Detection of Ultralow-Frequency ctDNA Mutations Using a Dual Hairpin-Competition CRISPR/Cas14a System Authors: Zhaocheng Liu, Rui Zhang, Chunyan Chang, Daxing Xu, Dongsheng Mao, Xiaoli Zhu, Hongyang Xu Published: 2026 ABSTRACT Circulating tumor DNA (ctDNA) mutation profiling is essential for guiding targeted therapy and monitoring cancer recurrence, yet its clinical adoption is constrained by overwhelming wild-type DNA background and the limited sensitivity of existing platforms. Here, we introduce a dual hairpin-competition CRISPR/Cas14a (DHCC) system that integrates two sequential layers of hairpin competition: selective enrichment of mutant DNA during asymmetric PCR, followed by suppression of nonspecific sgRNA binding during Cas14a detection. This design dramatically enhances mutant-wild-type discrimination, elevating the discrimination factor from 2.48 to 145a 58-fold improvement. While previous Cas14a methods achieve detection limits of 0.5–0.1% variant allele frequency (VAF), DHCC delivers a 250-fold sensitivity gain, routinely detecting four clinically relevant mutations (EGFR T790M, L858R, G719A, and NRAS Q61K) at VAFs as low as 0.002%. In multiplexed format, sensitivities of 0.005–0.01% VAF are maintained. Clinical validation using 22 plasma ctDNA samples demonstrated 100% concordance with droplet digital PCR for EGFR L858R detection. Compared to ddPCR and next-generation sequencing, DHCC substantially reduces turnaround time and cost while operating on standard qPCR instruments, eliminating the need for specialized infrastructure. By combining ultrahigh sensitivity, PAM independence, multiplexing preamplification capability, and practical affordability, DHCC provides an accessible platform for ctDNA-based liquid biopsy in clinical settings. Introduction Liquid biopsy (LB) has revolutionized cancer management through noninvasive interrogation of circulating biomarkers in peripheral blood, such as circulating tumor DNA (ctDNA), circulating tumor cells (CTCs), and exosomes. This paradigm-shifting technology allows for early tumor detection, real-time therapeutic evaluation, and precise recurrence monitoring. − Among these biomarkers, ctDNAextracellular tumor-derived DNA fragments detectable in blood and urinehas become pivotal in molecular oncology due to its noninvasive accessibility, reduced tumor heterogeneity, and dynamic correlation with disease progression. The clinical significance of ctDNA mutation profiling is broad, ranging from guiding targeted therapies to monitoring cancer recurrence through the detection of low-frequency variants. , Nevertheless, the abundance of mutant ctDNA released from small solid tumors or minimal residual disease (MRD) is minimal, with a significantly high background of wild-type templates. For instance, the ctDNA fraction corresponding to a tumor volume of 1 cm3 is only 0.002% to 0.03%. , Presently, detection platforms like next-generation sequencing (NGS), electrochemical biosensors, and various PCR-based strategies are employed. − NGS, known for its high-throughput capabilities, is widely used in variant detection. By enhancing sequencing depth, such as methods of personalized deep sequencing analysis (CAPP-Seq) , and Signatera platform, , these methods achieved the identification of low-abundance mutations. Nevertheless, deep sequencing approach had drawbacks such as multiple operational steps, lengthy analysis times, and high costs. In addition to NGS, quantitative PCR (qPCR) ,,− and digital PCR , are essential techniques for mutation detection. However, traditional qPCR lacks the sensitivity required for low-frequency variant detection, while digital PCR necessitates specialized droplet generation and fluorescence reading equipment, leading to increased expenses. The CRISPR/Cas system is a prominent molecular diagnostic technology platform known for its user-friendly application and minimal equipment requirements. − Cas proteins, guided by sgRNA for target-specific recognition, demonstrate high specificity and sensitivity through their trans-cleavage activity. Unlike Cas12a, which requires PAM sequenceslimiting its applicability in mutation detectionCas14a (Cas12f1) offers the advantage of PAM-independent recognition of single-stranded DNA targets and exhibits robust activity in detecting single nucleotide variations. , Previous studies have demonstrated that the Cas14a system, when coupled with ligase chain reaction (LCR), can detect the BRAF V600E mutation at variant allele frequencies (VAF) of 0.5%. The incorporation of an enrichment step using blocker displacement