Amplifying mutational profiling of extracellular vesicle mRNA with SCOPE

Abstract

Sequencing of messenger RNA (mRNA) found in extracellular vesicles (EVs) in liquid biopsies can provide clinical information such as somatic mutations, resistance profiles, and tumor recurrence. Despite this, EV mRNA remains underused due to its low abundance in liquid biopsies, and large sample volumes or specialized techniques for analysis are required. Here, we introduce Self-amplified and CRISPR-aided Operation to Profile EVs (SCOPE), a platform for EV mRNA detection. SCOPE leverages CRISPR-mediated recognition of target RNA using Cas13 to initiate replication and signal amplification, achieving a sub-attomolar detection limit while maintaining single-nucleotide resolution. As a proof-of-concept, we design probes for key mutations in KRAS, BRAF, EGFR, and IDH1 genes, optimize protocols for single-pot assays, and implement an automated device for multi-sample detection. We validate SCOPE’s ability to detect early-stage lung cancer in animal models, monitor tumor mutational burden in colorectal cancer patients, and stratify glioblastoma patients. SCOPE can expedite readouts, augmenting the clinical use of EVs in precision oncology.

INTRODUCTION

Liquid biopsies present opportunities in precision oncology by providing dynamic molecular information of the entire tumor through minimally invasive and repeatable tests^1‑3. As such, both researchers and clinicians use liquid biopsies to monitor tumor evolution and heterogeneity⁴. Accessing real-time insights can impact patient care through early lesion detection, minimal residual disease tracking, personalized therapeutic decision-making informed by resistance profiles, and tumor recurrence monitoring^5,6.

Extracellular vesicles (EVs) are analytical targets within the liquid biopsy arena^7‑9. These small particles (<1 μm in diameter) carry diverse molecular cargo, including nucleic acids, proteins, and metabolites, thus serving as cellular surrogates^10,11. Detecting EV mRNA can generate rich and actionable clinical information. EV mRNA can reflect somatic driver mutations (e.g., KRAS^G12D, BRAF^V600E) that are crucial for tumor initiation and growth that inform treatments¹². Additionally, EVs rarely contain the nuclear proteins that confer drug resistance, but they harbor corresponding mRNA counterparts to inform resistance status^13‑15. Since EV mRNAs are enclosed within vesicles, they are shielded from nucleases in biofluids, allowing for the isolation of intact, high-quality nucleic acids¹⁶. These collective advantages position EVs as a promising source of nucleic acids, complementing the benefits of circulating tumor DNA.

However, EVs’ clinical potential remains underused, primarily due to technical drawbacks. The majority of EV RNA is non-coding, and mRNA copy numbers in EV samples can be very low. For instance, even abundant mRNA species, such as GAPDH, are only detected at a frequency of one copy per 10⁴ – 10⁶ EVs, compared to microRNAs at one copy per 10² EVs¹⁷. This disparity has led most proof-of-concept assay development to conveniently focus on microRNA detection^18‑20. Moreover, tumor-derived EVs constitute a minor fraction (<5%) of total circulating EVs²¹. The scarcity of EV mRNAs, combined with the low abundance of tumor-derived EVs, necessitates large sample volumes (over 2 mL of plasma) and sophisticated resources (e.g., droplet-digital PCR and next-generation sequencing). This diminishes the competitive advantages of EV tests and complicates their integration into routine preclinical and clinical assays.

We develop a streamlined, rapid EV-mRNA test, drawing inspiration from the clustered regularly interspaced short palindromic repeats (CRISPR) method. CRISPR technologies are increasingly adopted in molecular diagnostics for their sequence-specific nuclease activity^22,23. CRISPR-associated (Cas) proteins become active endonucleases upon recognizing target nucleic acids²⁴. This property has been exploited to amplify signals through the promiscuous cleavage of reporter probes (e.g., single-stranded DNAs tagged with a fluorescent dye and quencher pair). However, applying CRISPR assays to EV mRNA proves challenging due to the low abundance of targets. Separate pre-amplification is usually needed to replicate mRNA targets and improve assay kinetics²³. Replication errors and biases inherent in such amplifications can propagate^25,26, leading to confounding results. To address this fundamental challenge, we repurpose the Cas activity, where the high-precision Cas system both recognizes target mRNA and then triggers its replication in situ, bypassing the need for pre-amplification and associated errors. This robust and efficient strategy could ensure high analytical sensitivity while maintaining sequence specificity, a crucial capability for identifying scarce mRNA targets.

Here, we present the SCOPE (Self-amplified and CRISPR-aided Operation to Profile EVs) platform, an integrated assay for accurate EV mRNA detection and monitoring. SCOPE leverages the synergy between the Cas13a machinery and signaling templates. Cas13a first recognizes its target RNA, and then both RNA targets and fluorescent signals are amplified. The high selectivity afforded by Cas13a allows SCOPE to resolve single-nucleotide polymorphisms. The assay further achieves a sensitivity down to a sub-attomolar detection limit, owing to its built-in dual amplification mechanism and further optimizations. We applied SCOPE for various applications such as detecting early-stage lung cancer in animal models, tracking cancer mutational burden in patients with colorectal cancer undergoing standard of care, and identifying mutations in glioblastoma multiforme (GBM) for patient stratification. The developed assay can promote preclinical and clinical opportunities to dissect cancer growth processes, identify the emergence of resistance, and gauge tumor response to treatments, including clinical trials. Ultimately, SCOPE offers to expedite standard clinical and drug trial decisions and enhance the use of EVs in the liquid biopsy space.

RESULTS

Overall SCOPE flow

The SCOPE strategy combines CRISPR recognition with RNA replication within a single assay format. The assay is initiated by introducing samples to an all-in-one SCOPE mixture containing the complex of Cas13a and CRISPR RNA (Cas13a/crRNA), T7 polymerase, signal template, and deoxyribonucleotide triphosphate (Fig. 1a). The signal template is crucial for coupling Cas13a/crRNA and T7 polymerase reactions. It is designed explicitly as an RNA-DNA hybrid comprising i) a fluorescent RNA segment cleavable by Cas13a/crRNA, ii) a fluorescent quencher, and iii) a DNA template for target RNA transcription by T7 polymerase (Supplementary Fig. 1).

Figure 1. Self-amplified and CRISPR-aided operation to profile extracellular vesicles (SCOPE).

(a) SCOPE mechanism. The assay couples Cas13a/crRNA and polymerase reactions through a signal template (bottom). The signal template contains an RNA segment with a quenched fluorescent dye and a DNA template for T7 polymerase. The fluorescent signal is initially quenched (F, fluorescent dye; Q, quencher). (b) Assay flow. (i) Cas13a/crRNA first recognizes the EV RNA target, which activates its ribonuclease function. (ii) The activated enzyme cuts the RNA segment in the signal template, generating a fluorescent signal and exposing the DNA template. (iii) T7 RNA polymerase binds to the promoter region of the DNA template and synthesizes target RNA replicas. (iv) The amplified target replicas bind to free Cas13a/crRNA, enforcing the overall assay process. Note that the initial CRISPR-mediated target recognition is necessary to unlock RNA replication and signal generation. (c) SCOPE workflow for onsite diagnostics. EVs are isolated from clinical samples and lysed. EV lysates are then placed in surface-treated tubes for RNA extraction (10 min). Subsequently, the SCOPE reaction is performed in a compact, portable device (30 min). The assay produces molecular information within our hour, enabling same-day clinical decisions.

The major steps of the SCOPE reaction unfold as follows (Fig. 1b, Supplementary Fig. 2). (i) Cas13a/crRNA recognizes and binds to RNA targets. (ii) The complex becomes an active ribonuclease and promiscuously degrades the RNA segment in the signal template. This trans-cleavage turns on fluorescent signals by releasing dye molecules from quenchers and exposes the DNA template complementary to the target RNA. (iii) T7 RNA polymerase attaches to the promotor region of the template and replicates RNA targets. (iv) Unreacted Cas13a/crRNA further recognizes RNA target replicas, reinforcing the overall reaction rather robustly.

