Microfluidic Print-to-Synthesis Platform for Efficient Preparation and Screening of Combinatorial Peptide Microarrays

Abstract

In this paper, we introduce a novel microfluidic combinatorial synthesis platform, referred to as Microfluidic Print-to-Synthesis (MPS), for custom high-throughput and automated synthesis of a large number of unique peptides in a microarray format. The MPS method utilizes standard Fmoc chemistry to link amino acids on a polyethylene glycol (PEG)-functionalized microdisc array. The resulting peptide microarrays permit rapid screening for interactions with molecular targets or live cells, with low nonspecific binding. Such combinatorial peptide microarrays can be reliably prepared at a spot size of 200 μm with 1 mm center-to-center distance, dimensions that require only minimal reagent consumption (less than 30 nL per spot per coupling reaction). The MPS platform has a scalable design for extended multiplexibility, allowing for 12 different building blocks and coupling reagents to be dispensed in one microfluidic cartridge in the current format, and could be further scaled up. As proof of concept for the MPS platform, we designed and constructed a focused tetrapeptide library featuring 2560 synthetic peptide sequences, capped at the N-terminus with 4-[(N′-2-methylphenyl)ureido]phenylacetic acid. We then used live human T lymphocyte Jurkat cells as a probe to screen the peptide microarrays for their interaction with α4β1 integrin overexpressed and activated on these cells. Unlike the one-bead-one-compound approach that requires subsequent decoding of positive beads, each spot in the MPS array is spatially addressable. Therefore, this platform is an ideal tool for rapid optimization of lead compounds found in nature or discovered from diverse combinatorial libraries, using either biochemical or cell-based assays.

The solid-phase peptide synthesis (SPPS) method, invented by Merrifield in the early 1960s, allows for rapid peptide synthesis on solid supports with high efficiency and reduced side reactions and is very amenable to automation.¹ SPPS established a solid foundation for the field of combinatorial chemistry,² which was initially pioneered by Geysen et al., who synthesized peptides using a 96-well plate multipin system.³ The multipin method was later modified by Frank to perform peptide synthesis on cellulose papers, referred to as the “SPOT library” method.⁴ Both multipin and SPOT synthetic methods were robust and could be fully automated, but the array footprints were large. In 1991, a light-directed, spatially addressable parallel chemical synthesis method was first reported by Fodor et al.,⁵ using light-activated chemical protecting groups exposed through a prefabricated photomask to implement selected activation and coupling on a solid substrate. Light-directed synthesis can achieve a high-array density, yet for each amino acid coupling, a specific photomask is required. Later, multimirror techniques were applied to direct light exposure patterns, eliminating the need for a photomask.⁶ However, each amino acid coupling still required an individual synthetic step, limiting throughput. Thus, light-directed synthesis tended to develop toward applications requiring fewer building blocks, such as DNA or RNA microarray synthesis.^7,8 For example, Beier and Hoheisel⁹ reported a light activated approach for oligonucleotide synthesis but reported a yield of <90% per coupling cycle. Pellois et al. introduced a novel light activated chemistry-based approach,⁶ but it remains to be seen whether acceptable yields can be achieved for certain amino acid residues that are typically more difficult to couple, particularly for longer peptides.

The aforementioned approaches all synthesize and screen peptides in an array format on fixed solid supports; thus, each peptide in the library is spatially addressable. Yet, even with high density microarrays, the diversity of peptides in one chip is generally less than 50 000. In 1991, Lam and co-workers have reported the development and application of the “one-bead one-compound” (OBOC) combinatorial peptide library method, in which a huge diversity of peptide bead libraries (with 10⁶ to 10⁷ members) can be rapidly generated such that each 90 μm bead displays a unique peptide entity.^10,11 Arrays of peptide library beads, either in suspension or immobilized,¹² can be screened for biological activities.¹³ While powerful, the main drawback of the OBOC method is that the chemical library is not spatially addressable. Each positive bead has to be physically isolated for decoding. For peptide libraries, decoding is achieved with Edman microsequencing or with mass spectroscopy.¹⁴

