Synthesis of plasmonic hybrid nanogels (PHNs) via photo-initiated one-pot polymerization
To obtain PHNs composed of light-responsive gold nanoparticles (GNPs) and thermo-responsive polymeric nanogels, we employed a photo-initiated one-pot synthesis method. The reaction mixture was prepared by blending a thermo-responsive monomer (i.e., NIPAM), several linker molecules (Table S1), a gold ion precursor (i.e., HAuCl4), and a photoinitiator (i.e., Darocur® 1173). Under exposure to 365 nm light, the reaction components quickly formed a globular structure through radical polymerization of the monomers and simultaneous self-integration of reduced gold ions into the polymeric network (Fig. 2a). The optimal condition for GNP synthesis was determined by altering the concentration and mixing ratio of the gold precursor and photoinitiator (Fig. S1). After 10 min of illumination, small-sized ligand-free GNPs less than 10 nm in size were successfully formed through ion reduction by the free radicals generated from the photoinitiator [46, 47]. During the polymerization of NIPAM and linkers, the GNPs that were readily synthesized and integrated with the polymeric network via weak interactive forces between the GNP and hydrophilic moieties of the polymeric networks (i.e., -OH, -NH, etc.) [48, 49]. Consequently, simultaneous self-integration was achieved to form hybrid nanogels composed of a polymeric network (i.e., PNIPAM) and GNPs.
The reaction time for the synthesis of PHNs was optimized by monitoring the hydrodynamic diameter and absorbance at the peak at each time point (Fig. 2b). After sampling small aliquots, an equivalent volume of deionized (DI) water was added to the products to terminate the free radical-mediated reaction. When N, N’-methylene bisacrylamide (MBA) was used as a linker molecule, the size of the resulting MBA-linked PHN (M-PHN) rapidly increased and reached a plateau within 10 min. Compared to the early-stage aliquots, the matured M-PHNs exhibited a monodisperse peak (Fig. S2). The diameters of the PNIPAM nanogels without GNPs increased to 400 nm as the reaction time increased (Fig. S3); however, the M-PHNs reached 80 nm within the same reaction time. A difference was observed in the growth pattern between that of PNIPAM nanogel and PHNs; this could be attributed to the effect of the consumption of radicals during GNP formation on the growth of the PHNs. Moreover, the absorbance spectra of the PHNs broadened and shifted to a longer wavelength compared with that observed in the reaction with gold ions only (Fig. S4). Thus, GNPs in the M-PHNs were well-incorporated into the polymeric networks during the reaction. The resulting M-PHNs exhibited relatively good stability under repetitive cycles of thermal stress than that of the PHNs without linker molecules (Fig. S5). This indicates that GNPs reinforced the structural stability of the PHNs by being anchored to the polymeric networks and regulating the growth kinetics of the nanogels.
The TEM images in Fig. 2c (i) show the representative morphology of the M-PHNs exhibiting GNPs embedded into the globular hybrid nanogel structure. As shown in the high-resolution image (Fig. 2c (ii)), sub-10 nm GNPs were assembled with a small gap. The typical lattice length and diffraction patterns for the assembled GNPs revealed a strong correspondence with the metallic GNPs, which have a face-centered-cubic facet (Fig. 2c (iii)) [50]. The polymeric network of the PHNs was not visible in the TEM images, which might be attributed to the thinness of the layer and low electron density, unlike in the GNPs. X-ray photoelectron spectroscopy (XPS) measurements were performed on the outermost layer of the PHNs to characterize the polymeric layer of the PHNs with precision. The wide-scan survey spectrum of the PHNs showed peaks corresponding to C1s, N1s, O1s, and Au4f at the characteristic binding energies (Fig. 2d). The relative atomic percentage of Au was almost indiscernible because the surface of the GNPs was mostly covered with thin polymeric layers. Nonetheless, the metallic Au0 peaks at 88.28 eV and 84.58 eV were measured from the narrow scan (Fig. S6). This indicates that the Au ions were successfully reduced to metallic Au during polymerization. Moreover, the peak shifted slightly to higher binding energy, indicating that the GNPs were embedded in the polymeric network through weak interactive forces (e.g., Van der Waals, dipole-dipole, induced dipole, etc.) [46, 47, 49, 51, 52]. In the C1s region, the spectrum was deconvoluted into three peaks (i.e., 284.78 eV for the C-C peak, 286.2 eV for the C-N peak, and 287.48 eV for the C = O), which indicate the presence of PNIPAM structures on the Au surface. In addition, the existence of PNIPAM was double-checked by observing at 532 eV for the C = O peak in the O1s spectrum. These XPS results indicate that the GNPs were successfully entrapped inside the thin PNIPAM-based nanogels.
