Circulating N6-methyladenosine RNA as a diagnostic and a stage-associated prognostic biomarker in colorectal cancer

Abstract

Background

Extracellular vesicle (EV)-based liquid biopsy is increasingly recognized as a promising strategy for cancer diagnosis and prognosis, as EVs carry abundant, stable biomolecular cargo. N6-methyladenosine (m6A), the most prevalent modification in eukaryotic intracellular RNA, plays a critical role in regulating diverse cellular processes and has been implicated in tumor initiation and progression. However, the potential of EV-associated m6A-modified RNA (EV-m6A RNA) as a clinically useful biomarker for cancer detection remains unclear.

Methods

EV-RNA was isolated from tumor tissues and matched serum samples collected from 76 colorectal cancer (CRC) patients. Serum samples from 30 healthy donors were included as controls. Serum EVs were prepared using the ExoQuick™ precipitation reagent. EV-RNA was extracted using Trizol reagent, and EV-m6A RNA levels were quantified using an enzyme-linked immunosorbent assay (ELISA).

Results

Precipitation-enriched serum particles (PESPs) isolated from CRC patients and healthy donors were identified as CD9(+)/CD63(+)/ALIX(+) small particles, with a mean particle diameter of 60–70 nm. The ApoB expression was detectable in PESPs. The mean PESP concentration and PESP-RNA yield were significantly higher in CRC patients than in healthy controls (PESP concentration: 11 ± 8 × 10¹²/mL vs. 5 ± 2 × 10¹²/mL, p < 0.001; PESP-RNA yield: 57 ± 52 ng/µL vs. 22 ± 15 ng/µL, p < 0.001). Through m6A modification-specific ELISA quantification, receiver operating characteristic (ROC) curve analysis demonstrated good discriminatory performance of normalized PESP-m6A RNA levels for CRC detection (AUC, 0.8346; 95% CI, 0.7583–0.9110; p < 0.0001). PESP-m6A RNA levels were noted to be higher in late-stage (III/IV) CRC patients than in those with early-stage (I/II) disease (0.015 ± 0.001% vs. 0.009 ± 0.005%, p < 0.001) with no significant correlation with intratumoral METTL3 expression, a m6A writer. Increased PESP-m6A RNA abundance further predicted a worse overall survival (OS) in CRC patients (p = 0.0107; HR, 3.938), while m6A RNA levels in tumor tissues showed no significant prognostic value (p = 0.7765; HR, 1.153).

Conclusion

These findings support circulating PESP-m6A RNA levels as a feasible and noninvasive biomarker for CRC diagnosis, with additional potential for liquid biopsy-based prognostication and longitudinal disease monitoring.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-026-16290-x.

Introduction

Extracellular vesicles (EVs) are a heterogeneous population of membrane vesicles with lipid-bilayer and nanoscale properties released from various cell types and present in body fluids. EVs can be classified into diverse subtypes based on biogenesis and size distribution. Exosomes are one of the small EVs (sEVs) with a diameter range of 30–150 nm [1]. Exosomes present several markers on the surface like tetraspanins (CD63, CD9, and CD81) [2] and major histocompatibility complex (MHC) class I [3], and carry a diverse cargo of biomolecules including proteins, lipids, and nucleic acids: DNA, messenger RNA (mRNA), noncoding RNA (ncRNA) [4], which enable them to deliver messages to recipient cells, mediating in cell-cell communication.

EVs enable liquid-biopsy detection of nucleic acid and protein biomarkers, with plasma and serum favored for their higher EV abundance [5]. In colorectal cancer (CRC), EV-derived DNA (evDNA) outperforms plasma cell-free DNA (cfDNA) by droplet digital PCR (ddPCR) for KRAS G12D/G13D detection across TNM stages [6]. By next-generation sequencing (NGS), evDNA achieves higher sensitivity and specificity than cfDNA by ddPCR [7]. The non-coding EV markers include serum exosomal miR-125a-3p [8], lncRNA NAMPT-AS [9], and circ-KLHDC10 [10]. EV protein cargo is informative: plasma EV fibrinogen α chain (FGA) rises with progression and outperforms CEA/CA19-9 for early-stage detection [11], while targeted proteomics identified EV proteins, with > 80% specificity for stage I/II [12]. EV-implemented assays complement traditional methods; multi-analyte liquid biopsy using cfDNA, evDNA, and EV-associated RNA (evRNA) capture cancer evolution [13].

