Biotin-decorated fl uorescent silica nanoparticles with aggregation-induced emission characteristics : fabrication , cytotoxicity and biological applications †

Biotin-decorated fluorescent silica nanoparticles (FSNPs) were successfully fabricated by a sol-gel reaction of silole-functionalized siloxane followed by a sequential reaction with tetraethoxysilane, (3-aminopropyl)triethoxysilane and biotin. The FSNPs were uniformly sized, spherical in shape and monodispersed. While their silole precursor was non-emissive in solution, the suspension of the FSNPs emitted strong green light upon photoexcitation due to the aggregation-induced emission characteristics of the silole aggregates in the hybrid nanoparticles. Morphology study and cell viability, trypan blue exclusion, Annexin V-FITC/PI apoptosis and ROS generation assays showed that the FSNPs showed low toxicity to living cells. The FSNPs worked as fluorescent visualizers for selective imaging of the cytoplasm of tumor cells with over-expressed biotin receptors. The fluorescent nanoparticles were lastingly retained inside the living cells, thus enabling long-term tumor cell tracking over multiple passages and quantitative analysis of tumor cell migration.


Introduction
Fluorescent silica nanoparticles (FSNPs) are promising materials for bioanalysis and bioimaging as well as for disease diagnosis and therapy because of their characteristic features including biocompatibility, inertness, hydrophilicity, size tuning, ease of surface modication, etc. 1 Since hundreds of uorescent molecules are encapsulated in a single nanoparticle, the detection sensitivity of FSNPs is much higher than that of direct uorophore labeling. To date, FSNPs have been used widely in DNA microarrays, 2 immunouorescence technique, 3 analogue luminescence detection, 4 tissue imaging, 5 cell imaging 6 and especially tumor cell imaging. 1b,5,7 As silica nanoparticles have been found to exhibit no remarkably toxic effect on living cells, 8 they thus possess the potential prospect in tumor cell imaging for early diagnosis and cell tracking.
Conventional dyes have been used for the fabrication of FSNPs. 9 These uorophores emit strongly in solutions but become weakly emissive or non-luminescent in the solid state due to the strong p-p stacking interactions, which promote the formation of detrimental species such as excimers or exciplexes. 10 The weak emission offered by the FSNPs has resulted in poor sensitivity in uorescence sensory systems, especially in bioassays of trace amounts of biomolecules. The sensitivity cannot be enhanced by using higher uorophore concentrations due to the aggregation-caused quenching (ACQ) effect. The development of inorganic quantum dots can surmount these disadvantages but create new problems such as difficult synthesis, limited variety and high cytotoxicity. 11 We observed a novel phenomenon of aggregation-induced emission (AIE) that is exactly opposite to the ACQ effect in a series of propeller-like molecules. Instead of quenching, aggregate formation has boosted their uorescence quantum yields by more than 300-fold, turning them from weak uorophores in the solution state to strong emitters in the solid state. 10,12 The restriction of intramolecular rotation in the aggregated state has been proposed as the main cause for the AIE effect. 10, 13 Because of their efficient solid-state emission, AIE luminogens are thus ideal uorophores for the fabrication of FSNPs. In our previous work, we succeeded in creating FSNPs hybridized with AIE luminogens by a one-step surfactant free sol-gel process. These highly emissive FSNPs pose good colloidally stability and are non-toxic to living cells, thus making them promising uorescent visualizers for intracellular imaging. 14 By modifying the fabrication procedure, silica nanoparticles with both strong light emission and high magnetization are prepared, which can function as selective cell imaging bioprobes and immobilize bovine serum albumin efficiently. 15 To widen the applicability of the AIE-loaded silica nanoparticles, their selectivity should be further improved. It is known that conjugation of targeting ligands such as proteins, peptides and aptamers on the surface of nanoparticles can enhance their binding affinity and facilitate receptor-mediated internalization, 16 thus enabling selective targeting and efficient intracellular uptake. Biotin (vitamin B7 or vitamin H), a member of the vitamin family for growth promotion at the cellular level, is one of the most common tumor recognition moieties because tumor cells require extra biotin to sustain their rapid growth and compared to the normal cells, biotin-specic receptors are thus generally over-expressed on their surface. 17 Thus, biotin has attracted particular interest in cancer-targeting drug carrier and uorescent nanoparticles for specic tumor cell targeting. For example, Panyam and colleagues prepared a biotin-functionalized drug delivery carrier, which signicantly improved the therapeutic efficacy of tumors. 18 Kwon also modied gold nanoparticles with biotin and rhodamine B to interact selectively with cancer cells for diagnosis and therapy. 17d However, examples of AIE-active silica nanoparticles functionalized with biotin have not been described.
In this paper, we aim to synthesize such nanoparticles for specic living tumor cell imaging. A series of experiments had been carried out to evaluate the toxicity or biocompatibility of the synthesized nanoparticles. The biotin-decorated FSNPs emitted strong photoluminescence (PL) under UV excitation and interacted selectively with tumor cells with over-expressed biotin receptors and internalized into cytoplasmic regions. They can stay inside the live cells for long periods of time, thus enabling long term cell tracing over multiple passages as well as analysis of tumor cell migration quantitatively.

