Targeted delivery of 5-fluorouracil-1-acetic acid (5-FA) to cancer cells overexpressing epithelial growth factor receptor (EGFR) using virus-like nanoparticles

Chemotherapy is widely used in cancer treatments. However, non-specific distribution of chemotherapeutic agents to healthy tissues and normal cells in the human body always leads to adverse side effects and disappointing therapeutic outcomes. Therefore, the main aim of this study was to develop a targeted drug delivery system based on the hepatitis B virus-like nanoparticle (VLNP) for specific delivery of 5-fluorouracil-1-acetic acid (5-FA) to cancer cells expressing epithelial growth factor receptor (EGFR). 5-FA was synthesized from 5-fluorouracil (5-FU), and it was found to be less toxic than the latter in cancer cells expressing different levels of EGFR. The cytotoxicity of 5-FA increased significantly after being conjugated on the VLNP. A cell penetrating peptide (CPP) of EGFR was displayed on the VLNP via the nanoglue concept, for targeted delivery of 5-FA to A431, HT29 and HeLa cells. The results showed that the VLNP displaying the CPP and harboring 5-FA internalized the cancer cells and killed them in an EGFR-dependent manner. This study demonstrated that the VLNP can be used to deliver chemically modified 5-FU derivatives to cancer cells overexpressing EGFR, expanding the applications of the VLNP in targeted delivery of chemotherapeutic agents to cancer cells overexpressing this transmembrane receptor.


Analysis of EGFR expression via immunofluorescence microscopy, and selective internalization property of CPP (NRPDSAQFWLHH) in cell lines expressing different levels of EGFR.
To detect the expression of EGFR in A431, HT29 and HeLa cells, the cell lines were incubated with the rabbit anti-EGFR monoclonal antibody, Cetuximab (C225), followed by incubation with the goat anti-rabbit IgG conjugated to FITC. Figure 1a shows that A431 cells exhibited the highest green fluorescent intensity, indicating the highest expression level of EGFR in A431 cells. The fluorescent intensity in HT29 cells is lower as compared to A431 cells, while HeLa cells showed the lowest fluorescent intensity. This indicates that the expression level of EGFR in HT29 cells is lower than that in A431 cells, while HeLa cells expressed the lowest level of EGFR among the three cell lines. In order to elucidate the targeting property of CPP with the sequence NRPDSAQFWLHH towards EGFR, the FITC-conjugated CPP was incubated simultaneously with three types of cell lines expressing different levels of EGFR: A431, HT29 and HeLa. As shown in Fig. 1b, A431 cells exhibited an intense green fluorescence as compared with HT29 cells, while HeLa cells showed the lowest intensity of green fluorescence. These results indicated that the CPP internalized efficiently into A431 cells expressing the highest level of EGFR, followed by HT29 cells, which have intermediate amount of EGFR. The uptake of the CPP reduced drastically in HeLa cells, which have the lowest expression level of EGFR among these cell lines. The results demonstrated that the CPP is targeted towards EGFR for cell internalization.
Conjugation of CPP to tHBcAg VLNP via the nanoglue, and cellular uptake of the conjugated nanoparticle into A431, HT29 and HeLa cell lines. The CPP (NRPDSAQFWLHH) was synthesized together with the nanoglue (SLLGRMKGA), and separated by a linker (GGG) between these two sequences. The resulting peptide, NRPDSAQFWLHHGGGSLLGRMKGA, was covalently cross-linked to the carboxyl group at the spikes of tHBcAg VLNP using Sulfo-NHS and EDC. The tHBcAg monomer shifted approximately 1 kDa on an SDS-polyacrylamide gel, demonstrating that the peptide was successfully cross-linked to the monomer (Supplementary Fig. S1). The tHBcAg VLNP displaying the CPP (namely, CPP-tHBcAg VLNP) was then incubated with A431, HT29 and HeLa cells. The internalization efficiency of the CPP-tHBcAg VLNP into the three cell lines was examined with immunofluorescence microscopy, using the mouse anti-HBcAg monoclonal antibody as the primary antibody and the FITC-conjugated goat anti-mouse antibody as the secondary antibody. Figure 2 shows that A431 cells gave rise to the highest green fluorescence signal as compared with HT29 cells, while HeLa cells had the lowest signal. This indicates that the internalization of CPP-tHBcAg VLNP into these cells is in EGFR-dependent manner. The result also demonstrated that the CPP was capable to deliver tHBcAg VLNP into cells expressing EGFR.
