6% (data not shown) showed peaks at 16.4 • , 17.4 • , 20 • and 22 • (2θ), corresponding to crystalline IBU [52,53]. However, IBU/OLC complex with a loading of 20.7% did not show any diffraction pattern corresponding to crystalline IBU. Therefore, all subsequent studies were performed on PA/OLC complex-11.5% and IBU/OLC complex-20.7% only, which showed no crystallinity in the sample.

Drug-Carrier Interactions
It is important to consider possible chemical interactions between drugs and OLC, since these interactions may affect the chemical nature and stability of drugs. FTIR spectra of PA/OLC complex 11.5% and PA/OLC physical mixture containing an equivalent amount of drug as that of the complex showed peaks corresponding to an NH amide band stretch at 3320 cm −1 and a broad phenolic OH stretch at 3129 cm −1 , similar to that of Pure PA ( Figure 6A). Spectra of PA/OLC complex did not show any significant shift in the existing peaks or new peaks, confirming physical adsorption of PA. Also, spectra of IBU/OLC complex-20.7% and IBU/OLC physical mixture containing an equivalent amount of drug as complex showed peaks corresponding to carbonyl CO stretch at 1694 cm −1 , similar to pure IBU, and no new peaks were found, indicating the absence of any chemical interactions ( Figure 6B) [54]. Thus, the interaction of drug/OLC is inferred to be through physical adsorption.

In Vitro Drug Release Studies
Release profiles of pure drug and drug/OLC complex in the presence and absence of SDS are compared in Figure 7. In the absence of SDS, drug release from PA/OLC complex was incomplete, with only 11.4% release in 10 min and could be due to poor wettability of carbon onion (particles were found sticking to the surface of the vessel). Similarly, drug release from IBU/OLC complex was also incomplete in the absence of SDS; however, this was slightly higher, with 17.4% release in 10 min

In Vitro Drug Release Studies
Release profiles of pure drug and drug/OLC complex in the presence and absence of SDS are compared in Figure 7. In the absence of SDS, drug release from PA/OLC complex was incomplete, with only 11.4% release in 10 min and could be due to poor wettability of carbon onion (particles were found sticking to the surface of the vessel). Similarly, drug release from IBU/OLC complex was also incomplete in the absence of SDS; however, this was slightly higher, with 17.4% release in 10 min and could be attributed to the higher solubility of ibuprofen in the sodium phosphate buffer media. In the presence of SDS, drug release from PA/OLC complex was complete in 15 min, and was significantly faster (p < 0.0001, two-way ANOVA) compared to pure PA. Similarly, drug release of IBU/OLC complex was complete in 30 min in the presence of SDS and was significantly faster (p < 0.0001, two-way ANOVA) compared to pure crystalline IBU.
Crystals 2020, 10, x FOR PEER REVIEW 11 of 19 and could be attributed to the higher solubility of ibuprofen in the sodium phosphate buffer media.
In the presence of SDS, drug release from PA/OLC complex was complete in 15 min, and was significantly faster (p < 0.0001, two-way ANOVA) compared to pure PA. Similarly, drug release of IBU/OLC complex was complete in 30 min in the presence of SDS and was significantly faster (p < 0.0001, two-way ANOVA) compared to pure crystalline IBU. Complete release in the presence of SDS could be due to the increased wettability of OLC particles. Compared to the pure drug, faster release in drug/OLC complex could be attributed to the amorphous nature of the drug loaded in complex and higher surface area that is in contact with the dissolution media, supporting the advantage of using OLC in amorphous drug delivery.
To better understand the kinetics of drug release, drug release profiles obtained in the presence of SDS for PA and IBU were fitted with the simplified Higuchi model based on Fick's law of diffusion, which describes drug release from an insoluble matrix [55,56]. Higuchi square root of time plots for both drug loaded complexes (Figure 8 and Table 2), display a two-step release, with an initial burst effect, which could be attributed to the drug release from superficial pores, followed by a slow release, which could be attributed to the drug release from deeper micropores (Section 3.7).
Crystals 2020, 10, x FOR PEER REVIEW 12 of 19 Complete release in the presence of SDS could be due to the increased wettability of OLC particles. Compared to the pure drug, faster release in drug/OLC complex could be attributed to the amorphous nature of the drug loaded in complex and higher surface area that is in contact with the dissolution media, supporting the advantage of using OLC in amorphous drug delivery.
To better understand the kinetics of drug release, drug release profiles obtained in the presence of SDS for PA and IBU were fitted with the simplified Higuchi model based on Fick's law of diffusion, which describes drug release from an insoluble matrix [55,56]. Higuchi square root of time plots for both drug loaded complexes (Figure 8 and Table 2), display a two-step release, with an initial burst effect, which could be attributed to the drug release from superficial pores, followed by a slow release, which could be attributed to the drug release from deeper micropores (Section 3.7).

