MUTE‐Seq: An Ultrasensitive Method for Detecting Low‐Frequency Mutations in cfDNA With Engineered Advanced‐Fidelity FnCas9

Abstract

In this study, we present the development of the Mutation tagging by CRISPR‐based Ultra‐precise Targeted Elimination in Sequencing (MUTE‐Seq) method. We engineered a highly precise advanced‐fidelity FnCas9 variant, named FnCas9‐AF2, to effectively discriminate single‐base mismatches at all positions of the single guide RNA (sgRNA) target sequences. FnCas9‐AF2 exhibited significantly lower off‐target effects compared to existing high‐fidelity CRISPR‐Cas9 variants. MUTE‐Seq leverages FnCas9‐AF2 for the enrichment of mutant DNA through the exclusive cleavage of perfectly matched wild‐type DNA, allowing for sensitive detection of low‐frequency cancer‐associated mutant alleles. MUTE‐Seq enabled sensitive monitoring of minimal residual disease (MRD) from the bone marrow of patients with Acute Myeloid Leukemia (AML). Furthermore, MUTE‐Seq was applied in a multiplexed manner on cell‐free DNA (cfDNA) from patients diagnosed with non‐small cell lung cancer (NSCLC) and pancreatic cancer. This approach demonstrated a significant improvement in the sensitivity of simultaneous mutant detection and highlighted its clinical utility for early‐stage cancer patients with extremely low levels of circulating tumor DNA (ctDNA). We anticipate that the FnCas9‐AF2‐based MUTE‐Seq could offer a valuable clinical tool to facilitate improved molecular diagnosis, prognosis evaluation, and treatment planning for cancers in various stages.

An advanced‐fidelity CRISPR nuclease, FnCas9‐AF2, is rationally engineered to discriminate single‐base mismatches with unprecedented level precision. Integrated into the MUTE‐Seq workflow, FnCas9‐AF2 depletes wild‐type cell‐free DNA, thereby exposing rare tumor‐derived mutant DNA (a molecular “needle in a haystack”). The method enables ultrasensitive, cost‐efficient detection of low‐frequency mutations for liquid‐biopsy cancer diagnostics, supporting early detection, companion diagnosis, and MRD monitoring.

1. Introduction

Cell‐free DNA (cfDNA) comprises highly fragmented DNA molecules that are released into the circulatory system from cells.^{[ 1} , ² , ^{3 ]} Circulating tumor DNA (ctDNA) specifically refers to the cfDNA that originates from tumor cells.^{[ 4} , ^{5 ]} During the early stages of cancer, ctDNA in the blood can be exceedingly minute (as low as 0.01% of total cfDNA), posing a significant challenge for detection.^{[ 6} , ⁷ , ⁸ , ^{9 ]} Contemporary diagnostic techniques utilizing Next‐Generation Sequencing (NGS) often struggle with sensitivity, frequently failing to detect minute levels of ctDNA below 0.1%.^{[ 9} , ^{10 ]} Since low levels of ctDNA are common for many early‐stage cancers and minimal residual disease (MRD), it has been difficult to obtain reliable detection results for such cases.^{[ 9} , ¹⁰ , ¹¹ , ^{12 ]}

CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) based ctDNA enrichment methods have shown promising enhancement of detection sensitivity.^{[ 13} , ¹⁴ , ¹⁵ , ¹⁶ , ^{17 ]} The methods involved the depletion of regular major alleles prior to conducting NGS, thereby enriching for the minor alleles. Innovations in this domain include the highly efficient CRISPR‐based sequence depletion methods, including CARM (Cas9 Assisted Removal of Mitochondrial DNA),^{[ 13 ]} DASH (Depletion of Abundant Sequences by Hybridization),^{[ 14 ]} MAD‐DASH(miRNA and Adaptor Dimer‐DASH)^{[ 15 ]} and CUT‐PCR (CRISPR‐mediated, Ultrasensitive detection of Target DNA using PCR).^{[ 16 ]} These methods facilitated the enrichment of minor allelic DNA by depleting major allelic DNA, thus enabling accurate detection of minor DNA with fewer NGS reads, suggestive of immense potential in the field of molecular diagnostics.

However, while these methods provided unprecedented effectiveness in selectively enriching the sequences of interest in NGS libraries, the applicability of the current CRISPR‐based mutant detection methods was limited by the specific requirements.^{[ 13} , ¹⁴ , ¹⁵ , ^{16 ]} The previous studies demonstrated that the eliminations required the sequences to contain significant sequence differences, or the mutations to be positioned within the protospacer adjacent motif (PAM) site. A primary reason for this is that the specificities of the current CRISPR systems have been shown to be insufficient to effectively distinguish single‐ or double‐base mismatches in the target DNA sequence.^{[ 18} , ¹⁹ , ^{20 ]} These inaccuracies in the CRISPR cleavage events constituted a significant barrier, both in vivo and in vitro, for medical and industrial applications that required exact base pair discrimination.^{[ 16} , ¹⁷ , ²¹ , ^{22 ]} The sequence limitation has been a major hindrance in applying CRISPR‐based enrichment for diagnosis through detecting rare mutant alleles in cfDNA.

Several studies reported enhanced CRISPR systems with improved specificity, such as eSpCas9 and SpCas9‐HF.^{[ 18} , ^{20 ]} However, while these engineered CRISPRs showed considerably higher accuracies compared to wild‐type SpCas9 (SpCas9‐WT), significant cleavages still occurred sporadically at off‐targets that had single‐base pair mismatches. To overcome this precision issue, we sought to develop a CRISPR system capable of effectively distinguishing a single‐base mismatch across all the sgRNA target sequences. To this end, we selected Francisella novicida Cas 9 (FnCas9), which is known to be one of the more precise Cas9s investigated in previous studies,^{[ 23 ]} and we developed an FnCas9 with advanced fidelity (FnCas9‐AF2), that could efficiently discriminate mutations across all positions of a sgRNA target sequence. In vitro cleavage and genome‐wide analyses demonstrated that FnCas9‐AF2 had negligible off‐target activity.

Next, we applied FnCas9‐AF2 to develop a CRISPR‐based sequence depletion approach for detecting low‐frequency cancer‐associated mutations, and we named it MUTE‐Seq (Mutation tagging by CRISPR‐based Ultra‐precise Targeted Elimination in Sequencing). Then we observed that MUTE‐Seq markedly increased the detection sensitivities of low‐frequency mutant alleles via wild‐type allele depletion in biopsy samples from acute myeloid leukemia (AML), non‐small cell lung cancer (NSCLC), and pancreatic cancer patients. MUTE‐Seq significantly increased the detection rates of low‐frequency NRAS mutations in AML patients monitored for minimal residual disease (MRD). The increment of detected mutant allele frequencies was apparent in both chromatograms in Sanger sequencing and NGS. Subsequently, we conducted multiple sgRNA‐based simultaneous depletions of wild‐type alleles in cfDNA using FnCas9‐AF2, revealing that the system could be utilized in a multiplexed manner. Multiplexed MUTE‐Seq significantly enhanced the concordance of detected mutations between tissue and blood samples from NSCLC and pancreatic cancer patients. Notably, MUTE‐Seq was also effective for very low‐frequency mutations in early‐stage NSCLC and pancreatic cancer patients. The findings suggested that the MUTE‐Seq method has considerable potential for developing diagnosis panels aimed at detecting multiple low‐frequency ctDNA.