The SCOPE mechanism offers significant technical advantages for RNA detection. First, it capitalizes on the high selectivity of Cas13a/crRNA complexes^23,27. The crRNA design incorporates a strategically placed synthetic mismatch, which empowers Cas13a to discriminate between RNA sequences, even down to single nucleotide variations²⁷. In SCOPE, this highly specific Cas13a acts as a powerful filter – Cas13a activates target amplification only in the presence of a bona fide target RNA. This strategy enriches the desired target RNA signal within a complex background of other RNAs. Second, SCOPE achieves high sensitivity (low detection limit) through dual amplifications: activated Cas enzymes keep cleaving RNA segments to boost fluorescent signals, and RNA polymerases make multiple copies of the target RNA. Third, the reaction is fast and carried out in a single tube at a constant temperature (40 °C). These features render SCOPE simple and readily applicable across routine laboratory settings (e.g., academia, hospitals, primary care clinics, industry).

These advantages inspired us to streamline the workflow of detecting RNA targets from EV lysates. Specifically, SCOPE simplified RNA extraction and detection steps for clinical EV analysis (Fig. 1c). In this procedure, RNA is swiftly (10 min) extracted from EV lysates in a polymer-coated tube. Subsequently, isothermal SCOPE reactions occur in that same tube, producing analytical results within 30 min. Starting from sample loading, the entire process is completed within 1 hour, facilitating same-day preclinical and clinical molecular diagnostics. Moreover, the assay uses standard laboratory equipment (e.g., size exclusion column for EV isolation, thermal cycler) and only requires small volumes of samples (e.g., EV isolates from <100 μL of plasma).

SCOPE instrumentation

We implemented compact modules to streamline the assay process and enhance end-user friendliness. The first module was a polymer-coated tube for nucleic acid extraction (Fig. 2a). Specifically, we prepared PCR tubes coated with a charge-shifting polymer, poly[(2-dimethylamino)ethyl acrylate] (pDMAEA), through initiated chemical vapor deposition (see Methods for details; Supplementary Fig. 3a). Upon addition of an aqueous sample, the polymer inside the tube dissolved and became positively charged, allowing for electrostatic interactions between cationic polymers and negatively charged nucleic acids. This reaction resulted in the formation of polyplexes, which could be collected via centrifugation (5 min, 3000 g). Subsequently, the SCOPE mixture was added to the precipitates, and the temperature was raised to 40 °C. Under this condition, the polymer underwent self-catalyzed hydrolysis of the ester groups located in its side chains. This process induced negative charges on the polymer, resulting in the disassembly of polyplexes and the release of nucleic acids. The SCOPE reaction then continued within the same tube. We observed no significant difference (P > 0.05; unpaired two-sided t-test) in SCOPE signals when using pDMAEA-coated tubes compared to standard PCR tubes (Supplementary Fig. 3b). In addition, self-catalyzed hydrolysis of pDMAEA was confirmed through nuclear magnetic resonance spectroscopy (Supplementary Fig. 4).

Figure 2. SCOPE device engineering.

(a) Single-tube system for SCOPE assay. We used PCR tubes coated with poly[(2-dimethylamino)ethyl acrylate] (pDMAEA) polymer. With EV lysates added, the dissolved polymer was positively charged initially, forming polyplexes with negatively charged nucleic acids. Following brief centrifugation to collect polyplexes, we then added SCOPE reagents. The polymer underwent charge-shifting hydrolysis to release nucleic acids, and the SCOPE reaction continued in the same tube. (b) Comparison of RNA extraction yield between pDMAEA tubes and standard silica-based columns. The pDMAEA method demonstrated similar extraction yields as the column devices, with an average extraction yield of 96% across different input RNA amounts (3, 30, 60, and 100 ng). Extracting nucleic acids with pDMAEA tubes was fast (10 min) and compatible with SCOPE reactions. Data are shown as mean ± s.d. from technical triplicates. (c) Integrated device for onsite SCOPE tests. This compact device accommodated 16 samples in a tray-like sample holder. The tray was wrapped with a flexible heater and docked to the mainframe that regulated the heater. Fluorescent signals were detected by an optical module (top right). The system was connected to a tablet via wireless communication. A user-friendly graphical interface (lower right) in a tablet provided intuitive windows for system operation and data analysis. (d) Upon completion of the reaction, the optical module quantified fluorescent signals emitted from the samples. A single module linearly scanned the sample tray, simplifying the system design. (e) The SCOPE device was used to process samples containing different amounts of synthetic CD63 RNA. The SCOPE results were highly consistent for a given RNA concentration, regardless of the sample location in the heating block. The coefficient of variation was <1.5%. The bar represents a mean fluorescent signal from four different sites in the heating block. Color codes matched the sample location with RNA concentration. a.u., arbitrary unit.

Figure 2b compares the RNA extraction efficiency between pDMAEA tubes and standard silica-based columns (see Methods)²⁸. The pDMAEA approach achieved similar extraction yields as the column devices, with an average extraction yield of 96% across different input RNA amounts (3, 30, 60, and 100 ng). Using pDMAEA tubes, however, streamlined the overall assay workflow. It facilitated rapid nucleic acid extraction (10 min) and enabled the execution of SCOPE reactions within a single closed-tube format, eliminating the need for sample transfer in between steps.

The second assay module was a compact device for parallel SCOPE measurements (Fig. 2c, Supplementary Fig. 5). It integrated a tray-type heating block, a fluorescent-optical detector, and a line scanner. The heating block could hold up to 16 PCR tubes and docked to the main frame for an electrical connection. The system then generated pre-programmed temperature profiles (e.g., isothermal, thermal cycling). Once the reaction was complete, the optical detector measured fluorescent signals (Fig. 2d). The detector comprised two independent fluorescent excitation-detection headers, with each header performing independent measurements. Two identical headers could be installed for redundant one-color measurements, or two different modules could be utilized for two distinct fluorescent dyes. The optical detector executed a linear motion to scan all 16 samples. System operation was controlled through a graphical user interface on an external terminal device.

The developed SCOPE system demonstrated high uniformity and reproducibility in parallel detection. The sample holder heated all 16 samples to 40 °C, with well-to-well temperature variations of less than 0.5 °C (Supplementary Fig. 5). We also observed highly uniform SCOPE signals for samples containing the same amount of mRNA, regardless of their location in the heating block (Fig. 2e).

Validating SCOPE concept

We systematically examined SCOPE assay kinetics through analytical modeling (see Supplementary Note for details). SCOPE couples two distinct catalytic reactions through the signal template (Fig. 3a): i) target-bound Cas13a/crRNA turns on the fluorescent signal by cleaving the RNA segment in the signal template and ii) RNA polymerase replicates target mRNAs using the DNA sequence in the signal template. When executed separately, each type of catalytic activity generated the end product (i.e., fluorescent signal or mRNA replica) that increased linearly with time (Fig. 3b, left); these observations matched the zeroth-order nature of Cas13a and polymerase reactions in our assay conditions (Supplementary Note)^29,30. In contrast, coupling these two processes approximated a first-order rate kinetics (Supplementary Note), speeding up the overall reaction. Indeed, SCOPE signals rose exponentially and reached a plateau within 30 min (Fig. 3b, right).

Figure 3. SCOPE assay kinetics.

(a) SCOPE integrates two enzymatic reactions via the signal template: i) Cas13a/crRNA generates fluorescence by degrading the RNA segment in the signal template, and ii) T7 polymerase replicates RNA targets. SCOPE reactions commence upon initial recognition of mRNA targets by Cas13a/crRNA. (b) When performed independently, Cas13a/crRNA and T7 reactions exhibited a linear increase in end product over time (left panels). Coupling these two reactions enabled SCOPE to emulate first-order rate kinetics. The analytical signals amplified exponentially, reaching a steady state within 30 min (right panels). Data are shown as mean ± s.d. from technical triplicates. (c) Assay validation. Signal intensity was maximized when all SCOPE assay components were present (red bar). Removing T7 polymerase diminished the signal due to the absence of target RNA amplification (blue bar). Assay conditions that hindered the initial mRNA recognition produced negligible signals (grey bars). The initial target RNA (KRAS^G12D) concentration was fixed at 1 nM. Data are shown as mean ± s.d. from technical triplicates. a.u., arbitrary unit.

Notably, off-target RNA amplification was suppressed. For RNA amplification to proceed, SCOPE first required Cas13a/crRNA to recognize its initial RNA target. The activated Cas enzyme then degraded the RNA segment in the signal template, which triggered downstream polymerase activities (Supplementary Fig. 6). Conversely, the polymerase failed to replicate RNA targets when signal templates had an intact RNA segment (Supplementary Fig. 7); the overall loop structure presumably hindered the polymerase from docking the promoter region of the signal template.

