Introduction

The increasing complexity of biological research necessitates advanced multiplex immunoassays for comprehensive analysis of diverse molecules1,2,3,4, providing valuable insights into biological functions and diseases through a single assay. These multiplex immunoassays are broadly categorized into planar5,6,7,8 and suspension arrays9,10,11,12. Planar arrays, such as protein microarrays, are notable for their ability to analyze thousands of targets simultaneously in a single experiment. However, planar arrays often face challenges related to poor solution kinetics, resulting in prolonged incubation time and reduced sample throughput13,14. In clinical settings, the focus tends to be on high-throughput analysis of a few dozen targets, making planar arrays less suitable for such requirements.

Suspension arrays, in contrast, are preferred for analyzing multiple targets from large sample sets. These assays use barcoded particles as microcarriers, with probes attached to them, enabling precise targets identification. These barcoded microcarriers can be freely combined, allowing for flexible and customizable assay configurations to meet various clinical needs12,15. Technologies like Luminex10 and Cytometric Bead Array (CBA)16, which utilize spectral-coded microbeads, are among the most common. Compared to planar arrays, suspension arrays exhibit faster reaction kinetics due to the shorter diffusion path analytes must travel to interact with microcarriers, resulting in higher reaction efficiency and reduced incubation time13,14,15. Repeated measurements using hundreds of identical barcoded particles provide lower coefficients of variation (CVs) and improved data quality12. Despite their widespread use, suspension arrays still encounter challenges in demanding scenarios that require higher sensitivity, faster processing times, and smaller sample volumes14,17.

Microfluidic systems18,19,20 present a solution to some of these challenges by miniaturizing immunoassays, as exemplified by Micromosaic Immunoassays (μMIA)21,22. These systems offer numerous benefits, including faster antibody-antigen reactions due to enhanced surface area-to-volume ratios and reduced sample consumption due to smaller assay dimensions. The Ella Automated Immunoassay System23,24, for instance, showcases these advantages by performing multiplex immunoassays using microfluidic glass reaction tubes (Glass Nano Reactor, GNR) with specific capture antibodies immobilized inside the tubing. Despite these benefits, these methods are limited by low array densities—typically only a few dozen data points per mm²—and are hindered by complex, serial manufacturing processes and high production costs. Therefore, ongoing research and innovations are actively seeking solutions to these challenges25.

In this study, we introduce the “Lab-in-a-Tip” (LIT) system, which addresses the major challenge of achieving high-density array integration within confined spaces. By transforming suspension-based Graphically Recognizable Array of Suspension Particles (GRASPs)26 into a self-assembled planar array confined within a capillary, this system challenges the conventional paradigm that suspension arrays inherently offer superior performance. LIT effectively combines the advantages of planar arrays, suspension arrays, and microfluidic systems, expanding the scope of multiplex immunoassays. This integrated approach not only enables large-scale production and cost efficiency, but it also offers specialized modes that surpass the capabilities of Luminex, the current gold standard in multiplex detection. Among these modes, Ultrasensitive LIT can detect concentrations as low as the fg/ml level, achieving a sensitivity two orders of magnitude higher than Luminex. Speedy LIT reduces incubation time to just 15 min—compared to Luminex’s 210 min—achieving at least a tenfold reduction in processing time. Additionally, Microvolume LIT operates with as little as 10 µl of sample, just one-fifth of the typical volume required by Luminex. These innovations promise to significantly enhance both the efficiency and analytical capabilities of clinical and laboratory immunoassays.

Results

Principles of LIT immunoassay

Suspension array technologies, like Luminex10, Cytometric Bead Array (CBA)16, and our previously developed GRASPs multiplex platform26, are regarded as superior to traditional planar arrays in performance. The LIT immunoassay technology transforms suspension-based GRASPs into a high-density, self-assembled planar array confined within a capillary, challenging the conventional paradigm of suspension arrays’ performance advantages. In brief, the GRASPs consist of a 25 × 14 µm silica substrate, which is coated with high-reflection optical films at specific locations to form a two-dimensional (2D) barcode pattern. The encoding system is organized into a 2 × 4 grid (Fig. 1a), with eight dots in total. Seven of these dots (3 × 3 µm each) serve as encoding positions, together creating a 7-bit code, while the eighth, larger dot (5 × 4 µm), acts as an alignment marker to assist with machine vision recognition. Each encoding dot is visible under a microscope using bright-field imaging, attributable to the high-reflection thin-film coating, which reflects light and distinguishes these dots as representing a bit value of 1. In the absence of the high-reflection film, the bit value is 0. Each differently coded GRASPs is conjugated with specific capture antibodies. Approximately 10,000 GRASPs are uniformly self-assembled within a square quartz capillary, which is integrated with a standard pipette tip. Critical reagents, including streptavidin phycoerythrin (SAPE) and biotin-conjugated detection antibodies, are precisely freeze-dried at designated locations along the interior surface of the pipette tip (Fig. 1a and Supplementary Fig. 1a), thereby encapsulating all necessary immunoassay components within the LIT.

Fig. 1: Principles of LIT immunoassay.
figure 1

a Schematic illustration of “Lab-in-a-Tip” comprising self-assembled GRASPs, detection antibodies, and SAPE. GRASPs, which provides 128-plex coding via seven distinct coding dots, are conjugated with specific capture antibodies and self-assembled within a square quartz capillary integrated into a standard pipette tip. The LIT is loaded onto a liquid handling workstation, which allows for precise volume control. b The key steps of LIT multiplex immunoassay workflow. The assay procedure includes incubations of sample, biotin-conjugated detection antibodies, and SAPE. Incubation procedures were facilitated by dynamic aspiration, propelling the solution across the self-assembled GRASPs. Fluid heights are precisely controlled: the sample incubation fluid does not exceed the biotin-conjugated detection antibody level, the detection antibody solution is maintained below its midpoint with SAPE, and the SAPE solution is kept above its designated spot. After the assay, the barcodes on the GRASPs in the LIT are microscopically imaged, and the encoded data is decoded using a software algorithm. This algorithm performs image preprocessing, code recognition, and statistical extraction to quantify the fluorescence intensity of the captured antigens. Figure was created with MedPeer (medpeer.cn).