In the last two decades, additional methods have been developed to generate and screen high-density peptide microarrays. Stadler and co-workers reported a particle-based synthesis method, where amino acids are incorporated into solid particles and deposited onto specific spots using laser printing or fusing. The particles are then melted down to liquid form to initiate the coupling reaction.^15–17 Recently, inkjet printing¹⁸ has gained attention for combinatorial synthesis, since it is capable of depositing reagent droplets on demand, with high temporospatial control and compatibility with standard synthetic chemistry. Yet, industrial inkjet printers are typically too complex and inflexible to modify for the purpose of chemical synthesis. Moreover, the integrated fluidic components can be problematic for cross-contamination, a limitation that has largely prevented inkjet printing from being widely adopted. Modified ink jet printers have been applied for peptide library synthesis. For example, Antohe et al. reported use of a Microfab printer head to synthesize peptides on cellulose paper,¹⁹ and Kim and co-workers developed an acoustic printing technique for peptide synthesis.^20,21 Our group has recently introduced a microfluidic impact printing platform.^22–24 Compared to conventional inkjet printing, our design features a detachable, disposable cartridge that greatly reduces routine costs and also avoids possible contamination. The small spot size (100 pL) yields a high peptide array density with minimal reagent costs. Furthermore, the flexible cartridge design allows multiple channels to be integrated in one cartridge, greatly increasing the flexibility and throughput of printing, affording high-speed high-density peptide microarray synthesis via sequential printing of amino acid building blocks and coupling reagents.

Here, we introduce a novel microfluidic-enabled combinatorial synthesis platform, referred to as Microfluidic Print-to-Synthesis (MPS). The MPS platform, derived from our established microfluidic impact printer,^22–24 aims to establish an array of specifically designed peptide sequences in a high-throughput and automated fashion. Unlike light-activated or particle-based arrays, which require specialized amino acids, the MPS method utilizes standard Fmoc-protected amino acids and SPPS chemistry, performed on a polyethylene glycol (PEG)-functionalized microdisc array fabricated on a glass slide. In this way, the MPS approach can guarantee high coupling yield due to the proven, robust Fmoc-chemistry, while the microdisc array affords ease of visualization and low nonspecific binding. The combinatorial peptide array can be reliably printed at a spot size of 200 μm and spot density of 1 mm center-to-center distance, dimensions that yield favorable reagent consumption (less than 30 nL per coupling cycle per spot). Unlike conventional inkjet printing, our microfluidic printing platform allows for up to 12 independent chemical dispensing channels to be assembled in one disposable microfluidic cartridge. As a demonstration of the MPS synthetic platform, we have designed, constructed, and screened a 1280-membered combinatorial tetrapeptide library N-capped with 4-[(N′-2-methylphenyl)ureido]phenylacetic acid (UPA), to identify peptidomimetic molecules targeting the α4β1 integrin overexpressed on a lymphoid cancer cell surface.²⁵ High consistency with our previously reported cell binding results has been found.¹³ Thus, we expect our new and inexpensive MPS platform to enable rapid discovery and optimization of high binding affinity ligands against a variety of biological target molecules, using biochemical or cell-based assays.

EXPERIMENTAL SECTION

Glass Silane Modification.

Standard glass slides (2 in. × 3 in.) were preactivated by oxygen plasma (Diener Electronic) for 15 min at 99 W, before immersion in a 1% solution of (3-acryloxypropyl)trichlorosilane in anhydrous toluene for 12 h under nitrogen environment.^26,27 Upon removal from the solution, the slide was thoroughly rinsed with toluene and ethanol and then baked at 115 °C for 2 h to fully cross-link the silane coating.

Microdisc Array Fabrication.