Fabrication of biotin-decorated FSNPs
With a view to fabricate FSNPs with strong light emissions in the solid state and biotin moieties on the surface for specic cancer cell targeting, we synthesized a silole-containing siloxane (1) by stirring a dimethylsulfoxide (DMSO) solution of 2 and (3-aminopropyl)triethoxysilane (APS) at room temperature for 24 h (Scheme 1). The sol-gel reaction of 1 followed by reaction with tetraethoxysilane afforded FSNP-1 with a coreshell structure. 14,15 Addition of APS into the reaction mixture generated FSNP-1-NH 2 with numerous amino groups decorated on the surface, enabling it to undergo amidation reaction with biotin in the presence of 1,3-dicyclohexylcarbodiimide (DCC) and 4-(dimethylamino)pyridine (DMAP) to furnish FSNP-1biotin.

Structural characterization
We rst characterized FSNP-1-biotin by IR spectroscopy. The IR spectrum of FSNP-1-biotin is given in Fig. S1 in the ESI; † for comparison; the spectra of FSNP-1-NH 2 and biotin are also provided in the same gure. The Si-OH, Si-O and N-H stretching vibrations of FSNP-1-NH 2 occurred at 951, 1707 and 3321 cm À1 , respectively. Aer biotin modication, the absorption at 3321 cm À1 was enhanced and new peaks associated with C]O stretching vibrations emerged at 1638 and 1707 cm À1 in FSNP-1-biotin, revealing that biotin was covalently graed on the surface of FSNP-1-NH 2 through amidation reaction. Analysis by transmission electron microscopy (TEM) showed that both FSNP-1-NH 2 and FSNP-1-biotin were spherical in shape and uniform in diameters with narrow size distributions ( Fig. 1). Compared to FSNP-1-NH 2 , FSNP-1-biotin possessed a much rougher surface. The average particle size of FSNP-1-NH 2 was measured to be 48.19 AE 2.82 nm, while that of FSNP-1biotin was slightly larger (50.50 AE 2.91 nm). Similarly, the images from a scanning electron microscope (SEM) given in Fig. S2 † showed that the surface morphology of FSNP-1-NH 2 altered little aer the biotin conjugation. To realize the composition of the nanoparticles, X-ray photoelectron spectroscopy (XPS) and energy dispersive X-ray spectroscopy (EDX) were carried out and the results are summarized in Table S1. † XPS analyses revealed no sulfur atom on the surface of FSNP-1-NH 2 , while a trace amount of sulfur element (0.3%) was detected in FSNP-1-biotin. Similar results were also obtained from the EDX analysis, proving that biotin is successfully conjugated on the surface of FSNP-1-NH 2 . The thermal stability of the FSNPs was investigated by thermogravimetric analysis (TGA). As shown in Fig. S3, † FSNP-1-NH 2 possessed high thermal stability and started to degrade at a temperature of $300 C. Even when heated up to 800 C, more than 70% of its Scheme 1 Fabrication of biotin-decorated fluorescent silica nanoparticles. weight was retained. FSNP-1-biotin was also thermally quite stable and degraded at a similar temperature with high residual yield at 800 C. Since biotin decomposed completely at 650 C, the amount of biotin graed on FSNP-1-biotin could be calculated as the weight difference between the thermograms of FSNP-1-NH 2 and FSNP-1-biotin at this temperature and was equal to 6.92 wt%.

