CRISPR‐Enhanced Hydrogel Microparticles for Multiplexed Detection of Nucleic Acids

Abstract CRISPR/Cas systems offer a powerful sensing mechanism to transduce sequence‐specific information into amplified analytical signals. However, performing multiplexed CRISPR/Cas assays remains challenging and often requires complex approaches for multiplexed assays. Here, a hydrogel‐based CRISPR/Cas12 system termed CLAMP (Cas‐Loaded Annotated Micro‐Particles) is described. The approach compartmentalizes the CRISPR/Cas reaction in spatially‐encoded hydrogel microparticles (HMPs). Each HMP is identifiable by its face code and becomes fluorescent when target DNA is present. The assay is further streamlined by capturing HMPs inside a microfluidic device; the captured particles are then automatically recognized by a machine‐learning algorithm. The CLAMP assay is fast, highly sensitive (attomolar detection limits with preamplification), and capable of multiplexing in a single‐pot assay. As a proof‐of‐concept clinical application, CLAMP is applied to detect nucleic acid targets of human papillomavirus in cervical brushing samples.


Introduction
Nucleic acids (NAs) are essential biomarkers for disease detection and monitoring, commonly detected through a quantitative polymerase chain reaction (qPCR). Recent advances in gene Scheme 1. Overview of the multiplexed CLAMP assay. i) Spatially-coded hydrogel microparticles (HMPs) are loaded with Cas12a/gRNA complexes, with each code denoting a different nucleic acid (NA) target. These particles are mixed with a sample, and Cas12a/gRNA complexes are activated when target NAs are recognized. ii) HMPs are introduced into a fluidic device along with fluorescent-quencher (F-Q) reporters. Activated HMPs become fluorescent with F-Q reporter cleavage. iii) A machine-learning (ML) algorithm recognizes HMP codes and quantifies particles' fluorescent intensities, generating assay results on multiple NA targets.
unmet need when expanding CRISPR/Cas's reach in biosensing. Technical hurdles, particularly for single-pot, multi-target detection, continue to challenge progress. Unlike qPCR wherein different F-Q probes can be turned on in a target-specific manner, CRISPR/Cas assays lose sequence specificity due to non-selective probe cleavage by Cas nucleases. [2] This issue has been partly addressed by screening Cas effectors with differential preferences towards cleavage motifs, [13] but the availability of orthogonal Cas effectors limits multiplexing to four targets. An appealing alternative is to compartmentalize Cas reactions and run parallel assays. Droplet fluidics have been adopted for such purposes, [14] demonstrating high-throughput NA detection. The droplet approach, however, can be process-intensive and requires sophisticated tools to generate, coalesce, and detect individual droplets. Another fluidic approach used multiple fluidic chambers to contain target-specific CRISPR/Cas12 complexes; [15] this configuration is essentially a microwell-type assay, requiring different CRISPR/Cas12 loading per well.
Here, we report an HMP-based CRISPR/Cas system, termed CLAMP (Cas-Loaded Annotated Micro-Particles), for multiplexed NA detection (Scheme 1). The assay integrated several technical advances. First, we immobilized Cas proteins in spatiallyencoded HMPs, turning each HMP into a discrete vessel for NA detection. Each HMP type, marked with a spatial code, was specific to a unique NA target; upon recognizing its intended NA target, the Cas12a/gRNA complexes cleaved F-Q probes, turning on fluorescence signals inside HMPs. Second, we designed a microfluidic device to facilitate signal detection. The fluidic device made it easy to locate individual HMPs and, more importantly, allowed us to surround HMPs with immiscible oil to improve the signal-to-background contrast. Third, we trained a machine-learning algorithm to identify the HMP type and quantify its fluorescence intensity. Combining these features into an assay system offered the advantages of i) high sensitivity (attomolar detection limits), ii) simple assay processes with no additional labeling and washing steps needed, and iii) multiplexed NA detection in a single pot. To explore its potential clinical use, we applied CLAMP to detect high-risk human papillomavirus (HPV) DNA targets within cervical brushing samples.

