Magnetic carbon nanotubes for self-regulating temperature hyperthermia

Magnetic hyperthermia can enhance the anti-tumor effects of chemotherapy. As carbon nanotubes are ideal drug carriers for chemotherapy, their combination with magnetic nanoparticles provides a novel chance for multi-modal thermo-chemotherapy. Most related work focuses on attaching Fe3O4 nanoparticles on carbon nanotubes, however the hyperthermia temperature for this combination can not be self-regulated due to the high Curie temperature of Fe3O4. In this work, magnetic Zn0.54Co0.46Cr0.6Fe1.4O4 nanoparticles with low Curie temperature were attached onto carbon nanotubes to obtain the magnetic carbon nanotubes. The morphology, formation mechanism, magnetic properties, heat generation ability and cytotoxicity of the magnetic carbon nanotubes were investigated. These magnetic carbon nanotubes show a Curie temperature of 43 °C and a self-regulating temperature at 42.7 °C under clinically applied magnetic field conditions (frequency: 100 kHz, intensity: 200 Oe). The evaluation of in vitro cytotoxicity suggests no obvious toxicity effects under the concentrations of 6.25 μg ml−1 to 100 μg ml−1. This study proposed a methodology for the bespoke synthesis of magnetic carbon nanotubes with a low Curie temperature for self-regulating magnetic hyperthermia, which may be used for further research on loading drugs for multi-modal cancer therapy.


Introduction
Magnetic nanoparticle-based hyperthermia is considered as a green method for treating cancer due to its low side effects. [1][2][3] Basically, magnetic nanoparticles are injected into the tumor region and generate heat upon the application of an external alternating magnetic eld. Since normal cells possess higher heat resistance than tumor cells, the tumor cells can be killed selectively without affecting adjacent normal tissue if the temperature can be maintained at 42-45 C. [4][5][6] Carbon nanotubes (CNTs) are well ordered, hollow graphitic materials with high aspect ratio. [7][8][9] Due to the ability to move easily among tissues and compartments of body and come across the membrane into the cell via endocytosis and diffusion, CNTs are considered as perfect carriers for drugs, nucleic acid and imaging agents for targeting therapy. [10][11][12] The large surface area of CNTs, together with their hollow structure, enables them to be loaded with a large quantity of drug molecules, which can effectively prolong the circulation time of drug molecules in blood and enhance cellular uptake of the drug by cancer cells. [13][14][15][16] Graphic surface of CNTs can also be modied in order to gra specic antibodies, avoiding immune action and reducing the possibilities of undesired cytotoxicity. 17 These fascinating properties make CNTs emerge as promising therapy-enhancing nanomaterials.
Hyperthermia can enhance the anti-tumor effects of chemotherapy by increasing the uptake of carcinostatics into tumor cells and inhibiting the repair of tumor cells. 18,19 The dose of anti-tumor drugs, therefore, can be decreased by the combination with hyperthermia, and the undesirable side effects of chemotherapy drugs can be also minimized. As CNTs are ideal carriers of drugs, their combination of magnetic nanoparticles may provide a possibility for further research on multi-modal thermos-chemotherapy. However, most of related work focus on attaching Fe 3 O 4 nanoparticles on CNTs, but the hyperthermia temperature in vivo is hard to monitor and control. And a temperature higher than 45 C may cause damage to normal tissue. 