GQDs-MSNs nanocomposite nanoparticles for simultaneous intracellular drug delivery and fluorescent imaging

GQDs were prepared via the hydrothermal method (Permatasari et al. 2016; Qu et al. 2014), with the minor modification, using NaCitr (instead of citric acid) as a carbon precursor and TU as a base. NaCitr and TU reagents were mixed in the molar ratio of 1:3 (1.3 g:1.15 g), respectively. The reaction was run in water-based (25 ml) solution of both reagents in Teflon-lined stainless autoclave at 180 °C for 8 h. The formation of GQDs was indicated by the color change of the reaction solution to the orange one. Obtained postreaction solution was then given to vacuum drying at 60 °C and pressure of 200 mbar in order to remove the excess water. The final product was precipitated by adding EtOH to the solution and collected by centrifugation (13.2 rpm, 15 min), and dried at 60 °C. The obtained solid material can be easily dispersed in water.

MSNs, as a carrier for GQDs, were synthesized using the previously reported template-directed sol-gel method, with the minor modifications (Nooney et al. 2002). In a typical used synthesis procedure 0.5 g CTAB, as the structure-directing agent, was dispersed in 240 ml of deionized water (H₂O_d) in a closed vessel under vigorous stirring at room temperature. After the solution became homogenous, its pH was adjusted to 11 with the 1 M (molar) NaOH solution and heated up to 80 °C. Subsequently, 2.5 ml of TEOS as the silicon source was dropwise added and the reaction solution was further stirred continuously for 3 h, until white precipitated was obtained. As-obtained product was collected by centrifugation (24,000 rpm, 15 min) and washing with MetOH and then dried overnight at 60 °C. In order to remove CTAB template, the final dried product was refluxed for 24 h in a mixture of 160 ml MetOH and 9 ml HCl (~37%). The obtained MSNs were centrifuged and washed with MetOH and H₂O_d, and finally dried at 60 °C.

Prior to the GQDs-MSNs nanocomposites preparation, as-prepared MSNs were firstly amino-functionalized. For this purpose, 0.04 g MSNs was dispersed in EtOH and then treated with 0.8 ml APTES under the stirring. After 24 h, the suspension was centrifuged (13.2 rpm, 15 min) and washed with EtOH repeatedly for three times, in order to remove unreacted APTES. The prepared GQDs were then covalently immobilized onto amino-functionalized MSNs (NH₂-MSNs) with the use of carbodiimide crosslinker chemistry. The amino-functionalized MSNs were dispersed in PBS-based solution (pH = 5.8) of GQDs (1 mg/ml) and added with 0.2 ml of EDC linker solution (20 mg/ml). After 24 h, the unbound QDs were removed by successive centrifugation and washing with EtOH and H₂O_d.

Doxorubicin HCl (DOX, Carbosynth Limited), a model chemotherapy drug, was non-covalently loaded onto prepared GQDs-MSNs nanocomposite nanoparticles, as well as onto un-modified MSNs. Briefly, suspension of nanocarrier was mixed with DOX solution (both prepared in PBS pH = 7.4, in a weight ratio of 2:1, respectively). A resulting reaction solution was given to continuous stirring for 24 h at the room temperature and in the dark. An obtained red-purple precipitate was separated by centrifugation, washed with PBS several times, and dried in air at 60 °C. The efficiency of DOX loading was determined by finding the difference in the DOX concentration in the solution before and after loading. DOX concentration was measured spectrophotometrically at 485 nm. Percentage of the loaded drug was calculated according to the following equation:where C_{DOX 0} and C_{DOX F} are the initial and final DOX concentration in the reaction mixture.

To investigate drug release behavior from prepared GQDs-MSNs nanocomposites, as well as MSNs, 0.5 mg DOX-loaded nanocarrier was dispersed in 2 ml of PBS with pH 4.5, 5.0, 6.5, and 7.4 and given to continuous shaking under dark conditions. The temperature of PBS solution was kept constant 37 or 50 °C. After predetermined time intervals, the samples were centrifuged, then supernatant samples were withdrawn and replaced with fresh PBS for continuous drug release. The amount of released DOX was determined spectrophotometrically at 485 nm.