N6-methyladenosine (m6A) is a prevalent internal RNA modification across mRNA and ncRNA [14]. It is dynamically installed by writers, such as the METTL3-METTL14 complex, removed by erasers, such as FTO, and interpreted by readers, including YTHDF1 and IGF2BP2, to shape RNA fate and gene expression collectively [15]. Intracellular METTL3 drives metastasis via an m6A-IGF2BP2-SOX2 axis and is associated with worse patient survival [16, 17]. Cell-intrinsic expression of YTHDF1 further serves as a prognostic and diagnostic factor [18, 19]. Despite these associations, the clinical significance of m6A-modified evRNA remains unclear.

In this study, we isolated m6A RNA from tumor tissue and precipitation-enriched serum particles (PESPs) across tumor stages to investigate m6A RNA levels and compare them with those from healthy donors. The increased PESP-m6A RNA content in colorectal cancer (CRC) patients suggested its potential for noninvasive diagnosis. Additionally, CRC patients with high PESP-m6A RNA levels were associated with poor survival, supporting its potential as a biomarker for advanced CRC.

Methods

Human specimens

Paired 76 sera and tissue samples from CRC patients used in this retrospective study were obtained from the biobank of Taipei Veterans General Hospital, with approval from the institutional review board (IRB) (Approval number: 2023-01-014BC), following the ethical principles of the Declaration of Helsinki. A total of 30 healthy sera were used. 29 sera from healthy donors were collected at National Yang Ming Chiao Tung University under the IRB approval (Approval number: NYCU111182AE). Informed consent is waived. One commercially available pooled human serum was purchased from Rockland Immunochemicals (Catalog number: D119-00-0050).

Isolation of precipitation-enriched serum particles (PESPs)

PESPs were isolated from serum samples using the ExoQuick™ kit (System Biosciences, Palo Alto, CA, USA). 100 µL of serum was mixed with 25 µL of 1x phosphate-buffered saline (PBS, Bioman, New Taipei City, Taiwan) and 31.5 µL of ExoQuick reagent. The mixture was incubated on ice for 1 h. After incubation, the sample was centrifuged at 1,500 × g for 30 min at 4 °C, and the supernatant was discarded. The pellet was centrifuged at 1,500 × g for an additional 5 min at 4 °C, then the supernatant was removed. The resulting PESP pellet was collected for further analysis. The LipoMin™ reagent (Reliance Biosciences, Taipei, Taiwan) was used to remove lipoprotein contamination from PESPs. According to the manufacturer’s instructions, 0.5× LipoMin magnetic beads were added to the PESP at the recommended ratio for as size exclusion chromatography (SEC) samples. The mixture was incubated with rotation at 4 °C for 10 min. Following magnetic separation for 3 min, the supernatant was collected as lipoprotein-depleted PESPs.

Total RNA extraction

The PESP pellets or CRC tissues were resuspended in 50 µL of Dulbecco’s phosphate-buffered saline (DPBS) (Cytiva, Marlborough, MA, USA), added with 950 µL TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA), and incubated at room temperature for 10 min. Following this, 200 µL of chloroform (Sigma-Aldrich/Merck KGaA, Darmstadt, Germany) was added, and the mixture was incubated at room temperature for 3 min. The mixture was centrifuged at 12,000 × g for 15 min at 4 °C to obtain a transparent aqueous phase containing RNA. An equal volume of isopropanol (Sigma-Aldrich) was added, and the mixture was incubated overnight at -20 °C for RNA precipitation. The sample was centrifuged at 13,000 × g for 30 min at 4 °C, and the supernatant was discarded. The RNA pellet was washed with 1 mL of 75% cold ethanol and centrifuged at 13,000 × g for 10 min at 4 °C. The supernatant was discarded, and the pellet was air-dried. Finally, the RNA pellet was resuspended in 30 µL of nuclease-free water (Bioman) for downstream applications.

m6A RNA quantification

EpiQuik™ m6A RNA Methylation Quantification Kit (EpiGentek, Farmingdale, NY, USA) was used for m6A RNA quantification. 200 ng of RNA was used as the input for each reaction. Standards of different concentrations and working solutions were prepared, and the experimental steps were carried out according to the manufacturer’s instructions. After the reactions, the absorbance was measured at 450 nm using a microplate reader (Infinite M200 Pro, Tecan, Switzerland). A standard curve was generated for the absolute quantification of m6A levels.

Transmission electron microscope (TEM)

PESP pellets were resuspended in DPBS and filtered through a 0.45 μm filter (Millipore, Burlington, MA, USA), fixed with an equal volume of 4% paraformaldehyde (PFA, Sigma‒Aldrich/Merck KGaA) for 30 min at room temperature. The fixed PESP samples were diluted using nuclease-free water to prepare 100× and 1000× dilutions, then loaded onto a formvar/carbon-coated grid (Pelco, Fresno, CA, USA) for 40 min at room temperature. The grids were rinsed with nuclease-free water and dried for over 5 days. Images were captured using a JEOL JEM-1400plus (JEOL, Tokyo, Japan) at magnifications of 20,000×.

