2.1. Model Development and Optimization
To optimize the preparation of MLs, we chose the lipid composition and CMNP concentration during liposome preparation as the experimental variables and studied the effects on the encapsulation efficiency (EE) of CMNPs and liposome size. By incorporating the experimental design and non-linear regression, those effects were investigated within the context of response surface methodology (RSM), which enables experimental investigation of the interactions of factors simultaneously as opposed to one factor at a time approach. Models of EE and liposome size were constructed, which allow the evaluation of the significance of the parameters and provide the prediction capability for EE and size. The utilization of RSM is a powerful methodology in the enhancement of the optimization of the liposomal preparation within a designed experiment [
29,
30,
31,
32]. Based on the statistical theory with 2
5−1 fractional factorial design, a central composite rotatable design (CCRD) was adopted to design the experiments using four variables and five levels (
Table 1). The effects of four independent variables (mass of DPPC, mass of DDAB, mass of CH, and concentration of CMNP, coded separately as X
1, X
2, X
3, and X
4, respectively) were studied to assess the encapsulation efficiency (EE) of CMNPs and size of MLs. The structure of CCRD comprises 31 experimental runs with seven replicates at the center to estimate the pure error sum of squares, 16 factorial points (−1 and +1), and 8 axial points (−2 and +2). Each experiment in the design was carried out randomly from statistical points of view. The individual experimental condition and their corresponding observed response values from experiments are tabulated in
Table 1.
The Pareto chart was used to study the effects of variables on EE and size, in which the length of each bar was proportional to the absolute value of the estimated effect (associated regression coefficient) (
Figure 1). The vertical line represents the 95% significance limit, which declared statistical significance when the bar associated with a factor crossed this line [
32]. From
Figure 1A, one cross-product (X
2X
4), one linear-coefficient (X
4), and two quadratic coefficients (X
12 and X
32) are found to be significant (
p < 0.05). Similarly,
Figure 1B indicates that one quadratic coefficient (X
32) and three linear coefficients (X
2, X
4 and X
3) are significant (
p < 0.05).
For validation of the significance of the RSM model, the trial consequences in
Table 1 were applied for the multiple regression analysis and the response surface 3D plots, and predictions over observed values were generated. The fitting of experimental data with a polynomial equation representing the response surface model was used to evaluate the relationship between the factors and the response values statistically. Second-order regression models correlating responses and variables were obtained according to the estimation of data:
where X
i is the coded value of each variable, Y
1 is the EE response value, and Y
2 is the size response value. From the regression results, the R
2 values are 0.744 for EE and 0.765 for size, which reveal the level of variation of all responses that is predicted by the model. Considering the percentage of variability, the models were used for subsequent prediction of response values during the validation step.
The optimum conditions for preparation of MLs were obtained through regression models in accordance with the limit criterion of the maximum EE response and minimum size response using Statistica software. The optimized conditions estimated by the model equations were DPPC (X
1) = 2 μmol, DDAB (X
2) = 0.5 μmol, CH (X
3) = 1 μmol, and CMNP (X
4) = 0.25 mg/mL (
Table 2). The theoretical predicated values of EE and size under the above conditions were 86.9% and 121.1 nm, respectively. The accuracy of the model was validated with four repeated experiments under aforementioned optimum conditions not employed in the model (
Table 2). The experimentally observed values for EE and size were 84.2 ± 4.7% and 124.3 ± 6.5 nm (mean ± SD,
n = 4), respectively. Therefore, the verification experiments confirmed the validity of the predicted model, with the predicted value being reasonably close to the mean of the experiment value and within its mean ±SD range for both EE and size. Taken together, the results of model validation experiments endorse the model from CCRD as being accurate and reliable for prediction of EE and size.
Further evaluation of the optimization technique was carried out, where the observed value was compared with the observed value for EE (
Figure 2A) and size (
Figure 2B). The experimental observed values accord well with the predicted values from the model development, where clustering points around the diagonal line confirms a satisfactory correlation between the experimentally observed values and the predicted values. It should be noted that even the model coefficients are empirical, which may not be assigned to any physical or chemical effects. They have been shown to be very useful for predicting outcomes of untested experimental conditions [
32].