Characterization of HMPs Functionalized with Cas12a/gRNA
We prepared HMPs specific to different NA targets (see Experimental Section for details). We first fabricated patterned-HMPs using mask photolithography (Figure 1a). A prepolymer solution containing acrylic acid, polyethylene glycol diacrylate (PEGDA), and a photoinitiator was exposed to UV through a photomask, which photocured the solution into disk-shaped particles engraved with identity codes. The optimal concentration of crosslinking monomers (PEGDA) was found to be 5% (volume fraction), which produced structurally robust, highly permeable HMPs. [24] To further validate that HMPs had large enough pores, we incubated particles with fluorescent-dextran whose molecular weight (150 kDa) was similar to that of Cas12a (145 kDa). The partition coefficient (K) was then estimated by taking the fluorescent intensity ratio between HMPs and the bulk solution ( Figure S1, Supporting Information). The observed K was 0.89, indicating HMP's high permeability.
Cas12a proteins were coupled to carboxylic acid groups in HMPs through carbodiimide chemistry that was previously used to immobilize proteins in HMPs. [25] Such immobilization was necessary to maintain high Cas12a concentration inside HMPs and prevent inter-particle crosstalk; without immobilization, proteins would diffuse out of HMPs, leading to signal loss (Figure S2, Supporting Information). Cas12a-HMPs were then incubated with gRNA, making the particles (Cas12a/gRNA-HMPs) ready to recognize target NAs. We also prepared the F-Q reporter probe by conjugating a fluorophore and a quencher at each end of a single-stranded DNA. This probe generated the fluorescence signal when cleaved by activated Cas12a/gRNA complexes (Figure 1b). Preparation and characterization of Cas12a/gRNA HMPs. a) Schematic of the HMP synthesis containing Cas12a/gRNA complexes. HMPs were produced by photocuring a precursor solution. UV light was exposed through a photomask to pattern codes in HMPs. Cas12a proteins were immobilized in HMPs by reacting carboxyl groups in the hydrogel with amine groups in Cas12a. Coded HMPs were rendered target-specific by loading appropriate gRNA. b) Generation of the analytical signal. When the NA target forms a ternary structure with the Cas12a/gRNA complex, the HMP becomes active in cleaving F-Q reporters, increasing fluorescence. c) Cas12a concentration ([Cas12a]) was varied when reacting with HMPs for loading. The resulting HMPs were used for the CLAMP assay, and the fluorescent intensity was measured. To determine the optimal condition for Cas12a-HMP preparation, we varied Cas12a concentration [Cas12a] and its conjugation time with HMPs. The prepared Cas12a-HMPs were then used to detect synthetic NA targets. Both target NAs and fluorescent reporters are diffused into HMPs to react with Cas12a/gRNA complexes. [26] The CLAMP fluorescence signal was higher when we used Cas12-HMPs prepared under higher Cas12a concentration (Figure 1c), but the protein aggregated past the maximal [Cas12a]. Similarly, we found that the optimal reaction time was about 2 h for Cas12a conjugation with HMPs ( Figure S3, Supporting Information). The protein binding was stable with negligible leaching of Cas12 proteins from HMPs; compared to the strong CLAMP signal from HMPs, those from supernatants were close to the background during 24-h incubation in a buffer solution ( Figure S4, Supporting Information).
