Quantitative performance of online SPE-LC coupled to Q-Exactive for the analysis of sofosbuvir in human plasma

Abstract

A quantitative strategy for the analysis of sofosbuvir (SF) in human plasma has been developed with an online solid phase extraction (SPE)-liquid chromatography (LC) coupled with high resolution mass spectrometry (HRMS) system using a MS/MS targeted ion fragmentation scan (t-MS²), targeted-selected ion monitoring (t-SIM) and full MS-SIM (F-SIM) mode. The sample was pretreated automatedly and analyzed within 10 min via an online SPE system equipped with an Oasis®HLB (2.1 × 20 mm, 5 μm) SPE column and LC system with a ZORBAX SB-C18 (4.6 × 250, 5 μm) analytical column. A Q-Exactive hybrid quadrupole-Orbitrap HRMS was utilized for positive identification and quantification. Method detection limits ranged between 0.5 and 2000 ng mL⁻¹ in human plasma for both t-MS² and t-SIM mode and between 2 and 2000 ng mL⁻¹ for F-SIM mode. In all three modes, the overall intra-day and the inter-day variations were less than 8.07%. The recovery of all three methods was in the range of 92.06–107.20% with RSD% less than 3.90%. The matrix effect of all three methods was in the range of 94.82–101.89% with RSD% less than 7.69%. The optimized two methods (t-MS² and t-SIM) demonstrated good performance in terms of specificity, lower limit of quantification (LLOQ), linearity, recovery, precision and accuracy.

---

1. Introduction

Bioanalysis is one of the crucial processes in the area of pharmaceutical research, including research in pharmacokinetics (PK) and therapeutic drug monitoring (TDM). Bioanalysis of a drug in a matrix, such as plasma, serum, urine and tissue, has contributed substantially to establish dosage schemes, to reveal metabolic variability and to minimize adverse effects.¹ Liquid chromatography-tandem mass spectrometry (LC-MS/MS) has been widely used as the most reliable technology platform in the field of qualitative and quantitative bioanalysis.^2–7 MS/MS triple quadrupole spectrometry (QqQ) has revolutionized trace analysis in complex samples and replaced many traditional techniques based on fluorescence, ultraviolet, or single stage mass spectrometry detectors.⁸ Through the analysis of the precursor ions and product ions of target compounds, QqQ could achieve remarkable sensitivity and quantification capability. However, one of the major drawbacks of the conventional QqQ system is its relatively low resolution, which could not differentiate a target compound from an interfering compound with a decimal difference. The inherent limitation of LC-MS/MS is its total dependence on the separation of the target molecule from interfering compounds by LC.⁹ Metabolites with similar molecular masses to their parent analytes always perform as a common source of interference.¹⁰ For instance, a metabolite with a gain of 1 Da could be attributed to an enzymatic conversion of a primary amide on the parent compound to its carboxylic acid counterpart. The metabolite with this carboxylic acid group is isobaric with the [M + 1] isotope of the parent compound containing the amide group.¹¹ This phenomenon could interfere with the quantification of the metabolite. To overcome the aforementioned limitations, rapid development of a high resolution mass spectrometer (HRMS) has been offering new possibilities. Accurate mass measurement with high resolution could narrow down the number of possible molecular formulas that might be represented by a particular mass so as to avoid mass false positives.⁸ Selectivity could be significantly improved by the high resolving power (RP) of HRMS. Various analyzers with high resolution power, including time of flight (TOF) instruments and instruments with Orbitrap technology, have been developed and commercialized.^12,13 When HRMS was coupled with a quadrupole mass filter, high sensitivity and high selectivity could be achieved through combining the high resolving power of HRMS with the selectivity of the quadrupole. Q-Exactive is a hybrid benchtop Orbitrap mass spectrometer coupled with a quadrupole and a higher-energy collisional dissociation (HCD) cell. The ions trapped by the C-trap may be subjected to HCD, generating specific fragment ions.¹⁴ The device could achieve a mass accuracy lower than 2 ppm and resolving power up to 140 000 full width at half maximum (FWHM) at m/z 200.¹⁵ The device allows not only the high resolution full scan (FS) but also a MS/MS targeted ion fragmentation scan (t-MS²). An increasing number of applications of LC-HRMS utilizing an Orbitrap mass spectrometer for the analysis of complex samples has been recently published, including those of plasma,¹² kidney,¹³ manure,¹⁵ urine,¹⁶ milk,¹⁷ fruits and vegetables.¹⁸

