Boronate-mediated covalent and oriented immobilization of antibodies on the PDMS surface toward improved capture of circulating tumor cells

Abstract

Microfluidic immunoassays are crucial for early detection and capture of circulating tumor cells (CTCs). The method of immobilizing functional receptors, such as antibodies (Abs), plays a critical role in determining the effectiveness of these systems. In this study, we present a microfluidic channel functionalized with boronic acids to facilitate the directed immobilization of native Abs, thereby improving their interaction with target antigens and cells. We evaluated the selectivity and efficiency of CTC capture using the anti-epithelial cell adhesion molecule (EpCAM) as the capture Ab. Using EpCAM-positive PC-9 human pulmonary adenocarcinoma cells and EpCAM-negative HeLa cervical cancer cells as models, our comparisons revealed that oriented Ab immobilization through covalent boronate formation resulted in approximately 5.2 times more PC-9 cell capture compared to random covalent Ab immobilization. Additionally, directional Ab immobilization demonstrated a roughly 30.8-fold increase in selectivity for EpCAM-expressing CTCs. This versatile Ab immobilization platform offers a promising approach for selective cell capture under dynamic flow conditions.

---

Introduction

Circulating tumor cells (CTCs) are crucial biomarkers for predicting cancer recurrence and monitoring disease progression.¹ Despite their rarity in physiological samples, advancements in microfluidics, nanomaterials, and biofunctional surfaces have significantly improved CTC capture efficiency, overcoming major technical challenges.^2,3 These innovations have facilitated the development of sensitive and efficient enrichment and detection strategies, offering valuable diagnostic insights for precision medicine and guiding personalized therapeutic interventions. Consequently, there remains an urgent need for rapid, cost-effective, and high-throughput methods to efficiently capture and analyze CTCs.⁴

Microfluidic devices, renowned for their ability to process minimal sample volumes with high capture efficiency, provide a powerful platform for isolating and monitoring viable biomarkers at a single-cell resolution.⁵ Among the various CTC isolation approaches, adhesion-mediated cell immobilization has gained prominence due to its high specificity and reproducibility.⁶ Immunoaffinity-based capture, which utilizes antibodies (Abs) targeting surface markers on CTC membranes, has demonstrated superior efficiency in numerous studies.⁷ One of the most commonly targeted biomarkers for immunoaffinity-based CTC isolation is the epithelial cell adhesion molecule (EpCAM),⁸ a transmembrane glycoprotein highly expressed on most solid tumor cells but absent in hematologic cells.⁹ This specificity makes EpCAM an ideal target for CTC isolation across various microfluidic and bioanalytical platforms.^10–12 While immunoaffinity-based approaches offer high purity and sensitivity, their overall effectiveness depends on the stability, homogeneity, and orientation of the immobilized Abs.¹³

A range of Ab immobilization techniques have been developed, each varying in complexity and immobilization efficiency.¹⁴ Physical adsorption, the simplest method, is widely used but suffers from instability and loss of Ab functionality due to weak interactions with the substrate.¹⁵ In contrast, covalent immobilization via amide bond formation enhances stability by creating irreversible linkages between Abs and surface functional groups.^16,17 However, this conventional approach often results in random Ab orientation, reducing antigen accessibility and capture efficiency.^13,14 To address these limitations, bioaffinity-based oriented immobilization strategies have been developed.¹⁸ The avidin–biotin interaction, a well-established bioaffinity approach, has been extensively employed for EpCAM-based CTC capture.^6,19–23 Although this method improves efficiency and selectivity, it requires multiple surface functionalization steps, increasing complexity, production time, and cost.²⁴

Boronic acids (BAs) have emerged as promising biofunctional ligands due to their ability to form reversible covalent bonds with Lewis bases under physiological conditions.^25,26 BA-based probes have been widely used for carbohydrate detection, molecular recognition, and bio-orthogonal protein conjugation.^27,28 In our previous studies, we developed BA-based affinity ligands capable of forming covalent and irreversible bonds with native, unmodified full-length Abs.^29–31 These ligands facilitate site-specific attachment of capture agents (e.g., Abs) on solid supports, such as magnetic nanoparticles and glass substrates. BAs and their derivatives form reversible boronate diester complexes with N-glycan chains of Abs, specifically at the Fc C_H2 domain, where glycan moieties contain N-acetylneuraminic acid, galactose, mannose, and fucose residues.^26,32 Thus, BA-functionalized microfluidic substrates present an attractive strategy for regioselective Ab immobilization.

In this study, we introduce a novel boronate-affinity-based oriented Ab immobilization strategy within PDMS-based microfluidic channels. Unlike conventional immobilization methods that rely on random amide bond formation, we employed a BA-tosylate (BA-tosyl) probe (2)³¹ functionalized with an azide group. This probe was grafted onto the microfluidic device via highly efficient click chemistry (Fig. 1). The resulting BA-derivatized PDMS surface facilitates specific and oriented Ab immobilization through a two-step process: (1) reversible boronate formation and (2) irreversible covalent bond formation via proximity S_N2 substitution. This strategy ensures precise Ab orientation, preserving antigen-binding activity while enhancing capture selectivity and efficiency. We evaluated the capture performance of our platform using anti-EpCAM Ab for selective CTC isolation. PC-9 pulmonary adenocarcinoma cells (EpCAM-positive) and HeLa cervical cancer cells (EpCAM- negative) were employed as model systems. Our findings demonstrated a 5.2-fold increase in PC-9 cell capture efficiency with boronate-affinity oriented Ab immobilization compared to the random covalent method. Additionally, directed Ab immobilization resulted in a 30.8-fold higher selectivity for EpCAM-expressing CTCs. This platform offers a robust and cost-effective approach for selective CTC capture under dynamic flow conditions, paving the way for enhanced Ab-based bioanalytical applications.