Since the light-responsive property is closely related to the absorption and scattering cross sections [53], the integration density of GNPs in the PHN structure or the diameters of PHN should be further optimized. For this purpose, five kinds of other linker molecules, tryptophan (T), sucrose (S), polyethylene glycol diacrylate (P), alginate (A), and gelatin(G) were tested to control the density of the GNPs and the size of the PHNs. Figure 2e shows the difference in the diameter of the synthesized PHNs, which might be attributed to the different physicochemical properties of the linker molecules such as hydrodynamic diameters, functional groups, and the presence of hydrophilic pockets. The sugar-ring structure is known to stabilize the GNPs owing to their high hydrophilicity; hence, the structural stability of PHNs can be improved by embedding GNPs into additional polysaccharides [47, 51, 52]. As expected, both the density of the GNPs and the diameter of the PHNs depended on the linker molecules (Fig. S7). In the case of M-PHN, P-PHN, and A-PHN, smaller GNPs (i.e., less than 10 nm) tended to be more closely packed in the structures. In particular, the PHN fabricated with alginate as the linker (i.e., A-PHN) exhibited the highest density of the small GNPs. Consequently, the absorbance peaks shifted to longer wavelengths upon changing the linker molecules (Fig. S8a). Although the PHNs with large-sized GNPs (e.g., T-PHN and S-PHN) had red-shifted absorption bands, they showed poor colloidal stability (Fig. S8b), whereas the A-PHN, P-PHN, and G-PHN were stable after overnight storage at room temperature (i.e., 25℃). Moreover, A-PHN exhibited the highest scattering peak at approximately 700 nm (Fig. 2f), which could be attributed to the large optical cross-sectional area induced by the high density of GNPs. Based on these results, A-PHN was selected for further characterization and drug delivery applications.
To verify the existence of the alginate, we used calcium ions, which are well-known gelation linkers of alginate-based hydrogels and form a calcium-alginate egg-box structure [54]. As shown in Fig. S9, alginate-containing dispersions (i.e., A-PHN and alginate only) formed hydrogels in the presence of Ca2+, whereas gel formation was not observed for M-PHN. The formation of the purple-colored gel (A-PHN) indicated that alginate molecules were successfully incorporated into the GNPs and PNIPAM network because the alginate-only hydrogel exhibited a white color. Elemental mapping (Fig. 2g) and energy-dispersive X-ray spectroscopy results (Fig. S10) for A-PHN also showed the successful integration of each component. These results confirmed that high-density GNPs are embedded into alginate-linked PNIPAM nanogels, and the resulting A-PHN possesses a large optical cross-sectional area to the incident light.