The relationships between the factors and responses were studied from the surface 3D contour plots by plotting the response model against two of the factors, while keeping the other two factors constant (
Figure 3). From
Figure 3(A1,B1), the EE and size showed similar dependence, with low DDAB and high DPPC leading to lower response values, but stronger dependence was found for DDAB. From
Figure 3(A2,B2), both the low DPPC and CH could maximize EE, while simultaneously minimizing size. For DPPC and CMNPs, low DPPC combined with high CMNPs leads to higher EE. Nonetheless, a small liposome size could only be achieved using low DPPC and low CMNPs (
Figure 3(A3,B3)). Overall, the interactions between factors and their influence on the responses support the choice of the optimum formulation from the regression model.
2.2. Characterization of Physico-Chemical Properties
The CMNPs were prepared by chemical co-precipitation of Fe
+2 and Fe
+3 ions and surface-modified by citrate coating. The MLs were prepared by using the optimum formulation of lipids and CMNPs (
Table 2) by the thin film hydration–extrusion method to load CMNPs inside the aqueous cores of the liposomes. The hydrodynamic size distribution was analyzed by dynamic light scattering (DLS) together with the zeta potential values using a Zetasizer (
Figure 4A,B). As shown in
Table 3, the average particle size (diameter) of CMNPs is 36.5 ± 5.4 nm and for MLs is 124.3 ± 6.5 nm. The polydispersity index (PDI) is below 0.30, providing evidence of uniform particle size distribution in addition to the good suspension stability of MLs (
Table 3) [
33]. For drug delivery using liposomes, a PDI value of 0.3 or below is considered to be acceptable, which indicates a homogenous distribution of the lipid vesicles [
34]. The loading percentage of CMNPs in MLs (weight percentage of CMNPs in MLs) was determined to be 21.8% ± 1.7% (
w/w). The average zeta potentials are −19.4 ± 4.3 mV and 16.3 ± 3.7 mV from electrophoretic mobility measurements (
Table 3). The average zeta potential of CMNPs is negative due to the carboxylate groups of the coated citrate, which changes to a positive value for MLs due to the presence of cationic lipid DDAB in the lipid bilayer. Intracellular uptake will be facilitated by taking advantage of this cationic nature of MLs for targeted magneto-photothermal cancer therapy originated form the MH–PT nature of the CMNPs cargo inside MLs.
The TEM images of CMNPs (
Figure 4C) and cryo-TEM imaged of MLs (
Figure 4D) show close to spherical morphology, while the particle size is generally consistent with that from DLS measurements. The size of CMNPs was below 20 nm, as assessed by ImageJ analysis of discrete nanoparticles in the agglomerate and shown in the size histogram of CMNPs (
Figure 4C). This size is consistent with that observed for the individual CMNPs within the aqueous core enclosed by a lipid bilayer in MLs (
Figure 4D). The magnetic responsiveness of MLs was confirmed from the time-lapsed images in
Figure 4E, where a 3 mg/mL suspension of MLs in PBS was guided by a magnet to the side of the tube to result in a more transparent solution with time. In general, this behavior will provide additional magnetic targeting ability for MLs through their guidance with an external magnet to the tumor site, followed by MH–PT-induced by AMF–NIR lasers for tumor therapeutic functions [
35].
The stability of MLs in vitro was determined using nanoparticle tracking analysis (NTA) after incubating MLs in PBS (3 mg/mL) at 37 °C for different durations and tracking the movement of MLs with a 405 nm laser. Using NTA, MLs were observed as scattering points moving under Brownian motion, with larger particles scattering more light and appearing to be bigger [
36]. As shown in
Figure 5A, no significant change of the size of the scattering light was observed from the screenshot images at any time points. An estimation of the sample polydispersity at any given time due to the high resolution of the captured NTA images also indicated that the sample is fairly monodispersed, which is consistent with the PDI value from the DLS (
Table 3). For quantitative comparison, the concentration of particles (MLs) was determined as a function of the particle diameter after different incubation times (
Figure 5B). At time 0, the peak particle diameter was close to the value predicted from the RSM model (121 nm) and the experimental value from DLS (124 nm). After incubation for 1 h, the peak particle diameter increased to ~200 nm but the peak particle concentration decreased. Further increase in incubation time did not result in apparent change of peak particle diameter, but the peak particle concentration decreased. No MLs were above 400 nm, even after 48 h incubation. The stability of liposomes depends on many factors, while the shift in particle concentration may be due to destruction of MLs [
37]. Particle aggregation may also lead to an increase in size and decrease in counts from NTA analysis [
38]. Nonetheless, the remaining MLs of acceptable average size (~200 nm) will not impair the capability for intracellular uptake by cancer cells [
39].