We further optimized the SCOPE template design and reaction conditions to maximize its signal intensity (Supplementary Fig. 8). The optimal assay was then validated by controlling the reactant composition (Fig. 3c). The analytical signal was highest when all assay components (e.g., target RNA, Cas13a, crRNA, T7 RNA polymerase) were present. The signal intensity decreased when RNA polymerase was absent; therefore, no additional targets were synthesized. The signals were negligible in those cases wherein initial target recognition by Cas13a/crRNA was missing.

SCOPE assay characterization

SCOPE achieved high sensitivity through its dual amplification scheme (fluorescent signal and mRNA target). In RNA-titration experiments, SCOPE’s limit of detection (LOD) reached sub-attomolar ranges, which was >10⁴-fold lower than RT-PCR’s LOD (Fig. 4a for KRAS^G12D; Supplementary Fig. 9 for CD63). Moreover, SCOPE displayed a broader dynamic range than RT-PCR. We observed a similar enhancement in sensitivity when analyzing EVs as mRNA sources (Fig. 4b); the LOD by SCOPE was about 10⁵ EV/mL, which was >10³-fold lower than the LOD by RT-PCR.

Figure 4. SCOPE assay characterization.

(a) SCOPE (30 min) and conventional RT-PCR (80 min) were employed to analyze samples containing varying concentrations of synthetic KRAS^G12D RNA. SCOPE exhibited a detection limit (LOD) of 8.5 copies/ μL (14.2 zM), outperforming RT-PCR (LOD of 1.1 × 10¹⁰ copies/ μL, 18.6 pM). Data are displayed as mean ± s.d. from technical triplicates. Some error bars are too small to be visible. The net intensity ( ΔIntensity) was obtained by removing background fluorescent signals (no-target controls). a.u., arbitrary unit. (b) EV samples were prepared from KP1.9 cell culture and analyzed for KRAS^G12D mRNA. SCOPE’s LOD was approximately 10⁵ EVs/mL, whereas RT-PCR’s LOD was about 10⁸ EVs/mL. In addition, SCOPE displayed a broader dynamic range than RT-PCR. Data are shown as mean ± s.d. from technical triplicates. (c) The specificity of SCOPE probes targeting KRAS wild type (WT) and its mutated subtypes (G12C, G12D, G12S, G12V) was evaluated. SCOPE displayed high signal contrast between on-target and off-target samples. At a target concentration of 1 nM, the contrast ratio exceeded 30. The heatmap shows mean values from technical triplicate measurements. (d) A mixture of KRAS^WT and KRAS^G12D synthetic RNA samples was prepared, varying the variant allele fractions (VAFs). SCOPE detected KRAS^G12D down to 0.01% VAF, with a signal significantly different from the background (p = 0.011; unpaired, two-sided t-test). The bars represent mean values from technical triplicate measurements.

SCOPE also exhibited superb sequence-specificity, allowing for precise differentiation of point mutations. We chose KRAS, the most frequently mutated oncogene, as a representative target and prepared SCOPE mRNA probes for wild-type KRAS (KRAS^WT) and its mutated subtypes, KRAS^G12C, KRAS^G12D, KRAS^G12S, and KRAS^G12V (see Supplementary Table 1). SCOPE consistently demonstrated high signal-to-background ratios between on-target and off-target samples (Fig. 4c, Supplementary Fig. 10); for 1 nM target concentrations, the ratio was >30 across different KRAS targets. SCOPE’s high selectivity enabled reliable analyses of samples with low variant allele fractions (VAFs). For instance, we could detect KRAS mutant alleles down to 0.01% VAF (Fig. 4d), a performance on par with digital PCR and sequencing^31,32. In contrast, RT-PCR assays displayed poor specificity due to off-target crosstalk, even when a commercial KRAS kit was used (Supplementary Fig. 11).

Early detection of lung cancer in an animal model

We hypothesized that SCOPE’s high sensitivity could facilitate EV-based early cancer detection. To test the hypothesis, we employed a genetically engineered mouse model of lung adenocarcinoma (KP model), characterized by Cre-induced Kras^G12D+/− p53^−/− mutation³³. This autochthonous model allowed us to regulate tumor initiation and collect blood samples at different stages of tumor development (Fig. 5a), an approach that effectively replicated the challenge of detecting EVs originating from early lesions.

Figure 5. Early cancer detection with SCOPE.

(a) Experimental design. A genetically engineered mouse model was used to mimic the development of non-small cell lung cancer. The animal had a Cre-activatable Kras^LSL-G12D/+; Trp53^flox/flox genetic background (KP model). Tumor growth was initiated through intranasal administration of adenoviruses expressing Cre recombinase (Ad-Cre). Serial blood samples were collected from the animals before tumor induction and at weekly intervals after that. (b) In vitro validation of SCOPE for KRAS genotyping. Cell lines with distinct KRAS genetic profiles were used: H2228 (WT/WT), KP1.9 (G12D/WT), and A549 (G12S/G12S). EVs were isolated from cell culture media. SCOPE KRAS assays on these EV samples yielded results consistent with the cellular KRAS status. Data are presented as mean ± s.d. from biological triplicates. a.u., arbitrary unit. (c) Tumor development in the Cre-treated mice was corroborated via immunohistochemistry on lung tissue specimens. Lesions positive for Kras^G12D protein were detectable one week after tumor induction, and the lesions enlarged in subsequent weeks. Staining was performed two different samples, and representative images are shown. (d) SCOPE analyses on EV Kras^G12D mRNA during tumor development. Serial blood samples were collected over eight weeks from the same individual mice. In the non-cancer cohort (n = 3), the SCOPE signal remained at background values throughout the duration. In the tumor-bearing cohort (n = 10), the SCOPE signal increased from background values one week after tumor induction and reached saturation approximately three weeks after the onset of tumor growth. The heatmap shows mean values from technical triplicate measurements. (e) Estimation of Kras^G12D mRNA copy numbers. SCOPE results in (d) were converted based on the titration curve (Fig. 4a). In the tumor-bearing cohort, the copy number was below the baseline (< 1) initially and then significantly increased (p = 0.042, paired two-sided t-test) one week after tumor induction. Each data point is a mean value from technical triplicate measurements.

We first validated whether EV mRNA profiling can infer cellular genotypes. For this validation, we used cell lines with different KRAS genetic backgrounds, specifically H2228 (KRAS^WT/WT), A549 (KRAS^G12S/G12S), and a well-characterized cell line derived from the KP mouse model (KP1.9; Kras^G12D/WT)³⁴. EVs were collected from cell culture media and subjected to the SCOPE KRAS mRNA assay (see Methods). We observed that EV profiling results consistently matched cellular KRAS status (Fig. 5b). EVs from KP1.9 cells exhibited high SCOPE signals for Kras^WT and Kras^G12D, reflecting the cell line’s Kras heterozygosity. In contrast, EVs from homozygous cell lines were exclusively positive for one of the KRAS targets.

We next applied SCOPE to analyze EVs in mouse plasma samples. We prepared two cohorts using KP mice with identical genetic backgrounds (Kras^G12D+/−). The tumor-bearing group received an intranasal inoculation of Cre adenovirus to induce oncogene expression and initiate tumor growth, whereas the non-cancer control group did not undergo the procedure. Immunohistochemistry performed on lung tissue specimens confirmed tumor growth in the Cre-treated mice (Fig. 5c). Following oncogene induction, multifocal disease develops that heterogeneously progresses from atypical adenomatous hyperplasias and small adenomas into larger adenomas and eventually invasive and disseminated adenocarcinoma³⁵. Notably, Kras^G12D protein expression in lung epithelium was observed within one week after Cre induction, followed by tumor formation in the subsequent weeks (Fig. 5c).

For both cancer (n = 10) and control (n = 3) cohorts, we collected longitudinal blood samples and analyzed plasma EVs for Kras^G12D mRNA; such serial EV profiling was feasible because SCOPE required only 40 μL of plasma samples. For tumor-bearing animals, we also obtained computerized tomography (CT) scans on a weekly basis. Figure 5d and Supplementary Fig. 12 display Kras^G12D SCOPE signals tracked over eight weeks. In the non-cancer mice, the signal remained at background values throughout the duration. Conversely, in tumor-bearing mice, the signal increased from the background once oncogene expression was induced and reached saturation at about five weeks of tumor growth. CT analysis performed on the same animals demonstrated tumor growth over time, although the lesions became detectable approximately two weeks after tumor induction (Supplementary Fig. 13).

