Controlling Surface Wettability for Automated In Situ Array Synthesis and Direct Bioscreening

Abstract

The in situ synthesis of biomolecules on glass surfaces for direct bioscreening can be a powerful tool in the fields of pharmaceutical sciences, biomaterials, and chemical biology. However, it is still challenging to 1) achieve this conventional multistep combinatorial synthesis on glass surfaces with small feature sizes and high yields and 2) develop a surface which is compatible with solid‐phase syntheses, as well as the subsequent bioscreening. This work reports an amphiphilic coating of a glass surface on which small droplets of polar aprotic organic solvents can be deposited with an enhanced contact angle and inhibited motion to permit fully automated multiple rounds of the combinatorial synthesis of small‐molecule compounds and peptides. This amphiphilic coating can be switched into a hydrophilic network for protein‐ and cell‐based screening. Employing this in situ synthesis method, chemical space can be probed via array technology with unprecedented speed for various applications, such as lead discovery/optimization in medicinal chemistry and biomaterial development.

An amphiphilic coating for glass surfaces is reported, on which small droplets of polar aprotic organic solvents can be deposited with enhanced contact angle (small feature size) and inhibited motion (strong binding), permitting fully automated multiple rounds of combinatorial synthesis of small‐molecule compounds and peptides. The amphiphilic coating can be switched to hydrophilic network for protein‐ and cell‐based screening.

1. Introduction

The detection of protein–ligand or cell–ligand interactions on a planar surface can facilitate the most direct measurement for screening a library of chemical compounds since the binding of cells or fluorescently labeled proteins can be visualized optically. Further, high‐density arrays of biomolecules can be constructed (100–10 000 features cm⁻²), thus ensuring the reduced volume of the assay, as well as the consumption of reagents.^{[ 1 ]} However, although oligonucleotide arrays for analyzing DNA and RNA represent one of the most valuable applications of array technology,^{[ 2 ]} the one‐well‐one‐assay (96‐ and 384‐wells) remains the most common method in the field of drug screening and discovery.^{[ 3 ]} Most cell–material interactions have been studied employing well formats, while only a few biomaterial screenings have been studied employing array formats.^{[ 4} , ^{5 ]} As an alternative approach to addressable arrays, the one‐bead‐one‐compound synthesis has been developed to generate large peptide and peptoid libraries.^{[ 6 ]} The compounds on the selected beads can be identified by peptide sequencing and mass spectrometry (MS). The synthesis of an addressable array or the screening of compounds on the beads is more sophisticated than aliquoting solutions into different wells. Moreover, similar to other screening technologies, most array‐based screenings^{[ 7 ]} are also limited to a collection of presynthesized chemical compounds. To overcome such limitations, we developed a technology for fully automated in situ synthesis of the high‐density array, which was compatible with peptide and combinatorial chemistry, and possessed suitable surface properties for protein‐ and cell‐based screening assays. The technology can avail a flexible and versatile tool for drug discovery in chemical biology and biomaterial development.

There are two general approaches for synthesizing molecule arrays; they include the immobilization of presynthesized compounds and the in situ (on‐demand) synthesis on surfaces. The first approach, which has been employed to prepare most oligonucleotide/peptide/protein/antibody arrays, is relatively simple. The resulting arrays exhibit superior quality since the compounds can be purified before spotting/immobilization. Moreover, this approach is more economical for preparing many array chips in parallel compared with the in situ approach. Further, the preparation of the array is relatively straightforward since it involves only one reaction step (spotting/immobilization); it does not involve the complications that accompany multistep in situ syntheses, as would be discussed. DNA microarray printing has been widely employed to quantitatively analyze gene expression patterns,^{[ 8 ]} while peptide microarray printing has been developed to profile protein‐binding, enzyme–substrate specificity, and isozyme‐selective inhibitors.^{[ 9 ]} However, since in situ array synthesis can be employed to explore very large combinatorial chemical spaces, which are not limited to presynthesized libraries, it can offer a powerful and flexible tool for ad‐hoc ligand design and synthesis. Regarding in situ oligonucleotide syntheses, a PicoArray reactor was developed for the parallel synthesis of DNA, which can then be assembled into multiple genes.^{[ 10 ]}