Fig. 1 (A) Schematic illustrating the combined use of the surface functionalization by the CuAAC reaction using a boronate-affinity probe on an alkynylated PDMS surface followed by a regioselective S_N2-type reaction between nucleophile amino acid side chains from the Fc region resulting in irreversible attachment of a full-length Ab to enable capture of CTCs using a microfluidic device. (B) Structures of BA-containing probes and small-molecule compounds used in this study.

Materials and methods

Reagents and instruments

3-(Dimethyl-(3-(trimethoxysilyl)propyl)ammonio)propane-1-sulfonate (sulfobetaine siloxane, SBS) was purchased from Gelest Inc. (3-Aminopropyl)-triethoxysilane (APTES) was purchased from Sigma-Aldrich. Herceptin (trastuzumab, anti-HER2-Ab) was obtained from Roche. Anti-human IgG (Fab' specific) Ab conjugated to Cy3 was purchased from Jackson Immunoresearch (#109-165-170). Human EpCAM/TROP-1 Ab (anti-EpCAM-Ab) was sourced from R&D systems (# MAB960). Bovine serum albumin (BSA) and Cy3-N₃ were purchased from Sigma-Aldrich. Polydimethylsiloxane (PDMS) was purchased from Dow Corning. Fluorescence imaging was performed using a SpinScan microarray scanner (Caduceus Biotechnology Inc.) and ChemiDoc MP imaging system (Bio-Rad).

Fabrication of the PDMS microfluidic device. PDMS microfluidic devices were fabricated using two distinct master molds (Fig. 2): (1) a single-channel serpentine mold, designed for individual sample analysis and (2) a multi-channel mold, capable of conducting up to eight tests simultaneously to enhance experimental efficiency. Samples (0.5–1 mL) were introduced at a rate of 200–400 μL h⁻¹.

Fig. 2 Whole image of two different mold types of PDMS microchannels used.

Master mold preparation. A 3 mm thick PMMA sheet was milled using a 3-axis CNC machine (Roland EGX-400) with a 1 mm diameter square end mill. After machining, the PMMA mold was cleaned with isopropanol (IPA) and vapor-polished with dichloromethane (DCM) for 5 seconds to smooth the surface. The prepared master molds were secured inside a 15 cm Petri dish before PDMS casting.

PDMS casting and device assembly. A PDMS base and a curing agent were mixed at a 10 : 1 ratio, poured over the master molds, and degassed in a vacuum desiccator for 2 hours to remove air bubbles. The mixture was then cured at 80 °C for 2 hours. Once polymerized, the PDMS layer was peeled off, and inlets and outlets were punched. Both the PDMS membrane and glass slide were cleaned with IPA, sonicated in an ultrasonic cleaner (DELTA DC150H) for 5 minutes, and dried under nitrogen gas. Oxygen plasma treatment was performed using a plasma cleaner PDC-32G (Harrick Plasma PDC-32G) at 0.6 Torr for 2 minutes, enabling permanent bonding between the PDMS slab and the flat glass slide.

General surface modification procedure. Plasma-activated PDMS microfluidic channels were treated with a 100 mM solution of alkynyl triethoxysilane 1 in ethanol (EtOH) and a zwitterionic molecule, SBS (CAS:151778-80-2) in water (v/v = 98/2) at room temperature (rt) for 3 hours. The device was washed with 200 μL EtOH and baked at 80 °C for 30 minutes to ensure silane polymerization. Next, the alkynylated surface was incubated in a reaction mixture containing: 5 mM BA-tosyl 2, 2 mM THPTA, 1 mM CuSO₄, 1 mM sodium ascorbate (NaAsc), and 1 mM triethylamine in 75% DMSO and 25% H₂O at rt for 16 hours to enable the click chemistry reaction. After incubation, the device was thoroughly washed with 200 μL of DMSO, and unreacted alkynes were capped with 100 mM MeO-PEG-N₃3 in 50% DMSO/50% H₂O for another 16 hours at rt. The final surface was washed with 500 μL of DMSO and deionized water (ddH₂O) before incubation with Herceptin® in 20 mM HEPES buffer (pH 8.4, containing 150 mM NaCl) at rt. Afterwards, the device was sequentially washed with PBST (1% Tween20 in PBS), PBS, and ddH₂O (500 μL each). To prevent non-specific adsorption, the device was blocked with 5% BSA in 100 μM dextran solution at rt for 2 hours, followed by final washes with PBST (0.05% Tween20 in PBS), PBS, and ddH₂O (200 μL each) before drying under N₂ gas.