Tunable resistive pulse sensing (TRPS) particle analysis

PESP pellets isolated from 50 µL of serum were resuspended in DPBS and filtered through a 0.45 μm filter before analysis. Tunable resistive pulse sensing (TRPS) was conducted by an Exoid system employing an NP100 Nanopore and CPC100 standard calibration beads (Izon Bioscience, Christchurch, New Zealand). The Nanopore was stretched to 47 mm and wetted to verify the baseline current at 100 mV. It was then coated and rinsed with various buffers from the supplied Izon reagent kit. Before analysis, the NP100 Nanopore was calibrated with diluted CPC100 standard calibration beads (1500×) provided in the kit. Sample exosomes were diluted in 2× PBS, loaded into the Nanopore, and analyzed under the same pressure and voltage conditions as the calibration beads. The measurement conditions were as follows: stretch 47 mm, pressure 1000 cm H₂O, and a current of about 110–130 nA. Data recording was performed using the Exoid Control Suite software (version V1.0.0.181), and particle size ranges and concentrations were calculated using the Izon Data Suite (version V1.0.2.32).

Immunohistochemistry (IHC) staining

Paraffin-embedded tissue slides were deparaffinized by heating at 65 °C for 10 min, followed by immersion in ultra-clear solution (J.T. Baker, Phillipsburg, NJ, USA) three times, each for 10 min. The slides were rehydrated through a graded ethanol series: 100% ethanol (Nihon Shiyaku, Japan) for 5 min, twice; 90% ethanol for 5 min; 70% ethanol for 5 min; and 50% ethanol for 5 min, followed by rinsing with ddH₂O. For antigen retrieval, the slides were autoclaved at 120 °C for 10 min and immersed in 10 mM citrate buffer (Honeywell, Morris Plains, NJ, USA). After cooling, the slides were washed with 1× PBS. The tissue area intended for staining was encircled with a hydrophobic pen. Blocking was performed by treating the tissues with 3% hydrogen peroxide for 10 min, followed by two 5-minute washes with 1× PBS. Cell membrane permeabilization was achieved using 0.1% Triton X-100 (Bionovas, North York, ON, Canada) for 5 min, followed by washing with 1× PBS twice for 5 min. The tissues were then incubated overnight at 4 °C with the anti-METTL3 antibody (ABclonal, cat A8370) diluted 1:400 in antibody dilution buffer (Ventana). The tissues were washed twice with 1× PBS for 10 min each before incubation with anti-rabbit immunoglobulin (BioGenex, Fremont, CA, USA) for 30 min at room temperature. Next, the tissues were washed with 1× PBS for 10 min, twice, before being applied with streptavidin peroxidase (BioGenex, Fremont, CA, USA) for 20 min at room temperature, followed by 1× PBS for 10 min, twice. Next, tissues were treated with DAB solution (Epredia, Kalamazoo, MI, USA) for 45 s for visualization, and the reaction was stopped with 1× PBS. Finally, the tissues were counterstained with Mayer’s hemalum solution (Sigma-Aldrich/Merck KGaA) for 20 s and rinsed with ddH₂O. The slides were then mounted using Kaiser’s glycerol gelatin (Sigma-Aldrich/Merck KGaA). Images were captured under an Olympus BX43 microscope equipped with a DP22 CCD camera (Olympus, Tokyo, Japan), and the histology score (H-score) was calculated to quantify staining intensity. The H score was defined as the percentage of the METTL3-positive immunostained region (0 to 100) multiplied by the degree of METTL3 staining (0, 1, 2, and 3).

Western blot

PESP pellets isolated from 20 µL of serum were lysed with 40 µL of 1× radioimmunoprecipitation assay (RIPA) lysis buffer prepared from ddH₂O diluted 5× RIPA lysis buffer (T-Pro Biotechnology) containing 1% protease inhibitor (ThermoFisher, Waltham, MA, USA) on ice for 1 h, then centrifuged at 13,000 rpm for 15 min at 4 °C to extract protein supernatant. Subsequently, 40 µL of 2× sample buffer was added to the supernatant. The mixture was heated at 99 °C for 10 min. SDS-PAGE-separated samples were then transferred to a PVDF membrane (Millipore, Burlington, MA, USA). The membranes were blocked with 5% skim milk for 1 h, followed by three washes with 1× PBS containing 0.1% Tween 20 (Bioshop, Burlington, ON, Canada). The membranes were then incubated with the primary antibodies at 4 °C overnight in PBS-Tween20 (0.1%), followed by incubation with the relevant HRP-conjugated secondary antibodies (GeneTex, Irvine, CA, USA) at room temperature for 1 h on an orbital shaker. The membranes were soaked with chemiluminescent HRP substrate (Bioshop, Burlington, ON, Canada) and visualized using an ImageQuant LAS 4000 chemiluminescence detection system (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA). The primary antibodies used are as follows: Alix (Cell Signaling, cat 2171), CD9 (Abcam, cat ab92726), CD63 (EMD Millipore, cat CBL551), Calreticulin (Abcam, cat ab39897), TOM20 (ABclonal, cat A16896), and ApoB (ProteinTech, cat 20578-1-AP).

