A Multiplexed Digital Platform Enables Detection of Attomolar Protein Levels with Minimal Cross-reactivity

Abstract

Protein-based biomarkers are essential for disease diagnostics, yet their low abundance in biofluids often presents significant detection challenges for traditional enzyme-linked immunosorbent assay (ELISA) techniques. While various ultrasensitive methods such as digital ELISA have improved sensitivity, multiplex assays still suffer from considerable cross-reactivities that can compromise result accuracies. To address this challenge, we have developed barcoded Molecular On-bead Signal Amplification for Individual Counting (barcoded MOSAIC), a multiplexed digital ELISA technology that markedly reduces cross-reactivity by pairing barcoded detection antibodies with specific bead types. This approach enables the simultaneous detection of eight analytes from less than 9 microliters of blood, with sensitivities ranging from mid-picomolar to low-attomolar levels, and a collective dynamic range exceeding seven logs across multiple analytes within a single multiplex assay. Additionally, barcoded MOSAIC is compatible with standard immunoassay reagents and workflows, utilizing a rapid, automatable flow cytometric readout for quantification, which makes it a highly accessible benchtop platform that is readily adoptable by both research and clinical laboratories, setting the stage for future translation into point-of-care applications.

Graphical Abstract

Protein-based biomarkers hold great promise for disease diagnostics, therapeutic monitoring, and treatment,^1–3 yet their low abundance in biofluids often poses detection challenges for traditional enzyme-linked immunosorbent assay (ELISA) techniques. To address this analytical barrier, various ultrasensitive methodologies such as digital ELISA have been developed, achieving up to 1000-fold increases in sensitivity.^4–8 Among these methods, a notable example is the Single-Molecule Arrays (Simoa) technology, which isolates individual protein molecules within femtoliter-scale reaction chambers, facilitating signal generation at locally high concentrations for single molecule counting. While Simoa has markedly advanced biomarker detection, it often suffers from inefficient bead capture in a limited number of microwells,⁹ which reduces analytical sensitivity and limits the sampling of rare molecules. Building upon Simoa, several digital ELISA platforms have been developed, including Molecular On-bead Signal Amplification for Individual Counting (MOSAIC),¹⁰ dropcast single-molecule assays,¹¹ and droplet digital ELISA.⁷ These platforms have improved sampling efficiencies and analytical sensitivity tenfold by transforming single-molecule measurements into simplified assay formats through localizing non-diffusible signals on beads, thereby eliminating the need for microwells for signal compartmentalization.

Such advancements have enabled the detection of many rare molecules, thereby facilitating the analysis of complex multifactorial diseases, such as cancers, autoimmune diseases, and neurodegenerative diseases, which require the analysis of multiple biomarkers to effectively tailor therapeutic regimens.¹² Accordingly, the validation of biomarkers and their integration into multiplex immunoassay panels confer the attractive prospect of simultaneous measurements of multiple analytes from a single patient sample, significantly reducing assay costs and sample volume.^13–15 Notably, technologies like Simoa and MOSAIC have been adapted to simultaneously measure multiple low-abundance cytokines in biological fluids,^{9–10, 16–17} enhancing the potential for comprehensive disease analysis for improved therapeutic outcomes. Nevertheless, the increased complexity of multiplex assays introduces greater potential for cross-reactivity, which can lead to false signals between binding antibodies and off-target proteins^{10, 18–19} and compromise result accuracy.

To mitigate the higher risk of cross-reactivity inherent in multiplex assays due to the involvement of multiple antibodies, proximity-based detection approaches have been used to reduce cross-reactivity,^19–20 enabling the generation of detectable signals only when both the correct capture and detection antibodies —each conjugated with specific DNA sequences— specifically bind to the target protein in proximity. This intrinsic pairing mechanism ensures that signals are produced exclusively by correct antibody-DNA pairs, effectively minimizing cross-reactivity.^21–22 However, this approach entails multiple complex steps such as ligation, amplification, and sequencing, which can be mid-throughput²³ and costly, especially for high-depth applications. Additionally, methods employing spatial or temporal separation have been developed, ensuring that each capture antibody is incubated only with its corresponding detection antibody, thereby reducing cross-reactivity^24–26 and enabling the accurate quantification of two or more biomarkers simultaneously without additional sample dilutions. Despite these advancements, adapting such approaches for ultrasensitive multiplex assays with minimal cross-reactivity remains challenging. Our lab has previously addressed this challenge by either running singleplex assays or by developing a multiplex Simoa assay utilizing sequential protein capture to mitigate cross-reactivity and enable higher-order multiplexing.²⁷ However, this method is limited by sample loss and bead carryover during multiple sequential incubations.