Optimization of silanization on PDMS. To assess silanization efficiency, alkynylated PDMS channels were subjected to a copper-catalyzed azide–alkyne cycloaddition (CuAAC) with 50 μM Cy3-N₃ in DMSO (containing 2 mM THPTA, 1 mM CuSO₄, 1 mM NaAsc) at rt for 16 hours. Negative controls were prepared by (1) incubating alkynylated devices with Cy3-N₃ but without click reagents and (2) incubating bare (unfunctionalized) PDMS channels with Cy3-N₃ in the presence or absence of click reagents. After incubation, all devices were washed with DMSO and dried with N₂ gas, and fluorescence signals were measured.

Suppression of non-specific Ab adsorption. To minimize non-specific binding, alkynylated PDMS devices were subjected to CuAAC reactions using capping agents, either HO-PEG-N₃4 or MeO-PEG-N₃3 at rt for 16 hours. The device was thoroughly washed with DMSO and ddH₂O, followed by Herceptin incubation (1.6 μM at 4 °C for 6 hours). After washing, the device was treated with 5% BSA. After an additional washing step, the device was incubated with anti-human IgG (Fab' specific) Cy3-labeled Ab (2 μg mL⁻¹, rt for 2 hours). Fluorescence signals were measured after thorough washing.

Oriented immobilization of Ab. Alkynylated PDMS microfluidic channels were modified via a CuAAC reaction with BA-tosyl 2, BA-N₃5, and linker-tosyl 6, respectively, at rt for 16 hours. Unreacted alkynes were capped under CuAAC conditions followed by thorough washing. The devices were incubated with Herceptin (1.6 μM) at 4 °C for 24 hours, and washed. Surfaces were subsequently blocked with 5% BSA in 100 μM dextran solution at rt for 2 hours, followed by washing with PBST (1% Tween20 in PBS, pH 6.8). For downstream applications, devices were either probed with anti-human IgG (Fab' specific) Ab-Cy3 (2 μg mL⁻¹) for fluorescence measurements, incubated with biotinylated HER2 (10 μg mL⁻¹) for antigen binding assays, or used directly for cell-capture studies.

Proof of the covalent immobilization of Ab. To confirm covalent bonding, two alkynylated PDMS devices were subjected to a CuAAC reaction with BA-tosyl 2 and BA-N₃ (5) probes at rt for 16 hours, washed and capped with MeO-PEG-N₃3 (100 mM in click reaction solution). Devices were incubated with Herceptin (1.6 μM) at 4 °C or rt for 24 hours. After washing, 60 mM fructose in PBST (1% Tween20 in PBS, pH 6.8) was pumped through at 300 μL h⁻¹ at 37 °C for 3 hours to assess the stability of immobilized Abs.

Random Ab immobilization by amide bond formation. A plasma-activated microfluidic device was treated with 10% APTES in EtOH (rt for 3 hours), washed, and baked at 80 °C for 30 minutes. The amine-functionalized surface was treated with 2.7 mM disuccinimidyl substrate (DSS) in DMSO (rt for 2 hours), washed, and incubated with Herceptin or anti-EpCAM Ab (100 μg mL⁻¹ in PBS, pH 7.4, rt for 24 hours). After blocking with 5% BSA in PBS (2 hours), the device was used for cell-capture studies.

Cell culture, capture, and imaging. The non-small cell lung cancer cell line PC-9 was kindly shared by Dr. Yu-Ju Chen (Institute of Chemistry, Academia Sinica) and was initially obtained from RIKEN BioResource Research Center (Japan, catalogue number: RCB4455). HeLa cells were a gift from Dr. Jye-Chian Hsiao (Institute of Chemistry, Academia Sinica) and were originally obtained from the American Type Culture Collection (ATCC). PC-9 (lung cancer) and HeLa (cervical cancer) cells were cultured in RPMI-1640 medium supplemented with 10% (v/v) fetal bovine serum (Invitrogen, San Diego, CA, USA) and 1% (v/v) penicillin–streptomycin (P/S, Hyclone) in a humidified atmosphere at 37 °C and 5% CO₂. Cells were stained (with 0.1 μM HCS NuclearMask™ Red) and harvested by using accutase (Thermo Fisher). Cells were counted before being spiked into functionalized microfluidic devices. Cells were incubated under controlled flow conditions, washed and imaged using a laser scanning confocal microscope (LSM 700, Zeiss). Captured cells were quantified using the Image J (Fiji) cell counter plugin.