Local heat generation by photothermal conversion of GNPs in the PHNs
A wave optic simulation was performed to predict the local heat generation from the PHNs under light illumination by calculating the absorption power of GNPs in the PHNs. First, it was considered that a single PHN structure was immersed in water, and the light at an intensity of 3.5 W/cm2 was continuously irradiated at wavelengths from 460–790 nm. For this modeling, the GNPs were arranged at regular intervals based on the GNP size (r) and interparticle distance between the GNPs (d) at the differential PHN diameter (R) (Fig. 3a and Fig. S11). The optical properties of the PHN structure were evaluated at different values of r (3, 4, and 5 nm), d (2, 3, 4, and 5 nm), and R (60, 80, and 100 nm). The number of GNPs was set to the maximum number that could be enclosed in the PHN. Figure 3b shows the absorption cross-section (σabs) of the PHN according to the wavelength of the incident light. The spectrum of σabs varies according to r and d. When the radius of the PHN was 100 nm, σabs increased as r increased, and d decreased. These results show a similar tendency as R changed from 80 nm to 60 nm. The absorption power was calculated at a wavelength of 550 nm, where PHN showed the maximum σabs. As shown in Fig. 3c, the absorption power also increased as r increased, and d decreased when R = 100 nm. Moreover, the absorption power at 550 nm increased with the GNP packing density in the PHNs (Fig. 3d). Therefore, a small number of GNPs with larger r values show higher absorption power than a large number of GNPs with smaller r when the interparticle distance was the same. As a result, the increased density of the proximate GNPs in the large PHNs induced more intensive heat generation.
To experimentally validate this prediction, three types of PHNs (i.e., M-PHN, A-PHN, and G-PHN) with different GNP densities, as shown in Fig. 3e, were tested. Considering the similar size and absorption peak wavelength of the PHNs, an 80-nm gold nanosphere (GNS) was used as a control to evaluate the photothermal conversion efficiency of the PHNs (Fig. 3f). As shown in Fig. 3g, the solution temperature of the colloidal A-PHN drastically increased by 11℃ on exposure to a 532 nm laser (3.5 W/cm2). Notably, the temperature increment depended on the density of the GNPs (i.e., A-PHN > M-PHN > G-PHN). In particular, the heat increment level was two-fold higher for the A-PHN, which had high-density GNPs, compared with that of the GNS. Furthermore, the photothermal conversion efficiency (η) of the GNS and PHNs was calculated to quantitatively compare the heat generated by the colloids [30, 35]. The rate of heat transfer was measured by removing the light source and monitoring the decrease in temperature (see Supporting Information for detailed calculations). From the calculations, η was found to be 15.69, 8.54, and 3.99% for A-PHN, M-PHN, and GNS, respectively (Fig. S12). A higher η for A-PHN than for M-PHN indicates that the packing density of GNPs in the PHN is crucial for effectively generating heat in response to light. For lyophilized PHN powder, temperature increments prominently occurred up to 80℃ even under exposure to laser at a relatively low power density (0.8 W/cm2). This can be attributed to a high-density state in the powder than in the colloidal state of the PHNs (Fig. S13). Temperature elevation higher than 20℃ was achieved in the A-PHN colloids by increasing the illumination time to 1800 s (Fig. S14) using a commercial LED lamp (0.8 W/cm2), which emitted light with a broad wavelength (480–700 nm).
Photothermally-driven conformational changes of PHNs
As the resonant laser was capable of increasing the solution temperature by more than 10℃ within 10 min, we supposed that it was sufficient to induce the conformational changes in the PNIPAM network at A-PHN. The heat- or light-responsive properties of hydrated A-PHNs in an aqueous solution were evaluated by increasing the solution temperature and light exposure time. Under suitable stimulation, the A-PHN structure became unstable owing to the increased hydrophobicity of the PNIPAM side chains, which led to interparticle aggregation (Fig. 4a). The hydrodynamic diameters of A-PHNs gradually increased upon exposure to both light (Fig. 4b (i)) and heat (Fig. 4b (ii)). This indicates that either local or global heat can induce conformational changes in the A-PHNs. Since the hydrophilic PNIPAM chains dehydrate as the solution temperature exceeds the LCST [37, 38], their hydrophobic isopropyl branch and backbone form a globular structure owing to the hydrophobic interaction, which results in the agglomeration of A-PHNs in colloids.