2. Results

2.1. Comparison of Single‐Base Mismatch Discrimination between SpCas9, SpCas9 Variants, and FnCas9

We sought to examine the mismatch tolerance of SpCas9, high‐fidelity SpCas9 variants, and FnCas9. As a part of the study, 20 single guide RNAs (sgRNAs) were prepared with target and single‐base mismatches, each at a different position of the KRAS target sequence (Figure 1A). These were then tested in in vitro cleavage assays using SpCas9‐WT, SpCas9‐HF1, SpCas9‐HF4, eSpCas9(1.0), eSpCas9(1.1), and FnCas9‐WT.^{[ 18} , ^{20 ]}

Figure 1.

Enhancing the Base Mismatch Sensitivity and Specificity of FnCas9 Through Protein Engineering. A) Sequences of sgRNAs targeting a KRAS site with a single mismatch and in vitro cleavage efficiencies of SpCas9‐WT, SpCas9‐HF1, SpCas9‐HF4, eSpCas9(1.0), eSpCas9(1.1), and FnCas9‐WT proteins. B) Positions of amino acids within the crystal structure of FnCas9 (PDB 5B2O) that were changed in the 11 highly precise mutants; six with modified residues in the Rec and PLL domain (R455, R785, K721, K789, R919, and R1241) that interacted with the phosphate backbone of the target DNA strand, and five with modified residues in the HNH, RuvC and PLL domain (R939, K941, K1189, R1226, K1228) that interacted with the phosphate backbone of the non‐target DNA strand. C) Quantification of target cleavage efficiencies of the 11 FnCas9 variants with single amino acid substitution. (T: On‐target sequence; Numbers: Single‐base mismatch positions counted from PAM).

SpCas9‐WT induced significant DNA cleavage not only with perfectly matched sgRNA, but also with sgRNAs that contain single‐base mismatches in all 20 positions. SpCas9‐HF1, SpCas9‐HF4, eSpCas9(1.0), and eSpCas9(1.1), the engineered variants of SpCas9 designed for higher precision, also exhibited a noticeable cleavage for most mismatched sgRNAs. However, FnCas9 presented a tendency for lower rates of in vitro cleavage with mismatched sgRNAs. Together, the in vitro cleavage assays suggested that FnCas9 could potentially discern single‐base mismatches more efficiently than SpCas9 and its high‐fidelity variants (Figure 1A; Figure S1, Supporting Information).

2.2. Engineering FnCas9 for Enhanced Sensitivity to DNA Mismatches

Following the initial studies, the aim was to engineer the FnCas9 protein to produce optimized FnCas9 variants that would exhibit enhanced sensitivity to mismatches. The hypothesis was that by mitigating the interactions between positively charged amino acids and the nucleotide implicated in conformational shift during Cas9‐sgRNA‐DNA binding, it might be possible to destabilize the R‐loop formation in the context of mismatched DNA. This potential instability could subsequently curtail nuclease activity directed toward the mismatched DNA sequences.^{[ 18} , ²⁰ , ²⁴ , ^{25 ]} To examine this, 49 recombinant FnCas9 proteins were prepared, each containing a single amino acid substitution to alanine in positions expected to interact with the phosphate backbone of the target DNA (Figure S2A, Supporting Information). Subsequently, in vitro cleavage assays were carried out to test the ability of these 49 FnCas9 single‐substitution variants to cleave the target sequence with sgRNAs, each bearing a single‐base mismatch at a different position. The assays revealed that certain amino acid substitutions in FnCas9 led to a significant reduction of cleavage with the single‐base mismatched sgRNAs, while preserving on‐target cleavage activity (Figure S2B, Supporting Information).

To identify the most precise FnCas9 variant, their specificity scores were calculated by their abilities to discriminate single‐base mismatches (Figure S2C, Supporting Information). Eleven variants demonstrated specificity scores above 60%: six with modified residues in the Rec and PLL domain (R455, R785, K721, K789, R919, and R1241) that interacted with the phosphate backbone of the target DNA strand, and five with modified residues in the HNH, RuvC and PLL domain (R939, K941, K1189, R1226, K1228) that interacted with the phosphate backbone of the non‐target DNA strand (Figure 1B). Of these 11 variants, it was found that the K1189A and R1241A mutants presented the highest specificity scores, leading to further optimization of FnCas9 based on these mutations (Figure 1C; Figure S2C, Supporting Information).

2.3. Multiple Amino Acid Substitutions Enhance Specificity of FnCas9 Across Multiple Target Sites and Mismatch Conditions

We sought to ask if the specificities of the FnCas9 variants could be further enhanced. We observed that the single substitution variants showed residual off‐target cleavage at the KRAS target site. In order to improve the specificity, we conducted combinatorial alanine substitutions of FnCas9 at the identified amino acid positions and evaluated the variants for specificity at both the KRAS and NRAS target sites (Figure 2A,B). As a result, two FnCas9 variants (FnCas9‐K1189A/R1241A [termed FnCas9‐AF1], FnCas9‐R785A/K1189A/R1241A [termed FnCas9‐AF2]) demonstrated undetectable off‐target cleavages while preserving on‐target activities for both target sites.

Figure 2.

Comparative Cleavage Efficiencies and Specificities of FnCas9 Variants against Single‐Mismatched sgRNA Targets. A) Assessment of specificity of FnCas9 variants with single, double, and triple mutations. Heatmaps show cleavage efficiencies of FnCas9 variants with single‐mismatched sgRNAs against a KRAS target. B. Specificity assessment of FnCas9 variants against an NRAS target. C. Cleavage efficiencies of FnCas9 and FnCas9‐AF2 (R785A, K1189A, R1241A triple mutant) for sgRNAs targeting the KRAS with single‐base mismatches at all 20 nucleotide positions. D. Cleavage rates of FnCas9 and FnCas9‐AF2 for sgRNAs with single‐base mismatches at all 20 positions of the NRAS target. The perfect base‐pair between the sgRNA and the target DNA in each nucleotide position is marked by the cross mark. The cleavage rates are shown in shades of blue.

