All the blood samples of CRC patients were before treatment, and the specimens were collected on the same day when the patients finished collecting the test. The collection of serum specimens was between March 2021 and April 2022; we collected 205 samples from CRC patients and 201 samples from healthy donors (screened from a population of healthy physical examination) from The Affiliated Taian City Central Hospital of Qingdao University. The selected sample type is 3mL of coagulation promoting whole blood containing separation gel and centrifuged at 4°C for 1000g for 10 minutes. We collected the supernatant into a 1.5ml EP tube and stored it in a -80°C refrigerator. The case details of colorectal patients were reviewed: gender, age, cancer type, patient’s lifestyle habits, and tumor metastasis.

In the training stage, these 20 dysregulated Ex-tRF RNAs were determined by qRT-PCR using 48 HDs and 48 CRC samples. In the testing stage, the identified Ex-tRF RNAs were validated in serum EV samples of 72 HDs and 72 CRC patients. In the last stage, 166 samples (81 Health donors and 85 CRCs) were collected to further evaluate the Ex-tRF RNAs in CRC.

All collected serum samples 1.5mL were thawed on ice and centrifuged at 10,000g for 30 min at 4°C. The supernatant was then centrifuged at 100,000g for 2 hours at 4°C using a Type 50.4 TiRotor (Beckman Coulter) to isolate the EV. The extraction methods followed those described by Zhang ZJ, Xie L et al. (23, 24). Extracted EV size and distribution were determined through a particle size potentiostat qNano (New Zealand) The parameters set for measuring particle size in qNano include a size determination accuracy of 1nm, a concentration range of 10 ^ ⁵-10 ^ ¹³ particles/mL, and a sample size of 40μL. The EV were placed on a copper grid with a 50 µl drop of 1% glutaraldehyde, and transferred to 100 µl distilled water after 5 min. The grids were left undisturbed for 2 min and stained with 50 µl oxalyl uranyl solution (pH 7) for 5 min. The grids were rinsed 7 times with distilled water for 2 min. We observed their physical morphology using a transmission microscope FEI Tecnai T20e (FEI Company, USA).

Western blotting was performed for the identification of EV proteins on PVDF membranes (Bio-Rad, America, 1620177). Reference (25) was consulted to obtain EV surface high expression of proteins TSG101(Proteintech, America, 14497-1-AP), CD63(Proteintech, America, 25682-1-AP) and CD81(Proteintech, America, 66866-1-Ig). The secondary antibody followed a 1h incubation at 37°C, using an ECL reagent (Bio-Rad, America, 1705060) to visualize the bands.

The extracted EV were resuspended with 1 mL of PBS (Solarbio, China, P1022-500ml), adding 1 mL of Trizol (Invitrogen, America, 10296010-100ml) to fully lyse the EV to extract RNA, reversely transcribing the extracted RNA into cDNA (AG, China, 11701) for the subsequent qPCR (AG, China, 11717). The experimental procedure was susceptible to RNA degradation during operation; all the procedures were always performed on ice.

Herein, we sampled EV from three groups of healthy controls and six groups of CRC patients by microarray (Arraystar Small RNA Expression Chip, Kangcheng Biotechnology, China). In total, 22.5μL 2X Hybridization buffer (Agilent) was combined with the completed labeling process, resulting in a 45μL final volume. The mixture was subjected to heat of 100°C for 5 min and subsequently rapidly cooled to 0°C. The labeled sample mix, measuring 45 μL, was subjected to hybridization on a microarray at 55°C for 20 h. The slides were immersed in a solution of 6X SSC with 0.005% Triton X-102 at room temperature for 10 min, followed by a solution of 0.1X SSC with 0.005% Triton X-102 for 5 min. The slide scan was conducted using an Agilent G2505C microarray scan, analyzing the acquired array images via Agilent Feature Extraction software (version 11.0.1.1). The raw intensities were subjected to log2 transformation and quantile normalization. Following normalization, we only retained the probe signals that had Present (P) or Marginal (M) QC flags in at least three out of nine samples. Differentially expressed small RNAs were compiled by fold change and P-value cutoffs and annotated with genomic and biological info, scatter plots, volcano plots, and hierarchical clustering heatmap analyses. The FDR correction algorithm for P-value is the Benjamini-Hochberg method, and the computation is done in the programming language (R).

The qPCR was performed to detect differentially expressed tRF RNA in healthy donors and CRC patients. The internal reference was U6, and we made statistical analyses based on ΔCT values (ΔCT=CT_CRC-CT_U6) (23, 26). The reagents used are SYBR Green (AG, China). Each data set was repeated independently three times (P<0.001). Statistical analyses were conducted through GraphPad Prism 6.0 (CA, USA) and SPSS 22.0 (Ehningen, Germany). The normally distributed numeric variables were analyzed by t-test, whereas non-normally distributed variables were evaluated by Mann–Whitney test. Analyzed the statistical significance of health donors and CRC patients. P < 0.05 indicated statistically significant. The diagnostic efficiency was evaluated using the receiver operating characteristic curve (ROC), represented by AUC curve.

Herein, we used ultracentrifugation to extract EV, which are vesicles that are the basis for the study of tumor metastasis and information transfer. We characterized and tested the properties of EV. The morphology of EV showed a rounded spherical shape, with a diameter distribution between 80–130 nm and a normal distribution of diameter size, mainly concentrated around 100 nm ( Figures 1A, B ). Western blot experiments verified the surface protein expression of EV vesicles, which was significantly different from the cell surface protein expression. The surface of EV highly expressed TSG101, CD81/63 proteins ( Figure 1C ). The above results showed that we successfully extracted EV. The clinical data of the patients were collected for pathological analysis, and 3’tRF-ThrCGT and 3’tRF-mtlleGAT did not statistically significantly differ in the patients’ clinicopathological information in terms of gender, age, life habits, pathology type, and tumor metastasis.

