DNA Hanger: Surface‐Minimized Single‐Molecule Immunoassay Platform

Abstract

A novel single‐molecule immunoassay platform, termed DNA Hanger, is developed to address the limitations of conventional surface‐based assays. By suspending biotinylated λ‐phage DNA across microfabricated quartz barriers, this method enables high‐specificity protein detection with minimal nonspecific binding. DNA Hanger significantly reduces background signals, achieving nonspecific binding rates as low as one protein per 236 µm of DNA. Quantification of mNeonGreen‐tagged human poly(A)‐binding protein C1 (mNG‐PABP) and single‐molecule fluorescence‐linked immunosorbent assay (FLISA) of human tumor necrosis factor α (TNF‐α) demonstrates the assay's specificity and sensitivity at the single‐molecule level, with a detection limit of 0.90 pM in buffer, 38‐fold lower than that of conventional FLISA, and 20.6 pM in 70% fetal bovine serum, an 8‐fold improvement. DNA Hanger also enables the detection and quantification of endogenous TNF‐α in human serum, highlighting its clinical potential. The DNA Hanger assay eliminates the need for surface blocking and simplifies workflow, resulting in completing the immunoassay process within 1 hour. DNA Hanger offers broad applicability for biomolecular interaction studies and clinical diagnostics.

The DNA Hanger platform suspends long DNA molecules biotinylated along their length away from the surface, effectively minimizing nonspecific protein binding. The assay achieves high sensitivity and specificity in single‐molecule FLISA down to subpicomolar levels. The platform also enables accurate analysis of protein‐protein interactions, such as single‐molecule pull‐down assays, with minimal background interference.

1. Introduction

Single‐molecule immunoassays represent an advanced evolution of traditional immunoassay techniques, which have long been indispensable tools for detecting and analyzing biomolecules.^{[ 1 ]} Unlike conventional assays that measure ensemble‐averaged response,^{[ 2 ]} this cutting‐edge approach offers direct observation of the dynamics and interactions of individual molecules within complex molecular environments.^{[ 3} , ^{4 ]} This resolution enables a comprehensive understanding of their functional roles and regulatory mechanisms, made possible by high‐sensitivity fluorescence microscopy and single‐molecule detection technologies.^{[ 5 ]}

In typical single‐molecule immunoassays, biomolecules are immobilized on a surface, such as a specialized glass slide or membrane. While the assay is designed to detect specific molecular interactions, target molecules, and detection antibodies often exhibit nonspecific binding to the surface. These nonspecific interactions can significantly distort the readout, leading to reduce the accuracy and reliability of the assay. Various surface passivation strategies, including surface blocking with agents like polyethylene glycol (PEG), lipid bilayer coating, and surface charge or hydrophobicity modifications, have been employed to suppress such nonspecific binding.^{[ 6} , ⁷ , ⁸ , ⁹ , ¹⁰ , ^{11 ]} However, residual surface binding remains a persistent issue. As a result, minimizing nonspecific binding continues to be a critical challenge for improving single‐molecule assay performance.^{[ 12 ]}

To address this limitation, we recently developed a single‐molecule fluorescence imaging platform to minimize interference from nonspecifically surface‐bound molecules called DNA Skybridge.^{[ 13 ]} In this platform, long DNA molecules are stretched and aligned across adjacent ridges of microfabricated quartz barriers under laminar flow, resulting in being suspended away from the surface. DNA Skybridge enables to observe individual fluorescently labeled proteins moving on suspended DNA using a thin light sheet produced by interference from excitation beams refracted by the barriers. This setup enables real‐time tracking of individual proteins on DNA without interference with the surface.

In this study, we introduce a novel single‐molecule immunoassay termed DNA Hanger, building on this system. In this platform, λ‐phage DNA molecule (48.5 kilo‐base pairs) is randomly biotinylated to its bases to provide binding sites for target proteins and/or antibodies. The modified DNA molecules are suspended between two adjacent quartz barriers, creating surface‐minimized environment. This configuration substantially minimizes nonspecific binding and allows quantitative measurement of target proteins of interest simply by counting fluorophore‐labeled antibodies or directly fluorophore‐labeled proteins with high specificity and sensitivity. We demonstrate the DNA Hanger's capability by monitoring mNeonGreen‐tagged human poly(A)‐binding protein C1 (mNG‐PABP) and its antibody, which suggests potential extension to pull‐down‐like applications.

Finally, we validate DNA Hanger as a robust platform for a highly sensitive and specific single‐molecule immunoassay using mNG‐PABP and human tumor necrosis factor α (TNF‐α) as model targets. We evaluate the assay's performance compared to conventional flat PEG‐biotin surfaces. Furthermore, we assess its functionality in complex biological envionments, including fetal bovine serum and human serum, to explore its potential for clinical diagnostics.

2. Results and Discussion

2.1. Design and Validation of DNA Hanger

In our previous study,^{[ 13 ]} we demonstrated that the unique 3D structure of the quartz slide could generate a light sheet with a full width half maximum of ≈1.1 µm from the incident beam, depending on the incident angle (Figure 1a). This light sheet formed an image plane on top of the barriers when the excitation beam was an incident angle of θ–75°. This configuration dramatically reduces background signals from analytes or ligands that nonspecifically bind to the bottom surface, as emitters on the bottom surface appear to be out of focus. Consequently, the signal‐to‐noise ratio of the probe on suspended DNA molecules is sufficiently high to allow for the visualization of individual fluorophores.

Figure 1.

