The purity of linezolid and tedizolid (used as internal standard, IS, Figure 1C) were >98%, and were offered from Beijing sunflower and technology development CO., LTD. (Beijing, China). PNU-142300 (purity >98%) was synthesized from WuXi AppTec (Jiangsu, China). Methanol and acetonitrile in this study were LC grade, and both of them were supplied by Merck Company (Darmstadt, Germany). Beijing sunflower and technology development CO., LTD. (Beijing, China) also provided formic acid, which was analytical grade. In addition, ultrapure water was made by a Water Purification System from Milli-Q (Millipore, Bedford, United States) to meet the need of the whole experiment.

Liquid chromatographic analysis was carried out by a Waters ACQUITY UPLC I-Class system (Milford, MA, United States) fitted with a Waters Xevo TQ-S triple quadrupole tandem mass spectrometer equipped with an electrospray ionization (ESI) source (Milford, MA, United States). Masslynx 4.1 software and Quanlynx programme (Milford, MA, United States) in this experiment were used to acquire and process the data of this whole experiment and control the instruments.

The UPLC experiments on the chromatographic separation of linezolid, PNU-142300 and IS was performed on an Acquity UPLC BEH C18 column (2.1 mm × 50 mm, 1.7 μm). In addition, the mobile phase in this study, consisting of solvent A (acetonitrile) and solvent B (0.1% formic acid in water), was employed to separate the analyte and IS using a linear gradient program with a flow rate of 0.40 ml/min as follows: 10–10% A (0–0.5 min), 10–90% A (0.5–1.0 min), 90–90% A (1.0–2.0 min), and 90–10% A (2.0–2.1 min). Then, the post time was 0.9 min. The injection volume of each needle was 0.2 µl, and the entire time for each analysis was about 3.0 min. The autosampler was set at 10°C for each sample analysis, and the column temperature was maintained at 40°C temperature.

Detection was finished for the measurement of experimental samples in the positive ion mode through an ACQUITY UPLC system equipped with a Xevo TQ-S triple quadrupole tandem mass spectrometer. Multiple reaction monitoring (MRM) method was selected to detect the analytes and IS in serum, in which the precursor-to-product ion transitions for linezolid, PNU-142300 and IS were m/z 338.01 → 296.03, m/z 369.96 → 327.98, and m/z 370.98 → 342.99, respectively. The collision energy and cone voltage were respectively 15 eV and 30 V for linezolid, PNU-142300 and IS. The general MS parameters were optimized as follows: desolvation temperature 600°C, capillary voltage 2.0 kV, collision gas 0.15 ml/min, cone gas 200 L/h, and desolvation gas 1000 L/h.

Each stock solution of linezolid, PNU-142300 and IS at the corresponding concentration level of 1.00 mg/ml was individually obtained by dissolving each compound into an accurate amount of methanol for quality control (QC) and calibration curve samples. A mixed-standard working solutions were prepared by series dilution of the stock solution of each analyte with methanol. Then, calibration curves spiked with blank human serum were operated by ten-fold dilution of the mixed-standard working solutions in order to get final concentration levels in the calibration curve range of 0.01–20 μg/ml for linezolid, and 0.05–100 μg/ml for PNU-142300, respectively. Similary, QC samples at LLOQ (lower limit of quantification), low (LQC), medium (MQC) and high (HQC) concentration levels were also prepared as follows: 0.01, 0.02, 2.0, 16 μg/ml for linezolid, and 0.05, 0.1, 10, 80 μg/ml for PNU-142300, respectively. The IS working solution was obtained by diluting the stock solution with methanol to produce the required concentration of 5.0 μg/ml. All prepared solutions in this experiment including the stock and working solutions were ready in advance and placed at −40°C for further use.

All the samples in our study were processed by using a simple and rapid protein precipitation approach with acetonitrile. To each serum sample with a volume of 100 µl in 1.5 ml EP centrifuge tube, 20 µl IS working solution at the concentration level of 5.0 μg/ml was spiked and mixed for 0.5 min. Then, plasma proteins were immediately precipitated by adding 300 µl of acetonitrile, and the mixture was vigorously vortexed for 1.0 min, and centrifugated at 13,000 g for 10 min at room temperature. Finally, 100 µl aliquot of the clear supernatant was used to transfer into the new auto-sampler vial, and 0.2 µl of the solution was added into the UPLC-MS/MS system for quantitative determination and data acquisition.