We further characterized Cas12a proteins inside HMPs. The protein amount was first assessed by incubating Cas12a-HMPs with fluorophore-labeled gRNA (Figure 1d). The resulting HMP's fluorescence was then referenced to the gRNA calibration curve ( Figure S5, Supporting Information). About 5.2 fmol of Cas12a was immobilized per HMP (or 4.9 μM per HMP), assuming that the gRNA binds to the Cas12a in a 1:1 ratio. [27] We next evaluated Cas12a-HMP's catalytic efficiency for trans-cleavage. We activated Cas12a/gRNA complexes in HMPs by incubating particles with target NAs. Catalytic Cas12a-HMPs were then mixed with the F-Q reporter probe, and the fluorescence signal resulting from the probe cleavage was monitored (see Experimental Section for details). Figure 1e shows the initial cleavage rates measured at varying reporter concentrations. Applying the Michaelis-Menten model to the results, we could estimate the enzyme turnover rate (k cat = 0.15 s −1 ) and the Michaelis-Menten constant (K M = 1.73 μM). The catalytic efficiency (k cat /K M ) of HMP-bound Cas12a was 1.0 × 10 5 M −1 s −1 . This value was in the same order of magnitude as the catalytic efficiency (3.3 × 10 5 M −1 s −1 ) of free Cas12a in solution ( Figure S6, Supporting Information) [28] but was slightly lower, presumably due to the structural confinement coming from Cas12a immobilization. Figure 2. CLAMP assay optimization. a) Optimization of the reaction time between Cas12a/gRNA HMPs and NA targets. The reaction time was varied, and the resulting CLAMP signal was measured. The signal gradually reached a plateau with the estimated half-time (T 1/2 ) of 31 min. The reaction time was set to 60 min, which would yield 86% of the final signal. Error bars indicate the standard deviation from ten HMPs. Scale bar, 50 μm. b) Optimization of F-Q probe concentration. The CLAMP assay was carried out while varying the F-Q probe concentration ([F-Q probe]). The signal-to-noise ratio (SNR) was estimated as the ratio of fluorescent intensity between targeted and control (no target NA) HMPs. The highest SNR was achieved at [F-Q probe] = 2 μM. Data are displayed as mean ± sd from ten replicates. c) Serially diluted NA targets were measured by CLAMP (green) and a conventional solutionbased Cas12a assay (black). CLAMP had a lower detection limit (3 pM) than the solution-based assay (3.6 pM). Error bars are standard deviations from ten HMPs (CLAMP) and three replicates (solution-based assay). d) Assay validation. The fluorescent signal was observed only when all reaction components were included. Data are displayed as mean ± SD from three replicates.
Cas12/gRNA-HMPs and target NAs (Figure 2a) and measured the fluorescence signal. The optimal reaction time was set to 1 hour, at which point the analytical signal started to plateau. The reporter probe's concentration was another factor optimized because it directly influenced the background signal; we selected the probe concentration of 2 μM, which maximized the signal-to-noise ratio ( Figure 2b). Containing DNA probes inside individual HMPs was also crucial to retaining high sensitivity and minimizing inter-particle crosstalk. We could achieve such isolation by immersing HMPs in immiscible oil.
Applying the optimal CLAMP protocol, we analyzed serially diluted NA targets ( Figure 2c). The analytical detection limit was about 3 pM, and the dynamic range spanned two orders of magnitude. The CLAMP showed higher sensitivity than the solutionbased Cas12a assay. Note that Cas12a concentration was higher in HMPs (4.9 μM) than in solution (0.64 μM), which boosted the trans-cleavage of F-Q reporters. We further validated that the CLAMP signal was generated only when all assay components were present (Figure 2d).

CLAMP Detection System for Multiplexing
We engineered the detection system for the multiplexed CLAMP assay. To distinguish target-specific HMPs, we imprinted particles with spatial codes comprising three dots and an L-shaped indicator (Figure 3a). The dots were placed among eight possible positions at the HMP's circular edge; the indicator was at the HMP's center to set the readout direction. This rule can generate up to 168 binary codes. For quadruple multiplexing, we chose four equidistant codes; the Hamming distance was the maximal (=4) between codes.
We next designed a microfluidic chip to arrange individual HMPs (Figure 3b). The chip contained an array of HMP-trapping structures. Each trap had a small rear opening to allow fluid to flow through. Once a particle was trapped, the opening was blocked to increase the flow resistance; other HMPs would then bypass the occupied trap. [29,30] We first confirmed the working principle through numerical fluidic simulation (Figure 3c and Figure S7, Supporting Information) and laid out staggered trap columns across the flow direction to minimize channel clogging. The prepared chip demonstrated > 90% trapping efficiency. Importantly, it prevented inter-particle interaction (i.e., signal bleeding from the direct particle contact), simplified the buffer exchange with oil, and expedited HMP imaging.
We adopted an ML approach to analyze individual HMPs automatically. We selected Mask R-CNN as an ML engine; [31] this neural network can classify objects in an image while simultaneously generating a high-quality segmentation for each instance ( Figure S8, Supporting Information). Our strategy was to use i) the classification to identify HMP types and ii) the segmentation to define a mask for fluorescent detection (Figure 3d). We trained the Mask R-CNN model with brightfield images of each HMP type (110 images per type). The trained model differentiated 4 different HMP types (Figure 3e) in bright field images, achieving high accuracy (0.979) and F 1 score (0.978) in a separate validation. After HMP identification, the code and the segmentation were transferred to fluorescence images, and the fluorescence intensity of each particle was extracted (see Movie S1, Supporting Information).