The morphology and size of prepared GQDs were investigated using the transmission electron microscopy (TEM) and atomic force microscopy (AFM). TEM images were recorded with a high-resolution and analytical HRTEM Jeol ARM 200F operating at 200 kV or TEM Jeol JEM-1400 operating at 120 kV. AFM images were recorded in PeakForce Tapping® mode with the Icon Bruker microscope. Silicon nitride ScanAsyst-Air probes were used during the experiment and a freshly cleaved mica as a substrate. The crystal structure was characterized by X-ray diffraction (XRD) using Empyrean (PANalytical) diffractometer with Cu Kα-filtered radiation (1.54 Å). The vibrational properties of prepared samples were analyzed via attenuated total reflection (ATR) technique with a Tensor 27 spectrometer (Bruker), equipped with a MCT detector and a horizontal triple reflection PIKE MIRacle™ ATR accessory. Raman spectra were collected using an in Via Renishaw Raman Microscopy system with a 633-nm He/Ne laser and 1800 g/mm grating. The laser light was focused on the sample with a × 50/0.75 microscope objective (LEICA). N₂ adsorption-desorption isotherms were obtained on Sorptometr Quantachrome NOVA 1000e. The surface area (SSA) and pore volume have been determined based on low temperature N₂ sorption and estimated applying Brunauer-Emmett-Teller (BET) method and Barrett-Joyner-Halenda algorithm. Zeta potential and particle size distribution of prepared GQDs were measured on Zetasizer Nano ZS (Malvern Instruments Ltd.) based on electrophoretic light scattering (ELS) and non-invasive light scattering method (NIBS), respectively. The chemical composition of prepared GQDs was investigated using X-ray photoelectron spectroscopy (XPS). XPS spectra were obtained with an Sphera II photoelectron energy analyzer (Scienta Omicron) with the monochromatized Al Kα X-ray source mounted inside the UHV system. The absorbance and fluorescence excitation and emission spectra were acquired on UV-Vis-NIR spectrophotometer (Perkin Elmer lambda 950) and spectrofluorometer (FluoroSENS Gilden Photonics). The fluorescent quantum yield (QY) of prepared GQDs solutions in water was determined by applying the following equation:where QY is the quantum yield, I is the measured integrated emission, A is the absorbance at the excitation wavelength, and n is refractive index of the solvent. As a standard, the quinine sulphate solution in 0.1 M sulfuric acid solution was used (QY = 54.6%).

For the synthesis of GQDs, the bottom-up approach was chosen, as it is reported to yield better quality GQDs concerning their morphology, size distribution, and optical properties (Bacon et al. 2014). After the hydrothermal reaction, a bluish solution with blue emission under UV light was obtained. From the HRTEM images (Fig. S1a,b in ESM), a clear interplanar distance of about 0.35 nm can be found as marked particularly in the Fig. S1b, which corresponds to that of the (002) d-spacing of graphite (Qu et al. 2013) confirming that prepared GQDs have a graphite-like nature. As presented in the HRTEM images, the crystal structure of GQDs is strongly defected; hence, it is difficult to indicate accurate hexagonal ordering typical for graphene. Prepared GQDs are uniform in size (ranging from 1.5 to 7.6 nm), with an average particles diameter of 3.65 nm (Fig. S1c in ESM). AFM images (Fig. S1 e, f in ESM) confirm that uniform GQDs were prepared, as they show a topographic height of nanostructures up to 1.15 nm. Raman spectrum of prepared GQDs reveals prominent D, G, D’, and 2D bands at 1328.6 cm⁻¹, 1575.8 cm⁻¹, 1611.2 cm⁻¹, and 2642.6 cm⁻¹, respectively. G band is a result of in-plane vibrations of sp² bonded carbon atoms, whereas D and D’ bands are due to the out of plane vibrations attributed to the presence of structural defects. Found I_D/I_G ratio is around 0.86, which indicates that sp² bonds of the carbon are disrupted by the presence of oxygen-containing functional groups (originating from the precursor), and therefore prepared GQDs are considered as highly defected materials.