A prepolymer mixture of 200 μL of PEGDA [(polyethylene glycol)-diacrylate (MW 700 Da)], 112 μL of cross-linker (trimethylolpropane ethoxylate triacrylate), 8 μL of photo initiator (PI, 2-hydroxyl-2-methylpropiophenone), and 7.2 mg of 2-aminoethyl methacrylate·HCl was first dissolved in 300 μL of deionized water. Fifteen μL of this mixture was dropped onto a photomask and covered by the silane-treated glass slide. The sandwiched structure was then set up horizontally for 15 min to allow the solution mixture to spread evenly between the slide and the photomask. Using a mask aligner (ABM, Inc.), the uniform layer of the prepolymer solution was exposed through the photomask to 365 nm, 14 mW/cm² UV light for 8 s. The design of the photomask determines the pattern of microdisc array, which is 200 μm in diameter and 1 mm in center-to-center distance. Microdiscs were then formed via polymerization of the prepolymer mixture and anchored by the acrylate group on the silanized glass surface. Afterward, unreacted mixture was removed by thorough washing with ethanol.^26–28

PET/PDMS Frame Attachment to the Microdisc Array.

PET (0.3 mm thick) was pretreated with (3-acryloxypropyl)-trichlorosilane using the same protocol described for glass silane modification. PDMS prepolymer was prepared using a 25:1 (w/w) mixture of base to curing agent. The liquid PDMS was spin-coated on silane treated PET at 1000 r.p.m. to form a 90 μm thick layer and then cured on a hot plate at 120 °C. The PET/PDMS membrane was subsequently cut with a CO₂ laser machine (Universal Laser Systems, Inc., VersaLaser 2.30) into a frame featuring 2560 through holes (500 μm × 500 μm each). This frame was then attached to the glass by the self-adherent properties of PDMS such that each disc aligned with a corresponding through hole on the frame.

Microfluidic Printing Setup.

The MPS printing platform is composed of a printer head, an X–Y traveling stage (DDS 220, Thorlabs Inc.), controlling circuits, and software. The traveling stage (2 × DDS 220) and controller (BBD202) were purchased from Thorlabs Inc., US, while the controlling circuits and software were customized at MiNILab, UC Davis. These custom components are responsible for coordinating the traveling stage and the printer head, in order to generate droplet arrays of any desired pattern. The printer head consists of a microfluidic cartridge, which was fabricated by layer-to-layer bonding of laser-machined PDMS (polydimethylsiloxane), and an electromagnetic valve array (the Lee Co.). The valve array is connected to a compressed air pump (Barnant Co.) via a pressure regulator (Omega) that regulates the air pressure.

Microfluidic Print-to-Synthesis.

Coupling solution was prepared by dissolving the Fmoc-protected amino acid of choice (200 μM), 2-(6-chlor-1H-benzotriazol-1-yl)-1,1,3,3-tetramethylaminium-hexafluorophosphate (HCTU) (200 μM), and N,N-diisopropylethylamine (DIPEA) (400 μM) in anhydrous N-methyl-2-pyrrolidone (NMP).²⁹ The Fmoc-amino acid coupling solution was then loaded onto the cartridge of the print head. Twenty nanoliters of coupling solution was printed onto the preset spots on the substrate. After 1 h, the chip was washed copiously with dimethylformamide (DMF), dichloromethane (DCM), and ethanol (EtOH) (repeated three times per solvent), followed by drying with compressed air. To prepare the library molecule for the next coupling, the Fmoc protecting group was cleaved by addition of 20% (v/v) 4-methylpiperidine in DMF for 30 min, during which the solution was refreshed at 15 min. Finally, the DMF/DCM/EtOH washing step was repeated to wash the chip. The traveling stage was at rest during the coupling or Fmoc-deprotection reaction. Coupling and Fmoc-deprotection was repeated in sequence for each amino acid. After the last Fmoc-deprotection step, the peptide chains were N-capped with UPA, using the same coupling procedure. Following completion of the desired sequence, side-chain protecting groups were then removed by submerging the entire chip in a modified Reagent K solution, composed of 5% phenol (w/v), 5% deionized (DI) water, 5% thioanisole, 2.5% triisopropylsilane (TIS), and 82.5% trifluoroacetic acid (TFA) (all % are v/v except phenol that is w/v) for 3 h. Final DMF/DCM/EtOH washing steps were done following two rounds of washing with 2% DIPEA in DMF which quenches the TFA.