DNA Hanger. a) The 3D‐etched quartz surface features narrow ridges (≈1 µm wide) with a height of 4 µm, which enables the formation of a thin excitation light sheet near the ridges under prism‐based TIRF illumination.^{[ 13 ]} A 3′digoxigenin oligonucleotides were ligated to 12 nt cohesive ends of λ‐phage DNA. The λ‐phage DNA molecules are suspended between two adjacent ridges via specific binding to an anti‐digoxigenin antibody, effectively forming a surface‐suspended bridge structure. b) Biotin groups were introduced onto nucleic acid bases using a photo‐coupling reagent (photobiotin), which covalently attaches biotin upon 405 nm irradiation. In an alternative approach, biotin was site‐randomly incorporated into double‐stranded DNA using a pSoralen–biotin conjugate, a photoactivatable reagent that intercalates into DNA and covalently crosslinks to thymine bases upon UV exposure (365 nm). c) Representative images showing Cy5‐streptavidin binding to biotin‐coated λ‐phage DNA (upper left). The bottom right panels display 5′ biotin‐ssDNA‐Cy5 binding to streptavidin ‐coated λ‐phage at various concentrations of the ssDNA‐Cy5. d) A plot showing fluorescence intensity of Cy5 as a function of 5′biotin‐ssDNA‐Cy5 concentrations using DNA Hanger. Each concentration was repeated in two independent experiments (N = 2), with the number of field of view (FOVs) as follows: n = 4, 7, 8, 13, 10, 7 and 9 for 6.4 pm, 3.2 × 10 pm, 1.6 × 10² pm, 8.0 × 10² pm, 4.0 × 10³ pm, 2.0 × 10⁴ pm, and 1.0 × 10⁵ pm, respectively.

To extend the functionality of the system to single‐molecule immunoassay, we functionalized λ‐phage DNA with biotin along its bases using the photocoupling reaction of pSoralen‐biotin or photobiotin (Experimental Section, Figure 1b, Table S1, Supporting Information). Additionally, digoxigenin‐labeled oligonucleotides were ligated to both cos ends of the λ‐phage DNA, which is tethered on the quartz barrier ridges via specific interactions with anti‐digoxigenin antibody (Figure 1b).

To evaluate the extent of biotin incorporation along the λ‐phage DNA, 50 µg ml⁻¹ of Cy5‐streptavidin was injected into the flow chamber. The DNA imaging substrate was saturated with Cy5‐labeled streptavidin at this concentration (Figure 1c top), indicating full coverage of available biotin sites. This saturation condition can be used to control ligand density for subsequent binding studies. To further assess the binding capacity of the DNA Hanger assay, we performed binding experiments using 5′biotinylated‐single stranded DNA (100 nt) labeled with Cy5 at the 3′ ends (5′biotin‐ssDNA‐3′Cy5, Table S1, Supporting Information). The modified ssDNA in DNA binding buffer (Experimental Section) was injected into the chamber across a concentration range of 5 pm–20 nm. As the concentration increased, the DNA Hanger's binding capacity approached saturation at ≈4 nM, with EC90 calculated to be 26.28 nm (Figure 1c). This result confirms that 50 µg mL⁻¹ (917.4 µm) of streptavidin is sufficient to fully coat the λ‐phage DNA and allow maximal binding of biotinylated substrates. Taken together, these results validate DNA Hanger as a reliable and quantitative platform for molecular tethering, supporting its application in a wide range of single‐molecule immunoassays.

2.2. Single‐Molecule Capture‐Based Immunoassay Using DNA Hanger

Single‐molecule immunoassays represent a significant advancement for studying molecular interactions at the single‐molecule level. However, their specificity and sensitivity are often compromised by substantial nonspecific antibody binding, which hinders precise quantitative analysis. Even with stringent surface passivation strategies, conventional PEG‐biotin–coated quartz slides frequently produce inconsistent or heterogeneous results across experiments, which underscores the limitations of traditional approaches in effectively minimizing nonspecific surface interactions (Figure S1, Supporting Information).^{[ 14 ]}

To test the applicability of DNA Hanger for a single‐molecule immunoassay, we first examined the binding of human PABP to poly(A) RNA. A Cy5‐labeled DNA/RNA duplex containing a 114‐nucleotide poly(A) tail was injected into the DNA Hanger chamber, resulting in tethered Cy5‐DNA/RNA duplexes. Subsequently, 100 nm mNG‐PABP was introduced, followed by washing to remove unbound mNG‐PABP molecules (Figure 2a). Alexa Fluor 594‐conjugated anti‐PABP (AF594‐anti‐PABP) antibody was then added to detect mNG‐PABP bound to the poly(A)₁₁₄ of the Cy5‐DNA/RNA duplex, and the chamber was washed again to remove free antibodies in solution (Figure 2a). We repeated the same experiment using a conventional flat PEG‐biotin coated quartz slide to compare it with DNA Hanger.

Figure 2.

Capture of PABP bound to RNA. a) Schematic (top) and corresponding representative fluorescence images (bottom) showing the stepwise assembly of the assay: Cy5‐DNA/RNA duplex containing a poly(A)₁₁₄ RNA tail (red) tethered to the λ‐phage DNA bridge via a streptavidin–biotin interaction (left), binding of mNG‐PABP (green) to the duplex (middle), and subsequent binding of AF594‐anti‐PABP antibody (pink) to mNG‐PABP (right). Magnified views of the white boxed regions are shown on the far right. White arrows indicate colocalized spots exhibiting all three fluorescence signals. b) Non‐colocalization ratios between mNG‐PABP and AF594‐anti‐PABP antibody in the flat PEG–biotin‐coated quartz surface (n = 41, N = 3 for both mNG‐PABP and AF594‐anti‐PABP antibody) and the DNA Hanger platform (n = 18, N = 3 for mNG‐PABP, n = 17, N = 3 for AF594‐anti‐PABP antibody). Yellow and green diamonds represent the normalized non‐colocalization ratios for mNG‐PABP and AF594‐anti‐PABP antibodies, respectively (see the Experimental Section for the mathematical definition). Error bars represent standard deviations (S.D.).