To confirm the conformational changes of the PNIPAM units in A-PHN, Raman measurements were performed to distinguish the hydrated and dehydrated status. As the alginates were linked with PNIPAM in the A-PHN, the shrinkage of the PNIPAM chain could be preserved with the introduction of Ca2+ in the turbid state (i.e., dehydrated A-PHN gel). As shown in Fig. S15, the dehydrated A-PHN gel presented fingerprint peaks for alginate (ν = 1240, 1350, and 1440 cm− 1) and hydrophobic moieties of PNIPAM (i.e., C-C ring stretching and breathing, ν = 980, 1050, and 1580 cm− 1), unlike the hydrated gel. This can be interpreted to indicate that the alginate-linked PNIPAM network in the PHN was proximate to the GNPs by conformational changes induced due to dehydration. In Fig. 4c, the dark-field scattering images show the spatially-controlled formation of agglomerates under light illumination. With increasing light exposure time, the size of the agglomerate increased, and the resulting Raman signal for the A-PHN dynamically changed to a spectral pattern similar to that of the dehydrated gel. This means that the photothermal conversion of GNPs can induce conformational changes in the thermo-responsive PNIPAM at A-PHN. Furthermore, the change in the surface charge with increasing temperature clearly shows evidence of such changes in the PNIPAM structure at the surface of A-PHN (Fig. 4d). The surface charge of A-PHN was initially negative (c.a. -39 mV) owing to the negatively charged alginate, and it changed to a more negative value of -54 mV as the solution temperature increased to 45℃. In the case of M-PHN, the same tendency toward negative charges was observed. Subsequent investigations focused on finding an optimal temperature range for the conformational changes in PHNs. Based on the change in transmittance and turbidity of the colloidal PHNs with increasing temperature, LCST values of approximately 38.4℃ and 35.6℃ for the A-PHN and M-PHN, respectively, were obtained (Fig. 4e). These temperatures were higher than that obtained for PNIPAM-based polymers in general (i.e., 32–33℃). This is probably due to additional intermolecular interactions (i.e., hydrogen bonds) between PNIPAM and alginate or GNPs in A-PHN. Moreover, the 1H-NMR study using A-PHN showed a temperature-dependent peak shift to downfield, which is strong evidence of the conformational changes in the PNIPAM nanogel (Fig. 4f and Fig. S16) [55]. Since those changes of A-PHN occur close to 38.4℃, which is an easily achievable range through the photothermal conversion of GNPs, photothermally triggered spatiotemporal controllability of A-PHN in the cellular delivery can be easily accomplished.
Photothermally-driven spatiotemporally controlled drug delivery using A-PHN
To verify the controlled drug release by photothermally driven conformational changes of the A-PHN (Fig. 5a), two model drugs, doxorubicin (dox) and paclitaxel (ptx), were loaded into the A-PHN during the calcium-induced gelation method. The loading efficiency of dox was calculated to be 60% (v/v) according to the standard curve for dox absorbance (Fig. S17). As shown in Fig. 5b, the loaded dox was released faster from the A-PHN in the presence of light. Moreover, the release profile drastically increased under acidic conditions, which resulted from the loosened polymeric network at low pH (Fig. S18) and the synergistic effect of protonated dox to boost the release [56]. Additionally, photothermally induced drug release was monitored using the intrinsic Raman signals of dox (Fig. 5c). According to the illumination time, the Raman signals of dox emerged gradually, whereas the same peaks were also observed in the drug-loaded A-PHN itself (See Fig. 4c). This attribute of the structural changes in A-PHN also occurred following the loading of dox with calcium ions. Moreover, a series of Raman spectra were obtained to check the temporal controllability of drug release from A-PHNs under discrete light illumination. Figure 5d (i) shows the Raman signals obtained alternatively in the presence and absence of laser illumination. The released dox signals drastically increased initially during the “laser on” condition (Fig. 5c and Fig. S19), whereas only minimum signals were observed during the “laser off” condition. The increase in the Raman signal slowed as the amount of remaining dox inside the A-PHN decreased after the first sequence (Fig. 5d (ii) and (iii)). This clearly demonstrates the temporal controllability of drug release in response to light.
