Applying DNA rolling circle amplification in fluorescence imaging of cell surface glycans labeled by a metabolic method

We report herein imaging cell surface glycosylation by using click chemistry and DNA rolling circle amplification (RCA) to improve detection sensitivity..


Introduction
Protein glycosylation is an important post-translational modi-cation. Aberrant alterations in N-and O-linked glycans are involved in many diseases. 1,2 Consequently, research on glycan structure and function, particularly visualization of glycans on the cell surface, is critical for elucidating the biological role of protein glycosylation in the dynamic environment of cells. However, glycosylation is a complex process due to its intrinsic heterogeneity. 3 Glycans cannot be visualized alone due to a lack of chromophores or uorophores. Moreover, unlike proteins or nucleic acids, it is difficult for glycans to be detected using specic antibodies or complementary sequences. 4,5 Although, there are some reports on the evaluation of cell surface N-glycan expression based on the recognition of lectin, 6,7 this method is unsuitable for analyzing O-linked glycans because their biosynthesis is complex and requires extensive prior digestion with glycosidases. 8 Thus, glycan detection remains the main obstacle in glycosylation studies.
There are several publications of detecting protein glycosylation in cells using in situ proximity ligation assay. [9][10][11] These methods offer new opportunities for glycopeptide identication, provided that antibodies for both protein and carbohydrate are available. 9 But as we know, it is difficult to obtain antibodies against glycans, and their binding constants are rather low. In fact, so far very few anti-glycan antibodies are commercially available. Even for lectins, it is difficult for antibodies to be widely used due to their low availability and weak binding capacity. This is a bottleneck for investigating the process of protein glycosylation. Recently, a metabolic glycan labeling technique has been used to hijack a cell's biosynthetic process. This was rst reported by Reutter's group 12 and then extended by Bertozzi's group. 13 This strategy has been used to visualize glycans on cells, [14][15][16][17] virus 18 and living organisms, such as zebrash, [19][20][21] Caenorhabditis elegans 22 and mice. 3,23 Carbohydrates functionalized with azide groups can be incorporated into the glycans of an organism via the cell's own metabolic machinery. Next, the azide group assembled in the glycan facilitates further coupling with a second bioorthogonal reagent through a covalent bond. Azide groups can directly react with alkyne-functionalized dyes 15,16,[24][25][26] or with labeled antibodies and lectins through two sequential bioorthogonal reactions. 3,23 In the latter case, the antibodies or lectins can be labeled with more than one uorophore, and the uorescence intensity is enhanced accordingly. However, this enhancement is restricted due to the limited number of uorophores conjugated to antibodies or lectins. On the other hand, carbohydrate metabolism can change all corresponding carbohydrates on a cell surface to the unnatural sugars without discrimination. If the concentration of added unnatural sugar is too high, the normal physiological function of a cell can be interfered with. 27 Therefore, the methods for sensitive visualization of glycans on a cell surface by using as few azido-sugars as possible would be greatly helpful for understanding the glycosylation process.
Rolling circle amplication (RCA) is an isothermal DNA replication technique, which can generate a long single DNA strand with multiple repeat units. [28][29][30] Recently, RCA has been used in uorescence in situ imaging of microRNA in tumor cells, [31][32][33] and tumor-specic delivery of drugs. 34 However, until now, there has no report on the uorescence imaging of carbohydrates on a cell surface using the RCA method. Here, we applied nucleic acid technology to glycochemistry, which could amplify the detection signal of a carbohydrate through nucleic acid amplication. Alkyne-functionalized DNA reacts with an azide group on a cell-surface glycan through click chemistry. The DNA assembled on the cell surface can further produce a long single DNA strand aer RCA. Since exponential amplication by RCA allows to detect even a single DNA or RNA molecule, 31 we hypothesized that this strategy can be used to investigate the subtle progress of glycosylation with more sensitivity. In addition, due to the lack of selectivity of the metabolic glycan labeling method, which leads to labeling of all glycoproteins and glycolipids containing the added unnatural sugar, 27,35,36 we further utilized uorescence resonance energy transfer (FRET) to specically image glycoform-bearing proteins of interest.

