Multivalent Aptamer/Gold Nanoparticle–Modified Graphene Oxide for Mass Spectrometry–Based Tumor Tissue Imaging

The protein mucin1 (MUC1) is an attractive target for cancer biomarkers because it is overexpressed in most adenocarcinomas. In this study, we exploited a MUC1-binding aptamer (AptMUC1) as a targeting agent for nanoparticle-based imaging systems coupled with laser desorption/ionization mass spectrometry (LDI-MS). We found that AptMUC1-conjugated gold nanoparticles immobilized, through hydrophobic and π–π interactions, on graphene oxide (AptMUC1–Au NPs/GO) bound effectively to MUC1 units on tumor cell membranes. The ultrahigh density and high flexibility of AptMUC1 on the GO surface enhanced the platform’s cooperative and multivalent binding affinity for MUC1 on cell membranes. After we had labeled MUC1-overexpressing MCF-7 cells (human breast adenocarcinoma cell line) with AptMUC1–Au NPs/GO, we used LDI-MS to monitor Au cluster ions ([Aun]+; n = 1–3), resulting in the detection of as few as 100 MCF-7 cells. We also employed this AptMUC1–Au NPs/GO–LDI-MS system to analyze four different MUC1 expression cell lines. In addition, the AptMUC1–Au NPs/GO platform could be used further as a labeling agent for tumor tissue imaging when coupled with LDI-MS. Thus, Apt–Au NPs/GO can function as a highly amplified signal transducer through the formation of large Au clusters ions during LDI-MS analysis.

In this study, we found that graphene oxide (GO) modified with Mucin1 (MUC1) binding aptamerconjugated gold nanoparticles (Apt MUC1 -Au NPs/GO) coupled with LDI-MS is a facile platform for the detection of tumor cells and for tissue imaging (Fig. 1). MUC1 is a large transmembrane glycoprotein (250-500 kDa) of the mucin family; it consists of a varying amplified repeat sequence of 20 amino acid residues that is rich in serine, threonine, and proline residues in its extracellular domain, a hydrophobic membrane-spanning domain of 31 amino acid residues, and a cytoplasmic tail comprising 69 amino acid residues 13 . The overexpression (> 10-fold) of MUC1 is often associated with a variety of malignant tumors, including breast, gastric, colorectal, lung, prostate, ovarian, pancreatic, and bladder carcinomas 14 , making MUC1 an ideal target molecule for tumor labeling and chemotherapeutics. Apt MUC1 (for its sequence, see the Method section) binds to synthetic peptides (K d : ca. 0.1 nM) containing MUC1 tandem repeated sequences of the extracellular domain 15 . In addition, Apt MUC1 has been employed to recognize and bind to native MUC1 at the surfaces of MCF-7 breast cancer cells 16 . In this study, we used LDI-MS analysis to demonstrate that the nanocomposite Apt MUC1 -Au NPs/GO binds selectively to MCF-7 cells. In addition, we employed Apt MUC1 -Au NPs/GO in combination with LDI-MS, monitoring the Au cluster ions ([Au n ] + ; Fig. 1c), for tumor tissue imaging. We conclude that the binding of Apt-Au NPs/GO with tumor cells can be transduced to highly amplified Au cluster ion signals during LDI-MS analysis.

