An RCA-Coupled CRISPR/Cas12a Cascade-Enhanced SERS Platform for Ultrasensitive Detection of Exosomal miRNA-21 in Colorectal Cancer Serum Authors: Jian Tao, Binbin Zeng, Caili Bi, Minyi Xu, Xiaxian Gu, Dingding Jing, Jun Sun Published: 2026 ABSTRACT AbstractPurposeUltra-sensitive detection of exosomal miR-21 is crucial for the early diagnosis of colorectal cancer. This study aims to develop a cascade amplification platform based on surface-enhanced Raman scattering (SERS), combining rolling circle amplification (RCA) and CRISPR/Cas12a, to enable ultra-sensitive quantitative detection of exosomal miR-21.MethodsA SERS substrate was constructed using streptavidin-modified concave gold nanocubes (CGNs), and gold double-stranded probes loaded with Raman reporter molecules were used as SERS probes. The two components are controllably linked via a single-stranded DNA molecule with a thiol group at one end and biotin at the other, forming a stable detection system. Target-triggered RCA generates a repeat sequence, which, under the mediation of crRNA, activates Cas12a cleavage, causing the probe to be released from the substrate surface and resulting in a significant attenuation of the SERS signal.ResultsThis platform achieves three-stage cascading signal amplification via RCA, CRISPR/Cas12a, and SERS, with a detection limit at the femto-mole (fM) level and a total detection time of approximately 80 minutes. The results were consistent with those of qRT-PCR, demonstrating high sensitivity, high specificity, and reliability.ConclusionThis study proposes an RCA-CRISPR-SERS cascade amplification platform that provides a highly sensitive and specific method for detecting miR-21 in exosomes, offering potential clinical value for the early diagnosis and dynamic monitoring of colorectal cancer.Keywords: surface-enhanced Raman scattering, rolling circle amplification, CRISPR/Cas12a, exosomes, colorectal cancer Introduction Colorectal cancer (CRC), a highly aggressive malignant tumour, is currently the second leading cause of cancer-related deaths worldwide1 and accounts for 9.2% of all deaths,2 primarily due to the lack of prominent early symptoms and late diagnosis. Colonoscopy is widely regarded as the gold standard for early screening of CRC, offering high sensitivity and specificity.3 However, this method relies on expensive equipment, specialised technical personnel and a high level of patient compliance, which limits its widespread implementation. As real-time indicators of tumour progression, serum biomarkers such as proteins,4 DNA (mutations and methylation markers),5 RNA (particularly microRNAs)6 and volatile organic compounds7 have been extensively studied and are of significant importance for achieving early cancer screening and precision intervention. Exosomes, a class of membrane-bound vesicles widely present in body fluids,8 contain miRNAs that are highly stable and closely associated with the development and progression of colorectal cancer, making them highly promising non-invasive biomarkers.9,10 Among these, miRNA-21 plays a central role in the development of colorectal cancer by regulating the PI3K/AKT, STAT3 and PDCD4/TNF-α signalling pathways,11 and its detection in plasma can achieve a diagnostic accuracy and sensitivity of approximately 90%.12 However, the abundance of exosomal miRNAs in body fluids is extremely low.13 Existing methods such as qRT-PCR14 and microarray chips15 are complex to operate and lack reproducibility, whilst digital PCR16 and RNA sequencing17 are costly and difficult to apply on a large scale. Consequently, the development of cost-effective screening tools holds significant clinical importance. Surface-enhanced Raman spectroscopy (SERS), as an ultra-sensitive spectroscopic technique that relies on localized surface plasmon resonance, can significantly enhance Raman scattering signals. It possesses exceptional molecular fingerprinting capabilities and non-destructive detection properties,18 and shows broad application prospects in the field of bioanalysis.19,20 However, in complex biological sample systems, nonspecific adsorption and background interference severely compromise signal accuracy and stability, posing a key bottleneck that limits the further application of SERS in nucleic acid detection.21 In recent years, the clustered regularly interspaced short palindromic repeats (CRISPR) and their associated protein (Cas) systems, originating from the adaptive immune systems of bacteria and archaea,22 have been widely applied in molecular diagnostics23 due to their highly specific recognition and cleavage capabilities toward target nucleic acid sequences. Combining CRISPR technology with SERS detection strategies24 not only enables precise recognition and cleavage of targets25 but also effectively enhances the specificity and sensitivity of signal responses. However, a single signal amplification mechanism remains insufficient to meet the demands of ultra-sensitive detection of low-abundance nucleic acid molecules. Rolling circle amplification (RCA), as a circular-template-based isothermal nucleic acid amplification technique, offers advantages such as ease of operation, high amplification efficiency, and strong signal amplification capabilities,26 and has been widely used in the development of nucleic acid signal amplification systems.27 Therefore, incorporating RCA into the CRISPR/SERS detection system to achieve triple-cascade signal amplification is expected to enable ultra-sensitive detection of low-abundance exosomal miR-21. By organically integrating RCA, CRISPR, and SERS, we have established a multi-stage signal amplification strategy that achieves cascading signal enhancement, thereby significantly improving detection sensitivity. This method holds promise for the ultra-sensitive detection of low-abundance exosomal miR-21 and provides a new technical approach for the efficient analysis of nucleic acids in complex biological samples. In this study, we introduced the RCA nucleic acid amplification strategy into the detection of exosomal miRNA and developed a SERS detection platform driven by RCA-mediated CRISPR cascade amplification for the ultra-sensitive detection of colorectal cancer-associated exosomal miRNA-21 (Figure 1). As shown in Figure 1A, this study first constructed a highly ordered and reproducible array of concave gold nanocubes (CGN) as a substrate and coated its surface with streptavidin. Next, Au nanobipyramids (Au NBPs) were synthesized and modified with DTNB to serve as SERS probes. A thiol-modified single-stranded DNA (ssDNA) segment was first conjugated to the SERS probe, and then successfully immobilized on the concave gold nanocube array via biotin at the other end, thereby establishing the detection platform. When the target is present in patient serum exosomes, the lock-on probe initiates RCA, generating a large amount of RCA products. These are recognized by crRNA, which activates Cas12a to cleave the ssDNA linker connecting the SERS probe to the concave gold nanocube array. The SERS probe detaches from the array, causing the signal to decay (Figure 1B). As shown in Figure 1C and D, when miR-21 is detected in patient serum exosomes, RCA products are generated, triggering Cas12a