amplification (BDA) further enhanced detection sensitivity to 0.1% VAF. These findings underscore the effectiveness of mutation-enrichment strategies in reducing wild-type DNA background. However, the degree of wild-type suppression achieved by BDA remains insufficient for the reliable detection of ultralow frequency mutations. Therefore, a novel CRISPR/Cas14a approach capable of substantially mitigating wild-type interference and enabling robust identification of ultralow variants is urgently needed. Here, we developed a dual hairpin-competition CRISPR/Cas14a (DHCC) system for the ultrasensitive detection of low-frequency circulating tumor DNA (ctDNA) mutations. The DHCC system incorporates engineered hairpin structures to initiate a two-step competitive mechanism: the first step enriches mutant DNA during asymmetric PCR, while the second minimizes nonspecific hybridization between mutant-specific sgRNA and wild-type DNA. This dual-hairpin competition strategy achieved detection of mutations at variant allele frequencies (VAFs) as low as 0.002% across four distinct mutation types. Furthermore, multiplexed preamplification DHCC demonstrated robust sensitivity, detecting variants at VAFs of 0.005% to 0.01%. In a comparative study using 22 clinical ctDNA samples, DHCC and droplet digital PCR (ddPCR) showed 100% concordance in identifying EGFR L858R mutations. Given its ultrahigh sensitivity, independence from protospacer-adjacent motif (PAM) constraints, and capability to simultaneously enrich multiple mutation targets, DHCC represents a promising platform for guiding targeted therapies and monitoring cancer recurrence in clinical practice. Experimental Section DNA Synthesis and Extraction All gene-specific primers and probes were synthesized and purified by Genewiz (Suzhou, China). Genomic DNA was extracted from 293T and H1299 cells using the Universal Genomic DNA Kit (CWbio, Jiangsu, China) according to the manufacturer’s instructions. H1299 DNA carries a heterozygous NRAS Q61K mutation. Plasmid DNA containing EGFR L858R, T790M, and G719A mutations was synthesized by Genewiz and spiked into 293T genomic DNA at final concentrations of 30 or 60 ng/μL to prepare template samples. DNA concentrations were measured using a DeNovix DS-11 spectrophotometer (DeNovix Inc., Wilmington, DE, USA). Circulating tumor DNA (ctDNA) was extracted from approximately 3 mL of plasma from lung cancer patients using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany). sgRNA Production by PCR, Gel Extraction, and T7 Transcription sgRNA was prepared through sequential steps of PCR, gel extraction, in vitro transcription, and RNA purification. All PCRs were performed on SLAN-96S real-time PCR system (Hongshi Medical Technology, Shanghai, China) in 30 μL reactions. The reaction mixture contained 0.03 U/μL Taq polymerase (TransGen, Beijing, China), 2 mM Mg2+, 0.2 μM dNTPs, 0.25 μM primers, and 10 μL template. The cycling protocol was: 95 °C for 3 min, followed by 50 cycles of 95 °C for 10 s and 55 °C for 20 s. Templates were plasmid DNA (104 copies per reaction) encoding the sgRNA sequence. Primer sequences for generating sgRNAs are listed in Table S1. PCR products (15 μL per sample) were electrophoresed on a 3% agarose gel. Bands corresponding to the expected size were excised and purified using a Gel Extraction Kit (Cwbiotech, Beijing, China). Purified dsDNA was then transcribed in vitro using the HiScribe T7 Quick High Yield RNA Synthesis Kit (New England Biolabs, USA) according to the manufacturer’s instructions. The resulting RNA was purified with an RNA purification kit (Sangon Biotech, Shanghai, China). sgRNA concentrations were determined using a DeNovix DS-11 spectrophotometer. Asymmetric PCR All asymmetric PCRs were performed on the SLAN-96S system in 30 μL reactions containing 1× PCR buffer, 0.03 U/μL Taq polymerase, 2 mM Mg2+, 0.2 μM dNTPs, and 10 μL template. The cycling conditions were: 95 °C for 3 min; 50 cycles of 95 °C for 10 s and 58 °C for 30 s; followed by 95 °C for 10 s and rapid cooling to 10 °C. For sensitivity assays, hairpin primers were used at 0.3–0.4 μM and competitive primers at 0.04–0.1 μM. Multiplex PCR mixtures contained 1× PCR buffer, 0.05 U/μL Taq polymerase, 3.5 mM Mg2+, 0.2 μM dNTPs, and 10 or 5 μL template. Cycling conditions were: 95 °C for 3 min; 55 cycles of 95 °C for 10 s and 58 °C for 30 s; followed by 95 °C for 10 s and rapid cooling to 10 °C. Hairpin/forward primers were used at 0.1–0.3 μM and competitive/reverse primers at 0.05–0.15 μM. Primer sequences are provided in Table S2. Cas14a Detection CRISPR/Cas14a