There are two major techniques for in situ array syntheses, and they are both associated with advantages and disadvantages. SPOT technology consists of the stepwise synthesis of peptides on a cellulose membrane utilizing a standard fluorenylmethyloxycarbonyl protecting group (Fmoc)‐based solid‐phase peptide synthesis (SPPS).^{[ 11 ]} The high porosity of cellulose makes it an ideal substrate for SPPS, which can absorb reactants in the matrix and be easily washed. However, owing to the high porosity, it is challenging to achieve a small feature size and high‐density array (up to 25 cm⁻²) on the cellulose membrane. The large feature size and the porous matrix also cause high protein consumption in the screening experiments. To image cells on the surface, the strong light scattering by the cellulose fibers interferes with the fluorescence‐based imaging and data analysis.^{[ 4 ]} Regarding the second approach, high‐density peptide arrays (>100 000 cm⁻²) can be generated on glass by printing the polymer particles containing preactivated building blocks,^{[ 12} , ¹³ , ¹⁴ , ^{15 ]} utilizing photolabile protecting groups,^{[ 16} , ^{17 ]} or generating prepatterned omniphilic/omniphobic surfaces.^{[ 18 ]} However, they are also associated with drawbacks. Many methods require special chemistry or in‐house‐developed instrumentations, such as amino acid carrier ink,^{[ 13 ]} as well as laser‐based transferring,^{[ 14 ]} scanning probe lithography‐based,^{[ 15 ]} or maskless lithography techniques.^{[ 17 ]} Second, automation is relatively more challenging than SPOT technology.

To realize in situ high‐density array synthesis on glass with the flexibility for SPOT technology, a seemingly straightforward but challenging approach is to spot the reactant solutions directly onto the glass. This approach requires the deposition of a small volume of an organic solvent on the glass surface exhibiting a large contact angle and inhibited motion. However, polar aprotic solvents, such as N,N‐dimethyl formamide (DMF) and dimethyl sulfoxide (DMSO), cannot form droplets on glass or amino‐functionalized glass surfaces. Their high wettability on the substrate induces a small contact angle (θ ≪ 10°) and causes the spread of liquid over a large area (Figure  1A). This issue cannot be easily circumvented by modulating the contact angle by varying the surface or solvent. The weakened attractive force between the solvent and surface will cause the undesired motion of the droplets (a common phenomenon in liquid droplet deposition).^{[ 19 ]} Therefore, the modulation of surface wettability has been explored to increase the spot density for array synthesis.^{[ 18 ]}In addition to the chemical synthesis, the compatibilities of the resulting molecule array with various biochemical and cell‐adhesion assays are also important.

Figure 1.

Amphiphilic coating of on glass surface with switchable wettability to organic solvents. A) The relationship between droplet contact angle and droplet cap base radius (r) on a planar surface. Assuming a droplet possesses the shape of a perfect spherical cap, the ratio of droplet cap base radius (r_θ ) to the droplet cap base radius in the shape of perfect semisphere (r_θ , θ = 90°) is plotted versus the contact angle. When θ = 19°, ratio = 2. B) Scheme of the synthesis of switchable amphiphilic coating. i) After the reaction of the amino‐functionalized glass slide with succinic anhydride, the resulting carboxyl groups were activated by EDC/NHS, followed by crosslinking with chitosan. ii) PG‐NH‐linker‐COOH (PG: amino protecting group) was activated by HATU/NMM and coupled to the remaining amino group of chitosan. The density of the amino group in the coating could be tuned by repeating cycles (i) and (ii). The lipid was coupled to the –OH group of chitosan using DIC/DMAP (iii), while the PG‐NH‐linker is used in the following solid‐phase synthesis (iv,v). After the synthesis and side‐chain deprotection (vi), the lipid ester could be hydrolyzed in ammonia solution (vii). C) Structure of chitosan coating (G3 in 1B).

Thus, in this work, we developed a glass surface coating, rendering the substrate compatible with solid‐phase combinatorial chemistry, including Fmoc‐based peptide syntheses and multicomponent Ugi reaction. By introducing a lipid chain into the coating to modulate the surface wettability of polar aprotic solvents, the in situ array synthesis with a small feature size (down to 50 µm) was efficiently achieved. Moreover, the amphiphilic coating can be switched into a hydrophilic matrix after the synthesis to make the molecule array suitable for protein‐binding and cell‐adhesion assays.

2. Results

2.1. Amphiphilic Surface with Low Wettability to Polar Aprotic Solvents

To depict the relationship between the surface wettability to polar aprotic solvents and the spot size, we plotted the contact angle, θ (in the shape of a circular arc), of the droplet against the spot radius (r_θ ) (Figure 1A). When θ was small, e.g., <15°, r_θ increased rapidly with decreasing contact angle. A surface with a relatively large θ for polar aprotic solvents, such as DMF, DMSO, and sulfolane, would favor array synthesis. For example, at θ = 20°, the resulting spot radius was only approximately two times as large as that from a hemisphere droplet (θ = 90°). However, the droplet should not possess a very large θ. Reducing the attractive force and wettability to a surface, e.g., with Teflon coating, will cause the droplet of the solvent to move on the surface with minimum resistance to small disturbances in their environment. Such undesired motion of the droplets must always be prevented in the printing process.