Here, we demonstrate a near cross-reactivity-free digital ELISA technology that achieves attomolar sensitivities—an order of magnitude improvement over the gold standard digital ELISA methods. Building upon MOSAIC, we have employed barcoded detector antibodies with distinct DNA sequences and implemented a ‘parity check’ mechanism, where ‘parity’ refers to the precise match between the bead type (determined by its color, size, or intensity) and the DNA barcode on the detector antibody, while ‘check’ refers to the verification of this match. This approach effectively identifies and eliminates cross-reactive binding events by detecting mismatched bead-probe pairs. We also integrated multispectral probe colors using ratiometric combinations, enhancing multiplexing capabilities to measure a panel of five biomarkers with minimal cross-reactivity. Additionally, this approach has been expanded to quantify an eight-biomarker panel commonly used in cancer detection and monitoring, utilizing just under a quarter of a blood drop (approximately 9 μL).

RESULTS AND DISCUSSION

Development of Ultrasensitive Multiplex Digital ELISA Platform with Minimized Cross-reactivity.

We have developed an ultrasensitive multiplex digital ELISA platform that substantially reduces cross-reactivity, leveraging the foundations of MOSAIC technology. The previous iteration involved the formation of single immunocomplex sandwiches on beads, comprising capture antibody-coated beads, protein analytes, and biotinylated detector antibodies. These complexes were then targeted with streptavidin-conjugated DNA primer-template pairs, followed by rolling circle amplification (RCA) to generate extended DNA concatemers. Subsequent incorporation of fluorescently labeled DNA probes during RCA facilitated in situ hybridization, producing robust fluorescent signals from beads bound with target molecules—these were designated as the “on” bead population. Flow cytometric counting of “on” and “off” beads provided digital quantification of target molecules, quantified as “average molecules per bead” (AMB), measuring the average number of target molecules detected per bead. By localizing a non-diffusible signal to each captured target molecule on a bead, MOSAIC technology removes the need for traditional signal compartmentalization, which restricts the number of analyzable bead types due to physical constraints on compartment numbers. This multiplex MOSAIC assay expands multiplexing capabilities by enabling the use of more bead types; however, it is still limited by antibody cross-reactivities as the level of multiplexing increases. Cross-reactivity primarily arises from non-specific interactions, where detection antibodies bind to off-target analytes or beads, generating false-positive signals. This phenomenon is more pronounced in multiplex assays due to the presence of multiple antibody pairs and the increased potential for combinatorial interactions. However, it is less likely to occur in single-plex assays, where the absence of competing targets and detection antibodies reduces the likelihood of non-specific interactions.

In the improved MOSAIC assay, instead of using the same streptavidin-DNA conjugate and the corresponding fluorescent probe for a single-color readout, we directly conjugate each detector antibody to a distinct DNA template sequence (Figure 1). Each DNA template-barcoded detector antibody is paired with a capture bead, containing specific features such as color, intensity, or size, resulting in a distinct identifier for each analyte. In flow cytometry, these pairs can be differentiated using multiple detection channels.

Figure 1. Schematic illustrating the barcoded multiplex MOSAIC.

In this format, each detector antibody in a multiplex MOSAIC assay is conjugated to a distinct DNA template, which is then paired with a corresponding fluorescent dye-conjugated probe. Each analyte corresponds to a specific pair of: (1) capture bead color, fluorescence intensity, and/or size; and (2) fluorescent probe color. Consequently, only “correct” matched pairs of capture bead and probe signal are classified as “on” beads for each analyte. Any mismatched “wrong” pairs of capture bead and probe colors, indicative of cross-reactive binding events, are eliminated from the analysis.

Consequently, only “correct” matched pairs of capture beads and probe signals are classified as “on” beads for each analyte, with cross-reactive binding events, represented by mismatched pairs, being eliminated from analysis. Termed “barcoded” MOSAIC, this approach marks a significant advancement in reducing cross-reactivity. The sequences designed for this system can be integrated into existing sandwich ELISA configurations using simple copper-free click chemistry. Given the widespread availability of flow cytometers, our barcoded MOSAIC technology facilitates high-throughput, multiparametric single-molecule measurements and is adaptable to the dynamic landscape of molecular amplification methods.

Parity Matching of the Bead Type and Probe Color Greatly Reduced Cross-reactivity in Multiplex Assays.

To test the feasibility of pairing capture bead type and labeling probe color for the identification and elimination of mismatched cross-reactive binding events, we developed a proof-of-concept three-plex barcoded MOSAIC assay using three representative cytokines (Figure 2A): interleukins 8, 10, and 12p70 (IL-8, IL-10, and IL-12p70), which play important roles in regulating the immune response,^28–29 and are often present at low levels in many biological samples.^{10, 28} In this assay, each cytokine was assigned a distinct fluorescent dye-coded bead, and a detector antibody conjugated to a distinct DNA template sequence matching a specific fluorescently labeled DNA probe. To test whether barcoded multiplex MOSAIC assays can successfully measure multiple protein analytes with high sensitivities, we compared the three-plex calibration curves in the barcoded MOSAIC assay against the corresponding non-barcoded MOSAIC assay (Figure 2B). We observed comparable sensitivities (attomolar levels) across both assay formats. Specifically, the barcoded approach exhibited a 4-fold and 16-fold reduction in limit of detections (LODs) for IL-10 and IL-12p70, respectively, whereas IL-8 displayed a 5-fold increase in LOD compared to non-barcoded MOSAIC (Table 1). The observed variations in LODs between the barcoded and non-barcoded assays may not signify a definitive advantage in sensitivity for either method. Rather, these differences may reflect intrinsic disparities in antibody performance, which are hypothesized to arise from modifications associated with specific labeling reagents. The covalent attachment of azide-DNA to DBCO-modified antibodies is posited to alter binding kinetics and antigen affinity, while the DNA template could introduce steric hindrance, affecting epitope accessibility compared to smaller biotin molecules. Additionally, barcoded MOSAIC reduces cross-reactive signals arising from mismatched detection antibodies binding to off-target analytes, which primarily improves the accuracy of the assay rather than the sensitivity. While this refinement improves quantification by reducing false positives, it does not imply that the LOD for non-barcoded MOSAIC is inherently worse—rather, the barcoded system reduces misleading signals caused by cross-reactivity, yielding more accurate and reliable measurements.