Using colocalization analysis,^{[ 15 ]} we determined the non‐colocalization ratios of mNG‐PABP (0.18 ± 0.04, mean ± S.D.) and AF594‐anti‐PABP antibody (0.14 ± 0.04) on DNA Hanger (Figure 2b). The normalized non‐colocalization was defined as the fraction of mNG‐PABP signals not colocalized with Cy5‐DNA/RNA duplexes, and similarly, AF594‐anti‐PABP antibody signals not colocalized with mNG‐PABP (see Data analysis in Methods). In contrast, the PEG‐biotin quartz surface exhibited significantly higher non‐colocalization ratios (0.88 ± 0.05 for mNG‐PABP, 0.94 ± 0.03 for AF594‐anti‐PABP antibody) (Figure 2b). However, we confirmed that specific interactions were not impaired on the PEG‐biotin surface: 79% of Cy5‐DNA/RNA and 71% of mNG‐PABP bound to Cy5‐DNA/RNA were colocalized with mNG‐PABP and AF594‐PABP antibody, respectively. The high non‐colocalization ratios of PABP or its antibody on the PEG‐biotin surface are primarily due to nonspecific adsorption of PABP and its antibody, which dominates the contribution from specific binding to poly(A) tail and PABP. This substantial level of nonspecific binding can distort quantitative analysis on the conventional flat PEG‐biotin quartz surface.

To investigate the origin of the modest non‐colocalizations measured for mNG‐PABP and AF594‐anti‐PABP antibody on DNA Hanger, we performed additional control experiments. In the control experiment assessing nonspecific binding of mNG‐PABP in the presence of Cy5‐DNA, but without poly(A) RNA, we observed a low binding level of 0.07 ± 0.03 (Figure S2, Supporting Information). Similarly, in the absence of mNG‐PABP, in the control condition containing only the Cy5‐DNA/RNA duplex (without mNG‐PABP), the nonspecific binding of the AF594‐anti‐PABP antibody was measured to be 0.09 ± 0.04 (Figure S2, Supporting Information). These results suggest that the 18% (mNG‐PABP) and 14% (AF594‐anti‐PABP antibody) non‐colocalization ratios in Figure 2b cannot be solely attributed to nonspecific binding.

We infer that non‐fluorescent (photobleached or dysfunctional) Cy5 or non‐fluorescent (photobleached or mutated) mNG likely contributed to the observed non‐colocalization. This is further supported by instances in which the fluorescence signals of AF594‐anti‐PABP antibody and Cy5‐DNA/RNA duplex are colocalized, but the mNG‐PABP signal is absent, which suggests the presence of non‐fluorescent mNG‐PABP at the site. If such colocalization is assumed to represent successful binding of AF594‐anti‐PABP antibody to non‐fluorescent mNG‐PABP, the recalculated non‐colocalization ratio of AF594‐anti‐PABP antibody drops to 0.05 ± 0.04 (Figure S3, Supporting Information). In conclusion, the 18% and 14% non‐colocalization ratios do not substantially affect the sensitivity of our assay.

2.3. Single‐Molecule FLISA

The Fluorescence‐Linked Immunosorbent Assay (FLISA) is a widely used technique for detecting and quantifying specific protein biomarkers, offering both specificity and sensitivity using antibody‐based recognition.^{[ 16 ]} We applied the DNA Hanger platform to perform a single‐molecule FLISA targeting the well‐characterized cytokine, human tumor necrosis factor α (TNF‐α). For this assay, fluorophore‐labeled TNF‐α antibodies were prepared for both capture (Bio‐AF647‐MAb1, cAb) and detection (AF568‐MAb11, dAb) (Experimental Section/Methods). The cAb was introduced into the DNA Hanger chamber and incubated for 5 min to fully cover streptavidin‐coated λ‐phage DNA through biotin‐streptavidin interaction. After washing excess cAb, TNF‐α was infused at concentrations ranging from 0 to 1 nm and maintained for 5 min. Following removal of unbound TNF‐α, the dAb was added and incubated for 10 min to complete the sandwich immunoassay (Figure 3a).

Figure 3.

Single‐molecule FLISA of TNF‐α in buffer. a) Schematic of TNF‐α detection using DNA Hanger. AF647‐cAb is immobilized on streptavidin‐coated λ‐phage DNA through the biotin‐streptavidin interaction. TNF‐α is captured by the cAb and subsequently recognized by dAb. b) Quantification of nonspecific binding of dAb (10784 λ‐phage DNA molecules, n = 30, and N = 3) and TNF‐α (6894 DNA, n = 30, and N = 3) to DNA Hanger or to cAb (9954 DNA, n = 30, and N = 3). Each diamond indicates the average of the number of dAb per DNA in a field of view. Statistical analysis was performed using one‐way ANOVA followed by Tukey post hoc test ( ^**p < 0.01). “ns” denotes no statistically significant difference (p ≥ 0.05). c) Representative fluorescence images showing cAb on λ‐phage DNA and dAb bound to TNF‐α at a concentration range (0 – 1000 pm). d) Quantification of dAb binding per λ‐phage DNA at varying TNF‐α concentrations (mean ± S.D.): 0.07 ± 0.04 at 0 pm (9951 DNA, n = 30, and N = 3), 0.31 ± 0.11 at 3 pm (3465 DNA, n = 17, and N = 3), 1.31 ± 0.47 at 10 pm (3714 DNA, n = 21, and N = 4), 3.41 ± 0.45 at 30 pm (1769 DNA, n = 18, and N = 3), 9.92 ± 1.04 at 100 pm (473 DNA, n = 9, and N = 3), 16.3 ± 1.93 at 300 pm (518 DNA, n = 6, and N = 2), and 17.5 ± 1.60 at 1000 pm (736 DNA, n = 6, and N = 3).