The introduction of lipid chains of different lengths to hydrogel coating could vary its amphiphilicity, thus tuning θ of the printed droplet on the surface (as shown later). As shown in Figure 1B,C, chitosan hydrogel coating was synthesized on the glass surface. The amino‐functionalized surface was converted into a carboxylic acid‐functionalized one by treating the amino‐silanized glass slide with succinic anhydride. After activating the carboxylic groups with N‐(3‐dimethylaminopropyl)‐N′‐ethylcarbodiimide/N‐hydroxysuccinimide (EDC/NHS), chitosan was added to form a hydrogel coat by crosslinking the carboxylic groups on the glass surface with the amino groups of chitosan (G1). Thereafter, the remaining amino groups were coupled with Fmoc‐Gly‐OH employing Hexafluorophosphate Azabenzotriazole Tetramethyl Uronium/N‐Methylmorpholine (HATU/NMM) as the coupling reagents (G2), while the hydroxyl groups were employed to induce the lipophilic modifications to tune the wettability to organic solvents (G3). As an alternative to chitosan coating, a surface comprising hydroxy and amino groups could be generated by coupling Fmoc‐Ser‐OH to the amino‐modified glass surface (Figure S1, Supporting Information). We selected three fatty acids with 8, 12, and 16 carbons to modify the hydroxyl groups of chitosan (Figure 1C) or serine (Figure S1, Supporting Information) utilizing N,N′‐Diisopropylcarbodiimide/4‐Dimethylaminopyridine (DIC/DMAP) as the coupling reagent, while the resulting ester bonds were labile to base‐catalyzed hydrolysis. Alternatively, the hydroxyl groups could be modified through an acid‐catalyzed addition reaction to form ether bonds, which are labile to acid‐catalyzed hydrolysis (Figure S2B, Supporting Information). In summary, the amino groups of chitosan or serine in the coating were utilized for the subsequent solid‐phase combinatorial synthesis (G5 and G6 in Figure 1C and Figure S1 in the Supporting Information), while the lipid modifications could be removed after the combinatorial synthesis (G7 in Figure 1C and Figure S1 in the Supporting Information).

The amino‐functionalized lipid‐modified surfaces (G4) exhibited remarkably reduced wettability to various polar aprotic organic solvents, including DMSO and DMSO/sulfolane (Figure  2 and Table  1 ). The high boiling points of DMSO and sulfolane make them particularly interesting for array synthesis because of the slow evaporation of the droplets after their deposition on the surfaces. Expectedly, the modification of C16 exerted the most remarkable effect on the wettability with advancing angles of 46° ± 1° and 48° ± 2° for DMSO and DMSO/sulfolane, respectively. Contrarily, without the lipid modification, θ  was <<10° and could not be correctly measured. The decrease in wettability was also reflected by that in the calculated surface energy (Tables S1 and S2, Supporting Information). Despite the relatively low wettability, the droplets could bind strongly to the substrate, as indicated by their relatively small receding angles and large θ hysteresis^{[ 20 ]} (the difference between the receding and advancing angles, which reflected the resistance of the lateral movement) (Table 1). Consequently, the printed droplets did not move even when the slide was tilted to 90° (Figure S3A, Supporting Information). When the hydroxy and amino groups of the chitosan coating were both modified with a C16 lipid chain, the resulting surface exhibited a further increase in θ (advancing angles of 67° ± 2° and 66° ± 4° for DMSO and DMSO/sulfolane, respectively). However, the deposited droplets were quite unstable on the surface; irregular shifts were observed, particularly in the small droplet printing (Figure S3B, Supporting Information). Controlled droplet movement is valuable in microfluidic liquid handling, on self‐cleaning surfaces, and in heat transfer.^{[ 19 ]} However, undesired droplet movement is a major obstacle to array printing and synthesis. The utilization of polymer particles as a reaction medium represents an indirect strategy for avoiding the motion and spread of droplets.^{[ 13 ]} Employing the C16‐modified amphiphilic G4 surface, organic solvents can be deposited by contact or piezo inkjet printings with a relatively large θ, as well as completely inhibited droplet motion. Droplets of different diameters ranging from 120 µm (40 pL, employing piezo inkjet printing) to 1 mm (50 nL, employing contact printing) could be deposited onto the surface when different printing methods were applied. Notably, the surface wettability to the polar aprotic solvent remained unchanged after multiple reaction steps, e.g., 16 cycles of coupling and deprotection in peptide synthesis (as shown later).

Figure 2.

Surface wettability of sulfolane/DMSO (6:4) mixture on different chitosan‐coated surfaces. With the increasing length of lipid chain modification, the wettability decreases. Both without lipid modification and lipid chain cleavage via hydrolysis of ester bond showed good solvent wettability.

Table 1.