Dox is well-known for specifically targeting the cell nucleus; therefore, the released drugs from the A-PHN would travel rapidly into the nuclei. To visualize this, dox@A-PHN (i.e., 80 µM dox) was used to treat A375P melanoma cells. The cells were placed under an LED lamp for predetermined periods (i.e., 0, 1, 10, and 30 min). Non-light illuminating groups were placed in the dark to avoid undesired leakage of dox from the A-PHN. As a result, colocalization of dox within the nuclei was more noticeable in the cells in the 30-min light exposure group than in the cells of the control group that were incubated in the dark (Fig. 5e (i)). To quantify the delivery efficiency, Pearson's correlation coefficient (k) was used to examine the degree of colocalization between dox and Hoechst. As expected, this value tended to increase in a time-dependent manner from 0.07 (at 0 min) to 0.645 (30 min). The value of the control group was significantly lower than that of the light-exposure groups (Fig. 5e (ii) and Fig. S20). A cytotoxicity test was performed under the same conditions to further confirm drug delivery efficacy. To avoid side effects on the cells from the remaining dox@A-PHN, the incubated cells were carefully washed and replaced with fresh media. Since dox does not kill the cells immediately, additional incubation was provided overnight before assaying. For the cells treated with dox@A-PHN under 30 min of light illumination, a significant decrease (48.9 ± 4.6%) in cell viability was observed (Fig. 5f). Notably, no significant cell death was observed in the control groups, including light exposure only without dox@A-PHN (101.3 ± 5.9%), dox@A-PHN in the dark (97.0 ± 1.8%), and A-PHN only (95.7 ± 4.8%) groups with and without light exposure.
Photothermally facilitated the endosomal escape of cellular internalized A-PHN
For a more nuanced understanding of the sophisticated drug delivery mechanism, we investigated the temperature distribution inside vesicles containing A-PHNs. Through observation of fluorescent and dark-field scattering images, we confirmed that A-PHN with 150–200 nm in diameter could be internalized into cells via the endocytosis pathway [15, 16]. As shown in Fig. S21, the diameters of vesicles including A-PHNs ranged between 200–800 nm and are visualized as red fluorescent dots (i.e., endo-lysosomes) and orange scattering dots (i.e., A-PHN-containing vesicles). The broad distribution of vesicle sizes can be attributed to the additional packing process during the proximate vesicle fusion in the cytosol after internalization. Based on the experimental observation of endocytosis, a simulation was performed to predict the temperature profiles by changing the number of A-PHNs in the endocytic vesicle under the light. As shown in Fig. 6a, the simulation domain consisted of a vesicle that enclosed the A-PHNs. The outer layer of the vesicle was a lipid bilayer membrane with a thickness of 5 nm, and the inside of the vesicle was assumed to be filled with water. We conducted a simulation by varying the size of the vesicles (Sv) and A-PHNs (SA−PHN) and the number of encapsulated A-PHNs (n). In the simulation, various Sv (200 nm, 400 nm, and 600 nm), SA−PHN (100 nm, 150 nm, and 200 nm), and n (1 to 6) were applied. The temperature changes inside the vesicles were calculated using the maximum absorption power obtained from the wave optics simulation as the heat source for the heat-transfer simulation. The initial ambient temperature was set to 37.5℃, which is the common incubation temperature for cellular experiments. The average temperature inside the vesicles rapidly increased in the presence of A-PHN (< 500 µs) and could be modulated according to the changes in parameters (Fig. 6b and Fig. S22). For example, the mean temperature increased from the initial temperature of 37.5℃ to 42.6℃ when n = 4, Sv=400 nm, and SA−PHN=150 nm (Fig. 6c). The result of the heat transfer simulation (Fig. 6d) showed that A-PHNs in the endocytic vesicle could generate sufficient temperature for vesicle rupture (i.e., ΔT > 4℃) [57, 58, 59]. This result indicates that spatially controlled drug delivery is possible after the progression of endocytic vesicle growth.