Figure 2. Comparison of sensitivities between barcoded and non-barcoded MOSAIC assays.

(A) Schematic comparison of the barcoded to the non-barcoded multiplex MOSAIC assays. In the latter, all detector antibodies are biotinylated and labeled with the same streptavidin-DNA conjugate and fluorescent probe, thereby masking cross-reactive binding events. Representative plots show beads differentiated by a series of gates in different fluorescence channels, with the AMB for each bead type calculated based on probe-specific fluorescence intensities. Q750 and Q700 represent Quanterix Homebrew beads conjugated to 750 nm and 700 nm fluorescent dyes, respectively. SP denotes singleplex beads without an encoded dye. (B) Calibration curves for IL-8 (green), IL-10 (orange), and IL-12p70 (blue) in the non-barcoded multiplex MOSAIC assay (left) and the barcoded multiplex MOSAIC assay (right). Curves are fitted using four-parameter logistic (4PL) regression, highlighting the overall attainment of attomolar sensitivities across assays. Error bars represent the standard deviation of triplicate measurements.

Table 1. Comparison of the limit of detection (LOD) and lower limit of quantification (LLOQ) values for the three-plex barcoded MOSAIC assay and the corresponding non-barcoded MOSAIC assay.

LOD and LLOQ values were calculated as concentrations corresponding to three and ten standard deviations, respectively, above the background (established by running 3–6 blank samples through the assay). Reproducibility was evaluated by performing two independent calibration curves on separate days, which yielded consistent AMB values and LODs. The reported LOD values represent the [range] of averaged results from these independent measurements, reflecting day-to-day variability.

Analyte	Limit of Detection (aM) (3x)		Lower Limit of Quantification (aM) (10x)
	Barcoded MOSAIC	Non-barcoded MOSAIC	Barcoded MOSAIC	Non-barcoded MOSAIC
IL-10	5±2 [4 – 6]	20±5 [16 – 23]	49±38 [22 – 76]	79±10 [72 – 85]
IL-12p70	2±1 [0.8 – 4]	32±17 [20 – 44]	34±37 [8 – 60]	119±59 [77 – 161]
IL-8	546±188 [414 – 681]	108±79 [52 – 163]	3322±1726 [2102 – 4542]	373±270 [182 −563]

We next assessed whether the barcoded multiplex MOSAIC assay reduces cross-reactivity by performing protein dropout experiments, where increasing concentrations of a single interfering protein were measured to identify potential false positive “on” beads arising from cross-reactive binding events. In theory, any parity-mismatched color combinations were excluded from the analysis to enhance the accuracy of multiplex measurements in barcoded MOSAIC. We observed significant cross-reactivity for all three protein dropout assays in the non-barcoded multiplex MOSAIC assay, with observable false positive signals on off-target beads at protein concentrations above 8 fM (Figure 3A–i). We further assessed the specific binding and cross-reactivity for each analyte concentration by calculating the signal-to-background ratios (Figure S1 and Table S1), which we defined as the ratio of the AMB of a given sample concentration to the AMB of the blank sample, or background. Above 100 fM of IL12p70, for example, the false positive signal exceeds 28-fold and 42-fold above background for IL-8 and IL-10, respectively (Table S1). The non-barcoded multiplex MOSAIC, which utilizes a single-color readout through streptavidin-DNA conjugates targeting all biotinylated detectors, is prone to false positives primarily because proteins that bind cross-reactively to off-target beads may still be recognized by their specific detection antibodies, generating false signals. In contrast, the barcoded multiplex MOSAIC assay showed no detectable false signals arising on off-target beads across all three dropout assays, even at high protein concentrations, with the off-target bead signals remaining at the background levels (Figure 3A–ii). Thus, the inclusion of only correct capture bead-probe color pairs in barcoded multiplex MOSAIC enables the effective elimination of cross-reactive signals and thus ensures more accurate multiplex measurements.

Figure 3. Three-plex barcoded MOSAIC assay with markedly reduced cross-reactivity.