Full size image

The assay procedure consists of sequential incubation steps involving the sample, biotin-conjugated detection antibodies, and SAPE, with each step succeeded by a washing step (Fig. 1b). A custom-built liquid-handling workstation (Fig. 1a and Supplementary Fig. 1b) enables precise control of liquid volumes and the dissolution of reagents at specific heights during each assay step (Fig. 1b). In addition, direct comparison data by performing LIT manually versus automating system demonstrates that automation also significantly reduces the CV (Supplementary Fig. 2). Incubation and washing are optimized through dynamic aspiration, propelling the solution across the self-assembled GRASPs for efficient interaction and cleaning. After the assay, the GRASPs in LIT are examined using a custom-built imaging system that captures images of the GRASPs, decodes the encoded GRASPs, and quantifies bound antigens by measuring fluorescence intensity in the PE channel (Fig. 1b and Supplementary Fig. 3a). The decoding process extracts the data of the captured images in three stages: first, image preprocessing, where an edge-preserving smoothing filter, unsharp masking, and morphological closing operations are applied to remove noise and enhance contrast. In the second stage, code recognition is performed using the R-HOG (Round Histogram of Oriented Gradients) operator to locate each microparticle, followed by measuring its tilt angle and rotating it to a horizontal position. The code on each microparticle is then recognized using a template-matching algorithm, and the decoding efficiency is approximately 80% (Supplementary Fig. 3b). Finally, in the statistical extraction stage, fluorescence values such as mean, median, and standard deviation are calculated from the corresponding fluorescent images of the recognized microparticles (Supplementary Fig. 3c).

Establishment of LIT immunoassay

A major challenge in LIT immunoassay is achieving high-density GRASPs integration within the limited space of a capillary. To ensure the reliability of this integration, it is critical that GRASPs, once self-assembled within the LIT, can withstand multiple washing cycles. To verify the self-assembly of GRASPs, we assessed their retention under high-speed aspiration (50 μl/s) in various buffers, including PBST and serum, over 30 min. Our results showed that over 90% of GRASPs were retained, regardless of coding (Fig. 2a and Supplementary Fig. 4). Notably, most GRASPs loss occurred during the first three washing cycles, with negligible losses in subsequent cycles (Fig. 2b). These findings confirm that the quantity and diversity of GRASPs self-assembled in the LIT are sufficient to ensure robust detection throughout the assay (Supplementary Fig. 5a–c). Further refinements in the bioanalytical performance were achieved by adjusting the concentrations of biotin-conjugated detection antibody and SAPE, enhancing the median fluorescence intensity (MFI) and optimizing the signal-to-background ratios. The optimal concentrations were determined to be 5 μg/ml for both the detection antibody and SAPE (Fig. 2c, d).

Fig. 2: Establishment of LIT immunoassay.
figure 2

a Representative bright-field images of self-assembled GRASPs in LIT before and after rapid aspiration with different buffer solutions (PBST and serum). Scale bar = 50 µm. b The recovery rate after different cycles of washing. Median fluorescent intensity, background signal, and signal-to-noise ratios measured in LIT immunoassays at varying concentrations of detection antibodies (c) and SAPE (d). The standard curves for IL-8 measured by LIT (e) and suspension GRASPs assay (f), with representative bright-field and fluorescent images of GRASPs from both assays (g, h). Scale bar = 10 µm. i The cross-binding percentages of six cytokine analytes at 1 ng/ml were measured in a six-plex LIT assay. j Representative bright field and fluorescent images of cross-binding assay. Scale bar = 20 µm. k Standard curves plotted for a six-plex LIT assay (IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF). Median fluorescent intensities of IL-5 (l) and IL-8 (m) at concentrations of 1000 pg/ml, 200 pg/ml, and 40 pg/ml measured in two-plex LITs over 180 days of storage. n Stability of IL-5 and IL-8 over 6 months, as evidenced by the estimated error of MFI at the concentration of 1 ng/ml. All images shown are representative images from three independent experiments. Data shown are means ± standard deviation (S.D.) for three independent experiments. General data were analyzed with Origin 2021. The five-parameter logistic regression model was used to fit the standard curves. LOD and LOQ were defined as the concentrations corresponding to signal values of mean(blank) + 3σ(blank) and mean(blank) + 10σ(blank), respectively, where mean(blank) is the average signal and σ(blank) is the standard deviation of the blank measurements.

Full size image

To validate the bioanalytical performance of LIT, we developed a single-plex LIT for IL-8 detection (Supplementary Fig. 6). A standard curve for IL-8 LIT was established. The LOD and LOQ of IL-8 LIT were calculated to be 0.8 pg/ml and 6.6 pg/ml, respectively (Fig. 2e, g). For comparison, a single-plex suspension GRASPs assay for IL-8 detection was also developed using the same antibodies, resulting in LOD and LOQ values of 0.8 pg/ml and 5.9 pg/ml, respectively (Fig. 2f, h). Furthermore, the incubation time of LIT was only about half of that of suspension array. The reaction kinetics experiments confirmed the faster reaction rate of LIT assay (Supplementary Fig. 7a, b). The linear range and CV were also compared between the two methods (Supplementary Fig. 8a–c), indicating the bioanalytical performance of LIT is on par with suspension GRASPs assay. Additionally, we developed single-plex LIT assay and suspension GRASPs assay for other cytokines, comparing their LODs, LOQs, and CVs (Supplementary Figs. 9–11). The results further confirmed the detection performance is comparable between single-plex LIT and suspension GRASPs assay.