The specificity of these variants was further assessed using 60 sgRNAs containing all possible single‐base mismatches at all 20 positions within the target sequences. FnCas9‐AF1 exhibited superior specificity to the NRAS target sequence in comparison to the K1189A single substitution (Figure S4, Supporting Information). FnCas9‐AF2 showed significantly reduced cleavage rates for all mismatched base positions and base mismatch types for KRAS and NRAS target sequences (Figure 2C,D). In summary, the specificities of FnCas9‐AF2 with the KRAS and NRAS sgRNAs were significantly higher than FnCas9‐WT, suggesting that its sensitivity to single‐base mismatches is applicable for multiple target sequences.

2.4. Comparative Specificity Assessment Between FnCas9‐AF2 and SpCas9 Variants

We further evaluated the precision of FnCas9‐AF2 on various targets, with clinical implications, comprising single nucleotide variants (SNVs) or indel such as EGFR p.L858R(c.2573 T>G), EGFR p.T790M(c.2369 C>T), EGFR p.C797S(c.2389 T>A), KRAS p.G12D(c.35 G>A), MET p.Ex14 skipping(c.3028+1 G>A), and EGFR p.749_750 del(c.2236_2250 del). These mutational positions are hotspots where mutations are detected in significant proportions of cancer patients, and notably, mutations at MET c.3028G position comprise more than 95% of MET exon 14 skipping cases of NSCLC.^{[ 26} , ^{27 ]} For each site, we designed sgRNAs targeting the wild‐type sequences and carried out in vitro cleavage on both the wild‐type and mutant DNAs (Figure 3A). The wild‐type DNA was completely cleaved by FnCas9‐AF2 and was observed as two DNA fragments on the gel. However, in the case of the mutant DNA, FnCas9‐AF2 did not produce detectable cleavage in targets with both indel and single‐nucleotide variants. We performed a quantitative analysis of the in vitro cleavage efficiency using FnCas9‐AF2 and SpCas9 variants on both wild‐type and mutant targets, including the 5 SNVs and 1 indel (Figure 3B). When compared to the SpCas9 variants, FnCas9‐AF2 showed an extremely low cleavage rate of mutant DNA (0.21% in average) while preserving highly efficient cleavage activity on wild‐type DNA (96.84% in average) (Figure 3B; Figure S5, Supporting Information).

Figure 3.

Genome‐wide assessment of effective discrimination of mutations by FnCas9 and SpCas9 Variants. A) In vitro cleavage efficiency of FnCas9‐AF2 against wild‐type and mutant DNA targets. SNVs (EGFR c.2573 T>G, EGFR c.2369 C>T, EGFR c.2389 T>A, KRAS c.35 G>A, MET c.3028+1 G>A) and an indel variant (EGFR c.2236_2250 del) were assessed. Cleavage of wild‐type DNA by FnCas9‐AF2 resulted in two distinct DNA fragments, while mutant‐type DNA remained intact, illustrating FnCas9‐AF2's specificity for wild‐type over mutant DNA targets. B) Comparative in vitro cleavage efficiencies of FnCas9‐AF2, eSpCas9(1.1), and SpCas9‐HF4. While SpCas9 variants show variable levels of cleavage efficiencies, FnCas9‐AF2 demonstrates consistent, high cleavage rates of wild‐type DNA (96.84% in average) and markedly low cleavage rates for mutant DNA (0.21% in average). Data for independent biological replicates (n = 3) are depicted by black dots, and error bars represent SEM. C) Genome‐wide analyses of off‐target cleavage sites were conducted with Digenome‐seq for SpCas9‐WT, eSpCas9(1.1), SpCas9‐HF4, FnCas9‐WT, FnCas9‐AF1, and FnCas9‐AF2, using NRAS‐targeting sgRNA. The graphs show the genomic locations of off‐target sites cleaved by the Cas9 variants. D) Genome‐wide multiplex analyses of off‐target cleavage sites were conducted with Multiplex Digenome‐seq for SpCas9‐HF4 and FnCas9‐AF2, using gRNAs targeting five genomic locations (FANCF, NRAS, AAVS1, HBB, DMD). The graphs show the genomic locations of off‐target sites cleaved by the Cas9 variants.

2.5. Genome‐Wide Analysis of Off‐Target Effects in FnCas9 and SpCas9 Variants

We then sought to determine whether the higher sensitivity of FnCas9‐AF2 is associated with reduced off‐target effects in the whole genome by Digenome‐seq analyses^{[ 28 ]} using NRAS‐targeting sgRNA (Figure 3C). While SpCas9‐WT exhibited 654 potential off‐target sites, FnCas9‐WT had 77 potential off‐target cleavage sites, underscoring FnCas9's higher specificity. The high‐fidelity SpCas9 variants further minimized off‐target cleavage sites, with eSpCas9(1.1) registering 37 sites and SpCas9‐HF4 having 13. Remarkably, FnCas9‐AF1 presented only one potential off‐target cleavage site, while FnCas9‐AF2 showed even higher precision with zero detectable off‐target sites. Furthermore, we conducted multiplex Digenome‐seq analysis using sgRNAs targeting an additional five genomic locations^{[ 25 ]} —FANCF, NRAS (different location from previous analysis), AAVS1, HBB, and DMD simultaneously (Figure 3D). This analysis was performed for SpCas9 and FnCas9 variants, which were identified as having the highest fidelity. The results indicated that SpCas9‐HF4 was associated with 383 potential off‐target sites, whereas FnCas9‐AF2 had only six potential off‐target sites—none associated with FANCF, NRAS, and DMD, two related to HBB‐targeting sgRNA, and four to AAVS1‐targeting gRNA. These findings demonstrate that FnCas9‐AF2 has extremely low potential off‐target sites even on multiple sites, suggesting a strong potential for further application.

2.6. Development of MUTE‐Seq, A High Fidelity FnCas9‐AF2‐Based Wild‐Type Allele Depletion Technique, and Its Application to MRD Detection for AML Patients

Utilizing FnCas9‐AF2, we introduce MUTE‐Seq (Mutation tagging by CRISPR‐based Ultra‐precise Targeted Elimination in Sequencing), a novel technique that overcomes the limitations of applicable sequences in enriching minor alleles through the highly precise elimination of the major counterparts. MUTE‐Seq incorporates high‐fidelity in vitro depletion of wild‐type DNAs using FnCas9‐AF2, and therefore increases the sensitivities of the subsequent analyses of the remaining mutant DNAs via sequencing (Figure 4A).

Figure 4.