Statistical analysis

GraphPad Prism (version 10.1.2) was used for statistical analysis, including two-tailed unpaired Student’s t-test, Pearson correlation analysis, ROC curve analysis, chi-square test for correlation of clinicopathological variable and m6A abundance Survival differences were evaluated using Kaplan–Meier analysis with the log-rank (Mantel–Cox) test. Hazard ratios were estimated using univariate and multivariate Cox proportional hazards regression models, with p-value calculated by the Wald test. The Wilson/Brown test performed 95% confidence intervals (95% CI). A p-value less than 0.05 was considered statistically significant. Python visualized the confusion matrix for a cancer prediction model on Google Colab. The calculating formula was following: accuracy = (TP + TN)/(TP + TN+FP + FN), precision = TP/(TP + FP), sensitivity = TP/(TP + FN), specificity = TN/(TN + FP).

Characterization of PESPs from CRC patients and healthy donors

A total of 76 CRC patients were included in the analysis. The early-stage group contained 16 of stage Ⅰ and 20 of stage Ⅱ; the late-stage group contained 20 of stage Ⅲ and 20 of stage Ⅳ; and their median ages were 66 years (range, 22–84 years) and 62.5 years (range, 34–84 years), respectively.

All PESPs were isolated using ExoQuick™ precipitation solution and characterized according to MISEV2023 guidelines, including quantification, protein markers, and single-vesicle analysis [20]. Two healthy controls and four tumor samples were randomly selected for PESP characterization. For protein marker identification, small extracellular vesicle (sEV) markers, including ALIX, CD9, and CD63 [21], were enriched in PESPs from patients and healthy controls. PESPs were free from organelle contamination, including mitochondria (TOM20) and endoplasmic reticulum (Calreticulin) (Fig. 1A). The lipoprotein marker, ApoB, was detected in PESPs (Supplementary Fig. 1A). Under the transmission electron microscope (TEM), the images showed that PESPs were about 60 nm in diameter and had bilayer membranous structures (Fig. 1B). Single-vesicle analysis of PESPs from representative samples using tunable resistive pulse sensing (TRPS) showed a size distribution of 50–70 nm, consistent with TEM results (Fig. 1C). The mean particle diameters of PESPs from CRC patients and healthy donors were about 70 nm (Fig. 1D), and the mode particle diameters were about 60 nm (Fig. 1E). The d90/d10 ratio showed no statistically significant differences between the two groups (Fig. 1F). The particle concentration of PESPs from CRC patients was higher than that from healthy donors (Fig. 1G). Moreover, PESP RNA amounts were elevated in CRC patients (Fig. 1H). These findings support PESP RNA as a viable target for detection. However, depletion of lipoproteins using LipoMin™ reagent reduced the abundance of sEV amount as indicated by reduced CD9 expression (Supplementary Fig. 1A-1B).

Fig. 1.

Biochemical characterization of purified PESPs. A: Western blot analysis shows the expression of sEV markers (ALIX, CD9, and CD63) and the organelle markers, including calreticulin (endoplasmic reticulum) and TOM20 (mitochondria), in PESPs. 10 µg of whole-cell lysate (WCL) from 293T cells is used as a control. The equivalent volume of PESP proteins is taken for each lane. H1 and H2, healthy PESPs; T1 to T4, tumor PESPs. M.W., molecular weight. B: Representative TEM images of PESPs from two cases. Scale bar: 100 nm. C: The particle size distribution of PESPs was analyzed using TRPS with Exoid. Mean, average value of recorded particle diameter. Mode is the particle diameter that occurs most frequently. D90, the particle diameter at the 90th percentile of the distribution. D10 is the particle diameter at the 10th percentile of the distribution; D90 / D10 index is the ratio of the 90th percentile diameter to the 10th percentile diameter, representing the relative distribution range. D-F: Histograms demonstrate the mean diameter (D), the mode diameter (E), and the D90 / D10 index (F) of PESPs. G: Histogram exhibits the particle number of PESPs. H: Histogram displays the RNA amount of PESPs. Data represented mean ± SD. *p < 0.05, ***p < 0.001; ns, not significant (Student’s t-test)