activity, resulting in the cleavage of the ssDNA and a decrease in signal intensity; conversely, the signal increases in the absence of the target. Compared to existing technologies, this method significantly enhances detection sensitivity through triple amplification via RCA-CRISPR-SERS, while also providing a novel approach for the detection of miRNAs in serum exosomes from CRC patients. (A) Preparation and assembly of Au BPs and CGNs; (B) Detection principle of the sensor, namely RCA amplification and CRISPR-Cas12a-mediated targeted cleavage; (C) Detection of clinical samples; (D) Evolution of the SERS detection signal. Experimental Section Experimental Materials All chemical reagents used in this study were of analytical grade. Sodium borohydride (NaBH4), tetrachloroauric acid (HAuCl4), sodium hydroxide (NaOH), sodium oleate (NaOL), hydrochloric acid (HCl), AgNO3, and ascorbic acid were purchased from Aladdin (China). Cetyltrimethylammonium bromide (CTAB), cetyltrimethylammonium chloride (CTAC), trisodium citrate dihydrate, anhydrous ethanol, and phosphate-buffered saline (PBS) were all obtained from Sinopharm Chemical Reagents Co., Ltd. (Suzhou). The nucleotide sequences used in this experiment (Table 1) were synthesized by Shanghai Bioengineering Co., Ltd. In Table 1, the bolded and underlined text indicates the specific base types substituted in the single-base mismatch (MT1) and three-base mismatch (MT3) sequences, respectively. Silver nitrate (AgNO3), ascorbic acid, cetylpyridinium chloride (CPC), and other relevant reagents were purchased from Sigma-Aldrich. The markers used for the two electrophoresis runs were ordered from Nearshore Proteins (item no.: 05371704) and Jinsha Biotech (SM811) respectively; the band ranges for both were 100–1000 bp. The real-time quantitative polymerase chain reaction (qRT-PCR) kit was purchased from Jiduan Biotechnology Co., Ltd. The Cas12a protein was ordered from Integrated DNA Technologies (product number: 1081068), with a Cas12a/crRNA concentration ratio of 1:1.5; the reaction conditions were as follows: pH 7.5, temperature 37 °C, and reaction time 60 min. The exosome extraction kit was purchased from Biotek Biotechnology Co., Ltd. (Wuxi). All solutions were prepared using ultrapure water with a resistivity of 18.2 MΩ cm to ensure the stability and reproducibility of the reaction system. Table 1Nucleotide Sequences Used in the ExperimentNameSequencesmiR-21UAGCUUAUCAGACUGAUGUUGAssDNABiotin-TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT-SHcrRNAUAAUUUCUACUAAGUGUAGAUCGUCGCCGUCCAGCUCGACCPadlock probeCTGATAAGCTAAGATACCCTAACCATCGATCGTCGCCGTCCAGCTCGACCTCAACATCAGTMT1TAGCTTATGAGACTGATGTTGAMT3TAGCTTAACAGTCAGATGTTGARandom sequenceGGUUUACUUUAGAGACUAGAUC Nucleotide Sequences Used in the Experiment Collection and Processing of Clinical Samples Participant enrollment and eligibility criteria were established in accordance with the Chinese Expert Consensus on Early Diagnosis and Treatment of Colorectal Cancer (2023 Edition). Patients with CRC were comprehensively classified based on colonoscopic findings, computed tomography (CT) imaging, and histopathological confirmation. All serum samples were obtained from the Changshu Hospital Affiliated with Nanjing University of Chinese Medicine. Written informed consent was obtained from all participants prior to inclusion in the study. The study was conducted in strict compliance with the ethical principles outlined in the Declaration of Helsinki and was approved by the Medical Ethics Committee of the Changshu Hospital Affiliated with Nanjing University of Chinese Medicine (Ethics No. 2025178). A total of 90 subjects were enrolled (Table 2), including 30 healthy controls (14 males and 16 females) and 60 CRC patients (32 males and 28 females). Peripheral blood samples were collected and subjected to standardized processing immediately after collection. The processed samples were then aliquoted and stored at −80°C until further analysis. This standardized protocol was designed to preserve sample integrity to the greatest extent possible, minimizing degradation of target molecules and reducing potential variability across experimental batches. Table 2Baseline Characteristics of Study ParticipantsGroupCRC Patients (n = 60)Healthy Controls (n = 30)Mean age (years)5957SexMale3214Female2816Clinical stageStage I11—Stage II19Stage III14Stage IV16Total6030 Baseline Characteristics of Study Participants Bioinformatics Analysis Identification of Differentially Expressed miRNAs and Data Processing miRNA sequencing data for colorectal adenocarcinoma were retrieved from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/), specifically from the COAD and READ projects. The dataset comprised 542 tumor samples and 9 normal tissue samples. Differential expression analysis between tumor and matched normal tissues was performed using the R package edgeR. Differentially expressed miRNAs (DEmiRNAs) were identified based on the thresholds of |log2fold change| ≥ 1 and false discovery rate (FDR) < 0.05. The results were visualized using volcano plots. Prediction of miR-21 Target Genes Potential target genes of hsa-miR-21-3p were predicted using two established databases, TargetScan 8.028 (https://www.targetscan.org/vert_80/) and miRDB29 (https://mirdb.org/), with the species restricted to Homo sapiens. The predicted target gene lists from both platforms were exported and integrated. To enhance prediction reliability, a Venn diagram was constructed to identify the overlapping target genes shared by the two databases, which were retained for subsequent analysis. Identification of Target Genes Among Differentially Expressed Genes Transcriptomic RNA-Seq data for TCGA-COAD and TCGA-READ were obtained from the TCGA database under the parameters: Experimental Strategy = “RNA-Seq” and Data Category = “Transcriptome Profiling”. A total of 701 samples were included, comprising 51 normal and 650 tumor tissues. Differential gene expression analysis was conducted using the R package DESeq2, with screening criteria set as |log2 fold change| ≥ 1 and P < 0.05. A total of 1,542 differentially expressed genes (DEGs) were identified, including 426 upregulated and 1,116 downregulated genes. The intersection between the predicted miR-21 target genes and the identified DEGs was determined and used for downstream analyses. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway Enrichment Analyses Functional enrichment analyses of the intersecting target genes were performed using the R package clusterProfiler. Gene Ontology (GO) analysis categorized genes into three domains: biological process (BP), cellular component (CC), and molecular function (MF). For each category, the top 10 significantly enriched GO terms (ranked by P-value) were selected for visualization. In parallel, KEGG pathway enrichment analysis was conducted, and the top 10 significantly enriched pathways were similarly visualized to elucidate the potential biological functions and signaling pathways associated with miR-21 in colorectal cancer. Synthesis of Concave Gold Nanocubes The synthesis of CGNs was performed based on a previously reported protocol with deliberate optimization and refinement to enhance structural uniformity and reproducibility.30 Briefly, 0.6 mL of freshly prepared NaBH4 solution (10 mM) was precooled and rapidly injected into a vigorously stirred reaction mixture containing 10 mL of CTAC (100 mM) and 0.25 mL of HAuCl4 solution (10 mM). This rapid reduction step facilitated the formation of gold seed nanoparticles. The reaction system was then allowed to stand undisturbed for 120 min to ensure complete nucleation and stabilization of the seeds. Subsequently, the obtained seed solution was diluted 1,000-fold using 0.1 M CTAC. For the growth phase, a seed-mediated growth solution was prepared by mixing 120 mL of CTAC (100 mM), 6 mL of HAuCl4 (10 mM), 1.2 mL of AgNO3 (10 mM), 1.8 mL of HCl (1 M), and 1.2 mL of ascorbic acid (100 mM). Under continuous vigorous stirring, 1.2 mL of the diluted seed solution was slowly introduced into the growth solution, initiating the controlled anisotropic growth of concave gold nanocrystals. The reaction mixture was then left undisturbed at room temperature overnight to allow for full crystal maturation and structural stabilization. Upon completion, the resulting CGNs were collected and redispersed in CPC (10 mM) solution, which improved their colloidal stability during storage and facilitated subsequent applications. Synthesis of Gold Nanobipyramids Au NBPs were synthesized via a seed-mediated growth approach, following a modified protocol to achieve enhanced morphological uniformity and anisotropic control.31 First, thoroughly mix 8 mL of CTAC aqueous solution (66 mM) with 2 mL of HAuCl4 solution (1 mM), then add 2 mL of HNO3 (10 mM) and allow the mixture to react for 15 minutes at 800 rpm. Next, introduce 2 mL of a mixed solution of NaOH and NaBH4 (volume ratio 1:1) as a strong reducing system, and add 200 μL of trisodium citrate dihydrate solution (1 M) to obtain the gold seed solution. For the growth stage, 38 mL of ultrapure water, 1.82 g of CTAB, and 0.076 g of NaOL were sequentially added to the reaction system. The mixture was vigorously stirred at 800 rpm until complete dissolution, establishing a stable dual-surfactant environment conducive to anisotropic crystal growth. Subsequently, 12.5 mL of HAuCl4 (0.5 mM) and 3 mL of AgNO3 (2 mM) were introduced, where silver ions played a critical role in directing longitudinal growth. When the solution gradually changes from yellow to colorless, add 1 mL of hydrochloric acid (0.5 M), 20 mL of hydroquinone (0.1 M), and 300 μL of the pre-prepared seed solution in sequence. The system is allowed to grow for one day under mild reducing conditions to form well-defined bipyramidal structures. After the reaction, the mixture was purified three times by centrifugation to remove unreacted materials and excess surfactant. Finally, the resulting Au NBPs were redispersed in 20 mL of ultrapure water for future use. Exosome Isolation and Extraction of miRNA Exosome Isolation This study employed a protocol combining differential centrifugation with an EPF purification column to isolate serum-derived exosomes. This method relies on the differences in sedimentation coefficients between cells, cell debris, apoptotic bodies and exosomes to achieve graded separation; low-speed centrifugation is used to remove large cellular impurities, whilst ultra-high-speed centrifugation is employed to precipitate nanoscale exosomes. The procedure is as follows: serum samples from colorectal cancer patients were placed in a 37 °C water bath to thaw completely, and impurities were removed stepwise using gradient low-speed centrifugation. First, the samples were centrifuged at 300 ×g for 10 minutes (4 °C) to precipitate residual intact blood cells; the supernatant was then transferred to a new sterile centrifuge tube, and the cell pellet was discarded; Collect the resulting supernatant and centrifuge further at 2,000 ×g for 10 minutes (4 °C) to remove cell debris and apoptotic bodies, then collect the supernatant again; centrifuge the second supernatant at 10,000 ×g for 30 minutes (4 °C) to remove coarse cell debris, and filter the resulting liquid through a 0.45 μm microporous filter membrane to remove larger vesicles and impurities. Transfer the clear filtrate to an ultracentrifuge tube; after weighing and balancing, perform ultracentrifugation at 100,000 ×g for 120 minutes (4 °C). This ensures that soluble contaminating proteins remain in the supernatant. Upon completion of centrifugation, discard the supernatant. Gently resuspend the pellet at the bottom of the tube in pre-chilled (4 °C) sterile phosphate-buffered saline (PBS) to obtain a crude exosome extract. To further enhance purity, add the crude exosome suspension to the upper chamber of an EPF molecular sieve purification column and centrifuge at 3,000 ×g for 4 hours (4 °C) until the liquid has completely passed through the column bed. Collect the eluted purified exosome solution, sterilise it using a 0.22 μm filter membrane, and immediately aliquot it into sterile cryovials; store the exosomes at −80 °C for long-term storage, whilst avoiding repeated freeze-thaw cycles to prevent damage to the morphology of the exosomes and their internal miRNAs. Extraction of Exosomal miRNA Exosomal miRNA was extracted using a lysis–precipitation-based method. In brief, 150 μL of purified exosome suspension was mixed thoroughly with 3 μL of carrier RNA, 100 μL of lysis buffer, and 10 μL of proteinase K, followed by incubation at 65 °C for 10 min to ensure complete vesicle disruption and protein degradation. Subsequently, 550 μL of absolute ethanol was gradually added to facilitate nucleic acid precipitation, and the mixture was centrifuged at 12,000 rpm for 60s at 4 °C to remove the supernatant. The pellet was then washed with 500 μL of wash buffer A (composed of 80% absolute ethanol and 20% wash buffer), incubated at room temperature for 120 s, and centrifuged again under the same conditions to eliminate residual contaminants and enhance miRNA purity. Finally, miRNA was eluted using 30 μL of preheated elution buffer, followed by centrifugation at 12,000 rpm for 120 s at 4°C. This elution step was repeated once, and the eluates were combined. The extracted miRNA samples were stored at −80°C until further use. Rolling Circle Amplification To construct the circular DNA template required for RCA, specific hybridization between miRNA-21 and a padlock probe was first established. Briefly, 4 μL of padlock probe (100 nM), 1 μL of 10× T4 DNA ligase buffer, 2 μL of miRNA-21, and 12 μL of ultrapure water were thoroughly mixed and subjected to a brief denaturation at 90°C for 180 s to eliminate secondary structures and facilitate precise hybridization. The reaction mixture was then rapidly cooled, followed by the addition of 1 μL of T4 DNA ligase, and incubated at 16°C for 60 min. This step enabled efficient ligation of the padlock probe guided by the miRNA template, resulting in the formation of a covalently closed circular DNA structure. Upon completion of ligation, the reaction was heated to 65°C for 5 min to inactivate the ligase, yielding a stable circular DNA template suitable for subsequent amplification. For the RCA reaction, an isothermal amplification system was prepared by mixing 1 μL of dNTPs (10 mM), 0.5 μL of bovine serum albumin (20 mg/mL), 0.5 