Evaluation of MUTE‐Seq and Clinical Validation in AML patients. A) Schematic diagram demonstrating the process of MUTE‐Seq. This includes the in vitro digestion of wtDNA using FnCas9‐AF2, and the subsequent analysis of the enriched mutant DNA via Sanger and Next Generation Sequencing (NGS). B) Table of eight clinical samples showing NRAS mutations and corresponding Sanger sequencing VAF(%) calculated as the “Mutant peak height/(Mutant peak height + Wild‐type height) x 100(%)”. C) Sanger sequencing chromatograms of eight clinical samples, both before (control) and after MUTE‐Seq treatment, demonstrating the presence of NRAS mutations. D) VAFs from Sanger sequencing of eight clinical samples, before and after MUTE‐Seq treatment. The error bars represent the SEM from three independent experiments. E) Box plot showing the VAFs NRAS mutation of all patients before and after MUTE‐Seq treatment, respectively. Gray lines demonstrate pairing the same patient samples before and after MUTE‐Seq treatment. F) Scatter plot showing the correlation between Sanger sequencing VAFs and NGS VAFs. Circles and diamonds represent the MUTE‐Seq and control group, respectively. The line represents the linear regression (R ² = 0.958). G) Sanger sequencing chromatograms after enrichment for reference material created by mixing patient sample with the NRAS c.38G>A mutation and wild‐type gDNA to reach initial mutant ratios of 2.5, 1.25, 0.5, and 0.25%. H) Scatter plot for original VAF and detected VAF (R ² = 0.977).

Bone marrow samples of eight AML patients were analyzed with the MUTE‐Seq technique to validate its clinical efficacy utilizing Sanger sequencing. We obtained human genomic DNA (gDNA) from the bone marrow of eight AML patients who were being monitored for minimal residual disease (MRD). In the control group, NRAS mutations (G12D, G12C, and G13D) in DNA samples from patients A1–6 ranged from 3 to 16.9%, whereas samples from patients A7 and A8 exhibited no detectable NRAS mutations in the clinical laboratory (Figure 4B).

Our primary objective was to assess the increment of detected Variant Allele Frequency (VAF) using MUTE‐Seq followed by Sanger sequencing compared to the control group. In Sanger sequencing chromatograms, if there are two or more peaks at the same position, it indicates the presence of a mutation. VAF is typically inferred from the relative height of chromatogram peaks corresponding to mutant and wild‐type alleles at specific positions. The chromatograms of the control group displayed relatively lower peak sizes for mutant bases compared to wild‐type base peaks. In contrast, the MUTE‐Seq group exhibited distinct double peaks at the positions of mutant bases in the chromatograms (Figure 4C). VAFs were significantly higher in the MUTE‐Seq group, with a substantial level of agreement observed between the groups (Figure 4D,E).

For cross‐validation, NGS was employed to measure the VAF in both the control and MUTE‐Seq groups, revealing relatively similar detected VAFs (R ² = 0.958) between Sanger and NGS in both groups (Figure 4F). Notably, for samples A7 and A8, both groups exhibited VAFs below the detectable limit, consistent with the Sanger sequencing results.

Next, we tested whether MUTE‐Seq can be applied to lower VAF, which is hard to discern with Sanger sequencing. For this test, we prepared blended gDNA samples by mixing NRAS wild‐type and G13D‐mutant gDNA from 0.25 to 2.5% VAFs. Quantification of NRAS G13D allele frequencies showed that the MUTE‐Seq group consistently exhibited elevated VAFs compared to controls: rising to 11.1 from 0.25%, 16.4 from 0.5%, 31.8 from 1.25%, and 44.6 from 2.5% (Figure 4G). The increased levels of VAF in MUTE‐Seq enabled the detection of samples with a low mutant ratio. Significantly, the correlation between pre‐ and post‐enrichment was highly maintained with a R ² = 0.977 (Figure 4H). The results suggested that MUTE‐Seq methods could enhance the detection limit of Sanger sequencing for identifying low‐prevalence mutations in cancer samples.

2.7. Multiplexing MUTE‐Seq and Increasing the Sensitivity of Mutant Allele Detection by NGS

To evaluate the capacity of MUTE‐Seq by precisely eliminating wild‐type DNAs of the target sequence for detecting very low VAF in a multiplexed manner, we first conducted a multiplexed MUTE‐Seq using serially diluted plasmid DNAs of known VAF of target mutations (EGFR exon 19 deletion and BRAF V600E) from 0.5 to 0.005% along with 0% (wild type) sample. The samples were divided into two groups: the MUTE‐Seq group, where FnCas9‐AF2 was used to eliminate wild‐type DNAs, and a control group that remained untreated (Figure 5A). We then analyzed the VAFs of both groups with NGS to assess the feasibility of the MUTE‐Seq technique for identifying very low VAFs and to determine if there is a consistent elevation of original VAF by MUTE‐Seq. The targeted VAFs used in the study were 0.5, 0.25, 0.1, 0.05, 0.025, 0.01, and 0.005% in the order of dilution. Multiplexed sgRNAs were used to simultaneously target the EGFR and BRAF genes. The MUTE‐Seq group consistently showed elevated VAFs for EGFR exon 19 deletion compared to the controls, with an average 23.5‐fold increase across all tested VAFs. Notably, a strong correlation between pre‐ and post‐enrichment VAFs was maintained. Similarly, for the BRAF V600E allele, consistent VAF elevation post‐enrichment was also observed, with an average 19.2‐fold increase. In the contrary, no enrichment was detected for 0% VAF samples.

Figure 5.

Enhancing the Detection Sensitivity of Low‐frequency Mutations using Multiplexed MUTE‐Seq. A) Detected VAFs from diluted plasmid EGFR exon 19 deletion (top) and BRAF V600E (bottom) DNA samples of 0.5 to 0.005% VAFs for three MUTE‐Seq groups and three control groups. The linear regression line of best fit, along with the regression coefficient, is presented. Right side panels are magnified graphs focusing on the original VAFs ranging from 0.1 to 0.005% with a separate linear regression line of best fit and a regression coefficient. The dotted line indicates 0.1% VAF. B) Detected VAFs using NGS for control and multiplexed MUTE‐Seq on cfDNA reference material containing EGFR exon 19del, EGFR T790M, EGFR L858R, and KRAS G12D mutations of 1, 0.1, and 0% (n = 3, respectively). C) Box‐and‐whisker plot displaying the comparison of combined VAFs detected by control and MUTE‐seq. D) Probability‐of‐detection curves for 5 ng (left) and 50 ng (right) DNA inputs with probit regression fits (solid lines). The estimated 95% limits of detection were 0.15% VAF [95% CI, 0.08–0.22%] for 5 ng and 0.034% VAF [95% CI, 0.018–0.051%] for 50 ng.