We evaluated the extent of nonspecific bindings of dAb by quantifying the number of dAb bound to the modified λ‐phage DNA (Figure 3b). In the absence of both cAb and TNF‐α, dAb binding was negligible (0.031 ± 0.016). When TNF‐α was added without cAb, the nonspecific binding remained nearly unchanged (0.038 ± 0.020), which indicates that TNF‐α did not bind nonspecifically to streptavidin‐coated λ‐phage DNA. When cAb was introduced in the absence of TNF‐α, dAb binding increased slightly (0.070 ± 0.043), suggesting minor interaction between dAb and cAb. However, this level remained very low compared to the binding observed in the presence of TNF‐α (Figure 3c,d).

Based on the barrier spacing of 13 µm and a nonspecific binding rate of 0.055 dAb molecules per λ‐phage DNA in the presence of cAb, we estimate that, on average, one dAb molecule is nonspecifically bound per 236 µm of DNA. Moreover, nonspecific binding of dAb increased slightly with longer incubation, but remained low even after 4 h (0.063 ± 0.025, Figure S4, Supporting Information). These results confirm that nonspecific binding to DNA Hanger is minimal and unlikely to interfere with quantitative analysis.

Next, we explored the specificity of dAb binding to TNF‐α in the presence of cAb at various concentrations (Figure 3c,d). In DNA binding buffer, the average number of dAb per λ‐phage DNA was 0.31 ± 0.11 at 3 pM TNF‐α, which increased ≈10‐fold at 30 pM TNF‐α (3.41 ± 0.45, Figure 3d). At concentrations above 30 pm, individual dAb could no longer be reliably counted due to dense binding (Figure 3c). Alternatively, the total fluorescence intensity of dAb bound to each λ‐phage DNA was divided by the intensity of a single dAb to estimate the number of dAb per DNA. Using this method, we obtained values of 9.92 ± 1.04 at 100 pm, 16.3 ± 1.93 at 300 pm, and 17.50 ± 1.60 at 1 nm TNF‐α. The data were fitted to a Langmuir isotherm model, commonly used in a FLISA analysis, yielding maximum specific binding (B _max, 20.36 ± 1.31) and an equilibrium dissociation constant (K_d, 107.56 ± 23.54) (Figure 3d). To evaluate an analytical sensitivity of DNA Hanger, the limit of detection (LOD) was calculated based on the Langmuir fit and these parameters, resulting in an LOD of 0.90 ± 1.13 pm, which is ≈38‐fold lower than that of conventional FLISA (34.24 ± 14.11 pm) (Figure S5a, Supporting Information).

2.4. Single‐Molecule FLISA in Complex Biological Fluids

We conducted the same experiments in 70% Fetal Bovine Serum (FBS) to simulate physiologically relevant conditions while avoiding interference from endogenous human TNF‐α. TNF‐ α was tested at concentrations ranging from 0 to 30 nm in 10‐fold increments. As expected, the high protein content in FBS partially hindered the specific binding of TNF‐α (Figure 4a). However, the nonspecific binding of dAb was comparable to that observed in DNA binding buffer (0.062 ± 0.12), which indicates that nonspecific interactions in DNA Hanger are minimal even in complex biological environments. The average number of dAb per λ‐phage DNA increased to 0.39 ± 0.09 at 10 pM TNF‐α and reached 23.50 ± 2.20 at 30 nM TNF‐α. Langmuir isotherm fitting yielded a slightly higher B _max (28.00 ± 0.55) and an increased K _d (5856.90 ± 410.20 pM). The resulting LOD is 20.57 ± 2141.00 pM, still 7.7‐fold lower than that of conventional FLISA (157.90 ± 12.73 pM) (Figure S5b, Supporting Information).

Figure 4.

Single‐molecule FLISA of TNF‐α in serum. a) Detection of TNF‐α in 70% FBS. The average number of dAb per DNA was quantified at various TNF‐α concentrations: 0.06 ± 0.01 (mean ± S.D.) at 0 pm (1674 DNA, n = 8, and N = 3), 0.39 ± 0.11 at 10 pm (1435 DNA, n = 6, and N = 2), 0.87 ± 0.11 at 100 pm (1711 DNA, n = 7, and N = 3), 4.12 ± 0.70 at 1000 pm (591 DNA, n = 8, and N = 3), 17.6 ± 2.29 at 10000 pm (1425 DNA, n = 9, and N = 2), 23.5 ± 2.20 at 30000 pm (1229 DNA, n = 8, and N = 3). (b) Detection of TNF‐α in 70% human serum. The average number of dAb per DNA was measured at the following concentrations: 0.57 ± 0.13 at 0 pm (2144 DNA, n = 12, and N = 2), 1.17 ± 0.17 at 20 pm (2557 DNA, n = 11, and N = 2), 2.50 ± 0.25 at 100 pm (2511 DNA, n = 11, and N = 2). P‐values were calculated using one‐way ANOVA followed by Tukey's post hoc test ( ^**p < 0.01).