Results and Discussion
LDI-MS of Apt MUC1 -Au NPs/GO. We synthesized GO from graphite (7-11 μ m) using the modified Hummers method 17 . Atomic force microscopy (AFM) and transmission electron microscopy (TEM) images revealed that the average size of a single-layer GO (thickness: ca. 1.4 nm) was approximately 235 nm (see Supplementary Fig. S1a and S1b). The 5´-thiolated Apt MUC1 , which comprised the functionality of a MUC1-binding aptamer (25-mer) and a polythymidine (T 15 ) linker, was bound to the surfaces of the Au NPs (diameter: ca. 13 nm) through Au-S bonding, resulting in the preparation of Apt MUC1 − Au NPs (ca. 50 Apt MUC1 molecules per Au NP). The 25-mer Apt MUC1 contains three Watson-Crick pairs, three consecutive mispairs, and four Watson-Crick pairs capped by a TTT triloop motif (Fig. 1a); their base-paired nucleotides melt cooperatively at approximately 47 °C. The TTT triloop structure is the primary binding site for MUC1 protein 15 . Apt MUC1 -Au NPs/GO was prepared through simple mixing of Apt MUC1 -Au NPs (1.0 nM) with GO (0.25 g L -1 ) in 5 mM sodium phosphate (pH 7.4). Apt MUC1 -Au NPs were absorbed on GO with stabilization mainly through multivalent nucleobase-graphene π -π stacking between the linear Apt MUC1 and GO (Fig. 1b), although hydrophobic and hydrogen bonding interactions between Apt MUC1 and GO could not be excluded 18,19 . The conformation of Apt MUC1 was predominantly a stretched linear structure on the Au NP surfaces in a solution of low ionic strength (5 mM sodium phosphate, pH 7.4). Linear single-stranded DNA interacts with graphene more strongly than do hairpin and double-stranded DNA 20,21 . After the Apt MUC1 -Au NPs had self-assembled on GO, we added phosphate-buffered saline (PBS; 10 mM sodium phosphate, 140 mM NaCl, 2.70 mM KCl; pH 7.6) to the Apt MUC1 -Au NPs/GO solution to induce folding of the Apt MUC1 units into hairpin structures on the Au NP surfaces. As a result, the Apt MUC1 molecules could be arranged in an appropriate orientation and conformation for the binding of MUC1.
TEM imaging of Apt MUC1 -Au NPs/GO revealed that the Apt-Au NPs had assembled homogeneously on the GO (see Supplementary Fig. S1c). UV-Vis absorption and DLS measurements (see Supplementary  Fig. S2) revealed that Apt MUC1 -Au NPs/GO dispersed well (no aggregation) in PBS solution. We used LDI to further study the formation efficiency of Au cluster ions from Apt MUC1 -Au NPs and Apt MUC1 -Au NPs/GO (Fig. 2). The photoabsorption of the Au NPs (13 nm) induced the desorption and ionization of the surface Au atoms under SmartBeam laser irradiation (355-nm Nd:YAG; 100 Hz; pulse width: 6 ns; 2.84 × 10 4 W cm -2 ; beam diameter: ca. 40 μ m). Photothermal evaporation and Coulomb explosion are two well-established mechanisms for the fragmentation of metallic NPs under pulsed laser illumination 22 . The Coulomb explosion mechanism involves the ejection of a large number of electrons to generate multiple ionized metallic clusters/nanoparticles 23 . In contrast, the photothermal mechanism suggests that the degree of surface evaporation is highly determined by laser-induced thermal energy transfer into the lattice system 22,23 . The formation of Au cluster ions from Au NPs under pulsed laser irradiation is very sensitive to the composition, size, and surface properties of the particles and the wavelength, intensity, and pulse width of the laser 22−25 . Similar to our previous findings 25 , the formation efficiency of the Au cluster ions [Au n ] + (n = 1-3) from the Au NPs decreased to 16.7% after the Au NPs had been modified with Apt MUC1 (Figs. 2A, B). The Au NPs may have transferred their absorbed energy to the surface Apt MUC1 molecules, decreasing the lattice temperature (T l ) of the Au NPs, after laser excitation 25 . As a result, the evaporation or ionization of the surface Au atoms/ions into the gas phase was suppressed. In addition, the dense mantle of Apt MUC1 units may have strongly inhibited electron ejection from the Au NPs, decreasing the electron temperature (T e ) of the Au NPs; therefore, the explosive fragmentation-induced formation of Au cluster ions was strongly inhibited 22,23 .