The XPS elemental analysis of prepared GQDs reveals the presence of carbon, oxygen, nitrogen, sulfur, and sodium. The survey and relevant core level C 1s, O 1 , N 1s spectra and S 2s, S 2p region spectra are given in Fig. S2 in ESM. C 1s core level spectrum is composed of three peaks at around 284.5 eV, 286.0 eV, and 288.2 eV that can be assigned to the sp² C in graphene (C-C, C=C bonds), sp³ C in C-O, C-N bonds and O-C=O, respectively (Qu et al. 2013). The O 1s core level peaks at around 531.3 and 532.6 eV correspond to oxygen in states of C=O, C-O-C/C-OH, respectively (Qu et al. 2013). As shown, C=O-related oxygen states are dominating. The C=O groups are formed by the intramolecular dehydrogenation of carbon precursor, and the other C-O-C/C-OH are remaining surface and edge-related epoxy groups and carboxylic functional groups, resulting from the incomplete dehydrogenation and carbonization during the synthesis. Moreover, N 1s core level spanning from around 396.0 to 403.0 eV is clearly resolved, as well as S 2s and S 2p core levels at around 227.0 and 163.0 eV, respectively (Qu et al. 2013). Elemental XPS analysis reveals that prepared GQDs contain 72.34 at% of C, which is a high content in comparison to other results (Hao et al. 2015), 13.97 at% of O, 4.97 at% of N and 5.39 at% of S, hence confirming N and S-doping of GQDs.

These results confirm that obtained GQDs are rich in carbon, doped with N and S atoms, but still contain number of oxygen-containing functional groups and exhibit disordered and disrupted structure, what is likely for zero-dimensional (0D) nanostructures in comparison to two-dimensional (2-D) graphene nanosheets (Kozak et al. 2016).

During the synthesis of MSNs, the pH and temperature were carefully controlled in order to obtain nanostructures with size below 50 nm, as their eventual biomedical application is foreseen. As shown in the TEM images (Fig. 1a), obtained silica nanoparticles are spherical and dendritic structures, with the average particle size of 44.08 nm and quite uniform size distribution (Fig. 1b). Their average particle size slightly increased up to 49.53 nm upon GQDs immobilization (Fig. 1d, e). N₂ adsorption-desorption isotherms (Fig. 1f) also confirm that mesoporous silica nanoparticles were obtained, as they can be characterized with type IV isotherm, exhibiting a hysteresis loop. After immobilizing the GQDs onto MSNs, obtained GQDs-MSNs maintained mesoporous structure; however, the N₂ adsorption amount decreased sharply. Similarly, upon GQDs immobilization, the surface area (SSA) and pore volume decreased from 442.33 to 197.40 m²/g and from 1.06 to 0.65 cm³/g. These results confirm indirectly that GQDs were successfully immobilized onto MSNs. The broad diffraction peak at 15–30 2θ° indicates that prepared MSNs have an amorphous structure (Fig. 1c). The XRD pattern does not reveal any structural changes of MSNs upon GQDs immobilization. The mesoporous structure with high surface area and pore volume indicate that prepared GQDs-MSNs have a great potential for drug-loading efficiency.