Results and discussion

Fabrication and characterization of the PDMS microfluidic device

Oxygen plasma-activation is a widely employed method to facilitate permanent covalent bonding in PDMS-based microfluidic devices,³³ enabling surface functionalization. In this study, oxygen plasma-activated PDMS channels were incubated with a 100 mM solution of alkynyl triethoxysilane³¹ at room temperature (rt) for 3 hours, forming an alkynylated layer on the device surface (Fig. 1, STEP 1). Initially, silanization was conducted in solvents such as DMSO or ethanol.^34,35

To further minimize non-specific protein adsorption, a 2 mol% aqueous solution of SBS, a zwitterion siloxane known for antifouling properties, was simultaneously introduced during the silanization step to ensure uniform incorporation into the modified PDMS layer (Fig. S1). Under appropriate conditions, the alkynyl functionalized PDMS surface provides reactive sites for the covalent attachment of affinity molecules. To evaluate the effectiveness of surface functionalization, a CuAAC reaction (Fig. 1, STEP 2) was employed to incorporate a fluorescent dye (Fig. S2).^36,37 For the CuAAC reaction, the alkynylated PDMS microfluidic devices were incubated with Cy3-N₃, a fluorophore-containing azide, in the presence of CuSO₄, sodium ascorbate, and tris(3-hydroxypropyltriazolylmethyl)amine (THPTA), a Cu(I)-stabilizing ligand.³⁸ The observed high fluorescence intensity confirmed the successful incorporation of alkynes on PDMS channels (Fig. 3A). Furthermore, it was found that ethanol was a more efficient solvent than DMSO,³⁴ particularly when heated to 80 °C for 30 minutes, significantly improving alkyne immobilization,³⁵ and yielding higher fluorescence intensity (Fig. 3A, right panel). Additionally, control experiments confirmed that the fabricated PDMS surface exhibited minimal or no physical adsorption of azide-containing fluorophores, suggesting that non-specific binding of dye was negligible (Fig. 3B). Fluorescence signals only appeared when CuAAC reactions were performed (facilitating 1,2,3-triazole formation), further confirming successful surface modification (Fig. 3B).

Fig. 3 Silanization conditions on the PDMS surface. (A) Impact of the silanization process in DMSO or EtOH at varying temperatures. (B) Evaluation of the surface click reactions on alkynylated and bare PDMS substrates under different reaction conditions. Portion of the fluorescence images of the corresponding microchannels are shown above each bar graph. (C) Assessment of non-specific binding in PDMS based immobilization of Abs using BA-tosyl modification and Cy3 conjugated detection. Error bars indicate the standard deviation (SD) from three replicate experiments. Representative fluorescence images for each condition are shown.

Ab immobilization

Following alkynylation, the PDMS microfluidic channels were functionalized using 5 mM BA-tosyl probe 2 (Fig. S3) under click reaction conditions (Fig. 1, STEP 2). Subsequently, Herceptin (1.6 μM in 20 mM HEPES buffer containing 150 mM NaCl and 0.05% Tween20, pH 8.5) was employed as a model Ab to initiate the proximity-driven S_N2 reaction (Fig. S4), leading to covalent and irreversible immobilization on the BA-functionalized surface (Fig. 1, STEP 3). To minimize non-specific interactions, residual BA groups on the surface were blocked with a 100 μM dextran solution⁴⁰ containing 5% bovine serum albumin (BSA) (Fig. 1, STEP 4). This blocking step is essential for preventing unintended protein adhesion, which could otherwise interfere with Ab-specific interactions. By using the optimized surface fabrication and washing conditions to reduce any unbound Abs (see below for detailed studies), successful Ab immobilization was validated by incubating the modified surface with Cy3-conjugated anti-human IgG (2 μg mL⁻¹ in PBS buffer containing 0.05% Tween20, pH 6.8), following fluorescence imaging. The strong fluorescence signals observed on BA-tosyl modified surfaces (Fig. 3C, right panel) confirmed effective Ab immobilization and retention of the Ab's structural integrity.

However, before optimization, a significant fluorescence signal was observed on the unmodified control surface (Fig. 3C, left panel), suggesting strong physisorption of the Ab onto native PDMS. This phenomenon was well-documented and arose from the hydrophobic nature and porosity of PDMS, which allows for the non-specific absorption and adsorption of biomolecules.³⁹ To circumvent this problem, we introduced additional surface capping agents and optimized washing conditions to reduce non-specific binding (see below).

To investigate the impact of capping reagents on Ab adhesion, we tested two polyethylene glycol (PEG)-based azides: methoxy (MeO-PEG-N₃) 3 and hydroxyl (HO-PEG-N₃) 4. These molecules contain a tri-ethylene glycol spacer, which increases surface hydrophilicity and was expected to reduce undesired Ab adhesion. The capping agents were applied after the CuAAC reaction with BA-tosyl probes, reacting with any remaining alkyne groups on the PDMS surface. To assess their effectiveness in reducing non-specific interactions, control experiments were performed in which capping agents were applied in the absence of BA-tosyl probe 2 (Fig. S5). While these capping agents demonstrated potential in decreasing non-specific adsorption, they did not completely eliminate Ab adhesion (Fig. 4B). As a result, we shifted our focus to optimizing the post-attachment washing conditions. Initially, a 0.05% Tween 20 (PBST) solution was used: increasing the Tween 20 concentration to 1% proved to be significantly more effective in removing non-covalently bound Abs. This adjustment led to an 81% reduction in fluorescence signal intensity, indicating a substantial decrease in non-specific Ab adsorption.

Fig. 4 (A) Schematic of the fabrication of microfluidic channels. (B) Optimization of the washing conditions to minimize non-specific Ab adsorption on alkyne-modified surfaces. (C) Comparative analysis of Ab orientation on microchannels via BA–probe interactions. (D) Fructose competition assay confirming a stable covalent and irreversible Ab immobilization via BA-tosyl probe 2. Error bars indicate the SD from three replicate experiments. Representative fluorescence images for each condition are shown.