We used X-ray diffraction (XRD) to characterize CMNPs, MLs, and blank liposomes (MLs without CMNPs), the diffraction patterns for which are shown in
Figure 6A. For CMNPs and MLs, six diffraction peaks were observed at 2
θ = 30.3°, 35.5°,43.4°, 53.5°, 57.2°, and 62.6°, which could be indexed to the (220), (311), (400), (422), (511), and (440) planes of a cubic cell. The crystalline structure was affirmed to be comparable with that of magnetite (JCPDS card number 19-0629), which indicates pure Fe
3O
4 associated with the spinal structure of magnetite in CMNPs [
15,
32,
40]. The absence of the six diffraction peaks in the patterns shown for blank liposomes, which were prepared similarly to MLs but without CMNPs, further supports the encapsulation of CMNPs in MLs. From the strongest diffraction (311) peak, the average crystal grain sizes calculated from the Scherer equation were 14.7 and 17.1 nm for CMNPs and MLs, respectively, by assuming spherical crystals from XRD line broadening, which is consistent with the nanoparticle size of CMNPs observed from cryo-TEM.
Figure 6B shows the Fourier transform infrared (FTIR) spectra of CMNPs, MLs, and blank liposomes (MLs without CMNPs). In CMNPs, the strong absorption peak at 572 cm
−1 corresponds to the Fe–O bond in the nanoparticles, which also appeared in MLs. The –OH vibrations are confirmed from the peak at 3424 cm
−1 [
41]. The characteristic peaks of citrate at 1394 and 1735 cm
−1, due to the symmetric vibration and asymmetric stretching of C–O from –COOH group and C=O from –COOH, were shifted to 1381 and 1637 cm
−1 in CMNPs to successfully support citric acid coating. The encapsulation of CMNPs in MLs could be also confirmed from the absence of the Fe–O stretching bond at 574 cm
−1 by comparing the spectra of blank liposomes against those of MLs. In both liposomal preparations, characteristic peaks at 1097 cm
−1 and 1246 cm
−1 could be assigned to the phosphate group (P=O), while the CH
2 symmetric and asymmetric stretching provided peaks at 2848 cm
−1 and 2916 cm
−1, respectively [
42].
Based on superconducting quantum interference device (SQUID) analysis, the magnetization curves determined at room temperature indicate that the saturation magnetization value of CMNPs is 60.5 emu/g (
Figure 6C). The saturation magnetization value of MLs is lower than that of CMNPs (13.0 emu/g). The drop in magnetization strength may be associated with encapsulation of CMNPs within the liposomes, as the magnetization sensitivity may decline when the surfaces of CMNPs are occupied by lipids [
43,
44]. Nonetheless, decreased saturation magnetization may largely stem from the reduced weight percentage of CMNPs in MLs, as the magnetic moment is based on the unit weight of MLs, while all other components in MLs are diamagnetic [
45,
46]. This could be supported by comparing the expected saturation magnetization value of MLs (13.2 ± 1.0 emu/g), calculated from the loading percentage of CMNPs in MLs (21.8% ± 1.7%), and the value measured using SQUID analysis (13.0 emu/g). The remnant magnetization and coercivity could be determined from the insert of
Figure 6C, which are 0.25 and 0.9 emu/g for CMNPs and 13.2 and 12.9 Oe for MLs, respectively. Superparamagnetic properties could be suggested for CMNPs and MLs, with the remnant magnetization being close to zero and the coercivity being low from the magnetization curve. Combined with the sufficient magnetization value, MLs will be bestowed with the critical properties required for magnetically targeted cancer therapy.
The results from thermogravimetric analysis (TGA) of blank liposomes, CMNPs, and MLs are shown in
Figure 6D, together with the differential thermal analysis (DTA) curves shown in the figure insert to elucidate the thermal properties and loading percentage of CMNPs in liposomes. The early weight loss before 200 °C for all samples may be assigned to the losses of residually bound water and absorbed CO
2. For CMNPs, some weight loss was recorded starting from ~300 °C, reaching a final residual weight of ~94% (
w/w) at 650 °C from the thermal decomposition of coated citric acid on the surface of CMNPs [
47]. For blank liposomes, substantial weight loss occurs within 260 °C to 380 °C, with a sharp decomposition peak temperature at ~360 °C and a shoulder peak temperature at ~290 °C shown from the DTA curve. The residual weight of blank liposomes was 10.3% (
w/
w) at the end of temperature increase to 650 °C. In comparison, MLs displayed a similar decomposition peak, temperature, but the shoulder temperature transformed into a more distinctive peak temperature possibly due to the interactions of lipids with CMNPs [
48]. From the residual weight difference between MLs and blank liposomes at 650 °C, the weight percentage of CMNPs in MLs was estimated to be ~21.0% (
w/
w), assuming negligible weight loss for CMNPs, which compared favorably with that obtained from chemical analysis (21.8% ± 1.7%).