The matrix of biological samples may contain various components such as salts, proteins and phospholipids. Biological matrices likely cause LC column efficiency degradation, ion source contamination, signal interference and matrix effects, resulting in poor bioanalysis of samples.¹⁹ Co-eluting matrix components could lead to enhancement or suppression of analyte ion intensity. These are referred to as matrix effects.²⁰ To prolong the service life of a LC column and MS instrument, as well as generate reliable and reproducible data, sample clean-up is crucial before bioanalysis. Liquid–liquid extraction (LLE), protein precipitation (PP) and offline-solid-phase extraction (SPE) are considered as conventional sample preparation techniques.¹ However, all these methods can be time-consuming and inconvenient due to heavy involvement of manual processes. In addition, the manual processes involved in these methods are prone to generate errors, which may lead to high variation of quantification. Recently, online SPE utilizing column-switching techniques coupled with LC-MS/MS have rapidly gained acceptance in bioanalytical applications to automate the sample clean-up process. The integrated process to separate a target compound from a matrix could not only reduce both time and manpower required for offline techniques but also increase the data quality of the bioanalysis.^21–23 To date, only an online SPE system coupled with Orbitrap has been used to analyse small molecules in aqueous matrices.^24–26 To the best of our knowledge, an analysis platform for the analysis of drug concentrations in plasma with integrated online SPE-HPLC and Q-Exactive has not yet been reported in literature.

In our research group, we have successfully developed various methods for bioanalysis of plasma samples using online-SPE coupled with an API 4000+ triple quadrupole mass spectrometer.^4,5 With an aim to explore the possibility of using HRMS as an alternative to MS/MS for bioanalysis of plasma samples, we attempted to establish such a platform cascaded online SPE system with LC-Q-Exactive. To verify the feasibility of our quantification strategy, sofosbuvir (SF) was chosen as the target drug for bioanalysis. Approved by the US FDA in 2013, SF, as the active principal ingredient of Sovaldi and Harvoni, has become a breakthrough in the curing of Hepatitis C virus (HCV) infections.^27,28 Previously, the available treatment for patients with HCV infection was a combination therapy using pegylated interferon (IFN) and ribavirin (RBV).²⁹ Unfortunately, the treatment had only moderate efficacy and is poorly tolerated by patients due to significant adverse effects.³⁰ SF was designed as a phosphoramidate prodrug of a cytidine nucleoside analog (Fig. 1) and provided a higher cure rate with fewer side effects. A reliable analysis method has been of importance for both drug development and clinical research. To date, there is only a single report on a method for the determination of SF in biological matrices with liquid–liquid extraction using UPLC-MS/MS.³¹ In the clinic research, the concentration of SF and its metabolite GS-331007 in the plasma were used as surveillance data.³² Minimal accumulation and linear pharmacokinetics of GS-331007 and SF were observed when SF was administrated in multiple doses.³³ Therefore, only SF was selected for our study. An UltiMate 3000*2 Dual-gradient HPLC system was introduced in the sample pre-treatment and separation of analyte via online-SPE-HPLC, and then Q-Exactive was used in the identification and quantification of the product ion with t-MS², t-SIM and F-SIM modes. The quantification performance of each of the three modes was studied. The validation of each of the methods was assessed by the determination of the method linearity, lowest limit of quantification (LLOQ), accuracy, extraction recovery and matrix effect.

Fig. 1 The structure of sofosbuvir.

2. Materials and methods

2.1. Chemicals and reagents

SF (purity > 99%) was synthesized and purified according to literature.²⁸ Chemical structure of SF is shown in Fig. 1. Carbamazepine (CBZ, purity 99.0%), purchased from Shanghai Oriental Pharmaceutical Science and Technology Co. Ltd (Shanghai, China), was used as the internal standard (IS). Acetonitrile and methanol were of HPLC grade and purchased from Fisher Scientific (Whitby, ON, Canada), and water used in the experiment was prepared by a Milli-Q Ultrapure purification water system (Bedford, MA, USA). Ammonium acetate and acetic acid were purchased from J&K (Beijing, China). Chromatographic mobile phases were prepared daily. Under the approval of the ethical committee of Tianjin AnDing Hospital, all blank plasma samples were kindly donated with informed consent of healthy volunteers from Tianjin AnDing Hospital, collected in K₂EDTA-treated tubes and stored at −80 °C.