After addressing non-specific Ab adsorption, we aimed to verify the specific and oriented immobilization of Abs on microchannel surfaces (Fig. 4A). This immobilization strategy was based on the specific interaction between the BA head group in the BA-tosyl probe and the diol groups on the glycan moieties of Abs, facilitating boronate diester formation. To evaluate the role of BA in Ab attachment, we employed control probes: BA-N₃ (5), which lacks the reactive tosyl moiety, and linker-tosyl group (6), which lacks the BA component and serves as a negative control. Following surface modification with these probes, Herceptin was immobilized, and the attached Ab was detected using Cy3-conjugated secondary Abs. As shown in Fig. 4C, surfaces functionalized with both BA-tosyl (2) and BA-N₃ (5) generated strong and comparable fluorescence signals, indicating effective Ab immobilization. The slightly lower intensity observed for 2 likely reflects differences in the effective surface density of immobilized Abs, as the S_N2-driven conjugation at the tosyl moiety affords irreversible attachment.⁴¹ However, BA-N₃ (5) appears to enable a more flexible mode of conjugation, which could enhance fluorophore accessibility, leading to a marginally higher apparent fluorescence signal. In contrast, the linker-tosyl probe (6) resulted in a three-fold decrease in signal intensity, suggesting that the BA affinity head group is essential for mediating Ab attachment, whereas the linker-tosyl alone is insufficient for efficient immobilization. Furthermore, the unmodified control exhibited negligible fluorescence, confirming minimal non-specific binding to the PDMS surface. Collectively, these results validate the effectiveness of boronate-affinity-driven Ab modification on microfluidic surfaces and demonstrate that BA-mediated diol recognition plays a crucial role in enabling site-selective Ab immobilization.

To confirm the formation of an irreversible covalent bond between the BA-tosyl probes and conjugated Abs via S_N2 substitution, we performed comparative analyses using a control BA-N₃ probe 5. As depicted in Fig. 4D, Abs were immobilized on microchannels modified with either BA-tosyl (2) or BA-N₃ (5) at 4 °C for 24 hours, followed by treatment with 60 mM fructose,³¹ a competitive polyol, for 3 hours. The assay results revealed distinct differences in fluorescence signal retention between the two modifications. A 40% reduction in fluorescence intensity was observed for BA-N₃ (5), whereas the BA-tosyl-functionalized surface exhibited a 33% decrease. These results indicate that Ab attachment via glycan-boronate formation, in the absence of covalent S_N2 bond formation, is susceptible to displacement by fructose. Upon increasing the reaction temperature to rt to enhance S_N2 reactivity, the BA-tosyl (2) functionalized substrate exhibited superior signal retention, with only an ∼8% decrease in fluorescence intensity after fructose treatment. This suggests that covalent S_N2 substitution led to irreversible Ab immobilization, preventing significant displacement by fructose. In contrast, the BA-N₃ (5) group continued to exhibit a 34% signal loss, further confirming that Ab immobilization via BA-N₃ (5) occurs through reversible glycan-boronate interaction rather than permanent covalent bonding. These findings strongly support our strategy for oriented, irreversible Ab immobilization within the microfluidic channel, ensuring high stability under competitive binding conditions.

Comparison of the antigen capture efficiency of Ab immobilization methods on microfluidic channels

To assess the accessibility of antigen-binding sites using different Ab immobilization methods, we employed the human epidermal growth factor receptor 2 (HER2) antigen as a model system. We systematically compared the antigen capture efficiency of oriented Ab immobilization with that of random immobilization on PDMS microfluidic channels. For random immobilization via amide bond formation, PDMS channels were treated with (3-aminopropyl)trimethoxysilane (APTES), a widely used organo-silane for silicon surface functionalization. APTES was chosen due to its ability to generate a densely packed monolayer of uniformly distributed amine functional groups on PDMS substrates, enabling further chemical modifications. The amine-functionalized surfaces were then treated with disuccinimidyl substrate (DSS), a homo-bifunctional covalent cross-linker, which introduces N-hydroxysuccinimidyl (OSu)-activated esters. These activated groups form stable amide bonds upon reaction with primary amines on Abs. After immobilization, biotinylated HER2 antigens were introduced to bind specifically to the immobilized Abs.

The presence of a biotin tag on HER2 allowed the use of Cy3-streptavidin as a fluorescent probe, enabling quantitative comparison of antigen binding efficiency between the oriented and random Ab immobilization strategies. Fluorescence intensity analysis revealed a four-fold increase in HER2 antigen binding when using oriented Ab immobilization compared to the random approach (Fig. 5A). This substantial enhancement is attributed to the directional orientation of Abs, which optimizes the accessibility of their Fab domains for antigen binding. By immobilizing Abs in a predefined orientation, antigen–Ab interactions are significantly more efficient, leading to improved binding affinity. Control experiments conducted without probes, Abs, or antigens exhibited minimal fluorescence signal (Fig. 5B), confirming the specificity of the assay. Additionally, antigen concentration assays demonstrated a clear dose–response relationship, with fluorescence intensity decreasing proportionally with antigen concentration (Fig. 5C). Notably, the oriented immobilization technique displayed superior binding efficiency compared to random immobilization, particularly at lower antigen concentrations, highlighting its potential for applications requiring heightened sensitivity.