We further estimated mRNA copy numbers (Fig. 5e, Supplementary Fig. 14) by referencing the SCOPE results to the titration curve (Fig. 4a). In cases without tumors, we observed a consistent baseline (copy number <1). However, the copy number became significantly different from the baseline as early as one week after oncogene induction (p = 0.042, unpaired t-test). These findings support SCOPE’s potential to detect early tumors, minimal residual disease, and/or recurrences, especially in cases involving known mutations.

Monitoring colorectal cancer patients during clinical care

RAS is frequently mutated in colorectal cancer (CRC), with about 40% of CRC cases showing these mutations, primarily in codon 12 (65%)^13,36. Assessing and monitoring KRAS mutations remains crucial for prognosis evaluation and treatment decisions. SCOPE could streamline this process by enabling minimally invasive, accessible, and highly accurate KRAS profiling. To test this approach, we applied SCOPE to analyze EVs for KRAS^WT and major mutations in codon 12 (KRAS^G12C, KRAS^G12D, KRAS^G12S, and KRAS^G12V). The KRAS SCOPE probes were first validated with EVs derived from a panel of CRC cell lines, each representing a different KRAS codon 12 mutation status (Supplementary Fig. 15).

Following probe validation, we proceeded to a clinical study as outlined in Fig. 6a (see Supplementary Table 2 for the patient information). We collected pre-surgical plasma samples from CRC patients who had not undergone any treatment. Concurrently, tissue samples were acquired and analyzed for KRAS mutations. For those patients with codon 12 KRAS mutations confirmed in tissue samples, we further examined longitudinal plasma samples collected after surgery and during standard adjuvant chemotherapy (one year). For EV isolation, we processed plasma samples through size exclusion columns (see Methods). The collected EV fraction showed positive signals for established EV markers (CD63, Alix) while lacking a non-EV marker (histone H2B), indicating EV enrichment in our samples (Supplementary Fig. 16a)³⁷. Furthermore, when we applied the SCOPE assay to both EV and non-EV fractions, only the EV fraction reported high SCOPE signals for mRNA targets (Supplementary Fig. 16b). Treating the EV isolates with RNase did not elicit significant changes to the SCOPE signal (Supplementary Fig. 16c). Collectively, these findings strongly suggested that EV-associated mRNA was the primary source of signals detected by the SCOPE assay.

Figure 6. Monitoring colorectal cancer patients.

(a) Plasma samples were collected from colorectal cancer (CRC) patients before and after surgery, as well as during standard care. Tissue samples were obtained during surgery and analyzed for the KRAS gene. (b) Plasma EVs were analyzed via SCOPE for a panel of KRAS mRNA mutations. Samples were from non-CRC controls (n = 15) and pre-surgery CRC patients (n = 107). SCOPE results reflected KRAS mutation status from tumor tissue analysis. KRAS signals from EVs were normalized against that of GAPDH. The heatmap shows mean values from technical triplicates. (c) Images of tissue staining (immunohistochemistry). For a patient positive with KRAS^G12D EV mRNA, the tumor tissue was also positively stained for KRAS^G12D protein. The staining experiments were performed on two different tissues, and representative images are shown. Scale bar, 20 μm. (d) The results from (b) were converted into the ratio (R_KRAS = KRAS^MT/KRAS^WT) of SCOPE signals between specific KRAS mutation (KRAS^MT) and KRAS wild type (KRAS^WT). The threshold for KRAS^MT positivity was established using profiling data from non-CRC controls and KRAS^WT CRC patients. For each specific KRAS mutation, the R_KRAS values of CRC patients harboring that particular mutation exceeded those of non-CRC controls and KRAS^WT CRC patients. Half-filled circles (◒) indicate patients whose KRAS mutation type is concordant between SCOPE and clinical tissue analyses. Each data point is a mean value from technical triplicates. (e) Seventeen KRAS^MT CRC patients were followed up for one year. For all patients, R_KRAS values immediately after surgery were lower than their initial values. Among the non-recurrent patients (n = 14), R_KRAS decreased below the threshold (5.5%). In the recurrent patients (n = 3), R_KRAS values eventually returned to pre-surgery levels. Each data point is a mean value from technical triplicates. (f) In the non-recurrent CRC patients, the R_KRAS values would fall below the threshold approximately ten days following surgery. The dotted line is the optimal fit for the post-surgery data, R_KRAS ~ d^0.3, where d is the number of days after surgery. The shaded region represents the 95% confidence interval. Data are shown as mean ± s.d. from 14 patients.

Figure 6b shows the initial EV profiling results of samples from pre-surgical CRC patients (n = 107) and non-cancer control subjects (n = 15; see Supplementary Fig. 17 for the complete heatmap). The SCOPE assay identified a subgroup of CRC patients (n = 17) whose mutation types matched their tissue KRAS status. This finding was further corroborated through tissue staining (immunohistochemistry), which detected mutated KRAS protein in tumor tissues (Fig. 6c and Supplementary Fig. 18). These observations underlined the rationale for assessing tumor KRAS status through EV profiling.

To evaluate the prognostic potential of detecting KRAS mutation via SCOPE, we used the ratio R_KRAS (= KRAS^MT/KRAS^WT) between KRAS mutation (KRAS^MT) and KRAS^WT signals. We then set a threshold for this ratio to determine whether a sample is positive for KRAS mutation. The data from EV samples without KRAS mutation (i.e., non-cancer subjects and KRAS^WT CRC patients) were pooled and fitted to a normal distribution (Supplementary Fig. 19). We then set the threshold for KRAS mutation negativity as the 99th percentile of this distribution (R_KRAS = 5.5%). Applying this threshold, we could effectively categorize samples by their KRAS mutation status (Fig. 6d). In samples with KRAS mutation, only the R_KRAS value of the mutated gene exceeded the threshold. In contrast, the values of non-mutated genes remained below it.

We further monitored the R_KRAS of the 17 patients with KRAS mutations. Fourteen patients were eventually deemed non-recurrent, with the remaining three considered recurrent; these were confirmed by surgical resection or radiological detection of lesions. Figure 6e summarizes the changes in R_KRAS values of these patients (see Supplementary Fig. 20 for individual patient results) as they received standard clinical care, including curative surgery and chemotherapy. For all 17 patients, R_KRAS initially decreased after surgery, presumably due to the reduction in CRC-derived EVs in circulation following tumor resection. However, a noticeable divergence emerged over time. Among the non-recurrent patients (Fig. 6e, left), R_KRAS continued to decrease and eventually fell below the non-KRAS^MT threshold, aligning with favorable clinical outcomes (tumor-free status). In contrast, R_KRAS values rebounded in recurrent patients and ultimately returned to their pre-surgery levels (Fig. 6e, right).

R_KRAS values for mutated mRNA were higher than the threshold (5.5%) in the days soon after surgery. We estimated EV clearing by modeling the R_KRAS trend in non-recurrent patients (Fig. 6f). The R_KRAS values decreased over time according to a power law (R_KRAS ~ d ^0.3), where d is the number of days after surgery. The model suggested that R_KRAS values would fall below the mutation threshold about ten days after surgery, which underscored the potential advantage of delaying EV analysis for a more accurate assessment of minimal residual disease and post-surgical prognosis. Furthermore, these analyses present an opportunity to develop an EV-based prognostic metric to guide the selection and duration of adjuvant treatment.

Stratifying glioma patients

We used SCOPE to detect a panel of genetic alterations in glioma patients, specifically IDH1 mutation, EGFR amplification, and EGFRvIII deletion. Obtaining this information can be crucial to classifying gliomas and guiding effective treatment based on their molecular traits³⁸. For instance, patients with IDH1-WT and EGFR amplification (classified as glioblastoma IDH1-wt) are linked to poor prognoses and resistance to radiotherapy³⁹. On the other hand, patients with the IDH1 mutation (classified as astrocytoma or oligodendroglioma depending on other markers, including 1p19p co-deletion) exhibit prolonged overall survival and are more likely to respond to temozolomide treatment^40,41. The analysis of extracellular vesicles (EVs) in peripheral blood holds particular promise for gliomas, considering the risks associated with complications and morbidity from brain tissue biopsies. Analyzing EVs in peripheral blood is particularly appealing for gliomas, considering the risk of complications and morbidity from brain tissue biopsies.