Light emission
The PL spectra of a solution of 1 and suspensions of FSNP-1-NH 2 and FSNP-1-biotin in ethanol are displayed in Fig. 2. Nearly no uorescence signals were recorded when the ethanol solution of 1 was photoexcited. In the solution state, the multiple peripheral phenyl rings in the isolated molecules of 1 undergo active intramolecular rotation, which effectively annihilates their excited states and hence renders the luminogen nonemissive. 10 When the molecules of 1 are covalently incorporated into and aggregate in the silica networks of FSNP-1-NH 2 and FSNP-1-biotin, strong PL spectra peaked at $490 nm are recorded. Evidently, 1, similar to its congener 2, is AIE-active. The rigid silica network largely restricts the intramolecular rotations of the luminogen. This blocks the nonradiative relaxation channels and populates the radiative excitons, 10,12a,19 thus making the FSNPs highly luminescent. Under the same measurement conditions, the PL intensities of FSNP-1-NH 2 and FSNP-1-biotin were 95-and 87-folds higher than that of 1 in an ethanol solution, respectively. The photographs of a solution of 1 and the suspensions of FSNP-1-NH 2 and FSNP-1-biotin taken under UV exposure from a hand-held UV lamp are shown as an inset in Fig. 2. While intense light was emitted from FSNP-1-NH 2 and FSNP-1-biotin, the solution of 1 was invisible under the UV illumination. This visual observation further substantiates that the intramolecular rotation of 1 is restricted by its covalent melding with the silica matrix. The light emission was very stable, with no detectable change in the PL spectrum aer FSNP-1-biotin was put on shelves for 6 months without protecting from light and air ( Fig. S4 †).

Cytotoxicity
To nd useful biological applications, a luminogen should neither inhibit nor promote the growth of living cells. To evaluate the toxicity of FSNP-1-biotin, we rst studied the morphology change of HeLa and mouse broblast NIH 3T3 cells cultured in the presence of FSNP-1-biotin. As shown in Fig. S5, † only a small amount of cells showed vacuole shrinkage and chromatin condensation, suggesting that FSNP-1-biotin possessed good biocompatibility. A cell viability assay was also carried out at nanoparticle concentrations of 0.1, 1, 10, 100 and 200 mg mL À1 for 48 h using a WST-8 cell counting kit. The cell viability decreased with an increase in the nanoparticle concentration. Compared with the control group, the nanoparticles had no discernible deleterious effects on the viability of both HeLa and 3T3 cells (P > 0.05) at concentrations below 100 mg mL À1 (Fig. 3A). At 100 mg mL À1 , the viability decreased to 81.60% and 85.55% for HeLa cells and 3T3 cells (P < 0.05), respectively. To further verify the cytotoxicity of FSNP-1-biotin, the cell survival was examined by the trypan blue exclusion assay. 8a The viable cells exclude the trypan blue dye and are not stained to identify living cells. At a nanoparticle concentration of 100 mg mL À1 , the cell survival of HeLa cells and 3T3 cells was reduced to 88.49% and 86.74% (P < 0.05), respectively, which echoed in the foregoing data (Fig. 3B). Consequently, the cells were exposed to up to 80 mg mL À1 of FSNP-1-biotin for the experiments aerward in accordance with the preliminary cytotoxicity effect and requisite nanoparticle concentration for tumor cell imaging.
Cell apoptosis is an important parameter for the toxicity of nanomaterials. Nanoparticles could induce apoptosis through  many kinds of pathways. The apoptotis-related proteins, such as Bcl-2, Bax, CytC and Caspase-3, were up-regulated when HepG2 cells were exposed to silica nanoparticles. 20 Herein, the quantication of apoptosis of HeLa and 3T3 cells was studied by ow cytometry (FCM) with Annexin V-FITC/propidium iodide (PI) double staining assay at different concentrations of FSNP-1biotin. The cells stained with Annexin V-FITC alone represent early apoptosis, while those labelled with PI demonstrate necrosis. The late apoptotic cells are stained with both uorescent dyes. As depicted in Fig. 4, FSNP-1-biotin could induce apoptosis in a dose-dependent manner. Compared with the control group, the number of apoptotic cells increased with increasing nanoparticle concentration. In the presence of 80 mg mL À1 of nanoparticles, obvious apoptosis was observed in both HeLa and 3T3 cells (P < 0.05).
The oxidative stress-mediated pathway is demonstrated as one of the apoptotic mechanisms due to the induction of nanomaterials. 21 Nanoparticles can induce production of intracellular reactive oxygen species (ROS), which can change the permeability of a mitochondrial membrane, damage the ultrastructure of mitochondria 21a and then trigger secondary damage effects such as mitochondrial dysfunction and DNA damage. 22 The ROS assay was based on the peroxide-dependent oxidation of DCFH-DA to form a uorescent compound named dichlorouorescein. Herein, the HeLa and 3T3 cells were treated with various concentrations of FSNP-1-biotin for 24 h. Aerwards, they were washed with a buffer solution and incubated with DCFH-DA for 20 min. The uorescence from the cells was then immediately measured by a FCM. The ROS generation and hence the uorescence from the solution became higher when HeLa and 3T3 cells were cultured with an increasing amount of FSNP-1-biotin nanoparticles (Fig. 5). The difference in the ROS generation between the control and treated cells was small at a nanoparticle concentration of 40 mg mL À1 but was obvious at 80 mg mL À1 (Fig. 5C). However, when compared to the positive (Rosup) control, the extent of ROS generation was not large at such nanoparticle concentrations. On the other hand, the intuitive uorescent images shown in Fig. 5D also According to the emission intensity from the fluorescence dye ("+" ¼ high, "À" ¼ low), four regions, named Q1 (Annexin V-FITC À, PI +), Q2 (Annexin V-FITC +, PI +), Q3 (Annexin V-FITC À, PI À) and Q4 (Annexin V-FITC +, PI À) were divided, which represented the necrotic, late apoptotic, living and early apoptotic cells, respectively. (B) Quantitative analysis of cell apoptosis from three independent experiments. The total apoptotic cells (Q2 + Q4) were shown in the histograms, analyzed by Dunnett's test and compared with the control (0 mg mL À1 ). * Means P < 0.05 and indicates that the total number of apoptotic cells is significantly different from that of control. indicated that the nanoparticles induced ROS generation when the cells were exposed to a high concentration of FSNP-1-biotin.