To evaluate the potential of the LIT technology for multiplex immunoassay applications, we utilized the LIT platform to simultaneously detect a panel of human cytokines: IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF. Initially, we validated the specificity of the LIT platform for detecting multiple cytokines. Specifically, we developed a six-plex Standard LIT, which includes six uniquely encoded microparticles, each conjugated with specific capture antibodies for individual cytokines. This system was tested using a mixture of six different detection antibodies and SAPE (Supplementary Fig. 12a). To evaluate potential cross-reactivity, we applied the LIT platform to detect individual cytokine analytes at a concentration of 1 ng/ml. For example, when IL-4 was the single analyte, GRASPs targeting IL-4 exhibited signal intensities as high as 15,423, while signal intensities for GRASPs targeting other cytokines remained as low as 30 (Fig. 2j and Supplementary Fig. 12b). This resulted in a specificity ratio exceeding 500. Similarly, when other cytokines such as IL-1β, IL-5, IL-6, IL-8, and GM-CSF were tested individually, the specificity ratios consistently exceeded 100 (Supplementary Fig. 12b), with cross-binding rates remaining below 1% (Fig. 2i). These results strongly confirm the high selectivity and specificity of our platform for detecting multiple biomarkers without cross-reaction. Furthermore, to ensure the selectivity of the platform in detecting mixed analytes, we conducted additional experiments with a mixture of six cytokines as analytes and individual detection antibodies. The results, presented in Supplementary Fig. 13a, b, demonstrate that the cross-binding rates remained below 1% under these conditions, further validating the platform’s robustness and reliability. Finally, to provide a comprehensive performance comparison, we evaluated the specificity of the LIT platform relative to the suspension GRASPs assay. As shown in Supplementary Fig. 14a, b, the LIT platform demonstrated comparable levels of specificity, underscoring its robustness and versatility.

Subsequently, we validated the performance of the Standard LIT in multiplex immunoassay. The six-plex LIT was utilized to evaluate its detection capabilities across varying concentrations of a six-cytokine analyte mixture. The LODs for IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF were determined to be 0.8–1.1 pg/ml (Fig. 2k). When compared to the suspension GRASPs assay (Supplementary Fig. 15a), the LODs were similar (Supplementary Fig. 15b). The linear dynamic ranges and CVs of both methodologies were fundamentally comparable (Supplementary Figs. 16–18). Furthermore, to evaluate the stability of LIT assays, we prepared a two-plex LIT for IL-5 and IL-8 detection. After storage at −20 °C for 180 days, the MFI values for high, medium, and low concentrations of IL-5 and IL-8 remained relatively stable, with CVs below 10% (Fig. 2l, m, n, and Supplementary Fig. 19). This stability highlights the robustness of Standard LIT assays, thus facilitating further research into their diverse applications.

Three modes of LIT: Ultrasensitive LIT, Speedy LIT and Microvolume LIT

Among the various multiplex immunoassays, Luminex, regarded as the gold standard, and Cytometric Bead Array (CBA) technologies are the most widely used. However, these technologies often underperform in specific scenarios, such as detecting low-abundance analytes27, requiring rapid results28,29, or handling analyses with limited biological material30,31. Consequently, to address these diverse application needs, we developed three modes of the multiplex LIT platform: Ultrasensitive LIT, Speedy LIT, and Microvolume LIT. Each mode optimizes the LIT platform’s capabilities to meet unique challenges posed by different research and diagnostic requirements, enhancing its applicability across a broad range of biomedical fields.

In scenarios where the concentration of the analyte of interest is exceedingly low, detection methodologies must meet heightened requirements. The Ultrasensitive LIT, designed to detect trace analytes, incorporates a tyramide signal amplification (TSA) system32 to enhance sensitivity (Fig. 3a, b). Standard curves for IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF demonstrated that the LODs were improved by approximately 100-fold compared to Standard LIT (Fig. 3c, d). This method extends detection capabilities to levels comparable with single-molecule technology33, achieving sensitivities as low as fg/ml while maintaining a dynamic range of over three logarithmic units (Fig. 3c, d and Table 1).

Fig. 3: The three LIT modalities enable applications requiring high sensitivity, rapid analysis, and minimal sample volume.
figure 3

Schematic illustrations of key workflow steps of Ultrasensitive LIT (a) and rationale behind tyramide signal amplification technology (b). Figure was created with MedPeer (medpeer.cn). c LODs of IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF measured by four LIT modalities: Ultrasensitive LIT, Speedy LIT, Microvolume LIT and Standard LIT. d-f. Dose dependent median fluorescence intensities of IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF at different concentrations were compared between Standard LIT (black) and each of the other LIT modalities: Ultrasensitive LIT, Speedy LIT, and Microvolume LIT (red or green). Data presented as mean ± SD from three replicates. General data were analyzed with Origin 2021. The five-parameter logistic regression model was used to fit the standard curves. LOD and LOQ were defined as the concentrations corresponding to signal values of mean(blank) + 3σ(blank) and mean(blank) + 10σ(blank), respectively, where mean(blank) is the average signal and σ(blank) is the standard deviation of the blank measurements.

Full size image
Table 1 Comparative performance of four LIT modes, Luminex, and ELISA
Full size table

In addition to detecting trace analytes, obtaining rapid results is crucial. Typically, the Standard LIT protocol reduces incubation time to 1 h, faster than methods such as Luminex10 and ELISA34 (Table 1). However, in urgent diagnostic scenarios, even an hour can be too long. Compared to single-plex immunoassays such as chemiluminescence35,36 and lateral flow assays28, which provide results in less than 30 min, there remains a need for Standard LIT to achieve even faster processing. To address this, we developed Speedy LIT, which reduce the incubation time to just 15 min (Supplementary Fig. 20 and Table 1). Remarkably, Speedy LIT not only delivers rapid results but also maintains high sensitivity and precision, with only a minor reduction in LODs and LOQs compared to the Standard LIT approach (Fig. 3c, e and Supplementary Fig. 21).

In addition to speed, the LIT platform addresses the challenges of handling small or valuable samples, a limitation often encountered with traditional technologies. Technologies such as Luminex10 and ELISA34 typically require a minimum sample volume of 50 µl, which can be restrictive when dealing with precious samples. To enhance the capability for analyses requiring small sample volumes, we developed Microvolume LIT. It takes advantage of the LIT platform’s microfluidic environment, allowing for efficient testing with just 10 μl of sample. Standard curves were generated using sample volumes of 50 μl, 25 μl, and 10 μl. The results indicated that both the LODs and LOQs for 25 μl and 10 μl were only slightly compromised compared to those for 50 μl (Fig. 3c, f and Supplementary Fig. 21).

