Chromatographic grade acetonitrile was bought from Merck KGaA (Darmstadt, Germany). Methanol of HPLC grade was purchased from J.T.Baker (Radnor, PA, United States). Formic acid (HCOOH) was bought from Fluka (Muskegon, United States). Ammonium formate, malic acid and 5-methylcytidine (m⁵C) were bought from Sigma-Aldrich (St Louis, MO, United States). 2′-O-methyladenosine (Am), N ⁶-methyladenosine (m⁶A), 2′-O-methylguanosine (Gm), 1-methylguanosine (m¹G), 2′-O-methylcytidine (Cm), 2′-O-methyluridine (Um), 5-methyluridine (m⁵U) and isotopically labeled standards [D₃]m⁶A, [D₃]Um and [¹³C₅]m⁵U were obtained from Toronto Research Chemicals (Toronto, Canada). [¹³C¹⁵N₂]Gm, [¹³C¹⁵N₂]m¹G and [¹³C₅]Cm were synthesized according to literature (Fu et al., 2015). Water was purified by a Milli-Q water purification device (Millipore, Milford, MA, United States).

Acquity UPLC system (Waters, Milford, MA, United States) achieved by Empower Pro 6.0 software was applied for analysis. A Waters Acquity BEH Amide column (100 mm × 2.1 mm, 1.7 μm) was applied for chromatographic separation. 4000 QTRAP mass spectrometer (AB SCIEX, Foster City, CA, United States) was applied for MS detection. The mass spectrometer was equipped with electrospray ionization (ESI) positive ion mode. Multiple-reaction monitoring (MRM) was chosen to acquire data. Data acquisition and processing were controlled by Analyst 1.6.3 software.

The Ethics Committee of Medical Research of the Second Affiliated Hospital, Zhejiang University School of Medicine (SAHZU) approved our study. A total of 69 healthy volunteers (mean age of 43.9 ± 11.1 years, range from 30 to 70 years) and 61 patients with breast cancer (mean age of 52.2 ± 11.9 years, range from 29 to 80 years) were recruited from SAHZU. All breast cancer patients had a diagnosis report with a pathological stage of stage I or stage II at SAHZU between June 2020 and December 2020. The exclusion criteria were as follows: 1) Co-suffering other malignant tumors. 2) Having received any type of treatment for tumors. 3) Suffering from metabolic diseases, kidney diseases or liver diseases. 4) Taking any drugs for a long time. All participating volunteers agreed the informed consent in advance. Then, the serum samples were collected in the early morning and reserved at −80°C.

At first, 10 μl of isotope-labeled internal standards (IS) mixed solution was added into 100 μl serum samples which were naturally thawed in ice. And then 330 μl pre-refrigerated acetonitrile/methanol of 2:1 (v/v) was added. After vortexed for 60 s, let it stand at −20°C for 2 h and centrifuged at 13,000 rpm, 4°C for 15 min orderly, 352 μl of the supernatant were taken out and then drain under vacuum. Then, 80 μl acetonitrile/water of 9:1 (v/v) was used to redissolve the dried samples. After vortexed for 10 s, ultrasonicated for 15 s and centrifuged at 13,000 rpm for 15 min at 4°C, 70 μl of the supernatant fraction were sucked into the sample bottle for subsequent HILIC-MS/MS detection.

The mobile phase was (A) H₂O containing 10 mM ammonium formate, 0.2% formic acid and 0.06 mM malic acid, and (B) acetonitrile containing 2 mM ammonium formate, 0.2% formic acid and 0.06 mM malic acid. The eight analytes were perfectly separated at a flow rate of 0.4 ml/min by the optimized LC gradient program as follows: 0 min, 94% B; 4 min, 94% B; 6.1 min, 75% B; 6.5 min, 94% B; 8 min, 94% B. The BEH Amide column was set at 40°C and the samples temperature was maintained at 4°C. 5 μl of sample was injected each time and each sample was measured twice. To minimize the interference of the mass spectrometer, a switching valve was used and the eluents from the column were introduced into the ion source during 1.0–6.5 min.

The ion spray voltage was kept at 5.5 kV and the ion source temperature (TEM) was maintained at 550°C. Ion source gas 1 (GS1), ion source gas 2 (GS2) and curtain gas (CUR) were all set at 45 psi. The ion transitions of these eight ribonucleosides and corresponding isotope labeled internal standard (IS) were shown in Supplementary Table S1. The optimized MRM parameters of them including declustering potential (DP), entrance potential (EP), collision energy (CE) and collision cell exit potential (CXP) were also listed in Supplementary Table S1.