We found that the formation of Au cluster ions from Apt MUC1 -Au NPs was enhanced in the presence of GO. As revealed in Fig. 2B-F, the intensities of the Au cluster ions in the LDI-MS spectra increased upon increasing the concentration of GO (0-0.25 g L −1 ). Because of their high specific surface areas (ca. 2600 m 2 g -1 ), high UV-region absorption coefficients, and energy transduction capabilities, graphene and GO have been used recently as nanomaterials in surface-assisted LDI-MS for the analysis of oligonucleotides, proteins, lipids, and small molecules [26][27][28][29][30] . We suspect that energy (heat) transfer from GO to the Apt MUC1 -Au NPs may have been the main reason for the enhanced formation of [Au n ] + cluster ions; GO-assisted electron ejection from the Apt MUC1 -Au NPs may also have contribute to the higher formation efficiency of the [Au n ] + cluster ions. The aggregation of Apt MUC1 -Au NPs (1.0 nM; ~0.014 g L -1 )/GO resulted in a slight decrease in the signal intensity of [Au n ] + when the concentration of GO was greater than 0.25 g L -1 (Fig. 2F). Aggregation occurred when the GO concentration was > 0.25 g/L presumably because of the relative low number of adsorbed Apt MUC1 -Au NPs on each GO and therefore, their stability in PBS solution decreased (see Supplementary Fig. S3 and Fig. S4). In addition, a low ratio (< 0.06) of the concentrations of Apt MUC1 -Au NPs (1.0 nM; ~0.014 g L -1 ) to GO (> 0.25 g L -1 ) may have resulted in the formation of crosslinks (Apt MUC1 -Au NPs/GO), thereby inducing aggregation of the nanocomposites. Such aggregation of Apt MUC1 -Au NPs/GO nanocomposites made it difficult to fragment, desorb, and ionize the surface Au atoms/ions on the Au NPs. The combination of Apt MUC1 -Au NPs/GO composites and LDI-MS monitoring of the [Au 1 ] + signals allowed detection of the Apt MUC1 -Au NPs at concentrations as low as 10 fM. In contrast, the equivalent concentration from the UV-Vis absorption measurements, monitoring the surface plasmon resonance (SPR) absorption of the Au NPs at 520 nm, was approximately 10 pM (see Supplementary Fig. S5). In addition, the Au ion-driven signals arising from the high surface ratio of Au atoms/ions on Au NPs are much stronger than those from proteins in the LDI-MS analysis because of the low ionization/desorption efficiencies of proteins. Therefore, we suspected that our Apt MUC1 -Au NPs/GO would be an exceptional labeling agent for the analysis of marker proteins in tumor cells when coupled with LDI-MS. Fig. 3a, b, the [Au n ] + signals obtained using Apt MUC1 -Au NPs/GO, prepared from Apt MUC1 -Au NPs (1.0 nM) and GO (0.25 g L -1 ), when coupled with LDI-MS for the detection of MUC1-overexpressing MCF-7 cells were approximately three times stronger than those obtained using Apt MUC1 -Au NPs. In a control experiment, polythymine (T 45 )-conjugated Au NPs provided low-intensity [Au n ] + signals, compared with those from Apt MUC1 -Au NPs and Apt MUC1 -Au NPs/GO, when analyzing MCF-7 cells. The detection of MCF-7 cells using Apt MUC1 -Au NPs/GO was superior to that using Apt MUC1 -Au NPs, presumably because the former displayed a higher formation efficiency of Au cluster ions and/or stronger binding affinity toward MUC1 on the cell membrane. In a quantitation analysis performed using inductively coupled plasma mass spectrometry (ICP-MS), we found that the number of Au NPs bound or uptaken by MCF-7 cells from Apt MUC1 -Au NPs/GO was approximately four times higher than that from Apt MUC1 -Au NPs (Fig. 3c), revealing that Apt MUC1 -Au NPs/GO had relatively stronger binding affinity toward MUC1 on the cell membrane. The signal enhancement obtained by LDI-MS (~3-fold) was lower than that of by ICP-MS (~4-fold) probably due to the incomplete fragmentation, desorption, and ionization of Apt MUC1 − Au NPs to Au cluster ions in the LDI-MS analysis. The ultrahigh density of Apt MUC1 on the surface of the Au NPs and the ultrahigh density of Apt MUC1 -Au NPs on the GO surface provided a high local concentration of flexible Apt MUC1 ligands, enhancing the cooperative and multivalent binding affinity toward MUC1 units on the cell membrane 31,32 . The nanosheet structure of GO, allowing ready adhesion to the plasma membrane and high uptake by mammalian cells through phagocytosis and clathrin-mediated endocytosisi 33−35 , may have contributed to the strong affinity of Apt MUC1 -Au NPs/GO toward MCF-7 cells. In addition, high density of Au NPs on GO can generate greater number of Au cluster ions for signal enhancement. Our Apt MUC1 -Au NPs/GO-LDI-MS system could detect as few as 100 MCF-7 cells when monitoring the [Au 1 ] + signals (see Supplementary Fig. S6). Compared to other mass spectrometry based tumor cell analysis techniques 36−38 , our Apt MUC1 -Au NPs/GO-LDI-MS platform is relatively simple, rapid, and selective and shows great legibility for tumor cells quantification.  