Fig. 5 Efficacy of functionalized microchannel surfaces for antigen binding: oriented and random covalent Ab immobilization strategies. (A) HER2 antigen binding assay, (B) control experiments, and (C) HER2 concentration-dependent binding assay. Error bars indicate the SD from three replicate experiments. Representative fluorescence images for each condition are shown.

To further contextualize the benefits of our immobilization strategy, we performed a comparative evaluation using a widely adopted Fc-directed approach. Protein G was covalently introduced onto N-hydroxysuccinimide (NHS)-activated PDMS microchannels via amide coupling, followed by Herceptin adsorption, HER2 antigen capture, and streptavidin-Cy3 detection. Abs immobilized through boronate-mediated covalent linkage exhibited efficient HER2 binding and delivered ∼1.2-fold higher fluorescence intensity than the protein G-based method (Fig. S6). This enhanced performance was likely due to the higher affinity-ligand density achievable on BA-tosyl-modified surfaces, resulting in more uniform Ab orientation and improved stability under flow.

Importantly, this comparison provides several practical advantages of the boronate-affinity chemistry. While protein A/G facilitates oriented binding via Fc recognition, its non-covalent interactions are vulnerable to dissociation under washing or hydrodynamic shear. Peptide- or tag-based conjugation strategies, although site-selective, require Ab engineering or chemical modification, which limits applicability to native IgGs. In contrast, our boronate–S_N2 coupling simultaneously affords directional immobilization and irreversible covalent fixation under mild aqueous conditions, without altering the Ab structure. This universal and reagent-free strategy is thus well-suited for functionalizing native full-length Abs directly on soft polymeric substrates such as PDMS, providing robust performance for microfluidic CTC capture applications.

CTC capture efficiency and selectivity with anti-EpCAM Ab modified microfluidic channels

To evaluate the impact of Ab immobilization methods on circulating tumor cell (CTC) capture performance within microfluidic channels, we conducted cell capture assays using EpCAM-positive PC-9 human pulmonary adenocarcinoma cells. A cell suspension (10⁴ cells per mL in DMEM medium) was prepared, and a 1 mL aliquot was introduced into microfluidic channels modified with anti-EpCAM Ab via boronate-mediated immobilization. The channels were incubated for 1 hour to allow sufficient cell–surface interactions. To compare the efficiency of oriented versus random Ab immobilization, a control microchannel was functionalized with anti-EpCAM Ab using random amide bond formation. After incubation, unbound cells were removed by washing, and the captured cells were imaged and quantified using a fluorescence microscope (Fig. S7). As shown in Fig. 6A, boronate-mediated Ab immobilization resulted in a 5.2-fold increase in captured cells compared to the random covalent immobilization method. This substantial enhancement suggests that the oriented Ab immobilization significantly improves cell capture efficiency by optimizing Fab domain accessibility for target cell recognition.

Fig. 6 (A) Fluorescence microscopy images of PDMS microfluidic channels showing PC-9 cell capture under different Ab immobilization strategies, along with quantitative analysis of cell-capture efficiency. (B) Selective capture of target PC-9 cells using oriented Ab immobilization. (C) Capture performance using random Ab immobilization. (D) Negative control without Ab immobilization. Oriented Ab immobilization on the microfluidic devices showed reduced non-specific binding of negative cells. (E) Oriented Ab immobilization enables highly specific capture of PC9 cells from PC9 : HeLa mixtures.

To further assess the selectivity of the Ab functionalized microfluidic channels, we evaluated the performance of anti-EpCAM Ab-modified surfaces. As a negative control, EpCAM-negative HeLa cervical cancer cells, which include both adherent and suspension cell types, were used for comparison. Following the cell capture process and washing steps, representative fluorescence images were obtained (Fig. 6). As expected, the highest number of captured cells was observed for EpCAM-positive PC-9 cells, with a 30.8-fold higher capture efficiency compared to EpCAM-negative HeLa cells (Fig. 6B). In contrast, the surface modified with randomly immobilized Abs exhibited only a two-fold difference in captured cells (Fig. 6C), indicating that random immobilization increases non-specific interactions. The importance of Ab orientation was further validated in control assays where Abs were omitted, resulting in indiscriminate cell attachment to the surface (Fig. 6D). These findings demonstrate the high specificity of boronate affinity-based Ab modification, ensuring preferential capture of EpCAM-positive CTCs while minimizing non-specific binding.

Next, to further verify the specificity towards targeted cell capture under the mixed cell population, PC-9 cells were pre-stained with a nucleus dye (Hoechst 33342), and mixed with HeLa cells in the ratios of 1 : 10 and 1 : 100. These cells were then subjected to the aforementioned capture process and washing operations on the anti-EpCAM Ab-modified surfaces. By counting both the PC-9 and HeLa cells before and after washing with the help of the nucleus dye in PC-9, our results showed that the capture efficiency of PC-9 cells when compared with Hela cells was substantially higher in both the 1 : 10 and 1 : 100 PC-9/Hela cell mixtures (Fig. 6E). Altogether, these results support the notion that EpCAM-positive PC-9 cells can be selectively enriched on the oriented Ab-functionalized surfaces despite being present at a smaller population.