Tumor cell targeting and long-term cell tracking
In our previous study, we found that FSNPs hybridized with AIE luminogens can be utilized as uorescent visualizers for intracellular imaging. 14 Would FSNP-1-biotin show similar or even better performance? To test this, the targeting property of FSNP-1-biotin was investigated using HeLa and BEL-7402 cells, which are typical cell lines for cervical carcinoma and hepatocellular carcinoma, respectively, with over-expressed biotin receptors. Normal liver cells LO2 containing less-expressed receptors were also used in the investigation for the purpose of comparison. Aer 3 h staining, strong PL was emitted from the HeLa and BEL-7402 cells, while dim uorescence was observed from LO2 cells (Fig. 6). This phenomenon should be attributed to receptor-mediated endocytosis. 23 For tumor cells with overexpressed biotin receptors on the surface, FSNP-1-biotin binds the cell membrane via ligand-receptor interaction (Scheme 2). The interaction of the nanoparticles with the biotin receptor triggers cell internalization into intracytoplasmic vesicles or the formation of clathrin-coated vesicles. The nanoparticles may be further processed in vacuoles and endosomes, which are then eventually released to the cytoplasm. On the other hand, the nanoparticles without biotin coating enter the cells mainly by caveolae-dependent endocytosis, whose rate, specicity and affinity are much lower than those of clathrin-dependent endocytosis. 23 The uptake of FSNP-1-biotin nanoparticles by LO2 cells is low due to the absence of biotin receptors on their surface. Moreover, they, unlike HeLa and BEL-7402 cells, require a small amount of biotin for proliferation, thus resulting in low nanoparticle uptake and weak uorescence.
To prove the existence of ligand-receptor interactions or the occurrence of receptor-mediated endocytosis, uorescence imaging of tumor cells by FSNP-1-biotin was carried out in the presence of free biotin. As shown by the images given in Fig. 7, the uorescence from the HeLa and BEL-7402 cells was markedly decreased when they were pretreated with a 10 mg mL À1 biotin solution prior to staining. This result suggests that the free biotin molecules competitively bind to the biotin receptors on the surface of tumor cells and interdict the receptor-mediated endocytosis of FSNP-1-biotin. Evidently, bio-conjugation of FSNP-1-NH 2 with biotin molecules has enhanced the targeting efficiency and endocytosis of the resulting FSNP-1-biotin.
The uptake of nanoparticles by cells involves several different pathways such as nonspecic diffusion, phagocytosis and receptor-mediated or uid phase endocytosis. 24 To conrm the uptake and distribution of FSNP-1-biotin, we analyzed the HeLa cells treated with FSNP-1-biotin for 12 h by TEM. The nanoparticles were distinguished based on their diameters and spherical shape. As displayed in Fig. S6, † the nanoparticles were  internalized into cells via receptor-mediated endocytosis just as depicted in Scheme 2. This facilitates their uptake and internalization in the cells (Fig. S6A †). The nanoparticles were then released into the cytoplasm (Fig. S6B †) and distributed as isolated species or clusters (Fig. S6C †). No nanoparticles were found in the nucleus, which is consistent with the literature results, and it was presumably due to the large size of the nanoparticles. 25 In our previous work, we found that a cytophilic uorescent bioprobe fabricated from 1-decyl-1-methyl-2,5-bis{4-[(N,Ndiethylamino)methyl]phenyl}-3,4-diphenylsilole could track living cells for four passages, while almost no PL signals were detected in cells labelled with MitoTracker Green FM, a commercial dye, aer two generations. 26 As an excellent tumor targeting probe, FSNP-1-biotin should also possess the property of high photostability and long-term tracking. Thus, the uorescence from the HeLa and BEL-7402 cells stained with FSNP-1-biotin was investigated in a continuous manner of cell culture. The tumor cells were observed aer incubation with 40 mg mL À1 nanoparticles for 24 h. When 70-80% conuence was reached, the cells were trypsinized, counted and subcultured at a density of 2 Â 10 5 cells per well into a 6-well plate. Generally, the cells grow into another generation within 1 day. Although the PL from the cells became weaker along the passage due to the division of the nanoparticles accompanied by cellular ssion, they were still visible even aer 5 day culture (Fig. 8). Evidently, FSNP-1-biotin is a superb long-term cell tracer. No uorescence was observed in the cell nucleus, which was consistent with the TEM analysis. We prolonged the culture to 7 days. Although dim uorescence was still observed from the cells, their morphology was hard to be discernible (Fig. S7 †). Thus, the retention time of the nanoparticles inside the cell is approximately one week.