The successful preparation of GQDs, MSNs, and GQDs-MSNs nanocomposite nanoparticles, as well as their successive modifications and loading with DOX are revealed and confirmed with the FTIR spectroscopy (Fig. 2). FTIR spectrum of GQDs (Fig. 2a) is featured IR absorption bands typical for graphene derived structures, namely (i) C-O-C stretching (1109 and 1254 cm⁻¹); (ii) phenolic -OH stretching (1395 cm⁻¹); (iii) C-C (1457 cm⁻¹); (iv) aromatic C=C (1572 cm⁻¹), but also C=O stretching (1726 cm⁻¹) of carboxylic groups (Yao et al. 2017a, b; Yao et al. 2017a, b). In the 2800–3000 cm⁻¹ region, the C-H-related bands are exhibited: aldehyde doublet, methylene asymmetric stretching, and methyl asymmetric stretching at 2829 cm⁻¹, 2845 cm⁻¹, and 2940 cm⁻¹, respectively. In turn, bands observed in the range of 3000–4000 cm⁻¹ could be attributed to hydroxyl and amine groups. The presence of -NH groups is strictly related to remaining impurities of thiourea applied during the synthesis process. This could be additionally confirmed by the presence of C-N at 1329 and 1179 cm⁻¹, as well as C-S at 690- and 650 cm⁻¹-stretching vibrations. Therefore, these IR spectra confirm that hydrothermally prepared GQDs are rich in oxygen-containing functional groups, thus have a potential to react further with NH₂-MSNs. In case of MSNs-based samples (Fig. 2b–f), all the spectra are featured by presence of Si-O-Si bending vibration, Si-O symmetric stretching, Si-OH symmetric stretching, and Si-O-Si asymmetric stretching in range of 680–695 cm⁻¹, 780–795 cm⁻¹, 945–990 cm⁻¹, and 1075–1090 cm⁻¹, respectively. Moreover, one can notice scissor-bending vibrations of adsorbed molecular water in range of 1615–1645 cm⁻¹, as well as H-bonded silanol -OH groups in range of 3365–3395 cm⁻¹ and 3635–3650 cm⁻¹. From the spectrum of bare MSNs (Fig. 2b), one can notice the presence of asymmetric -CH₂ and -CH₃ stretching vibrations at 2904 and 2979 cm⁻¹, respectively. The presence of methylene and methyl groups could be related to remaining impurities like unreacted Si-precursor TEOS or surfactant CTAB, used during the synthesis. In case of NH₂-MSNs sample (Fig. 2d), one can observe intensity decrease of the Si-O symmetric stretching vibration along with the disappearance of Si-OH symmetric stretching, which proves successful MSNs functionalization with aminosilane compound. However, the amine groups vibrations are overlapped by free or adsorbed -OH groups. Therefore, -NH₂ presence is confirmed by the broadening of the band in range of 3050–3700 cm⁻¹. Introduced amine groups were further used to covalently bind GQD to MSNs via amide bond (C-N), which is present in GQD@MSNs spectrum at 1695 cm⁻¹ (Fig. 2c), next to the bands typical for MSNs and GQD previously described. Finally, the adsorption of the doxorubicin onto MSNs was also investigated. Figure 2e shows the IR spectrum of DOX-MSNs, where the presence of drug is confirmed with the following bands: C-C stretching (1413 cm⁻¹), -OH bending (1444 cm⁻¹), aromatic C=C vibrations (1516 and 1580 cm⁻¹), N-H bending (1611 cm⁻¹), as well as C=O stretching (1726 cm⁻¹) of carboxylic groups. All of the doxorubicin-related bands can be also observed in the spectrum of DOX-GQDs-MSNs (Fig. 2f). The positions of the peaks are slightly up-shifted (+ 5 cm⁻¹), and their intensities are decreased. Moreover, the contribution of typical bands for graphene structures diminishes upon DOX loading, which indicates the stronger interaction between GQDs and DOX molecules. Nevertheless, these results confirm successful formation of designed drug delivery system.

Due to the foreseen biomedical application of prepared GQDs-MSNs nanocomposite nanoparticles, they should form stable suspensions in water-based media. Therefore, the zeta potential was controlled at each stage of their preparation, as well as after surface functionalization and further doxorubicin drug attachment. The results are given in Table 1. Both prepared bare GQDs and MSNs exhibit moderately high zeta potential in water-based suspension and are negatively charged, − 24.2 and − 30.0 mV, respectively. This is in agreement with reported isoelectric point and zeta potential of MSNs (Wu et al. 2013a) and of GQDs (Wu et al. 2013b). The negative charge is resultant of number of oxygen-containing functional groups at both GQDs and MSNs, in agreement with FTIR spectroscopy. After modifying with amino groups, the zeta potential of NH₂-MSNs changed to the positive one with higher value of + 34.1 mV. In turn, after covalent immobilization of GQDs, the zeta potential remained positive and exhibited value of + 31.1 mV for GQDs-MSNs. Hence, these results confirm successful functionalization and preparation of GQDs-MSNs nanocomposite nanoparticles. Their high zeta potential is an advantage, as they are considered to form relatively stable suspensions. Moreover, the positive charge of GQDs-MSNs will enable their efficient cellular uptake via the electrostatic interaction with negatively charged cell membranes (Frohlich 2012).