reactions (20 μL) were prepared with 1× reaction buffer, 3 μL synthesized ssDNA or PCR product, 0.04 μM target-specific sgRNA, 0.03 μM Cas14a protein (Tolo Biotech, Shanghai, China), and 0.15 μM fluorescent reporter probe. For ACTB detection, 0.08 μM sgRNA and 0.08 μM Cas14a were used. Reactions were incubated at 42 °C for 40 min on the SLAN-96S system. Synthetic ssDNA was diluted to 0.05 μM in 1× PCR buffer, denatured at 95 °C, and rapidly cooled to 10 °C prior to use. ddPCR Assays ddPCR was performed using the AD3200 Droplet Digital PCR System (Pilot, Zhejiang, China). Each 15 μL reaction mixture consisted of 3 μL reaction mix (containing dNTPs and Taq polymerase) and 5 μL DNA sample, along with 1 μM forward primer, 1 μM reverse primer, and 0.3 μM target-specific TaqMan probe. The mixture was loaded onto a microfluidic chip for droplet generation. Thermal cycling was carried out as follows: initial denaturation at 95 °C for 10 min, followed by 50 cycles of 95 °C for 30 s (denaturation) and 58 °C for 50 s (annealing/extension). After amplification, the chip was placed into the AD3200 Reader (Pilot, Zhejiang, China), and droplet fluorescence was measured at 50 °C. Primer and probe sequences are listed in Table S2. Hairpin Folding and Parameters Calculation Structures of hairpin in amplicons were predicted and parameter of T m, ΔG values were calculated by mfold server at the condition of 50 mM Na+, 2 mM Mg2+ and 58 °C. Results Overview of the Dual Hairpin-Competition CRISPR/Cas14a (DHCC) System A novel CRISPR/Cas14a-based technique, named the dual hairpin-competition CRISPR/Cas14a (DHCC) system (Figure A), was developed for detecting ultralow-frequency mutations in circulating tumor DNA (ctDNA) from blood or urine. This method significantly enhances discrimination between mutant and wild-type DNA through a unique dual hairpin-competition mechanism operating in both PCR amplification and CRISPR/Cas14a detection. The core component of DHCC is a specially designed hairpin primer, consisting of a wild-type flap sequence linked to a primer sequence. During amplification, this flap sequence is incorporated into the amplicons of both mutant and wild-type templates. In single-stranded form, the flap sequence is complementary to the wild-type DNA, enabling the formation of a stable hairpin structure prior to the annealing step. This stable hairpin impedes the binding of a competitive primer, thereby greatly suppressing the amplification of wild-type DNA. In mutant amplicons, however, the mismatch between the wild-type flap sequence and mutant templates results in an unstable hairpin. The competitive primer can thus bind via toehold displacement, unfold the weakly structured hairpin, and efficiently initiate amplification of mutant DNA. Asymmetric PCR was employed to generate single-stranded products for CRISPR/Cas14a detection, with the competitive primer being depleted earlier than the hairpin primer. After PCR, a denaturation step at 95 °C followed by rapid cooling to 10 °C ensured hairpin formation. Schematic of the dual hairpin-competition CRISPR/Cas14a (DHCC) system for detecting ultralow-frequency mutations. (A) Incorporating hairpin structures of differential stability into mutant and wild-type amplicons enables two sequential hairpin-competition steps. The first layer of hairpin competition, taking place during asymmetric PCR, selectively enriches mutant DNA by suppressing wild-type amplification. The second layer of hairpin competition, within the CRISPR/Cas14a detection stage, minimizes nonspecific binding of the single-guide RNA (sgRNA). (B) Schematic illustration of the distinct thermodynamic processes governing sgRNA binding to hairpin versus linear DNA. Hairpin-mediated competition between sgRNA and template DNA improves the discrimination of mutant from wild-type alleles. The first layer of hairpin-competition during PCR thus effectively enriches mutant DNA. Although wild-type amplification is largely suppressed, product of wild-type DNA is still present. However, the stable hairpin structure in wild-type amplicons serves an additional function during the subsequent CRISPR/Cas14a detection: it physically occludes the target sequence, preventing nonspecific binding of mutant-specific sgRNA. In contrast, mutant amplicons retain an unstable hairpin, allowing mutant-specific sgRNA readily access its targets and activates the trans-cleavage activity of Cas14a. This leads to cleavage of a fluorescent reporter and a measurable fluorescence signal. Thus, a second layer of hairpin-competition during CRISPR/Cas14a detection would