Supplementary  Fig. S7c). Notably, the sample preparation for ICP-MS analysis required relatively complicated and tedious pretreatment steps. We also observed that the intensity of the [Au 1 ] + signals increased upon increasing the ratio of MCF-7 to MCF-10A cells in a co-culture system containing these two cell types (see Supplementary Fig. S8). MUC1 is widely expressed by normal epithelial cells (e.g., MCF-10A); this expression increases significantly when the cells become malignant, such as in breast, pancreatic, or ovarian cancer 14 . MUC1 is overexpressed in approximately 90% of human breast cancers; the expression of membrane-associated MUC1 in MDA-MB-231 cells is, however, lower than that in MCF-7 cells 39 . The low level of MUC1 expression in some breast adenocarcinoma cells is associated with decreased cytokeratin expression and increased vimentin expression 40 . In fact, the MUC1 protein is expressed aberrantly in many solid tumor cell lines 41 . For example, the 293T cell line is MUC1-negative because it does not expresses endogenous MUC1 41 . Our results from Apt MUC1 -Au NPs/GO-LDI-MS analysis platform (Fig.  S7) are consistent with these reports and the results of the western blotting analysis (see Supplementary  Fig. S9) 40,41 . MUC1 is associated with cellular transformation and tumorigenicity; it is considered as both a potential cancer biomarker and an attractive target for cancer immunotherapy 14 . Therefore, we suspect that our proposed assay will have practical applications in the detection of MUC1 expression in tumor cells, offering high accuracy and reliability with comprehensive information for the early detection of MUC1-related cancers.

Specificity of Apt
Analysis of MUC1 expression. Furthermore, we applied our Apt MUC1 -Au NPs/GO-LDI-MS platform to the analysis of MCF-7 cells after treatment with apigenin (4´,5,7-trihydroxyflavone; 0-75 μ M) for 24 h. Apigenin, a natural polyphenol product belonging to the flavone class, is abundant in some fruits (e.g., cherries, apples, grapes) and vegetables (e.g., parsley, Chinese cabbage, bell peppers, celery). Apigenin has been studied widely for its anti-inflammatory, antioxidant, and antitumor properties; it can inhibit tumor cell invasion and metastases by regulating protease production 42 . Recently, it was demonstrated that apigenin can inhibit mucin gene expression and the production of mucin protein in epithelial and tumor cells 43 . Using our Apt MUC1 -Au NPs/GO-LDI-MS detection system, we also observed an apigenin-induced downregulation of MUC1 expression: the intensity of the [Au 1 ] + signals decreased upon increasing the concentration of apigenin used for treatment of the MCF-7 cells (Fig. 4a). This downregulation of MUC1 in MCF-7 cells was consistent with the results of western blotting (Fig. 4c), revealing that our LDI-MS-based assay has great potential for use in the analysis of membrane protein expression levels after labeling with functional Au NPs. Compared with western blotting analysis, our assay is relatively simple, sensitive, and capable of high throughput. Tissue imaging. The development of a rapid, high-throughput method for MUC1 expression imaging in a typical tissue microarray would be a significant advance over the current practice. Therefore, we further employed our Apt MUC1 -Au NPs/GO-LDI-MS platform for tissue microarray imaging by monitoring the [Au 1 ] + signals after the Apt MUC1 -Au NPs had targeted cells in a tissue microarray (US Biomax). The tissue microarray, comprising breast carcinoma or normal tissues (diameter: 1.5 mm) was labeled with Apt MUC1 -Au NPs/GO and subsequently studied using LDI-MS. Representative images for breast carcinoma and normal microarray tissues sections (see Supplementary Fig. S10) revealed that the total intensity of the [Au 1 ] + signals from the breast carcinoma tissue was much higher (> 10-fold) than that from the normal tissue, suggesting that our Apt MUC1 -Au NPs/GO-LDI-MS system is capable of analyzing MUC1 expression in tissue samples. The spatial resolution of our LDI-MS imaging technique (~80 μ m) is limited by laser beam diameter size (40 μ m) which let the spatial resolution inferior to the current practices such as fluorescence imaging 44 . However, Apt MUC1 -Au NPs/GO-LDI-MS platform provides the great specificity and sensitivity for the molecular imaging due to the specific and strong binding of Apt MUC1 -Au NPs/GO to targeted molecules on cell membrane and the high signal enhancement from Au NPs.