To investigate EV-based glioma typing, we analyzed EVs in plasma samples taken before surgery. Results were compared to those derived from corresponding tissue samples collected during surgical removal of tumors (Extended Data Fig. 1a). We first designed and validated a new set of SCOPE probes for mRNA targets (Extended Data Fig. 1b, Supplementary Fig. 21): wild-type IDH1 (IDH1^WT) and its point mutation (IDH1^R132H) most frequently found in gliomas; WT EGFR (EGFR^WT) and EGFRvIII. The designed probes were highly specific (Extended Data Fig. 1c) and enabled SCOPE to genotype different GBM cell lines through EV profiling (Extended Data Fig. 1d).

Next, we analyzed plasma samples from 15 non-cancer controls and 60 glioma patients with different tumor subtypes (see Supplementary Table 3 for patient information). We applied SCOPE to profile EVs for GAPDH, CD63, IDH1^WT, IDH1^R132H, EGFR^WT, and EGFRvIII mRNA. Given that both EGFR^WT and EGFRvIII were designated as target markers, we used the GAPDH gene as a reference for normalization purposes^14,42. Specifically, we scaled signals from other markers against the GAPDH signal rather than estimating VAFs. We also molecularly assessed and classified patients’ tumor specimens (see Methods). The SCOPE results demonstrated the potential of robust and highly sensitive glioma typing through EV analysis (Extended Data Fig. 1e, Supplementary Fig. 22). We observed differential expression patterns of EGFR^WT, IDH1^R132H, and EGFRvIII across different glioma subtypes. For example, EGFR^WT expression in EVs was higher in GBM patients with EGFR amplification or EGFRvIII mutation than in IDH1^R132H mutant patients. IDH1^R132H and EGFRvIII mRNA were only detected in the EVs from patients with matching mutations in tumor tissue.

DISCUSSION

EVs are a promising source of cell-free RNA for liquid biopsy applications both in preclinical and clinical contexts. Nucleic acids within EVs are shielded from degradation and are more abundant than ctDNA, which renders EV RNA robust analytical targets for molecular diagnostics^14,43. Among the various EV RNA species, microRNAs are highly abundant and thus more commonly studied^18,19. In contrast, due to its lower concentration, detecting EV mRNA often requires large sample volumes and advanced instrumentation. The SCOPE technology can overcome these limitations by improving target amplification and signal generation. i) SCOPE exhibits specificity down to a single nucleotide. This precision is achieved through an innovative concept where RNA replication is initiated solely upon recognition of target RNA by Cas13a. ii) SCOPE’s concurrent amplification of RNA targets and fluorescent signals leads to analytical sensitivity far surpassing other isothermal amplification methods (see Supplementary Table 4 for comparison)⁴⁴. iii) SCOPE produces these results within a short timeframe (30 min) through a one-pot assay that can be conducted using readily available laboratory tools. These advantages would facilitate SCOPE’s integration into routine preclinical and clinical applications across academic and industrial laboratories.

A key innovation of SCOPE lies in its signal template, a hybrid design that combines a cleavable, fluorescent RNA probe with a DNA template. This structure synergistically couples two enzymatic activities (Cas13a collateral cleavage and T7 polymerase reaction) and establishes a self-amplifying cycle to accelerate signal growth. Notably, this reaction is conditional — Cas13a first needs to recognize the RNA target to unlock the subsequent SCOPE reaction. Our rationale for this design was supported by kinetic modeling and experimental data. When Cas13a and T7 reactions were executed individually, their assay products increased linearly over time, consistent with zeroth-order kinetics. Conversely, SCOPE signals displayed exponential growth and rapidly reached saturation, closely resembling first-order rate kinetics. However, we acknowledge that the reaction scheme may be susceptible to false positives. A potential contributing factor is the degradation of the fluorescent RNA segment in the signal template. This segment, designed for specific cleavage by the activated Cas13a, can be degraded by other RNases present in samples. In the current work, we mitigated such an effect by including a broad-spectrum RNase inhibitor in the SCOPE mixture. We also analyzed technical replicates of each sample to guard against false positives. In future work, we may consider chemically modifying the RNA segment, for instance, with phosphorothioate to enhance nuclease resistance⁴⁵. Further investigations would be necessary to evaluate Cas13a trans-cleavage efficacy on these modified targets.

We validated SCOPE’s performance through systematic pre-clinical and clinical studies. SCOPE’s limit of detection extends into the sub-attomolar ranges. Equally impressive was SCOPE’s high selectivity, enabling the precise identification of point mutations. In a noteworthy example, SCOPE effectively differentiated KRAS^WT mRNA from its mutated subtypes (G12C, G12D, G12S, G12V) with minimal crosstalk, surpassing the performance of a commercial KRAS mutation assay. Furthermore, SCOPE detected KRAS mutant alleles at levels as low as 0.01% VAF, a sensitivity comparable to high-end techniques such as digital PCR and BEAMing PCR⁴³. The SCOPE assay achieved this sensitivity without sample partitioning and remained compatible with conventional thermal cyclers, which would enhance its accessibility to a wider research community.

SCOPE demonstrated potential for various clinical applications, including early tumor detection, recurrence monitoring, residual disease assessment, and tumor subtyping. The assays were fast (<30 min for signal generation) and cost-effective (<$4 per marker). These findings illustrate SCOPE’s versatility and affordability as a diagnostic tool to expedite clinical decision-making (same-day turnaround), track cancer recurrence/minimal residual disease, gauge responses to therapies (both conventional and experimental under trial auspices), and achieve timely pre-clinical readouts to support biological and drug development efforts. The current results are pertinent to SEC-processed plasma samples, a commonly used approach across EV studies⁴⁶. We employed SEC throughout the study for consistent sample handling, but this method may co-isolate additional extracellular RNA carriers, such as lipoproteins and platelet-derived EVs, that could contribute to the overall SCOPE signals. To further elucidate specific mRNA carriers, we may consider a single-vesicle imaging study; this approach will involve fluorescent labeling of vesicle surface proteins (e.g., CD9 for EVs and CD31 for platelets) to identify their origins^9,47 while simultaneously labeling target mRNA with molecular beacons. Subsequent imaging of these vesicles should readily reveal the origin of mRNA-carrying vesicles⁴⁸. We could also conduct studies that evaluate SCOPE’s applicability with other sample types and pre-processing methods (see Supplementary Fig. 23 for examples) to examine SCOPE’s robust performance across diverse use cases.

By employing the SCOPE technique, researchers could expand their knowledge of EV characteristics. In this work, we detected circulating EVs harboring genetic mutations that mirrored tumor tissues, thereby reaffirming EVs’ potential as biomarkers. Longitudinal monitoring yielded further insights into EV dynamics. i) Mutation targets in EVs progressively increased during tumor initiation and growth. ii) EV mutational loads declined from pre-surgery values following curative surgery in CRC patients. This decline persisted in patients responsive to chemotherapy and receded to the background value. These observations contrasted with non-responsive CRC patients whose EV mutational loads stayed above the background value. These results underscore EVs’ potential utility for tumor monitoring, akin to the capabilities demonstrated with ctDNA⁴⁹. We also note unexpected observations that warrant further investigation. Within a week of CRC surgery, EV mutational loads may remain elevated, potentially due to residual microscopic tumors or the release of tumor-derived materials during surgery. In our pilot study, this uncertainty resolved approximately ten days after surgery, a potentially optimal window to begin EV analysis to inform post-surgery prognoses and shape the type and duration of adjuvant interventions. A valuable next step would be to broaden SCOPE to include diverse tumor types and assess its applicability within multimodal treatment contexts. This expansion could establish cancer-specific timeframes for optimal EV analysis, furthering its utility in treatment monitoring and prognostication.

Another potent strategy entails analyzing both EV mRNA and ctDNA in plasma to obtain maximal molecular information. EV mRNA possesses the advantage of higher abundance compared to ctDNA, thereby improving sensitivity for cancer diagnostics⁴³. Detecting EV mRNA can also be practical for certain targets. For example, EGFRvIII deletion involves variable breakpoints in the EGFR gene⁵⁰, which complicates the design of specific ctDNA assays. Alternatively, EV mRNA analysis can target the conserved EGFRvIII transcript sequence, facilitating a more straightforward assay. In comparison, ctDNA analyses can provide molecular insights unavailable at the transcript level, such as promoter alterations⁵¹ and methylation patterns⁵², which can enhance diagnostic accuracy and indicate the tumor’s tissue of origin. Conveniently, both EV nucleic acids and ctDNA can be extracted from plasma using the same isolation process to expedite these synergistic analyses⁴³.