Surface passivation with carbonyl groups and heteroatom doping, revealed by FTIR and XPS, find the reflection in the optical properties, as they change the electronic density of bulk semiconductor materials (Zhu et al. 2012; Permatasari et al. 2016). The UV-Vis absorption and photoluminescent (PL) excitation and emission properties were investigated using water-based suspensions of both GQDs and GQDs-MSNs (Fig. 3). The PL quantum yield of GQDs for the emission at 440 nm was calculated to be 39.5%. In the UV-Vis absorbance spectra of GQDs, a strong absorption peak at 330 nm is observed, which corresponds to the n-π* transition of the conjugated C=O bonds in carbonyl groups present at the surface and edges of prepared GQDs and of possible C=N bonds (Qu et al. 2014), the presence of which was revealed by XPS and FTIR. By the contrast, the prepared GQDs-MSNs suspension reveal high absorbance background of decreasing intensity with increasing wavelength. The PL excitation spectra (Fig. 3b, d) are consistent with the absorbance spectra, as their maxima overlap with the absorbance maxima. The highest PL emission intensity is obtained for the excitation between 340 and 360 nm (Fig. 3a, c). GQDs and GQDs-MSNs solutions exhibit the blue PL emission with the maximum peak located at around 440 nm under UV excitation from 320 to 360 nm. GQDs-MSNs solution reveals significantly lower PL emission intensity. Maximum intensity of PL emission is observed under excitation wavelength of 360 and 340 nm for GQDs and GQDs-MSNs, respectively. The Stokes shift is therefore about 80 nm for GQDs and 95 nm for GQDs-MSNs. Under excitation above 400 to 550 nm the maximum PL emission peak shifts towards longer wavelengths from 440 to about 600 nm, along with the intensity decrease. Therefore, the PL emission of prepared GQDs and GQDs-MSNs can be considered as excitation-dependent, with three main PL regions (blue emission for the excitation up to 400 nm, green emission for the excitation below 480 nm, and red emission for the excitation below up to 550 nm). This effect will be particularly presented in the bioimaging studies. The excitation-dependent PL emission has been previously observed in GQDs, and it was related to the presence of GQDs with different sizes or conjugated sp² domains (Sk et al. 2014). Therefore, we can conclude that the blue emission results from the chromophores related to C=O bonds (observed in absorbance spectrum). The green and red emission therefore must have an another chromophore origin, possibly related to the C=N bonds (Qu et al. 2014), but also due to the presence of different sized GQDs (Sk et al. 2014). This is also well reflected in the PL excitation spectra for the emission above 420 nm, which reveal more than one excitation maximum, dominating one at 330 nm (as in absorbance spectra), but also excitation peaks about 370 nm, in the dominating peak shoulder. This second excitation contribution is more pronounced (increasing for the emissions at longer wavelengths) in case of GQDs-MSNs. This suggests the scenario of forming clusters of GQDs at the MSNs surface, which retain the fundamental optical properties of the individual GQDs, however with an influence on their PL emission intensity in the water-based solutions. Effect of the clustering of QDs on their optical properties and resulting application as bioimaging labels has been investigated, e.g., for CdSe/CdS quantum dots and quantum rods (Rafipoor et al. 2015) or more recently in ZnAgInSe/ZnS QDs (Deng et al. 2017).

To study the anticancer drug-loading and release efficiency, doxorubicin hydrochloride was non-covalently loaded onto prepared GQDs-MSNs nanocomposite nanoparticles, as well as onto un-modified MSNs as a control drug carrier. The advantage of the adsorption interaction due to the high surface area of mesoporous carrier, as well as electrostatic attraction was exploited for this purpose. The DOX-loading efficiency (% DOX_loaded) was estimated spectrophotometrically to be 53.93 (which corresponds to the loading capacity of 269.7 μg/mg) and 78.88% (which corresponds to the loading capacity of 399.4 μg/mg) for GQDs-MSNs and MSNs, respectively. As expected, the drug-loading efficiency was higher in case of control un-modified MSNs, what can be assigned to their high surface area and pore volume, but also to the efficient electrostatic attraction between positively charged DOX molecules and negatively charged MSNs. When considering DOX-loading onto GQDs-MSNs, the drug-loading efficiency was reduced, firstly by their lower specific surface area and pore volume, which were blocked by the GQDs immobilized onto MSNs. Secondly, the drug-loading efficiency through the electrostatic attraction is unfavorable to take place owing to the positive charge of GQDs-MSNs. Results of the zeta potential measurements (see, Table 1) show that due to the DOX-loading the electrophoretic potential of DOX-GQD@MSN suspension does not change significantly and remain positive (+ 30.8 mV). Whereas for DOX-MSNs, the significant zeta potential decrease is observed; however, it remains negative value (− 6.9 mV). These results indicate that water-based suspensions of prepared DOX-GQDs-MSNs exhibit better stability than DOX-MSNs; hence, these nanocomposite nanoparticles can be considered as more preferred carrier for drug delivery application.