Cytotoxicity tests of azido-sugars in different doses
Studies have reported that per-O-acetylated azido sugars such as Ac 4 GalNAz can diffuse into the cells passively and are deacetylated under the action of intracellular esterases. The resulting deacetylated azido-sugars are converted to their high-energy donor forms (nucleotide sugars). These nucleotide sugars such as UDP-GalNAz are then directly exploited in glycan synthesis or are converted to other nucleotide sugars before being integrated into newly synthesized glycans. For example, UDP-GalNAz can be epimerized to UDP-GlcNAz by UDP-galactose 4 0 -epimerase. 15 As mentioned above, various uorescence imaging technologies have been developed for the detection of glycan biosynthesis using the azido-sugar metabolic labeling method. [23][24][25][26] However, in these strategies, the consumption of azido-sugars is considerably high to produce clear images. For example, 500 mM, 27 100 mM, 35 and 50 mM 15 azido-sugars have been used previously. The introduction of high concentrations of unnatural azido-sugars to the cell may interfere with the normal physiological processes of the cell. To test this viewpoint, here we test the effect of azido-sugar concentration on the growth of B16 cells. From the results shown in Fig. 1 we can see that in all three tests, per-O-acetylated azido-sugars (Ac 4 ManNAz, Ac 4 GlcNAz and Ac 4 GalNAz) show signicantly dose-dependent inhibition to the proliferation of B16 cells. Growth and proliferation rate of B16 cells can be affected signicantly aer treatment with 50 and 200 mM azido-sugar, while treatment with 5 mM azido-sugar has much less impact on the cell viability. In fact, similar results were also found in a previous study, in which Ac 4 ManNAz dose-dependently inhibited the growth of HeLa cells. 23 To decrease the adverse effects, it is essential to reduce the dose of azido-sugar as much as possible by increasing the sensitivity of the detection method. In this study, the RCA technique was adopted to solve this problem.
Design for RCA-assisted metabolic labeling of cell surface glycans As shown in Scheme 1, the azide group of azido-sugars integrated into glycans was covalently coupled with alkyne-functionalized DNA, which facilitated the introduction of DNA on the cell surface. The assembled DNA could initiate an in situ RCA reaction in the presence of phi29 DNA polymerase and dNTPs, generating a long tandem repeated sequence. The hybridization of the RCA product with high quantities of FITC-modied detection probe can produce a super-bright uorescence image under a uorescence microscope, and thus the proposed method can provide information about the localization and distributions of carbohydrates on the cell surface by using relatively low concentration of azido-sugar.