further minimize off-target recognition between mutant sgRNA and residual wild-type DNA. The thermodynamic basis for this improved discrimination is explained in Figure B, which contrasts sgRNA binding to linear vs hairpin-structured DNA. For linear DNA, both mutant and wild-type sequences hybridize to sgRNA with large negative free energy changes (ΔG a ′ and ΔG b ′ ≪ 0), resulting in high binding efficiency but poor discrimination between mutant and wild-type (the “conversion ratio” is high for both). When the same sequences are folded into hairpins, hybridization to sgRNA is obstructed, raising the free energy changes (now denoted ΔG a for mutant hairpin and ΔG b for wild-type hairpin) and lowering the conversion ratio. Importantly, the obstruction effect is much stronger for the wild-type hairpin because its stable structure is difficult to unfold. Specifically, the free energy change for wild-type hairpin hybridization (ΔG b = ΔG 4 – ΔG 3) becomes positive, placing it in the low-conversion regime (Figure B). For the mutant hairpin, although the free energy change increases relative to the linear form (ΔG a = ΔG 2 – ΔG 1), it remains negative because the mismatch destabilizes the hairpin, making it easier for sgRNA to invade. Consequently, the presence of hairpin creates a large differential in binding efficiency between mutant and wild-type, thereby substantially enhancing the discrimination ability. Together, the two sequential hairpin-competition reactionsfirst enriching mutations during PCR and then enhancing specificity during CRISPR detectionenable highly sensitive identification of mutations at extremely low allele frequencies. Furthermore, multiple mutation targets can be coamplified in a single PCR and subsequently distinguished using target-specific sgRNAs in separate detection reactions. This multiplexed preamplification approach significantly increases assay throughput while conserving limited sample materiala critical advantage for clinical applications. Enhanced Mutation Discrimination in CRISPR/Cas14a through Hairpin Competition We first examined how hairpin competition affects sgRNA recognition in the CRISPR/Cas14a detection system using synthesized linear and hairpin DNA templates. The sequences of wild-type and mutant DNA (EGFR L858R) are shown in Figure A, with the sgRNA-binding region highlighted in blue. As shown in Figure B, linear wild-type DNA generated high fluorescence due to nonspecific sgRNA binding, resulting in a discrimination factor (DF) of only 1.96. By contrast, when the same wild-type sequence was folded into a hairpin, fluorescence was significantly reduced, and the DF improved markedly to 5.37 (Figure C). These results indicate that hairpin competition effectively suppresses nonspecific sgRNA binding and enhances detection specificity in CRISPR/Cas14a assays, as illustrated in Figure B. Hairpin competition enhances mutation discrimination in CRISPR/Cas14a by reducing nonspecific sgRNA binding and enriching mutant DNA. (A) Design of mutant and wild-type DNA targeting the EGFR L858R mutation, with or without a hairpin structure. The sgRNA-binding region is underlined in blue. (B) Fluorescence curves from CRISPR/Cas14a detection of linear DNA. The discrimination factor (DF) was calculated as the ratio of fluorescence increase from mutant DNA to that from wild-type DNA: DF = (F M – F M0)/(FW – F W0), where F M and F W are the fluorescence signals from mutant and wild-type templates measured at 40 min, and F M0 and F W0 are the respective background fluorescence intensities recorded during the first minute of the reaction. (C) Fluorescence curves from CRISPR/Cas14a detection of hairpin-containing DNA. Schematic and fluorescence curves of three CRISPR/Cas14a assay formats: no hairpin (D), single hairpin-competition (E), and dual hairpin-competition assays (F). Samples with 50%, 1%, and 0% variant allele frequency (VAF), as well as a no-template control (NC) were detected by the three different CRISPR/Cas14a assays. All fluorescence detection assays were performed in duplicate. These samples were prepared by spiking mutant DNA into wild-type DNA (30 ng/μL 293T genomic DNA); the 1% VAF sample contained 900 copies of mutant DNA and 89,100 copies of wild-type DNA. Next, we compared three assay configurations: no hairpin competition, single hairpin competition, and dual hairpin competition (Figure D–F). The no-hairpin assay used asymmetric PCR to generate single-stranded products for sgRNA binding. The single hairpin-competition assay introduced a hairpin into the PCR product (without affecting