(A) Protein dropout curves depicting increasing concentrations of each individual target protein for the non-barcoded multiplex MOSAIC (i) and the barcoded multiplex MOSAIC (ii). (B) Recoveries of spiked recombinant proteins for the three-plex barcoded MOSAIC assay and corresponding non-barcoded MOSAIC assay in buffer (i) and human plasma at 16-fold dilution (ii), showcasing only the recoveries of IL-10 here; see detailed recoveries of other recombinant proteins summarized in Tables S2–3. Recoveries are calculated as the mean ± standard deviation from duplicate measurements and are defined as the ratio of the difference in interpolated concentrations post-spike to the baseline concentration—using the limit of detection (LOD) for buffer and the pre-spike plasma concentration at 16-fold dilution—divided by the indicated spiked protein concentration. The acceptable range of recoveries (70–130%) is highlighted in grey. Recovery rates higher than 100% can occur when the signal from the spiked sample exceeds the expected value, possibly due to matrix interference or non-specific binding, leading to overestimation of the analyte concentration.

To evaluate the measurement accuracy of the three-plex barcoded MOSAIC assay compared to that of the corresponding non-barcoded MOSAIC assay, we measured various spiked recombinant protein mixtures at different concentration levels of three analytes in both buffer and human plasma, where we compared recoveries according to the actual and measured protein concentrations. Initial evaluations in buffer established a baseline for cross-reactive performance without matrix interference. The barcoded MOSAIC assay consistently demonstrated higher accuracy and reduced cross-reactivity (Table S2). For instance, recovery rates for IL-10 in the barcoded assay ranged from 72% to 113%, while the non-barcoded assay showed extreme variations from 4% to 853% (Figure 3B–i). Notably, in conditions with low IL-10 and high concentrations of the other two proteins (IL-10, IL-12p70, IL-8 = 0.1, 40, 200 fM), the barcoded assay accurately recovered IL-10 at 105%, while the non-barcoded assay showed significant cross-reactivity, reaching a recovery of 853%. Subsequent tests in biological fluids reinforced these findings. Spike and recovery assays in plasma typically achieved recoveries within the acceptable range of 70−130% for both assays, except under conditions of low cytokine concentrations or significant concentration disparities exceeding three logarithmic scales (Figure 3B–ii and Table S3). With a low spiked concentration of IL-10 alongside comparable levels of IL-12p70 and IL-8 (0.2, 0.2, 4 fM respectively), the non-barcoded assay exhibited extreme variability, with recovery rates fluctuating from −28% to 28,400%. Negative recovery rates in the non-barcoded MOSAIC assay likely arise from background interference or cross-reactivity, where blank signals exceed those of spiked samples, leading to underestimation of analyte concentrations. In stark contrast, the barcoded assay displayed markedly stable recovery rates, spanning from 95% to 280%. With a spiked concentration of IL-10, IL-12p70, IL-8 at 0.2, 40, 180 fM, the barcoded assay demonstrated relatively consistent recoveries, while the non-barcoded assay showed poor performance, illustrating the barcoded MOSAIC assay’s substantial reduction in cross-reactivity and its ability to provide accurate measurements across varied concentration levels, even when lower concentrations of measured analyte coexist with much higher concentrations of other analytes. Although the barcoded system is highly effective in minimizing false positive signals from cross-reactivity, non-barcoded multiplex MOSAIC remains a useful approach in less complex multiplex assays with fewer analytes, particularly when carefully selected antibody pairs with low cross-reactivity are employed. However, as the degree of multiplexing increases, the likelihood of cross-reactivity between detection and capture antibodies rises, compromising assay accuracy.

Development of Higher-Order Multiplexed Barcoded MOSAIC assay with Minimal Cross-reactivity.

To develop a higher-order multiplexed panel in barcoded MOSAIC for N analytes, we require N spectrally distinct signatures, which translates to N distinct bead types —characterized by properties such as color, fluorescence intensity, and size— combined with N distinct DNA barcodes on detector antibodies. To expand beyond the number of available laser colors in conventional flow cytometry, instead of measuring the fluorescence intensity of a single-colored probe, we employ multispectral profiles resulting from ratiometric combinations of probes that label individual molecules derived from RCA. By differentiating DNA template sequences conjugated to antibodies using multispectral probes, we generate distinguishable optical barcodes, facilitating high-dimensional multiplexing via multispectral signatures. We screened various DNA sequences capable of hybridizing with different ratiometric combinations of probes (Figure S2). Utilizing two probe colors allowed us to generate four distinct non-overlapping signals. This strategy of employing ratiometric combinations of probes enabled a greater degree of multiplexing within a constrained spectral range, while retaining low to mid-attomolar sensitivities and minimal cross-reactivity for low-abundance cytokines (Figure S3 and S4).