Finally, we evaluated the specificity of TNF‐α in 70% human serum, a matrix that closely resembles clinical blood samples (Figure 4b). At 0 pM TNF‐α, we observed a non‐negligible number of dAb per λ‐phage DNA (0.57 ± 0.13), indicating that endogenous TNF‐α in human serum is captured by the cAb. Upon spiking TNF‐α into the serum, the average number of dAb per per λ‐phage DNA increased to 1.17 ± 0.17 at 20 pM and 2.50 ± 0.25 at 100 pM. These findings demonstrate that the DNA Hanger FLISA assay can detect and quantify endogenous target proteins at concentrations as low as 20 pM levels in the blood sample.

3. Conclusion

In this study, we developed and demonstrated the DNA Hanger assay, a novel single‐molecule immunoassay technique. This platform offers a significant advancement in both specificity and sensitivity for molecular detection and quantification, particularly in the context of single‐molecule FLISA. The most notable improvement of the DNA Hanger assay is its ability to reduce nonspecific binding. By suspending DNA molecules away from the surface on thin quartz barriers, we achieved nonspecific binding rates as low as one protein per 236 µm of DNA, even under prolonged incubation conditions (Figure S4, Supporting Information). This improvement is essential in overcoming the challenges posed by nonspecific binding in traditional immunoassay assays, which can lead to erroneous interpretations of molecular interactions.

DNA Hanger achieved an LOD of 0.9 pm for TNF‐α in buffer (Figure 3d), which is ≈30‐fold lower than that obtained using conventional FLISA (34.2 pM, Figure S5a, Supporting Information). This performance is comparable to that of state‐of‐the‐art commercial FLISA and ELISA assays, despite operating at the single‐molecule level and requiring significantly shorter processing times (Figure S6, Supporting Information). For instance, TR‐FLISA (Poly‐Dtech), a nanoparticle‐based fluorescence kit, reports an LOD for mouse TNF‐α as low as 16.2 pg mL⁻¹ (0.9 pm).^{[ 17 ]} Commercial ELISA kits report LODs ranging from 0.5 pg mL⁻¹ (28.7 fM) to 23 pg mL⁻¹ (1.3 pm).^{[ 18 ]}

In 70% Fetal Bovine Serum (FBS), DNA Hanger obtained an LOD of 20.6 pm, which is ≈20‐fold higher than in buffer. This increase suggests that serum components partially interfere with TNF‐α detection, possibly by affecting its interaction with the dAb. This result aligns with previous findings by Hariri et al.,^{[ 14 ]} who reported an LOD of 6.6 pm in buffer and 19.4 pm in 70% chicken serum using a single‐color measurement. Remarkably, their colocalization method further enhanced sensitivity, achieving LODs of 2.1 pm in buffer and 4.9 pm in serum. While our assay shows a 2‐fold lower LOD in buffer, our LOD in serum is 4‐fold higher. Their protocol included overnight incubation of TNF‐α and dAb in the imaging chamber, whereas our assay did not involve extended incubation. We found that incorporating a long incubation step in our assay increases dAb signals, with only a slight increase in nonspecific binding (Figure S4, Supporting Information). This suggests that DNA Hanger's LOD could be further enhanced by extending incubation time.

In conclusion, the DNA Hanger assay represents a versatile and powerful tool for single‐molecule studies such as FLISA and pull‐down assays. Its ability to minimize nonspecific binding and provide high sensitivity makes it an attractive platform for a broad range of applications, from basic research in molecular biology to potential clinical diagnostics. Notably, the DNA Hanger assay eliminates the need for conventional blocking steps, thereby simplifying the experimental workflow and significantly reducing the overall assay time (Figure S6, Supporting Information). Moving forward, we anticipate that this method will open up new possibilities for exploring the detailed mechanisms of biomolecular interactions and will contribute to the development of more accurate and reliable assays for biomolecule detection.

4. Experimental Section

Patterned Quartz Slide

The fabrication of 3D‐etched quartz slides was performed with slight modifications to a previously established protocol.^{[ 13 ]} Briefly, quartz wafers were thoroughly cleaned using a piranha solution (a 3:1 mixture of H₂SO₄ to H₂O₂) at 130 °C for 10 min. Following this, an adhesion promoter (hexamethyldisilane, J.T.Baker) and a photoresist (TDMR‐AR87, Tokyo Ohka Kogyo) were applied to the surface through the spin coating at 3,000 and 4,500 rpm, respectively, then baked at 88 °C for 60 s. A stripe pattern was created on the quartz surface using soft contact mode with a photomask aligned by an MA‐6 mask aligner and rinsing in a developer (AZ 300 MIF, Merck) for 60 s. The patterned surface underwent a hard bake at 113 °C for 90 s to develop the design. The final step involved etching the quartz surface by immersing it in a 6:1 diluted HF solution (Buffered oxide etch, Samchun) at room temperature without stirring. This fabrication process produced quartz slides with the following dimensions: barrier height of 4 µm and barrier intervals of 13.1 µm.

Quartz Slide Surface Cleaning and Passivation

The quartz slides were subject to a rigorous washing to ensure complete removal of impurities. The cleaning process began with incubation in a staining jar containing 10% Alconox for 20 min, followed by sequential washes in ddH₂O for 5 min, acetone for 15 min, KOH for 20 min, piranha solution for 20 min, and methanol for 10 min. Sonication was employed during these procedures to enhance impurity removal. After the washing steps, the slides were briefly exposed to a flame using a torch for 30 s to ensure thorough surface cleaning.