2.3. Heating Efficiency Induced by AMF, NIR, or Combined Laser Treament
The heating efficiency study was carried out by treating a suspension of MLs or CMNPs in water with AMF, NIR laser, or combined AMF + NIR laser treatment to examine whether dual MH–PT could amplify the heating power of MLs. For this purpose, samples were prepared in distilled water in Eppendorf tubes and subjected to AMF, NIR, or combined laser treatment up to 5 min (
Figure 7A). The thermal images of the solutions were captured from the bottoms of the tubes with an infrared thermal camera (
Figure 7B). The peak temperatures acquired from the time-lapsed thermal images were plotted vs. time to compare the heating efficiency subject to different treatments (
Figure 7C). A fixed base concentration of CMNPs (0.6 mg/mL) was used to compare the heating efficiency between CMNPs and MLs. There was no statistical difference in temperature change between MLs and CMNPs regardless of mode of treatment, indicating encapsulation of CMNPs in MLs did not influence the heating efficiency at the same concentration of heating agent in the solution. For AMF treatment at 52 kHz frequency for 5 min, the temperature reached 39 °C. The samples exposed to NIR laser (808 nm) for 5 min at 1.8 W/cm
2 showed better heating efficiency compared with AMF, where the temperature reached 43 °C. This behavior is consistent with a recent report that PT is far more efficient than MH using magnetite nanoparticles [
49]. Most importantly, in comparison to single-mode operation employing AMF or NIR lasers, dual-mode operation employing AMF + NIR lasers provides better heating efficiency, as the temperature reaches 56 °C at the end of the same treatment period. Taken together, the in vitro heating experiments support that CMNPs alone can increase the temperature from dipole interactions under AMF and with the absorbance of light near the NIR region [
50]. Nonetheless, the combined use of AMF and NIR lasers could enhance the heating efficiency of MLs originated from encapsulated CMNPs. By taking advantage of the combined effect of dual-mode treatment, it is expected that the efficacy of cancer hyperthermia treatment could be enhanced by using a lower concentration of the heating agent and a shorter treatment time [
51].
For quantitative comparison, the specific absorption rate (SAR, W/g CMNPs) was determined from the slope of each heating curve in
Figure 7C by linear regression of all data points in a single curve, using the specific heat capacity of water and concentration of CMNPs [
52,
53]. As shown in
Table 4, there is no significant difference in SAR between CMNPs and MLs (based on unit weight of CMNPs) regardless of the stimulus modality, supporting the observation that CMNPs in MLs could fully preserve the response to MH–PT. Comparing the single modality, PT shows significant improvement of SAR compared to MH. Indeed, dual AMF–NIR laser stimulus modality resulted in elevated SAR values, which were 2.0 and 1.5 times greater than that of AMF and NIR lasers alone, respectively.
2.4. Intracellular Uptake of MLs
Cationic liposomes have the potential to carry cargos (such as drugs) and accumulate in tumor tissues owing to their positive charge [
54]. This passive accumulation process can result in significant accumulation of loaded drugs compared with the administration of free drugs [
55]. Due to their targeted and drug release functions, cationic liposomes show the potential to increase drug efficacy in tumors [
56]. Considering the positive zeta potential of MLs (
Table 3) originated from the cationic lipid DDAB in the lipid composition, we expected the designed MLs could be able to selectively bind onto the surfaces of U87 glioblastoma cells, thus facilitating cellular trafficking through charge-mediated endocytosis for enhanced therapeutic effects from the thermally induced killing of cancer cells. The increased intracellular uptake of MLs as induced by the targeting function of the cationic MLs was confirmed from confocal laser scanning microscopy and compared with control MLs (MLs prepared without cationic lipid DDAB) (
Figure 8). The confocal microscopy images revealed intracellular red fluorescence corresponding to IR780-labeled MLs when U87 cells were exposed to MLs for 24 h. This illustrates endocytosis and subsequent accumulation of MLs in the cell cytoplasm after internalization. MLs (red fluorescence) were observed to be internalized in U87 cells and appeared to be co-localized with lysosomes stained with LysoTracker (green fluorescence). The contributing factor may be related to the intracellular localization of MLs in the acidic compartment in the cytosol. As with the merged images in
Figure 8, only MLs but not control liposomes showed high fluorescence intensity (yellow) corresponding to both MLs (red) and LysoTracker (green), indicating efficient uptake of MLs through endocytosis. Besides lysosomal co-localization, MLs not co-localized with LysoTracker were also observed to a large extent. This is consistent with the finding that cationic liposome can undergo electrostatic-interaction-mediated membrane fusion with the anionic endosome membrane, thus escaping the endosome [
57]. Therefore, co-staining of the acidic organelles (lysosomes) demonstrated that despite not all MLs being associated with lysosomes, an important proportion of them did end up in this acidic cell compartment.