We also applied our Apt MUC1 -Au NPs/GO-LDI-MS platform to analyze the tissue sections (US Biomax) of normal breast tissue (normal human mammary gland; female, 35 years old) and human breast cancer tissue (human breast invasive ductal carcinoma; female, 42 years old). As indicated in Fig. 5b, MSI revealed, through monitoring of the intensity of [Au 1 ] + signals, a dramatic difference between the normal and tumor tissues, suggesting high sensitivity of our LDI-MS-based tumor tissue system for the diagnosis of malignancies. We used immunohistochemistry (IHC) to further validate MUC1 expression and distribution in tumor and normal breast tissue. Figure 5c presents the immunostaining images of the sections; the results are in good agreement with those obtained through LDI-MS, with the heterogeneously developed brown color being deeper in the breast tumor tissue (Fig. 5cB). These IHC validation studies confirmed that the expression pattern of MUC1 reflects the tissue distribution in the malignant breast cells.
In summary, we have coupled an Apt MUC1 -Au NPs/GO probe with LDI-MS for the detection of breast cancer cells and for tissue imaging. The formation efficiency of [Au n ] + ions from Apt MUC1 -Au NPs under pulsed laser irradiation was enhanced after anchoring to GO. In addition, Apt MUC1 -Au NPs/GO could bind specifically to MUC1 units on cell membranes. The intensities of the signals for [Au n ] + cluster ions in the mass spectra revealed the levels of MUC1 expression in tumor cells; these signals behaved as a highly amplified target-labeling indicator for the targeted tumor cells. Our probe enabled the selective and sensitive detection of as few as 100 MCF-7 cells and facilitated efficient analyses of MUC1 protein expression in four different cell lines. We also applied the Apt MUC1 -Au NPs/GO probe to breast tumor tissue imaging, highlighting the potential for studying various cancer tissues with different aptamer-modified Au NPs/GO. To the best of our knowledge, this study provides the first example of combining Apt MUC1 -Au NPs/GO with LDI-MS for tumor tissue imaging. This high-throughput LDI-MS imaging system has great potential for the diagnosis of cancer.

Materials and reagents.