The amplication effect of RCA for glycan metabolic labeling
Due to the relatively high uorescence intensity of our method, we can conveniently detect biosynthetic glycans on cells by simply using a microplate reader. Fluorescence detection of the azido-sugar-labeled cells with or without (control) RCA amplication is illustrated in Fig. 2 with the error bars representing the standard deviation of three individual experiments. The cells were cultured with 0, 0.5 or 5.0 mM Ac 4 GalNAz for 3 days. As expected, the results show that the RCA reaction can signicantly enhance uorescence intensity. Although a higher azido-sugar concentration would produce a higher uorescence intensity, 5.0 mM unnatural sugar was selected for the cell incubation in our experiment because the uorescence intensity at this concentration fully meets the requirements for visualization (Fig. 3) and uorescence detection in this study. It is observed, by using the RCA technique, that we can use such a small quantity of unnatural sugar to produce high-resolution cell images and investigate the biosynthetic glycans on cells by simply using a microplate reader. The optimization of other experimental conditions, including the concentrations of alkyne-functionalized DNA S1 and FITC-DNA1, and a comparison of different catalysts, are shown in Fig. S1. † From Fig. 2 we can also found that the amplication effect of RCA is about 21 times that of cells treated with azido-sugar but without the RCA process, in which alkyne-functionalized DNA S1 was hybridized with FITC-DNA2. It has been reported that RCA is an isothermal DNA amplication procedure and can generate a linearly concatenated DNA molecule containing up to 1000 complementary copies of the circular DNA in 1 h. 37 The RCA reaction initiated by alkyne functionalized DNA S1 and padlock DNA was analyzed via electrophoresis, as shown in Fig. S2. † According to the result of agarose gel electrophoresis, the RCA reaction proceeded successfully in solution. Compared with DNA RCA in solution, the DNA RCA method in cell uorescence imaging is performed in a heterogeneous system and the environment of cells is quite complex. Owing to these reasons, the amplication effect is not as enormous as DNA RCA performed in buffer solution. However, there is still an obvious amplication of uorescence intensity for both confocal images (Fig. 3) and data obtained from a microplate reader (Fig. 2). Thus, DNA RCA can be effectively performed in such a complex biological environment. In addition, we also tried to directly use longer concatenated alkyne-DNA (S2) from commercial synthesis, which has four repeated FITC-DNA1 binding units (Fig. S3 †), or the DNA product of RCA reaction in solution for the click reaction (data not shown), but the staining results showed that this straightforward method did not work well. This may be caused by steric effects and high charge density of the RCA product, which is repulsed by the negative charge on the cell surface. In addition to the study on xed cells, we also tried to investigate the feasibility of this RCA method on live cells using a ow cytometric assay (Fig. S4 †), and the results showed that this method was also applicable for the glycan analysis on live cells.   The uorescence images for metabolically glycan-labeled 4TO7 cells were investigated. The cell nuclei were stained with DAPI. From the results shown in Fig. 3, virtually non-existent uorescence was observed in the azido-sugar-untreated cells. Also, relatively weak uorescence was observed in the azidosugar-treated cells without the RCA process, even under a high voltage photoelectric multiplication tube (PMT) of 850 V. In the latter case, FITC-DNA2 was simply hybridized with alkynefunctionalized DNA S1 bound to azido-sugar-labeled glycans on the cell surface but there was no signal amplication. The results suggest that azido-sugar-containing glycans are displayed on the cell surface but not at a high enough level to obtain a high quality image under the present experimental conditions. However, if RCA and hybridization with FITC-DNA1 processes were sequentially carried out aer metabolic glycan labeling, an intense uorescent staining signal was consistently observed under a relatively low voltage of PMT of 700 V, which indicated that the metabolically labelled glycans on the cell surface can be successfully detected by signal amplication using the RCA reaction. A similar phenomenon was also observed in B16 cells (shown in Fig. S5 †). It should be noted that, unlike many RCA reactions, which show bright dots for RCA products, 31,33 carbohydrate metabolism can change all corresponding carbohydrates on a cell surface to the unnatural sugars without discrimination. Thus, a large amount of azido sugars were labeled on the cell surface. Therefore, it is impossible to see bright dots on the cell surface in the confocal cell images of our RCA products.

Glycan metabolic labelling of different cells with different azido-sugars
It has been reported that different types of unnatural monosaccharide residues can be incorporated into glycoconjugates through the cells' own biosynthetic machinery. Ac 4 ManNAz, Ac 4 GlcNAz and Ac 4 GalNAz are the most common azido-sugars used in metabolic glycan labeling. In the sugar metabolic pathway, they can convert into the corresponding azido sialic acids (SiaNAz), GalNAz and GlcNAz, respectively. 8,15,25 Fluorescence analyses for the glycan biosynthesis of 4TO7 cells (Fig. 4A) and B16 cells (Fig. S6 †) were carried out using these three per-Oacetylated azido-sugars. The results show that the same cell line treated with different azido-sugars show very different uorescence intensities, and different cells treated with the same azido-sugar also show different uorescence intensities. This phenomenon suggests that the involvement of azido-sugars in glycan synthesis is not random and the glycan type is cell type dependent. Thus, using this method we can trace the variation in carbohydrate levels at different stages of the physiological process. These phenomena can be directly observed in the uorescence images shown in Fig. 4B.