The drug release behavior was investigated in response to different pH (4.5, 5.0, 6.5, 7.4) and temperature (37 and 50 °C) in PBS-based release solution. The acidic and neutral pHs of a release solution were chosen as they reflect best the tumor environment (endosomes and lysosomes of cancer cells) and a physiological environment in normal tissue, respectively (Upreti et al. 2013). Figure 4 shows the cumulative DOX release from DOX-GQDs-MSNs and DOX-MSNs under different conditions. All profiles exhibit gradual release over the time, without the rapid initial DOX release. The release of DOX was fastest in the medium with pH 4.5 and 5.0 and at 50 °C for both prepared carriers. Obviously, the release of DOX was more efficient in case of DOX-MSNs, and 90.02 and 89.12% of loaded drug was released within 48 h at 50 °C, respectively at pH 4.5 and 5.0. By contrast, DOX-GQDs-MSNs exhibited slower release rate, and 36.55 and 28.42% of loaded drug was released within 48 h at 50 °C, respectively at pH 4.5 and 5.0. This can be ascribed to the lower drug-loading efficiency, but also to the stronger type of interaction between DOX molecules and GQDs-MSNs, as shown with FTIR analysis, in comparison to the weak adsorption and electrostatic attraction in case of DOX-MSNs. Therefore, also the temperature effect was more significant in case of DOX-MSNs, where the temperature increase could weaken the interaction and facilitates DOX molecules release. It can be also concluded that both DOX-GQDs-MSNs and DOX-MSNs carrier exhibit a pH-dependent drug release behavior, as the DOX release can be triggered by acidic conditions, due to the weakening of the interaction between DOX molecules and GQDs-MSNs and MSNs. Hence, the drug delivery and release in cancer cells could be enhanced by the use of proposed nanocarriers, and further could result in diminishing the side effects of DOX. Although the low DOX-release efficiency is observed for DOX-GQDs-MSNs, the released amount of drug is enough to trigger the cytotoxic effects and what will be further presented.

In order to estimate the cytotoxicity of GQDs, MSNs as well as GQDs-MSNs nanocomposite nanoparticles, the WST-1 assay was firstly performed. This colorimetric assay is based on the cleavage of a tetrazolium salt by mitochondrial dehydrogenases forming formazan in viable cells. Therefore, the greater the amount of formazan is produced following the addition of WST-1, the greater the number of metabolically active viable cells. Figure 5a, b shows the cell viability of cervical cancer cells (HeLa) and normal human fibroblasts (MSU1.1) after incubation for 24 h with a series of concentrations of GQDs, MSNs, and GQDs-MSNs, also loaded with doxorubicin. The results show that GQDs, MSNs, and GQDs-MSNs only slightly affect the viability of both cells type, which remained above 90%. Only the highest concentration (250 μg/ml) of nanoparticles was more deleterious to cells; however, even in this case, the viability remained above 80%.

The loading of doxorubicin onto MSNs or GQDs-MSNs generally results in a decrease of cells viability in a concentration-dependent manner. In case of HeLa cells, the decrease after incubation with DOX-MSNs proceeds very rapidly from the 25 μg/ml nanoparticles concentration, and more than 80% of cells are indicated as death. Whereas, for DOX-GQDs-MSNs-treated cells a marked decrease in viability is observed at higher concentration of 250 μg/ml. This is in agreement with the presented above DOX-release profile, showing that the amount of released DOX from DOX-GQDs-MSNs is much lower than from DOX-MSNs, therefore justifying the need for higher nanocarrier concentration in this case, but not a higher dose of DOX itself.

In comparison, for human fibroblasts, the viability decrease is milder; however, for DOX-MSNs-treated cells, the 50% inhibition of the cell viability is observed already at the concentration of 10 μg/ml and then decline gradually. The cytotoxicity of DOX-GQDs-MSNs initially seems to be lower, but ultimately at higher concentrations (from 50 μg/ml) reaches higher value than for DOX-MSNs.