Performance comparison of LIT with Luminex and ELISA

To further assess the analytical performance of the LIT platform, we conducted a comparative analysis with Luminex, which utilizes spectral-coded microbeads for multiparameter analysis (Fig. 4a), and ELISA, the traditional gold standard for single protein analyte detection (Fig. 4b).

Fig. 4: Performance comparison of LIT with Luminex and ELISA.
figure 4

Schematic diagrams illustrating the methodologies used in Luminex® Assay (a) and ELISA (b). Figure was created with MedPeer (medpeer.cn). c Dose dependent median fluorescence intensity of IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF at different concentrations were compared between Standard LIT (black) and Luminex (red). d Dose dependent median fluorescence intensity and absorbance of IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF at different concentrations were compared between Standard LIT (black) and ELISA (red). Data presented as mean ± SD from three replicates. General data were analyzed with Origin 2021. The five-parameter logistic regression model was used to fit the standard curves. LOD and LOQ were defined as the concentrations corresponding to signal values of mean(blank) + 3σ(blank) and mean(blank) + 10σ(blank), respectively, where mean(blank) is the average signal and σ(blank) is the standard deviation of the blank measurements.

Full size image

Both Standard LIT and Luminex support multiplex immunoassays, providing comparable LODs and LOQs for various analytes (Fig. 4c, Table 1, and Supplementary Fig. 22). The analytical performance, including linear dynamic ranges, is thoroughly documented in Supplementary Fig. 23 and Supplementary Table 1. The LIT platform, particularly its Ultrasensitive, Speedy, and Microvolume LIT modalities, significantly outperforms Luminex in specific application situations. Ultrasensitive LIT achieves sensitivity enhancement of 1-3 orders of magnitude, a capability unmatched by Luminex for ultra-sensitive detection. Speedy LIT enables analysis 14 times faster, and Microvolume LIT requires only one-fifth of the sample volume typically needed (Table 1).

Compared to ELISA, the Standard LIT exhibits superior LODs and analytical performance (Fig. 4d, Table 1, and Supplementary Figs. 22, 24, Table 1). The Ultrasensitive LIT offers a sensitivity enhancement of 2-3 orders of magnitude, Speedy LIT enables analysis 20 times faster, and Microvolume LIT reduces the required sample volume to one-tenth. Unlike ELISA, which is restricted to single-analyte assessments, LIT enables the simultaneous detection of multiple analytes, providing a broader scope of analysis in bioanalytical applications.

Performance comparison of LIT with Luminex and ELISA in detecting serum samples

After evaluating the detection capabilities of LIT, Luminex, and ELISA using protein standards in PBST, we extended our analysis to assess their performance in detecting targets within serum samples. Initially, we investigated the impact of serum matrix effects on LIT by employing negative human serum as the dilution buffer for standard curve generation. Our results indicated that the serum matrix did not compromise sensitivity for six cytokines (Fig. 5a, b). In spike recovery experiments conducted at high, medium, and low analyte concentrations, all six cytokines were detectable in serum samples, with recovery rates ranging from 90% to 110% (Fig. 5c).

Fig. 5: Performance comparison of LIT, Luminex, and ELISA in detecting targets in serum samples.
figure 5

a Dose-dependent median fluorescence intensity of IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF at various concentrations compared between Standard LIT assays with (red) and without (black) added serum. b Comparison of LODs for IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF with (pink) and without (blue) serum. c Spike recoveries for cytokines at high, medium, and low analyte concentrations. Linear regression plots comparing serum concentrations of IL-6 (d) and IL-8 (e) as determined by Luminex and LIT. Linear regression plots for IL-6 (f) and IL-8 (g) serum concentrations, determined by ELISA and LIT. Linear regression plots for IL-6 (h) and IL-8 (i) serum concentrations, determined by chemiluminescence and LIT. Data are presented as mean ± SD; n = 3 repeated tests. General data were analyzed with Origin 2021. The five-parameter logistic regression model was used to fit the standard curves. LOD and LOQ were defined as the concentrations corresponding to signal values of mean(blank) + 3σ(blank) and mean(blank) + 10σ(blank), respectively, where mean(blank) is the average signal and σ(blank) is the standard deviation of the blank measurements.

Full size image

We collected serum samples from patients to measure cytokine concentrations (IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF) using LIT, Luminex, and ELISA methods. Due to the inherent properties of each cytokine, IL-1β, IL-4, IL-5, and GM-CSF were mostly undetectable by all these methods in 50 samples. IL-6 was detected in 30 samples by LIT, 23 by Luminex, 19 by ELISA, and 28 by clinical chemiluminescence assays. Similarly, IL-8 was detected in 45, 49, 47, and 44 samples by each respective method (Supplementary Table 2). The measured concentrations of cytokines were comparable consistent across LIT, Luminex, ELISA, and clinical chemiluminescence assays. Among the samples in which cytokines were detected, the concentrations demonstrated a high correlation between the results obtained with LIT and those with ELISA (Fig. 5d, e), as well as between LIT and Luminex (Fig. 5f, g). Furthermore, a strong correlation was also observed between the measurements from LIT and those obtained clinically via chemiluminescence assays (Fig. 5h, i). However, compared to these methods, LIT offers significant advantages in terms of reduced assay time and enhanced automation capabilities, making it a superior option for clinical and research applications.