CLAMP Assay to Detect Human Papillomavirus DNA
To explore CLAMP's potential for clinical application, we configured CLAMP for multiplexed detection of high-risk HPV, a lead-ing cause of cervical cancer. [32] Cervical cancer incurs high incidence and mortality rates in many resource-limited countries which often lack early screening infrastructures. We reasoned that the CLAMP system could serve as a portable, rapid diagnostic system to effectively triage suspicious or high-risk cases.
We chose high-risk HPV subtypes (HPV16, HPV18) as viral targets. Since our goal was to detect HPV genes in human cells, we also added GAPDH as a target for positive control of cellularity. We first designed a set of gRNAs for each target and compared their specificity in the presence of off-targets ( Figure S9, Supporting Information). From this comparison, we selected the gRNA set allowing for multiplexed NA detection in a single pot (Figure 4a). We also added an upstream NA amplification to enhance the overall detection sensitivity. The recombinase polymerase amplification (RPA) method was used because its reaction temperature was close to that of the Cas12a trans-cleavage assay, allowing for the isothermal reaction for the entire assay. We validated that the optimized RPA primers amplified only targeted genes ( Figure S10, Supporting Information). By introducing RPA, we could lower the detection limit to an attomolar range (2 aM; Figure 4b). In comparison, the detection limits were about 145 aM with the RPA alone and about 10 aM by the conventional qPCR ( Figure S11, Supporting Information). . CLAMP assay for multiplexed HPV detection. a) gRNAs were selected for HPV16, HPV18, and GAPDH targets. The selected gRNAs were highly specific only to their intended NA target. The signal was from solution-based Cas12a assays without RPA. Error bars indicate the standard deviation from three replicates. b) The CLAMP assay analyzed serially diluted synthetic HPV16 DNA samples. Incorporating the RPA step lowered the detection limit to 2 aM (1.2 copy μL −1 ). Data are displayed as mean ± SD from three replicates. c) Schematic of the CLAMP cellular assay. Genomic DNA was extracted and processed by RPA, followed by multiplexed CLAMP detection. d) Example of CLAMP multiplexing. Three types of coded HMPs were rendered to detect GAPDH (yellow circle), HPV16 (pink), and HPV18 (cyan), respectively. HMPs with the fourth code were used as a negative control (white circle). Micrographs show HMPs detecting genomic DNA from HPV16 + cells. e) A varying number of cervical cancer cells were analyzed by CLAMP targeting HPV16. With CaSki (HPV16 + ) cells, the detection limit was about 10 cells mL −1 , whereas the CLAMP signal was negligible with C33A (HPV16 − ) cells. Error bars indicate the standard deviation from ten HMPs. f) The multiplexed CLAMP analyzed four different cell lines (CaSki, HeLa, SiHa, C33A) for HPV16, HPV18, and GAPDH. The CLAMP results matched the known genotype of these cells.
We further challenged CLAMP to differentiate a singlebase mismatch. As a model system, we used a non-modified HPV16 DNA target (matched) and a set of modified HPV16 DNAs with a single-base mismatch at varying positions (Figure S12a, Supporting Information). The CLAMP assay differentiated the matched target from single-base mismatched ones ( Figure S12b, Supporting Information) when the mismatch was within the seeding region (≈8 bp away from the PAM). The results agreed with the previous report on Cas12a's sequence specificity. [33] The PAM-adjacent region (5-8 bp away from the PAM) forms multiple contacts with several domains of a Cas protein (WED, REC1, and RuvC). A perfect www.advancedsciencenews.com www.advancedscience.com base-pair match within this region is essential for Cas12a's catalytic activity.
We next applied the CLAMP assay to detect HPV DNA in cervical cancer cells. We extracted genomic DNA (gDNA) from cell lysates, amplified NA targets via RPA, and performed the CLAMP assay (Figure 4c). The assay procedure was close to the clinical HPV test wherein samples (e.g., Pap smear, cervical brushings) are assessed for HPV DNA targets inside cells. To enable multiplexing, we assigned each coded HMP type a different target (Figure 4d). Testing CaSki (HPV16 + ) for HPV16, we observed that the cellular detection limit was about 10 cells mL −1 (Figure 4e). In contrast, the signal was negligible with HPV16 − C33A cells. We further expanded the multiplexed CLAMP assay to a panel of cervical cancer cell lines. All samples passed the quality check ascertained by the GAPDH signal, and the CLAMP results matched with known HPV16 and HPV genotypes of cells (Figure 4f).

