Fabrication of a protein microarray by fluorous-fluorous interactions

Fluorous-modified surfaces have emerged as a powerful tool for the immobilization of fluorous-tagged biomolecules based on their specificity and the strength of fluorous-fluorous interactions. To fabricate a fluorous-based protein microarray, we designed two strategies for site-specific modification of proteins with a fluorous tag: attaching the fluorous tag to the C-termini of expressed proteins by native chemical ligation (NCL) or to the Fc domain of antibodies through boronic acid (BA)-diol interactions. The perfluoro-tagged proteins could be easily purified by fluorous-functionalized magnetic nanoparticles (MNPs) and immobilized on a fluorous chip with minimal non-specific adsorption. Importantly, proteins immobilized on the solid support through non-covalent fluorous-fluorous interactions were sufficiently stable to withstand continuous washing. We believe that this fluorous-fluorous immobilization strategy will be a highly valuable tool in protein microarray fabrication.

interactions. These fluorous-based microarrays have the following advantages: (i) high signal-to-noise ratios, (ii) low non-specific binding, (iii) low and uniform background fluorescence, and (iv) simple fabrication workflows.
Boronic acids (BAs) are known to rapidly form reversible cyclic boronate esters with cis-diols 25 . Accordingly, BA has been applied as a probe for recognizing various saccharides and glycoproteins in numerous applications 26 . Previously, we reported the use of BA-functionalized glass slides to fabricate oriented protein microarrays by targeting the glycan chain of Fc-fusion proteins 27 . Herein, we developed a straightforward protein microarray fabrication method utilizing a fluorous-based surface for the oriented and noncovalent immobilization of proteins on the surface through fluorous-fluorous interactions. Two types of fluorous probes were designed for the site-specific labeling of proteins with a fluorous tag: one involving the combination of expressed protein through ligation with their N-terminal cysteine residue and the other via boronate ester formation between a BA-probe and an antibody's glycan moieties (Fig. 1). To make the ligation feasible in buffer solution, ethylene glycol and cysteic acid were incorporated into the fluorous probe to improve its hydrophilicity. Proteins could be easily modified with the fluorous probes by either native chemical ligation (NCL) (for expressed proteins, Fig. 1a) or boronate formation (for antibodies, Fig. 1b) and then immobilized on fluorous surfaces via fluorous interactions with minimal background.
As a proof of concept, eGFP was chosen as a target protein for a model study. eGFP was expressed using the IMPACT system (New England Biolabs) 28 and was modified by NCL to incorporate a fluorous tag on the protein C-terminal end (F tag -protein). To achieve a high ligation yield, the fluorous probe must have sufficient water solubility because the protein labeling yield is positively correlated with the probe's hydrophilicity.

Results and Discussion
Design and synthesis of fluorous tagged probes 1-4. Five probes, as shown in Fig. 1c, were designed and synthesized. In the design of probes 1-4, we incorporated a short-chain tri(ethylene glycol) linker within the probe, a common reagent generally used for increasing solubility in buffered-aqueous solutions. We also insert polar and charged amino acid residues including cysteic acid to further reduce the hydrophobicity of the fluorous-tagged probes. The preparation of probes 1-4 started from fluorous carboxylic acid 6, which was first coupled to N-Boc-2,2′-(ethylenedioxy)bis(ethylamine) (7) to provide compound 8 in 81% yield (Fig. 2). The Boc-protected derivative 8 with a fluorous tail could serve as a common building block for further functionalization. Thus, removal of the Boc group from 8 by trifluoroacetic acid (TFA) generated an amine, which was reacted with 9 (Supplementary Scheme 1) to give 10 in 73% yield over two-steps. Deprotection of trityl ether was smoothly occurred under acidic conditions (aq. TFA, CH 2 Cl 2 ) in the presence of 2.5% triisopropylsilane (TIPS) as a scavenger, which also affected cleavage of the Boc group in 10, producing probe 1 in essentially quantitative yield. To access BA-containing probe, Boc-protected cysteic acid linker (11) was incorporated on amine form of 8 to afford an intermediate 12 (68% over two steps), which was followed by Boc deprotecion and subsequent 1-ethyl-3-(3-dimethylaminorpopyl)carbodiimide hydrochloride (EDC-HCl)/1-hydroxybenzotriazole (HOBT)mediated coupling with 3-carboxyphenylboronic acid to afford fluorous-tagged BA probe 2 in a modest 18% yield over two-steps. For the synthesis of arginine-containing probe, Fmoc-Arg(Pbf)-OH was first reacted with the amine generated from flourous tag 8, and extension of the synthesis of Fmoc-protected intermediate was then carried out using 20% piperidine in DMF as the Fmoc deprotecting agent followed by coupling to the carboxylic acid of the incoming amino acid Boc-Cys(Trt)-OH affording 13 (53%, 4-steps). In a similar manner, compound 14 was synthesized in 69% yield starting from 8 and using Boc-Cys(Trt)-OH. Simultaneous cleavage (aq. TFA in CH 2 Cl 2 , 2.5% TIPS) of the trityl ether, Pbf, and Boc protecting groups in 13 and in 14 straightforwardly produced the target probes 3, and 4, respectively. The fluorous-tagged small molecule 5 was used for purification of fluorous-modified proteins in this study (see below and, for the synthesis, see Scheme S1).