μL of Phi29 DNA polymerase, 2 μL of 10× reaction buffer, 2 μL of circular DNA template, and 4 μL of ultrapure water. The mixture was incubated at 30°C for 30 min. Owing to its high processivity and strong strand displacement capability, Phi29 DNA polymerase continuously extends along the circular template, generating long concatemeric DNA strands composed of tandem repeats, thereby achieving substantial signal amplification. Following amplification, the reaction was terminated by heating at 65°C for 5 min to deactivate the enzyme. The resulting high-copy RCA products provide a robust molecular basis for subsequent detection with enhanced sensitivity. Agarose Gel Electrophoresis A 4% agarose gel was prepared by dissolving 2 g of agarose in a solution consisting of 49 mL ultrapure water and 1 mL of 50× TAE buffer, followed by heating until complete dissolution. The gel was allowed to cool and solidify prior to use. For sample preparation, 10 μL of the reaction product was thoroughly mixed with 2 μL of 6× loading buffer and then loaded into the gel wells. Electrophoresis was performed in 1× TAE running buffer at a constant voltage of 140 V for 35 min, enabling effective separation of nucleic acid fragments of different lengths. Instrumentation UV-vis absorption spectra were recorded using a Cary UV-5000 spectrophotometer (Agilent Technologies, USA). Raman measurements were performed with an InVia confocal Raman microscope (Renishaw, UK). The morphology and surface characteristics of nanomaterials were characterized using a field-emission scanning electron microscope (SEM, S-4800 II, Hitachi, Japan). Raman imaging was conducted on a DXR3xi Raman imaging microscope (Thermo Fisher Scientific, USA). High-resolution structural analysis was carried out using a field-emission transmission electron microscope (TEM, Tecnai G2 F30S-TWIN, FEI, USA). Results and Discussion Characterization of Gold Nanobipyramids To systematically evaluate the structural characteristics and SERS enhancement properties of the prepared Au NBPs, this study conducted multi-scale characterization of their morphology, crystal structure, and optical properties. As shown in Figure 2A, SEM images reveal that the mass-produced Au NBPs exhibit a highly uniform bipyramidal morphology with good interparticle dispersion; no significant agglomeration or morphological collapse was observed. TEM results (Figure 2B) reveal that the Au NBPs have distinct structural contours and sharp tips, with a concentrated distribution of aspect ratios. This tip region is capable of generating significant localized surface plasmon resonance (LSPR) effects. Clear lattice fringes were resolved in HRTEM, with spacings of approximately 0.238 nm and 0.208 nm, corresponding to the {111} crystal plane of gold, indicating that the material possesses good crystallinity (Figure 2C). Selective area electron diffraction (SAED, Figure 2D) reveals clear, regular diffraction rings. This excellent crystal structure not only contributes to the stability of the LSPR properties but also provides a foundation for SERS enhancement. The UV-vis absorption spectrum (Figure 2E) shows a distinct absorption peak at 732 nm. Furthermore, the solution appears uniformly colored, further indicating that the nanoparticles possess good dispersion stability. Particle size distribution analysis (Figure 2F and G) reveals that the length of Au NBPs is primarily distributed within the range of 65 ± 5 nm, with a width concentrated at 25 ± 2 nm, indicating a narrow size distribution and good uniformity. X-ray diffraction (XRD) analysis (Figure 2H) reveals diffraction peaks corresponding to the Au (111), (200), (220), and (311) crystal planes at 2θ values of 38.2°, 44.4°, 64.6°, and 77.5°, respectively, with the (111) peak exhibiting the highest intensity, indicating preferential crystal growth along this direction. To evaluate the SERS enhancement capability, DTNB was used as the Raman signal molecule for testing (Figure 2I). The results showed that DTNB labeled Au NBPs produced a significantly enhanced Raman signal, far exceeding the intensity of high-concentration DTNB without enhancement. In summary, the prepared Au NBPs exhibit excellent performance in terms of morphological uniformity, crystal structural integrity, and optical response characteristics. (A) SEM image and 3D structure of batch-prepared Au NBPs; (B–D) TEM, HRTEM, and SAED images of Au NBPs, where ① and ② are magnified regions in the HRTEM images, and the distances between the parallel lines correspond to lattice spacings of 0.238 nm and 0.208 nm. (E) UV-vis absorption spectra and photographs of Au NBPs in solution; (F and G) Statistical analysis of the length and width size distributions of Au NBPs; (H) XRD patterns of Au NBPs; (I) Comparison of SERS spectra between DTNB-modified Au NBPs and pure DTNB. Characterization of Concave Gold Nanocubes We prepared highly ordered concave gold nanocube arrays as highly efficient SERS-active substrates. As shown in Figure 3A, SEM images and 3D structures reveal that the CGNs exhibit a regular tetrahedrally concave structure overall, forming distinct synergistic features of edges and concave faces, with a highly anisotropic curvature distribution. TEM images (Figure 3B) further confirm that CGNs possess good monodispersity and a uniform size distribution, with no obvious agglomeration observed. Clear lattice fringes are resolvable in the HRTEM image (Figure 3C), with interplanar spacings of approximately 0.217 nm and 0.156 nm, corresponding to characteristic Au crystal planes. SAED reveals regular and distinct diffraction spots, indicating a well-developed polycrystalline structure that can significantly enhance SERS intensity. The UV-Vis absorption spectrum (Figure 3D) shows a distinct absorption peak at 664 nm, and the solution appears dark green in the photograph. To further construct a highly efficient SERS substrate, CGNs were assembled into an ordered array (Figure 3E). In the array structure, the particles are densely packed, forming numerous nanoscale gaps that provide the spatial conditions for the formation of electromagnetic “hotspots”. The electromagnetic field distribution was simulated using the FDTD method (Figure 3F–H). The results show that the electric field is primarily concentrated at the edges of the concave surfaces, the tips, and the inter-particle gap regions, with higher field intensities observed in the coupling regions between adjacent particles. This indicates that the CGNs structure is capable of generating localized SERS enhancement effects at multiple locations. Testing was conducted using DTNB as the Raman signal molecule (Figure 3I). The results show that pure DTNB exhibits only a very weak Raman signal, whereas after CGNs modification, the intensity of its characteristic peak is significantly enhanced, demonstrating that CGNs possess excellent SERS enhancement capabilities. (A) SEM images and 3D structures of batch-prepared CGNs; (B) TEM images of CGNs; (C) HRTEM and SAED images of CGNs; ① and ② represent two randomly selected close-up views, and the distance between the parallel lines represents the lattice spacing; (D) UV-Vis absorption spectra and a photograph of CGNs; (E) SEM image of a CGN array