Utilizing an expanded palette of ratiometric probe combinations, we conducted proof-of-concept experiments to quantify a panel of five cancer-relevant markers—IL-6,^30–32 IL-12p70,³⁰ HE4,^33–34 CA-125,³⁵ and ORF1 protein (ORF1p)^36–37—in plasma samples (Figure 4A). This approach, leveraging distinct bead and probe colors for each target, effectively minimizes false signals from cross-reactive binding, enabling accurate quantification of each analyte. Calibration curves for this five-plex assay indicated sufficient sensitivities to measure all analytes at concentrations well below their physiological levels, although the LOD for IL-12p70 (around 20 aM) approached the measured concentrations (Figure 4B). Notably, the LOD for LINE1 ORF1p, a recently identified biomarker for multiple cancers,^36–37 improved by two orders of magnitude compared to the previously reported LOD from a singleplex Simoa assay.^{7, 36} By decreasing the detector concentration to moderate the assay’s sensitivity for HE4, which is highly abundant in plasma in healthy conditions (hundreds of pM) and further elevated in diseases such as ovarian cancer, we extended the assay’s dynamic range to accommodate both low and high abundance markers. This demonstrates the barcoded MOSAIC platform’s ability to quantify biomarkers with varying concentrations, addressing the typical variability in dynamic range required for clinical samples. We applied the assay across various dilutions—8-, 12-, and 16-fold—in human plasma samples, observing acceptable recoveries in spike and recovery experiments and consistent dilution linearity (Table S5 and Figure S5), with minimal cross-reactivity across the target analyte concentration ranges (Figure S6). The exceptional sensitivity of barcoded MOSAIC enables the use of higher dilution factors to minimize matrix interference and potential false signals due to cross-reactivity. In conjunction with parity checking, these features collectively further enhance the assay’s resistance to cross-reactivity-related false signals. We further measured the five analytes in two healthy plasma samples, diluted eight-fold and requiring less than 25 μL across duplicates. Despite the low sample volume, the assay’s performance remained robust, with quantification of analytes (Figure 4C) comparable to those previously obtained using Simoa^{28, 36} and non-barcoded MOSAIC methods¹⁰. This multiplexing capability, coupled with minimal cross-reactivity and reduced sample volume requirements, underscores the assay’s potential as a valuable tool for accelerating biomarker signature discovery, particularly in applications with limited sample volumes like finger-prick blood.

Figure 4. Five-plex barcoded MOSAIC with minimal cross-reactivity.

(A) Schematic of multiplexing using barcoded MOSAIC. Beads coated with antibodies specific to different target analytes are differentiated by utilizing fluorescent dyes of varying wavelengths, intensities, and multiple bead sizes. Upon capture of the target analytes, single immunocomplex sandwiches are formed and labeled with detector antibodies that are barcoded with different DNA primer sequences (T14, 15, 20, 21, 25, Table S9), hybridized with corresponding fluorophore-labeled DNA probes (yielding specific ratios of ATTO565 to ATTO647 fluorescence of 1:0, 0:1, 4:1, 1:4, and approximately 5:1. The “approximately 5:1” ratio reflects the combined fluorescence contributions from ATTO590, ATTO565, and ATTO647, with a precise ratio of 3:2:1, which can appear as roughly 5:1 for ATTO565 to ATTO647. Table S9). RCA is carried out and the mixture of beads is analyzed by flow cytometry. Beads are differentiated by a series of gates in different fluorescence channels, and the AMB for each bead type is then determined from the intensities in the fluorescence channel corresponding to the probe color(s). (B) Calibration curves and LODs for five analytes. Reproducibility was evaluated by performing three independent calibration curves on separate days, which yielded consistent AMB values and LODs. The reported LOD values represent the [range] of averaged results from these independent measurements, reflecting day-to-day variability. The LODs and LLOQs in both units (fM and pg/mL) are summarized in Table S4. (C) Measured concentrations of five protein analytes in human plasma using a five-plex MOSAIC assay. Concentrations shown are the measured concentration values in the 8-fold diluted plasma samples. Assay LODs are denoted by the blue dashed lines.

To enhance the utility of our platform in cancer diagnostics, where high multiplexing is essential for delineating both sensitive and specific biomarker signatures, we expanded the assay to include three additional cytokine markers— IL-7,³⁸ IL-18,³⁹ and IFN-γ,⁴⁰—resulting in an eight-plex panel. This assay incorporates three primary markers (CA-125, HE4, and ORF1p) with minimal cross-reactivity using distinct bead types and probe signatures. The auxiliary cytokines were analyzed using distinct bead types, with shared probe colors selectively applied to enhance multiplexing efficiency, particularly for less cross-reactive pairs or secondary markers where relative quantification is sufficient (Figure 5A). The LODs for the three primary markers remained comparable to those measured from the five-plex assay, even with the inclusion of additional cytokines (Figure 5B). After confirming satisfactory recovery for each analyte (Table S7), we measured these eight proteins in a cohort of 10 plasma samples using less than 17 μL of plasma, inclusive of two replicates. The eight-plex barcoded MOSAIC assay successfully quantified endogenous proteins across a broad concentration spectrum, from picomolar to attomolar (Figure 5C), demonstrating consistency with the general concentration ranges of these markers in plasma, as observed in previous measurements.^{10–11, 28, 36} The strategic use of shared probe colors for auxiliary markers underscores the flexibility of the barcoded system to adapt to varying degrees of cross-reactivity, enabling efficient multiplexing while maintaining high specificity where necessary. Thus, the expanded multiplexing capability of the barcoded MOSAIC platform not only allows for accurate quantification of primary markers but also broadens the diagnostic potential by enabling the profiling of a more comprehensive spectrum of clinically relevant biomarkers.