Cell-Binding Screening for MPS Library.

Immortalized malignant human T lymphocytic cell line Jurkat cells, stably transfected with green fluorescent protein (GFP), were maintained at 5% CO₂, 37 °C in a humidified incubator (VWR International, LLC) in RPMI-1640 media supplemented with 10% (v/v) FBS (fetal bovine serum) and 100 units/mL penicillin.³⁰ For screening, Jurkat-GFP cells, suspended at 2 × 10⁶ cells mL⁻¹, were added on top of the library chip and allowed to mix with gentle agitation for 2 h at 5% CO₂, 37 °C. Nonbinding cells were removed by gentle washing with phosphate buffered saline (PBS) thrice. Remaining cells were considered bound to positive sites via interaction with cell surface molecules. The chip was kept in fresh PBS and promptly examined by confocal laser scanning microscopy (CLSM). A Zeiss LSM 800 inverted microscope with a 5× EC-Plan Neofluar objective was used to scan the entire chip surface to produce a tiled map. Green fluorescence was used to detect whole cells (via GFP signal, Figure 5a), and brightfield images were used to localize the position of each microdisc. A custom Matlab (Mathworks) script was written to automate the quantification of cell binding to each disc (Figure 5b).

Figure 5.

Fluorescent image of Jurkat cell binding to the peptide microarrays, under different magnifications (scale bar: 0.5 mm).

RESULTS AND DISCUSSION

Microfluidic Print-to-Synthesis Operations.

The MPS system (Figure 1) includes a microfluidic printing platform controlled by custom digital circuitry connected to a computer station with a programmable user interface, which allows for fully automated synthetic operation. The key components of the MPS platform are (i) a multichannel microfluidic cartridge made of silicone elastomer (polydimethylsiloxane, PDMS) and (ii) a pneumatic driving mechanism that directs individual microchannels through a three-way electromagnetic switch (Figure 1b). In response to electrical commands, each switch can direct compressed air to the corresponding microfluidic channel for on-demand droplet printing. Accordingly, the droplets are ejected from the microfabricated nozzle by the pneumatic pressure pulses. Our previous reports on a similar microfluidic printing platform describes in detail our ability to control droplet size by modifying the geometric designs of both the microchannels and nozzles.^22,23 Moreover, the droplet size can also be tuned by modifying the compressed air pressure and the pressure pulse waveforms (i.e., the duration of valve opening).

Figure 1.

Scheme for the microfluidic print-to-synthesis (MPS) platform. (a) A computer interface is used to automate the position of a stage (holding the library chip) and the pressure at each solenoid valve positioned above a microfluidic cartridge. (b) The pneumatic-driven droplet mechanism features an electromagnetic valve that can be closed (idle state) or open (printing state) to compressed air that drives droplet formation through a capillary tube separating the valve from the microfluidic cartridge. The compressed air pressure, duration of valve opening, and cartridge dimensions each contribute to droplet sizes.