Dynamic contact angles of H₂O, DMSO, and DMSO/sulfolane (40/60) on 1‐round or 3‐round chitosan‐coated surfaces

Solvent	Lipid modification	Advancing angle [°]	Receding angle [°]
H2O	No lipid	28 ± 2	Not detectable
	C8	63 ± 4	19 ± 1
	C12	69 ± 3	30 ± 3
	C16	76 ± 5	30 ± 5
	C16 (ammonia)	30 ± 1	Not detectable
DMSO	No lipid	Not detectable	Not detectable
	C8	25 ± 1	Not detectable
	C12	35 ± 1	13 ± 1
	C16	46 ± 1	18 ± 1
	C16 a)	44 ± 2	18 ± 1
	C16 (ammonia)	Not detectable	Not detectable
40% DMSO, 60% Sulfolane	No lipid	Not detectable	Not detectable
	C8	25 ± 2	Not detectable
	C12	37 ± 1	12 ± 0
	C16	48 ± 2	21 ± 1
	C16 (ammonia)	Not detectable	Not detectable

^a) With three‐round chitosan coating.

2.2. Switching Surface Wettability

The amphiphilic surfaces could be switched into hydrophilic ones by incubating the glass slide in an ammonia solution to hydrolyze the lipid ester bonds. Polar aprotic solvents possess very small θ on the resulting surfaces, and this reflects their high wettability after saponification (Figure 2 and Table 1). Similar to the surface without lipid modification, θs were ≪10° and could not be correctly measured. Alternatively, the acid‐labile lipid modification could be cleaved by Trifluoroacetic acid (TFA) to switch the wettability of DMF on the surface (Figure S2C,D, Supporting Information).

To determine the efficacy of the saponification reaction, we coupled a NeutrAvidin (NA) binder, desthiobiotin, to the hydroxyl group utilizing DIC/DMAP as the coupling reagent (Figure S4A, Supporting Information). DyLight633‐labeled NeutrAvidin (DL633‐NA) was utilized to monitor the formation and hydrolysis of the ester bond. As a positive control, desthiobiotin was coupled to the amino group (Figure S4B, Supporting Information). After treating the glass slide with ammonia solution, the signal intensity of the G3‐desthiobiotin surface decreased by 80.4% (Figure S4C,D, Supporting Information) compared with that of the G2‐desthiobiotin surface. Therefore, saponification with an ammonia solution could efficiently hydrolyze the ester bond that was connected to the chitosan matrix without affecting the amide bond.

2.3. Optimization of the Hydrogel Coating for Solid‐Phase Synthesis

Although the amphiphilic coating (G4) was compatible with spotting technology employing different polar aprotic organic solvents, the concentration of the amino group per area in the coating (C _area) for solid‐phase synthesis could be further enhanced. C _area could be increased by repeating the chitosan coating cycle (Figure S5A, Supporting Information) to obtain 0.072, 0.364, and 1.289 nmol cm⁻² of the free amino group after one, two, and three cycles, respectively (Figure S5C, Supporting Information). C _area was measured by the Fmoc release method employing ultraperformance liquid chromatography (UPLC) to quantify the amount of the dibenzofulvenepiperidine adduct, which was produced by Fmoc deprotection (Figure S5B, Supporting Information). The thickness of the chitosan coating after three cycles was analyzed by atomic force microscopy employing the scratch‐and‐scan method; it revealed a homogeneous coat with a thickness of ≈8.9 nm (Figure S6, Supporting Information). The growth of the chitosan hydrogel coating slightly decreased the surface θ of DMSO (44° ± 2° after three coating cycles and lipid modification; Table 1). However, it did not remarkably affect the printing of the solvent droplets. To optimize the linker for solid‐phase synthesis, as well as the subsequent biochemical assays, the design and synthesis of the linker were investigated (Figure S7, Supporting Information). The produced B₆ linker (six repeated β‐alanine units) could generate a spacer between the synthesized compound and matrix, thus minimizing the impact of the solid support on the protein–ligand interaction.

2.4. Multicomponent Ugi Reaction

To determine if the G4 substrate could be utilized for other chemical reactions apart from peptide‐bond formation, we conducted the multicomponent Ugi reaction. This reaction can introduce structural diversity by combining four reactants in a single step; it has been demonstrated as an efficient method for combinatorial synthesis and drug discovery.^{[ 21 ]} Figure  3A shows the design of the solid‐phase synthesis of a chemical structure consisting of six building blocks. The amino building block was on the solid support, while the aldehyde, carboxylic acid, and isocyanide ones were in the solution. Since the carboxylic acid building block possessed a protected amino group, an additional amide bond, which was formed with a fluorescent dye (R5), was employed to evaluate the reaction yield after the Ugi reaction. The direct formation of an amide bond with R5 without the Ugi reaction was employed as the positive control.

Figure 3.