Porosity Analysis
N2 sorption analysis was performed to understand the changes in the porosity of OLC before and after drug loading. The adsorption/desorption isotherm (Figure 9) of OLC before and after loading exhibited a typical type II isotherm and H3 hysteresis loop, corresponding to the presence of aggregate created porosity.
The pore size distribution ( Figure 10) of pure OLC shows the presence of pores in the range of 2-14 nm. The pore size distribution for micropores could not be determined, however the micropore volume (Table 3) of pure OLC was determined to be 0.03 cm 3 /g.  Table 2. Kinetic parameters of drug release from paracetamol and ibuprofen loaded carbon onion corresponding to a 2-step release with an initial burst effect during 3 to 10 min.

Porosity Analysis
N 2 sorption analysis was performed to understand the changes in the porosity of OLC before and after drug loading. The adsorption/desorption isotherm (Figure 9) of OLC before and after loading exhibited a typical type II isotherm and H3 hysteresis loop, corresponding to the presence of aggregate created porosity.
The pore size distribution ( Figure 10) of pure OLC shows the presence of pores in the range of 2-14 nm. The pore size distribution for micropores could not be determined, however the micropore volume (Table 3) of pure OLC was determined to be 0.03 cm 3 /g. Table 3. Surface areas and pore volumes obtained from nitrogen sorption of carbon onion (OLC), ibuprofen-loaded carbon onion (IBU/OLC complex) and paracetamol loaded carbon onion (PA/OLC complex).

Sample
Specific Surface Area 1 (m 2 /g)   of 2-5 nm completely disappeared and could be due to the adsorption of gaseous IBU on OLC that could be attributed to the sublimation of IBU during N2 sorption analysis [57]. Post loading, in PA/OLC complex, a reduction in the pore size was observed ( Figure 10) in pores in the range of 2-14 nm and the pore size distribution broadens in the range of 7.5-10.5 nm, suggesting the adsorption of PA into the pores of OLC. In the case of IBU/OLC complex, pores in the range of 2-5 nm completely disappeared and the pore size distribution broadens for pores in the range of 5-14 nm, which could be due to the reduction of pore size from drug loading For PA/OLC physical mixture, a reduction in surface area and pore volume (Table 3) was observed but the pore size distribution (Figure 10) was similar to that of OLC, indicating that there was no drug present in the pores of OLC. However, for IBU/OLC physical mixture, pores in the range of 2-5 nm completely disappeared and could be due to the adsorption of gaseous IBU on OLC that could be attributed to the sublimation of IBU during N2 sorption analysis [57].
Post loading, in PA/OLC complex, a reduction in the pore size was observed ( Figure 10) in pores in the range of 2-14 nm and the pore size distribution broadens in the range of 7.5-10.5 nm, suggesting the adsorption of PA into the pores of OLC. In the case of IBU/OLC complex, pores in the range of 2-5 nm completely disappeared and the pore size distribution broadens for pores in the range of 5-14 nm, which could be due to the reduction of pore size from drug loading