Synthesis and characterization of 5-FA. The conversion of 5-FU to 5-FA was performed as described in Sun et al. 37 . 5-FU reacted with α-chloroacetic acid in the presence of potassium hydroxide (KOH) and water at 60 °C (Scheme 2). The 5 h reaction yielded 83% of 5-FA. Nuclear magnetic resonance (NMR) spectroscopy www.nature.com/scientificreports/ and electrospray ionization-high resolution mass spectrometry (ESI-HRMS) were used to characterize the 5-FA. The substituted elements and the chain carbon atoms are highlighted and numbered according to their positions in the 5-FA molecule to ease the analysis of NMR spectroscopy spectra. According to the 1 H-NMR spectrum (Fig. 3a), there is a peak with δ H at 13.23 ppm, which corresponds to a carboxyl group in C2′. The presence of two oxygen atoms at C2′ causes deshielding of the hydrogen atom, and consequently moving the signal to the downfield. The area under the curve (AUC) of the peak is approximately 1, showing that the peak was generated by one hydrogen atom. The doublet with δ H at 11.93 and 11.92 ppm seems to correspond to the hydrogen atom bound to N 3 , considering that it was created by only one hydrogen atom based on the AUC of the peak. This hydrogen atom is under influence of strong electronegative atoms in close proximity with it, which are the two oxygen atoms linked to C 2 and C 4 , and caused the downfield position of the hydrogen atom. Additionally, there is no other hydrogen atom in the neighbourhood of this hydrogen atom according to the N + 1 rule. Another doublet with δ H at 8.09 and 8.08 ppm also corresponds to one single hydrogen atom as predicted from the AUC. The chemical shift to downfield suggests the proximity to an electronegative atom. By comparing the J 3 coupling constant (6.7 Hz) with J 4 from the previous doublet, it can be concluded that this is larger, thus indicating a hydrogen or fluorine atom in position J 3 . The peak with a signal of δ H at 4.36 ppm, and AUC of 2 indicates that the signal corresponds to two hydrogen atoms. Considering the chemical shift displacement, it is reasonable to conclude that there is a weak electron withdrawer neighbour, or it suffers the influence of both strong electronwithdrawing and strong electron-donating neighbours. This peak was then assigned to hydrogen atom in C1′ since it is influenced by both the electron-donor N 1 and C 2′ (Fig. 3a). The hydroxyl peak with δ H at 13.23 ppm corresponds to the carboxyl group of 5-FA. Additionally, the signals of δ H at 11.92 and 8.08 ppm, respectively, refer to the NH and CH protons in the 5-FA ring. These data are consistent with the number of atom carbon resonances observed in the 13 C-NMR spectrum (Fig. 3b). In 13 C-NMR spectrum, C 5 represents a doublet. The Scheme 1. Schematic representation of dual conjugated tHBcAg VLNP for targeted drug delivery. tHBcAg VLNP exhibits a number of Glu, Asp and Lys residues accessible for the conjugation of drugs and cancer targeting moieties using EDC and sulfo-NHS. Carboxyl groups of Glu and Asp located at the spike of the dimer of tHBcAg VLNP were covalently linked to the primary amines of Lys residue on the nanoglue (SLLGRMKGA) co-synthesized with the CPP (NRPDSAQFWLHH) targeting EGFR. The two sequences are separated by a linker (GGG). The CPP-tHBcAg VLNP was then conjugated with 5-FA at Lys residues exposed