Peripheral blood samples of 120 AML patients were collected from the Hematology Department in Zhongnan Hospital of Wuhan University under an approved Institutional Review Board protocol. For DNA template preparation, about 500 µl of peripheral blood was mixed well with four times the volume of the RBC lysis buffer (Biosharp, Hefei, China). After 1 min of lysis, WBCs were precipitated by 1 min of mini centrifugation and then incubated with 150 µl nucleic acid releaser (GenDx, Suzhou, China) at 95°C for 3 min to release genomic DNA. Two microliters of the released DNA were used for subsequent RPA.

A 630-bp DNA fragment covering the WT FLT3-691 site was amplified by PCR with primers P1 and P2 from a WT patient’s genomic DNA, then cloned into the pGem-T vector (Takara, China). After transformation into Escherichia coli DH5α, the T-vec-F691 plasmid was extracted by an AxyPrep Plasmid Miniprep Kit (Axygen, CA, United States). For constructing the FLT3-L691 plasmid, primers P3 and P4 carrying a T > G base mutation were used to amplify the T-vec-F691 plasmid circularly. Then the amplified product was self-ligated and cloned into the pGem-T vector to construct the T-vec-F691 plasmid. For identification of the specificity of crRNAs, 350-bp F691 and L691 DNA fragments were produced by PCR with primers P5 and P6 using T-vec-F691 and FLT3-L691 plasmids as templates. The quantification of the above plasmids and DNA fragments were conducted by a Nanodrop2000 (Thermo Fisher Scientific, MA, United States). The sequence information of PCR primers are listed in Supplementary Table S1.

The RPA assay was conducted by a GenDx ERA Kit (Suzhou GenDx Biotech, China). Briefly, the 50 µl RPA system included 2 µl DNA template, 2.5 µl forward primer (10 nM), 2.5 µl reverse primer (10 nM), 11 µl ERA basic buffer, 20 µl reaction buffer, 2 µl activator, and supplementary ddH2O. After sufficient mixing, the mixture was incubated at 37°C for 15 min. Five microliters of the RPA product were transferred to subsequent Cas12a reaction. The RPA primers are listed in Supplementary Table S1.

Cas12a fluorescence reaction was conducted according to a previous description (Wang et al., 2020a). The 20 µl reaction mixture included 200 ng Cas12a, 0.1 pmol crRNA, 2 µl 10 × NEBuffer 3.1 (NEB, MA, United States), 1 µl RNase inhibitor (Novoprotein, China), 25 pmol ssDNA-FQ reporter (Genewiz, NJ, America), an appropriate amount of DNA fragments or 5 µl RPA product, and supplementary ddH2O. Among these components, crRNAs were directly synthesized by GenScript (Nanjing, China). And the sequence information of F691-crRNA and L691-crRNA are listed in Supplementary Table S2. Cas12a protein was produced as described previously (Creutzburg et al., 2020). The mixed Cas12a reaction system was incubated at 37°C for 15 min, and then the result was directly observed under a 485-nm blue lamp (Sangon, Shanghai, China). Fluorescence kinetics were monitored using a monochromator with excitation at 485 nm and emission at 520 nm.

For FGS detection, 30 µl PCR products or 30 µl RPA products were purified by the AxyPrep PCR Clean-up Kit and quantified by the Nanodrop2000. Approximately 200 ng DNA products were sent to FGS conducted by Tsingke (Beijing, China). Different barcoded primers were designed and synthesized for the PCR amplification of the FLT3-691 region of genomic DNA samples for NGS detection. The amplified products were then purified and mixed equally for NGS detection by the Illumina NextSeq 500 (2 × 150) platform at the CAS-MPG Partner Institute for Computational Biology Omics Core, Shanghai, China. The PCR primers for the preparation of NGS samples are listed in Supplementary Table S3.

We have repeated all experiments three times. Statistical analysis was performed by GraphPad Prism software version 8.0. Unpaired two-tailed Student’s t-test was used for comparison between two groups. Quantitative data are expressed as mean value ± SE. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, no significance.

Commonly-used CRISPR detection includes two steps, isothermal amplification and Cas12a fluorescence reaction. Recombinase polymerase amplification (RPA) is a highly selective and efficient isothermal amplification technique, which allows the amplification of single-digit DNA target copies under 37–42°C (Lobato and O’Sullivan, 2018). In this study, RPA was adopted to amplify the FLT3-691 region of genomic DNA. Then RPA product was detected by L691-crRNA induced Cas12a reaction, wherein fluorophore-labeled ssDNA existed. Upon recognizing F691L mutant amplicons, Cas12a will generate nonspecific ssDNAase activity to cleave fluorophore-labeled ssDNA and release fluorescence signal (Figure 1A).

We developed a complete diagnosis process from drawing blood to receiving the clinical application of this CRISPR detection system. First, the blood drawn from the patient was mixed with red blood cell (RBC) lysis buffer for 1 min. Then 1 min of centrifugation was adopted to obtain white blood cell (WBC) precipitation. Next, the genomic DNA of these WBCs was released by a nucleic acid releaser under 95°C for 3 min. Then, the DNA sample was amplified by RPA under 37°C for 15 min, and detected by L691-crRNA induced Cas12a reaction under 37°C for 15 min. Finally, the result can be analyzed by a fluorescence reader or directly observed by naked eyes under a blue lamp. Increased fluorescence intensity and green fluorescence signal indicate a F691L positive result. On the contrary, unchanged fluorescence intensity and no green fluorescence signal suggest a F691L negative result (Figure 1B). Thus, the whole diagnosis can be completed within 40 min, without the need for large equipment or expensive reagent.