Pilot HPV Detection in Clinical Samples
We then applied the CLAMP assay to detect HPV targets in clinical samples. We obtained excess cervical brushing specimens collected during routine gynecologic evaluation. Samples were selected by a pathologist based on high-risk HPV status as determined by routine HPV testing (Cobas; Roche, Indianapolis, IN); the test results were blinded until the CLAMP assay was completed. For each given sample, we extracted its gDNA, amplified NA targets (HPV16, HPV18, GAPDH), and performed the multiplexed CLAMP detection (see Experimental Section for details). Figure 5a summarizes the CLAMP-HPV results. All samples showed a high enough GAPDH signal to pass quality control ( Figure S13, Supporting Information), and no statistical difference was observed between HPV-positive and negative samples (p = 0.74, unpaired two-sided t-test). Importantly, CLAMP results agreed with conventional pathology-derived HPV diagnoses. We set the threshold for marker positivity as 3• , where is the standard deviation of a blank sample. All marker-positive samples showed signals higher than the threshold, whereas marker-negative samples had signals lower than the threshold (Figure 5b). The CLAMP results further matched those by qPCR (Figure 5c), which confirmed the CLAMP's capability for quantification.

Conclusion
Cas systems are increasingly adopted for biosensing; [34][35][36] they offer new ways to amplify analytical signals with precision down to single nucleotide variants. In this work, we encapsulated the Cas assay into biocompatible, coded HMPs. The resulting CLAMP assay brought in unique advantages: i) Casfunctionalized HMPs could be specific to different NA targets (down to single-base mismatch) while being identified by unique spatial codes, ii) HMPs locally maintained a high dose of the detection agent (Cas12a/gRNA) to boost the analytical signal, and iii) with Cas12a/gRNA immobilized inside HMPs, each particle can detect its designated NA target. These advantages enabled CLAMP to detect different NA targets (multiplexing) in a singlepot assay scheme. Furthermore, immobilizing Cas12a/gRNA in HMPs allowed us to concentrate the complex without compromising its catalytic activity, which rendered the CLAMP assay more sensitive than solution-based Cas assays. These aspects (multiplexing, high sensitivity, scalability) distinguish CLAMPS from other CRISPR/Cas assays (see Table S1, Supporting Information for comparison). The CLAMP assay successfully detected high-risk HPV targets in the ensuing pilot study, demonstrating a potential for on-site HPV screening.
Activated Cas proteins cleave reporter probes indiscriminately, making it difficult to perform multiplexed detection. Our solution to this formidable challenge was to transform HMPs into an independent vessel for the Cas reaction; we exploited HMPs' superior hydrophilicity and compatibility with microfabrication. The fluidic chip for HMP capture was another critical component. Spatially isolating HMPs facilitated their detection via imaging. More importantly, we could confine cleaved fluorescent probes inside HMPs to prevent signal bleed-through among HMPs. This top-down approach simplified the assay development. It required only a single pair of a Cas effector and a fluorescent reporter, contrasting with other multiplexing approaches (e.g., multiple Cas effectors and fluorescent reporters). [13,14] The CLAMP method is also economically scalable to accommodate many targets. HMPs can be mass-produced, and the spatial codes are readily expandable to incorporate more binary bits. A major challenge, however, would be designing primer sets for multiplexed amplification. As shown in Figure S9, Supporting Information, primers designed for singleplex assays may fail to work in one-pot assays due to off-target amplification, necessitating the process of iterative primer design and screening. To our best knowledge, up to nine targets were amplified in a single pot via multiplexed RPA. [15] We expect to further develop CLAMP for use in point-of-care diagnostics. First, a compact system could be built for onsite operations. Two functions are required in the CLAMP assay: i) maintaining a constant temperature for the isothermal CLAMP reactions and ii) taking photos of HMPs. A smartphone could be a powerful base platform for such operations. A phone can be programmed to control a simple heater, and, more importantly, it can be coupled with simple optics to enable high-resolution and sensitive imaging, effectively replacing a bulky microscope. [37][38][39][40][41] Indeed, using a smartphone camera-based system, we could identify HMPs with their codes and measure the fluorescent signals ( Figure S14, Supporting Information). Second, we should explore lyophilizing CLAMP reagents to simplify transport logistics and extend shelf life. Previous reports demonstrated that lyophilized Cas proteins and HMPs could be resuscitated after storage in ambient conditions. [42,43] We need a similar test on HMPs preconjugated with Cas12/gRNA complexes. Finally, we should consider incorporating DNA extraction into the CLAMP workflow, which will minimize the risk of errors from sample contamination and user intervention. Heat-based cell lysis is promising [44,45] as it can be executed with an existing CLAMP device. With these improvements, we believe the CLAMP assay would be a portable and field-usable platform that facilitates NA profiling for clinical diagnoses.