Anti‐Digoxygenin Antibody Stamping with PDMS

A mixture of PDMS base and curing agent (7:1, SYLGARD 184 SILICONE ELASTOMER KIT, Dow Corning) was poured onto a flat plate. The mixture was degassed in a vacuum desiccator (≈0.2 atm) until all air bubbles were eliminated. The plate was then baked in an oven at 80 °C for 1 h to cure the PDMS. After the curing process, the PDMS was cut into pieces matching the dimensions of the pattern on the quartz slide, with a size of 11 mm × 4 mm. To prepare the stamping side on the PDMS pieces, 10 µL of anti‐digoxigenin solution (200 µg mL⁻¹, diluted in PBS buffer, 11333089001, Roche) was placed and spread on the PDMS surface inside a moisture box. After a 30‐min incubation, the solution on the PDMS surface was gently removed by nitrogen gas. The prepared PDMS stamp was then carefully applied onto the patterned surface of the quartz slide. Finally, the quartz slide, and a cover glass were assembled into a flow chamber using double‐sided tape.

Biotinylated λ‐Phage DNA

The DNA substrate was prepared using bacteriophage λ‐phage DNA (48.5 kbp, New England Biolabs) through ligation with 14 nt‐length 3′ digoxygenin‐labeled oligonucleotides, using T4 ligase (NEB) at room temperature overnight (Table S1, Supporting Information). After ligation, excessive oligonucleotides were removed by dialysis (Spectra‐Por Float‐A‐Lyzer, G2) or ultrafiltration (10 kDa Amicon, Millipore). Subsequently, biotin was conjugated to the λ‐phage DNA using either photobiotin (PHOTOPROBE, Vector) or pSoralen‐biotin (HOOK‐pSoralen‐PEO‐Biotin, G‐biosciences). The biotin reagent solution was added to the purified λ‐phage DNA solution in a 50:1 volume ratio for photobiotin and a 100:1 for pSoralen‐biotin. The solutions were gently mixed by inverting the tubes, and the tubes were placed 2 cm below a 365 nm hand‐held mercury lamp (8 W, Sanyo) for 1 h to facilitate the photo‐coupling reaction. In this imaging condition, purification of the DNA after the photo‐coupling reaction was not a mandatory step. Additionally, long‐range PCR was explored using LA Taq DNA polymerase (Takara) in combination with biotinylated dUTP (Biotin‐11‐dUTP Solution, Thermo Scientific) as an alternative method for fabricating the DNA substrate. However, this method resulted in low yield due to the lengthy DNA template used in the reaction.

mNeonGreen‐PABP

The pQE31 vector expression system was used to construct plasmids for the recombinant protein expression of full‐length human PABP.^{[ 19 ]} The mNeonGreen sequence was amplified by PCR from a plasmid (Addgene, #98877), and the PCR product was inserted at the N‐terminus of the pQE31 expression vector. Additionally, a 3xFLAG tag was subcloned into the C‐terminus of the construct. The PABP gene, obtained through PCR amplication, was then inserted between mNeonGreen and the 3xFLAG sequence in the plasmid. Recombinant proteins were overexpressed in NEBExpress Iq Competent E. coli (C3037I, NEB) cultured in LB medium. Protein expression was induced by adding 0.4 mm IPTG, and the culture was incubated for 4 h at 37 °C. After cell harvesting, the recombinant proteins were purified following the manufacturer's instructions for the NI‐NTA Fast Start Kit (30600, QIAGEN) and Anti‐DYKDDDDK Affinity Resin (101274‐MM13‐RN, Sino Biological). All steps of the protein expression and purification procedures were carried out on ice to preserve protein stability. The concentration of purified recombinant protein was measured using a nanophotometer (IMPLEN). Proteins were aliquoted and stored at −80 °C.

RNA/DNA Partial Duplex with Poly(A) Tail

To construct the partial duplex, single‐stranded polyadenine RNA (114 nt) was synthesized using an in vitro transcription assay. Initially, the pBluescript SK(‐) vector was digested at two restriction sites and ligated with oligonucleotides to form the polyadenine DNA construct template (Table S1, Supporting Information). After ligating the oligonucleotide to the vector, the remaining gap on the upper strand was filled with T4 DNA polymerase (M0203S, NEB) according to the manufacturer's protocol. The resulting template DNA was iterated twice to reach a poly(A) length of 114 nt according to the methods previously described.^{[ 15 ]} The final RNA template was synthesized using the HiScribe T7 Quick High Yield RNA Synthesis Kit (E2050S, NEB), following the manufacturer's instructions. The construct used was 5′‐ctcgaggtcgacggtatcgataagcttgaagac(a)₂₅‐3′ and the complementary DNA oligonucleotide was modified with Cy5 at the 5′ end and biotinylation at the 3′ end (IDT, Coralville, USA). The partial duplex was prepared by annealing the DNA oligo with the RNA transcript at a 1:9 molar ratio in an annealing buffer (1 mm MgCl₂, 10 mm Tris‐HCl, 100 mm NaCl, pH 7.4) to a final concentration of 100 nm. The mixture was incubated at 85 °C for 10 min and then allowed to cool gradually to room temperature. The resulting partial duplex (Cy5‐DNA/RNA duplex) was stored at 4 °C for further use.

Fluorophore‐Labeled Antibodies and TNF‐α

For the preparation of fluorophore‐conjugated antibodies, the protocol described in a previous study was followed.^{[ 14 ]} Briefly, purified human TNF‐α antibodies (MAb1 and MAb11, BioLegend) were modified for either capture or detection of TNF‐α. MAb1, the capture antibody (cAb), was site‐specifically biotinylated using the SiteClick Antibody Labeling Kit (Thermo Fisher Scientific) according to the manufacturer's instructions. MAb11, the detection antibody (dAb), was buffer exchanged with PBS, and both MAb1 and MAb11 were separately mixed with 5 µL of Alexa Fluor 647 NHS Ester or Alexa Fluor 568 NHS Ester (Thermo Fisher Scientific) in DMSO (10 mg mL⁻¹) for 1 h at room temperature, respectively. The modified antibodies were subsequently purified using PD MiniTrap G‐25 (GE Healthcare) and buffer‐exchanged with PBS containing 0.09% sodium azide using an Amicon Ultra‐10k Centrifugal Filter Unit. The concentration and degree of labeling of the antibodies were measured using a Nanophotometer. All antibodies were stored at 4 °C. TNF‐α (5 µg, R&D Systems) was reconstituted in PBS to 5 µm and stored at −80 °C.