on the surface of tHBcAg VLNP. The three-dimensional structure of tHBcAg dimer was prepared using PyMOL 68   www.nature.com/scientificreports/ exposed on the surface of tHBcAg VLNP and CPP-tHBcAg VLNP, to form an amide bond. The nanoparticles conjugated with 5-FA, namely 5-FA-tHBcAg VLNP and 5-FA-CPP-tHBcAg VLNP, were purified using sucrose density gradient ultracentrifugation, and the absorbance at 275 nm (A 275 ) was measured spectrophotometrically. In comparison to unconjugated tHBcAg VLNP, 5-FA-tHBcAg VLNP and 5-FA-CPP-tHBcAg VLNP showed a remarkably higher absorbance at 275 nm, indicating that the nanoparticles were successfully conjugated with 5-FA (Fig. 5a). The conjugation efficiency of 5-FA (CE 5FA ) was 3.86 ± 0.43%, amounting to approximately 833 5-FA molecules per tHBcAg VLNP. Transmission electron microscopic analysis of the 5-FA-tHBcAg VLNP and 5-FA-CPP-tHBcAg VLNP revealed that the nanoparticles were intact with icosahedral shape, demonstrating that they were stable throughout the 5-FA conjugation process (Fig. 5b).  6). The IC 50 value of 5-FA in all the three cell lines could not be determined even though its concentration was increased to 1 mM, indicating that the 5-FA was significantly less toxic compared to 5-FU. After the conjugation of 5-FA to the nanoparticles, the IC 50 values of 5-FA-tHBcAg VLNP in A431, HT29 and HeLa cells were 3.26 ± 0.29 µM, 71.96 ± 1.97 µM and 36.94 ± 2.52 µM, respectively. This shows that 5-FA-tHBcAg VLNP is more toxic than 5-FU, and the cytotoxicity of these two compounds is not in EGFR-dependent manner as their inhibitory activities on HeLa cells (containing the least EGFR) are higher than those on HT29 cells (containing the intermediate level of EGFR). Interestingly, when CPP was conjugated to 5-FA-tHBcAg VLNP, the resulting formulation, 5-FA-CPP-tHBcAg VLNP, demonstrated EGFR-dependent inhibitory activities, although its toxicity is lower than the former. A431 cells expressing the highest number of EGFR had the lowest IC 50 value, 21.98 ± 3.8 µM, followed by HT29 cells which showed an IC 50 value of 215.16 ± 1.89 µM. The IC 50 value of 5-FA-CPP-tHBcAg VLNP could not be determined up to 1 mM in HeLa cells, which contain the lowest level of EGFR among the three cell lines (Fig. 7). The negative controls, tHBcAg VLNP and CPP-tHBcAg VLNP, were not toxic to the three tested cell lines.
Apoptotic activity of 5-FA formulations on A431 cells. Flow cytometry was performed to study the induction of A431 cell apoptosis by various 5-FA formulations. FITC annexin V staining was used as a marker for apoptotic cell death, and 5-FU served as the reference drug. The results showed that very few cells were apoptotic in untreated cells, as well as cells treated with tHBcAg VLNP and 5-FA (Fig. 8). The cells treated with 5-FU, which served as a positive control, underwent apoptosis. The percentage of early apoptosis increased significantly in the cells treated with 5-FA-tHBcAg VLNP, 5-FA-CPP-tHBcAg VLNP and 5-FU (Fig. 8g). Furthermore, the percentage of cells at late apoptotic stage increased significantly to 68.5%, 62.3%, and 72.6%, in the cells treated with 5-FA-tHBcAg VLNP, 5-FA-CPP-tHBcAg VLNP and 5-FU, respectively. Overall, these data demonstrated that 5-FA after being conjugated to tHBcAg VLNP has an anti-proliferative effect on A431 cells by inducing apoptosis.