Discussion

Achieving multiplexing ability in confined spaces is key to microfluidic chips37. μMIA is the most commonly used technology38. The basic principle involves immobilizing different capture antibodies in parallel microchannels and changing the microchannels orthogonally to flood samples across the immobilized capture antibodies. However, this method cannot array antibodies within the tips. Our LIT technology addresses this challenge by simply aspirating antibody modified GRASPs and self-assembling them onto the inner surface of the capillary. Our results suggest that the performance of assembled GRASPs is not affected due to the mild conditions. Furthermore, a high array density of up to 500 particles/mm² is achieved, which is about 20 times higher than that of microarray made by robotic spotting39,40. Unlike the spherical microparticles commonly used in suspension multiplex immunoassays, our GRASPs have a planar configuration with a much larger interacting area to keep them firmly attached to the inner surface of the capillary. The mechanism of this immobilization results from synergistic physical and chemical interactions. Specifically, hydrophobic and polar interactions between antibodies on the GRASPs’ surfaces and the capillary complement the formation of covalent bonds between the amine groups on the GRASPs and the aldehyde groups, creating a stable assembly41. The robust self-assembly withstands repeated buffer washing, enabling LIT to integrate directly with liquid handling workstations. This integration addresses the challenge of automating the multiple washing steps required in suspension arrays42. By leveraging the graphic encoding capability of GRASPs, rather than relying on location information of protein microarray, it facilitates encoding without the need for bulky devices to precisely deposit specific probe molecules onto each spot in a flat array. A variety of differently encoded GRASPs can be arbitrarily selected as in the suspension multiplex immunoassays, which greatly improves the panel flexibility of the method14. Each encoded GRASP can deliver hundreds of repeats rather than the several repeated spots of protein microarray, thus ensuring that the same target is detected hundreds of times, thereby guaranteeing the data quality of LIT with a low CV26. Additionally, for each batch of GRASPs, we conduct a rigorous fluorescence-based quality control process to ensure that only batches with a PE MFI ≥ 8000 are selected for further experimentation. This threshold ensures the high quality of each batch of GRASPs, thereby minimizing batch-to-batch variability and controlling variability in the final LIT preparation.

The design of integrating all reagents into the tip of the LIT enhances both portability and usability. It significantly reduces the costs and package size associated with the cold chain transportation of reagent kits. Additionally, this design simplifies the operational process by eliminating the need to mix different reagents when testing various panels, which greatly facilitates ease of use for operators. Furthermore, the streamlined design reduces the cost per six-plex LIT to approximately 0.3 dollars, making it an economically viable option for large-scale applications.

In addition to these advantages, these designs also enhance performance. Within the LITs, dynamic aspiration facilitates incubation and washing by driving the solution across the self-assembled GRASPs. As samples are pipetted back and forth through the capillary, they interact with capture antibodies conjugated to the GRASPs. The depleted analytes near the GRASPs are replenished by the continuous flow, and the diffusion boundary layer thickness is compressed to a thin layer due to the rapid, constant back-and-forth pipetting43,44,45,46. The thinning of the boundary layer can be explained using principles of fluid dynamics. According to the theory of laminar boundary layers, the thickness \(\delta \left(x\right)\) is given by the formula:

$${{{\rm{\delta }}}}\left({{{\rm{x}}}}\right)=\frac{5.0x}{\sqrt{R{e}_{x}}}$$

where the Reynolds number

\({{{\rm{R}}}}{e}_{x}\) is defined as:

$${{{\rm{R}}}}{{{{\rm{e}}}}}_{{{{\rm{x}}}}}=\frac{\rho {U}_{\infty }x}{u}$$

with \(\rho\) being the fluid density, \({U}_{\infty }\) the free-stream velocity, \(x\) the distance from the leading edge, and \(u\) the dynamic viscosity of the fluid. This relationship indicates that \({{{\rm{\delta }}}}\left({{{\rm{x}}}}\right)\) is inversely proportional to the square root of \({U}_{\infty }\)47. The rapid and repetitive back-and-forth pipetting implemented in the system induces an increased velocity, leading to a reduction in boundary layer thickness, thereby shortening the diffusion distance. This mechanism facilitates enhanced molecular interactions and contributes to the reduction in incubation time. Consequently, the total incubation time for the Standard LIT is decreased to one hour and for the Speedy LIT to just 15 min, compared to approximately 3.5 h required by traditional suspension arrays. The Standard LIT achieves sensitivity comparable to chemiluminescence48,49, establishing a reliable baseline for detection. Furthermore, Ultrasensitive LIT extends detection to a level comparable to single-molecule methods33, achieving sensitivities as low as fg/ml through the use of TSA technology32. Although TSA technology is widely used in pathological analysis, its primary application is in qualitative research due to inherent challenges in quantitative assessments. These challenges stem from the technology’s high sensitivity to the thoroughness of washing steps: inconsistencies in residual solutions can significantly impact CV, thereby affecting its practical applications in quantitative analysis. In suspension arrays, the washing process may leave residual enzymes, leading to exponential signal amplification and increased background noise. In contrast, the LIT platform precisely controls the duration of the washing steps to the second with the help of liquid handling workstations, ensuring thorough cleaning, minimizing solution residue, and maintaining consistent CV across different assay batches. In Microvolume LIT mode, we have tested a minimum sample volume as low as 10 μl. However, theoretically, Microvolume LIT could be extended to even smaller sample volumes, such as 1 μl. This may require addressing challenges associated with sample evaporation during repeated flow over attached GRASPs during incubation. Introducing a small volume of air after sample loading to seal the liquid may potentially resolve this issue. Expanding on these innovations, we systematically assessed our platform against both established and emerging technologies. The comparative findings are detailed in Supplementary Table 3 demonstrating that the LIT platform frequently outperforms existing methods, thereby confirming its transformative potential in the field.

In summary, LIT technology not only preserves many advantages of microfluidic chips such as high sensitivity, fast processing, and low-sample consumption, but also supports large-scale and low-cost manufacturing. It can easily be scaled up to handle 96 or even 384 samples simultaneously to achieve ultrahigh throughput. Furthermore, our preliminary experiments suggest that the LIT technology can be readily extended to multiplexed nucleic acid detection50,51. These features of LIT technology make it a potential powerful multiplex tool for a variety of scientific and diagnostic scenarios.

Methods

Design, fabrication, and modification of GRASPs

For fabricating GRASPs, a 25 × 14 µm silica substrate was chosen. High-reflection optical films made of alternating Si3N4 and SiO2 layers, with thicknesses finely tuned for over 90% reflectance at the reference wavelength, were deposited at specific sites to form a 2D barcode pattern in a 2 × 4 grid. Seven 3 × 3 µm dots served as encoding positions for a 7-bit code, and an 8th, 5 × 4 µm dot was for alignment. Encoding dots with the film were visible as bit value 1 under a microscope. In the assay, for fluorescence readout, high-reflection dot regions were excluded from pixel extraction. During imaging, light transmitted through the SiO2 substrate with little loss but was strongly reflected at dot regions, creating a dark pattern on the particle’s transmission image for identification.