Next, we tested multiplexed MUTE‐seq to cfDNA reference materials, containing mutations including EGFR, KRAS that are common cancer‐associated mutations with VAFs of 1, 0.1, and 0% (Figure 5B). MUTE‐Seq resulted in average VAF increment, for the examined mutations, of 34.2‐ and 66.2‐fold for 1 and 0.1% VAF samples (Figure 5C). We found that the sensitivity of MUTE‐Seq outperformed that of the control (1.0 vs 0.65), despite that the specificity of MUTE‐Seq remained comparable to the control (0.95 vs 0.95) (Figure S6, Supporting Information). In comparison, the VAF increments by enrichment using SpCas9‐WT, eSpCas9(1.1), SpCas‐HF4, and FnCas9‐WT were less than 4‐fold for both 1 and 0.1% samples (Figure S7, Supporting Information).

Next, the limits of detection (LODs) of multiplexed MUTE‐Seq were measured using cfDNA reference material diluted to VAFs of 1, 0.05, 0.02, and 0.01% (Figure 5D). The LODs were established as the VAFs at which the detection probability reaches 95%, as determined through probit regression analyses. The tests were conducted using input DNA amounts of 5 ng and 50 ng, considering that typical cfDNA concentrations range from 5 to 10 ng per mL of plasma.^{[ 29 ]} The probit‐derived LOD₉₅ values were 0.15% [95% CI, 0.08–0.22%] for 5 ng and 0.034% [95% CI, 0.018–0.051%] for 50 ng of input DNA. These values are closely aligned with the theoretical 95% LODs based on the Poisson distribution (0.2% for 5 ng of input DNA and 0.02% for 50 ng of input DNA).^{[ 30 ]} These findings suggest that the multiplexed MUTE‐Seq technique possesses a robust capability for detecting mutations at very low prevalence and holds significant potential for clinical application due to its consistent VAF enrichment ability.

2.8. Application of Multiplexed MUTE‐Seq Using NGS for Non‐Small Cell Lung Cancer (NSCLC) and Pancreatic Cancer Patients

Next, we evaluated the clinical utility of multiplexed MUTE‐Seq using plasma samples from 10 patients with NSCLC. Multiple sgRNAs were designed to simultaneously detect 323 clinically relevant cancer‐related mutation loci, based on the COSMIC database, in a single multiplexed MUTE‐Seq. Among the mutation loci, we compared the EGFR mutation profiles (EGFR exon 19 deletion, EGFR T790M, and EGFR L858R) identified in tissue biopsies with those detected in cfDNA from the NSCLC patients. The detected VAFs of EGFR variants were overall higher in the MUTE‐Seq group compared to the control group (Figure  6A). Among the 30 tested sites, tissue biopsy indicated 11 positives and 19 negatives. The control group detected 6 out of 11 true positives and one false positive. In contrast, the MUTE‐Seq group detected 10 out of 11 positives and one false positive (Figure 6B). The MUTE‐Seq consistently identified tissue specific‐mutations within cfDNA, highlighting its utility for liquid biopsy‐based detection of cancer mutations. Only two cases exhibited discordant mutational profiles: one case had a mutation uniquely identified in the tissue, while the other had a mutation exclusively detected in the cfDNA. Notably, all 18 cfDNA of the biopsy‐negative samples were concordantly called negative by both the control assay and MUTE‐Seq. In summary, MUTE‐Seq achieved showed higher sensitivity (90.9%), compared to the 54.5% sensitivity of the control group. The specificity was identical for both control and MUTE‐Seq groups (94.7%). However, given the ≈15% false‐negative rate reported for NSCLC tissue genotyping attributable to sampling bias, low tumor content, and intratumor heterogeneity,^{[ 31} , ^{32 ]} the interpretation of negative results must also consider the possibility that the results can be confounded by a concomitant biopsy miss and the presence of plasma variants below the detection threshold.

Figure 6.

Evaluation of Multiplexed MUTE‐Seq and Clinical Validation in Non‐Small Cell Lung Cancer (NSCLC) and Pancreatic Cancer Patients. A) VAF (%) of EGFR variants in cfDNA of NSCLC patients in control and MUTE‐Seq groups. B) Detection of EGFR mutations in control and MUTE‐Seq (EGFR Exon 19 deletion, EGFR T790M, and EGFR L858R) in tissue and blood specimens from 10 NSCLC patients. C) ROC analysis between tissue and blood samples from 10 NSCLC patients in relation to EGFR mutations. Lines colored in blue and red denote the ROC curves of multiplexed MUTE‐seq and control, respectively. D) Detection of pancreatic cancer‐associated mutations in tissue and cfDNA of 20 pancreatic cancer patients with MUTE‐Seq.

We generated a receiver‐operating‐characteristic (ROC) curve from the ten‐patient dataset encompassing stages I–IV (Figure 6C). The area under the ROC curve (AUC) for the combined stages was 0.96 with MUTE‐Seq, markedly higher than the 0.72 obtained with the control assay. Previous studies have shown that ctDNA detection is less sensitive in early‐stage cancer than in later‐stage disease.^{[ 4} , ⁵ , ³³ , ^{34 ]} To assess early performance, we examined the stage I cases; MUTE‐Seq achieved 100% tissue–plasma concordance. These findings indicate that MUTE‐Seq can sensitively detect mutations even at the earliest stage of cancer.

Next, the multiplexed MUTE‐Seq technique was applied to analyze the total of 716 clinically relevant mutation loci, based on the COSMIC database, in cfDNA of 20 pancreatic cancer patients who had undergone cancer‐removal surgeries (Figure 6D). Seven of the most prevalent mutation loci associated to pancreatic cancer were analyzed for both cancer tissue and cfDNA samples. We asked if MUTE‐seq could facilitate sensitive detection of the seven pancreatic cancer‐related mutations in patient cfDNA. Among the 29 detected mutations in cancer tissues, 24 detected mutations were in patient cfDNAs, with a sensitivity of 82.8% and a specificity of 100%. Considering that 17 of the 20 patients were at early stage (I and II) the results suggested that the MUTE‐Seq technique can be effectively utilized for detecting pancreatic cancer‐associated mutations in cfDNA for low‐grade cancer patients.

3. Discussion

Various CRISPR systems have been widely applied for biological and medical research.^{[ 11} , ¹² , ¹⁶ , ¹⁹ , ^{35 ]} Accordingly, previous studies reported the development of high‐fidelity CRISPR‐Cas variants.^{[ 18} , ²⁰ , ^{36 ]} Nonetheless, even with engineered Cas variants, it has been difficult to effectively discriminate off‐targets with a single‐base mismatch positioned within the target sequence of the sgRNA. In this study, we addressed this issue and utilized rational design to engineer a highly accurate FnCas9‐AF2 variant that effectively discriminates single‐base mutations at all 20 positions of the DNA that are complementary to sgRNA. We found that the precision of the engineered FnCas9‐AF2 exceeds that of eSpCas9 and SpCas9‐HF variants. In vitro cleavage assays with a DNA sample of meticulously designed SNVs and Digenome‐seq analyses for multiple genomic locations showed that FnCas9‐AF2 is capable of inducing DNA cleavage exclusively for targets exhibiting perfect base matches, ensuring single‐base precision and effective discrimination against off‐targets with single‐base mismatches.