Discussion
Chemotherapy is one of the most widely used methods for cancer treatments. However, low specificities of chemotherapeutic agents towards cancer cells in the human body always lead to poor cancer prognosis and mortality 38,39 . Therefore, specific delivery of chemotherapeutic drugs to cancer cells and tissues with minimal side effects on healthy tissues is a promising field for cancer treatments. EGFR is expressed abundantly in a broad spectrum of human cancers such as ovary, breast, lung, bladder, colon and skin cancers [40][41][42] . Hence, ligands that interact specifically with EGFR have gained popularity. Previously, we have isolated a CPP with the amino acid sequence NRPDSAQFWLHH from a phage displayed peptide library via biopanning 13 . The internalization of the CPP into A431 cells was inhibited by the anti-EGFR antibody 13 . Fluorescence activated cell sorting (FACS) analysis on EGFR performed by Beusechem et al. 36 revealed that A431 cells expressed the highest level of EGFR with a relative median fluorescence (RMF) value of 13.5, followed by HT29 cells (RMF: 7.3) and HeLa cells (RMF: 3.5). This is in good agreement with our study using immunofluorescence microscopy, which demonstrated that the EGFR expression level was the highest in A431, followed by HT29 and HeLa cells. In the present study, www.nature.com/scientificreports/ we have further demonstrated that the degree of CPP internalization is proportional to the amount of EGFR expressed on different cancer cells, with A431 showing the highest fluorescence intensity, followed by HT29 and HeLa cells. This further justifies that the internalization of the CPP into these cell lines is EGFR-dependent.
With an aim to enhance the therapeutic efficacy while minimizing undesired side effects of chemotherapy, researchers have designed various types of nanoparticles for incorporation of anticancer drugs, and specific drug delivery. VLNPs have attained much interest in the smart drug delivery system. The advantages of VLNPs over synthetic nanomaterials are their stable and highly ordered structural architecture in nanosize scale, which enhance tumor permeability and retention 43 , monodispersity and ease of production 44 , and well-defined interfaces for functionalization. Overall, VLNPs are less toxic, more stable, and more uniform as compared to nonviral based nanoparticles such as metal nanoparticles, liposomes and polymer particles 20,45,46 . To demonstrate the efficiency of the CPP as a targeting moiety to deliver VLNPs into cells expressing different levels of EGFR, the CPP was co-synthesized with the nanoglue, and covalently linked to tHBcAg VLNP. Immuno-fluorescence microscopy demonstrated that tHBcAg VLNP displaying the CPP successfully delivered the nanoparticle into A431, HT29 and HeLa cells in EGFR-dependent manner, which paves the way for the application of the CPP to deliver VLNPs to cancer cells overexpressing EGFR.
As one of the most commonly prescribed anti-cancer agents, 5-FU, either alone or in combination with other anticancer agents, has been commonly applied for the treatments of various types of cancers 47 . A variety of chemically modified 5-FU derivatives have been synthesized to enhance their antitumor efficacy 48,49 , but their irregular oral absorption, low bioavailability and lack of specificity often result in poor clinical therapeutic outcomes [5][6][7] . In order to improve the oral bioavailability, a wide variety of polymers have been synthesized for direct or indirect attachment of 5-FU and its derivatives 50,51 . In the present study, 5-FA was introduced as a 5-FU derivative as it has been reported to be highly effective and less toxic 52 . To synthesize 5-FA, 5-FU was reacted with chloroacetic acid in the presence of KOH. NMR and ESI-HRMS analyses confirmed the successful conversion of 5-FU to 5-FA. In order to deliver the 5-FA to cancer cells overexpressing EGFR, it was then conjugated to tHBcAg VLNP displaying the CPP. The conjugation was confirmed with spectrophotometer, and TEM analysis revealed that conjugation of 5-FA and CPP to tHBcAg VLNP has no undesired effect on the icosahedral structure of the nanoparticle.