Analysis of serum microarray results from CRC patients and normal controls revealed that 12 tRF RNAs were up-regulated and 12 tRF RNAs were downregulated ( Table 1 ). Figure 2 presents an analysis of the following three groups: healthy controls, nonmetastatic CRC patients, and metastatic CRC patients. The clustering diagram results show differentially expressed tRF RNAs. In the subsequent experiments, we validated tRF RNA expression in the serum of 205 cases of CRC through quantitative PCR analysis. The microarray data in this study has been stored in genome sequence archives (Genomics, proteomics and bioinformatics) Bioinformatics/Beijing Institute of Genomics, Chinese Academy of Sciencesaccessible to the public https://ngdc.cncb.ac.cn/omix/preview/KYegl8Sk (OMIX007492).

We have collected serum EV for reverse transcription and real-time quantitative PCR (ΔCT=CT_CRC-CT_U6) to explore the expression of serum EV 3’tRF-ThrCGTand 3’tRF-mtlleGAT in 205 colorectal cancer patients and 201 normal healthy donors. We verified the non-specific nature of the primers, and the melting curve showed a single peak with good specificity. The results are shown in Supplementary Data Sheet 1 . The results of quantitative PCR elucidated 3’tRF-ThrCGT significant down-regulation while 3’tRF-mtlleGAT significant overexpression in the sera of cancer patients ( Figures 3A, B ). In order to display the research data more intuitively, we used the -ΔCT=-(CT_CRC-CT_U6) formula calculations for 3’tRF-ThrCGT. The data we obtained through quantitative PCR is consistent with the data from the microarray in Figure 2 .

To ascertain the correlation between patients’ survival and tumor stage grading of CRC, early diagnosis of CRC occurrence has an important role in patients’ survival improvement. We screened that 3’tRF-ThrCGT was not statistically significant with T-stage grading of CRC primary foci ( Figure 4A ). CRC tumor metastasis is more pronounced in advanced stages, and tumor findings are generally more frequent in these patients compared to those in earlier stages. Clinical data show that lymph node metastasis (LNM) on N0/N1 was statistically significant (P <0.05) ( Figure 4B ). The number of patients with LNM found at the initial stage was higher compared to those without LNM. In the staging of the tumor stage I/III, there was statistical significance *P<0.05 ( Figure 4C ). The results showcased no statistical significance between stages I, II, and IV. 3’tRF-mtlleGAT was not statistically significant in the analytical data of the clinical staging of CRC patients in comparison with the tumor of the primary location. Stage T ( Figure 4D ), LNM ( Figure 4E ), and the clinical stage of the tumor ( Figure 4F ) did not significantly differ. Reviewing the lifestyle habits and case information of CRC patients revealed no statistically significant differences ( Table 2 ).

3’tRF-ThrCGT and 3’tRF-mtlleGAT were used as diagnostic CRC indicators, assessing the diagnostic rate using AUC. The 3’tRF-ThrCGT diagnostic rate as a predictor for qPCR screening to diagnose CRC was 0.669 ( Figure 5A ), and the 3’tRF-mtlleGAT diagnostic rate alone was 0.656 ( Figure 5B ). The combined diagnosis of 3’tRF-ThrCGT and 3’tRF-mtlleGAT was 0.777 {95% confidence interval (CI) 0.732–0.822} ( Figure 5C ). The combined diagnosis of CRC by tRF RNAs is more efficient and provides a basis for minimally invasive diagnosis in the clinic.

Currently, tumor-related antigens are mostly utilized as a diagnostic basis for CRC clinical prediction. Carcinoembryonic antigen (CEA) is a tumor-related antigen with low specificity for the prediction of CRC, and it is difficult to determine the organ of origin of the cancer using CEA alone, which has an elevated expression in the process of numerous tumorigeneses. Therefore, it is necessary to combine with other tumor markers for diagnosis. Glycoantigen724 (CA724) is a laboratory marker for detecting gastric cancer and various digestive cancers, which is mainly seen in the gastrointestinal tract and has a higher sensitivity for gastric cancer, ovarian mucinous cystadenocarcinoma, and non-small cell lung cancer, biliary system tumors, CRC, and pancreatic cancer. Combined diagnostic significance, the 3’tRF-ThrCGT combined diagnostic test has significantly improved CRC diagnostic efficiency. 3’tRF-ThrCGT and CEA diagnostic efficiency was 0.827 ( Figure 6A ), 5 ‘tRF-GlyGCC and CA724 diagnostic efficiency was 0.848 ( Figure 6B ), 3’tRF-ThrCGT combined with the two tumor markers diagnostic rate was 0.911{95% CI 0.880–0.942} ( Figure 6C ). 3’tRF-mtlleGAT combined with CEA and CA724 to detect their AUCs, respectively, and it was found that the results of ( Figure 6D ) study showed that 3’tRF-mtlleGAT and CEA diagnostic efficiency was 0.838, 3’tRF-mtlleGAT and CA724 diagnostic efficiency was 0.851 ( Figure 6E ), and 3’tRF-mtlleGAT combined with CEA and CA724 diagnostic rate was 0.919 {95%CI 0.889–0.948} ( Figure 6F ). The diagnostic rate of 3’tRF-ThrCGT and 3’tRF-mtlleGAT combined CEA was 0.878 ( Figure 6G ), and the diagnostic rate of 3’tRF-ThrCGT and 3’tRF-mtlleGAT combined CA724 diagnostic rate was 0.896 ( Figure 6H ), 3’tRF-ThrCGT and 3’ tRF-mtlleGAT combined with both swelling markers diagnostic rate was 0.938 {95% CI 0.913–0.963} ( Figure 6I ).