The microfluidic cartridge is fabricated (details can be found in the Experimental Section) by serial stacking and bonding of three PDMS layers: the top and bottom structural layers and the middle microchannel layer. The top layer features access reservoirs for chemical loading, while the bottom layer is embedded with printing nozzles. All synthetic reagents are dissolved with either NMP or DMF, both compatible with the hydrophobic PDMS and thus automatically prime the microchannels. Therefore, unlike PDMS channels that must carry aqueous solutions, our channel surfaces can be left chemically unmodified after bonding. The geometry of the microfluidic cartridge is designed such that up to 50 μL of Fmoc-protected amino acid solution can be loaded into each channel, with reliable and consistent printing until the residual volume reaches 1 μL. This allows for minimal chemical waste without reloading during the printing of the entire microarrays. The microfluidic cartridge is connected to the solenoid valve array via capillary tubes in a plug-and-play assembly to form a detachable and low-cost PDMS cartridge design, highly desirable for biochemical synthesis where contamination and cross-talk pose a considerable concern. Traditional inkjet cartridge replacement cost is high, such that, in practice, the cartridge is often recycled, making it difficult to avoid contamination. In our MPS platform, the cartridge is inexpensive and disposable and, thus, a good fit for biomedical applications. Another reason to use PDMS is due to its chemical inert property to Fmoc chemistry in addition to its layer-by-layer processing.³¹ Other materials have to satisfy these requirements to be considered as alternatives. In addition, the microfluidic synthetic system is flexible with regards to multiplexing and scaling up, since one can add and expand the number of input channels (up to 24 in our current setup) to accommodate various chemical inputs for synthesis, e.g., coupling reagents, washing solvents, and amino acid building blocks.

In brief, the proposed microfluidic print-to-synthetic platform comes with fully automated operation, multiplexed chemical reactions, high synthetic throughput, and low chemical consumption and is free of cross-contamination of chemical reagents. To illustrate the automated synthetic procedure enabled by the MPS system, Figure 2 displays the step-by-step chemical reactions of the droplet-based peptide synthesis. A cross-sectional illustration of the process can be seen in Figure S1. The chemical printing is conducted onto the PEG microdisc array immobilized on a silanized glass substrate (Figure 2a), and each microdisc is surrounded by a plastic microwell to prevent lateral diffusion of chemical reagents into adjacent microdiscs. PEG-based polymer has been chosen as the synthetic matrix in this study due to its well-known property of nonspecific binding to biomolecules,³² a desired feature for minimizing false positive results during cell screening. In addition, PEG-based resins, such as ChemMatrix resin, have previously been successfully used for SPPS. In order to form a strong bonding of PEG microdiscs via an acrylate functional group, the glass slide was first coated with (3-acryloxypropyl)trichlorosilane. PEG-containing solution was then casted and subsequently photopolymerized on the silanized substrate by UV exposure, resulting in the formation of a microdisc array chemically bonded to the substrate (Figure 2b). By controlling the thickness of the casting layer, we were able to achieve a minimal PEG matrix thickness of 5 μm, in which the coupling reagent can quickly diffuse through to facilitate synthetic coupling.

Figure 2.

Schematic of the step-by-step chemical reactions of droplet-based peptide synthesis enabled by the MPS system. (a) A silane-coated slide is used as a substrate for (b) an array of PEG-based microdiscs featuring a high density of free amine groups. A PET frame is used to separate each microdisc for subsequent chemical functionalization procedures, including (c) MPS droplet printing of Fmoc-protected amino acid coupling solutions, (d) washing, and (e) Fmoc deprotection solution. Steps (c–e) are repeated until the desired sequences are fully synthesized. (f) The side-chain protecting groups are removed from the library molecules using a modified Reagent K cocktail to generate the final library. The chip remains stable for several months, during which it can be utilized for screening assays, e.g., for binding tumor cells.

To quantify the primary amine loading of the resulted microdisc array, we first removed ten microdiscs from the glass slide with a razor blade and then dried them completely prior to reacting with 9-fluorenylmethyl N-succinimidyl carbonate (Fmoc-OSU). After thorough washing, the coupled Fmoc was removed by reaction with 20% 4-methylpiperidine and collected. UV absorption of the resulted solution, which correlates well with Fmoc concentration, was measured at 276 nm (Nanodrop). Using this method, the amine loading of microdisc was calculated to be 0.21 mmol/g, while the theoretical loading was estimated to be 0.24 mmol/g, assuming that 100% of 2-aminoethyl methacrylate used in the polymerization reaction was incorporated into the PEG microdisc. The amine loading was comparable to that of microbeads commonly used in SPPS, which ranged from 0.1 to 0.6 mmol/g.