Multicomponent Ugi reaction on G4 surface. A) Ugi reactions on G4 surface. i) Different amino acids on the substrate G4 were spotted with aldehyde/piperidine mixtures to form imine, followed by spotting the mixtures of Boc‐β‐Ala/isocyanide (molar ratio 1:4). ii) After overnight reaction, the glass slide was treated with Tris‐buffered saline/DMSO (volume 1:1) to hydrolyze the imine. iii) The free amino groups were capped by 5% acetic anhydride/2% NMM in DMF. iv) The Boc protection group was removed by TFA and the exposed amino groups were coupled to DyLight633‐NHS. B) 64 Ugi reactions on G4 surface with four different amino acids, four different aldehydes, and four isocyanides. The positions spotted with Boc‐β‐Ala/HATU/NMM was used as the positive control. Compared to the positive control, the Ugi reaction spots showed similar fluorescence intensity. C) Mass spectrometry analysis of Ugi4‐BBX (R₁: A4; R₂: I4, R₃: Met, R₄: Boc‐β‐Ala; X is a photocleavable linker) synthesized on glass surface after Boc deprotection and release by photolysis. The calculated monoisotopic mass with one charge state is 588.3543 and the observed is 588.3536.

A mixture of Boc‐β‐Ala/aldehyde/isocyanide (molar ratio 1:1:1) was spotted on the substrate containing the amino group. After the overnight reaction, the remaining free amino groups were capped by 5% acetic anhydride/2% NMM in DMF. After Boc deprotection, the exposed amino group was coupled to a DyLight633‐NHS ester. No fluorescent signal was detected for the spots of the Ugi reaction dissimilar to the observation during the direct formation of the amide bond (Figure S8A, Supporting Information). This can be attributed to the inefficient formation of the intermediate imine.^{[ 22 ]} To improve the formation, the substrate containing an amino group was first spotted with an aldehyde/piperidine mixture, followed by the spotting of the Boc‐β‐Ala, isocyanide, or Boc‐β‐Ala/isocyanide (molar ratio 1:1) mixture. Compared with the positive control, the Ugi reaction achieved a reaction yield of 26%, while no reaction was detected when only the carboxylic acid or isocyanide building block was spotted (Figure S8B, Supporting Information).

To prove that the Ugi reaction could be applied to combinatorial array synthesis, we synthesized a 64‐member library with four aldehydes, four isocyanides, and four amino acids. The relatively low reaction yield might be due to the high volatility of the reagents, e.g., piperidine, aldehydes, and isocyanide (Figure S8C, Supporting Information). We increased their concentrations (final molar concentrations of piperidine, aldehyde, Boc‐β‐Ala, and isocyanide were 0.25, 0.25, 0.125, and 0.5 m, respectively) in the spotted solutions and obtained remarkably improved reaction yields (Figure 3B; Table S3, Supporting Information). Furthermore, no reaction was detected in the absence of aldehyde or isocyanide (Figure S8D,E, Supporting Information). Four compounds (A1‐I1‐βAla, A2‐I2‐Phe, A3‐I3‐Ile, and A4‐I4‐Met) were selected to characterize the products of the Ugi reaction by MS. For each compound, 300 spots (diameter = 1 mm for each spot) were synthesized on one slide, which had been premodified with a photocleavable amino linker. The irradiation of the slide with light at 366 nm facilitated the release of sufficient substances for MS measurement.^{[ 23 ]} Thus, the four products of the Ugi reaction were confirmed by MS (Figure 3C; Table S4, Supporting Information). Therefore, the G4 substrate can be utilized for the combinatorial array synthesis via the Ugi reaction, while other multicomponent reactions will be explored in the future.

2.5. Adaptation to Standard SPPS and Automation

To fully automate peptide synthesis on glass surfaces employing standard SPPS chemistry and the spotting technique, we designed a flow cell system to accommodate the glass slide (Figure S9A, Supporting Information). The various reaction mixtures could be printed on the surface of G4 by piezo inkjet or contact printing (Figure S9B, Supporting Information, blue arrow). After each round of the coupling reaction, an acidic inactivation solvent (10% acetic acid in DMF/ethanol) was added to the pool to diminish the cross‐reactions between the spots, followed by washing, capping, and Fmoc deprotection (Figure S9B, Supporting Information, green arrow). The inactivation, washing, capping, and deprotection solutions could be drained under negative pressure (Figure S9B, Supporting Information, red arrow). A peptide was synthesized at each position by spotting/coupling the amino acids separately according to the peptide sequence (Figure S9C, Supporting Information). The spotting/coupling can be repeated multiple times to increase the coupling yield after acidic inactivation and washing. The amino groups, which were not in the spotted area, were acetylated in the first capping step. Therefore, the formation of an amide bond was impossible in those areas in the subsequent syntheses, and this rendered the array synthesis tolerant of small positional variations during high‐resolution spotting.