Detection of PESP-derived m6A RNA abundance as diagnostics

To investigate the clinical significance of circulating m6A RNA in CRC, whole PESPs were used to maximize sEV yield. The abundance of PESP-m6A RNA was quantified in all stages of CRC and in healthy controls. The receiver operating characteristic (ROC) curves were used to assess sensitivity and specificity across all possible threshold values [22]. The confusion matrix calculated the accuracy and precision at a fixed threshold [23]. Regardless of tumor staging, when the cut-off value of normalized PESP-m6A RNA level was 0.00867%, the PESP-m6A RNA level could differentiate CRC patients and healthy controls with an area under curve (AUC) of 0.8346 (95% CI, 0.7583–0.9110), a sensitivity of 78.95% (95% CI, 68.50% – 86.60%), a specificity of 83.33% (95% CI, 66.44% – 92.66%) (Fig. 2A). An accuracy of 80.19% (95% CI, 71.60% – 86.66%), and a precision of 92.31% (95% CI, 83.22% – 96.67%) were observed under the setting (Fig. 2B), suggesting PESP-m6A RNA level could be considered a diagnostic molecule Furthermore, the PESP-m6A RNA level of CRC was increased from early-stage (I/II) to late-stage (III/IV) (Fig. 2C).

Fig. 2.

Increased PESP-m6A RNA abundance in CRC patients. A: ROC curve analysis for distinguishing PESP-m6A RNA-positive and -negative patients. p < 0.0001, standard approximation method. B: A confusion matrix to visualize calculated accuracy, precision, sensitivity, and specificity. TN, true negative. FN, false negative. TP, true positive. FP, false positive. C: The scatter plot shows the PESP-m6A RNA contents of CRC patients and healthy donors. Data represented mean ± SD; **p < 0.01, ***p < 0.001 (Student’s t-test)

Correlation between m6A RNA levels in CRC tissues and circulating PESPs

Given the increased abundance of PESP-m6A RNA in CRC patients, we sought to elucidate potential sources. As METTL3 is the dominant m6A RNA writer [16, 17], we examined METTL3 protein expression by immunohistochemistry (IHC) (Fig. 3A). We found that in situ METTL3 expression positively correlated with m6A RNA levels in late-stage (III/IV) tumor tissues, but not in early-stage (I/II) CRC tissues (Fig. 3B). Interestingly, in early-stage CRC patients, METTL3 expression was positively correlated with PESP-m6A RNA levels, but not in late-stage samples (Fig. 3C). Furthermore, no significant association between m6A RNA in CRC tissues and PESPs was noted (Fig. 3D), suggesting highly dynamic m6A RNA biogenesis and transportation during cancer progression.

Fig. 3.

Correlation between m6A RNA abundance and METTL3 immunoreactivity in CRC patients. A: Representative images of METTL3 IHC staining. The photos on the upper right show enlarged views of the representative areas. Scale bar: 100 μm. B: The scatter plot depicts the correlation between m6A RNA contents in CRC tissues and METTL3. C: The scatter plot illustrates the correlation between PESP-m6A RNA contents and the METTL3 expression. D: The scatter plot shows the correlation between m6A RNA contents in PESPs and CRC tissues. *p < 0.05, t-test. r: Pearson correlation coefficient

Circulating PESP-m6A RNA status is associated with overall survival and advanced clinicopathological features in CRC patients

To assess the clinical relevance of PESP-m6A RNA status, patients were classified into low- and high-m6A groups and analyzed against overall survival (OS) and clinicopathological features. KM analysis demonstrated that patients with high circulating PESP-m6A RNA had significantly worse OS than those with low circulating PESP-m6A RNA (log-rank p = 0.0107) (Fig. 4A). In contrast, CRC patients with high m6A RNA levels in CRC tissues showed no difference in OS compared with those with low m6A RNA levels (log-rank p = 0.7765) (Fig. 4B), highlighting the prognostic potential of PESP-m6A RNA. Consistent with this finding, univariate Cox regression showed that high circulating PESP-m6A RNA was associated with poorer survival (HR 3.938, 95% CI, 1.366–14.13, p = 0.018) (Supplementary Fig. 2A). In the same univariate analysis, late-stage disease was also significantly associated with worse OS (HR 12.92, 95% CI, 2.618–233.6, p = 0.0132), whereas gender, age ≥ 65 years, and tumor location were not statistically significant (Supplementary Fig. 2A).

Fig. 4.