The WST-1 results on the cytotoxicity of prepared nanomaterials, reflecting the metabolic activity of cells, were then compared to plasma membrane integrity of cells, indicated by the intracellular esterase activity with the LIVE/DEAD assay (Fig. 5c, d). Here, two dyes were used, which quickly discriminates live from dead cells by simultaneous staining with green-fluorescent calcein-AM, indicating intracellular esterase activity and red-fluorescent ethidium homodimer-1, indicating a loss of plasma membrane integrity. In addition, the fluorescence images acquired with the InCell Apparatus (Fig. 5e) give an overview on the cell viability but also proliferation. In case of nanoparticles without doxorubicin, obtained results are comparable with those from WST-1 assay. The only difference is observed in case of fibroblasts, where at the highest concentration of MSNs and GQDs-MSNs the viability does not exceed 70%. Moreover, evaluation of DOX-MSNs and DOX-GQDs-MSNs cytotoxicity based on the cell membrane integrity analysis shows that these drug carriers are not as cells deleterious as it was presented above with the metabolic activity analysis. It is presented (particularly for fibroblasts), that even at the highest applied doses of 250 μg/ml, the viability was maintained at 80 and 60% for DOX-MSNs and DOX-GQDs-MSNs, respectively. However, as presented in Fig. 5e, the treatment with DOX-loaded nanoparticles has a strong effect on the cell proliferation, which is particularly the case of DOX-MSNs, in comparison to control non-treated cells. Therefore, found with these two assays differences in the cytotoxicity of prepared nanoparticles, arise from the fact that different aspects of cell functions were analyzed, cell viability and cell vitality, by live/dead assay and WST-1, respectively (Kwolek-Mirek and Zadrag-Tecza 2014). However, both are required for the estimation of the physiological state of a cell after exposure to various types of stressors, including nanoparticles. Whereas, the fluorescent staining distinguishes live from dead cells, the WST-1 assay gives the information about their ability to reproduction. Hence, the lower number of living cells upon DOX-MSNs and DOX-GQDs-MSNs indicates the action of successfully delivered and released doxorubicin. Further, both assays confirmed that conjugation of biocompatible GQDs with nontoxic MSNs, leads to the creation of low-cytotoxic GQDs-MSNs nanocomposite nanoparticles, making them particularly good candidates for drug-delivery system.

The use of GQDs with their PL properties allows to monitor cellular uptake and distribution of GQDs-MSNs nanocomposite nanoparticles directly. Moreover, the GQDs entrapment onto mesoporous silica nanoparticles structure resulted in the increasing fluorescent signal, especially in the red channel in comparison to bare GQDs (Fig. S3 in ESM). The intensification of obtained fluorescence signal could be related with immobilization and aggregation of GQDs onto MSNs, of what was previously suggested (Rafipoor et al. 2015; Deng et al. 2017), but also to their more effective penetration into cells, enabled by their preferred positive surface charge for the interaction with negatively charged cell membrane (Frohlich 2012). To demonstrate the fluorescence imaging performance of GQDs-MSNs nanocomposite nanoparticles, the in vitro cellular uptake experiments were performed using HeLa cells. Cells were exposed to 50 μg/ml of nanocomposite nanoparticles for 4 h and afterwards analyzed by CLSM. The results are presented in Fig. 6. Here, the excitation-dependent PL behavior of the GQDs can be observed. Confocal images of HeLa cells incubated with nanoparticles upon 405 nm excitation exhibit the fluorescence emission in blue region, which could be observed as signals inside the cells. When the excitation light changes to 488 nm, green fluorescence is observed; finally, upon 559 excitation, the red fluorescence is emitted. A similar effect was observed by Zhu et al. (Zhu et al. 2011) for GQDs, which however emitted green and yellow fluorescence. According to the merged images, GQDs-MSNs after 4 h of co-incubation were internalized by HeLa cells and localized in the cytoplasm. The internalization was clearly confirmed in the 3D images (Fig. 6b). It has been already proved that GQDs can be easily internalized into cells (Markovic et al. 2012; Yuan et al. 2014; Schroeder et al. 2016). Likewise, Lu et al. (Lu et al. 2007) demonstrated that MSNs nanoparticles are readily endocytosed by cells and usually enter cells in an energy-dependent manner.