primer binding during PCR). Therefore, this format provides only the second layer of competitionnamely, competition between sgRNA and the hairpin ampliconwhile lacking the mutation enrichment step during PCR. This allows us to isolate the contribution of hairpin-mediated sgRNA occlusion from that of PCR-based enrichment. In the dual hairpin-competition assay, competitive primers target the hairpin region to enrich mutant DNA, while the sgRNA subsequently binds to the hairpin product to further minimize nonspecific recognition. Without hairpin competition, the 0% variant allele frequency (VAF) sample produced fluorescence similar to the 1% VAF sample, yielding a DF of 2.48 and reflecting poor discrimination between wild-type and mutant DNA (Figure D). Introducing single hairpin competition improved differentiation: the 1% VAF sample showed distinct fluorescence compared to the 0% VAF control, with a DF of 11.7 (Figure E). Remarkably, under dual hairpin competition, the fluorescence signal from the 1% VAF sample increased substantially, while that from the 0% VAF sample dropped to a level comparable to the no-template control (NC). This combination produced an exceptional DF of 145, far exceeding the values from the other two formats (Figure F). Thus, dual hairpin competition achieves a 58-fold improvement in discrimination relative to the no-hairpin assay (145/2.48 ≈ 58). These findings highlight the superior discrimination ability of the dual hairpin-competition design and its strong potential for detecting ultralow-frequency genetic variants. Optimization of DHCC System To systematically optimize the CRISPR/Cas14a detection system, a panel of linear and hairpin-structured fluorescence probes was designed and synthesized. As illustrated in Figure A, linear probes LP1 and LP2 contain 12 and 20 nucleotides, respectively, while hairpin probes HP1, HP2, and HP3 feature loop lengths of 6, 12, and 20 nt. In CRISPR/Cas14a assays, HP3 consistently generated the strongest fluorescence signal among all probes evaluated (Figure B). Among the hairpin probes, fluorescence intensity increased progressively with loop length. Notably, although LP2 and HP3 share the same 20 nt linear sequence, HP3 generated a substantially greater fluorescence increase and a higher signal-to-noise ratio (Figure B). Based on its superior performance, HP3 was selected as the fluorescent probe for all subsequent DHCC assays. Optimizing DHCC by investigating fluorescence probes, hairpin primer and competitive primer-to-hairpin primer ratio. (A) Five linear and hairpin-structured fluorescence probes were designed. (B) CRISPR/Cas14a assays using the five probes. (C) Various hairpin produced by hairpin primers targeting the EGFR T790 M mutation. ΔG (D) and T m (E) for these hairpins in wild-type and mutant DNA were calculated and presented. (F) Fluorescence intensities for samples with 50%, 0.005%, and 0% VAF measured using these different hairpin primers. (G) Detection of samples with 50%, 0.005%, and 0% VAF using varying concentrations of competitive primers. All fluorescence detection assays were performed in duplicate. Next, optimization of DHCC focused on hairpin length and competitive primer-to-hairpin primer ratio. Hairpin primers for EGFR T790 M were designed with flap sequences ranging from 10 to 16 nucleotides, producing amplicons capable of forming hairpins with stem lengths between 10 and 17 base pairs in mutant and wild-type DNA. The toehold region for competitive primer binding, highlighted in pink, correspondingly varied from 11 to 8 nucleotides (Figure C). The calculated thermodynamic parameters (ΔG and T m) for each hairpin are shown in Figure D,E. Due to the sequence mismatch, mutant amplicons form unstable hairpin structures, exhibiting higher ΔG and lower T m values than their wild-type counterparts. A clear correlation was observed between increased hairpin length and enhanced stability, as evidenced by the less negative ΔG values for both wild-type (from −3.47 to 8.87 kcal/mol) and mutant DNA (from 0.11 to −5.2 kcal/mol). This enhanced stability was consistently mirrored by a corresponding increase in T m values. As shown in Figure F, the 10 bp hairpin failed to identify the 0.005% variant allele frequency (VAF) sample, as its fluorescence signal was indistinguishable from that of the 0% VAF control. Comparative analysis revealed that the 12 bp hairpin provided superior discrimination of the 0.005% VAF sample compared to the 14 bp variant. In contrast, the 17 bp hairpin yielded no detectable fluorescence across any sample, indicating failed