CRC patients with high PESP-m6A RNA abundance are associated with poor overall survival. A: Kaplan-Meier survival plot showing the overall survival rate of 76 CRC patients with high and low PESP-m6A RNA abundance. B: The overall survival rate of CRC patients with indicated tumor m6A RNA abundance. C: Utilization of liquid biopsy-based PESP-m6A RNA for CRC detection. PESP-m6A RNAs serve as a minimally invasive biomarker option for cancer detection, enabling cancer diagnosis and post-treatment cancer management. All components in the schematic are sourced from BioRender. PESP, precipitation-enriched serum particle

To further determine whether circulating PESP-m6A RNA status represented an independent prognostic factor, multivariate Cox regression was performed. In this model, late-stage disease remained significantly associated with inferior OS (HR 14.84, 95% CI, 1.993–304.9, p = 0.0204), whereas PESP-m6A RNA status lost statistical significance (HR 1.076, 95% CI, 0.3145–4.737, p = 0.9133) (Supplementary Fig. 2B). Tumor location showed a trend in the right/transverse colon subgroup, containing ascending colon, cecum, and hepatic flexure compared with the left/rectosigmoid subgroup, containing rectum, sigmoid colon, splenic flexure, descending colon, rectosigmoid junction, and transverse colon (HR 0.5738, 95% CI, 0.1665–2.286, p = 0.3929), but this did not reach statistical significance (Supplementary Fig. 2B).

Clinicopathological association analysis further showed that circulating PESP-m6A RNA status was strongly associated with tumor stage. Among patients with low m6A, 29 of 37 cases (78.4%) were early stage and 8 of 37 cases (21.6%) were late stage, whereas among patients with high m6A, only 7 of 39 cases (17.9%) were early stage and 32 of 39 cases (82.1%) were late stage (chi-square p < 0.001). In contrast, m6A RNA status was not significantly associated with gender (p = 0.259) or age ≥ 65 years (p = 0.362) and tumor location (p = 0.057) (Table 1). In contrast, m6A RNA levels in CRC tissues in situ showed no significant association with age, gender, tumor location, or tumor stage (Table 2). Taken together, these results suggest that circulating PESP-m6A RNA status primarily reflects disease aggressiveness and tumor stage in this dataset.

Table 1.

Association between clinicopathological characteristics and circulating PESP-m6A RNA abundance in CRC patients (N = 76)

		Low m6A (N = 37)a		High m6A (N = 39)a		p-valueb
Characteristic	Category	N	(%)	N	(%)
Age (years)	< 65	17	45.9	22	56.4	0.362
	≥ 65	20	54.1	17	43.6
Gender	Male	18	48.6	24	61.5	0.259
	Female	19	51.4	15	38.5
Location	Right/proximal	4	10.8	11	28.2	0.057
	Left/distal	33	89.2	28	71.8
Stage (AJCC VII)	Early (I/II)	29	78.4	7	17.9	< 0.001
	Late (III/IV)	8	21.6	32	82.1

^aMedian m6A RNA abundance is set for patient stratification

^bp-values were estimated using the chi-square test

Table 2.

Association between clinicopathological characteristics and tumor tissue m6A RNA abundance in CRC patients (N = 76)

		Low m6A (N = 38)a		High m6A (N = 38)a		p-valueb
Characteristic	Category	N	(%)	N	(%)
Age (years)	< 65	19	50.0	20	52.6	0.819
	≥ 65	19	50.0	18	47.3
Gender	Male	18	47.4	24	63.2	0.166
	Female	20	52.6	14	36.8
Location	Right/proximal	8	21.1	7	18.4	0.733
	Left/distal	30	78.9	31	81.6
Stage (AJCC VII)	Early (I/II)	18	47.4	18	47.4	> 0.999
	Late (III/IV)	20	52.6	20	52.6

^aMedian m6A RNA abundance is set for patient stratification

^bp-values were estimated using the chi-square test

Discussion

Minimally invasive strategies have been developed for CRC screening and early cancer detection with distinct advantages and limitations. Among these screening approaches, the fecal immunochemical test (FIT) detecting occult bleeding showed a pooled sensitivity of 79% and specificity of 94%, remaining the most widely implemented method because of its low cost, convenience, and suitability for large-scale screening [24]. The serum tumor biomarkers, like carcinoembryonic antigen (CEA), and carbohydrate antigen 19 − 9 (CA19-9), are commonly used for CRC screening. However, they may also be elevated in benign disease and generally show limited sensitivity. A systematic review reported a specificity of 89%, but a sensitivity of only 46% for CEA, and a lower sensitivity of 30% for CA19-9 [25].