20,21 Tuning the Curie temperature of magnetic media to a value just above the treatment temperature is considered as an expedient route to control the magnetic hyperthermia temperature and realize self-regulation. [22][23][24][25] Curie temperature is the temperature at which ferromagnetic materials lose their intrinsic permanent magnetic properties and consequently lose their ability to generate heat under alternating magnetic eld. The Curie temperature therefore gives an upper limit to the operational temperature for the magnetic media. In the present work, we prepared magnetic nanoparticles (Zn 0.54 Co 0.46 Cr 0.6 -Fe 1.4 O 4 ) with Curie temperature at 45.7 C and successfully decorated them on the surface of CNTs through hydrothermal method. Magnetic carbon nanotubes (MCNTs) with Curie temperature at 43 C were obtained. These MCNTs show a Curie temperature of 43 C and self-regulating temperature at 42.7 C under clinically applied magnetic eld conditions (frequency: 100 kHz, intensity: 200 Oe), which is suitable for hyperthermia use. Compared with pure nanoparticles, MCNTs have a better dispersive ability to avoid the inhomogeneous temperature distribution caused by the aggregation of nanoparticles. This study proposed a method on the bespoke synthesis of MCNTs with low Curie temperature for self-regulating magnetic hyperthermia, which may be used of further research on loading drugs for multi-modal cancer therapy. Then the NaOH solution (0.5 mol L À1 , 150 ml) was slowly dropped into the metal salts solution with vigorous stirring for 30 min to form the precursor. The precursor was sealed in autoclave and heated to 250 C (heating rate: 2.3 C min), maintained for 2 h and then allowed to cool to room temperature. The products were washed with deionized water and ethanol till neutral, and then dried at 60 C for 6 h in a vacuum drying chamber. Magnetic nanoparticles (MNPs) were also synthesized through the same way without adding carboxylic CNTs for comparison purpose.

Characterization
The crystal structure was characterized by PANalytical Empyrean X-ray diffractometer (XRD) (Netherlands) with Cu Ka radiation (l ¼ 0.15406 nm). The Raman analysis was conducted on a Renishaw Invia Laser Raman Spectrometer (UK) with a green laser (l ¼ 532 nm) incitation. Fourier transform infrared spectroscopy (FTIR) data were collected on a NEXUS 670 spectrometer. The transmission electron microscope (TEM) images were obtained on a FEI Tecnai G2 F30 (USA). Magnetization curves were recorded at 300 K on a Jilin University JDM-13 vibrating sample magnetometer (China).
The Curie temperature was measured by thermogravimetric analysis (TGA) using Mettler-Toledo TGA 851 (Switzerland) with a Nd-Fe-B magnet (100 Â 50 Â 5 mm 3 ) placed up over the samples about 10 cm. When the temperature is below Curie temperature, the weights of MCNTs recorded by TGA are less than real weights due to the attractive magnetic force between MCNTs and the magnet. When temperature increases to the Curie point, the MCNTs loses their magnetism and the weight becomes the real weight. The transition point is the Curie temperature.
The time-dependent temperature curves were determined by calorimetric measurements. The experimental setup is illustrated in Fig. 1. To make the performance of MCNTs under the alternating magnetic eld closer to the clinic treatment, the experimental condition was set up as the same as Jordan's clinical treatment. 26,27 All the MCNTs suspension were dispersed into deionized water with the same concentration (112 mg ml À1 ), and the frequency and intensity of the magnetic eld are set to 100 kHz and 200 Oe, respectively. Once being irradiated, the temperature of suspension was measured by an alcohol thermometer at 60 s intervals.

Cell culture and cytotoxicity assay
Human epidermal keratinocyte (HaCaT) cells were purchased from China Center for Type Culture Collection (Wuhan) and maintained in high glucose Dulbecco's Modied Eagle Medium (DMEM) (Gibco, USA) in a humidied 37 C incubator (Thermos HERAcell150, USA) at 5% CO 2 supplemented with 10% fetal bovine serum (FBS) (Gibco, USA) and 1% penicillin-streptomycin (Gibco, USA). Aer sterilized with 75% (v/v) ethanol/ deionized water for 24 h and rinsed with phosphate buffer saline (PBS) three times, the MCNTs were suspended in cell culture medium with a concentration of 0 mg ml À1 , 6.25 mg ml À1 , 12.5 mg ml À1 , 25 mg ml À1 , 50 mg ml À1 and 100 mg ml À1 . HaCaT cells were seeded in 24-well plates and 96-well plates with a density of 5 Â 10 4 cells per ml for evaluation of in vitro cytotoxicity of MCNTs.