Figure 5. Enhanced multiplexing with barcoded MOSAIC technology.

(A) Schematic representation of the eight-plex barcoded MOSAIC assay for the accurate measurement of key oncological biomarkers such as ORF1p, HE4, and CA-125, where each analyte is paired with specific bead types and probe colors to eliminate cross-reactivity. Additionally, auxiliary markers are measured using varied bead types that may share probe colors, thereby enhancing multiplexing capabilities while minimizing potential cross-reactive binding events among essential markers. (B) Calibration curves for each of the eight analytes, illustrating assay sensitivity and dynamic range. LODs in fM are summarized in the accompanying table. Detailed calibration curves can be found in Figure S7. (C) The measured concentrations of the eight protein analytes in human plasma, analyzed using this eight-plex MOSAIC assay. These concentrations are derived from measurements in 16-fold diluted plasma samples, with dashed blue lines indicating the LODs for each analyte. Error bars represent the standard deviation of replicate measurements (⩾2). LODs and LLOQs in both units (fM and pg/mL) for these analytes are provided in Table S6. The assay demonstrates reliable detection of ORF1p at sub-femtomolar concentrations. However, at the theoretical detection limit of 1 aM, the statistical variation due to Poisson noise becomes significant, as detecting such a small number of molecules introduces substantial variability. This inherent variability may partly contribute to the day-to-day fluctuations observed in Figure 4B.

CONCLUSIONS

Despite the discovery of numerous potential biomarkers, there are multiple barriers to biomarker validation, particularly for multiplex biomarker assays. One key technological barrier has been the lack of multiplex immunoassays that efficiently utilize samples, provide robust quantitative data without increasing background or cross-reactions, are rapidly customizable without relying on difficult-to-access reagents and hardware.⁴¹ The barcoded MOSAIC platform presented here addresses these issues by markedly reducing cross-reactivity and by being compatible with standard immunoassay reagents, workflows, and instrumentation, thus enabling straightforward implementation in clinical and laboratory settings using a flow cytometer. Moreover, unlike commercially available multiplex systems that require extensive validation for reagents due to potential cross-reactivity, barcoded MOSAIC supports a diverse array of detection affinity reagents, such as polyclonal antibodies that are more prone to cross-reactivity. Additionally, barcoded MOSAIC provides a cost-effective and high-throughput solution, with eight-plex measurements costing less than three dollars in consumables and procedures taking less than three hours —only thirty minutes of which involve hands-on time, along with 2.5 hours of incubation steps (Table S8). The current system is still limited by the fluorescence spectral overlap and the number of distinct detection probe colors that can be used simultaneously. Future enhancements will focus on applying algorithms to improve cluster delineation and broadening the spectrum of usable colors and ratios to extend multiplexing capabilities. The assay could also be adapted for DNA-based readouts such as qPCR or sequencing to achieve even higher levels of multiplexing. Although these higher-plex assays may show increased cross-reactivity and reduced sensitivity, implementing strategies like sequential target capture for spatial separation of detector antibodies can mitigate these effects, thereby enhancing the detection sensitivity and accuracy for critical analytes.

Although the assay achieves attomolar detection limits, this does not imply that the antibodies themselves exhibit attomolar affinities. Rather, theoretical models of digital ELISA demonstrate that antibodies with nanomolar to picomolar affinities, as measured by the dissociation constant (K_D), which typically ranges from 10⁻¹¹ to 10⁻⁹ M, can still achieve attomolar sensitivities.^42–43 The Kᴅ values reported in previous studies for antibodies used in highly sensitive immunoassays^44–45 align well with the performance of the current assay (Table S9), supporting the notion that antibodies with affinities in this range can robustly capture and quantify rare molecules. Furthermore, the surface avidity effect created by bead-immobilized antibodies strengthens target binding through multivalent interactions, helping to stabilize the immune complexes and prevent the loss of captured targets during washing steps, thereby ensuring the assay’s high sensitivity.

An additional application of the barcoded MOSAIC platform is its utility in streamlining the screening and optimization of antibody pairs for immunoassays. This platform enables simultaneous screening of multiple capture and detection antibody pairs in a single run, drastically reducing labor and resource expenditure. For instance, evaluating five antibodies as both capture and detection reagents would require 25 separate assays with a single-color readout. However, with the barcoded MOSAIC assay, this extensive process can now be accomplished in a single assay that utilizes 25 multispectral profiles, each pairing distinct beads with corresponding probe signatures. Barcoded MOSAIC not only accelerates the assay development process but also significantly cuts down on time and costs, making it a valuable tool for rapid assay optimization and validation.

In summary, the attomolar sensitivity, enhanced accuracy through reduced cross-reactivity, expanded multiplexing capabilities, streamlined workflow, and high-throughput flow cytometric readout of the barcoded MOSAIC technology hold vast potential for broad applications across various diseases. The platform also promises to enhance lab and clinic accessibility, potentially integrating into future point-of-care platforms.