Since FnCas9‐AF2 efficiently distinguished base mismatches in the target sequence of the sgRNA, we anticipated that it could be leveraged with flexible target selection for detecting low‐frequency mutations. As we anticipated, the FnCas9‐AF2‐based MUTE‐Seq facilitated sensitive MRD monitoring of AML patients. The results suggested that by integrating MUTE‐Seq methods with Sanger sequencing, we could enhance the detection limit of Sanger sequencing for identifying low‐frequency mutations in cancer samples. Moreover, MUTE‐Seq could be employed in a multiplexed manner, suggesting that this method could pave the way for the development of a cancer detection panel for the simultaneous detection of multiple cancer‐associated mutations in a single liquid biopsy analysis. Previous studies on mutant detection methods showed that it is more difficult to detect cancer‐associated mutations in blood samples of patients at the early stage.^{[ 33} , ³⁷ , ^{38 ]} Notably, we found that MUTE‐Seq enabled the sensitive detection of EGFR mutations in cfDNA from stage I NSCLC patients, as well as those in later stages. Similarly, for pancreatic cancer—which is notoriously difficult to detect in its early stages due to nonspecific symptoms and the absence of reliable, sensitive hematologic markers—we demonstrated that MUTE‐Seq could serve as a sensitive tool for early‐stage detection.

Over time, extensive efforts have been made to enhance the sensitivity of ctDNA detection.^{[ 39} , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ^{44 ]} Strategies employing NGS aimed to overcome the detection barrier by ultra‐deep sequencing. However, this approach not only increased costs but also appeared inefficient when the VAFs of ctDNA mutations dropped below the intrinsic error rates of NGS (0.1–1%).^{[ 11} , ^{12 ]} To overcome this sensitivity issue, enhanced technologies such as CODEC,^{[ 45 ]} MAESTRO,^{[ 46 ]} CAPP‐Seq,^{[ 47 ]} IDES,^{[ 48 ]} Safe‐seq,^{[ 49 ]} and PhasED‐seq^{[ 50 ]} were developed. Although these techniques vary in their details, they commonly employ unique molecular identifier (UMI) barcoding strategies that detect scant amounts of ctDNA through ultra‐deep sequencing with barcode to overcome the limitations of NGS errors.^{[ 9} , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ^{51 ]} Despite the efforts, the UMI method alone demonstrated limitations, with increased error rates observed when detecting below 0.1%.^{[ 48 ]} Droplet digital PCR is another technique with the potential for ctDNA detection; however, when used in isolation, its LOD is generally higher than 0.1%.^{[ 52 ]} In contrast to these methods, MUTE‐Seq does not require ultra‐deep sequencing nor UMI, as it uniquely leverages ultra‐precise CRISPR in the initial steps to reduce the wild‐type alleles in the samples. By implementing such noise canceling, MUTE‐Seq enables sensitive detection of mutant allele at a moderate sequencing depth, as the VAFs enter within the NGS confidence intervals. Furthermore, MUTE‐Seq may potentially be integrated with UMI or digital PCR‐based techniques to provide more synergies.

A tenacious effort to enhance the sensitivity even slightly is vital, as the applications of sequencing techniques are mainly for cancer diagnoses, where the tolerances for false negatives are minimal. The number of false negative patients is anticipated to be high, and its impact would be substantial given the large‐scale application in real‐world settings, which cannot be fully captured from common experimental studies, even for those involving patient samples. Therefore, the value of the MUTE‐Seq technique lies not only in the ability to detect low VAFs comparable to other novel sequencing methods, but also in achieving high sensitivity with a reasonable amount of required DNA input.

MUTE‐Seq sought to accomplish locus‐level precision with practical cost control: by depleting wild‐type molecules before sequencing. It achieves a tens‐of‐ppm–level limit of detection (≈0.005% VAF) at each chosen locus without the ultra‐deep read depths demanded by many UMI‐based assays, thereby cutting per‐sample cost dramatically. Because this per‐locus threshold is an order of magnitude lower than that of typical UMI workflows, the aggregate sensitivity across just a handful of loci readily falls into the sub‐ppm range, enabling tumor‐informed MRD monitoring without the many targets other platforms require. The same locus‐specific accuracy can also facilitate continuous monitoring of newly acquired drug‐targetable mutations that arise through clonal evolution. MUTE‐Seq can also be applied to multi‐cancer early detection with an expended hotspot panel. In addition, functioning as a front‐end “noise‐cancelling” step, MUTE‐Seq can be combined with existing laboratory pipelines: libraries may still be UMI‐tagged, analyzed with standard bioinformatics, and run on either high‐ or lower‐throughput sequencers, facilitating broad, cost‐effective deployment of liquid‐biopsy testing.

4. Experimental Section

Protein Engineering (Structural Analysis) and Cloning

The protein structure of FnCas9 (PDB ID 5B2O) was analyzed using PyMOL (Schrödinger, New York, NY, USA), and the FnCas9 component residues within hydrogen bonding distance of DNA were marked with spheres. These residues were changed to alanine using the QuikChange II Site‐Directed Mutagenesis Kit (Agilent, Santa Clara, CA, USA). Briefly, FnCas9‐WT (cloned in a pET28‐a vector) was used as a template to amplify FnCas9 variants using primers containing alanine point mutations. The FnCas9 variants were cloned according to the manufacturer's instructions, with a His×6 tag at the N‐terminus of each recombinant FnCas9.

Protein Purification

The pET vectors containing the FnCas9 variants under the T7 promoter were transformed into BL21‐DE competent cells (Novagen, San Diego, CA, USA) according to the manufacturer's instructions. The transformed cells were cultured in Luria–Bertani medium (Duchefa, Haarlem, Netherlands) at 37 °C. When the OD₆₀₀ (optical density at 600 nm) of the medium reached 0.5–0.7, the cells were treated with IPTG (Beams Biotechnology, Seongnam, Korea). Cells were harvested after overnight incubation at 18 °C and lysed in LYSIS buffer (50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, 1 mg mL⁻¹ lysozyme, 1 mM PMSF, 1 mM DTT, pH 8) using an ultrasonicator. The lysate was centrifuged at 15000 rpm to remove cell debris. The clear supernatant containing the FnCas9 variant protein was treated with Ni‐NTA beads (Qiagen, Hilden, Germany), which were then washed with WASH buffer (50 mM NaH₂PO₄, 300 mM NaCl, 20 mM imidazole, pH 8). Proteins were eluted with ELUTION buffer (50 mM NaH₂PO₄, 300 mM NaCl, 250 mM imidazole, pH 8), which was then exchanged with STORAGE buffer (50 mM HEPES, 200 mM NaCl, 20% glycerol, 1 mM DTT, pH 7.5) using an Amicon centrifuge filter (100 kDa; Merck, Kenilworth, NJ, USA).