Other liquid biopsy methods based on circulating tumor-derived biomarkers, such as circulating tumor cells (CTCs) and cfDNA, have emerged as promising alternatives for early CRC detection and provide further biological information. A meta-analysis demonstrated that CTC-based detection achieved sensitivity and specificity of 82% and 97%, respectively, whereas cfDNA-based detection achieved sensitivity and specificity of 76% and 88%, respectively [26]. Nevertheless, CTC detection remains technically challenging due to the low abundance and heterogeneity of CTCs in peripheral blood (PB). Similarly, cfDNA is often highly fragmented and present at low concentrations, which may compromise analytical sensitivity [27]. In contrast, EV possess several unique advantages, including high stability in biological fluids, protection of nucleic acids from degradation, and the ability to reflect dynamic tumor biological activity and cell-to-cell communication. Our study firstly compares m6A RNA abundance in CRC tumor tissue and paired circulating PESPs and highlights the potential of PESP-m6A RNA as a minimally invasive diagnostic option, with the findings that PESP-m6A RNA abundance was significantly elevated in CRC patients compared with healthy controls, with a sensitivity of 78.95%, specificity of 83.33%, and an AUC of 0.8346. CRC patients with high PESP-m6A levels were markedly enriched in advanced stages, and associated with poorer survival, indicating its prognostic value.

Although multiple studies have demonstrated that blood EVs are promising candidates for liquid biopsy, a standardized protocol for sample collection and EVs isolation remains poorly defined. According to the MISEV2023 guideline, improper blood collection, processing, and storage can introduce significant variability and affect EV quality. Freeze-thaw cycles of the blood sample may disrupt EV membranes or generate cellular debris of similar size to EVs. Second, the presence of abundant soluble proteins and lipoproteins, which share similar size or density with EVs, makes separation challenging [20]. An overnight fasting period before blood collection is recommended to avoid excess chylomicrons [28, 29]. Moreover, EV isolation remains a key methodological consideration in EV biomarker studies, as different strategies offer distinct balances of yield, purity, scalability, and clinical practicality. The MISEV2023 guideline emphasizes that there is no single optimal EV separation method and that the choice should be guided by the scientific question and downstream application [20]. Differential ultracentrifugation has long been a conventional EV enrichment approach and remains useful for mechanistic studies, whereas size-exclusion chromatography (SEC) has gained increasing attention as a clinically relevant alternative due to its favorable EV purity and representative marker profiles. Immunoaffinity-based approaches can further enrich specific EV subpopulations, although this selectivity may be less suitable for unbiased biomarker discovery [30]. Nevertheless, plasma EVs isolated by ultracentrifugation alone often retained albumin, whereas those isolated by SEC alone tended to retain lipoproteins [31]. Although combining more than two isolation techniques can significantly reduce these common contaminants, it often results in a substantial loss of EV yield [32].

Within this framework, precipitation-based methods remain attractive for exploratory biomarker studies because they are technically simple, do not require specialized equipment, and can recover EVs or evRNA and evDNA from limited input material, making them practical for targeted assays in early translational research [33, 34]. However, precipitation-based methods may also recover a broader range of extracellular components, including lipoproteins and soluble proteins, which should be considered when interpreting the data [35]. Lipoproteins were also detected in PESPs, and lipoprotein depletion using LipoMin™ reagent [36] may partially reduce the EV population within this fraction. Although PESP-m6A levels show promise as diagnostic and cancer progression biomarkers, potential contamination by free m6A-modified RNA from non-EV sources warrants further evaluation. The impact of co-precipitated non-EV components on downstream EV-RNA analysis should therefore be carefully assessed. Moreover, identifying the cell origin of PESP-m6A RNA and specific m6A-modified RNA species present in PESPs will be important for clarifying their biological relevance.

We observed that METTL3 expression in tumor tissues was positively correlated with PESP-m6A RNA levels in early-stage CRC, whereas this association was lost in late-stage disease. This decoupling suggests that circulating PESP-m6A RNA may not merely reflect passive levels of intracellular or extracellular m6A, but may instead be dynamically regulated during tumor progression. In early-stage CRC, METTL3-mediated modification of PESP-associated RNA may contribute to the early establishment of the tumor niche. Accumulating evidence suggests that m6A reader proteins, particularly members of the IGF2BP and YTHDF families, are associated with aggressive tumor behavior. Specifically, IGF2BP2 and IGF2BP3 have been reported to stabilize multiple oncogenic m6A-modified transcripts, including VEGFA, EphA2 [37], and CREB1 [38] in CRC cells, thereby promoting angiogenesis, proliferation, and metastasis in CRC. These findings raise the possibility that recipient cells may utilize PESP-m6A RNA following intercellular transfer, potentially amplifying oncogenic signaling pathways through m6A reader-mediated RNA stabilization and translation. Given that evRNA can remodel the tumor microenvironment (TME) through tumor-host interactions and promote cancer progression [39, 40], PESP-m6A RNA may help shape a distinct TME composition that favors early tumor development. Defining the role of METTL3 in shaping the intracellular and extracellular m6A epitranscriptome at the single-cell level may advance the development of m6A-based strategies for early cancer detection.