Tumor cell migration
Cell migration is an important pathological process in tumor cells and reects their ability of tissue invasiveness and metastasis. 27 In the above discussion, we found that FSNP-1biotin could selectively stain tumor cells with over-expressed biotin receptors and enable long-term cell tracing over multiple passages. With such regard, we explored the possibility of using FSNP-1-biotin to track the migration of tumor cells. Briey, different concentrations of serum were loaded to the medium in the lower compartment. FSNP-1-biotin labelled HeLa cells were then introduced into the upper compartment of a cell. Because the HeLa cells require additional nutrients to maintain their rapid proliferation, they are thus induced to move from the upper compartment to the lower one. The migrated cells were collected by trypsin and the uorescence of the cell suspension was then measured. As shown in Fig. 9A, the uorescence became stronger with an increase in the serum concentration. Since the PL intensity is associated with the number of labelled cells, this makes quantitative analysis possible. Same result was achieved by the conventional crystal violet staining method (Fig. 9B) but the uorescence-based technique was more objective and accurate.

Conclusions
In summary, biotin-decorated uorescent silica nanoparticles were successfully fabricated by a surfactant-free sol-gel reaction of silole-functionalized siloxane followed by esterication reaction with biotin. The structure of FSNP-1-biotin was investigated by IR, TEM, SEM and XPS analyses. The FSNP-1-biotin was uniformly sized, spherical in shape and monodispersed. Although their silole precursor was non-emissive, the FSNP-1biotin emitted strong green light upon photoexcitation, as a result of the aggregation-induced emission characteristics of the silole aggregates in the hybrid nanoparticles. Morphology study and cell viability, trypan blue exclusion, Annexin V-FITC/ PI apoptosis and ROS generation assays showed that FSNP-1biotin possessed low toxicity to living cells. The FSNP-1-biotin can be utilized to selectively image the cytoplasm of tumor cells with over-expressed biotin receptors. They can stay inside the living cells for a long period of time, thus enabling long term tumor cell tracing over multiple passages and quantitative analysis of tumor cell migration. These attributes make these AIE-active, low cytotoxic, strongly uorescent and photostable silica nanoparticles promising for an array of biomedical applications.