Herein, we tested multiple combinations of ethylene glycol, with the amino acids arginine and cysteic acid to find an ideal structure for improving the solubility of the fluorous probe (Fig. 1c). In the water-solubility Protein modification with fluorous tag and fluorous-directed immobilization on fluorous slides. After capturing the expressed fusion protein on a chitin column, 5-mM synthetic fluorous probe 1 was added to the column in the presence of an auxiliary thiol reagent (2-mercaptoethanolsulfonic acid) to allow on-column cleavage and subsequent covalent labeling of the protein with the fluorous probe via NCL (Fig. 1a). The reaction was performed in HEPES buffer (pH 8.0) at 4 °C for 16 h, and the resulting eluent contained F tag -protein.
To investigate whether the non-covalent F-F interaction is specific or not, a fluorous-modified glass surface was spotted with native eGFP and a mixture of F tag -eGFP and native eGFP via a microcontact printer. The results showed that non-covalent affinity was only detected in the presence of a C-terminal fluorous tag ( Supplementary  Fig. S1), suggesting that the immobilization of proteins using a fluorous tag is indeed highly specific.
Fluorous tag as a strategy for protein purification using Magnetic nanoparticles (MNPs). The labeling of a protein with a specific functional group is important for studying its interactions with small molecules, DNA, and other proteins. Thus, developing a tool to facilitate the isolation of such modified proteins is imperative because the labeling efficiency is often less than 100%. Magnetic nanoparticles (MNPs) are a well-developed platform for biochemical research 29, 30 because of their various advantageous properties, such as large surface-area-to-volume ratios, ease of modification, and magnetic behavior. As demonstrated previously [17][18][19][20][21][22][23][24] , non-covalent fluorous-fluorous interactions are highly specific and can effectively reduce the non-specific adsorption of proteins; therefore, fluorous MNPs were proposed as a purification platform for F tag -proteins. To make the fluorous functionalized MNPs (F tag @MNPs) water compatible, the nanoparticle surface was incorporated with ethylene glycol, which is also able to suppress non-specific protein adsorption. The F tag @ MNPs were synthesized by modifying our previously reported method 31 . A fluorous functionality was introduced onto the particle surface through a sol-gel process using tetraethyl orthosilicate (TEOS) followed by the addition of 2-[methoxy(polyethyleneoxy)propyl]trimethoxysilane (mPEG) and (tridecafluoro-1,1,2,2-tetrahydrooctyl) triethoxysilane (F tag -(OEt) 3 ) at various volume ratios-3:1, 4:1, 5:1, 6:1, 8:1, and 10:1-to give F tag @MNPs, as  Fig. 3. The solubility of the resulting F tag @MNPs increased as the amount of mPEG increased, and the non-specific adsorption decreased as the amount of (F tag -(OEt) 3 Fig. S2). At a mPEG/F tag -(OEt) 3 volume ratio of 6:1, the F tag @MNPs become soluble in buffer. It should be noted that the poor solubility of the F tag @MNPs can also be improved by the addition of detergent (discussed in the next paragraph). Therefore, to maintain the solubility of the MNPs while simultaneously maximizing the amount of available fluorous tags on the MNPs, F tag @MNPs with a mPEG/F tag -(OEt) 3 volume ratio of 5:1 were chosen for use in target protein purification.