Prior to printing, a laser-machined PET microwell frame was attached (via PDMS adhesion between the PET and silanized glass substrate) and aligned with the microdisc array in order to provide a physical barrier to prevent cross-contamination of chemical reagents. As PET is impermeable to NMP, no diffusion across adjacent spots can occur. Standard Fmoc chemistry was applied for peptide synthesis, during which mixtures of Fmoc-protected amino acids and appropriate coupling reagents were printed onto predetermined discs for amino acid coupling (Figure 2c). Between coupling steps, the microdiscs were thoroughly washed with DMF and dried under air (Figure 2d). Accordingly,¹¹ Fmoc-protecting groups were then removed by incubating the whole chip with 20% 4-methylpiperidine in DMF (Figure 2e). This process was repeated until the desired sequences were formed. Finally, the PET frame was removed, and Reagent K cocktail treatment was applied to cleave side-chain protecting groups (Figure 2f). To measure the efficiency of each coupling cycle, a standard bromophenol blue test was carried out. As shown in Figure S2, the presence and absence of blue color were reported on the state of amine functionality (i.e., coupling/deprotection efficiency) at each step. Thus, the bromophenol blue test can be conveniently used to monitor the progression of peptide synthesis on the microdisc.³³ According to the bromophenol blue test, the time required to ensure coupling for each cycle was 1 h. Since the PEG microdisc is as thin as 5 μm, the diffusion is probably not the limiting factor to the reaction, given the high permeability of PEG material.³⁴

Droplet Size Characterization.

As mentioned above, droplet size can be controlled by the geometrical design of the microfluidic cartridge, particularly, the nozzle diameter and length, which have been investigated in our previous efforts in characterizing microfluidic impact printing.^22–24 However, modification of microfluidic cartridge geometry involves resource-intensive fabrication. Instead, we have investigated additional control mechanisms in the pneumatic-driven printer, i.e., the air pressure ($P$) and the pulse duration ($T$) of the compressed air. On the basis of Poiseuille’s law for laminar flow, the pressure-driven fluidic volume ($V$) can be estimated as³⁵

$$V=QT=\frac{PT\pi r^4}{8\mu t}$$ (1)

where $\mu$ is fluid viscosity, and $t$ and $r$ stand for nozzle length and diameter, respectively. In order to determine the optimal printing speed and minimum droplet size for our MPS platform, we varied both pulse duration (from the smallest possible of 2 ms up to 12 ms) and also source pressure level. Figure 3a illustrates that droplet size can be manipulated by varying the compressed air pressure, which follows a linear relationship as evident by a well-fitting curve. In the range from 1.5 to 3.5 psi, the droplet size increased from approximately 6 to 15 nL. On the other hand, the pulse duration has a greater impact on droplet size, ranging from ~6 to 450 nL at a fixed pressure of 1.5 psi. (Figure 3b). The linear relationships show the experimental consistency with the theoretical prediction. The fitted slopes also match with calculated ones (0.60 for volume vs pressure, 41.6 for volume vs pulse) when the parameters are substituted (r = 40 μm, t = 250 μm). Errors of the slopes and the intercept are mainly due to valve response time, since the valve does not actually open with less than 2 ms of voltage pulse. Essentially, droplet size can be easily controlled by the pulse duration and source pressure, without varying the geometric design of the cartridge. In this study, 15 kPa pressure and 2 ms pulse were used to generate a droplet of 10 nL, while two droplets were printed on each coupling site, resulting in a total volume of 20 nL per spot.

Figure 3.

Droplet size was measured across various (a) air pressures or (b) pulse durations (valve opening).

Library Construction.