Sandwich Fluorescence‐Linked Immunosorbent Assay of Human TNF‐α

The capture antibody (Mab1, BioLegend) was diluted in 1 × PBS (137 mm sodium chloride, 2.7 mm potassium chloride, 10 mm sodium phosphate, 1.76 mm potassium phosphate) to a final concentration of 2 µg µL⁻¹. 50 µL of the capture antibody solution was added to each well of a 96‐well round‐bottom immunoplate (#32696, SPL). The plate was incubated at 4 °C for 12 h. After incubation, the capture Ab solution was removed by flicking on the sink, the plate was washed three times with 300 µL of 1 × PBST (1 × PBS buffer supplemented with 0.05% Tween‐20). To block nonspecific binding of TNF‐α, 300 µL of 1 × PBST containing 1% BSA was added to each well and incubated at RT for 2 h. After incubation, the plate was washed three times with 300 µL of 1 × PBST.

After blocking, the plate was washed three times with 300 µL of 1 × PBST, 50 µL of TNF‐α containing sample added to each well. The sample was prepared by dilution of TNF‐α (Recombinant human TNF‐α, 210‐TA, R&D Systems) in 1 × PBST, 70% fetal bovine serum, or 70% human serum. The plate was incubated at 4 °C for 16 h. After incubation, the plate was washed three times with 300 µL of 1 × PBST. For detection, Alexa Fluor 568‐conjugated TNF‐α detection antibody was diluted in 1 × PBST (Section: Fluorophore‐labeled antibodies and TNF‐α, 1 µg/ml). then 50 µL of detection antibody solution was added to each well. After addition, the plate was incubated at 4 °C for 16 h. After incubation, the plate was washed three times with 300 µL of 1 × PBST. Before the signal acquisition, an oxygen scavenging system [5 mm PCA (P5630, Sigma–Aldrich), 200 nm rPCO (46852004, Oriental Yeast), and 2 mm Trolox] was added in each well to prevent Alexa Fluor 568 fluorophore bleaching.

After each removal and washing step, the residual solution on the immunoblot was eliminated by firmly tapping the plate against absorbent paper towels to ensure complete removal. All incubation process was carried out on an orbital shaker (NB‐101S, N‐Biotek) at 200 rpm. Signal acquisition was carried out on a microplate reader (Synergy H1 microplate reader, BioTek) running Gen5 3.08 software. The fluorescence signal of Alexa Fluor 568 was measured, with an excitation wavelength of 568 nm and an emission detection wavelength of 619 nm by using monochromator mode.

Standard curves were generated by fitting the average intensity value of each concentration (x) into a four‐parameter logistic model:

$$I=\frac{A_1-A_2}{1+{\left(x/x_0\right)}^p}+A_2$$ (1)

where I is the intensity from the measurement, x ₀ is the point of inflection,A ₁ means background intensity, A ₂ means maximum intensity, p is the hill coefficient.

DNA Hanger Construction in Flow Chamber

To tether λ‐phage DNA in DNA Skybridge, modified λ‐phage DNA molecules were injected into the flow chamber at a flow rate of 0.07 mL min⁻¹ with a syringe pump (Harvard Apparatus). The DNA was diluted in DNA binding buffer (pH 7.5, 30 mm Tris‐HCl, 100 mm potassium glutamate, 5 mm MgCl₂, and 0.0025% Tween‐20) at 5 pm concentrations. It was confirmed that 5 mm Mg²⁺ effectively suppressed nonspecific binding by neutralizing the charge interaction between DNA and Mg²⁺. After tethering, 500 µL of streptavidin (20 ng mL⁻¹, S‐888, Invitrogen) or Cy5‐streptavidin (20 ng mL⁻¹, SA1011, Invitrogen) solution was injected into the flow chamber constructed with the 3D structure quartz slide and then incubated for 5 min to coat the λ‐phage DNA substrates containing biotin. The complete construction process typically requires ≈12 min.

Single‐Molecule Immunoassay Experiments

All procedures involving mNG‐PABP and its antibody immunoassay were conducted in a reaction buffer (10 mm Tris‐HCl, pH 8.0, 100 mm KCl. 5 mm MgCl₂, and 0.0025% Tween‐20). To immobilize ≈300 Cy5‐DNA/RNA duplexes per a field‐of‐view, 5 pM Cy5‐DNA/RNA duplexes with either 114 nt or 25 nt poly(A) tails were introduced into the DNA Hanger chamber and incubated for 3 min. After washing 100 nm mNG‐PABP was injected into the chamber and incubated for 3 min. Following two washes, 53.3 nm AF594‐anti‐PABP antibody (clone 10E10, Santa Cruz) was added and incubated for 8 min. The chamber was then washed five times, and a fluorescence image was performed under an oxygen scavenging system.