The heated electrospray ionization (HESI) source of Q-Exactive was tuned and calibrated once a week. The component of the positive ion calibration solution was singly-charged ions. The ions are as follows: n-butylamine (m/z 74), caffeine (m/z 195 and its fragment m/z 138), Ultramark 1621 (m/z 1022, 1122, 1222, 1322, 1422, 1522, 1622, 1722, 1822), and MRFA (m/z 524).¹⁷

2.2. Stock and working solutions

The standard stock solutions of SF and CBZ were prepared by dissolving accurately weighed SF and CBZ in methanol at the concentration of 1 mg mL⁻¹. Then, the stock solution of SF was diluted with methanol : water (50 : 50, v/v) for concentration series of 40 000 ng mL⁻¹, 20 000 ng mL⁻¹, 10 000 ng mL⁻¹, 4000 ng mL⁻¹, 1000 ng mL⁻¹, 400 ng mL⁻¹, 100 ng mL⁻¹, 40 ng mL⁻¹ and 10 ng mL⁻¹. The stock solution of CBZ was diluted with methanol : water (50 : 50, v/v) to 500 ng mL⁻¹ as internal standard working solution. All the standard stock solutions were stored at −20 °C and the working solutions were stored at 4 °C.

Calibration standard solutions were prepared by spiking the diluted working standard solutions into blank human plasma (1 : 19, v/v) to give concentration series of 0.5 ng mL⁻¹, 2 ng mL⁻¹, 5 ng mL⁻¹, 20 ng mL⁻¹, 50 ng mL⁻¹, 200 ng mL⁻¹, 500 ng mL⁻¹, 1000 ng mL⁻¹ and 2000 ng mL⁻¹. 10 μL of IS working solution was added to 200 μL human plasma samples. Then, the samples were vortexed for 30 s and centrifuged at 13 000 rpm for 15 min at 4 °C. Quality control (QC) samples were prepared at 0.5 ng mL⁻¹ (low quality control, LQC), 200 ng mL⁻¹ (middle quality control, MQC), and 2000 ng mL⁻¹ (high quality control, HQC) for SF.

2.3. Online SPE and LC parameters

A Thermo Scientific Dionex Ultimate 3000 pump coupled with a Thermo Scientific Dionex Ultimate 3000 RS column compartment and a Thermo Scientific Ultimate 3000 autosampler controlled by Chromeleon 7.2 Software (Thermo Scientific Fisher, Waltham, MA and Dionex Softron GMbH Part of Thermo Fisher Scientific, Germany) were used for analysis.

Sample pre-treatment was performed with an online SPE methodology with a SPE cartridge Oasis®HLB (2.1 × 20 mm, 5 μm) from Waters. The system setup for online SPE consisted of three steps (Fig. 2). In the loading step, the plasma sample was injected directly into the SPE cartridge and the biological matrix was washed out in 2 min with acetonitrile–5 mM ammonium acetate buffer 2/98 (v/v, pH 3.5) at a flow rate of 1 mL min⁻¹. SF was retained on the SPE cartridge. In the transfer step, a SPE cartridge was coupled with the analytical column. SF was transferred from the SPE cartridge to the analytical column by isocratic elution using acetonitrile–5 mM ammonium acetate buffer 40/60 (v/v, pH 3.5) at a flow rate of 1 mL min⁻¹ for 1 min. In the separation step, the SPE cartridge and analytical column were switched by a valve to the parallel state once again. The analytes were separated by gradient elution (shown in Table 1) at a flow rate of 1 mL min⁻¹. This type of technological process allows shortening of the cycle time, which in our approach is only 10 min per sample.

Fig. 2 The schematic of online- SPE-HPLC-HRMS system.

Table 1 Online-SPE and HPLC conditions^a

Time (min)	SPE cartridge			Analytical column
	Flow rate (mL min^−1)	A (%)	B (%)	Flow rate (mL min^−1)	A (%)	B (%)
a A, acetonitrile; B, 5 mM ammonium acetate buffer (pH = 3.5, adjusted by acetic acid).
0	1	2	98	1	40	60
3	1	2	98	1	40	60
6	1	100	0	1	70	30
7	1	100	0	1	70	30
7.5	1	2	98	1	40	60
10	1	2	98	1	40	60

---

A C18 (Agilent ZORBAX SB-C18, 4.6 × 250 mm, 5 μm particles) column was used for chromatographic separation of SF and IS at 30 °C. Mobile phase A was acetonitrile, whereas mobile phase B was 5 mM ammonium acetate buffer (pH 3.5). A flow rate of 1 mL min⁻¹ and an injection volume of 10 μL were chosen. Through a tuneable flow divider valve, 4/5 of the liquid passed through the Dionex Ultimate 3000 diode array detector, whereas 1/5 entered the Q-Exactive.