Conclusions

In this study, we successfully developed and validated an innovative antibody immobilization strategy for microfluidic platforms tailored for capturing CTC capture. This method leverages targeted PDMS surface modifications, coupled with optimized washing procedures, to effectively minimize non-specific antibody adhesion and enhance binding efficiency. Our findings demonstrate that the boronic acid-based probe significantly improve CTC capture efficiency and specificity compared to the conventional random immobilization method. This enhancement is attributed to the synergistic integration of bio-orthogonal boronate-mediated regioselective S_N2 reactions with boronic acid-functionalized microfluidic channels, ensuring irreversible and oriented antibody immobilization. This study underscores the critical role of precise antibody orientation and robust surface chemistry in the development of high-performance microfluidic platforms for biomedical applications. The proposed immobilization strategy offers a highly specific and efficient approach for CTC isolation. Future work will focus on clinical validation using patient-derived CTC samples and expanding the platform for multiplexed biomarker detection in complex biological fluids.

Author contributions

Conceptualization: C.-C. L. and H.-L. T.; data curation: K.-H. L., J.-W. C., D.-Y. J. and Y.-J. H.; data validation: K.-H. L. and A. K. A.; supervision: C.-C. L. and H.-L. T.; funding acquisition: C.-C. L. and H.-L. T.; writing – original draft: A. K. A.; writing – review & editing: A. K. A., C.-C. L. and H.-L. T.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data supporting this article have been included as part of the supplementary information (SI). Supplementary information: optimization of zwitterionic composition on PDMS microchannels (Fig. S1); silanization conditions for PDMS microchannels (Fig. S2); BA-tosyl probe concentration (Fig. S3); Ab concentration (Fig. S4); evaluation of capping reagents (Fig. S5); comparison with protein G-based Ab immobilization (Fig. S6); and representative fluorescence images of microfluidic channels (Fig. S7) See DOI: https://doi.org/10.1039/d5lc00862j.

Acknowledgements

This work was financially supported by Academia Sinica (AS-GC-111-M03) and the Ministry of Science and Technology (113-2113-M-007-024-MY3 and 113-2113-M-007-004) in Taiwan.