Design for RCA and FRET-based imaging of protein-specic glycan
Although the DNA RCA-assisted metabolic labelling method can detect glycan synthesis on cells with high sensitivity, it fails to distinguish the sugar chains of a specic glycoconjugate from all glycans. To overcome this shortcoming, the FRET technique has been used in several recent studies for imaging of protein-specic glycoforms.
For example, Haga et al. used trans-membrane FRET to visualize specic protein glycoforms. 27 Chen's group reported a cis-membrane FRET-based method for protein-specic imaging of cell-surface glycans. 35 Ju's group developed a FRET method for simultaneous imaging of two types of monosaccharides on a specic protein by single near-infrared excitation. 38 Here, we combined FRET with our DNA rolling circle amplied metabolic glycan labeling technique to investigate its applicability for imaging protein-specic glycans. In this study, glypican-3 (GPC3) was used as a model molecule to study glycosylation of a specic protein in living cells. GPC3 is an important cell surface proteoglycan, which has two potential heparan sulfate attachment sites 39 and is a specic biomarker of several common cancers, in particular hepatocellular carcinoma. 40 As shown in Scheme 2, 293T cells were transfected with the expression vector of HA-tagged GPC3. Aer metabolic glycan labeling, RCA amplication and hybridization with FITC-DNA1, the glycans were stained with a signicant amount of FITC molecules. On the other hand, an anti-HA antibody and a TRITC-conjugated secondary antibody were successively introduced to stain the core protein of GPC3 with TRITC. Because the HA epitope is fused to the N-terminal of the core protein, the glycans and HA-tag were located on the same protein and even the same side of the cell membrane. The closely associated two subunits can produce a strong FRETinduced uorescence signal, 41 which facilitates investigating the inuence of glycosylation on the functions of cell surface proteins. It should be noted that technically this method is not only limited to GPC3-like proteoglycans, but that other glycoconjugates such as glycoproteins and glycolipids can also be investigated by choosing appropriate azido-sugars and antibodies.

FRET imaging of 293T cells transfected with HA-tagged GPC3
FRET phenomena between the FITC-labeled glycans and the TRITC-stained HA on the same GPC3 molecule were investigated by confocal microscopy (excitation and emission spectra for FRET pairs were investigated in Fig. S7 †). As shown in Fig. 5A and B, FRET was not observed in the cells only treated with azido-sugar followed by RCA, or transfected with HAtagged GPC3 followed by TRITC-conjugated antibody staining, which only showed FITC or TRITC uorescence signals at the rim of the cells. Once the cells were treated with azido-sugar and transfected with HA-tagged GPC3, the cell surface could be stained by both the RCA process and anti-HA antibody (Fig. 5C). Notably, in addition to FITC staining for glycans on the surfaces of all cells and TRITC staining for GPC3 core protein on the surfaces of GPC3-expressed cells, a signicant FRET-induced uorescence on the surface of GPC3-expressed cells was observed due to the relatively near distance between the glycans and HA epitope of GPC3 molecules. The strong FITC uorescence intensity due to RCA can induce an intense FRET signal (Fig. 5C). However, without RCA reaction, the FITC uorescence was too weak to induce a signicant FRET signal, especially in the case of reducing the amount of azido sugar (Fig. 5D) (a zoomed-out image is shown in Fig. S8 †). Thus, by combining confocal FRET microscopy with an RCA reaction, researchers can conveniently and clearly image glycosylation of a specic protein. Compared with previous work, which developed a strategy for information liberation of protein specic glycosylation via an exonuclease III-aided recycling, 42 our work has two obvious advantages. First, unlike the RCA process, the reported work relies on the recycling "hybridization and cleavage" process of the protein probe with other adjacent glycan probes. Therefore, the amplication effect is restricted to the number of glycans on specic proteins. However, the number of glycans is uncertain for different glycoproteins. Second, although this work demonstrates a powerful homogeneous quantication tool for research of glycosylation, the reported dual-color confocal uorescence imaging produced a signal-off image aer the scission of Exo III, while our FRET signal can be enhanced signicantly aer RCA, which is more suitable for uorescence image study.

Verication of intramolecular FRET on the same GPC3
To further verify the FRET signals come from intermolecular interactions on the same molecule, we compared uorescence images of 293T cells transfected with wild-type GPC3 and its glycan deletion mutant GPC3DGAG in Fig. 6. As described above, HA-tagged wild-type GPC3 showed clear FRET images (Fig. 6A). However, because GPC3DGAG had no sugar chains, we only observed FITC-labeled glycans and TRITC-labeled GPC3 core proteins on the cell surface, respectively, and intramolecular FRET was absent (Fig. 6B). The results suggest that the FRET signals are not from the carbohydrates on adjacent proteins but from the same molecule. Therefore, this is a quite reliable method for glycan imaging of specic proteins. Filmus. DAPI (4 0 ,6-diamidino-2-phenylindole) was purchased from Solarbio (Beijing, China). DNA sequences used in this work are listed in Table S1. † The water used throughout this work was produced by the Milli-Q water purication system (Millipore, USA).