Compared to U87 cells treated with MLs, a drastically diminished intracellular red fluorescence signal was observed for control MLs prepared without DDAB (
Figure 8). This difference is due to the contribution of DDAB, which interacts with the cell membrane for passive targeting to facilitate intracellular uptake. As the U87 cell surface is negatively charged, the DDAB-containing cationic MLs rapidly interact with biological membranes to allow for facilitated passive targeting and particle internalization, in comparison with control liposomes [
58]. This induced specific cellular targeting and uptake of MLs will lead to an enhanced MH–PT effect for increased cytotoxicity and provide an efficient milieu for thermally induced cell death in vitro.
2.5. Thermally Induced Cancer Cell Killing In Vitro
After confirming the intracellular uptake, the biocompatibility of MLs was studied by exposing U87 (tumor) and 3T3 (normal) cells to various concentrations of MLs without thermal induction. As shown in
Figure 9A, the relative cell viability (compared to cell culture medium) remained above 95% for MLs concentrations up to 4.5 mg/mL (in cell culture medium), confirming that the cell viability was not affected by internalization of MLs and that the prepared MLs appear to be biocompatible. The thermally induced cancer cell killing effect was examined by incubating U87 with different concentrations of MLs, followed by AMF or NIR laser treatment. As showed in
Figure 9B, dose-dependent cytotoxicity was observed when using MLs as a thermal inductive agent for killing U87 cancer cells after subject to MH or PT treatment. There was no significant difference in cytotoxicity between the AMF and NIR laser groups, although dosage-dependent cytotoxicity occurred at higher concentrations of MLs. Most importantly, AMF + NIR laser dual-mode treatment led to a significant reduction in cell viability compared with single-mode AMF or NIR laser treatments. At the maximum concentration of 4.5 mg/mL, AMF and NIR lasers reduce cell viability by 32–35%, which could be significantly enhanced by 2.4-fold to 82% for AMF + NIR lasers. These outcomes demonstrated the synergistic impact with an improved thermal effect induced by magneto-photothermal treatment [
17].
We further evaluated the cytotoxicity of MLs after AMF and NIR laser treatments using flow cytometry analysis. The cells were stained with Annexin V/PI to examine the percentage of live (Q3), early apoptotic (Q4), late apoptotic (Q2), and necrotic cells (Q1) according to differences in plasma membrane permeability and integrity (
Figure 10). From
Table 5, the flow cytometry data confirm that cell death occurred primarily from apoptosis associated with MH [
59] or PT [
60]. Similar to the result obtained from the MTT assay (
Figure 9A), the control group without any treatment shows a minimum apoptotic rate with >95% viable cells. Single-mode treatment with AMF (or NIR laser) resulted in ~19% (~22%) early apoptosis and ~3% (~6%) late apoptosis. With AMF treatment, MH resulted in cell death and the live cell percentage decreased to 77.9%. With NIR laser treatment, a more pronounced cell apoptosis rate was observed due to higher thermal effects induced by PT (
Figure 7C), while the live cell percentage decreased to 71.9%. With combined AMF + NIR laser treatment, ~26% early apoptosis and ~34% late apoptosis rates were observed. The dual treatment also revealed a slightly increased necrosis rate not observed from single treatment and gave the most pronounced decrease of live cell percentage to 38%, due to combined MH–PT.
The in vitro cytotoxicity study, thus, suggests AMF–NIR laser treatment could induce dual hyperthermia modality for future translational study in vivo. This proof-of-concept study in vitro should endorse future clinical application of the optimized cationic MLs developed in this study. The high EE of CMNPs and their suitable size for internalization by the targeted cancer cells could facilitate the use of MLs developed here for dual MH–PT cancer therapy.