To confirm the simulation results experimentally, the cytosolic response was first monitored by applying laser illumination (532 nm, 3.5 W/cm2) to a vesicular A-PHN (Fig. S23). After the laser treatment, bubbles were generated immediately (< 1 s) around the A-PHN, which irreversibly destroyed the cytoskeleton. This also indicates that a high level of heat was generated around the vesicles [60, 61]. Since the cells maintain homeostasis under stress conditions, the photothermal-driven cytosolic temperature changes would also influence cellular components. For example, owing to the local heat generation in cells, heat shock proteins (HSPs) can be upregulated by light exposure in the presence of A-PHN. After incubating the A-PHNs in A375P cells for 4 h, low power LED lamp (0.8 W/cm2) was used to illuminate the cell. The level of HSPs markedly increased in the A375P cells with internalized A-PHN according to the illumination time (Fig. 6e (i)). As shown in Fig. 6e (ii), a drastic increase in the HSP intensity was observed in the A-PHN-internalized A375P cells with statistical significance within 30-min of light exposure. This is consistent with the observation presented in Fig. 5e, which indicates that the internalized A-PHNs could manipulate cellular responses by generating heat from the vesicles.
To effectively deliver drugs to targeted cellular sites, internalized drug-loaded carriers must escape from endocytic vesicles. To confirm this hypothesis, a membrane-impermeable fluorescent dye (i.e., calcein) was encapsulated into the A-PHN and delivered to the cells [20, 21]. As shown in Fig. 6f (i), the green fluorescent signals corresponding to the released calcein were spread into the whole cytosol in the light-exposed region of the cell, whereas only red dots (i.e., lysotracker, DND-99) were observed in the unexposed area. The calcein signals released into the cytosol highly overlapped with the lysotracker signals after full spread (k = 0.76, Fig. 6f (ii)), which indicates spatially controlled drug delivery only in the light-exposed region. Moreover, we observed widespread cytosolic distribution of calcein within 10 min of light treatment in a single cell (Fig. 6g). This reveals that pinpoint cytosolic manipulation can be achieved quickly and A-PHN can selectively influence the vesicular membrane by local heat generation for spatiotemporally controlled drug delivery.
Validation of enhanced drug delivery efficiency of PHNs using cellular 3D spheroids
To further examine the enhanced drug delivery efficiency using A-PHNs, a three-dimensional (3D) spheroid of cells, which mimics the microenvironments of the physiological system, was utilized. The delivery efficiency was evaluated based on the drug penetration depth and the ratio of live/dead cells. The penetration depth profiles were evaluated using a fluorescent dye (i.e., rhodamine 6G, R6G) loaded into the A-PHN (i.e., R6G@A-PHN). The results showed that R6G was successfully diffused into the deeper regions of the spheroid in the light-exposed group (Fig. 7a), whereas the spheroid in the dark exhibited a limited penetration depth (only observable near the outer surface, Fig. 7b). This is attributable to the generation of a temperature gradient from the A-PHN under light illumination.
In the live/dead cell assay, the non-fluorescent anticancer drug paclitaxel (i.e., ptx) was used to avoid the overlapping of fluorescence during the assay. Before performing spheroid experiments, phenotypic changes in the two-dimensional (2D) cell culture model were assessed by checking the ptx@A-PHN to determine whether the released drugs worked accurately (Fig. S24). Ptx-induced apoptotic nuclear fragmentation in A375P cells was observed in both ptx-and ptx@A-PHN (with light exposure)-treated groups, whereas no change was observed in the A-PHN treatment group. As ptx influences mitotic spindle assembly, chromosome segregation, and cell division by inhibiting microtubule assembly [62], nuclear fragments can be a result of the action of intracellular ptx released from A-PHN. After incubation of the spheroid with ptx@A-PHN under light exposure or in the dark, the live and dead cells of the spheroid were stained to evaluate drug delivery efficacy. As shown in Fig. 7c (i), the proportion of the red-colored dead cells in the spheroid was greater only in the light-illuminated group treated with ptx@A-PHN, indicating that the delivery efficacy was enhanced by the photothermally triggered release and rapid diffusion of ptx from A-PHN. To quantify this, the fluorescent signals of both live and dead cells were collected and the proportion of live/dead cells was calculated (Fig. 7c (ii)). In the case of the 30-min light exposure group using ptx@A-PHN, the ratio drastically decreased to 85.38% compared with that of the unexposed group. Collectively, drugs released from the light-responsive A-PHNs were observed to diffuse into the deeper region of the spheroid, killing more cells.