All procedures for the TNF‐α immunoassay were performed using DNA binding buffer, except for the TNF‐α injection step. To fully coat the λ‐phage DNA with cAb, 7.2 nm cAb was introduced into the DNA Hanger chamber and incubated for 3 min. After two washes, TNF‐α was injected and incubated for 5 min at varying concentrations depending on the matrix: 0 – 1 nm in DNA binding buffer, 0 – 30 nm FBS, 0 – 100 pm in human serum. The chamber was then washed once (for TNF‐α concentrations below 1 nm) or twice (for concentrations of 1 nm and higher). Next, 40 nm dAb was introduced into the chamber and incubated for 8 min. Following five additional washes, fluorescence imaging was performed in an oxygen scavenging system. It was noted that a buffer containing oxygen scavenging components (eg. PCA and rPCO) should be used immediately after cAb injection to prevent a photoconversion of AF647 under visible light to red visible light. All biomolecule injections and washes were carried out using a syringe pump at a flow rate 0.09 mL min⁻¹, with a total volume of 100 to 200 µL of volume was used.

Single‐Molecule Imaging

The flow chamber was mounted on a custom‐built prism‐type total internal reflection fluorescence (TIRF) microscope (Olympus IX‐71) equipped with a 60× water‐immersion objective (Numerical Aperture = 1.2). Samples were illuminated based on the excitation wavelengths of their conjugated fluorophores: mNeonGreen at 488 nm, Cy3 and SYTOX Orange (Invitrogen) at 532 nm, Alexa Fluor 594 at 561 nm, and Cy5 at 638 nm ((all lasers from Cobolt)). Fluorescence signals were captured using an EMCCD camera (ImagEM C9100‐13, Hamamatsu) and processed with camera link frame grabbers (AS‐FBD‐1XCLD‐2PE4L‐F, Active Silicon). Images were acquired at a frame rate of 100 ms per frame, which was maintained consistently across all experiments. Prior to image acquisition, an oxygen scavenging system was injected to minimize photobleaching and photoblinking during image acquisition.^{[ 20 ]}

Data Analysis

Colocalization analysis was performed using the ComDet v0.5.5 plugin in ImageJ (National Institutes of Health, Bethesda, MD). The normalized non‐colocalization ratio for mNG‐PABP, as shown in Figure 2b, was defined as:

$$\frac{\mathrm{PABP}-\mathrm{alone}\mathrm{signals}+\mathrm{PABP}-\mathrm{Ab}\mathrm{colocalized}\mathrm{signals}}{\mathrm{Total}\mathrm{PABP}\mathrm{signals}}$$ (2)

where total PABP signals include PABP‐alone signals, PABP‐RNA colocalized signals, PABP‐Ab colocalized signals, and PABP‐RNA‐Ab colocalized signals. Here, PABP refers to mNG‐PABP, Ab to AF594‐anti‐PABP antibody, and RNA to Cy5‐DNA/RNA duplex. Similarly, the normalized non‐colocalization ratio for the AF594‐anti‐PABP antibody was calculated as:

$$\frac{\mathrm{Ab}-\mathrm{alone}\mathrm{signals}+\mathrm{Ab}-\mathrm{RNA}\mathrm{colocalized}\mathrm{signals}}{\mathrm{Total}\mathrm{Ab}\mathrm{signals}}$$ (3)

where total Ab signals include Ab‐alone signals, PABP‐Ab colocalized signals, Ab‐RNA colocalized signals, and PABP‐RNA‐Ab colocalized signals.

The average number of dAb per DNA was calculated by dividing the total number of dAbs by the number of λ‐phage DNA within each field of view. To quantify the number of dAb bound to individual λ‐phage DNA, a line was drawn along each DNA strand, and the plot profile function in ImageJ was used to identify fluorescence peaks. Peak resolution was determined based on the Rayleigh criterion: if the intensity of the dip between two adjacent peaks was less than 80% of the weaker peak's intensity, the peaks were considered resolvable and counted as two distinct events. Due to excessive signal overlap at high analyte concentrations, accurate peak counting was not feasible for TNF‐α concentrations above 100 pm in DNA binding buffer conditions.

For fluorescence intensity measurements, images acquired under each analyte concentration condition using the DNA Hanger FLISA were processed with the Rolling Ball Background Subtraction function in FIJI (radius: 5–7 pixels). For each image, five consecutive frames were averaged starting from the first frame in which the fluorescent signal appeared. This post‐processing approach reduced background and noise. To minimize nonspecific fluorescence from the quartz surface, only regions corresponding to the DNA Hangers were selected as regions of interest (ROIs) for quantitative analysis of signal count and average fluorescence intensity. Intensity values were obtained by subtracting the background fluorescence from the signal in the DNA ROI (i.e., ROI – background). DNA molecules that overlapped or were poorly aligned were excluded from analysis.

Fluorescence signals obtained from individual λ‐phage DNA molecules within a single image were averaged to yield a representative intensity value for each field of view (FOV). Data collected from multiple FOVs were further averaged for each analyte concentration, and the corresponding standard deviations were calculated. Using these values, a standard curve was generated by fitting the data to a binding saturation model via nonlinear least‐squares regression using the Levenberg–Marquardt algorithm. The model was based on the following equation:

$$B=\frac{B_{max}\ast \left[Analyte\right]}{K_d+\left[Analyte\right]}$$ (4)

where B is the observed signal, B_max was the maximum binding signal, [Analyte] was the analyte concentration, and K_d ​ was the dissociation constant. The LOD was determined by calculating the analyte concentration corresponding to the signal intensity equal to the mean of the blank sample plus three times its standard deviation, using the inverse function of the fitted standard curve.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Seol J., Kim B., Yu E.‐S., Jeong C., Lee J.‐B., DNA Hanger: Surface‐Minimized Single‐Molecule Immunoassay Platform. Small 2025, 21, 2409933. 10.1002/smll.202409933

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.