To demonstrate the capability of the MPS platform to generate functional peptide libraries, we designed a model library based on a known motif previously reported to bind to α4β1 integrin overexpressed on the surface of human lymphoid cancer cells.¹¹ The membrane protein α4β1 integrin plays an important role in cancer metastasis and development.^36–39 In the past few years, α4β1 integrin has proven to be excellent therapeutic or imaging targets for lymphoid malignancies.^40,41 In particular, the molecule LLP2A (discovered via combinatorial library screening by Lam and co-workers) was reported to have very high binding affinity (IC₅₀ = 2 pM) to the activated form of the α4β1 integrin expressed on human malignant lymphoid cells.¹² LLP2A is a tripeptide-based peptidomimetic related to the well-known Leu-Asp-Val (LDV) motif,^42,43 with N-terminus capped by UPA. The tripeptide portion of LLP2A is comprised of unnatural amino acids: ε-6-(2E)-1-oxo-3-(3-pyridinyl-2-propenyl)-l-lysine]-[Aad]-[1-amino-1-cyclohexane carboxylic acid].

On the basis of the LDV motif in LLP2A, we designed a tetrameric focused peptide library capped at the N-terminus with UPA. The subscript notation followed the standard N to C terminus naming convention; thus, X₁, X₂, X₃, and X₄ represent the last, third, second, and first amino acids that were coupled to the substrate, respectively. For each position X₁, X₂, X₃, and X₄, we chose 8, 4, 8, and 5 different amino acids, respectively, resulting in 1280 permutations (full description of amino acid selection can be found in Table S3). Two copies of the library were synthesized on a 2″ by 3″ glass slide, resulting in a density of 2560 spots/slide.

Practically, different amino acids were printed under guidance of a matrix directed algorithm, in which repeating and alternating patterns were applied vertically and horizontally. Using this strategy, every possible combination of the amino acids in the library can be formed. To verify the four-step print-to-synthesis process, we visualized the MPS platform’s capabilities for depositing distinct coloring solutions representing different amino acids (Figure 4). It is important to point out that it is not necessary to use the same types or numbers of amino acids in each round.¹² In fact, this design strategy can be adapted to any peptide library with any length and unique library members with little or no sequence homology among them.

Figure 4.

Four-step design strategy for generating every possible library compound, illustrated by printing food coloring solutions. Library construction proceeds via printing (a) repeating patterns vertically, (b) alternating patterns vertically, (c) repeating patterns horizontally, and (d) alternating patterns horizontally. Scale bar is 2 mm.

Cell Binding Result.

Given that our test library features the LDV motif known to bind α4β1 integrin expressed on T lymphocytes, we screened our library for in situ binding to malignant Jurkat T-lymphoid cells. GFP expressing Jurkat cells were used to facilitate quantification of cell binding at the end of the binding experiment. Therefore, after incubation, cell attachment on microdiscs was observed by confocal laser scanning microscopy (CLSM) imaging. The binding results for the whole chip with 2560 spots were shown in Figure 5, where bound Jurkat cells were detected by green fluorescence. PEG microdiscs were circled in red for ease of visualization. A section of brightfield cell binding is shown in Figure S4. The resulting number of cells binding to each spot was estimated with a custom Matlab script and shown in Figure S5, where z axis was the cell count and x–y plane corresponded to the spatial information on each disc. Rectangular areas marked in red in Figure 5 correspond with the ones marked in Figure S5. We observed only little nonspecific cell binding to the chip, where 40% of the discs (negative spots) have 0 cells binding according to the cell counting result. This confirmed the presumed low nonspecific binding property of the PEG discs; thus, we are confident that this platform largely eliminates false positive binding.

In order to verify the stability and repeatability of the MPS synthesis method, the same chip was resynthesized twice more, each with two copies of the 1280-membered library. Cell binding assays were carried out on each of the resulting four sets of peptide microarrays; i.e., a total of 6 sets of 1280 peptides were screened. The repeatability of these binding experiments was then evaluated by the following method. First, the cell binding strength against each spot was categorized into weak (<5 cells/disc), medium (5–20 cells/disc), and strong binding (>20 cells/disc). If the binding affinity level for one spot stayed the same for all six copies, then the result was considered repeatable for that spot. Experimental repeatability was calculated to be 91.1% across the six copies (i.e., 91.1% of all spots exhibited the same binding affinity classification across all six copies of the same peptide), which indicated excellent repeatability of our MPS method in peptide microarray synthesis and screening.