amplification (Figure F). These findings underscore the necessity of optimizing hairpin stability for low-frequency mutation detection; excessively stable structuressuch as those with T m values exceeding 70 °C for both wild-type and mutant DNA (Figure E)can preclude efficient reaction. A consistent trend was observed for the EGFR L858R mutation, where the 14 bp hairpin achieved optimal discrimination, while the 16 bp hairpin failed to generate a signal (Figure S1A–D). To further optimize assay performance, we evaluated varying concentrations of competitive primers, thereby modulating the competitive primer-to-hairpin primer ratio to promote the efficient generation of single-stranded DNA during asymmetric PCR (Figure G).The results revealed that 0.1 μM competitive primers (corresponding to a competitive primer-to-hairpin primer ratio of 1:4) yielded the most pronounced fluorescence differential between the 0.005% and 0% VAF samples, indicating optimal discrimination (Figure G). Parallel optimization for the EGFR L858R mutation identified 0.05 μM competitive primers (competitive primer-to-hairpin primer ratio of 1:6) as the optimal condition (Figure S1E). Further fine-tuning of sgRNA and hairpin probe concentrations was also performed to maximize assay sensitivity and specificity (Figure S2A,B). Evaluation of Ultra-Sensitive Detection Performance of DHCC To determine the detection limit of the optimized DHCC system, we performed sensitivity analyses using samples with serially diluted variant allele frequencies (VAFs). The dilution series was prepared by titrating mutant DNA into a fixed background of wild-type 293T genomic DNA. For the EGFR T790 M mutation, even a sample with a VAF as low as 0.002% exhibited significantly higher fluorescence than the 0% VAF control (Figure A), corresponding to the detection of approximately 3.6 copies of mutant DNA among about 179,996.4 copies of wild-type DNA. The fluorescence curve of the 0% VAF sample closely matched that of the no-template control (NC), demonstrating minimal background from wild-type DNA. This ultrahigh sensitivity and excellent specificity are attributed to the unique dual-hairpin competition mechanism. Ultrasensitive detection performance of DHCC for diverse mutations. Fluorescence curves for serially diluted variant allele frequency (VAF) samples targeting (A) EGFR T790M, (B) EGFR L858R, (C) EGFR G719A, and (D) NRAS Q61K mutations. Statistical significance was assessed by unpaired t test (* p < 0.05, ** p < 0.01, *** p < 0.001). All samples were prepared by spiking mutant DNA into wild-type genomic DNA (60 ng/μL 293T). All fluorescence detection assays were performed in duplicate. (E) Summary of optimized reaction parameters for DHCC assays targeting each of the four mutations. Beyond EGFR T790M, DHCC consistently achieved a remarkable sensitivity of 0.002% VAF for other clinically relevant mutations, including EGFR L858R, G719A, and NRAS Q61K (Figure B–D). Despite differences in nucleotide changes and flanking sequences, DHCC maintained robust performance across all targets, highlighting its potential as a universal platform for ultrasensitive mutation detection. The optimized reaction parameters for each of the four mutation targetsincluding hairpin length, corresponding T m values, competitive primer-to-hairpin primer ratio and DHCC amplicon sizeare summarized in Figure E. Across these targets, the hairpin length employed in DHCC ranged from 12 to 16 bp, while the competitive primer-to-hairpin primer ratio varied between 1:3 and 1:8. Notably, the short DHCC amplicon (45–62 bp, excluding the introduced flap region) is ideally suited for the robust detection of highly fragmented ctDNA, a critical feature for liquid biopsy applications where DNA fragmentation is prevalent. These empirically derived parameters provide a practical framework for designing DHCC assays tailored to different mutation loci. Development of Multiplexed Preamplification DHCC and Application in Plasma ctDNA Samples The nonspecific cleavage of fluorescent reporters by the trans-cleavage activity of Cas14a limits the ability to identify multiple mutations within a single Cas14a reaction. However, multiplexed amplification can be performed in a single tube, with subsequent detection of different mutations carried out in separate Cas14a tubes. This multiplexed preamplification strategy not only increases target throughput but also conserves precious sample input, which is particularly important for clinical applications. Here, we established a multiplexed preamplification DHCC assay, in which