The encoded GRASPs were suspended in a 1.5 ml Protein LoBind microtube (Eppendorf, 022431081) at a concentration of 5×105 particles/ml, washed twice with 95% ethanol (SCR, 10009218), and then agitated in a 5% (3-Aminopropyl) dimethylethoxysilane (APDMS, J&K, 3069-29-2) solution (95% ethanol) for 30 min. Afterward, the GRASPs were washed once with 1 ml of N, N-dimethylformamide (DMF, Sigma Aldrich, 68-12-2) and agitated overnight in a 10% succinic anhydride solution (J&K, 185369). The GRASPs were subsequently washed three times with anhydrous ethanol and centrifuged at 3000 × g for 3 min to remove the supernatant. The carboxyl-functionalized encoded GRASPs were initially subjected to a single washing step utilizing 1 ml of 0.1 M 2-(N-morpholino) ethanesulfonic acid (MES, ThermoFisher, 28390) buffer (pH 5.0). Subsequently, 250 µl of N-hydroxysuccinimide (NHS, 25 mg/ml) and 250 µl of 1-Ethyl-3-(3-dimethylaminopropyl) carbodiimide (25 mg/ml, EDC, ThermoFisher, 229801) were added to the suspension, which was agitated for 30 min. The GRASPs were further washed twice with 0.1 M sodium acetate (NaAc, Sigma Aldrich, 127-09-3) buffer (pH 5.0). Consecutively, 10 μg of designated capture antibody dissolved in 500 μl of 0.1 M NaAc buffer (pH 5.0) was added to the encoded GRASPs, specifically #027 for IL-4 (R&D, MAB604-500), #034 for IL-6 (R&D, MAB206-500), #051 for IL-1β (R&D, MAB601-500), #080 for GM-CSF (R&D, MAB615), #088 for IL-8 (R&D, MAB208-500), and #089 for IL-5 (R&D, MAB405-500). After incubating overnight at 4 °C, the capture antibody conjugated encoded GRASPs were washed twice with PBST (ThermoFisher, 28360) and once with 1 M Tris, followed by a 30-min incubation in 1 M Tris (Solarbio, T1080) and a 1-h incubation in 1% BSA (Sigma Aldrich, 9048-46-8). Finally, after two additional PBST washes, the conjugated GRASPs were resuspended in PBST.

Quality control of GRASPs

Approximately 500 GRASPs, each conjugated with capture antibodies, are incubated with 500 µl of PE-conjugated fluorescent secondary antibody (PE-conjugated goat anti-rat IgG (H + L), Invitrogen, PA1-29628; R-phycoerythrin goat anti-mouse IgG, Invitrogen, P852; 1 µg/ml) for 1 h in the dark. After incubation, the GRASPs are washed three times with PBST. The quality of each batch of capture antibody labeled GRASPs is then evaluated by the MFI of the PE signal on the GRASPs. Only those batches of GRASPs demonstrating a PE MFI of ≥8000 are deemed to have passed quality control and are selected for further use in LIT fabrication.

Fabrication of LITs

A square capillary (Vitrocom, 8270) with an inner diameter of 0.7 mm was ultrasonicated sequentially in acetone (SCR, 10000418) and anhydrous ethanol for 10 min each. Following this, it was treated with a 5% aminosilane (APDMS) solution (prepared in 95% ethanol) and agitated for 30 min, and then rinsed twice with anhydrous ethanol. Subsequently, the capillary was incubated in a 2.5% glutaraldehyde (MACRUN, 111-30-8) solution (prepared in PBS) for 2 h. Finally, it was washed three times with deionized water and dried under a nitrogen flow.

The suspension of capture antibody-conjugated GRASPs was introduced into the aldehyde-modified capillary. Following overnight incubation at 4 °C, the GRASPs self-assembled on the interior surface of the capillary. To ensure a consistent number of microparticles (~10,000) in each planar array, we first resuspend the microparticles in PBS buffer and count them using a hemocytometer. Based on the counts, we adjust the concentration to 1.0 ×106 microparticles per ml. For single-plex LITs, 11.2 µl of the microparticles suspension (~11,200 microparticles) is introduced into the capillary. For 6-plex LITs, we take 1900 microparticles of each encoded type, mix them thoroughly, and adjust the concentration so that the total number of microparticles in mixed suspension is ~11,200. This mixed suspension is then assembled into the capillary using the same procedure as for the single-plex assay. After the capillary was then blocked with 1% BSA for 1 h and dried under a nitrogen flow, it was coupled with a standard 200 µl pipette tip using a latex tube (OUPLI). We have specially roughened the reagent deposition area of the custom-designed standard tips to facilitate the deposition of the reagents. Subsequently, 2 µl of 175 µg/ml mixed detection antibodies and 4 µl of 300 µg/ml SAPE in PBS (1%BSA + 2%Trehalose) were freeze-dried onto specific locations of the tip (4 cm and 5.5 cm from the tip’s end, respectively). LITs were prepared and stored for use.

Validation of the self-assembly stability of LITs

The LITs were loaded onto the homemade liquid handling workstation and treated with buffer at an aspiration and dispensing rate of 50 μl/s for 30 min. Both PBST buffer and serum buffer were evaluated. Before and after treatment, nine images across different areas of the LITs were taken in bright-field mode. The number and positions of GRASPs within the LITs were calculated and analyzed.