METHODS

Materials.

All affinity reagents, recombinant proteins, and DNA oligonucleotides used in this work are listed in the Supporting Information (Tables S10–11). Buffers and paramagnetic beads were purchased from Quanterix Corporation and Bangs Laboratories. Custom DNA oligonucleotides were purchased from Integrated DNA Technologies.

Preparation of capture and labeling reagents.

Capture antibodies were buffer exchanged with Bead Conjugation Buffer (Quanterix) using a 50K Amicon Ultra-0.5 mL centrifugal filter (MilliporeSigma). After adding Bead Conjugation Buffer to antibody solution in the filter up to 500 μL, buffer exchange was carried out by centrifuging three times at 14,000 xg for five minutes, with addition of 450 μL Bead Conjugation Buffer between centrifugation cycles. The buffer-exchanged antibody was recovered by inverting the filter into a fresh tube, centrifuging at 1000 xg for two minutes, rinsing the filter with 50 μL Bead Conjugation Buffer, and centrifuging one more time at 1000 xg for two minutes. The concentration of the buffer-exchanged antibody was then measured using a NanoDrop spectrophotometer. For each bead type, indicated starting number of beads (Table S12) were washed three times with 300 μL Bead Wash Buffer (Quanterix) and two times with 300 μL Bead Conjugation Buffer (Quanterix) before resuspending in cold Bead Conjugation Buffer. Bead number and conjugation conditions for each analyte are shown in Table S12. A 1 mg vial of 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC) (Thermo Fisher Scientific) was dissolved in 100 μL cold Bead Conjugation Buffer, and the desired volume was added to the beads. The beads were shaken for 30 minutes at either room temperature or 4 °C. After EDC activation of the carboxyl groups on the beads, the beads were washed once with 300 μL cold Bead Conjugation Buffer before resuspension in the buffer-exchanged antibody solution. Antibody conjugation was carried out by shaking the beads for two hours at either room temperature or 4°C, followed by washing twice with 300 μL Bead Wash Buffer. The antibody-coupled beads were then blocked for thirty minutes at room temperature with shaking in 300 μL Bead Blocking Buffer (Quanterix). After washing once each with 300 μL Bead Wash Buffer and Bead Diluent (Quanterix), the beads were resuspended in 200 uL Bead Diluent, counted with a Beckman Coulter Z1 Particle Counter, and stored at 4 °C. Detector antibodies in non-barcoded MOSAIC assay were obtained in biotinylated form as previously described.

Preparation of antibody-DNA conjugates.

Preparation of circularized padlock DNA template-primer hybrid. For each detector antibody, a 5’ azide-modified primer was annealed to a distinct DNA template by heating a solution of 30 μM primer and 30.3 μM template in NEBNext Quick Ligation Buffer (New England Biolabs) at 95 °C for two minutes and allowing to cool to room temperature over 90 minutes. Ligation was then performed with addition of T4 DNA ligase and incubation at room temperature for two hours. The ligation reaction was buffer exchanged into phosphate buffered saline (PBS) with 1 mM EDTA using a 7K MWCO Zeba spin desalting column (Thermo Fisher Scientific). Preparation of DBCO-modified antibody. For conjugation, the detector antibody was either reconstituted into PBS from lyophilized form or buffer exchanged into PBS using a 50K Amicon Ultra-0.5 mL centrifugal filter, incubated with a 20-fold molar excess of dibenzocyclooctyne-PEG4-N-hydroxysuccinimidyl ester (DBCO-PEG4-NHS, MilliporeSigma) for 30 minutes at room temperature, and purified with a 50K Amicon Ultra-0.5 mL centrifugal filter in PBS with 1 mM EDTA. The template-primer hybrid and DBCO-modified antibody can be prepared in either order. However, it is suggested that the DNA hybrid be prepared first, as the cooling (1.5 hours) and ligation (2 hours) steps provide a window for preparing the DBCO-modified antibodies. Conjugation of azide-modified DNA to DBCO-modified antibody via copper free click chemistry. A two-fold molar excess of the ligated primer-template was then added to the DBCO-modified antibody and incubated overnight at 4°C. The conjugate was stored in aliquots at −80 °C in PBS with 5 mM EDTA, 0.1% BSA, and 0.02% sodium azide. (Figure S8 provides a detailed schematic illustrating the workflow and key steps of the process.)

Barcoded MOSAIC assays.