In Vitro Transcription of sgRNA

Using an in vitro transcription method as previously described,^{[ 53 ]} sgRNAs with single‐base mutations were designed and synthesized for SpCas9 and FnCas9. Briefly, sgRNAs were transcribed by T7 RNA polymerase in a reaction mixture consisting of 40 mM Tris‐HCl (pH 7.9), 6 mM MgCl2, 10 mM DTT, 10 mM NaCl, 2 mM spermidine, NTPs, and an RNase inhibitor. The reaction mixture was incubated at 37 °C for 8 h, and the sgRNAs were purified using PCR purification kits (GeneAll, Seoul, Korea) and quantified using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA).

In Vitro DNA Cleavage Assay

A 3‐kb template including EGFR, KRAS, NRAS, and MET gene sequences (Table S1, Supporting Information) was cleaved with Cas9 proteins and sgRNAs (Table S3, Supporting Information). EGFR, KRAS, NRAS, and MET target sites were synthesized using IDT oligosynthesis platforms (Integrated DNA Technologies, Coralville, IA, USA) and cloned into a p3 vector, and the 3‐kb target DNA sequence was amplified from the vector by PCR, using two pairs of primers and Q5 DNA polymerase (New England Biolabs, Ipswich, MA, USA). The reactions were cleaned up using a PCR clean‐up kit (GeneAll, Seoul, Korea). The target DNA (100 ng) was incubated with 250 ng guide RNA and 500 ng Cas9 variant in CutSmart buffer (New England Biolabs) (100 mM potassium acetate, 20 mM Tris‐acetate, 10 mM magnesium acetate, 100 µg ml⁻¹ BSA, pH 7.9) for 1 h at 37 °C. The nuclease‐cleaved DNA fragments were run on a 1.5% agarose gel with TBE and stained with Midori Green (NIPPON GENETICS EUROPE, Duren, Germany).

Digenome‐seq

Digenome‐seq was carried out as described previously.^{[ 28 ]} Briefly, 8 µg genomic DNA (gDNA) was extracted from HEK293T using a DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany), then digested with 40 µg Cas9 and 10 µg sgRNA (target sequence: 5′‐TTGGACATACTGGATACAGC‐3′) for NRAS‐targeting Digenome‐seq analysis or 6 µg each of five sgRNAs (target sequence: 5′‐CTTTCTACCTACTGAGTCTG‐3′, 5′‐CTCCCTCCCAGGATCCTCTC‐3′, 5′‐CCACGTTCACCTTGCCCCAC‐3′, 5′‐GGAATCCCTTCTGCAGCACC‐3′, 5′‐TGGTTGGAGCAGGTGGTGTT‐3′ for DMD, AAVS1, HBB, FANCF, NRAS respectively) for multiplex Digenome‐seq analysis in 400 µL of 1× CutSmart buffer (New England Biolabs) at 37 °C for 16 h. Digested gDNA was isolated using a DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany) and then fragmented to a size of 500–600 bp using an M220 ultrasonicator (Covaris, Woburn, MA, USA). The NGS library for whole‐genome sequencing was prepared with a TruSeq Nano kit (Illumina, San Diego, CA, USA) and sequenced by NovaSeq (Illumina). The double‐strand break (DSB) score was measured and analyzed using the Digenome Sequencing analysis tool at CRISPR RGEN Tools (http://www.rgenome.net/). The loci with DSB scores over 1 were sorted, and their places in the human genome (hg38) were plotted in a Manhattan plot.

LOD Detection of MUTE‐Seq with Plasmid DNA and Reference cfDNA

DNA templates of wild type and mutant (EGFR 19del(ΔE746_A750) and BRAF V600E) were cloned into plasmids, linearized, and prepared at a concentration of 1 × 10⁷ copies per target, totaling 2 × 10⁷ copies. These samples were serially diluted to VAFs ranging from 0.5% to 0.25, 0.1, 0.05, 0.025, 0.01, and 0.005%.

Similarly, reference cfDNA samples were prepared by mixing sheared NA12878 cell line DNA with Seracare ctDNA Mutation Mix v2 AF2% (SeraCare, Milford, MA, USA) to achieve VAFs of 1, 0.05, 0.02, and 0.01%, using 5 and 50 ng of DNA for amplification.

Genes containing the hotspots of interest were amplified using Q5 DNA polymerase (New England Biolabs) and a multiplexed primer mix (Table S2, Supporting Information). To deplete wild‐type (WT) alleles, 1 µL of the multiplexed PCR product (diluted tenfold with DEPC‐treated water) was treated with 5 µg (25 pmol/10 µL) of FnCas9‐AF2 and 2 µg (66 pmol/10 µL) of 23 multiplexed sgRNAs (Table S3, Supporting Information) in 10 µL of 1× Remov RXN buffer (GeneCker, Seoul, Korea) at 37 °C for 1 h. The reaction was terminated by adding 10× STOP buffer (GeneCker, Seoul, Korea).

The WT‐depleted products were purified with AMPure XP beads (Beckman Coulter, Brea, CA, USA), then re‐amplified using Q5 DNA polymerase with an index primer. The index PCR amplicons were again purified with AMPure XP beads and sequenced using an Illumina sequencer (MiSeq for mutant plasmid DNA test and NovaSeq for reference cfDNA test).

Limits of detection at the 95% confidence level (LOD₉₅) were estimated by probit regression. For each DNA input (5 ng and 50 ng), four variant‐allele‐fraction (VAF) levels—1, 0.05, 0.02, and 0.01%— were spiked into wild‐type DNA and assayed at each level in 4–5 independent replicates.

Inclusion Criteria for Patient Samples

1. Individuals clinically suspected of having lung cancer, leukemia, or pancreatic cancer.

2. Patients who provided informed consent for sample collection and use.

3. Individuals with a confirmed diagnosis of NSCLC, AML, or pancreatic cancer based on pathological and/or molecular diagnostic criteria.

Bone Marrow Sampling and gDNA Extraction from AML Patients

Bone marrow samples were collected from AML patients who were suspicious of MRD. Bone marrow samples were collected in BD Vacutainer EDTA tubes (Becton Dickinson, Franklin Lakes, NJ, USA), and gDNA was extracted using DNeasy Blood & Tissue Kits (Qiagen, Hilden, Germany) according to the manufacturer's protocol. The concentration and purity of the gDNA were analyzed using the software associated with the Agilent TapeStation System.