Optimization of LIT parameters

To determine the optimal concentration of the biotin-conjugated detection antibody and SAPE, a single-plex LIT for human IL-8 was fabricated as a model. Approximately five thousand GRASPs conjugated with IL-8 capture antibody (R&D, MAB208-500) were self-assembled in the LIT. After being loaded on the liquid handling workstation, the LITs were incubated with 50 μl of 1 ng/ml of human IL-8 protein (R&D, 208-IL-010) for 40 min at a pipetting rate of 50 μl/s, followed by a washing step with PBST for 1 min. Incubation and washing within the LITs were facilitated by dynamic aspiration, which propelled the solution across the self-assembled GRASPs, ensuring efficient interaction and cleaning. The LITs were then incubated with 70 μl of different concentrations of IL-8 detection antibody (0.5, 1, 2, 3, 4, 5, and 6 µg/ml) at a pipetting rate of 50 μl/s, succeeded by a PBST wash. The LITs were then incubated with 120 μl of 6 μg/ml SAPE (ThermoFisher, SA10044) for 8 min at a pipetting rate of 50 μl/s, followed by another washing step. The LITs were then imaged using a homemade imaging system. To ensure comprehensive analysis, nine images were taken and analyzed for both bright-field and PE channels across different areas of the LITs. During image acquisition, the system was set to automatically capture these images, ensuring that at least 50 microparticles (GRASPs) of each encoded type were included in the analysis. This approach guarantees sufficient representation for each encoded microparticle type, which is crucial for maintaining data quality with a low CV. The analysis method of the images can be referred to in our previous work26. Briefly, the GRASPs were recognized and the PE fluorescence intensity of the GRASPs and background were measured. The signal-to-noise ratio (SNR) was calculated by dividing the signal strength by the background noise intensity. After determining the optimal detection antibody (R&D, BAF208) concentration, the images of LITs at SAPE concentrations of 0.5, 1, 2, 3, 4, 5, and 6 µg/ml were taken and analyzed.

Immunoassays with standard LITs

Immunoassays for individual analytes were performed using dedicated single-plex LITs, each specifically designed for detecting human cytokines IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF. Taking human IL-8 as an example, the immunoassays utilized IL-8 single-plex LITs, which were systematically processed following a standard procedure on a liquid handling workstation. Specifically, LITs were incubated with 50 μl of various concentrations of human IL-8 standard protein solutions (0.64 pg/ml, 3.2 pg/ml, 16 pg/ml, 80 pg/ml, 400 pg/ml, 2 ng/ml, 10 ng/ml) for 40 min at a pipetting rate of 50 μl/s, followed by a washing step with PBST for 1 min. Incubation and washing within the LITs were facilitated by dynamic aspiration, which propelled the solution across the self-assembled GRASPs, ensuring efficient interaction and cleaning. Subsequently, 70 μl of PBST was accurately dispensed to a predetermined location within the LITs, specifically for dissolving the lyophilized detection antibody powder. Post reconstitution, the LITs were incubated with 70 μl of the detection antibody for 12 min at a pipetting rate of 50 μl/s, succeeded by a PBST wash. Following this, 120 μl of PBST was precisely drawn to another designated location for reconstituting the lyophilized SAPE powder. After reconstitution, the LITs were incubated with 120 μl of SAPE for 8 min at a pipetting rate of 50 μl/s, followed by another washing step. Finally, the LITs were imaged using a homemade imaging system (Supplementary Fig. 2). To ensure comprehensive analysis, nine images were taken and analyzed for both bright-field and PE channels across different areas of the LITs. During image acquisition, the system was set to automatically capture these images, ensuring that at least 50 microparticles (GRASPs) of each encoded type were included in the analysis. This approach guarantees sufficient representation for each encoded microparticle type, which is crucial for maintaining data quality with a low CV.

In experiments of establishing standard curves for multiplex immunoassays, six-plex LITs were utilized to simultaneously assay six human cytokines: IL-1β (R&D, 201-LB-005), IL-4 (R&D, 204-IL-010), IL-5 (R&D, 205-IL-005), IL-6 (R&D, 206-IL-010), IL-8 (R&D, 208-IL-010), and GM-CSF (R&D, 215-GM-010), following a similar protocol to that used for single-plex LIT immunoassays (The six detection antibodies used in the six-plex LITs were: IL-1β, R&D, BAF201; IL-4, R&D, BAF204; IL-5, R&D, BAM6051; IL-6, R&D, BAF206; IL-8, R&D, BAF208; and GM-CSF, R&D, BAF215). The principal variation involved incubating the six-plex LITs with 50 μl of mixed cytokine solutions at various concentrations, containing all six cytokines, each ranging from 0.64 pg/ml to 10 ng/ml. To assess LITs specificity, the six-plex LITs were incubated with individual cytokine analytes at a concentration of 1 ng/ml. Furthermore, to ensure the selectivity of the platform in detecting mixed analytes, the six-plex LITs were incubated with a mixture of six cytokines as analytes and individual detection antibodies. In the linear dynamic range detection experiments, mixed cytokine solutions containing all six cytokines were serially diluted two-fold from 20 ng/ml to 0.6 pg/ml. The measured signal values were plotted against the corresponding concentrations to generate scatter plots, and linear regression was applied to determine the concentration range where the assay response remained linear. The dynamic range refers to the concentration range where the assay maintains a linear relationship between the measured signal and analyte concentration.

In the stability evaluation experiments for LITs, two-plex LITs targeting IL-5 and IL-8 cytokines were prepared and maintained at −20 °C in the dark. Subsequent assessments employed these LITs to measure mixtures of IL-5 and IL-8 at varying concentrations (1000 pg/ml, 200 pg/ml, and 40 pg/ml), over a span of six months: initially, after one day, one week, and monthly up to six months.

Manual and automated LITs processing by multiple operators

IL-8 capture antibody-conjugated GRASPs were self-assembled into LITs for evaluation. Each of eight operators incubated LITs with 100 µl of R-phycoerythrin goat anti-mouse IgG secondary antibody (1 µg/ml, Invitrogen, P852) using dynamic aspiration over a 5-min period. Immediately following the incubation, the LITs were washed three times with PBST. Three LITs were conducted per operator (n = 3). Simultaneously, these operators also processed three LITs using a custom-built liquid handling workstation under identical conditions. The mean fluorescence intensity (MFI) of the PE signal on the GRASPs was then quantified.