The relative binding affinity of Jurkat cells to the entire 1280 peptides on the microarrays is shown in Figure 6. The peptide sequence of each spot on the microarray can be addressed as X₁X₂ and X₃X₄ on the two axes, while the value (0–25) depicts the average number of cells bound to that spot according to Matlab calculation. The color contour map can provide useful information on the structure–activity relationship (SAR) of peptides to cell binding. It is clear from Figure 6 that some of the strongest binding peptides have the following motifs: UPA-Leu-homoAsp-XX, UPA-Nle-homoAsp-XX, UPA-Leu-Aad-XX, UPA-Nle-Aad-XX, UPA-Leu-Asp-XX, and UPA-Nle-Asp-XX. In addition, the Jurkat cell screening results demonstrated that the fourth position played an important role in determining the binding affinity. For example, the binding affinity of UPA-XXX-Glu and UPA-XXX-d-Asp is stronger compared with XXX-d-Glu and XXX-Asp. Proline addition to the C-terminus, i.e., X₄ position, further weakened cell binding, particularly when X₃ was D-Asp, Asp, Ile, or Leu. Of the 1280 peptides on the arrays, four peptides stood out as the top binders: UPA-Nle-Asp-Phe-Glu, UPA-Nle-homoAsp-Glu-Glu, UPA-Nle-homoAsp-Phe-d-Asp, and UPA-Leu-homoAsp-Phe-d-Asp.

Figure 6.

Binding profile of Jurkat cells to the 1280 peptide microarrays. The contour map shows the average number of bound cells per spot across six experimental repeats.

The MPS platform is highly complementary to the OBOC method. Although the OBOC method has the advantage of large diversity (10⁶–10⁷ peptide members), it is not spatially addressable, thus subsequent time-consuming and expensive decoding of positive beads is required. Due to those limitations, we consider OBOC to be most suitable for discovery of lead compounds. The MPS platform, on the other hand, is spatially addressable, but the diversity is significantly lower (10³–10⁴); therefore, it is more suited to lead compound optimization and SAR studies. In the past few years, we have expanded cell-based screening of OBOC libraries to functions beyond cell adhesion, such as intracellular cell signaling and apoptosis.⁴⁴ It is likely that our system could be applied to those cases in the near future or even further expanded to include rapid epitope mapping of antibodies and tumor antigenic peptides.

CONCLUSION

We have introduced the application of our newly developed MPS system for combinatorial peptide microarray synthesis. The MPS platform has several desirable features: (i) use of standard, robust Fmoc-chemistry with many commercially available Fmoc-amino acids, (ii) an inexpensive system and reagents with only a small amount of consumption needed, (iii) high printing speed/throughput, (iv) a modular microfluidic cartridge design that can accommodate multiple reagent reservoirs, without cross-contamination, (v) a small physical footprint, and (vi) the potential of being fully automated.

Applying the MPS platform, we have achieved microdisc array fabrication at high spatial density (2560 spots per 6 square-inches), with high peptide coupling efficiency. No specialized amino acid building blocks (e.g., Nvoc-protected amino acids) nor complicated or expensive instrumentation were needed. The resulting microdisc array exhibited very low nonspecific cell binding and, thus, is ideal for cell adhesion assays. We established proof-of-concept for our MPS platform by synthesizing a UPA-X₁X₂X₃X₄ library with 1280 permutations on a cross-linked PEG microdisc substrate. The binding profile of Jurkat human T lymphoid cancer cells to the entire microarrays of 1280 peptides provided a comprehensive SAR profile on interaction between a focused peptide library with activated α4β1 integrin protein overexpressed on Jurkat cells. Such cell-based assays can be readily applied to the new MPS platform. This simple yet robust MPS platform can be easily expanded to the preparation of larger sets of microarrays or the use of more amino acid building blocks. We expect the MPS technology to soon be invaluable for peptide drug discovery and peptide research.