Compared with standard SPPS in syringe reactors, array syntheses on planar surfaces are more exposed to the ambient atmosphere. Therefore, the reaction yields are generally lower than those of standard SPPS.^{[ 24 ]} To investigate the efficacy of peptide array synthesis on a G4 surface, two calcineurin binding peptides, GPVIVITO and GPRIEITO (where O is a short polyethylene glycol (PEG) linker, O2Oc), were synthesized by contact printing,^{[ 25 ]} and each coupling step was repeated two times to increase the coupling yield. Further, biotin was coupled to the N‐terminal to evaluate the coupling yields. Notably, only the full‐length peptides, which possessed biotin, could bind to DL633‐NA. By varying the concentrations of biotin, which was conjugated on the surface, it was observed that the fluorescence signal intensity of DL633‐NA correlated with the concentration of the biotin group on the G4 surface (Figure S10, Supporting Information). The signals of biotin‐GPVIVITO and ‐GPRIEITO were 8% and 14%, respectively, compared with the spots with biotin directly coupled to the (β‐Ala)₆ linker, which corresponded to average yields of 69.7% and 75.5% for each coupling step, respectively (Figure  4A). We suspected that the low coupling yield was due to the poor surface wettability of the deprotection solution on the coat. To increase the wettability (Figure S11, Supporting Information), we changed the piperidine and capping solvents from DMF to DMF/ethanol (volume = 50%/50%). The biotin‐GPVIVITO and ‐GPRIEITO signals increased to 29% and 59%, respectively, corresponding to the average yields, 83.8%, and 92.7%, for each coupling step, respectively (Figure 4B). The presence of ethanol in the capping mixture [5% acetic anhydride, 2% NMM in DMF/ethanol (volume 50%/50%)] did not affect the capping efficiency (Figure S12, Supporting Information).

Figure 4.

Fully automated peptide synthesis. A) Synthesis using 20% piperidine in DMF for deprotection. B) Synthesis using 20% piperidine in DMF/Ethanol (volume 1:1) for deprotection. C) Step yield quantification by Fmoc release method. Two peptides (GPRIEITO and GGGGGGGO) were synthesized and the amounts of released Fmoc group after each deprotection step were quantified by UPLC. D,E) Syntheses of two 16‐mer peptides and their shortened fragments. All the peptides were terminated with biotin, thus only the full‐length peptides will bind to fluorescently labeled NeutrAvidin, allowing for the evaluation of reaction yields.

Notably, the determination of the coupling yields via the biotin–Neutravidin method was not a quantitative chemical analysis. We compared this method with the quantification of Fmoc release through the stepwise syntheses of the two peptides (Figure 4C), and both methods agreed well. While the Fmoc release method requires an entire slide to perform the stepwise chemical quantification for the synthesis of a peptide, the biotin–Neutravidin method afforded a convenient tool for evaluating the overall yields of all the peptides on an array.

Next, we synthesized biotinylated peptides of different lengths (up to 16 amino acids) by contact printing. Figure 4D,E shows that as the peptide length increased, the signals of the DL633‐NA binding decreased gradually. After 16 coupling cycles, the signals of (Gly)₁₆ and the thrombin‐binding peptide were 69% and 32%, respectively, compared with the spots with biotin, which was directly coupled to the (β‐Ala)₆ linker, and this corresponds to the average yield, 98%, and 93%, for each coupling step, respectively. To characterize the peptide, which was synthesized as an array on glass, by MS, 300 spots (each spot was 1 mm in diameter) of peptide PRIEITBX, where B is β‐Ala and X is a photocleavable linker, were synthesized. The cleaved peptide, PRIEITB, was confirmed by MS (Figure S13, Supporting Information).

2.6. High‐Density In Situ Array Synthesis

To test the potential of this technology for high‐density array synthesis, we printed reagents by piezo inkjet printing. The biotin‐G₇O and Fluorescein‐G₇O arrays were synthesized, as shown in Figure  5 . Fluorescein was directly imaged, while biotin was visualized by DL633‐NA. At densities of up to 6400 spots cm⁻², every feature could be distinguished, while a small merging of the spots was observed at 8100 spots cm⁻². The statistical evaluation of the intensities of the spot revealed <10% and ≈30% deviations in the DL633‐NA and fluorescein signals, respectively (Figure 5B,D,F,H). Three peptides (biotin‐G₇O, biotin‐GPRIEITO, and biotin‐GPVIVITO) were synthesized as arrays with different densities (Figure S14, Supporting Information). Remarkably, the arrays with different densities (3600–8100 spots cm⁻²), which were synthesized in two independent experiments employing piezo inkjet printing, obtained higher coupling yields than those that were synthesized by contact printing (36 spots cm⁻²). The average coupling yields (97.8%, 94.5%, and 97.1%) of biotin‐GPRIEITO, biotin‐PVIVITO, and biotin‐G₇O, respectively, were comparable to those of conventional SPPS in syringe reactors. The small feature size of the high‐density array synthesis might facilitate the penetration of the reagents through the matrix, thereby further improving the reactivity, e.g., the Fmoc deprotection reaction. Conclusively, the fully automated in situ synthesis of the array with a density of up to 6400 spots cm⁻² was achieved by piezo inkjet printing on the G4 surface.

Figure 5.