We envision further technical improvements to enhance SCOPE’s preclinical and clinical impact. One straightforward approach is to encompass other major mutations for wider cancer coverage^53,54. These targets could include PIK3CA (H1047R; E545K), TP53 (R175H; R273C; G245S), NRAS (Q61R; Q61K), BRAF (V600E; V600K; V600R), and EGFR (L858R; E746_A750del). Additionally, the panel could be tailored for treatment-related targets, such as MGMT for temozolomide resistance¹⁴, EGFR (T790M) for resistance to EGFR tyrosine kinase inhibitors (TKIs)⁵⁵, and FGFR3 (R248C; S249C; G370C; Y373C) for sensitivity to FGFR TKI treatments⁵⁶. SCOPE’s high selectivity will help design these probes. In addition, the assay’s “one-target per tube” format will effectively eliminate potential crosstalk between probes within a reaction vessel; probe optimization can thus focus on attaining similar analytical performance among different probes. For example, we prepared BRAF mutation probes (Supplementary Fig. 24) within two weeks, demonstrating the assay’s efficiency for rapid probe development.

We can also incorporate protein detection into the assay workflow, capitalizing on EVs’ inherent advantage as a multi-cargo carrier. Detecting proteins will complement mRNA analyses, establishing greater connections between tumor genotypes and phenotypes. Protein markers can offer additional insights into tumors’ tissue of origin⁵⁷, crucial information often difficult to obtain solely from driver gene mutations. We propose using aptamers or DNA-barcoded antibodies as affinity ligands, which will make protein assays compatible with SCOPE. We can further explore employing orthogonal Cas proteins that exhibit distinct preferences for motif sequences during trans-cleavage⁵⁸. By designing signal templates that are fluorescently distinguishable and compatible with each Cas variant, it would be feasible to detect multiple targets (4 to 6) in a single reaction. In addition, adapting SCOPE to the digital PCR format (i.e., sample partitioning) would be an intriguing direction for absolute target quantification and further enhancements in VAF sensitivity. With these advancements, SCOPE would be positioned as a powerful liquid biopsy tool for precision oncology, enabling comprehensive and reliable molecular characterization of tumors in a single, cohesive, and accessible platform.

ONLINE METHODS

Materials

All oligonucleotides used in this study were purchased from Integrated DNA Technologies, Inc. (USA). The specific sequences of the oligonucleotides are listed in Supplementary Table 1 (SCOPE) and Supplementary Table 5 (RT-PCR). Monarch^® RNA Lysis Buffer, Monarch^® RNase A, RNase inhibitor (Murine), Proteinase K, DNase I, HiScribe^™ T7 High Yield T7 RNA synthesis kit, Luna^® Universal One-Step RT-qPCR kit, and Quick-Load^® Purple Low Molecular Weight DNA Ladder were obtained from New England BioLabs (USA). Qubit^™ RNA HS Assay kit and DEPC-DW were purchased from Thermo Fisher Scientific (USA). The RNeasy Micro kit was purchased from Qiagen (Germany), and the Total Exosome RNA & Protein Isolation kit was from Invitrogen (USA). GelRed nucleic acid stain (41003) was sourced from Biotium (USA). All other chemicals used in the experiments were of analytical grade and used without further purification.

Expression and purification of LwaCas13a

We prepared the LwaCas13a protein through bacterial expression and subsequent purification. A plasmid DNA carrying the LwaCas13a encoding sequence (plasmid #90097; Addgene, USA) was introduced into the E. coli strain Rosetta 2 (DE3) pLysS (Millipore Sigma, Burlington, MA, USA). The transformation process relied on the well-established heat shock transformation method, specifically optimized for bacterial expression. Post-transformation, protein expression was induced using isopropylthio-β-galactoside (0.5 mM; Millipore Sigma). LwaCas13a proteins were then isolated through a systematic multi-step process, commencing with cell lysis and initial nickel-nitrilotriacetic acid (Ni-NTA; Thermo Fisher Scientific) purification. The solution underwent the first buffer exchange to a small ubiquitin-like modifier (SUMO) cleavage buffer (30 mM Tris-HCl at pH 8.0, 500 mM NaCl, and 1 mM dithiothreitol/ DTT). This was followed by treatment with SUMO protease (Thermo Fisher Scientific), a second round of Ni-NTA purification, and a final buffer exchange to the storage buffer (50 mM Tris-HCl at pH 7.5, 600 mM NaCl, and 2 mM DTT). The purified LwaCas13a proteins were stored with 5% glycerol (Millipore Sigma) and a protease inhibitor (Roche, Switzerland) at −80 °C until use. To verify the successful purification of LwaCas13a proteins, the products obtained from each purification step were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by Ag staining (Supplementary Fig. 24).

Preparation of pDMAEA-coated PCR tubes

We used PCR tubes (diameter 6 mm) purchased from Bio-Rad (USA). Tert-butyl peroxide (TBPO, 98%) and dimethylaminoethyl acrylate (DMAEA, 98%) were purchased from Merck (Germany). The DMAEA monomer and TBPO were vaporized through heat exposure at 35 °C and 25 °C, respectively. Subsequently, the vaporized substances were introduced into an initiated chemical vapor deposition (iCVD) chamber (160 mTorr). The filament temperature of the iCVD was set at 140 °C to generate TBPO radicals. The tubes were held at 30 °C to facilitate vapor absorption. The pDMAEA thin film was produced as a result of the vapor-phase free radical polymerization, facilitated by a filament temperature of 140 °C.

Characterization of pDMAEA hydrolysis

For nuclear magnetic resonance (NMR) analysis, we dissolved pDMAEA in CDCl₃ or phosphate D₂O buffer at pH 8.0. Subsequently, 1H NMR spectra were collected using a Bruker Avance Neo 600 MHz spectrometer (Bruker BioSpin, Germany). The degree of hydrolysis was determined by comparing the integrated signals for the methylene of the ester in pDMAEA (~4.41 ppm) and the alcohol of dimethylaminoethanol (DMAE) (~4.00 ppm). For Fourier transform infrared (FTIR) analysis, we collected the spectra in the wavenumber range of 400 – 4000 cm⁻¹ and at a resolution of 1.4 cm⁻¹, using an Alpha instrument from Bruker Optics (USA). The zeta-potential values were determined using a Nano Zetasizer 3600 (Malvern, UK) equipped with a He-Ne laser source operating at 633 nm.

Characterization of pDMAEA tube

To evaluate the RNA extraction efficiency of the pDMAEA-coated tube, we added 100 μL of initial RNA (3, 30, 60, or 100 ng) solution to the tube and let the solution react for 10 min. We then centrifuged the tube for 5 min (123 ×g), collected the supernatant, and quantified the RNA concentration (supernatant) using the Qubit^™ RNA HS Assay kit. For the comparison, we processed the same RNA samples (100 μL with initial RNA amounts of 3, 30, 60, and 100 ng) using the RNeasy Micro Qiagen kit based on silica columns. Following the RNeasy protocol, the concentration of the extracted RNA was measured using the Qubit^™ RNA HS Assay kit.