mutation targetsincluding EGFR L858R, G719A, and NRAS Q61Kalong with a reference gene were coamplified in one tube and then detected individually (Figure A). The sensitivity of multiplexed preamplification DHCC for each mutation was evaluated independently. Results showed that EGFR L858R and G719A mutations could be detected at variant allele frequencies (VAFs) as low as 0.005%, while NRAS Q61K was detectable at 0.01% VAF (Figure B). Compared to the single-plex DHCC assay, which achieved 0.002% VAF sensitivity, a modest reduction in sensitivity was observed in the multiplexed formatlikely due to competition among primer pairs and limited reaction resources. Nevertheless, the detection limits remained within the ultrahigh sensitivity range. Importantly, good amplification balance was achieved, as demonstrated in reactions containing 50% VAF samples, where the fluorescence signals for ACTB and the three mutation targets consistently reached similar levels of approximately 6000 arbitrary units. Original fluorescence curves for these mutations are provided in Figure S3A–C. Application of multiplexed preamplification DHCC for mutation detection in ctDNA samples. (A) Schematic of multiplexed preamplification DHCC: EGFR L858R, G719A, NRAS Q61K mutations and a reference gene were coamplified and subsequently detected in separate reactions. (B) Sensitivity of multiplexed preamplification DHCC for different mutations. Statistical significance was assessed by unpaired t test (* p < 0.05, ** p < 0.01, *** p < 0.001). (C) Workflow for analysis of plasma ctDNA samples from lung cancer patients using DHCC and ddPCR. (D) DHCC fluorescence signals were measured for three target mutationsEGFR L858R, G719A, and NRAS Q61Kalongside the internal reference gene ACTB in ctDNA samples. MT-1, MT-2, and MT-3 denote simulated clinical samples harboring EGFR L858R, G719A, and NRAS Q61K mutations, respectively. Mutation-positive status was assigned according to the signal-to-noise ratio (signal-to-noise ratio = Increased fluorescence/background fluorescence), with thresholds of >3 for EGFR L858R, > 5 for EGFR G719A, and >3.2 for NRAS Q61K. These thresholds were determined based on sensitivity evaluation assays, in which the signal-to-noise ratio for the 0.005% VAF sample of EGFR L858R, the 0.005% VAF sample of EGFR G719A, and the 0.01% VAF sample of NRAS Q61K were 2.9, 4.9, and 3.1, respectively. Background fluorescence was the intensity recorded during the first minute, and increased fluorescence was the intensity at 40 min minus the background. (E) Comparison with ddPCR results for EGFR L858R detection. Next, 22 plasma ctDNA samples from lung cancer patients were analyzed using the developed multiplexed preamplification DHCC (Figure C). Fluorescence signals revealed that four ctDNA samples were positive for EGFR L858R mutations, whereas no samples carried EGFR G719A or NRAS Q61K mutations (Figure D). The corresponding raw fluorescence curves are shown in Figure S4. Notably, samples positive for EGFR L858R did not exhibit fluorescence signals for EGFR G719A or NRAS Q61K (Figures D, and S4), demonstrating high assay specificity. To confirm the accuracy of EGFR L858R detection, ddPCR was performed on the same samples, yielding results consistent with those from DHCC (Figure E). Furthermore, the Variant Allele Frequency (VAF) for EGFR mutations in the positive samples was determined by ddPCR, ranging from 0.48% to 14.1% (Figure E). Collectively, these findings underscore the reliability and precision of DHCC for identifying low-frequency mutations in plasma ctDNA samples. Discussion The detection of low-frequency single-nucleotide variants in plasma or urine circulating tumor DNA (ctDNA) remains a significant technical challenge, primarily due to the overwhelming background signal from abundant wild-type DNA. To overcome this limitation, we have developed a novel CRISPR/Cas14a-based strategy, termed the dual hairpin-competition CRISPR/Cas14a (DHCC) system, which integrates a unique two-stage competitive mechanism. This design achieves a 58-fold improvement in discrimination factor (from 2.48 to 145) and a 250-fold enhancement in sensitivity compared to previous Cas14a-based methods, , enabling routine detection of mutations at 0.002% variant allele frequency (VAF). The superior performance of DHCC stems from its integrated dual-layer hairpin competition strategy. In the first layer, engineered hairpin primers and competitive primers selectively suppress wild-type amplification while enriching mutant sequences during asymmetric PCR. This enrichment step alone