Mechanism and design of the PER-based approach
The working mechanism of the PER-based label-free miRNA detection approach is depicted in Fig. 1. Two split-DNAzyme sequences were first designed with three functional sections, respectively. The a sections in the two split-DNAzyme sequences could bind with miRNA-21; b section in the two sequences was complementary and could form a duplex only when a sections were hybridized with miRNA-21; the rest section in the two split-DNAzyme sequences could bind with loop section in the designed dumbbell probe. The dumbbell-shaped probe in the PER process was designed with two loop sections. In detail, the c sections could provide biding sites for primer; e section could transcribe G4 sequence; and d section assisted the formation of dumbbell probe. Based on the miRNA-21-assisted formation of correct secondary conformation of DNAzyme and Mg2+-assisted activation, split-DNAzyme-miRNA-21 complex bound with loop section in the dumbbell probe and generated a nicking site, exposing the recognizing section that could be identified by primer. When split-DNAzyme-miRNA-21 complex was discharged from a dumbbell probe, it could bind with a next dumbbell probe to form a signal cycle. With the aid of DNA polymerase (Klenow fragment, KF), a nascent ssDNA sequence was appended to the 3′ terminal of the primer, which was extended until the stand displacing extension reaction was haltered at the stop sequence on the dumbbell hairpin probe. The d section competed with copied d′ section and via the random walk process of three-way migration. As a result, the d section was released from dumbbell probe, and dumbbell probe was free to bind with another primer and induce a next PER. Therefore, a large amount of ssDNA sequences (e′ sequence) containing G4 sequences were generated, which could be identified by NMM to form G4-NMM complex with the enhanced fluorescence respond.
Feasibility of split-DNAzyme assembly and PER
Target miRNA-based assembly of split-DNAzyme determined the activation of DNAzyme and imitation of subsequent PER. The assembly of split-DNAzyme was firstly investigated through PAGE analysis. The Lane 1 in Fig. 2a was target miRNA-21, and Lane 2 was the complex of split-DNAzyme and miRNA-21. To test the cleavage activity of the active DNAzyme, a ssDNA sequence (m sequence) was synthesized to mimic the loop section in dumbbell probe. In Lane 3, a band was observed that moved faster than DNAzyme, indicating that m sequences were digested by the active DNAzyme. PAGE result of dumbbell probe-assisted formation of active DNAzyme is listed in Additional file 1: Fig. S1, indicating a same conclusion. Fluorescence assay was performed on FAM and BHQ-2 labeled m sequence, and the result showed a greatly enhanced signal when active DNAzyme existed (Fig. 2b). PER process was validated in the PAGE result in Fig. 2c. A hairpin probe was utilized to mimic the dumbbell probe which has a nicking site in the loop section. When the hairpin probe was mixed with primer, a new band appeared in Lane 3, moving slower than hairpin probe alone. After KF-based chain extension, a band between hairpin probe and primer appeared, which was considered the produced e′ sequences.
Analytical performance of the approach
For a better detection performance, several experimental parameters were optimized, including the length of b section in the two split-DNAzyme sequences, concentration of KF and incubation time. From the result in Additional file 1: Fig. S2, the length of b section was determined 5 nt; concentration of KF was 1 U/L; and the incubation time of the whole system was 1.5 h. In addition, the concentration ratio of split-DNAzyme for assembly was determined 1:1; the reaction temperature after adding the target miRNA-21 was optimized to 37 °C; and the concentration of NMM was determined 250 μM (Additional file 1: Fig. S3). Under the obtained optimized experimental conditions, we then studied the detection performance of the approach. To test the sensitivity of the approach, the approach was utilized to detect a series of samples containing synthesized miRNA-21 which was firstly diluted to different concentrations ranging from 10 fM to 1 nM. The detected fluorescence signal gradually increased when the concentration of miRNA in samples increased (Fig. 3a). The peak value of detected fluorescent signals was correlated with the logarithmic concentration of miRNA-21 in the sample. The correlation equation was determined Y = 1181*lgC + 10,135, with R2 of 0.9954 (Fig. 3b). In addition, the limit of detection (LOD) of the method was computed as 2.43 fM based on the general 3σ method. Selectivity of the approach was verified through detecting miRNA-21 and other miRNAs (miRNA-211, miRNA-155, Let-7a, 1 pM respectively). The result in Fig. 3c showed that fluorescence signal of the approach when detecting miRNA-21 was significantly high compared with control (without miRNA-21). When miRNA-211, miRNA-155 and Let-7a, the approach exhibited neglectable signal enhancements compared with control group, indicating a high selectivity of the approach. As a conventional approach, qRT-PCR is a gold-strand for miRNA quantification in clinical applications. Herein, we compared the detection performance of the established approach with qRT-PCR. The method and qRT-PCR were utilized to quantify 6 samples containing different concentrations of miRNA-21, respectively. The result in Fig. 3d showed a good correlation between the calculated miRNA concentrations by the established approach and by qRT-PCR, indicating a promising applicable potential of the approach.
Clinical application of the approach
As a crucial biomarker, the expression level of miRNA-21 is up-regulated in osteosarcoma patients. Thus, we applied the approach to quantify miRNA-21 in MG-63 cell extraction and normal cell extraction (hOB cell). The result in Fig. 4a showed that miRNA-21 level in MG-63 cell extractive is much higher than that in the hOB cell extractive. The repeatability was verified through detecting 10 sample duplicates containing 100 pM miRNA, respectively. The result in Fig. 4b showed a coefficient of variation (CV) of 4.65%. Compared with former approaches, the established method exhibited a low LOD in a label-free manner which could meet the requirements of clinical practice (Additional file 1: Table S2).