In order to evaluate the exact uptake mechanism of GQDs-MSNs, the inhibition of selected routes were performed. Results given in Fig. S5 (ESM) indicate that by lowering the temperature to 4 °C, the energy-dependent mechanisms is blocked. This is observed by lacking signal in cells, is comparison to control HeLa cells treated with 20 μg/ml GQDs-MSNs for 3 h at 37 °C, where the signal in cells is clear. Therefore, the endocytosis is considered as an exact uptake mechanism of studied nanoparticles. However, by using suitable inhibitors, different types of endocytosis were further investigated. Chlorpromazine is reported to specifically inhibit clathrin-dependent endocytosis (CDE), while methyl-β-cyclodextrin has been extensively used to inhibit clathrin-independent endocytosis (CIE), especially in caveolae/lipid raft-mediated endocytosis (Le et al. 2002; Parton and Richards 2003, dos Santos et al. 2011). Wortmannin, in turn, as covalent inhibitor of phosphoinositide 3-kinases (PI3Ks), plays a role in inhibition of micropinocytosis (Araki et al. 1996; Rupper et al. 2001). As can be seen in Fig. S5, chlorpromazine showed no inhibitory capacity in HeLa cells on studied nanoparticles uptake. Contrary to chlorpromazine, the MβCD inhibited the uptake of GQDs-MSNs into cells completely, suggesting the CIE as a main route of these nanoparticles internalization. Similar results were presented by Ekkapongpisit et al. (Ekkapongpisit et al. 2012), contrary to Hao et al. (Hao et al. 2012), who indicated that spherical mesoporous silica nanoparticles are internalized rather via the clathrin-mediated pathway. Interestingly, in our study, the wortmannin inhibits nanoparticles internalization partially, indicating that maybe larger nanoparticles or nanoparticles aggregates are internalized by micropinocytosis (Meng et al. 2011; Vollrath et al. 2013).

Additionally, to confirm that the uptake of GQDs-MSNs nanocomposites nanoparticles is mediated through the endosome-lysosomal mechanism, TEM analysis was performed (Fig. 7). Obviously, the GQDs-MSNs are uptaken by HeLa cells via their encapsulation into vesicular compartments. The beginning of the uptake occurs by the initiation of plasma membrane invagination, then the nanocomposite nanoparticles are transported to the early endosomes (Fig. 7a), and finally, they are found to be clumped in lysosomes (Fig.7c). Moreover, as presented in magnified images, GQDs-MSNs maintain the spherical and porous morphology. The abovementioned results indicated that the GQDs-MSNs possess both strong enough fluorescence emission and internalization ability, what makes them a promising material for bioimaging and biolabeling.

Finally, the cell uptake and DOX-MSNs and DOX-GQDs-MSNs and intracellular DOX release after 3 and 24 h of incubation with HeLa cells was also monitored with CLSM. As presented in the Fig. 8, after 3 h of incubation red fluorescence signals from DOX molecules in the DOX-MSNs and GQDs-MSNs nanoparticles can be observed mainly in the cytoplasm, suggesting the endocytosis of the nanoparticles by HeLa cells and DOX release. The observed bright red dots (marked with arrows in Fig. 8) are related to the aggregated carriers, whereas the homogenously spread red fluorescence signal throughout the cytoplasm is related to DOX molecules, which have been released from the nanoparticles within the acidic endosome and have diffused into the cytoplasm. Regarding the DOX-GQDs-MSNs, the use of fluorescent GQDs enables to monitor simultaneously the uptake of the DOX-GQDs-MSNs through channels corresponding to GQDs (mainly blue) and to DOX (red). This is consistent with the imaging results for HeLa cells treated with un-loaded GQD@MSNs presented in Fig. 6. Further incubation up to 24 h with higher dose of DOX-GQDs-MSNs (200 μg/ml) resulted in a marked increase in the releasing of DOX as observed with stronger fluorescent signal (Fig. S6), and these results are in accordance with the DOX-loading and release efficiency. Moreover, cytotoxicity studies indicated that the cytotoxic effect of DOX-GQDs-MSNs is observed for doses above 100 μg/ml in case of HeLa cells. Therefore, in case of DOX-GQDs-MSNs, use of higher doses is necessary to induce effects similar to those caused by DOX-MSNs.

In order to investigate the colocalization of the GQDs-MSNs nanocomposite nanoparticles with DOX, the analysis was performed using the Coloc2 tool and the Pearson’s correlation coefficients were found. These coefficients are 0.77 and 0.44 for cells incubated with the 20 μg/ml of DOX-GQDs-MSNs for 3 and 24 h, respectively. In case of the cell exposure to 200 μg/ml of DOX-GQDs-MSNs for 3 and 24 h, found coefficients were 0.39 and 0.37, respectively. These results indicate that colocalization occurs and is more significant in case of shorter time of incubation, whereas longer time of incubation results in the decreased colocalization coefficient. This confirms the time-dependent release of DOX from the GQDs-MSNs nanocomposite nanoparticles.

Summarizing, DOX-GQDs-MSNs nanocomposite nanoparticles are found to be efficient in DOX delivery via their internalization in HeLa cancer cells but also allow for simultaneous real-time optical tracking of the DOX during its delivery and release.