Immunoassay with suspension GRASPs

First, the suspension arrays were prepared by mixing the required GRASPs with different combinations of capture antibodies in a Protein LoBind microtube. Approximately 500 GRASPs per code were added to each 1.5 ml centrifuge tube. Then, 100 μl sample was added to each well. The centrifuge tubes containing the suspensions were placed on a shaker (Grant-bio, V-32) and incubated for 60 min. After incubation, the suspensions were washed three times by centrifugation at 2000 × g using washing solution (1×PBST, phosphate buffered saline, with 0.5% Tween 20). Then, 100 μl of a prepared cocktail of biotinylated detection antibodies were added, and the suspension was incubatated for 30 min at room temperature (R.T.) on a shaker. After centrifugation again for three washes, 100 μl of SAPE (Streptavidin/R-phycoerythrin conjugate, 0.5 μg/ml) was then added, and the reaction was kept at R.T. on a shaker for 15 min. Finally, the suspension was washed 3 times with 1×PBST and then transferred to the assay plate. Once the GRASPs settled to the bottom of the wells (usually taking less than 2 min), the plate is ready for image data acquisition.

Reaction kinetic experiments of the LIT assay

Initially, IL-8 capture antibody-conjugated GRASPs were self-assembled into the LIT configuration for subsequent analysis. These LITs were then processed through our custom-built liquid handling workstation, with the 50 µl of fluorescent secondary antibody (2 µg/ml; R-phycoerythrin goat anti-mouse IgG, Invitrogen, P852) applied at an aspiration and dispensing rate of 50 µl/s over intervals from 1 to 20 min. Following each interval, the LITs were washed with PBST for 1 min. We captured the images of GRASPs across different areas of each LIT to calculate the fluorescence intensity in the PE channel.

Reaction kinetic experiments of the suspension array

For suspension arrays, we added approximately 500 IL-8 capture antibody-conjugated GRASPs to each 1.5 ml centrifuge tube. We then introduced 100 µl of fluorescent secondary antibody (2 µg/ml; R-phycoerythrin goat anti-mouse IgG, Invitrogen, P852) to each tube and incubated them on a shaker for periods ranging from 1 to 20 min. Post-incubation, the suspensions were washed three times with PBST by centrifugation at 2000 × g. The washed GRASPs were resuspended in 50 µl of PBST and transferred to a 96-well plate for imaging and fluorescence intensity analysis in the PE channel.

Immunoassays with three modes of LITs

The Speedy LIT protocol adheres to a similar methodology as the single-plex LIT immunoassays but distinguishes itself by compressing the entire procedure to just 15 min. This rapid approach includes an 8-min incubation with the sample, a 4-min incubation with the detection antibody, and a 3-min incubation with SAPE, with each step followed by a 1-min wash. Similarly, the Microvolume LIT protocol follows the foundational approach of single-plex LIT immunoassays but uniquely minimizes the sample incubation volume to either 25 μl or 10 μl. For the Ultrasensitive LIT preparation, 0.5 μg/ml SA-HRP (Elabscience, E-AB-1043), 4 μg/ml biotin-tyramine (Sigma Aldrich, 41994-02), and 0.003% H2O2 (Sigma Aldrich, 7722-84-1) are prepared in advance. While the procedural sequence is largely similar to the standard LIT immunoassay, the Ultrasensitive LIT mode includes two additional steps: a 10-min SA-HRP incubation and a 2-min biotin-tyramine (in 0.003% H2O2) incubation. The whole workflow consists of a 40-min sample incubation, a 12-min detection antibody incubation, these two additional steps, and an 8-min SAPE incubation, with a 2-min washing step between each stage to ensure consistency and precision.

Immunoassay with Luminex kit

A Human Premixed Multi-Analyte kit (Luminex® Discovery Assay, LXSAHM-06) was purchased for conducting multiplex immunoassay of IL-1β, IL-4, IL-5, IL-6, IL-8, and GM-CSF. The experimental procedures followed the kit instructions. Briefly, samples were incubated with beads in 96-well microplates at room temperature for two hours. Post-incubation, the plates were subjected to a magnetic process and washed with a specific buffer before being exposed to a biotinylated detection antibody for an additional hour at room temperature. Following a subsequent filtration and wash step, streptavidin-phycoerythrin was added for a 30-min incubation. After three final washes, the samples were prepared in reading buffer for analysis, with measurements taken in duplicate to ensure precision. The Luminex 200 system was employed for reading the plates, ensuring a minimum of 100 beads per cytokine for each sample.

Immunoassay with ELISA kit

Human IL-1β ELISA kit (NOVUS, VAL101), Human IL-4 ELISA kit (NOVUS, VAL123), Human IL-5 ELISA kit (NOVUS, VAL125), Human IL-6 ELISA kit (NOVUS, VAL102), Human IL-8 ELISA kit (NOVUS, VAL103), and Human GM-CSF ELISA kit (NOVUS, VAL124) were purchased. The standard curves were generated according to the kit manual.

Effects of serum matrix on LITs

To validate the effect of serum matrix on the six-plex LITs, the standard curves were generated by adopting a similar protocol to that used for single-plex LIT immunoassays, except that the mixed cytokines were dissolved in cytokines-depeleted human serum matrix (SANTA SCOTT, S-4SP1) instead of PBST. In spike recovery experiments, the mixed cytokine solutions were spiked into a healthy human serum matrix to achieve high, medium, and low concentrations (1000 pg/ml, 200 pg/ml, and 40 pg/ml). These samples were then measured, and the recovery rate was calculated as Observed concentration/Expected concentration ×100%.

Patient serum acquisition

Serum samples for the study were obtained from the Chinese PLA General Hospital, utilizing tubes without additives (BD, 367812) for collection. Post-collection, samples were allowed to clot at room temperature for 30 min before being centrifuged at 12,000 × g for 10 min. The clarified serum was then pipetted into cryovials and preserved at −80 °C for future analysis.

Ethics

Every experiment involving animals, human participants, or clinical samples have been carried out following a protocol approved by an ethical commission. Each participant gave informed written consent. This study has been approved by the ethics committee of Chinese PLA General Hospital. Approved No. of ethic committee: S2022-722-01.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.