All thiolated oligonucleotides used in this study were purchased from Integrated DNA Technologies (Coralville, IA). Graphite powder (99%, 7-11 μ M) was purchased from Alfa-Aesar (Heysham, Lancashire, UK). Sulfuric acid, phosphoric acid, hydrogen peroxide, trisodium citrate, sodium phosphate tribasic, sodium phosphate dibasic, boric acid, tetrachloroauric acid, dimethyl sulfoxide, apigenin, xylene, ethanol, 3,3-diaminobenzidine tetrahydrochloride, hematoxylin, and isopropanol were purchased from Sigma-Aldrich (Milwaukee, WI). Citric acid, calcium chloride, magnesium chloride, potassium permanganate, tris(hydroxymethyl)aminomethane (Tris), and hydrochloric acid were purchased from Mallinckrodt Baker (Phillipsburg, NJ). Alpha-MEM and fetal bovine serum were purchased from GIBCO (Campinas, Brazil). An MTT assay kit was purchased from Thermo Fisher Scientific (Waltham, MA). MUC1 (VU4H5) mouse mAb, anti-mouse IgG, and HRP-linked antibody were  MUC1-binding aptamer (5´-HS-TTT TTT   TTT TTT TTT GCA GTT GAT CCT TTG GAT ACC CTG G-3´) was attached to the Au NPs following a slightly modified procedure reported elsewhere 45 . An aliquot of aqueous Au NP solution (980 μ L) in a 1.5-mL tube was mixed with the thiol-oligonucleotide (100 μ M, 20 μ L) to yield final concentrations of 15 nM Au NPs and 2 μ M oligonucleotides. After 2 h, the mixed solution was salt-aged in the presence of 200 mM NaCl for over 16 h. Salt-aging resulted in a greater number of thiol-modified oligonucleotides bound to the Au NPs 45 . The mixture was then centrifuged at an relative centrifugal force (RCF) of 30,000 × g for 20 min to remove the excess thiol-oligonucleotide. After removal of the supernatant, the oily precipitate was washed with 5.0 mM Tris-HCl (pH 7.6). After three centrifuge/wash cycles, the colloid was resuspended in 5.0 mM Tris-HCl (pH 7.6) and stored at 4 °C. The Apt MUC1 -Au NPs/GO nanocomposite was prepared by mixing the GO (0.25 g L -1 ) and Apt MUC1 -Au NPs (1.0 nM) in 5 mM sodium phosphate (pH 7.4). After incubation for 1 h, the Apt MUC1 -Au NPs/GO was transferred into PBS and incubated for another 1 h at room temperature. The samples were irradiated using a SmartBeam laser (355 nm Nd: YAG) operated at 100 Hz. Ions produced by LDI were energetically stabilized during a delayed extraction period of 30 ns and accelerated through the TOF chamber in the reflection mode prior to entering the mass-analyzer. The available accelerating voltages ranged from + 20 to -20 kV. The instruments were calibrated with Au clusters, using their theoretical masses, prior to analysis. A total of 500 pulsed laser shots were applied to accumulate signals from five MALDI target positions at a power density of 2.84 × 10 4 W cm -2 . For ICP-MS (Agilent 7700 Series, Agilent Technologies, Santa Clara, CA), the Apt MUC1 -Au NPs/GO-treated cells samples were prepared in 2% HNO 3 .

Analysis of cells by Apt
Tissue imaging. The human breast tumor and normal breast tissue microarray (BR811; 80 cases; diameter: 1.5 mm; thickness: 5 μ m), human normal breast tissue section (HuFPT130), and human breast cancer section (HuCAT298) were used for tissue imaging. The procedures of tissue sample treatments were performed in accordance with the protocols approved by Animal Administration Committee of National Taiwan Ocean University. After deparaffinization, the tissue samples were labeled with Apt MUC1 -Au NPs/GO, prepared from 1.0 nM Apt MUC1 -Au NPs and 0.25 g L -1 GO, in PBS for 1 h. The slides were washed three times with PBS before they were cast onto an MTP slide adapter II MALDI plate and air-dried at room temperature prior to LDI-MS measurements. LDI-MS imaging experiments were performed in the image reflectron positive-ion mode using an Autoflex III MALDI TOF/TOF mass spectrometer. The tissue slides were irradiated using a SmartBeam laser (355 nm Nd: YAG) operated at 100 Hz. A laser spot diameter of 40 μ m and a raster width of 50 μ m were employed. Ions produced by LDI were energetically stabilized during a delayed extraction period of 30 ns and accelerated through the TOF chamber in the reflection mode prior to entering the mass-analyzer. The available accelerating voltages ranged from + 20 to -20 kV. The instruments were calibrated with Au clusters, using their theoretical masses, prior to analysis. One spot was pulsed for a total of 500 laser shots at a power density of 2.84 × 10 4 W cm −2 .
Please see the Supplementary Information for the details on the preparation and characterization of GO, preparation of 13-nm spherical Au NPs, cell cultures, and immunohistochemistry with horseradish peroxidase.