The standard working solutions of m⁶A, Gm, m¹G, Cm, Um and m⁵U at different concentrations (1, 2.5, 5, 10, 25, 50, 100, 250, 500 nM), which were mixed with IS solution (final concentration: [D₃]m⁶A (5 nM), [¹³C¹⁵N₂]Gm (10 nM), [¹³C¹⁵N₂]m¹G (20 nM), [¹³C₅]Cm (30 nM), [D₃]Um (20 nM) and [¹³C₅]m⁵U (100 nM)), were made and analyzed. The calibration curves could describe as y = ax + b, where y represents the peak area ratio of the analyte to the corresponding IS and x denotes the concentration of the analyte. The limit of detection (LOD) and limit of quantification (LOQ) of each ribonucleoside were obtained by analyzing standard solutions with a signal-to-noise ratio of three and ten, respectively.

For the purpose of evaluating intra-day and inter-day precision, the quality control (QC) samples at three different levels of m⁶A (2.5, 5, 50 nM), Gm (5, 10, 50 nM), m¹G (5, 20, 100 nM), Cm (5, 30, 150 nM), Um (5, 20, 100 nM) and m⁵U (20, 100, 300 nM) were prepared in triplicate and were measured on the same day and three consecutive days, respectively. The accuracy was described as the ratio of measured value to the theoretical concentration.

For the purpose of evaluating the recovery of extraction, the serum samples were added with three different levels of m⁶A (2.5, 5, 30 nM), Gm (2.5, 10, 50 nM), m¹G (5, 20, 60 nM), Cm (6, 30, 150 nM), Um (5, 20, 60 nM) and m⁵U (50, 150, 300 nM). After 10 μl of IS solution (same as described above) was added, the serum samples were treated and analyzed as mentioned above. The recovery (R) of each analyte was calculated by (concentration in added serum sample–concentration in original serum sample)/added concentration × 100%.

The matrix effect was estimated by using a slope comparison method. By adding different concentrations (1, 2.5, 5, 10, 25, 50, 100, 250, 500 nM) standard solution and IS to the serum extracts, the calibration curve was obtained. The ratio value of its slope to the slope of the standard solution calibration curve was matrix effect.

Statistical analyses were achieved through SPSS 24.0 software (IBM, Armonk, NY, United States). The concentration differences of serum monomethylated modifications between healthy volunteers and breast cancer patients were accessed by Two-tailed Student’s t-test, where p value less than 0.05 was considered meaningful. The area under the curve (AUC) was acquired by receiver operating characteristic (ROC) curve analysis, and the optimal cut-off value of the methylation modification in serum for the diagnosis of breast cancer was determined by the Youden index (Youden index = sensitivity + specificity −1). Receiver operating characteristic (ROC) analysis was performed to evaluate the ability of monomethylated modifications to distinguish cancer patients from healthy controls.

The optimization of chromatographic conditions is mainly achieved by optimizing the type of chromatographic column and the composition of the mobile phase. In order to acquire symmetrical peak shape and great separation effect in a short time, it is very important to choose a column with high separation efficiency. The chemical structures of these monomethylated ribonucleosides were illustrated in Figure 1. In our previous study, we found malic acid could enhance the detection of methylated nucleosides in HILIC-MS/MS (Guo et al., 2018a). Therefore, a hydrophilic interaction column of BEH Amide (100 mm × 2.1 mm, 1.7 μm, Waters) was selected for analysis, and malic acid was added into the mobile phase. As showed in Figure 2, it could acquire satisfactory separation for these eight ribonucleoside modifications. Besides, the analysis could be accomplished less than 6.5 min. It meant that this analytical method was quick, high throughput and fit for large clinical practice.

For the purpose of optimizing the MRM parameters, the mass spectrometer analyzed the standard solution injected by the peristaltic pump. Abundant [M + H]⁺ ions were observed in full scan ESI-MS. Then, the collision induced dissociation (CID) experiment was used to calculate the ion transitions of all analytes. In MS/MS, the ribose group can be easily eliminated due to the cleavage of the C-N bond. Taking Am and [¹³C₅]Am as examples, abundant [M + H]⁺ ions at m/z 282.1 and 287.1 were observed for Am and [¹³C₅]Am, respectively. The [M + H]⁺ ion of Am lost 146 and 151 Da was lost from [M + H]⁺ ion of [¹³C₅]Am. Therefore, ion transition m/z 282.1→136.0 and m/z 287.1→136.0 was used for the quantification of Am and [¹³C₅]Am, respectively. The ion transitions of other ribonucleosides and corresponding IS were shown in Supplementary Table S1. In addition, the optimized parameter values of DP, EP, CE and CXP were also listed. Under these optimized conditions, The LODs and LOQs of these monomethylated ribonucleosides can reach sub femtomole level (Supplementary Table S2).