The quantitative evaluation of in vitro cytotoxicity of MCNTs against HaCaT cells were performed by using the cell counting kit-8 (CCK-8) assay (Dojindo Laboratories Kumamoto, Japan). HaCaT cells were seeded in 96-well plates with a density of 5 Â 10 4 cells per ml. Aer 3 h of culture, adherent cells were exposed to the fresh culture medium containing 0 mg ml À1 (control), 6.25 mg ml À1 , 12.5 mg ml À1 , 25 mg ml À1 , 50 mg ml À1 and 100 mg ml À1 MCNTs. At 24 h, 48 h and 72 h, the cell viability was assessed. The culture medium was briey refreshed, and 10 ml CCK-8 was added to each well containing 100 ml new medium. Aer incubated in dark at 37 C for 2 h, 100 ml of each MCNTs solution was transferred to a new 96-well plate to measure the absorbance of the formazan product at 450 nm by using a microplate Fig. 1 The experimental setup of calorimetric measurements. reader (SPECTRAFLUOR, TECAN, Sunrise, Austria). The cell viability was calculated through formula (1): where OD nanoparticles is the optical density of the well containing various concentration of nanoparticles, DMEM medium and cells. OD control is the optical density of the well containing DMEM medium and cells. OD blank is the optical density of the well only containing DMEM medium. In addition, cell viability was observed by staining with Calcein-AM, Hochest 33258, and propidium iodide (PI). Aer 72 h of culture, cells were rinsed with PBS for three times and stained with PBS containing 2 mmol L À1 Calcein-AM, 5 mg ml À1 Hochest 33258, and 4 mmol L À1 PI (Sigma, Mo, USA). Aer incubated in dark at 37 C for 15 min, cells were rinsed with PBS for three times. Then the viable and dead cells were imaged using a uorescent microscope (OLYMPUS BX71, Japan).

Results and discussion
The size, morphology and crystalline information have been investigated by TEM, XRD, Raman and FTIR analysis, given in   2c shows the XRD phase analysis. Patterns of MNPs can be easily indexed to cubic spinel structure of CoFe 2 O 4 -CoCrFe 2 O 4 (JCPDS-ICDD, 03-0864); and no obvious XRD peaks from impurities could be observed. Compared with MNPs, the patterns of MCNTs have an extra peak (002) arising from CNTs. 28 The element information was determined by an Electron-Probe Microanalyzer. The results, together with the molar ratio of metal elements initially added, are given in Table  1s. † Clearly, all metal elements in the samples do not deviate signicantly from their initial stoichiometry. Based on the Electron-Probe Microanalyzer and XRD experiments, it could be concluded that the all the metal ions successfully enter the lattice. The Raman spectra of MCNTs are given in Fig. 2e. D and G bands at 1346 cm À1 and 1575 cm À1 correspond to the sp 3 and sp 2 hybridized carbons respectively, 29 arising from CNTs. Peak at 477 cm À3 is assigned to T 2g (2) mode, corresponding to the local symmetry vibrations of metal ions in the octahedral site of spinel nanoparticles. 30 Peak at 678 cm À1 is related to A 1g mode arising from the stretching vibrations of Fe 3+ and O 2À in the tetrahedral site in spinel nanoparticles. 31 These active vibration modes show the spinel structure of nanoparticles attached on CNTs. A combination of TEM, XRD and Raman spectra analyses conrm the crystalline structure of MCNTs.