Synthesis of high‐density peptide array using piezo inkjet printing. Arrays of biotin‐G₇O (red, stained with DL633‐NA) and fluorescein‐G₇O (blue) were synthesized with different densities: A) 3600 spots cm⁻², C) 4900 spots cm⁻², E) 6400 spots cm⁻², and G) 8100 spots cm⁻²). Scar bar: 500 µm. Spot intensity statistical evaluation: (B), (D), (F), and (H) correspond to images (A), (C), (E), and (G), respectively.

2.7. Protein–Ligand Interaction

An array of four small molecular compounds (biotin, desthiobiotin, iminobiotin, and Fmoc‐CsA‐COOH, a cyclosporin A (CsA) derivative with a carboxylic acid group at position 1),^{[ 26 ]} and two cyclophilin A (CypA)‐binding peptides (CB1, AVRHFPRIWLH; and CB2, HFPRI) were synthesized. Biotin and desthiobiotin are potent binders to NeutrAvidin with K _d values in the pm and low nm ranges, respectively. Iminobiotin is a weak binder to NeutrAvidins with a µm‐range dissociation constant. CsA and the CypA‐binding peptides bind to their receptor (CypA) with low nm and mm affinities, respectively.^{[ 27 ]} After the syntheses, the array was probed with DL633‐NA or CypA (Figure  6A). CypA could bind selectively to CsA, but not to biotin and its derivatives. The binding of CypA to CB1 or CB2 was remarkably weaker than that to CsA. When probed by DL633‐NA, the fluorescent signals of the iminobiotin spots were much weaker than those of the biotin and desthiobiotin ones. Under this screening condition, the protein could not distinguish between the pm and low nm binders, although it could distinguish these two strong binders from the weak µm one (iminobiotin). The concentration of the amino group per area was lower when utilizing a serine‐modified slide. Weak interaction, such as that between NeutrAvidin and iminobiotin, could not be detected (Figure S15, Supporting Information). Therefore, we mainly focused on chitosan coatings in this study.

Figure 6.

Protein binding and epitope mapping on peptide arrays. A) Binding of neutravidin (up) and CypA (middle) to biotin, desthiobiotin, iminobiotin, CsA, and cyclophilin binding peptides (CB1 and CB2); binding of CaN to PRIEIT, PVIVIT, and CsA (bottom). B) Epitope mapping of monoclonal anti‐Flag antibody. Anti‐Flag‐tag antibody can bind specifically to the flag‐tag peptides, and Y₂ and K₃ are essential for the recognition, in good agreement with the published results. The original sequence NH₂‐DYKDDDDK‐COOH is marked by white circles. Red: Dylight633‐labeled anti‐Flag‐tag antibody. Green: Dylight550‐labeled anti‐His‐tag antibody.

Some promiscuous interactions were detected between SA and CB1/CB2 but not between SA and CsA. We also synthesized ligand‐containing hybrid compounds (biotin‐CsA, biotin‐CB1, and biotin‐CB2) for the two proteins. Expectedly, the hybrid molecules on the array could bind with both proteins. Additionally, the calcineurin (CaN)‐binding peptides, PVIVITO and PRIEITO, were also synthesized and probed with fluorescence‐labeled calcineurin (Fluo‐CaN). Figure 6A shows that CaN could bind with both peptides but not with CsA. Therefore, the in situ peptide array synthesis technology is compatible with the biochemical analysis of various protein–ligand interactions.

2.8. Peptide Epitope Mapping

Next, a peptide array was synthesized to map the antibody epitope. A FLAG‐tag peptide and an anti‐FLAG‐tag antibody were utilized as a model system. To obtain a longer and extensively flexible linker to facilitate antibody epitope recognition, a hybrid linker, O₂B₄ (two repeated short PEG chains (O2Oc) and four repeated β‐Ala) was utilized. The FLAG‐tag peptide and its mutations were synthesized, and the array was incubated with the fluorescence‐labeled anti‐FLAG‐tag antibodies. As shown in Figure 6B, the anti‐FLAG‐tag antibody could bind to the FLAG‐tag peptides, and Y₂ and K₃ were essential for the recognition; these findings are in good agreement with those of a previous report.^{[ 28 ]} The in situ peptide array synthesis technology could be employed to investigate antibody epitope recognition.