According to the aforementioned method, the prepared calibration curve showed excellent linearities (R ² < 0.999) in appropriate analytical ranges, and equations were showed in Table 1. The slope ratio values for these eight monomethylated ribonucleosides ranged from 94.7 to 104.7% (Table 1), which indicated the interference of matrix in this study was minimal.

As showed in Supplementary Table S3, the intra- and inter-day accuracy assays were in the range of 92.20–112.96% and 92.29–112.76%, respectively, and the precision of intra- and inter-day were both within 8.6%. These data indicated that sufficient reproducibility and accuracy were obtained. As showed in Supplementary Table S4, the recoveries ranged from 98.04 to 114.01% (RSD <10%), indicating an excellent recovery rate.

In a word, all these results mentioned above revealed that the established HILIC-MS/MS method could meet quantitative requirement of m⁶A, Gm, m¹G, Cm, Um and m⁵U in human serum samples, and it was quick, accurate, sensitive, reproducible and reliable.

By using this HILIC-MS/MS method, we detected these modified ribonucleosides in serum samples from 61 patients with breast cancer and 69 healthy volunteers. The results showed that m⁶A, Gm, m¹G, Cm, Um and m⁵U were detected in all the serum samples, whereas Am and m⁵C could not be monitored due to their extremely low levels. As demonstrated in Figure 3, the retention time of Um, m⁶A, m⁵U, Cm, m¹G and Gm were 1.44, 2.10, 2.42, 3.38, 4.34 and 4.91 min, respectively. Of note, the retention time of these compounds were consistent with their corresponding isotope-labeled internal standard. The same tandem mass spectrometry behaviors and chromatographic retention time of each modified ribonucleoside in serum as those of the isotope-labeled IS confirmed the existence of m⁶A, Gm, m¹G, Cm, Um and m⁵U in human serum indubitably. It is worth nothing that the presence of Gm and Um in human serum was revealed for the first time, as far as we known.

The detailed concentrations of m⁶A, Gm, m¹G, Cm, Um and m⁵U in all the serum samples were presented in Supplementary Table S5. The measured concentrations (nM) in serum samples of m⁶A, Gm, m¹G, Cm, Um and m⁵U from healthy volunteers were in the range of 0.93–6.01, 7.91–20.85, 20.29–39.74, 29.97–66.58, 14.47–36.83 and 101.97–260.34 nM respectively, and the average concentrations were 2.97, 13.47, 27.97, 22.60 and 186.43 nM, respectively (n = 69). In the serum of breast cancer patients, the concentration of m⁶A, Gm, m¹G, Cm, Um and m⁵U were in the range of 1.45–7.01, 7.84–21.39, 16.59–34.41, 24.61–57.14, 13.83–39.07 and 108.20–226.84 nM respectively, and the average concentrations were 4.48, 11.55, 25.15, 37.31, 20.27 and 159.40 nM, respectively (n = 61). As illustrated in Figure 4, it was obvious that the levels of Gm, m¹G, Cm, Um and m⁵U in serum were intensely decreased in patients with breast cancer compared to healthy controls (p < 0.0001 for Gm, p < 0.0001 for m¹G, p < 0.0001 for Cm, p < 0.01 for Um, p < 0.0001 for m⁵U), but the concentration of m⁶A in breast cancer patients was much higher than that in healthy volunteers (p < 0.0001).

To verify the significance of m⁶A, Gm, m¹G, Cm, Um and m⁵U as potential breast cancer biomarkers, receiver operating characteristic (ROC) curve was plotted. Based on the Youden index, the optimal cut-off values of m⁶A, Gm, m¹G, Cm, Um and m⁵U were 3.76, 12.99, 25.92, 37.31,21.62 and 169.58 nM, which indicates that if the detection values break through these thresholds, it increases the risk of breast cancer. Trapezoidal rule was used to calculate area under the curve (AUC). As demonstrated in Figure 5, the AUC values were 0.78, 0.70, 0.70, 0.78, 0.65, 0.78 and 0.68 for m⁶A, Gm, m¹G, Cm, Um, m⁵U and CA153, respectively. Most of the AUC values of ribonucleosides were higher than that of CA153, implying there might be better correlation between the levels of m⁶A, Gm, m¹G, Cm and m⁵U in serum and the incidence of breast cancer, compared with CA153. Moreover, when all these serum monomethylated ribonucleosides were combined, the AUC could reach 0.93, which suggested the sensitivity and specificity for breast cancer diagnosis were dramatically increased. These results indicated that lower levels of Gm, m¹G, Cm, Um and m⁵U and higher levels of m⁶A could be regarded as potential diagnostic indicators for the screening of breast cancer.