and a bar chart of particle size distribution; (F–H) Electromagnetic field simulation diagrams of CGNs; (I) Comparison of SERS spectra between DTNB modified CGNs and pure DTNB; (J) 3D superposition of SERS spectra obtained at different detection positions within CGNs; (K) Comparison of SERS spectra at 10 different detection positions within CGNs and a 2D distribution heatmap of Raman signal intensity in the 1330–1360 cm−1 range; (L) SERS spectra of the CGNs array and a two-dimensional distribution heatmap of Raman signal intensity at different storage times (0–30 days). Further evaluation of the substrate’s uniformity revealed good consistency in the Raman spectra obtained at different detection positions (Figure 3J). Statistical analysis of signal intensities at 10 random points (Figure 3K) showed consistent intensity at the characteristic Raman peak of 1344 cm−1, indicating that the substrate possesses good spatial uniformity. As shown in Figure 3L, during a storage period of 0–30 days, the Raman signal intensity initially decreased and then gradually stabilized, indicating that the substrate maintains stable enhancement performance under long-term storage conditions. In summary, the constructed CGNs array demonstrates satisfactory enhancement capability, spatial uniformity, and long-term stability. The Relationship Between Targets and Diseases Building on previous research, we conducted a systematic bioinformatics analysis of the miRNA expression profiles associated with CRC (Figure 4) to establish the potential regulatory role of miR-21 in CRC at the molecular level and to provide a basis for selecting targets for subsequent detection. As shown in Figure 4A, the miRNA expression profiles of normal and CRC tissues exhibited a distinct separation trend, with clear stratification observed between different samples. A volcano plot was constructed based on the differential analysis results (Figure 4B). After setting the log2(Fold Change) and significance thresholds, a total of 403 differentially expressed miRNAs were identified, including 232 up-regulated and 171 down-regulated miRNAs. Notably, hsa-miR-21-3p is located at the extreme end of the significantly upregulated region, exhibiting the highest level of expression shift, suggesting that it may occupy a central node in CRC-related regulatory networks. Bioinformatics analysis of miRNA-21 targets. (A) Heatmap showing expression differences of various types of microRNAs between normal tissue samples and colorectal cancer samples; (B) Volcano plot of differentially expressed miRNAs; (C) Venn diagram showing the intersection of differentially expressed genes in colorectal cancer and miRNA-21 target genes; (D) Bar chart of GO analysis; (E) Circle plot of GO analysis; (F) Sankey bubble chart of KEGG analysis. Based on these findings, we performed an intersection analysis between differentially expressed genes (DEGs) and the predicted target genes of miR-21-3p (Figure 4C). To further elucidate the functional characteristics of these candidate target genes, we conducted a GO enrichment analysis (Figure 4D and E). The results indicate that, at the BP level, these genes are primarily enriched in processes such as PDGF receptor signaling regulation, cyclic nucleotide metabolism, and GPCR signaling pathways; at the CC level, they are mainly localized to autolysosomes and related subcellular structures; and at the MF level, they are concentrated in cAMP-related enzyme activity and binding functions. Collectively, these enrichment results point to processes related to signal transduction and metabolic regulation. Further KEGG pathway analysis (Figure 4F) revealed that these genes are primarily involved in key pathways such as the cAMP signaling pathway, AMPK signaling pathway, insulin signaling pathway, and purine metabolism. Among these, the cAMP and AMPK pathways are closely associated with energy metabolism regulation, while the insulin signaling pathway is linked to cell proliferation processes, suggesting that miR-21 may participate in the development and progression of CRC through the synergistic regulation of multiple pathways. In summary, through differential expression screening, target gene intersection analysis, and functional enrichment analysis, miR-21-3p was progressively identified as a key candidate molecule. This provides a solid theoretical basis for its use as a key detection target and lays the molecular foundation for the subsequent development of a highly sensitive detection platform. Validation of RCA-CRISPR/Cas12a Cleavage Feasibility To validate the feasibility of the RCA amplification and CRISPR/Cas12a cleavage systems, the different reaction components were analyzed using agarose gel electrophoresis (Figure 5, Tables S1, S2 and Figure S1). As shown in Figure 5, lanes 1 and 2 contained miRNA-21 and the ssDNA linker, respectively, and both exhibited a single, distinct band, indicating that the nucleic acid structures were intact. In the absence of miRNA-21 (lane 3), the RCA system did not produce any detectable bands, indicating that the amplification reaction was not triggered; however, upon addition of miRNA-21 (lane 4), high-molecular-weight products were observed retained near the loading well, indicating the successful generation of long-chain RCA amplification products. In the Cas12a cleavage system (lanes 5–6), regardless of whether the RCA reaction mixture was added, the ssDNA conjugate band remained intact without significant degradation in the absence of miRNA-21-induced amplification products, indicating that Cas12a was not activated. In contrast, when both miRNA-21-induced RCA products and the Cas12a/crRNA system were present (Lane 7), the ssDNA ligand band was significantly attenuated or even disappeared, indicating that effective cleavage had occurred. Furthermore, this study validated RCA reaction systems lacking ligase, target or polymerase, respectively, by presenting control electrophoresis results in Figure S1. The results showed that the reactions failed to proceed normally in the absence of these key components, further confirming that all components are indispensable for the RCA amplification reaction. Validation of the RCA and CRISPR/Cas12a cleavage system. Analysis of RCA amplification products and Cas12a-mediated cleavage behavior under different component combinations via agarose gel electrophoresis. Among them, the sample in lane 1 was miRNA-21 only; the sample in lane 2 was ssDNA linker only; the sample in lane 3 was the RCA reaction system without miRNA-21; the sample in lane 4 was the RCA reaction system with miRNA-21 added; the samples in lanes 5 and 6 were Cas12a cleavage systems without amplification products induced by miRNA-21, and the difference between the two lay in whether the RCA reaction mixture was added; The sample in Lane 7, highlighted by the red box, contains all components, which fully confirms that the reaction has occurred. In summary, the electrophoresis results clearly demonstrate that RCA amplification occurs only in the presence of miRNA-21, generating a high-molecular-weight product; this product can further activate the CRISPR/Cas12a system, enabling effective cleavage of the ssDNA ligand, thereby validating the feasibility