In the present study, the cytotoxic effects of various 5-FA formulations on cells expressing different levels of EGFR were evaluated by MTT assay. The results clearly showed that the free 5-FA is less toxic than the reference drug, 5-FU. A study by Imoto et al. 53 revealed that 5-FU entered Caco-2 cells by passive diffusion. The terminal carboxyl group in 5-FA renders the molecule charged at physiological conditions, slightly more hydrophilic and higher polarity than 5-FU 54 . As a result, 5-FA is unable to diffuse easily through the lipid core of the plasma membrane 55,56 . These properties reduce 5-FA's cytotoxicity towards the cells. Moreover, the methyl-carboxyl group at N1 in the uracil group changes the 5-FU molecule in a way that makes it no longer a direct pyrimidine analog 57,58 . However, the cytotoxic effect of 5-FA was enhannced after being conjugated to tHBcAg VLNP, due to the non-specific binding and uptake of tHBcAg VLNP into the cells 59 , and subsequently the nanoparticle was hydrolyzed to release active 5-FU in the cells 60,61 . The non-specific distribution of 5-FA-tHBcAg VLNP in all the three cancer cell lines has become a drawback for this formulation, although the cytotoxic effect of 5-FA increased significantly upon conjugation onto tHBcAg VLNP. To address this problem, the CPP that targets EGFR was conjugated onto the 5-FA-tHBcAg VLNP. The resulting conjugate, namely 5-FA-CPP-tHBcAg VLNP displayed a much higher cytotoxicity in A431 cells compared to HT29 cells, followed by a very low cytotoxicity in HeLa cells. This EGFR-dependent manner was also observed for CPP-tHBcAg VLNP in the three cell lines (Fig. 2). The enhanced cytotoxicity of 5-FA-CPP-tHBcAg VLNP is believed to be caused by a specific interaction between the CPP displayed on the VLNP with EGFR.
5-FU induces apoptosis in cancer cells [62][63][64] , and 5-FA is believed to kill tumor cells with a similar mechanism as 5-FU 65 . Therefore, to confirm the apoptotic activity induced by various 5-FA formulations, FITC Annexin V/ PI (propidium iodide) flow cytometry assay was performed on A431 cells to identify externalization of phosphotidylserine (PS), a prevalent sign of apoptosis. FITC Annexin V is designed as a probe with high affinity for PS, for the detection of early apoptosis 66 . The results showed that the percentage of total apoptotic cells was far more than necrotic cells in A431 cells treated with the 5-FA formulations. This indicates that the 5-FA formulations markedly induced apoptosis, but to a much lesser extent, to cause necrotic cell death. Therefore, together with the targeting property of CPP, 5-FA-CPP-tHBcAg VLNP could be used as a therapeutic agent that induces apoptosis in cancer cells overexpressing EGFR.
In summary, 5-FA was successfully synthesized from 5-FU, and the former was significantly less toxic than the latter in A431, HT29 and HeLa cells. However, the cytotoxicity of 5-FA was greatly enhanced upon conjugation to tHBcAg VLNP. Besides, the specificity of tHBcAg VLNP increased significantly by conjugating the CPP that targets EGFR on the surface of the nanoparticle via the nanoglue concept. The resultant 5-FA-CPP-tHBcAg VLNP internalized cell lines expressing EGFR in relation to the amount of the receptor. Overall, tHBcAg VLNP together with the EGFR-CPP is favorable for targeted drug delivery system. Apart from the CPP and 5-FA, other tumor targeting moieties and chemotherapeutic drugs can conveniently be presented on the outer surface of tHBcAg VLNP using appropriate cross-linkers, expanding the variety of anti-cancer drugs to be delivered specifically to different types of cancer cells. www.nature.com/scientificreports/ Analysis of EGFR expression using immunofluorescence microscopy. A431, HT29 and HeLa cells (2 × 10 5 cells/well) were grown in six-well plates containing sterile glass coverslips, and incubated at 37 °C for 24 h. Subsequently, the medium was aspirated and the cells were washed three times with PBS ( Conjugation of peptide NRPDSAQFWLHHGGGSLLGRMKGA to tHBcAg VLNP, and cellular uptake of CPP-tHBcAg VLNP into A431, HT29 and HeLa cell lines. tHBcAg VLNP was produced and purified as described in Tan et al. 32 and Yoon et al. 67 . Peptide NRPDSAQFWLHHGGGSLLGRMKGA was conjugated at the spike of the tHBcAg VLNP by mixing tHBcAg: peptide (1:1) in phosphate buffer (25 mM NaH 2 PO 4 /Na 2 HPO 4 , pH 7) in the presence of EDC and Sulfo-NHS as described in Gan et al. 13 . The conjugated product was dialyzed against phosphate buffer (pH 7) to remove the excessive cross-linkers, and concentrated with VIVASPIN 6 (30 kDa cut-off polyethersulfone membrane; VIVASCIENCE, Germany) at 4500×g, 4 °C. The CPP-tHBcAg VLNP (250 µg/mL) was applied to A431, HT29 and HeLa cells in order to study its rate of internalization into these cells.  Fig. S2). The cooled reaction mixture was then adjusted to pH 5.5 with 2 M HCl solution.