Construction of SCOPE device

We designed the device using design software (Inventor; Autodesk, USA). The device’s primary components were fabricated via computer numerical control (CNC) machining on Al alloy (AL6061) and injection molding with polycarbonate. Heating block. The block was machined on an Al block to hold 16 standard PCR tubes (diameter, 6 mm). We used two polyimide heaters to heat the block: a high-powered 144-watt main heater and a 24-watt top heater. These heaters were driven by field-effect transistors connected to a 24 V power supply. The temperature of each heater was continually monitored using a 10-kΩ thermistor (NTCG103JF103F; TDK, Japan) and was maintained within a 0.3 °C tolerance through feedback control. Fluorescent optics. The device had an optical module consisting of two independent units, each configurable for a different fluorescent channel. For the green fluorescent channel (excitation 495 nm; emission 520 nm) used in this study, the components in the unit were a blue light-emitting diode (C503B-BAN-CZ0A0451; Cree LED, USA), a collimating lens (LC12; Roithner Lasertechnik, Austria), an excitation band-pass filter (FF01–474/27; Semrock, USA), a dichroic mirror (Di02-R488, Semrock), an emission band-pass filter (FF01–515/30, Semrock), and two focusing lenses (LC7, GS7020–2; Roithner Lasertechnik). The fluorescent emission from the sample was detected by a photodiode that had a broad spectral response range from 320 to 1100 nm (S2386–44K; Hamamatsu, Japan). The current from the photodiode was converted into a voltage signal by a trans-impedance amplifier (OPA320; Texas Instruments, USA) and further boosted by a non-inverting amplifier (AD8605; Analog Devices, USA). The resultant analog signal was converted into a digital format by a 16-bit analog-to-digital converter (ADS1115, Texas Instruments) that collected data at a 62.5 samples/sec sampling rate. Linear actuator. We mounted the optical module on a linear actuator to scan the module across 16 samples for fluorescent detection. The actuator was powered by a stepping motor (28BYJ-48; Kiatronics, New Zealand). The position of the optical module was initialized by moving the module until it contacted a limit switch. Microcontroller unit (MCU). We programmed an ESP32 MCU (Espressif, China) to operate the device. The MCU processed various data streams, including fluorescent measurement results, scanning locations, and heater temperatures. It also communicated with an external tablet through a Bluetooth connection. The MCU firmware was written in C++ (gcc version 12.1).

SCOPE App

The software was an Android App with an intuitive graphical user interface. We wrote the App using JAVA and the Android software development kit. The App saved all fluorescent measurements and test settings in a standardized comma-separated-value (CSV) format to ensure compatibility across various operating systems.

Cell culture

We purchased a panel of cell lines from the American Type Culture Collection (ATCC, USA) and grew them in the vendor-recommended media: H2228 (RPMI-1640, Cellgro); A549 (F-12K, ATCC); CCD-18 (DMEM, Corning); Colo205 (RPMI-1640); HCT116 (MacCoy’s 5a, Cellgro); HT29 (MacCoy’s 5a modified with 2% NaHCO₃, Cellgro); LS174T and LS123 (Eagle’s Minimum Essential Medium; Cellgro); SW480 and SW620 (DMEM, Corning). The KP1.9 cell line was derived from tumors of the KP mouse model on a C57BL/6 congenic background and was a generous gift from Dr. A. Zippelius, University Hospital Basel, Switzerland. KP1.9 cells were cultured in an Iscove’s DMEM medium (Corning). For GBM cell lines, we purchased GLI36 cells from ATCC and generated GLI36vIII and GLI36-R132H cell lines through lentivirus transduction. Both GLI36 and its variants were cultured in a DMEM (Corning). All media were supplemented with 10% fetal bovine serum (FBS) and penicillin-streptomycin (Corning). All cell lines were tested and determined to be free of mycoplasma contamination (MycoAlert Mycoplasma Detection Kit, LT07–418; Lonza, Switzerland).

EV preparation via size exclusion chromatography (SEC)

We cultured cells at passages 1–15 in a vesicle-depleted medium (with 5% depleted FBS) for 48 h. We then collected culture media from approximately 10⁷ cells and removed cellular debris through two successive rounds of centrifugation (10,000 ×g, 3 min). To isolate EVs, we first concentrated the conditioned culture medium by processing it with a centrifugal filter (Centricon Plus-70; Millipore Sigma). Subsequently, we loaded the concentrated medium (0.5 mL) to a size-exclusion column (SP1; Izon, New Zealand) and collected EV fractions (F7-F9; 1.5 mL) in low-protein-binding microcentrifuge tubes (90410; Thermofisher Scientific). Collected EVs were resuspended in a 1× phosphate-buffered saline (PBS) buffer and stored at −80 °C. We estimated EV concentrations in these stock solutions via nanoparticle tracking analysis (NTA). For each sample, we acquired three 30-second video clips (NanoSight LM10; Malvern) and analyzed them using NTA software (version 3.2).

EV preparation via ultracentrifugation

Cells were cultured until they reached approximately 70% confluency. At this stage, the supernatant was removed and replaced with F-12K Medium (ATCC, 30–2004) containing EV-depleted fetal bovine serum (FBS). After a 24-hour cell culture, the culture media supernatant was processed through sequential centrifugation. Initially, the supernatant underwent centrifugation at 300× g for 10 minutes and then at 2000× g for 10 minutes using a centrifuge to eliminate cells and debris. Subsequently, the supernatant was subjected to ultracentrifugation (Beckman, Optima L-100K) at 4 °C and 10,000× g for 30 minutes to remove proteins. The supernatant was ultracentrifuged at 4 °C and 100,000× g for 70 minutes to pellet EVs. The pellets were resuspended in cold PBS and ultracentrifuged again at 100,000× g for 70 minutes at 4 °C to eliminate residual impurities. Finally, the EV pellets were collected, resuspended in PBS (1.5 mL), and stored at −80 °C. For SCOPE analysis, we used 100 μL of this sample.

Western blotting

EVs were enriched from a plasma sample (3 mL) obtained from a colon cancer patient. The plasma was initially centrifuged at 2,000 × g for 3 min to remove cellular debris, and the resulting supernatant was processed using size exclusion chromatography (qEV, Izon Science). The isolated EVs were concentrated to 100 μL using Amicon Ultra-2 10K filters (Millipore Sigma). For western blot analysis, the concentrated EVs were lysed with ice-cold 10× RIPA buffer (Abcam) for 30 min, followed by centrifugation at 14,000 × g for 10 min at 4 °C. Protein concentration was determined via a BCA assay, and the lysates were analyzed via 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The following primary antibodies were used: Alix (1:1000, JM85–31, Invitrogen), biotinylated CD63 (1:1000, AHN16.1/46–4-5, Ancell), and Histone H2B (1:1000, D2H6, Cell Signaling Technology). Prior to use, the Alix and Histone H2B antibodies were biotinylated using sulfo-NHS-Biotin (Thermo Fisher Scientific) in accordance with the manufacturer’s protocol. Secondary detection was subsequently conducted using HRP-conjugated streptavidin (1:2000, R&D Systems).

SCOPE assay

We placed an EV sample (100 μL) and a lysis buffer (10 μL) in a pDMAEA-coated tube. The lysis buffer consisted of proteinase K (0.8 U/reaction), RNase inhibitor (1 U/reaction), and DNase I (2 U/reaction) in Tris-EDTA (TE; pH 8.0). After incubating the mixture at 25 °C for 10 min, we centrifuged the sample (123 ×g, 5 min) and discarded the supernatant, leaving about 1 μL of the lysate in the tube. We next added the SCOPE master mix (20 μL) to the same tube and let the mixture react (40 °C, 30 min) inside the SCOPE device. The master mix had the following components: 5 μM LwaCas13a, 10 μM crRNA, 50 μM Signal template, 16 U/μL RNase inhibitor, 10 mM deoxyribonucleotide triphosphate mixture (ATP, GTP, UTP, and CTP), 0.1 μM RNA polymerase, 0.5× Cas13a cleavage buffer (10 mM HEPES-Na at pH 6.8, 25 mM KCl, 2.5 mM MgCl₂, and 2.5% glycerol), and 0.5× RNAPol buffer (20 mM Tris-HCl at pH 7.9, 3 mM MgCl₂, 0.5 mM DTT, and 1 mM spermidine). Throughout the SCOPE reaction, we monitored the fluorescent signals using the device.

Quantitative PCR

We used quantitative reverse transcription-PCR (qRT-PCR) for comparison with SCOPE. We used either synthetic RNA samples (1 μL) or EV RNA samples (1 μL) as prepared in the SCOPE assay. A sample was mixed with the master mix (20 μL) containing 1× Luna Universal One-Step Reaction Mix, 1× Luna WarmStart^® RT Enzyme Mix, and 400 nM primers. We then carried out reactions on a CFX Connect 96 Real-Time PCR System (Bio-Rad) using the following thermal cycling schedule: 55 °C, 10 min (reverse transcription) and 95 °C, 1 min (RTase inactivation), followed by 40 cycles of 95 °C (10 sec), 60 °C (30 sec), and fluorescent measurements.

Gel electrophoresis

We resolved 20 μL of the reaction mixture on a gel composed of 15% polyacrylamide. We used a 1× TE solution as a running buffer and applied a constant voltage of 120 V throughout the electrophoresis process (2 hours). Upon the completion of this phase, the polyacrylamide gels were subjected to staining with GelRed. Subsequently, we scanned the stained gels using a Geldoc Go imaging system (Bio-Rad).