) increased (Supplementary
To evaluate the protein purification efficiency achieved using the F tag @MNPs, MBP was chosen as a model protein and modified with the fluorous NCL probe (1) as described previously. We noted that after concentrating the fluorous tag-modified eGFP to a high concentration, the proteins tended to precipitate, possibly because of fluorous tag self-assembly in the aqueous solution (data not shown). Native polyacrylamide gel electrophoresis (PAGE) analysis was used to investigate the purification results (Fig. 4). As expected, proteins containing the fluoro-tag (F tag -MBP) were captured by the F tag @MNPs (Fig. 4a, lane 4 vs. lane 6). Interestingly, the presence of 0.1% triton X-100 detergent in the washing buffer inhibited the fluorous-fluorous interactions, resulting in the removal of F tag -MBPs from the F tag -MNPs (lane 10). Encouraged by these results, we designed two strategies for eluting F tag -MBP from F tag @MNPs by the addition of 1-mM (or 0.65%) compound 5 (to give F tag -MBP F ) or 0.1% triton X-100 (to give F tag -MBP triton ) in the elution buffer (Fig. 4b, lanes 2 and 4). It should be noted that elution using compound 5 yielded better results than that using triton X-100 (Fig. 4b, lane 3 vs. lane 5), although multiple elutions were required to completely remove the F tag -MBPs from the MNPs. To further confirm that the eluted protein was indeed fluoro-tagged, the resulting eluent was analyzed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS); this analysis gave a molecular weight (MW) of 43971 Da, in agreement with the calculated MW of 43905 Da (Supplementary Fig. S3). We also noted that most of the MBPs were ligated with only one fluorous tag, with a smaller proportion ligated with two fluorous tags (because of the formation of the S-S bond), as revealed by the MALDI-TOF MS analysis. Additionally, 1 mg of F tag @MNPs can capture an estimated 3.1 μg of F tag -protein from solution ( Supplementary Fig. S4) with quantitative recovery.

Specificity of fluorous-fluorous interactions as a strategy for fabrication of fluorous protein microarrays.
To investigate non-specific adsorption of the fluorous lipids on the slide surface, MBP (Fig. 5, row 1), F tag -MBP F (purified by compound 5, Fig. 5, row 2), F tag -MBP triton (purified by triton X-100, Fig. 5, row 4), and a mixture of native MBP with either F tag -MBP F (Fig. 5, row 3) or F tag -MBP triton (Fig. 5, row 5) were printed on a fluorous slide. The presence and activity of immobilized MBP were visualized by staining with biotinylated anti-MBP antibody followed by the addition of streptavidin-Cy3. The fluorescence images reveal that only protein with the fluorous tag displayed an emission signal (Fig. 5, rows 2 and 3), while F tag -MBP triton exhibited very weak emission. These results indicate that the non-covalent fluorous-fluorous interactions are specific. The weak interactions between F tag -MBP triton and the fluorous slide may be attributable to triton X-100 contamination in the protein solution (Fig. 5, rows 4 and 5). Notably, we encountered difficulty in removing triton X-100 by PD MidiTrap G-25 centrifugation, whereas compound 5 was easily removed. Furthermore, the fluorescent signal obtained from a mixture of F tag -MBP F and MBP is similar to that of purified F tag -MBP F , indicating that the presence of non-fluorous modified proteins did not affect the fluorous-fluorous interactions (Fig. 5, rows 2 and 3). Indeed, the fluorescence interactions remained stable even after multiple washes followed by fluorescent staining (Fig. 5b) or incubation under complex biological sample conditions (10% fetal bovine serum [FBS]) for 1 h (Supplementary Fig. S5), showing that the affinity of the fluorous-tagged proteins for the fluorous surface is strong and irreversible, as previously suggested 32 . To find the suitable concentration of perfluoro-tagged-protein for the reliable fabrication of protein microarrays, various protein sample concentrations were spotted on slides for binding assays. We found that as little as 0.05-μg/μL perfluoro-tagged-MBP was sufficient to obtain a detectible signal and that a concentration of 0.1 μg/ μL resulted in a saturated signal (Supplementary Fig. S6). We further estimated that the binding affinity (Kd) of F tag -MBP to the fluorous slide surface is 1.04 × 10 −6 M, although the affinity is likely highly protein dependent ( Supplementary Fig. S7). To demonstrate generality of the developed method for multiplex protein microarray fabrication, MBP, GST and the anti-RCA 120 antibody were chosen as target proteins to be modified with a fluorous tag using probe 1 through NCL at the protein C-terminus or probe 2 through boronate formation at the carbohydrate moiety. The presence of F tag -protein on the slide was visualized using RCA 120 followed by the corresponding biotinylated antibodies and streptavidin-Cy3.