Materials and methods
The precipitate formed was removed, and the solution was then adjusted to pH 2.0 and kept at 4 °C for 18 h. The precipitate was collected as a crude product, and recrystallized with water to produce 5-FA as a white crystalline product.
Conjugation of 5-FA to tHBcAg VLNP and CPP-tHBcAg VLNP. The conjugation of 5-FA to CPP-tHBcAg VLNP was performed as described in Biabanikhankahdani et al. 34 with some modifications. The carboxyl group of 5-FA was activated by dissolving 5-FA (5 mg), sulfo-NHS (20 mg) and EDC (20 mg) in sodium phosphate buffer (25 mM NaH 2 PO 4 /Na 2 HPO 4 , pH 6.0; 5 mL) at RT for 8 h. After that, the pH of the solution was increased to 7.4 with NaOH, and the CPP-tHBcAg VLNP (3 mg) in sodium phosphate buffer was added. The mixture was then incubated with gently agitation at 4 °C overnight, followed by sucrose density gradient ultracentrifugation (8-40%; 210,000×g, for 5 h at 4 °C) as described in Tan et al. 32 . Simultaneously, activated 5-FA was also added to tHBcAg VLNP, and the same conjugation procedure was applied. UV-visible measurements of tHBcAg VLNP, 5-FA, 5-FA-tHBcAg VLNP, 5-FA-CPP-tHBcAg VLNP and CPP-tHBcAg VLNP were determined using a spectrophotometer (Jenway 7315, Staffordshire, UK). The absorbance at wavelength 275 nm, which corresponds to the amount of 5-FU was measured at RT. The conjugation efficiency of 5-FA (CE 5-FA ) and the number of 5-FA (N 5-FA ) molecules conjugated to the nanoparticle were calculated using Eqs. (1) and (2), respectively. Cytotoxicity of 5-FA formulations. Cytotoxicity of free 5-FU, 5-FA, 5-FA-tHBcAg VLNP, and 5-FA-CPP-tHBcAg VLNP was determined using the cell viability MTT assay. A431, HT29 and HeLa cells (2.0 × 10 4 cells/well) were seeded in 96-well plates, and incubated for 24 h. Subsequently, the culture media were discarded, and the cells were added with media (100 µL) containing free 5-FU, 5-FA, 5-FA-tHBcAg VLNP and 5-FA-CPP-tHBcAg VLNP at twofold serial dilution (0.049 µM-1000 µM), and incubated for 72 h. After the incubation, MTT reagent (5 mg/mL; 20 µL) prepared in PBS was pipetted into the wells and incubated for 3 h. Dimethyl sulfoxide (DMSO; 100 µL) was then added to the wells and incubated for 15 min to dissolve the formazan crystal in the viable cells. A 570 nm was measured using a microtiter plate reader (Elx800, Bio-Tek Instruments, USA). The cytotoxicity of tHBcAg VLNP and CPP-tHBcAg VLNP was studied as negative controls.  After the conjugation of CPP (NRPDSAQFWLHH) to the 5-FA-tHBcAg VLNP, the cytotoxic effect of 5-FA-CPP-tHBcAg VLNP became selective, in which it was more cytotoxic to A431 cells as compared to HT29 and HeLa cells. The cytotoxic effect of 5-FA-CPP-tHBcAg VLNP in HT29 was higher compared to HeLa cells. Small graphs on the right show that tHBcAg VLNP and CPP-tHBcAg VLNP were not toxic to the tested cells. Data are expressed as mean ± SD of triplicate measurements. www.nature.com/scientificreports/