Preparation of FSNP-1-biotin
A silole-APS conjugate (1) was synthesized by stirring a mixture of 6 mmol of 1,1-dimethyl-2,5-bis[4-(2-bromoethoxy)phenyl]-3,4diphenylsilole (2) and 16 mmol of APS in 50 mL of DMSO overnight according to the procedure published in our previous paper. 15 Then, FSNP-1 was fabricated by a two step sol-gel reaction. 14 Briey, 1 was added to a mixture of ethanol (64 mL), ammonium hydroxide (1.28 mL) and distilled water (7.8 mL). The solution was then stirred at room temperature for 1 h to prepare the silole-silica nanocores, aer which a mixture of 2 mL TEOS and 8 mL ethanol was slowly added. The reaction mixture was stirred at room temperature for 3 h to coat the nanocores with silica shells. FSNP-1-NH 2 was prepared by stirring a mixture of FSNP-1 and APS at room temperature for additional 24 h. The nanoparticles were centrifuged and washed with ethanol and water. Finally, the FSNP-1-biotin was fabricated by stirring a mixture of FSNP-1-NH 2 and biotin at room temperature overnight in the presence of DCC and DMAP. The nanoparticles were washed with deionized water and ethanol to get rid of unwanted substances and dispersed in deionized water or ethanol for further experiments.

Characterization
The IR spectra were recorded on a Perkin-Elmer 16 PC FT-IR spectrophotometer. The chemical compositions of the nanoparticles were determined by XPS and EDX analyses. Their morphologies and sizes were investigated using JOEL 2010 TEM and JOEL 6700F SEM (JEOL Ltd Tokyo, Japan) at accelerating voltages of 200 and 5 kV, respectively. Samples were prepared by drop-casting a dilute suspension of the nanoparticles onto copper 400-mesh carrier grids covered with carbon-coated formvar lms and the solvent was evaporated at room temperature in open air. The elementary particle size was estimated by measuring the diameter of 100 particles. The bioconjugation of biotin on FSNP-1-NH 2 was investigated by thermogravimetric analysis (TGA). PL spectra were recorded on a Perkin-Elmer LS 50B spectrouorometer (Perkin-Elmer Corporation, USA) with Xenon discharge lamp excitation.

Cell culture
HeLa cells were cultured in minimum essential medium. BEL-7402 cells, LO2 cells and NIH 3T3 cells were grown in an RPMI 1640 medium supplemented with 10% FBS, 100 U mL À1 penicillin G and 100 mg mL À1 streptomycin in a 5% CO 2 , 90% relative humidity incubator at 37 C. Suspensions of the nanoparticles were dispersed in 0.01 M phosphate buffer solution (PBS) by an ultrasonic bath Model SB5200 (Shanghai Branson, China), diluted to various working concentrations, and then added to cells immediately. Every experiment was performed in triplicate.

Hematoxylin and eosin (HE) staining
The morphology change of cells incubated with FSNP-1-biotin was investigated by HE staining. The HeLa cells and 3T3 cells were seeded into a 12-well plate overnight, incubated with nanoparticles at a concentration of 0, 40 or 80 mg mL À1 for 48 h, xed with 4% paraformaldehyde and then stained with hematoxylin and eosin. Hematoxylin mainly stains the chromatin in nucleus and ribosome in cytoplasm into blue and eosin mainly stains components in cytoplasm into pink or red. The cellular morphology change was observed under an Olympus CKX41 phase contrast microscope (Olympus, Japan).

Cell viability assay
The viability of cells treated with FSNP-1-biotin was measured using a WST-8 cell counting kit (CCK-8) according to the supplier's instruction. Cells were seeded into a 96-well plate at a density of 8000 cells per well and exposed to various concentrations of nanoparticles for 48 h. 10 mL of a CCK-8 solution was added into each well and the cells were incubated for additional 2 h at 37 C. The optical density was measured on a microplate reader (Thermo, USA) using a test wavelength of 490 nm and a reference wavelength of 630 nm.

Trypan blue exclusion assay
To further verify the cell survival rate, a trypan blue exclusion assay was performed. The viable cells exclude trypan blue dye and are not stained to identify living cells. Aer being exposed to the nanoparticles for 48 h, the cells were harvested and scored as alive or dead using trypan blue according to the manufacturer's instruction. The number of viable cells was counted using a conventional hemocytometer. The rate of viability was derived by comparing with the negative control.
with an Annexin V-FITC/PI double staining assay. Briey, aer incubation with various concentrations of nanoparticles for 48 h, the cells were harvested with an EDTA-free trypsin solution and then treated with 5 mL Annexin V-FITC and 5 mL PI for at least 10 min at room temperature in the dark. Immediate analysis was performed using FCM (BD Biosciences, USA). For each sample, 1 Â 10 4 cells were measured.