of the RCA-CRISPR/Cas12a cascade reaction system. Optimization of Experimental Parameters To achieve optimal detection performance, key parameters such as reaction temperature, reaction time, system pH, and the volume of ssDNA added were systematically optimized. All optimisation experiments were conducted in triplicate (n=3), and the results are presented as mean ± standard deviation (Mean ± SD). All optimisation of reaction conditions (such as temperature, time, pH and probe concentration) was carried out in serum-spiked solutions to eliminate background interference from complex biological sample matrices and ensure the accuracy of the screening of reference parameters. As shown in Figure 6A, the signal intensity at the characteristic peak of 1344 cm−1 exhibits a trend of “first increasing, then decreasing” with temperature, reaching a minimum at 37°C; the signal rises markedly when the temperature exceeds 37°C, confirming 37°C as the optimum temperature. As shown in Figure 6B, the SERS signal gradually increases with prolonged reaction time, weakens significantly between 0 and 60 minutes, and stabilises at 80 minutes; further extension of the reaction time resulted in a flatter trend in the signal. Eighty minutes was selected as the optimal detection time. As shown in Figure 6C, the pH of the system has a significant effect on the detection signal. As the pH increased from 6.5 to 7.5, the SERS signal gradually weakened, reaching a minimum at pH 7.5; a further increase in pH led to a marked rise in the signal, indicating a decrease in proteolytic activity. The results of the optimisation of the ssDNA volume are shown in Figure 6D. As the ssDNA volume increased, the SERS signal gradually decreased, reaching a minimum at 5 μL. Based on the above results, the optimal reaction conditions were finally determined to be: temperature 37°C, reaction time 80 min, pH 7.5, and ssDNA volume 5 μL. (A–D) Bar charts and line graphs of the SERS intensity at 1344 cm−1 for the CRISPR-SERS detection chip based on the RCA multi-cascade amplification strategy, under different temperatures, detection times, buffer pH values, and volumes of ssDNA. Performance Validation Under optimal experimental conditions, this study systematically evaluated the analytical performance of the SERS detection platform and conducted multidimensional validation of its reproducibility and specificity (Figure 7A–D). As shown in Figure 7A, the SERS spectra at 1344 cm−1 exhibited highly consistent peak shapes and signal intensities across detection platforms prepared from different batches, indicating that the platform preparation process is stable and controllable. Figure 7B shows that the signal intensity distribution across the surface of this SERS platform is generally uniform. Calculations indicate that the relative standard deviation (RSD) of the signals from parallel measurements across multiple batches was 2.32%, confirming that this SERS substrate exhibits good signal reproducibility. Functional validation of the CRISPR-SERS detection chip based on an RCA multi-stage amplification strategy. (A) SERS spectra of detection chips from different production batches; (B) Two-dimensional distribution heatmaps of SERS signals for detection chips from different production batches; (C and D) SERS signal intensities at 1344 cm−1 for various potential interferents and corresponding systems (blank, miR-21, MT1, MT3, Random, Glucose, Lipoprotein, Cholesterol, Mixture). To verify specificity, this study selected MT1, MT3, random RNA sequences, a pool of target miRNAs, a pool of non-target miRNAs, and common biological interferents (such as glucose, lipoproteins and cholesterol) as control groups, in order to systematically evaluate the sensor’s ability to recognise the target miR-21. In Table 1, use bold and underlining to indicate the specific types of bases that have been substituted in the MT1 and MT3 sequences. The results show that the signal at the 1344 cm−1 characteristic peak decreases significantly only in the presence of the target miR-21, while mismatched sequences and interferents have minimal effect on the signal (Figure 7C). When miR-21 is present in the system, the signal intensity is significantly lower than that of the blank group (Figure 7D). Quantitative Analysis After completing the system development and functional validation, the quantitative analytical capability of the detection platform in a serum system was evaluated under optimal conditions. miR-21 at different concentration gradients (1 fM-1 nM) was introduced into a simulated serum system for detection (Using simulated serum with varying target concentrations, this approach enables an accurate evaluation of the platform’s analytical performance under specific matrix conditions), and the corresponding SERS spectra are shown in Figure 8A. As the miR-21 concentration gradually increased, the intensity of the characteristic Raman peak at 1344 cm−1 exhibited a continuous and significant decreasing trend. To further illustrate the pattern of signal changes, a two-dimensional mapping of the SERS intensity corresponding to different miR-21 concentrations was performed (Figure 8B). As the miR-21 concentration increased, the signal gradually transitioned from high-intensity regions to low-intensity regions, exhibiting a clear gradient distribution, indicating that the system possesses excellent signal resolution across different concentration ranges. For quantitative analysis, a linear regression equation was established with the logarithm of the miR-21 concentration on the x-axis and the intensity of the characteristic SERS peak at 1344 cm−1 on the y-axis (Figure 8C). Each concentration data point was analyzed based on 3 independent parallel replicate experiments. The results show that within the range of 1 fM to 1 nM, the two variables exhibit a good linear relationship (R2 = 0.9858), with a detection limit as low as 0.54 fM, which is lower than that of existing detection methods (Table 3), indicating that this platform possesses stable quantitative analysis capabilities (Table S3). LOD is calculated using the classic international standard formula (LOD = 3SD/K),18,32,33 where SD is the standard deviation of the blank group and K is the slope of the standard equation. In summary, this SERS detection platform enables highly sensitive detection of miR-21 in complex serum environments, demonstrating excellent quantitative analytical performance and application potential. Characterization of the quantitative detection performance of miR-21. (A) SERS spectra of miR-21 at different concentrations (1 fM-1 nM) in serum; (B) Heatmap of the two-dimensional distribution of SERS signals for miR-21 at different concentrations; (C) Linear relationship curve between the intensity of the characteristic SERS peak at 1344 cm−1 and the logarithm of the miR-21 concentration. Table 3Comparison of the Methodology Employed in This Study with Other miRNA Detection TechniquesMethodTargetLODRef.FluorescencemiR-133a3.3 pM[34]PhotoelectrochemicalmiRNA-211 fM[35]ElectrochemicalmiR-37850 pM[36]SERSmiR-213.58 pM[37]RCA-CRISPR-SERSmiR-210.54 fMThis work Comparison of the Methodology Employed in This Study with Other miRNA Detection Techniques