Comparison of SEC-processed and direct plasma samples for the SCOPE assay

A plasma sample from a colorectal cancer (CRC) patient was divided into two aliquots of 40 μL each. One aliquot underwent size exclusion chromatography (SEC) to yield a final volume of 600 μL of EV isolate, from which RNA was extracted using the Total Exosome RNA & Protein Isolation Kit (Invitrogen, 4478545). The remaining plasma aliquot was processed for RNA extraction using QIAamp RNA Blood Mini Kit (Qiagen, 52304) in accordance with the manufacturer’s instructions. The extracted RNA was then quantified, and the final concentration was adjusted to 1 ng/ μL. The extracted RNA samples (50 μL) were then used for the SCOPE assay.

Comparative experiments with RNase treatment

We utilized 40 μL of plasma samples from CRC patients and isolated EVs via SEC. The final sample volume was 600 μL after SEC operation. We mixed 100 μL of this EV isolate with 3 μL RNase A (New England Biolabs, T3018L). The mixture was incubated at 25 °C for 5 minutes. Subsequently, the mixture was filtered through a 30 kDa Amicon Ultra centrifugal filter (Millipore, UFC503008) to remove RNase A. A control sample underwent the same processes, with RNase A replaced by PBS. All prepared samples were then lysed and subjected to the SCOPE assay.

EV profiling from mouse samples

We conducted the animal study following the guidelines provided by the Institutional Animal Care and Use Committee (IACUC) of Massachusetts General Hospital (Protocol 2019N000133). We used 129S1 mice, which carried the Cre-activatable conditional Kras^LSL-G12D/+; Trp53^flox/flox (KP mice) as an autochthonous model of lung adenocarcinoma^33,35. Mice were fed with autoclaved food and water and maintained in ventilated cages in a light-dark cycle, temperature (18–23 °C), and humidity (40–60%) controlled pathogen-free vivarium at Massachusetts General Hospital. To initiate the tumor growth, we intranasally instilled adenoviruses (3 × 10⁷ particles) expressing Cre recombinase (Ad-Cre, University of Iowa Viral Vector Core). We collected blood samples (~50 μL) from the mice before the virus instillation and weekly thereafter, using the retro-orbital bleeding method. Mice were monitored for signs of disease progression and distress, including according to their body condition score and body weight. To collect plasma, we centrifuged blood samples (2,000 ×g, 15 min) at 4 °C and recovered supernatant. Plasma samples were stored at −80 °C until use. For EV isolation, we diluted plasma samples in PBS to 100 μL and processed them using size-exclusion columns (SP2; Izon) as described above.

EV profiling with clinical samples

The study was approved by the Institutional Review Board (IRB) of Massachusetts General Hospital (Protocol 2017P001581, 2019P003441, 2019P003472) and the IRB of Kyungpook National University Chilgok Hospital (Protocol KNUCH 2021–10-025). The procedures followed were per the respective institutional guidelines. Informed consent was obtained from all subjects. We drew peripheral blood samples (~10 mL) from patients into blood collection tubes (Vacutainer 367525; BD). Samples were then centrifuged (2,000 ×g, 3 min) to separate plasma from red blood cells and buffy coat and stored at −80 °C until use. For EV isolation, stored plasma samples (40 μL) were diluted in PBS (60 μL) and processed with size-exclusion columns (SP1; Izon).

Tissue analysis

Colorectal cancer patients. Pathologic evaluation and KRAS analysis of colorectal cancer tissues were performed by the Department of Pathology at the Kyungpook National University Chilgok Hospital. Formalin-fixed, paraffin-embedded (FFPE) tissue blocks were used for DNA extraction. The PNAClamp^™ KRAS Mutation Detection kits Ver. 4 (Panagene, Korea), capable of detecting 40 KRAS mutation variants across codons 12, 13, 59, 61, 117, and 146, were employed for assays following the manufacturer’s instructions. Glioma patients. The Massachusetts General Hospital Pathology Service analyzed pathologic tissues from patients. The analysis involved the use of various methods, including immunohistochemistry for EGFR, Snapshot NGS for IDH1, Genexus NGS for IDH1, and the Solid Fusion assay for EGFRvIII.

Immunohistochemistry

Lung tissue from KP mice. We anesthetized mice with isofluorane and perfused them with 4% paraformaldehyde (PFA) through the trachea and by the tail vein. Lungs were further fixed for 24 h, washed with PBS, and soaked in 30% sucrose in PBS for 24 h. The lung tissue was then embedded in the optimum cutting temperature (OCT) compound (Sakura Finetek) and snap-frozen in an isopentane bath on dry ice. Once the tissue was frozen, we prepared serial 10- μm thick sections using a cryostat and placed them on glass slides. Samples were then sent to the Histopathology Research Core at Massachusetts General Hospital for staining: hematoxylin and eosin (H&E), and RAS G12D (GTX635362, GeneTex; 25× dilution). Tissue from colorectal cancer patients. IHC was performed on representative FFPE blocks using an automatic IHC staining instrument (BenchMark^®XT, Ventana Medical Systems) according to the manufacturer’s protocols. Briefly, 4 μm FFPE sections were transferred to poly-L-lysine-coated slides and dried in an oven at 65 ℃ for 2 h. After deparaffinization in xylene and rehydration through a series of graded alcohols, we retrieved antigens by boiling samples via microwave in 10 mM sodium citrate buffer (pH 6.0). Sections were then incubated with a primary antibody against RAS G12D (GTX635362, GeneTex; 200× dilution) overnight at 4 °C. Slides were then visualized using the UltraView Universal DAB kit (Ventana Medical Systems) and counterstained with Harris hematoxylin for nucleus. RAS G12D protein was considered positive when IHC staining was positive in the cytoplasm of tumor cells and was not expressed in normal non-neoplastic colonic epithelium cells.

Statistical analysis

We used GraphPad Prism version 10.2.3 (GraphPad Software Inc.) or R version 4.2.2 for statistical analyses. For two-group comparisons, we used an unpaired, two-sided t-test. For all statistical tests, p-values less than 0.05 were considered significant. Details on data presentation and the sample size are included in figure legends.

Extended Data

Extended Data Fig. 1 |. Stratifying glioma patients.

(a) SCOPE analyzed plasma samples collected from radiologically confirmed glioma patients. Tissue samples were used for clinical pathology. (b) Target RNA sequences for glioma-SCOPE analyses. IDH1 probes were designed to detect wild-type (WT) and single nucleotide mutation (R132H), and EGFR probes to detect WT as well as the variant III (EGFRvIII) resulting from genomic deletion of exons 2–7 in the EGFR gene. (c) The specificity of glioma-SCOPE probes was evaluated. The designed probes achieved a high signal contrast (>30) between on-target and off-target samples. Synthetic RNA samples (1 nM) were used. The heatmap displays mean values from technical triplicate measurements. a.u., arbitrary unit. (d) Application of SCOPE to profile EVs for IDH1 (WT, R132H) and EGFR (WT, vIII). EVs were harvested from glioma cell lines. The results confirmed that EVs reflected the genotype of parent glioma cells. Data are shown as mean ± s.d. from biological triplicates. a.u., arbitrary unit. (e) Plasma EVs were analyzed via SCOPE for IDH1 (WT, R132H) and EGFR (WT, vIII). Control samples were from healthy donors (n = 15). Glioma samples were from patients with different genotypes: EGFR amplification (GBM-WT; n = 20), IDH1-R132H mutation (n = 20), and EGFRvIII mutation (n = 20). The SCOPE assay identified glioma patients and correctly stratified them according to their molecular genotypes. The heatmap shows mean values from technical triplicate measurements.

DATA AVAILABILITY

The mRNA target sequences were obtained from the National Center for Biotechnology Information (NCBI) Reference Sequence database (accession codes: NC_000002.12, NC_00007.14, NC_000012.12). The primary data supporting assay characterization are accessible in a Source Data file and Supplementary Information. Raw patient datasets generated and analyzed during this study are available from the corresponding authors upon request, subject to approval from the Institutional Review Board of Massachusetts General Hospital (MGH) and MGH Innovation Office. Data access requests will be considered from academic investigators without relevant conflicts of interest, for noncommercial use, who agree to non-distribution terms.