To have a clear observation of the changes in functional groups on MCNTs, FTIR was performed. As shown in Fig. 2f, compared with the spectra of carboxylic CNTs, the band peaks at 582 cm À1 , 3433 cm À1 appeared and peaks at 1200 cm À1 , 1705 cm À1 vanished on MCNTs. The appearance of peaks at 582 cm À1 could be assigned to the stretching vibration of the metal-oxygen bond at tetrahedral site in spinel structure of Zn 0.54 Co 0.46 Cr 0.6 Fe 1.4 O 4 nanoparticles, formed on the surface of CNTs. 32 The appearance of a broad peak at 3433 cm À1 indicates the existence of hydroxyl group on MCNTs, which may be induced from alkaline solution during the hydrothermal process. 33 The bands at 1200 cm À1 , 1705 cm À1 and 724 cm À1 are all attributed to the carboxylic group on CNTs, corresponding to the bending of C-O bond, stretching and bending vibration of C]O bond respectively. Compared with CNTs, the disappearance of 1705 cm À1 , 1200 cm À1 and the appearance of 724 cm À1 on MCNTs shows change of carboxylic group, which mainly results from the formation of either a monodentate complex or a bidentate complex between the carboxyl group and metal atoms on the surface of nanoparticles. 34 These spectroscopic changes support that the nanoparticles are covalently bound to the carboxylic CNTs.
The formation mechanism of MCNTs was investigated and proposed, as depicted in Fig. 3a. The carboxyl groups on the surface of CNTs provide sites for the nucleation and growth of nanoparticles. In this experiment, aer dissolving the carboxylic CNTs into metal salt solution, the metal ions (Zn 2+ , Co 2+ , Fe 3+ and Cr 3+ ) are preferentially absorbed on these sites due to the electrostatic force. When a NaOH solution is added, the precursors, metal hydroxides attached on the CNTs, form. With increasing temperature, dehydration reactions take place on Co(OH) 2    To obtain evidence for the proposed mechanisms, two experiments were conducted. Raman spectroscopy is the rst one to characterize the structure of precursor taken at 100 C. The result is given in Fig. 3b. Raman modes at 1346 and 1575 cm À1 could be attributed to the D band and G band of CNTs. The other three Raman modes at 791, 678 and 447 cm À1 are assigned to CoOOH, Zn 0.54 Co 0.46 Cr 0.6 Fe 1.4 O 4 and CoO, respectively. 35 In the Raman spectra of MCNTs (Fig. 2e) In the second experiment, pristine multi-walled carbon nanotubes without any functional groups were used to prepared magnetic carbon nanotubes. Fig. 3c shows the TEM image of magnetic carbon nanotubes. No magnetic nanoparticles were found attached on the surface of carbon nanotubes. This supports the proposed mechanism that carboxylic groups provide sites for the nucleation and growth of nanoparticles due to the electrostatic force, and nanoparticles cannot attach on the surface of carbon nanotubes without carboxylic groups.
The magnetic properties of MNPs and MCNTs were displayed in Fig. 4. Room temperature (300 K) magnetization hysteresis loops were shown in Fig. 4a. The coercivity (H c ) and saturation magnetization (M s ) of MCNTs are 129 Oe and 15 emu g À1 respectively, which is lower than that of MNPs (172 Oe and 34 emu g À1 ). The difference of H c and M s mainly attributes the addition of nonmagnetic CNTs, which decrease the magnetic moment per unit mass.
The thermogravimetric curves are shown in Fig. 4b. As the temperature increases, the nominal weight of MNPs and MCNTs both increase initially, which means that the magnetization decrease with increasing temperature. When the temperature exceeds the Curie temperature, the nominal weight will not change. It is also observed that the temperature range of losing magnetization is broad. This is caused by the individual size differences of nanoparticles. According to Nikolaev's research, 36 the Curie temperature is directly proportional to the bulk density of exchange bonds. The thickness of surface layer increases with decreasing particle radius. Since the number of exchange bonds at the surface layer is less than that at the particle core, the smaller the nanoparticle size, the lower the Curie temperature. Thus, the temperature range of losing magnetization may be inuenced by the size distribution of nanoparticles.