Experimental Section
Materials: The following items were purchased for NA detection: oligonucleotides (Bioneer, Korea), gBlock gene fragments (Integrated All samples showed high GAPDH signals, passing the quality control. The HPV16 and HPV18 status by CLAMP matched with clinical diagnostics by pathology. b) HPV16 and HPV18 levels were normalized against GAPDH. The normalized HPV16 and HPV18 expressions were significantly higher in clinically-defined marker-positive samples (p < 0.001; unpaired two-sided t-test). c) The CLAMP results were compared with those by qPCR. Two methods showed a high correlation. r, Pearson's coefficient.  Table S2, Supporting Information.
HMP Permeability Test: Two types of HMPs with different crosslinking degrees were prepared by varying the amount of PEGDA in the precursor solution: highly porous HMPs (5% v/v PEGDA 700) and nonporous HMPs (95% v/v PEGDA 700). To test the HMP permeability, HMPs were contained inside a microfluidic chamber and injected with 150 kDa FITC-dextran (2.5 mg mL −1 ) at the flow rate of 50 μL h −1 . Fluorescent images were taken before and after 2 h of FITC-dextran injection. The partition coefficient (K = I gel /I sol ) was calculated by dividing the backgroundsubtracted fluorescent intensity of hydrogel (I gel ) by that of bulk solution (I sol ).
Fabrication of Microfluidic Chips: The device was laid out using design software (AutoCAD, USA) and had the master mold (SU-8) fabricated by a company (Microfit, Korea). The height of the fluidic channel was 85 μm. To make the fluidic device, a PDMS mixture (10:1 base to curing agent ratio) was prepared, poured over the master mold, and cured the polymer (60°C, overnight). The cured PDMS slab was then peeled off, cut into the desired size, and punched holes (0.8 mm biopsy punch; Miltex, USA) to define fluidic ports. As a fluidic bottom substrate, a glass slide was separately prepared; PDMS was coated on the slide and partially cured (60°C, 2 h). The fluidic device was assembled by attaching the PDMS slab to the glass slide. The final device was heat-cured (60°C, overnight) to ensure leaktight bonding.
Flow Simulation in Microfluidic Chips: The flow inside the fluidic chip was simulated using Comsol multiphysics 5.6 software (microfluidics module, COMSOL). The two designs were compared, one with an intertrap distance of 200 μm and the other with 280 μm. Other design parameters were identical. The fluid was assumed to be water (density, 1000 kg m −3 ; dynamic viscosity, 1 mPa·s) at ambient conditions (20°C, 101.3 kPa), and the flow speed at the inlet was set to 10 mm s −1 . A finer mesh and a stationary solver with no-slip boundary conditions were used. To visualize the flow, 100 streamlines evenly spaced across the fluidic channel were arranged. The HMP trapping was also simulated by introducing a disk (diameter, 140 μm; density 1120 kg m −3 ) inside the fluidic chip.