Clinical Sample Characterization To validate the applicability of the detection system in real clinical samples, we systematically characterized the histological features, imaging manifestations, and serum-derived exosomes of CRC patients at different stages. As shown in Figure 9A–D, tissue sections from CRC patients at different stages all exhibited typical morphological features of adenocarcinoma. With advancing stage, glandular structures gradually shifted from relatively regular to disorganized arrangements, accompanied by marked atypia and invasive growth, reflecting histological changes during tumor progression. Colonoscopic and rectoscopic images (Figure 9E–H) revealed abnormalities in the luminal structure in patients across all stages. Early-stage samples primarily exhibited localized lesions, whereas stages III–IV demonstrated more extensive tissue destruction and luminal deformation, suggesting that the severity of the disease increases with advancing stage. Characterization of clinical CRC samples. (A–D) Histological sections from patients with stage I, II, III, and IV colorectal cancer; (E–H) Colonoscopic images from patients with stage I, II, III, and IV colorectal cancer; (I–L) CT images of patients with stage I, II, III, and IV colorectal cancer; (M) TEM images of serum exosomes from CRC patients, and the red box highlights the exosomes; (N) Nanoparticle tracking analysis (NTA) results; (O and P) Western blot results; (Q) Zeta potential; (R–T) Nanoflow cytometry fluorescence results for the control group, CD9, and CD81. CT imaging results (Figure 9I–L) were highly consistent with endoscopic and pathological findings. Typical features such as irregular thickening of the intestinal wall, heterogeneous enhancement, and local structural disorganization were observed in patients across all stages; as the stage increased, the extent of the lesions expanded, and in some cases, invasion of surrounding tissues or blurred structures appeared, further reflecting the characteristics of tumor progression. This multimodal imaging evidence is consistent with previous clinical diagnostic criteria, indicating that the selected samples are highly representative. Based on these findings, serum-derived exosomes were characterized. As shown in Figure 9M, TEM images revealed that the exosomes exhibited a typical vesicular structure; particle size analysis (Figure 9N) indicated that their dimensions were primarily distributed within the range of approximately 30–150 nm; Western blot results (Figure 9O–P) demonstrated positive expression of the exosomal marker proteins CD63 and TSG101. Zeta potential measurements further indicate the stability of the serum exosomes isolated (Figure 9Q); their negative zeta potential is consistent with the general characteristic of exosomes, which is typically less than −10. Nano-flow cytometry was further utilised to quantitatively analyse the proportion of CD9- and CD81-positive exosome particles (Figure 9R–T). The non-specific positive signal in the blank control group was only 0.43%, demonstrating that background interference in the system was extremely low. Following immunofluorescence staining, CD9-positive particles accounted for 7.85% of the total particles detected, whilst CD81-positive particles accounted for 6.19%. The positive particle population was clearly distinguishable from the background noise, demonstrating the successful recovery of a large number of structurally intact exosome particles from patient serum, alongside high efficiency of the fluorescent labelling. In summary, through multi-level characterization involving pathology, endoscopy, imaging, and exosome analysis, the staging accuracy and biological characteristics of samples from CRC patients were systematically confirmed, providing a reliable clinical sample foundation for subsequent highly sensitive detection based on exosomal miRNA-21. Clinical Serum Sample Testing To evaluate the performance of the CRISPR-SERS detection platform in real clinical samples, serum samples from 90 healthy controls and patients with early-stage and advanced-stage CRC (These included 11 cases of stage I, 19 of stage II, 14 of stage III, 16 of stage IV, and 30 healthy controls) were analyzed (Table 2). As shown in Figure 10A, the intensity of the SERS characteristic peak at 1344 cm−1 exhibited significant differences across serum samples from different populations. The signal was generally higher in the healthy group, while it was significantly lower in the CRC patient group; furthermore, the intensity was lower in advanced-stage patients than in early-stage patients, demonstrating a decreasing trend with disease progression. Statistical analysis revealed significant differences between groups (P < 0.001), indicating that the platform can effectively distinguish between different clinical states. To further validate the accuracy of the test results, the miRNA-21 expression levels measured by the detection platform were compared with those obtained via the gold standard qRT-PCR method (Figure 10B and C). The results showed a good linear correlation between the two, indicating that the platform possesses stable quantitative consistency in the serum system (Table 4). The diagnostic performance of the detection platform was further evaluated using receiver operating characteristic (ROC) curves (Figure 10D). The results demonstrated that the system exhibits high discriminatory ability in distinguishing between early-stage and advanced-stage CRC (AUC values of 0.967 and 1.000, respectively), reflecting excellent sensitivity and specificity. (A) Distribution of SERS signal intensity at 1344 cm−1 in serum samples from healthy controls, patients with early-stage colorectal cancer, and patients with advanced-stage colorectal cancer (**** p<0.0001); (B) Correlation Analysis Between the CRISPR-SERS Platform and qRT-PCR for the Detection of miR-21 in Serum; each blue star represents one clinical serum sample; (C) Bland-Altman agreement analysis of CRISPR-SERS versus qRT-PCR; the dark green dashed line represents the mean deviation, the upper and lower light green dashed lines represent the 95% agreement limits, and each blue dot represents one clinical serum sample; (D) ROC curve distinguishing early-stage colorectal cancer (blue) from advanced-stage colorectal cancer (magenta); the red diagonal line represents the random classification reference line (AUC = 0.5), and the shaded area represents the 95% confidence interval. Table 4A Comparison of miR-21 Expression Levels in Serum Exosomes from Colorectal Cancer Patients Using SERS and qRT-PCRStageSERS (fM)qRT-PCR (fM)Relative Error (%)miR-21Health Group(9.63 ± 6.3)(10.2 ± 3.2)−5.58Early stage(66.46 ± 8.2)(61.16 ± 9.5)8.6Advanced stage(606.08 ± 10.5)(580.49 ± 11.7)4.4 A Comparison of miR-21 Expression Levels in Serum Exosomes from Colorectal Cancer Patients Using SERS and qRT-PCR In summary, this CRISPR-SERS detection platform enables effective differentiation and staging of CRC in clinical serum samples, demonstrating good quantitative consistency and diagnostic performance, and holds potential clinical application value. Conclusion This study established a CRISPR-SERS detection platform based on a multi-cascade RCA