The high specificity of the Cas12a reaction relies on a specific crRNA, which only recognizes target DNA while ignores non-targets. For the detection of mutation-type (MT) FLT3-L691 (c. TTG) from wild-type (WT) FLT3-F691 (c. TTT), we designed a crRNA with a sequence complementary to L691 genomic DNA (Figure 2A). To identify the specificity of L691-crRNA, we used L691-crRNA induced Cas12a reaction to detect 1e12 copies of WT and L691 DNA fragments, as well as other four mutations in the detection region, which were recorded in the catalogue of somatic mutations in cancer (COSMIC) database (Supplementary Table S4). After 15 min reaction under 37°C, the L691 sample showed a strong fluorescence signal, and no fluorescence was observed in WT and other mutation samples. These results confirmed the high specificity of L691-crRNA (Figures 2B,C). Then gradient copies (1e12 ∼ 1e9) of L691 DNA fragments were used to identify the sensitivity of L691-crRNA induced Cas12a reaction. The result showed that as low as 1e9 copies of L691 DNA fragments could be detected (Figure 2D). Finally, we mixed 1e12 copies of F691 and L691 DNA fragments with different proportions to simulate mutation rates (100, 50, 25, 10, 1, 0.1%, and WT). As expected, the fluorescence signals gradually decreased with mutation rates. A significant difference could be observed between 0.1% sample and the negative control, while no significant difference was found between WT sample and the negative control. Together, these results indicated that L691-crRNA is highly specific for the detection of FLT3-F691L mutation (Figure 2E). In the same way, we also designed an F691-crRNA for the detection of WT FLT3-F691 (Supplementary Figrue S1).

To achieve an ultrasensitive detection of FLT3-F691L mutation, a highly efficient amplification is needed to produce enough target copies. According to the design principles of RPA primers, we designed four forward primers RPA-F1 ∼ 4 and four reverse primers RPA-R1 ∼ 4 for the amplification of the FLT3-691 region (Figure 3A). Then 16 primer pairs were generated by pairwise collocation between forward and reverse primers, and they were tested to screen out the most efficient pair. In this test, 1e2 copies of L691 plasmid templates were used to simulate the genomic DNA, and the RPA products were detected by L691-crRNA induced Cas12a reaction. After 15 min of RPA with different primer pairs and 15 min of Cas12a reaction, the strongest fluorescence signal was observed in the F2R1 pair, indicating that F2R1 is most efficient to amplify the FLT3-691 region (Figure 3B). In order to analyze its amplification capacity more specifically, we used F2R1-mediated RPA to amplify 1e7 ∼ 1e1 gradient copies of L691 plasmid templates. After L691-crRNA induced Cas12a reaction, the results showed that as low as 1e1 copies of L691 plasmids could be successfully detected (Figures 3C,D and Supplementary Figure S2). Together, using 15 min of F2R1-mediated RPA and 15 min of L691-crRNA induced Cas12a reaction, the FLT3-F691L detection system can achieve detection of nearly single-digit copies of pure L691 DNA templates.

Detection of low-abundance FLT3-F691L mutation may help early clinical diagnosis and avoid ineffective treatment. In order to define the sensitivity of the FLT3-F691L detection system, 1e5 copies of plasmids with gradient F691L mutation rates were used as tested samples. The mutation rates included 100% (pure L691), 10, 1, 0.1, 0.01%, and 0 (WT, pure F691), generated by different proportions of L691 and F691 plasmids. These samples were amplified by RPA and then detected by L691-induced Cas12a reaction as mentioned above. The results showed that the detection system could achieve a sensitivity of 0.1%, which means detecting 100 copies of L691 templates among 99,900 copies of F691 templates. Besides, there is no significant difference between WT and NC samples, which further confirmed the high specificity of this detection system (Figures 4A,B and Supplementary Figure S3). The RPA products of these samples were also detected by F691-crRNA to show the amplification of WT templates (Figure 4A). To analyze the fluorescence change more clearly, we use a fluorescence reader to record the time course of the L691-crRNA induced Cas12a reaction. The gradually increased fluorescence intensity in 0.1% sample was observed, compared with unchanged fluorescence intensity in 0.01%, WT, and NC samples. Together, these results indicated that our FLT3-F691L detection system could achieve a sensitivity of 0.1% under 1e5 copies of templates, which is a 100-fold improvement compared with FGS.

We then applied CRISPR detection to screen AML patients for FLT3-F691L positivity. Briefly, blood samples of 120 AML patients were drawn and the genomic DNA was released, then the DNA templates were used to RPA and L691-induced Cas12a reaction, and finally, the results were observed by naked eyes under a blue lamp. The whole detection was completed within 40 min (Figure 5A). Out of the 120 samples, represented by P1∼120, we found only one FLT3-F691L positive case P33 (Figure 5B and Supplementary Figure S4–S6). This is consistent with the low frequency of FLT3-F691L mutation in AML populations. The NGS data of the 120 patients confirmed the CRISPR detection results, wherein P33 had a F691L mutation rate of 4.9%. However, this sample was not screened out using FGS, further indicating the low sensitivity of FGS (Figure 5C). Using NGS as the gold standard, we compared CRISPR detection with FGS to detect the 120 AML patients. The statistics showed that although CRISPR detection and FGS both performed a 100% specificity in this test, CRISPR detection displayed a much higher sensitivity than FGS (100 vs. 0%) (Figure 5D). Together, these results verified the high efficiency, specificity, and sensitivity of the CRISPR detection method for FLT3-F691L.