Estimating Cas12a Amounts in HMPs: A calibration curve was first generated for the FAM-labeled gRNA (FAM-gRNA). A known amount of FAM-gRNA was injected into the microfluidic device and measured the resulting fluorescent intensity ( Figure S5, Supporting Information). The height of the fluidic device was similar to HMP's height to minimize the risk of potential artifacts caused by a height difference. A varying dose of FAM-gRNA was injected into the microfluidic device and measured the resulting fluorescent intensity ( Figure S5, Supporting Information). Cas12a-HMPs reacted with the excess amount of FAM-gRNA (37°C, 30 min) and triplewashed the HMPs with 1× NEBT. Cas12a/FAM-gRNA HMPs were then injected into the fluidic device and their fluorescent intensities were measured. The amount of FMA-gRNA in an HMP was estimated from the calibration curve, assuming that gRNA would bind to Cas12a at a 1:1 ratio. [27] Solution-Based Cas12a Assay: In optimizing probe sequences and reaction conditions, the solution-based Cas12a assay was used. A reaction mixture was prepared by combining Cas12a (640 nM), target-specific gRNA (160 nM), F-Q probe (2 μM), RNase inhibitor (0.8 U μL −1 ), and target NA in 1× NEBuffer 2.1. The reaction was monitored by measuring fluorescence intensities at 37°C for 1 h on ViiA 7 Real-Time PCR System (ThermoFisher Scientific). Note that the fluorescence intensity was corrected by subtracting the background signal measured without target NA.
Generation of Calibration Curves: A solution-based assay using F-Q probes was performed at varying concentrations. The target NA concentration was set to 10 nM in a total volume of 20 μL. After the Cas12a reaction (37°C, 12 h), the fluorescence intensity of fully cleaved probes (ViiA 7 Real-Time PCR System) was measured to obtain a calibration curve for solution-based assays ( Figure S15, Supporting Information). The fully cleaved mixture was also injected into the microfluidic device and the resulting fluorescent intensity was measured, generating a calibration curve for the microfluidic device environment.
Cas12a Reaction in Solution: The reaction mixture was prepared as in the solution-based assay but the F-Q probe concentration was varied. The target NA concentration was set to 0.5 nM in a total volume of 20 μL. The Cas12a's trans-cleavage activity was monitored by measuring the fluorescent intensity for 10 min at every 30 s (37°C, ViiA 7 Real-Time PCR System). The initial reaction velocity (v) was then obtained by converting the fluorescent intensity to F-Q probe concentration ([S]). The molar reaction velocities were plotted against F-Q probe concentrations and fitted to the Michaelis-Menten model, v = E t ·k cat ·[S]/(K M + [S]), where E t is the concentration of catalytic sites (0.5 nM), k cat is the enzyme turnover rate, and K M is the Michaelis-Menten constant. Prism 9.3 (GraphPad, USA) was used for the fitting.
Cas12a Reaction in HMPs: Cas12a/gRNA HMPs were prepared and activated by adding 0.5 nM target NA. The activated HMPs were mixed with an F-Q probe solution, and the mixture was injected into the microfluidic device. The HMPs' fluorescent intensity for 10 min (37°C) was monitored and then the initial reaction velocity (v) was estimated. The F-Q calibration was used to convert the fluorescent intensity to the probe concentration ([S]). The data were analyzed to estimate k cat and K M .
CLAMP Assay: An RPA mix was prepared by combining 240 nM of target-specific forward and reverse primers, 0.8 U μL −1 RNase inhibitor, and target NA. Upon the addition of 14 mM MgOAc, the reaction proceeded at 37°C for 20 min. The amplified target NAs were then mixed with ≈50 Cas12a/gRNA-HMPs in 1× NEBuffer 2.1 containing the F-Q probe (2 μM) and RNase inhibitor (0.8 U μL −1 ). After incubation (1 h, 37°C), HMPs were introduced to the microfluidic device using manual pipettes or disposable eye droppers. After capturing individual HMPs onchip traps, the fluidic chamber was filled with HFE-7500 oil. After an additional incubation period (30 min, 37°C), HMPs were imaged using EVOS M7000 (ThermoFisher Scientific) and the image was analyzed with the customized ML algorithm. For a given NA target, the fluorescent signal was measured from about 15 target-specific HMPs and the mean value was used as an analytical metric. The CLAMP signal was corrected by subtracting the background signal measured without target NA.
ML Algorithm for Imaging Analyses: For each coded HMP type, 110 HMP images were used for training and additional 20 HMP images for validation. Data were augmented via image rotation and flipping. The Supervisely software (Supervisely OU, Estonia) was used to annotate HMPs