When the temperature reaches the Curie point of a majority of nanoparticles, it experiences the most rapid weight increase in the curves, and the Curie temperature can be determined by taking the maximum value of the rst derivative of the thermogravimetric curves. The Curie points of MNPs and MCNTs  are decided as 45.7 C and 43.0 C respectively. Compared with MNPs, the TGA curves of MCNTs shows a slight thermal degradation when the temperature increases over 200 C. This weight loss is mainly attributed to the decarboxylation of the carboxylic groups present on the CNTs. 37 The time-dependent temperature curves of MNPs and MCNTs suspensions were determined by calorimetric measurements. MNPs and MCNTs were dispersed into deionized water with the same concentration (112 mg ml À1 ) before exposure to the external alternating magnetic elds (200 Oe, 100 kHz), which is applicable to clinical therapy. 26,27 As can be seen in Fig. 4c (blue line and red line), aer being irradiated for 60 min, the temperature of suspension could rise to 44.0 C and 42.7 C respectively. In order to evaluate heat efficiency quantitatively, specic absorption rate (SAR) under alternating magnetic elds, which describes the energy converted into heat per time and mass, was calculated using the formula (2): 38 where C is the specic heat capacity of suspension (4.18 J gK À1 ); dT/dt is the slope of the time-dependent temperature curve; m s is the mass of suspension and m m is the mass of magnetic media in the suspension. As the temperature higher than room temperature will cause heat loss to surroundings which may signicantly inuence value of SAR, 39,40 here we calculated SAR just at room temperature and select dT/dt at the initial several seconds of the experiment, when heat transfer to surroundings can be ignored. The SAR of MNPs and MCNTs are 774 W kg À1 and 695 W kg À1 respectively. As the H c and M s of MCNTs are lower than that of MNPs, the area of hysteresis loop of MCNTs is smaller than that of MNPs, which result in lower heat efficiency of MCNTs. Aer a rapid increase of temperature, the MNPs and MCNTs suspensions reach a stable temperature of 44.0 C and 42.7 C, which is close to their Curie temperatures (45.7 C and 43.0 C). To investigate the ability of self-regulating temperature of MCNTs, additional magnetic heating experiment with higher SAR has been conducted. Here, time-dependent temperature curve under the magnetic eld of 400 Oe is recorded and the result is given in Fig. 4c (black line). Although SAR is much higher (1372 W kg À1 ), MCNTs suspensions reach the stable temperature of 43.1 C, still fairly close to the stable temperature under 200 Oe and Curie temperature. Only when the hyperthermia temperature is maintained below the Curie point, MCNTs keeps their magnetism and generate sufficient heat under an alternating magnetic eld in this self-regulating system. Otherwise, the magnetic media will lose their magnetism and not generate heat, which makes MCNTs suspension remain constant around its Curie temperature. These experiments demonstrate the self-regulating nature of MCNTs and self-regulating temperature around 43 C, meet the requirement for hyperthermia. Fig. 4d shows the dispersed states of MNPs and MCNTs in aqueous media at the moment of dispersion and aer $12 h. Most MNPs precipitated aer 12 h while MCNTs are still nely dispersed. The UV-vis spectra were performed to monitor dispersion of MNPs and MCNTs in aqueous media and the results are given in Fig. 1s. † The absorbance peak of MNPs decreased 65.2% aer sedimentation for 12 h, which is almost 5 times higher than MCNTs. Compared with MNPs, MCNTs have a better dispersibility. MNPs with great surface energy and the magnetic force have strong tendency to aggregate together. This is because the gravity of aggregate is stronger than the dispersive effect caused by Brownian motion. For MCNTs, the carboxylic groups on nanotube surface are negatively charged, which enable the MCNTs to repel from each other and keep the solution dispersed. MCNTs with a better dispersibility may avoid the inhomogeneous temperature distribution caused by the aggregation of MNPs in magnetic hyperthermia.