Cell-surface glycoprotein labeling
Mouse breast cancer 4TO7 cells were cultured in DMEM supplemented with 10% FBS, 2 mM L-glutamine and 2 mM nonessential amino acids in 5% CO 2 at 37 C. Mouse melanoma B16 cells were cultured in RPMI 1640 medium supplemented with 10% FBS in 5% CO 2 at 37 C. Stock solutions of azido sugars (10 mM) were obtained by dissolving the sugars in lter-sterilized dimethyl sulfoxide. Aer seeding at a density of 2000 cells in 100 mL of media with no azido-sugar or 5.0 mM of the indicated azido sugar, and incubating at 37 C for 72 h, the cells were xed using 30 mL paraformaldehyde (4%) for 20 min. For the click reaction, the cells were incubated with PBS containing 250 mM THPTA, 50 mM CuSO 4 , 500 mM SA, and 50 nM alkyne-functionalized DNA S1 for 2 h at room temperature. For the RCA reaction, alkyne DNA modied cells were rst treated with 50 nM padlock DNA, 1Â T 4 ligase buffer and 5 U T4 ligase in PBS buffer at 22 C for 1 h followed by treatment with PBS containing 5 U phi29 polymerase, 1Â phi29 polymerase buffer and 2.0 mM dNTP at 37 C for 1.5 h. Finally, the cells were reacted with 3.0 mM FITC-DNA1 at 37 C for 30 min. The cells were washed three times with PBS aer each step.
For the imaging experiments, the cells were cultured in 8well Millicell EZ slides. Aer the above assembly process, the cell nuclei were stained for 10 min with DAPI diluted 100 times. Next, the slide was washed with PBS, sealed and examined using a laser scanning confocal microscope.
For uorescence intensity detection using a microplate reader, the cells were cultured in a 96-well Assay Black Plate (Corning). The RCA reaction was performed as described above. Aer treating with 30 mL 0.5% Triton X-100 for 20 min and washing with PBS, uorescence intensity data were collected in triplicate.
During the control experiments, the cells were cultured without azido-sugar, or cultured with azido-sugar but without the RCA reaction, which reacted with FITC-DNA2 for uorescent labeling.
For FRET experiments, human embryonic kidney 293T cells were cultured in DMEM containing 10% FBS and 5.0 mM azido sugar for 24 h. Then, the cells were transfected with the expression vector of pEFGPC3-HA or pEFGPC3DGAG-HA using Lipofectamine 2000 reagent (Invitrogen). 39 Aer transfecting for 5 h, the medium was changed to the normal medium containing 5.0 mM of the indicated azido sugar and then incubated in 37 C for 48 h. The cells were xed using 100 mL paraformaldehyde (4%) for 20 min. Next, RCA reaction and FITC-DNA1 hybridization were performed as described above. Aer blocking with 200 mL 5% skim milk powder for 1 h, 100 mL of anti-HA antibody diluted 200 times was added, and the sample was incubated for 1 h in the dark. Subsequently, 200-fold diluted rhodamine (TRITC)-conjugated goat anti-mouse IgG (H + L) (100 mL) was added, and the sample was incubated for 1 h in the dark. Aer nuclear staining and slice sealing, the slide was used for imaging experiments. The cells were washed three times with PBS aer each fabrication process to remove nonspecic adsorption.

Cell imaging
Cell uorescence and FRET imaging were performed using a Zeiss LSM 700 laser scanning confocal microscope. FITC uorescence was excited at 488 nm and collected from 500 to 600 nm. Rhodamine (TRITC) was excited at 555 nm and data were collected from 560 to 700 nm. FRET-induced uorescence was excited using a 488 nm laser, and data were collected from 600 to 700 nm. Fluorescence intensity data were detected using a PerkinElmer Multiode Plate Reader (l ex ¼ 488 nm, l em ¼ 525 nm).

Conclusions
The ultrasensitive detection of glycans on a cell surface is challenging due to their low concentrations and structural complexity. Herein, a novel DNA-based RCA reaction was used for imaging cell surface glycans labeled by a metabolic method. Through the amplifying effect of the DNA cycling amplication, carbohydrates on a cell surface can be imaged using a confocal microscope or simply detected using a microplate reader with considerably less azido-sugar. By combining with FRET, the glycosylation of specic glycoconjugates can be detected using specic antibodies and appropriate azido-sugars. The technique developed herein is especially favourable for studying low-abundance cell-surface glycans, and thus, single-molecule glycan imaging may be possible.