In addition to METTL3, altered expression of other m6A writers and erasers has been reported to contribute to CRC progression. The m6A writer KIAA1429 promotes CRC progression by stabilizing oncogenic transcripts such as SIRT1 and SOX8 in an m6A-dependent manner [41], whereas the m6A eraser FTO enhances drug resistance and metastatic potential by inhibiting SIVA1-mediated apoptosis [42]. These suggest that global m6A regulation in late-stage CRC may no longer be predominantly determined by METTL3 alone. In parallel, malignant progression may also involve alterations in EV biogenesis and cargo sorting [43–45], leading to a diverse intracellular /extracellular m6A RNA distribution.

Recent automated EV-RNA extraction platforms using magnetic beads or digital microfluidics can process small plasma volumes within 30 min [46, 47], highlighting their potential for rapid clinical testing. In parallel, integrated nano-biochip systems capable of EV capture and marker detection have shown promise for early cancer liquid biopsy applications [48]. For m6A RNA quantification, m6A-ELISA may be more suitable for hospital-based implementation than LC-MS [49–51] or nanopore sequencing [52, 53] because of its lower cost, simplicity, scalability, and faster turnaround. Looking forward, integrating EV isolation, RNA extraction, and m6A detection into an automated microfluidic or lab-on-a-chip platform may improve the clinical feasibility of EV-based liquid biopsy by reducing sample and reagent consumption, shortening turnaround time, minimizing human error and analyte loss, and supporting standardized workflows for future point-of-care testing (POCT).

Several limitations should be considered when interpreting the present findings. First, the study cohort was relatively limited, particularly with respect to the number of healthy controls and overall survival events available for multivariable survival analysis, which may have reduced statistical power and limited generalizability. A large-scale validation with expanded control cohorts or an independent validation cohort is needed in future studies. Second, because this study was based on a Taiwanese population, caution is warranted when extrapolating the observed PESP-m6A RNA abundance to other geographic or ethnic populations. Third, although ExoQuick provides a practical and accessible platform for exploratory EV-RNA biomarker discovery, precipitation-based isolation methods may co-isolate non-EV extracellular components, potentially affecting analytical specificity. Finally, the biological functions of PESP-m6A RNAs were not directly investigated in the present study and require future mechanistic exploration.

Conclusion

Liquid biopsy has been considered a potential tool for cancer detection. This study demonstrated the value of PESP-m6A RNA as a biomarker potential for CRC diagnosis and prognosis, facilitating cancer management (Fig. 4C).

Abbreviations

AUC

Area under curve

CA19-9

Carbohydrate antigen 19 − 9

CEA

Carcinoembryonic antigen

cfDNA

Cell-free DNA

CTCs

Circulating tumor cells, CRC, colorectal cancer

ddPCR

Droplet digital PCR

ELISA

Enzyme-linked immunosorbent assay

EV

Extracellular vesicle

evDNA

EV-derived DNA

evRNA

EV-associated RNA

FGA

Fibrinogen α chain

FIT

Fecal immunochemical test

FN

False negative

FP

False positive

IHC

Immunohistochemistry

IRB

Institutional review board

m6A

N6-methyladenosine

NGS

Next-generation sequencing

OS

Overall survival

POCT

Point-of-care testing

ROC

Receiver operating characteristic

SEC

Size-exclusion chromatography

sEV

Small extracellular vesicle

PB

Peripheral blood

PESP

Precipitation-enriched serum particle

TEM

Transmission electron microscope

TME

Tumor microenvironment

TN

True negative

TP

True positive

TRPS

Tunable resistive pulse sensing

Authors’ contributions

ACC, HWT, and WLH conceptualized this study and directed the research; YTL and HYL conducted the experiments and analyzed the data with the support of CSC, CCL, HWT, and WLH. YTL wrote the original manuscript. ACC, CSC, CCL, HWT, and WLH revised the manuscript.

Funding

National Science and Technology Council (112-2314-B-038-146-MY3 to A-C.C., 113-2314-B-075-023 to H-W.T., and 112-2326-B-A49-002-MY3 and 111-2628-B-A49-017 to W-L.H). Ministry of Education, Higher Education SPROUT Project for Cancer and Immunology Research Center (114W031101 and 115W031101). A grant from the Yen Tjing Ling Medical Foundation (CI-111-15 and CI-115-24). Shin Kong Wu Ho-Su Memorial Hospital (2023SKHAND007).

Data availability

Data supporting the findings of this study are available within this article.

Declarations

Ethics approval and consent to participate

This study conforms to the principles of the Declaration of Helsinki and has been approved by the Institutional Review Board of Taipei Veterans General Hospital (2023-01-014BC) and the Institutional Review Board of National Yang Ming Chiao Tung University (NYCU111182AE). The informed consent has been waived.

Competing interests

The authors declare no competing interests.