MOSAIC assays were performed in a 96-well plate (Greiner Bio-One, 655096), with antibody-coated beads and detector antibodies diluted to the desired concentrations in Homebrew Sample Diluent (Quanterix). Assay conditions for each analyte are listed in Table S13. Sample volumes of 100 μL were used, with 10 μL of antibody-coated beads. The plate was sealed and shaken for one hour for target capture, followed by washing with System Wash Buffer 1 (Quanterix) using a BioTek 405 TS Microplate Washer. 100 μL detector antibody-DNA in Sample Diluent with 0.02 mg/mL heparin was added to the beads after target capture and washing steps. (Heparin, which mimics the polyanionic structure of nucleic acids, was included to mitigate non-specific interactions between the DNA-conjugated antibodies and non-target proteins.) The mixture was then incubated for 10 minutes. The samples were then washed with System Wash Buffer 1 for twelve cycles, transferred to a fresh 96-well plate, and washed an additional time with 180 μL System Wash Buffer 1 before being resuspended in 50 μL of the RCA reaction mixture. The RCA mixture consisted of 0.5 mM deoxynucleotide mix (New England Biolabs), 0.33 U/uL phi29 DNA polymerase, 0.2 mg/mL bovine serum albumin (BSA, Invitrogen), 1 nM fluorescently labeled DNA probe (Integrated DNA Technologies), and 0.1% Tween-20 in 50 mM Tris-HCl (pH 7.5), 10 mM (NH₄)₂SO₄, and 10 mM MgCl₂. Dye labeled DNA probes (Table S13) were used for single target and multiplex MOSAIC assays, respectively. Upon addition of the RCA mixture to each sample, the plate was shaken for 1.5 hours at 37 °C, followed by addition of 150 μL PBS with 0.1% Tween-20 and 5 mM EDTA to stop the reaction. Samples were washed one time with 200 μL of the same PBS-Tween-EDTA buffer and resuspended in 100 μL of the buffer with added 0.1% BSA. Samples were measured using a NovoCyte Flow Cytometer (Agilent Technologies) equipped with three lasers, in either tube or plate sampling mode. Bleach and buffer wells were included between different samples to minimize potential sample carryover. Multiplex MOSAIC assays were carried out following the same protocol as for the singleplex MOSAIC assays, with different fluorescent dye-encoded beads combined in the same sample. The incubation times are determined based on established protocols as well as prior theoretical modeling.^{10, 42}

Plasma samples were diluted in 20% Sample Diluent (Quanterix) with 1% Triton X-100 in 80% StartingBlock^™ Blocking Buffer (Thermo Fisher Scientific), with protease inhibitor (Halt^™ Protease Inhibitor Cocktail, Thermo Fisher Scientific). Plasma samples were obtained from BioIVT and the Mass General Brigham Biobank. All human samples were de-identified, and experiments were performed under Institutional Review Board approval by Mass General Brigham. All plasma samples were centrifuged at 2000 xg for 10 minutes at 4 °C before diluting for measurements.

Calibration curves were employed to ensure accurate quantification in both singleplex and multiplex MOSAIC assays. Although concentrations can theoretically be determined through direct counting of molecules captured on beads, variations in the number of beads counted per sample introduce discrepancies in these counts. Additionally, practical factors such as bead occupancy variability, non-linear signal responses at extreme concentrations, and amplification inconsistencies necessitate calibration. These curves also account for non-specific binding, bead aggregation, surface coating heterogeneity, and matrix effects in biological samples, providing reliable and reproducible quantification that direct counting alone cannot achieve. Additionally, due to interactions between analytes and background noise as the number of analytes increases, separate calibration and curve fitting were performed for each specific multiplex panel. This ensures that assay-specific conditions, which may not be fully captured by a single calibration curve, are properly addressed to maintain accuracy across different multiplex configurations.

Non-barcoded MOSAIC assays.

All non-barcoded assays were performed as previously described.¹⁰ Assay conditions for each analyte are listed in Table S13. The same antibody-coated beads used in the barcoded MOSAIC assays were used for the non-barcoded MOSAIC assays, with the same detector antibody concentrations.

Data analysis.

Flow cytometry data were first analyzed with FlowJo^™ Software (Becton, Dickinson and Company); beads were identified using gates on forward scatter, side scatter, and bead fluorescence. Single beads were additionally gated using forward scatter. The probe fluorescence intensities for each bead population were analyzed as previously described¹⁰ or using FlowJo^™ Software. The fraction of “on” beads was then converted to an average number of analyte molecules bound per bead using Poisson statistics, and mapped to concentration using a four-parameter logistic calibration curve. Calibration curves were fitted using a four-parameter logistic (4PL) regression model, a standard choice for ELISA assays due to its capacity to handle sigmoidal curves. Since most curves in this study did not reach full saturation, the distinction between 4PL and 5PL models, which address symmetric versus asymmetric sigmoidal behavior, was not significant. The 4PL model was chosen for its compatibility with the assay and aligns with prior work on similar immunoassay platforms.^{10–11, 27–28, 36} The LOD and LLOQ were calculated as three and ten standard deviations above the background, respectively. The background refers to the baseline signal measured in the absence of the analyte, established by running multiple (≥3) blank samples through the assay. The standard deviation of the background was calculated from these multiple blank measurements, with a correction factor c₄(n) applied for unbiased estimation, as described previously in the literature.⁴⁶

Supplementary Material

Supporting Information. The Supporting Information is available free of charge and includes additional experimental data, including calibration curves and assay validation, and detailed information on materials and assay conditions.

Data and Materials Availability.

All data are available in the main text or the Supporting Information. Raw flow cytometric data are available at the Open Science Framework (OSF): https://osf.io/ef2ny/?view_only=84ed8a62ef404bc7adb2a534f65d256d.