Toxicity Studies of Carbon Onion
Toxicity studies of these OLC aggregates were performed on Caco-2 cells to determine the feasibility of application of OLC in oral drug formulations. The efficiency of uptake of microparticles is much lower compared to nanoparticles, therefore the risk of toxicity associated with these OLC aggregates can be much lower.
However, studies have shown uptake of microparticles of size up to 10 µm by intestinal epithelium [58], suggesting the importance of considering the toxicity of microparticles. Hence, toxicity of OLC aggregates on Caco-2 cells was investigated by MTT assay.
OLC showed a significant effect (p < 0.0001, one-way ANOVA) on the survival of Caco-2 cells (Figure 11), with cell survival reduced when exposed to media containing OLC concentrations ≥400 µg/mL; however, the cell viability was still over 75%, suggesting OLC is a safe drug carrier for oral drug delivery. Since OLC aggregates are insoluble, the chance of internalisation is low, although synthesis conditions need to be optimised to eliminate the production of aggregates of size <10 µm, to avoid the possibility of internalisation.

Toxicity Studies of Carbon Onion
Toxicity studies of these OLC aggregates were performed on Caco-2 cells to determine the feasibility of application of OLC in oral drug formulations. The efficiency of uptake of microparticles is much lower compared to nanoparticles, therefore the risk of toxicity associated with these OLC aggregates can be much lower.
However, studies have shown uptake of microparticles of size up to 10 µ m by intestinal epithelium [58], suggesting the importance of considering the toxicity of microparticles. Hence, toxicity of OLC aggregates on Caco-2 cells was investigated by MTT assay.
OLC showed a significant effect (p < 0.0001, one-way ANOVA) on the survival of Caco-2 cells (Figure 11), with cell survival reduced when exposed to media containing OLC concentrations ≥400 µ g/mL; however, the cell viability was still over 75%, suggesting OLC is a safe drug carrier for oral drug delivery. Since OLC aggregates are insoluble, the chance of internalisation is low, although synthesis conditions need to be optimised to eliminate the production of aggregates of size <10 µ m, to avoid the possibility of internalisation. * * * * * * * * * Figure 11. Cytotoxicity of carbon onion against Caco-2 cells. MTT assay was used to analyse the survival rate of Caco-2 cells incubated with different concentrations of carbon onion. Statistically significant differences compared to control (0 µ g/mL) are noted for P < 0.05 (** P < 0.01; *** P < 0.001; **** P < 0.0001, one-way ANOVA and Dunnett's multiple comparison test). Results are the mean of triplicate experiments ± SD.

Conclusions
The current study investigated the application of carbon onion (OLC) aggregates as drug carriers using paracetamol and ibuprofen as model drugs. Nanodiamonds were annealed to 1100 °C and OLC aggregates produced exhibited very low cytotoxicity on Caco-2 cells with cell viability over 75% at all the concentrations tested (10−800 μg/mL). The solution adsorption method was employed for drug loading and the results demonstrated that loading efficiency increased with an increase in initial drug concentration, reaching a maximum when the initial drug concentration reached the saturation point. Results from thermal analysis and diffraction studies suggested that the drug was found to be completely amorphous in PA/OLC complex and IBU/OLC complex with the loading of 11.5% and 20.7%, respectively, indicating that the concentration of drug in the loading solution needs to be optimised to achieve complex without any crystalline drug. Drug release kinetics were studied using USP II dissolution method in sodium phosphate buffer with and without SDS. About 11.4% and 17.1% of the total loaded PA and IBU, respectively, was released in the absence of SDS; however, this Figure 11. Cytotoxicity of carbon onion against Caco-2 cells. MTT assay was used to analyse the survival rate of Caco-2 cells incubated with different concentrations of carbon onion. Statistically significant differences compared to control (0 µg/mL) are noted for P < 0.05 (** P < 0.01; *** P < 0.001; **** P < 0.0001, one-way ANOVA and Dunnett's multiple comparison test). Results are the mean of triplicate experiments ± SD.