elevates the effective mutation frequency to a level detectable by CRISPR/Cas14a. In the second layer, the retained hairpin structure in wild-type amplicons physically obstructs sgRNA binding, further minimizing nonspecific signal during Cas14a detection. Direct biophysical characterization of hairpin formation and primer displacement kinetics will be pursued in future work to further validate the proposed two-layer competition mechanism. While some PAM-independent CRISPR/Cas12a-based strategies have been developed to detect low-frequency mutationssuch as through elevated reaction temperature, single-stranded DNA generation, toehold-activator introduction, or sticky-end dsDNA production using NlaIIIthese approaches are often constrained by intricate reaction mechanisms or still demonstrate limited discrimination capability for single-nucleotide variants. In comparison to existing CRISPR/Cas-based detection platforms, DHCC offers three primary advantages: (1) Ultrahigh sensitivity: The dual hairpin-competition mechanism reliably detects mutations at VAFs down to 0.002%. (2) Flexible assay design: Detection is independent of protospacer-adjacent motif (PAM) constraints. (3) Multiplexed preamplification capability: Multiple targets can be coamplified within a single reaction and then detected in separate Cas14a reactions. When benchmarked against established methodologies such as droplet digital PCR (ddPCR) and next-generation sequencing (NGS), DHCC further demonstrates distinct practical benefits, including lower cost, shorter turnaround time, a simpler workflow and minimal reliance on sophisticated instrumentation (Table ). Collectively, these attributes position DHCC as a highly promising platform for the detection of low-frequency mutations, with significant potential to inform therapeutic strategies and monitor cancer recurrence in clinical practice. It is important to acknowledge, however, that DHCC currently does not support absolute quantificationa key feature offered by ddPCR and its multiplexing capacity is limited compared to NGS, which can identify hundreds of mutations simultaneously (Table ). 1Comparison of DHCC, ddPCR, and NGS for Low-Frequency Mutation DetectionparameterDHCC (this work)ddPCRNGS (targeted panel)Turnaround time<6 h∼1 day∼2–5 daysCostLowModerateHighInstrumentationStandard qPCR instrumentDroplet generator + special readerSequencer + bioinformatics infrastructureMultiplexing scope∼4–10 targets (coamplification, separate detection)2–4 colors (single reaction)Hundreds to thousandsSensitivity (VAF)0.002% (single-plex); 0.005–0.01% (multiplex)0.01–0.1%0.1–1% (standard); 0.01% (ultradeep)Absolute quantificationNoYesSemiquantitativeWorkflow complexityLowModerateHigh Comparison of DHCC, ddPCR, and NGS for Low-Frequency Mutation Detection Several improvements to the DHCC system are worth pursuing in future work. First, its two-step sequential format requires tube opening between stages, introducing a potential risk of cross-contamination. To enhance its suitability for clinical and routine laboratory use, future development should focus on transitioning DHCC toward a closed-tube or integrated microfluidics-based system, which would eliminate manual handling and substantially mitigate contamination risks. Second, polymerase fidelity is a critical factor in ultralow-frequency mutation detection. Although the Taq polymerase employed in our study did not produce false-positive calls, adopting a high-fidelity DNA polymerase would further minimize artifacts arising from misincorporation. Third, in this work we have provided empirically determined design parameters for DHCC, including a hairpin stem length of 12–16 bp and a competitive primer-to-hairpin primer ratio of 1:3 to 1:8. However, a unified, mechanistically predictive model capable of precomputing optimal hairpin parameters solely from sequence information has not yet been established. Establishing such a generalized design guideline for DHCC therefore remains a key goal of our future work. Finally, in this work we did not systematically evaluate cross-reactivity with neighboring mismatches or alternative clinically relevant variants within the same hotspot region. Although the hairpin-competition mechanism is expected to provide thermodynamic discrimination against nearby mismatches, experimental validation is still required. Future work will include cross-reactivity assays to confirm the specificity of DHCC for intended hotspot mutations, or to guide the development of improved versions tailored for more challenging hotspot regions. Supplementary Material