amplification strategy, enabling ultra-sensitive detection of miRNA-21 in colorectal cancer-associated exosomes. By integrating RCA, CRISPR/Cas12a-specific recognition, and SERS signal transduction, a detection system featuring hierarchical amplification and dual molecular recognition was developed. The engineered Au NBPs and CGN arrays provide abundant electromagnetic “hotspots”, significantly enhancing both signal intensity and detection stability. The platform exhibits a robust linear response over a wide concentration range (1 fM-1 nM) and maintains distinguishable signals even at femtomolar levels. In addition, it demonstrates excellent sequence specificity and strong resistance to biological interference. In recent years, CRISPR/Cas12a, RCA and SERS have been widely applied in biosensing research; however, single technologies or simple combinations of two technologies still have significant limitations. Firstly, under conditions where target concentrations are low, direct detection is susceptible to the influence of Cas12a background activity or the non-specific degradation of reporter molecules, thereby reducing the signal-to-noise ratio.38,39 Secondly, the initiation of RCA relies on specific hybridisation between the primer and the template; in the absence of strict regulation, non-specific amplification is likely to occur, generating false-positive signals.40 Furthermore, the uneven distribution of “hotspots” on conventional SERS substrates leads to poor signal reproducibility, whilst the non-specific adsorption of biomacromolecules onto the metal surface can mask the target signal, resulting in background interference.41 This study organically integrates these three approaches to construct a cascade-amplification biosensor, which not only overcomes the respective technical bottlenecks but also achieves ultra-sensitive, highly specific detection through multiple signal amplification mechanisms. Within this system, RCA, CRISPR/Cas12a and SERS each perform distinct yet interconnected functions. RCA is responsible for target-driven nucleic acid pre-amplification, generating ssDNA rich in repetitive sequences that serves as a functional scaffold for Cas12a activation and SERS probe assembly. Subsequently, under the guidance of crRNA, Cas12a specifically recognises the RCA products and activates its trans-cleavage activity, further converting target recognition into a cleavage event involving a large number of reporter molecules. SERS, in turn, converts this molecular process into a highly sensitive, quantifiable spectral signal. Consequently, RCA, CRISPR/Cas12a and SERS achieve nucleic acid amplification, specific recognition and enzymatic amplification, and highly sensitive readout, respectively, thereby establishing a multi-stage cascading signal amplification system that provides a powerful tool for the precise detection of low-abundance colorectal cancer-associated biomarkers. Importantly, when applied to clinical serum samples, the platform effectively differentiates healthy individuals from CRC patients at different stages, with results showing strong agreement with those obtained by qRT-PCR. ROC analysis further confirms its high diagnostic performance. This detection platform has demonstrated satisfactory performance, ranging from experimental results in laboratory synthesis systems (Validation of RCA-CRISPR/Cas12a Cleavage Feasibility–Quantitative Analysis) to detection results in real clinical samples (Clinical Serum Sample Testing). Overall, this RCA-coupled CRISPR/Cas12a-triggered SERS strategy provides a powerful approach for ultrasensitive miRNA detection in complex biological matrices and holds considerable promise for early screening of colorectal cancer. Nevertheless, the translation of this platform into routine clinical practice still faces a number of technical and methodological challenges. Firstly, exosome isolation and miRNA extraction are susceptible to factors such as isolation strategies, matrix effects, non-specific protein adsorption and nucleic acid degradation, which may lead to fluctuations in recovery rates and quantification biases.42–44 Therefore, establishing standardised pre-processing procedures and introducing a robust system of internal control calibration and quality control are key to improving inter-batch consistency. Secondly, complex biological samples may induce non-specific activation or excessive cleavage of Cas12a,45,46 whilst RCA efficiency is also affected by primer secondary structure, reaction conditions and mismatches in highly homologous miRNAs,40 which may result in false positives, signal attenuation or false negatives. At the same time, non-specific adsorption and background fluorescence on the SERS substrate may compromise signal stability and the linear range of quantification.47 In future, the thermodynamic design of the probes and their mismatch recognition performance should be optimised, and SERS substrates with enhanced resistance to contamination and interference should be developed to reduce detection bias. Thirdly, the limited number of clinical samples in this study is insufficient to fully reflect population and disease-stage heterogeneity; subtle differences in miRNA expression in early-stage lesions may therefore be underestimated, thereby limiting the generalisability of the sensitivity, specificity and AUC.48 Further work is required to confirm the robustness and universality of this platform for early CRC screening through multicentre, large-sample prospective cohorts, combined with cross-platform validation and long-term follow-up of clinical outcomes. Finally, multiplex testing continues to face challenges such as cross-reactivity between targets, serum matrix contamination and the standardisation of multi-channel signals. In particular, the combined detection of exosomal surface proteins and internal miRNAs, whilst providing richer biological information, places higher demands on the standardisation of pre-processing and detection procedures. In the future, the development of anti-interference SERS substrates, the optimisation of CRISPR/Cas12a orthogonality design, and the validation of multi-marker models in large-scale, multi-centre cohorts could facilitate the translation of this platform from laboratory research to a clinical screening tool, thereby providing technical support for the earlier, more precise and non-invasive detection of CRC. Acknowledgments Some of the material for Figure 1 was created using figdraw (figdraw.com). Funding Statement This work was supported by Soft Science Research Project of Yangzhou Association for Science and Technology (grant number 2025248), CSCO-Stone Medicine Cancer Research Fund Project (grant number Y-SY2024MS-0191), Yangzhou Hongquan Hospital Project (grant number yzhgyy 2022-3). Ethical Approval The study was conducted in strict compliance with the ethical principles outlined in the Declaration of Helsinki and was approved by the Medical Ethics Committee of the Changshu Hospital Affiliated with Nanjing University of Chinese Medicine (Ethics No.: 2025178). Author Contributions All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. Disclosure The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. References