Another concern about the medical application is the biocompatibility of MCNTs. In clinical trials, nanoparticles are injected into tumor regions and most of them are allocated in tumor tissue, which resulted in the several times higher uptake of nanoparticles in tumor than that in normal tissue. Only 2.7-10% nanoparticles could be detected in other normal tissue even aer 10 days' treatment. 41,42 Aer blood circulation and metabolism, the concentration of nanoparticles stay in normal tissue is low. For many cytotoxicity studies of hyperthermia, 100 mg ml À1 is considered as a very high concentration of nanoparticles that can extravasate into normal tissue. 43,44 Therefore, here we also select 100 mg ml À1 for cytotoxicity studies.
The cytotoxicity effect of MCNTs with the concentrations of 0 mg ml À1 , 6.25 mg ml À1 , 12.5 mg ml À1 , 25 mg ml À1 , 50 mg ml À1 and 100 mg ml À1 on HaCaT cells was examined aer co-culture of 24 h, 48 h and 72 h. As shown in Fig. 5a, the viability of HaCaT cells in all MCNTs concentrations are above 85%, and there is no signicant difference in cell viability between all MCNTs groups and control group (p-value > 0.05) aer co-culture of 48 h and 72 h. Aer co-culture of 24 h, all MCNTs groups show some enhancement of cell viability. The similar phenomenon has also been observed in the cytotoxicity assay of gold nanoparticles and the enhanced growth of cell may be related to the receptor-mediated control of cell membranes. 45,46 The more detailed mechanism is still under investigation.
In order to have more visualized observation of cell state, live-dead uorescent microscopic analysis was performed. Aer co-culture with MCNTs for 72 h, cells were stained with Calcein-AM, PI and Hochest 33258. Dead cells can be identied by a red uorescence generated by PI aer intercalating into DNA, which only occurs aer the entrance of the dead cell-permeable dye to the cell membrane. Cell nucleus can be identied by a blue uorescence generated by Hochest 33258 aer the binding of this living cell-permeable dye to DNA. Living cells can be iden-tied by a green uorescence generated by the enzymatic hydrolysis of Calcein-AM, which only takes place in living cells as a result of esterase activity. As can be seen in Fig. 5b, cells cocultured with all concentrations of MCNTs for 72 h show calcein-positive, Hochest 33258-positive and PI-negative state. This live-dead result is consistent with the quantitative CCK-8 assay in Fig. 5a. These results show that the MCNTs have low cytotoxicity for HaCaT in vitro under the concentrations of 6.25 mg ml À1 to 100 mg ml À1 . As cytotoxicity test is a quite complex systematic work, we will continue in-depth studies for other cell lines.

Conclusions
Magnetic carbon nanotubes with a Curie temperature of 45.7 C for self-regulating temperature hyperthermia were prepared through combining carboxyl carbon nanotubes and nanoparticles using hydrothermal method. Nanoparticles were successfully decorated on the surface of CNTs. The carboxylic groups on CNTs and the formation and hydration of CoO played important roles in the mechanisms of MCNTs formation, heterogeneous or homogeneous nucleation processes. MCNTs also show a better dispersibility than MNPs, which may avoid the inhomogeneous temperature distribution caused by the aggregation of nanoparticles in magnetic hyperthermia. The Curie temperature of MCNTs is 43 C while the temperature of MCNTs suspension can be self-regulated at 42.7 C under the clinically used magnetic elds, which is suitable for selfregulating temperature hyperthermia. The evaluation of in vitro cytotoxicity of MCNTs also suggests low toxicity under the concentrations of 6.25 mg ml À1 to 100 mg ml À1 . This study proposed a methodology on the bespoke synthesis of magnetic carbon nanotube for self-regulating hyperthermia, which may lead to further research on loading drugs for multi-modal cancer therapy.

Conflicts of interest
There are no conicts to declare.