2.9. Array of Cell‐Adhesive Biomatrices

The facile and rapid generation and screening of a library of biomolecule‐functionalized materials could facilitate the identification of tailored biomaterials and promote the in vitro adhesion and growth of specific cell types. We investigated the capacity of this method to generate arrays for the screening of cell‐adhesive biomatrices. Interestingly, when the fibroblast L929 cells were seeded onto a peptide array, the cells adhered tightly to the entire surface of the G7 substrate. The chitosan hydrogel coating is a favorable substrate for adhering to the L929 cells. We changed the surface conditions to reduce the affinity of the cells to hydrogel coating, thus facilitating the differentiation of the effects of the distinct peptide sequences. When the cells were seeded onto the substrate without cleaving the lipid (G6), their adhesion was remarkably reduced. To further reduce it, a PEG₃₀₀₀‐(β‐Ala)₄ linker was utilized instead of the (β‐Ala)₆ one. The linker did not remarkably affect the peptide array synthesis, and the adhesion of the cells to the peptide‐free area could be abolished. A cell‐adhesive peptide, RGDSP, and its mutant, GGDSP with a PEG₃₀₀₀‐(β‐Ala)₄ linker, were synthesized as a peptide array (on the G6 substrate), and their effects on the adhesion to L929, human dermal fibroblasts (HdFn), and primary human umbilical vein endothelial cells (HUVECs) were investigated. The area without a peptide, as well as the spots with the GGDSP peptide, exhibited poor cell adhesion. Contrarily, the spots with the RGDSP peptide exhibited strong adhesion (Figure S16, Supporting Information). The promiscuous interaction and background noise could profoundly affect the development of the screening assay. Although there is no universal substrate for all applications, the matrix coating system exhibits high flexibility to being optimized for binding proteins and adhering cells without compromising the synthesis of the array.

3. Discussion

Array syntheses on cellulose membranes offer the highest flexibility. However, it is challenging to produce small feature sizes and obtain high‐density arrays. Moreover, owing to the high protein consumption and light scattering that are caused by cellulose matrices, glass surfaces are superior to cellulose membranes for imaging and screening. However, it is very challenging to print droplets on a glass employing common polar aprotic organic solvents (DMF and DMSO). Thus, polymer particles have been developed to replace the organic solvents as the reaction media. Nonetheless, the preparation of building block precursors as the polymer particles is challenging, thus rendering the technology expensive and less flexible, challenging to be fully automatized, and consequently limited to a small number of commonly utilized building blocks, such as natural amino acids. Notably, an ideal surface for the in situ (on‐demand) array synthesis must fulfill the following requirements: 1) it must offer relatively low wettability to the polar aprotic organic solvents to print small droplets and generate high‐density arrays; 2) possess sufficient binding energy to the droplets to inhibit undesired droplet movements, such as vapor‐mediated droplet motion; and 3) its final surface must be compatible with various biochemical and cell‐adhesion assays.

The features of the various surfaces, which were generated in this study, are summarized in Figure S17 (Supporting Information). The coating of a glass surface with chitosan hydrogel can increase the ligand concentration (G1 and G2) for the subsequent synthesis, while hydrophilic coating can afford a suitable environment for developing biochemical and cell‐adhesion assays. However, the coating did not alter the high surface wettability of the glass to polar aprotic solvents. When a lipid chain was introduced in the coating (G3–G6), it remarkably reduced the wettability to water and the most commonly employed polar aprotic solvents (DMF, DMSO, and sulfolane) in peptide synthesis and combinatorial chemistry. The droplets can be deposited on the surfaces without spreading. Moreover, despite the reduced wettability and increased θ, the droplets still possessed sufficient binding energy to the amphiphilic surface, thus preventing the undesired droplet motion and fusion even when the droplets were very close (Figure 5) or when the slide was tilted to 90° (Figure S2, Supporting Information). Lipophilic modifications can cause the nonspecific absorptions of many proteins. For many biochemical analyses, it is necessary to switch the surface into a hydrophilic state. The resulting G7 surface exhibited high wettability to both polar aprotic solvents and water, and this makes it ideal for biological assays as biomimetic hydrogel coatings.

Although the switchable surface can be employed to construct a combinatorial chemical library via peptide chemistry and the Ugi reaction based on commercially available instrumentation, the synthetic methods require future exploration with other chemical reactions and solvent systems. Significant improvement is still required in the aspects of printing with precision and speed (resolution = 20 µm and speed = 5 features/second employing the current instrumentation). Moreover, the coupling of array technology with biophysical measurements, e.g., surface plasmon resonance or interferometry, on the chip would be relevant to quantify the binding kinetics of protein targets to ligands. Once the structural diversity and throughput of chemical synthesis are synergized with the power of screening, the chemical space can be probed to elucidate biomolecular interactions (classical medicinal chemistry and lead optimization; hit validation for DNA‐encoded chemical library^{[ 29 ]}; linker optimization for fragment‐based drug discovery^{[ 30 ]}; and the evolution of molecular structures employing computational tools, such as molecular docking,^{[ 31 ]} machine learning,^{[ 32 ]} or genetic algorithms^{[ 33 ]}) with unprecedented speed.

Conflict of Interest

W. Lin and Y. Zhang have applied for a patent application on this method.

Supporting information

Supporting Information

Lin W., Gandhi S., Oviedo A. R. Lara, Thomas A. K., Helbig R., Zhang Y., Controlling Surface Wettability for Automated In Situ Array Synthesis and Direct Bioscreening. Adv. Mater. 2021, 33, 2102349. 10.1002/adma.202102349

Data Availability Statement

Research data are not shared.