Notes and references

1. R. Lawrence, M. Watters, C. R. Davies, K. Pantel and Y.-J. Lu, Nat. Rev. Clin. Oncol., 2023, 20, 487–500.
2. H. Kang, Y. Xiong, L. Ma, T. Yang and X. Xu, RSC Adv., 2022, 12, 34892–34903.
3. H. Pei, L. Li, Z. Han, Y. Wang and B. Tang, Lab Chip, 2020, 20, 3854–3875.
4. A. Ring, B. D. Nguyen-Strauli, A. Wicki and N. Aceto, Nat. Rev. Cancer, 2023, 23, 95–111.
5. C. Lou, H. Yang, Y. Hou, H. Huang, J. Qiu, C. Wang, Y. Sang, H. Liu and L. Han, Adv. Mater., 2024, 36, 2307051.
6. S. Nagrath, L. V. Sequist, S. Maheswaran, D. W. Bell, D. Irimia, L. Ulkus, M. R. Smith, E. L. Kwak, A. Muzkansky, P. Ryan, U. J. Balis, R. G. Tomkins, D. A. Haber and M. Toner, Nature, 2007, 450, 1235–1239.
7. V. Murlidhar, L. R. Baez and S. Nagrath, Small, 2016, 12, 4450–4463.
8. Y. Song, Z. Zhu, Y. An, W. Zhang, H. Zhang, D. Liu, C. Yu, W. Duan and C. J. Yang, Anal. Chem., 2013, 85, 4141–4149.
9. P. T. H. Went, A. Lugli, S. Meier, M. Bundi, M. Mirlacher, G. Sauter and S. Dirnhofer, Hum. Pathol., 2004, 35, 122–128.
10. Y. Liu, W. Zhao, J. Hodgson, M. Egan, C. N. C. Pope, G. Hicks, P. G. Nikolinakos and L. Mao, ACS Nano, 2024, 18, 8683–8693.
11. K. Zhao, Y. Liu, H. Wang, Y. Song, X. Chen, C. Huang, Q. Niu, J. Cao, X. Chen, W. Wang, L. Wu and C. Yang, Lab Chip, 2022, 22, 367–376.
12. M. Nian, B. Chen, M. He and B. Hu, Anal. Chem., 2024, 96, 766–774.
13. N. G. Welch, J. A. Scoble, B. W. Muir and P. J. Pigram, Biointerphases, 2017, 12, 02D301.
14. A. K. Trilling, J. Beekwilder and H. Zuilhof, Analyst, 2013, 138, 1619–1627.
15. M. E. Wiseman and C. W. Frank, Langmuir, 2012, 28, 1765–1774.
16. K. C. Andree, A. M. C. Barradas, A. T. Nguyen, A. Mentrink, I. Stojanovic, J. Baggerman, J. Dalum, C. J. M. Rijn and W. M. M. Terstappen, ACS Appl. Mater. Interfaces, 2016, 8, 14349–14356.
17. U. Dharmasin, S. K. Njoroge, M. A. Witek, M. G. Adebiy, J. W. Kamande, M. L. Hupert, F. Barany and S. A. Soper, Anal. Chem., 2011, 83, 2301–2309.
18. S. Goo, J. M. Guisan and J. Rocha-Martin, Anal. Chim. Acta, 2022, 1189, 1–24.
19. K. A. Hyun, T. Y. Lee, S. H. Lee and H. Il Jung, Biosens. Bioelectron., 2015, 67, 86–92.
20. S. T. Wang, K. Liu, J. Liu, Z. T. F. Yu, X. Xu, L. B. Zhao, T. Lee, E. K. Lee, J. Reiss, Y. K. Lee, J. Huang, M. Rettig, D. Seligson, K. N. Duaiswamy, C. K.-F. Shen and H.-R. Tseng, Angew. Chem., Int. Ed., 2011, 50, 3084–3088.
21. B. D. Plouffe, M. Mahalanabis, L. H. Lewis, C. M. Klapperich and S. K. Murthy, Anal. Chem., 2012, 84, 1336–1344.
22. J. H. Kang, S. Krause, H. Tobin, A. Mammoto, M. Kanapathipillai and D. E. A. Ingber, Lab Chip, 2012, 12, 2175–2181.
23. J. Shi, C. Zhao, M. Shen, Z. Chen, J. Liu, S. Zhang and Z. Zhang, Biosens. Bioelectron., 2022, 202, 114025–114033.
24. C.-H. Lai, S. C. Lim, L.-C. Wu, C.-F. Wang, W.-S. Tsai, H.-C. Wu and Y.-C. Chang, Chem. Commun., 2017, 53, 4152–4155.
25. X. Sun, B. M. Chapin, P. Metola, B. Collins, B. Wang, T. D. James and E. V. Anslyn, Nat. Chem., 2019, 11, 768–778.
26. M. Zheng, L. Kong and J. Gao, Chem. Soc. Rev., 2024, 53, 11888–11907.
27. A. K. Adak, K.-T. Huang, C.-Y. Liao, Y.-J. Lee, W.-H. Kuo, Y.-R. Huo, P.-J. Li, Y.-J. Chen, B.-S. Chen, Y.-J. Chen, K. C. Hwang, W.-S. Wayne and C.-C. Lin, Chem. – Eur. J., 2022, 28, e202104178.
28. A. Akgun and D. G. Hall, Angew. Chem., Int. Ed., 2018, 57, 13029–13044.
29. P.-C. Lin, S.-H. Chen, K.-Y. Wang, M.-L. Chen, A. K. Adak, J. R. Hwu, Y.-J. Chen and C.-C. Lin, Anal. Chem., 2009, 81, 8774–8782.
30. C.-Y. Fan, Y.-R. Hou, A. K. Adak, J. T. Waniwan, M. dela Rosa, P. Y. Low, T. Angata, K.-C. Hwang, Y.-J. Chen and C.-C. Lin, Chem. Sci., 2019, 10, 8600–8609.
31. A. K. Adak, K.-T. Huang, P.-J. Li, C.-Y. Fan, P.-C. Lin, K.-C. Hwang and C.-C. Lin, ACS Appl. Bio Mater., 2020, 3, 6756–6767.
32. J. N. Arnold, M. R. Wormald, R. B. Sim, P. M. Rudd and R. A. Dwek, Annu. Rev. Immunol., 2007, 25, 21–50.
33. M. Amerian, M. Amerian, M. Sameti and E. Seyedjafari, J. Biomed. Mater. Res., Part A, 2019, 107, 2806–2813.
34. I. J. Bruce and T. Sen, Langmuir, 2005, 21, 7029–7035.
35. T. Dey and D. Naughton, J. Sol-Gel Sci. Technol., 2016, 77, 1–27.
36. S. Prakash, T. M. Long, J. C. Selby, J. S. Moore and M. A. Shannon, Anal. Chem., 2007, 79, 1661–1667.
37. Z. Zhang, X. Feng, F. Xu, X. Liu and B.-F. Liu, Electrophoresis, 2010, 31, 3129–3136.
38. T. R. Chan, R. Hilgraf, K. B. Sharples and V. V. Fokin, Org. Lett., 2004, 6, 2853–2855.
39. J. C. McDonald and G. M. Whitesides, Acc. Chem. Res., 2002, 35, 491–499.
40. M.-L. Chen, A. K. Adak, N.-C. Yeh, W.-B. Yang, Y.-J. Chuang, C.-H. Wong, K.-C. Hwang, J. R. Hwu, S.-L. Hsieh and C.-C. Lin, Angew. Chem., Int. Ed., 2008, 47, 8627–8630.
41. W.-H. Kuo, S. K. Kawade, A. K. Adak, D. S. Juang, C.-J. Liu, Y.-K. Wang, W.-H. Hsu, C.-L. Kao, D.-C. Wu and C.-C. Lin, Biosens. Bioelectron., 2025, 287, 117710.

---

Footnote

† The authors contributed equally to this work.

---