As shown in Fig. 6a, each protein produced its respective fluorescent signals, and no noticeable cross interactions were observed. We noted that there were a few inconsistencies in spot size and shape from batch to batch, probably due to inhomogeneity in the commercially purchased fluorous slides. To further prove that the signal from the anti-RCA 120 antibody was attributable to site-specific boronate formation through BA rather than non-specific interactions involving the fluorous tag, solutions of anti-RCA 120 , anti-RCA 120 pre-incubated with F tag -acid (compound 5) and anti-RCA 120 pre-incubated with F tag -BA (compound 2) and anti-RCA 120 were spotted on a fluorous slide. The fluorescence image-based binding results (Fig. 6b) show that only the antibody with F tag -BA gave a signal, indicating that the F tag was successfully assembled on the antibody by boronate formation. with probe 1 followed by F tag @MNP enrichment; and lanes 5, 6, 9, and 10: MBP-MESNa incubated with probe 1 followed by F tag @MNP enrichment. (b) Lane 1: MBP-MESNa incubated with probe 1 followed by F tag @MNP enrichment and washing with HEPES buffer; lanes 2 and 3: MBP-MESNa incubated with probe 1 followed by F tag @MNP enrichment and elution with compound 5; and lanes 4 and 5: MBP-MESNa incubated with probe 1 followed by F tag @MNP enrichment and elution with triton X-100. Lanes 3 and 5 were both obtained by eluting MNP only once.

Discussion
In summary, we demonstrated the use of a specific non-covalent fluorous-fluorous interaction to fabricate a stable protein microarray. The perfluoro tag can be easily attached to the C-termini of target proteins by NCL and the Fc domains of antibodies by boronate formation. Although the fluorous-fluorous interactions are non-covalent, we demonstrated that they are sufficiently stable to withstand vigorous washing processes and complex biological samples. Importantly, the fluorous surfaces of the slides resist non-specific adsorption, resulting in better detection sensitivity. Previously, the majority of approaches for enhancing the sensitivity of protein microarrays utilize methods that attempt to enhance the assay signal itself. Some examples include photo-initiated polymerization 33 , plasmonic substrates 34 , and silicon nanowires 35 . However in many scenarios, the main bottleneck in assay sensitivity is not due to a lack of signal strength, but becuase of the inability to discern the true signal from the false background due to non-specific binding. As a result, the limit of detection, usually defined as three times the standard deviation of the background signal, is often much higher in such systems when used under complex biological sample conditions. This study sheds light on an alternative approach in enhancing protein microarray sensitivity, in which we do not attempt to enhance the strength of the assay signal itself, but rather try to reduce the non-spcicific background signal. Thus, we believe that this fluorous-fluorous interaction strategy will be a valuable tool for facilitating the development of high-throughput protein microarray screening platforms.  Synthesis of fluorous-tagged probe 1. Compound 10 (100 mg, 0.01 mmol) was dissolved in a mixed solution of TFA/CH 2 Cl 2 /triisopropylsilane (TIPS)/H 2 O (94/2.5/2.5/1 v/v/v/v, 4 mL). The mixture was stirred at room temperature for 30 minutes to deprotect the triphenylmethyl (Trt) and Boc groups. The solution was concentrated under reduced pressure. The resulting residue was washed by hexane (Hex) and EA to give probe 1, which was used in the next reaction without further purification (97% yield, 70.9 mg). R f = 0.4 (1:3 MeOH-DCM) 1    General method for fluorous-tagged protein spotting. The perfluoro-tagged protein solution was prepared in a printing buffer (20-mM HEPES, 500-mM NaCl and 0.1-mM EDTA with a final glycerol concentration of 10%) and dispensed using a robotic contact arrayer (AD1500 Arrayer, BioDot) fitted with Stealth Pins SMP3 (Arrayit Corp.) onto a fluorous-coated slide. The printing process was performed at 90% relative humidity, and the temperature was maintained below 26 °C. The printed slide was kept in a humidified chamber and maintained at 4 °C for 12 h. Before incubation, the slide was left at room temperature for an additional 2 h to enhance the fluorous-fluorous interactions. The slides were then blocked twice with 1% bovine serum albumin (BSA) solution in phosphate-buffered saline (PBS) for 5 min each and then washed twice with deionized water for 5 min each.
General method for protein microarray image aquisition and data analysis. The functional protein microarray activities were characterized by using either a biotinylated antigen or a pair of antigens and the corresponding biotinylated secondary antibody and streptavidin-Cy3 in a sandwich detection format (see Supporting Information for details). All images were acquired using a VIDAR Revolution ® 4550 scanner with a Cy3 filter. The activity of fluorescent eGFP on the slide was directly measured with a NovaRay Microarray Scanner using a fluorescein isothiocyanate (FITC) filter. The fluorescence intensity of the spots was quantified