MUTE‐Seq for AML Patients

The gDNAs were extracted from the bone marrow of AML patients. The gDNA blends were prepared by mixing the gDNA from HEK‐293T cells (wild‐type DNA) and the patient's DNA (mutant DNA) for VAF 0.25, 0.5, 1.25, and 2.5%. 100 ng of gDNAs were digested from 500 ng (2.5 pmol/10 µL) FnCas9‐AF2 and 200 ng (8.4 pmol/10 µL) sgRNA (Table S3, Supporting Information) in 10 uL of 1× Remov RXN buffer (GeneCker, Seoul, Korea) at 37 °C for 1 h. The reaction was terminated by adding 10× STOP buffer (GeneCker, Seoul, Korea). For NRAS mutation enrichment for Sanger sequencing, the enrichment PCR was performed as follows: the enrichment 50 µL‐PCR reaction contained 2 µL digested product, 5 µL primers mix (Table S2, Supporting Information), 25 µL 2× Master Mix (Sungenetics, Daejeon, Korea), and 18 µl nuclease‐free water. The reaction was performed under the following conditions: 98 °C for 3 min followed by 42 cycles of 98 °C for 10 s, 55 °C for 40 s, and 72 °C for 30 s. The tubes were incubated at 72 °C for another 5 min before storing at 4 °C. The PCR product was purified with a PCR clean‐up kit (GeneAll, Seoul, Korea) and then eluted into DEPC‐water. The purified PCR products were sent to Macrogen (Seoul, Korea) for Sanger sequencing (Table S2, Supporting Information). For validation of the NRAS mutation‐enrichment coupled Sanger sequencing, the digested products were amplified using Q5 DNA polymerase (New England Biolabs, Ipswich, MA, USA) with an index primer. Index PCR amplicons were purified with AMPure XP beads (Beckman Coulter, Brea, CA, USA) and sequenced using the iSeq 100 Sequencing System (Illumina). All cell lines were obtained from ATCC and tested negative for mycoplasma contamination.

Blood Sampling and cfDNA Extraction from NSCLC and Pancreatic Cancer Patients’ Plasma

A total of 4 ∼ 20 mL of blood from each NSCLC/Pancreatic cancer patient was collected in Streck Cell‐Free DNA blood collection tube (cfDNA BCT; Streck, La Vista, NE, USA). The collected blood was transferred to a Falcon tube and centrifuged at 1900 × g. The supernatant (plasma) was then gathered into Eppendorf tubes and centrifuged again at 16 000 × g. The cfDNA was isolated using the QIAamp Circulating Nucleic Acid Kit (QIAGEN, Hilden, Germany), following the manufacturer's instructions. The cfDNA was eluted in 20–70 µL of elution buffer from the QIAamp Circulating Nucleic Acid Kit and was then processed with the Cell‐free DNA ScreenTape assay (Agilent 4150 TapeStation system). The concentration and purity of the cfDNA were analyzed using the software associated with the Agilent TapeStation System.

Multiplexed MUTE‐Seq for NSCLC and Pancreatic Cancer Patients

Ten (NSCLC) or eleven (Pancreatic cancer) sgRNAs were designed to simultaneously detect 323 or 716 clinically relevant cancer‐related mutation loci, based on the COSMIC database, in a single multiplexed MUTE‐Seq. For assessment of the multiplexed MUTE‐Seq, cfDNAs were extracted from the patient samples (10–50 ng). Hotspots of interest were amplified using Q5 DNA polymerase (New England Biolabs) and multiplexed primer mix (Table S2, Supporting Information). To remove the WT allele, a 1‐µL fraction of multiplexed PCR product (diluted tenfold with DEPC‐treated water) was treated with 5 µg (25 pmol/10 µL) FnCas9‐AF2 (or 8 ug Cas variant) and 2 µg (66 pmol/10 µL) multiplexed sgRNA mix (Table S3, Supporting Information) in 10 µL 1× Remov RXN buffer (GeneCker, Seoul, Korea) at 37 °C for 1 h; the reaction was terminated by adding 10× STOP buffer (GeneCker, Seoul, Korea). The WT‐depleted products were purified with AMPure XP beads (Beckman Coulter, Brea, CA, USA), then amplified using Q5 DNA polymerase (New England Biolabs) with an index primer. Index PCR amplicons were purified with AMPure XP beads (Beckman Coulter, Brea, CA, USA) and sequenced using an Illumina MiSeqDx sequencer (Illumina).

Next Generation Sequencing Data Analysis

Adapter sequences were removed from the raw sequencing data using BBDuk version 38.96. The trimmed reads were aligned against the human genome reference (GRCh38) with bwa‐mem version 0.7.17. The aligned reads were partitioned to each amplicon using a custom script prior to variant calling. Somatic variants and short indels were detected with Mutect2 version 4.2.6.1 and VarDict version 1.8.2. Variants were annotated using Variant Effect Predictor version 108. The ROC curve metric represents the concordance between tissue and blood, and the model was developed based on logistic regression.

Statistical Analysis VAFs were quantified from NGS read data by analyzing the wild type and mutant allele sequences. Data were presented as mean ± standard deviation or by showing the individual results of all samples. For performance evaluation, all experiments were conducted with a sample size of at least three (n ≥ 3). Probit‑link generalized linear models were fitted to probability of detection outcomes to estimate LOD95, using 4–5 independent replicates per VAF level. All analyses were conducted in Python (v3.10) using packages including NumPy, pandas, SciPy, statsmodels, and scikit‑learn.

Ethical Clearance Patients with AML were included in the study, which received approval from the institutional review boards of St Mary’s Seoul Hospital (IRB No. KC23SISI0242). Patients with NSCLC and pancreatic cancer were included in the study, which received approval from the institutional review boards of Korea University Anam Hospital (IRB No. 2020AN0005), Boramae Medical Center (IRB No. 20‐2017‐17) and Asan Medical Center (IRB No. 2022‐1106). All procedures adhered to the Declaration of Helsinki and applicable national regulations, and all datasets were anonymized prior to analyses.

Conflict of Interest

S.Y. filed patent applications based on this study.

Supporting information

Supporting Information

Ye S., Kim J.‐S., Kim M., et al. “MUTE‐Seq: An Ultrasensitive Method for Detecting Low‐Frequency Mutations in cfDNA With Engineered Advanced‐Fidelity FnCas9.” Adv. Mater. 37, no. 47 (2025): e05208. 10.1002/adma.202505208

Data Availability Statement

The data that support the findings of this study are openly available in NCBI SRA at https://www.ncbi.nlm.nih.gov/sra/PRJNA1023721, reference number 1023721.