Fully method validation procedures for this analytical method, including selectivity, the lineary of calibration curve, LLOQ, precision and accuracy, recovery, matrix effect, and stability, were conducted according to the FDA principles on the validation of the bioanalytical assay (Xu et al., 2019; Tang et al., 2020).

In this experiment, the selectivity of this method was investigated by checking the absence of endogenous interferences from the blank (neither analyte nor IS from six different human samples), standard solutions and real human serum at the corresponding retention time of the respective analyte of linezolid, PNU-142300, and IS, as previously described (Qiu et al., 2019).

Calibration curves of the analytes were evaluated by describing the relationship between the peak area ratio of each analyte to IS and the nominal concentrations of each analyte by least square regression mode with a 1/x ² weighting factor. The sensitivity of this assay was performed in terms of LLOQ, which was identified as the lowest point on the calibration curve, and should be investigated with precision below 20% and accuracy within ±20% of the nominal value.

Precision and accuracy of the analytes in the experiment were estimated by sextuple detection of three concentration levels of QC samples over three consecutive days. Relative standard deviation (RSD, %) stands for precision, and relative error (RE, %) stands for accuracy. Recovery from present approach of sample preparation was investigated by comparing the peak area ratio of each analyte before and after the extraction, respectively. Matrix effect (ME) was also analyzed in 6 replicates through the comparison of the response of each substance in serum matrix after extraction with that in neat solution.

The stability of each analyte in spiked serum was examined by detecting LQC and HQC samples under different storage conditions. Short-term stability was estimated at room temperature for 4 h, and long-term stability was evaluated at −40°C for 3 weeks. In addition, post-preparation stability was detected for QC samples placed in the autosampler at 10°C for 8 h. Moreover, three complete freez/thaw stability (−40°C to room temperature) was also studied.

The established UPLC-MS/MS bioanalytical assay was employed to determine simultaneously the serum concentration levels of linezolid and PNU-142300 in 115 samples from 89 patients who were treated with oral or intravenous linezolid at The First Affiliated Hospital of Wenzhou Medical University. The patients were segregated into five different groups according to the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation based on estimated glomerular filtration rate (eGFR, mL/min/1.73 m²): patients with eGFR≥90 (Group A, healthy), 60 ≤ eGFR < 90 (Group B, mild impaired), 30 ≤ eGFR < 60 (Group C, moderate impaired), eGFR < 30 (Group D, severe impaired), and continuous renal replacement therapy (CRRT, Group E, end-stage renal disease, ESRD). All patients were sampled at least after the 7th dose under steady state conditions. The serum samples of the analytes were collected, followed by centrifugation at 4,000 × g for 8 min at room temperature, and then immediately placed at −80°C until further analysis. The protocol of our study was assessed and approved by the Ethics Committee of The First Affiliated Hospital of Wenzhou Medical University (No. 2020-zz-019).

In our study, Statistical Package for the Social Sciences version 19.0 (SPSS, Chicago, IL, United States) software was used to calculate the results of all statistical analysis, and P < 0.05 was considered as meaningful in statistical analysis. Meanwhile, receiver operating characteristic (ROC) curve and its associated areas under the curve were used to assess the predictive power of the metabolite-to-parent concentration ratios for linezolid-associated myelosuppression; meanwhile, the cut-off point with the highest discriminative ability was chosen based on the maximum Youden’s index.

During method development, chromatographic conditions under different environments were investigated and assessed in order to get good results of the separation, peak shape, and mass response. In our study, five kinds of the solutions (including methanol, acetonitrile, water, ammonium acetate and formic acid), were evaluated, and it was found that the sensitivity was enhanced and the peak shape was improved when formic acid was added to aqueous phase. Thus, in our present experiment, acetonitrile as organic phase and 0.1% formic acid in water as aqueous phase were selected to offer higher intensity, lower background noise, and better peak shapes.

Given the complex nature of serum, a procedure for removing potential interferents and proteins in sample pretreatment is often required before LC-MS/MS analysis. Protein precipitation (PPT), liquid-liquid extraction (LLE) and solid-phase extraction (SPE) are the most common methods applied to the preparation of the biological samples (Gregoire et al., 2016; Hedaya et al., 2017; Paal et al., 2018; Ferrari et al., 2019). Compared with PPT for sample pretreatment, SPE is relatively time consuming and expensive, and LLE is relatively complicated. Thus, PPT method was adopted and investigated at the beginning of the method development. It was found that extraction recovery of each analyte was within the acceptable range, and no matrix effect was observed. Thus, in this present experiment, PPT method was adopted for it offered many advantages, such as low cost, quick and simple operation, and enough extraction recovery rates in the sample pretreatment.