Conclusions
The current study investigated the application of carbon onion (OLC) aggregates as drug carriers using paracetamol and ibuprofen as model drugs. Nanodiamonds were annealed to 1100 • C and OLC aggregates produced exhibited very low cytotoxicity on Caco-2 cells with cell viability over 75% at all the concentrations tested (10−800 µg/mL). The solution adsorption method was employed for drug loading and the results demonstrated that loading efficiency increased with an increase in initial drug concentration, reaching a maximum when the initial drug concentration reached the saturation point. Results from thermal analysis and diffraction studies suggested that the drug was found to be completely amorphous in PA/OLC complex and IBU/OLC complex with the loading of 11.5% and 20.7%, respectively, indicating that the concentration of drug in the loading solution needs to be optimised to achieve complex without any crystalline drug. Drug release kinetics were studied using USP II dissolution method in sodium phosphate buffer with and without SDS. About 11.4% and 17.1% of the total loaded PA and IBU, respectively, was released in the absence of SDS; however, this was incomplete due to poor wettability of OLC. Complete drug release was achieved in the presence of SDS and the dissolution rate was higher than that of the pure crystalline drug, establishing that drug loading was reversible and also faster due to amorphous nature of the drug loaded into OLC. The maximum initial concentration of drugs in the loading solution to ensure complete amorphous drug loadings is estimated at 11.5 PA and 20.7 IBU. These results suggest that the development of carbon onion based drug delivery systems is promising.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4352/10/4/281/s1, Figure S1: Scanning electron microscope pictures showing the morphology of (a) pristine nanodiamonds showing loosely bound agglomerates and (b) carbon onion showing tightly bound aggregates reaching several micrometers in size; Figure S2: Raman spectra of ND and OLC showing the F2g peak corresponding to the sp 3 diamond at 1325 cm −1 , D band and G band corresponding to sp 2 disorder at~1330 cm −1 and~1570 cm −1 , respectively. Post annealing, OLC spectra consists of a broad D band in the region 1332-1350 cm −1 , attributed to increase in the sp 2 disorder with a background of sp 3 diamond, indicating the transformation of diamond to graphite [37]; Figure S3: XPS spectra of pristine nanodiamonds showing an intense carbon peak at 284 eV and two weak nitrogen and oxygen peaks at 400 and 532 eV, respectively. (a) and carbon onion showing a decrease in intensity for carbon, oxygen and nitrogen peaks (b). The major peaks are marked for Carbon (C 1s), Nitrogen (N 1s) and Oxygen (O 1s) present on the surface.; Figure S4: XPS C KLL Auger Spectra (a) and their first derivatives (b) as compared with pristine nanodiamonds (ND) and carbon onion (OLC). D parameter represents the width between maximum and minimum excursions of the derivative of Auger spectra. D value was found to increase after annealing, indicating an increase in the content of sp2 carbon [36]; Figure Table S1: Raman shifts, areal intensities and peak widths obtained from the spectra of ND and OLC determined by fitting spectra with a Gaussian line shape using Origin software. After annealing, the FWHM of G band decreased, suggesting the formation of graphitic shells [38][39][40]; Table S2: The relative atomic contents of carbon, oxygen, and nitrogen, and the relative atomic contents of different chemical states of carbon determined from XPS analysis. The atomic percentage of carbon increased, while the atomic percentage of nitrogen and oxygen decreased post annealing, which could be due to the reduction in the surface dangling bonds, such as -C-H and -C-O-bonds, during annealing [36]. After annealing, the sp 2 content was found to increase to 68%, indicating the conversion of diamond to graphite and OLC still contains 32% of sp 3 carbon, indicating the presence of diamond core. Funding: This research was partially funded by Aston University through an overseas student scholarship.