Intracellular reactive oxygen species assay
Detection of intracellular reactive oxygen species (ROS) was based on the peroxide-dependent oxidation of DCFH-DA to form a uorescent compound named dichlorouorescein (DCF). Aer treatment with various concentrations of FSNP-1biotin or in the absence of nanoparticles as a negative control for 24 h, the cells were washed three times with PBS and incubated with 10 mM DCFH-DA at 37 C for 20 min. The cells incubated with 50 mg L À1 of a Rosup solution for 30 min were treated as the positive control. The uorescence of the cells was immediately measured using FCM (BD Biosciences, USA). For each sample, 1 Â 10 4 cells were collected. The visual image of the ROS generation in cells was taken on an Olympus BX41 inverted uorescence microscope (Olympus, Japan) at an excitation wavelength of 488 nm and an emission wavelength of 525 nm.

Tumor cell targeting
HeLa, BEL-7402 and LO2 cells were seeded on a round cover slip mounted onto a 6-well plate overnight. The living cells were incubated with a serum-free medium containing FSNP-1-biotin at a specic concentration (40 mg mL À1 ) with or without 10 mg mL À1 of a biotin solution for 3 h. The cells were then washed three times with 0.01 M PBS and imaged under a Nikon A1 confocal laser scanning microscope (CLSM, Nikon Corporation, Japan) at an excitation wavelength of 405 nm.

Cellular uptake of FSNP-1-biotin
The cellular uptake and distribution of FSNP-1-biotin in cells were analyzed by TEM (FEI Corporation, Netherlands) according to a modied procedure. 28 Briey, HeLa cells were treated with 40 mg mL À1 of FSNP-1-biotin for 12 h. Aerwards, the cells were washed three times with 0.01 M PBS to get rid of the unbound nanoparticles and xed with 2.5% glutaraldehyde buffered in 0.01 M PBS for 1 h at room temperature. Fixed cells were washed three times with 0.01 M PBS and collected in a centrifuge tube, and then post-xed in 1% osmium tetroxide for 1 h at room temperature. The sample was dehydrated by ethanol solutions at different concentrations (40, 50, 70, 80, 90, 95 and 100%), treated with propylene oxide and then embedded in Spurr's resin by inltration with a series of mixtures of resin and propylene oxide (ratios of propylene oxide to resin: 1 : 1, 1 : 2 and 1 : 3). The resin blocks were hardened at 70 C overnight. Ultrathin sections with a dimension of 70 nm were cut using glass knives and then stained with uranyl acetate and lead citrate prior to analysis under Tecnai G2 20 TEM (FEI Corporation, Netherlands) operating at 200 kV.

Tumor cell migration
The HeLa cells labelled with FSNP-1-biotin were utilized for the cell migration assay. Briey, the cells were incubated with 40 mg mL À1 FSNP-1-biotin for 24 h, aer which they were harvested, washed twice with 0.01 M PBS and resuspended in a serum-free medium at a density of 5 Â 10 5 mL À1 . Aer addition of 600 mL culture medium with 20% or 5% serum to each well of the 24well plate, 200 mL of the cell suspension was seeded into a millicell (8 mm pores). The cells were incubated for 24 h at 37 C and 5% CO 2 . Subsequently, the cell culture inserts were transferred into a fresh 24-well plate containing 500 mL of a prewarmed trypsin-EDTA solution per well for 10 min at 37 C. 100 mL of the trypsin-EDTA solution was then loaded into the quartz cell and the uorescence from the solution at 490 nm was measured using a Perkin-Elmer LS55 spectrouorometer (Perkin-Elmer Corporation, USA) at an excitation wavelength of 370 nm. The conventional method was performed according to the supplier's instruction with modication. Similarly, 200 mL of the cell suspension without pre-treatment with FSNP-1-biotin at a density of 5 Â 10 5 mL À1 was seeded into a millicell. The cells were induced to migrate by the serum medium in the lower compartment. Aer removal of non-migrated cells in the upper compartment, the cells were stained with 0.1% crystal violet and observed under an optical microscope (Olympus, Japan).

Statistical analysis
Data are expressed as mean AE standard deviation. Statistical analysis was performed using one-way analysis of variance (ANOVA) with PASW Statistics 18.0 (Chicago, USA). The level of signicance for all the comparisons was P < 0.05.