2.4. Electrospray ionization source

A HESI-II in positive mode was used for the ionization of the SF. Source parameters were as follows: spray voltage +3.5 kV, S-lens voltage 50 V, aux gas heater temperature 300 °C and capillary temperature 320 °C. Ultrapure liquid nitrogen was used as ion source gas and collision gas; the sheath gas flow rate and aux gas flow rate were 35 and 10 (arbitrary units), respectively. The parameters of different scan modes are shown in Table 2.

Table 2 The parameters of the three modes

Mode	RP	AGC	Maximum IT (ms)	Scan range (m/z)
t-MS^2	17 500	2 × 10^5	100	243.00756
t-SIM	70 000	5 × 10^4	200	530.16982
F-SIM	70 000	3 × 10^6	200	100–600

---

2.5. Data analysis and method validation

Xcalibur 3.0 (Thermo Fisher Scientific, San Jose, CA, USA) was used for instrument control and data acquisition. Positive identification of analyte was based on the accurate mass of the analyte with less than ±5 ppm error and retention time comparison within ±0.03 min. Mass Frontier 7.0 (HighChem, Ltd, Bratislava, Slovakia) was used to obtain the fragmentation pattern and accurate mass of the product ions in positive mode.

In the t-MS² mode, the presence of a product ion (m/z 243.07756) was used for confirmation and quantification. In the t-SIM and F-SIM modes, the precursor ion (m/z 530.16982) was used for quantification. The mass tolerance window (MTW) was set to 5 ppm for the three analysis modes.

2.5.1 Linearity and LLOQ. Nine-point calibration curves were obtained by carrying out the online SPE method on SF human plasma samples spiked between 0.5 and 2000 ng mL⁻¹ and analyzing them in quintuplicate.

A least squares linear regression model y = ax ± b weighted by 1/x² was used to fit the calibration curves, in which y is peak area ratio of SF and IS, a is slope of the calibration curve, b is the y axis intercept and x is the analyte concentration. The linear range was set as 0.5–2000 ng mL⁻¹.

2.5.2 Accuracy and precision. Accuracy values were determined by the relative standard deviation (RSD%). Intra-day precision was evaluated by analysis of the same spiked sample at three different concentrations (0.5, 200 and 2000 ng mL⁻¹) five times on a single workday. Inter-day precision was calculated at three different concentrations (0.5, 200 and 2000 ng mL⁻¹) prepared daily for a period of 3 days.

2.5.3 Extraction recovery and matrix effect. The extraction recoveries of SF were assessed by comparing the mean peak area of the QC samples spiked before SPE with the mean peak area of corresponding concentration samples spiked after SPE. The matrix effect of SF were determined by the method of Matuszewski,²⁰ comparing the mean peak area of the three different concentration samples (0.5, 200 and 2000 ng mL⁻¹) against the corresponding extracted samples through HPLC analysis.

2.5.4 Stability study. The stability of SF in plasma was evaluated by analyzing the three samples of different concentrations (0.5, 200 and 2000 ng mL⁻¹) from long-term (frozen at −80 °C), short-term (at room temperature 24 h) storage and freeze–thaw cycles.

3. Results and discussion

3.1. Optimization of SPE and LC parameters

In an online SPE process, one of the most important parameters is the selection of the SPE cartridge. The goals of SPE are the separation of analyte from the biological matrix and removal of interfering substances. Three cartridges were evaluated for their efficiency in the online SPE procedure: Oasis®HLB cartridge (2.1 × 20 mm, 5 μm), CAPCELL MF Ph-1 cartridge (4.0 × 10 mm, 5 μm) and HyperSep Hypercarb cartridge (2.1 × 20 mm, 30 μm).

To compare the performances of the different SPE cartridges, 10 μL of the standard sample (2000 ng mL⁻¹) was loaded on a cartridge directly linked to a flow divider valve and rinsed with pure water at 1 mL min⁻¹ flow rate. As shown with the results in Fig. 3, Oasis and Hypercarb exhibited satisfactory retentive efficiency compared to CAPCELL MF Ph-1.

Fig. 3 The retention effects of the different SPE cartridges.

After the endogenous matrix was cleaned up, the SF was eluted from the SPE cartridge and analyzed via an analytical column with acetonitrile–water (40/60). According to Fig. 4, Hypercarb gave a tailing shape peak of SF, whereas HLB gave a relatively better peak. Eventually, the HLB cartridge was selected for its trapping efficiency and elution ability for further research.

Fig. 4 The eluting effect of SF in HLB and Hypercarb cartridges.

Different ratios of water-acetonitrile solution were used to optimize the elution efficiency of the endogenous matrix with 1 mL min⁻¹ flow rate. The monitored wavelength of diode array detection (DAD) was set at 210 nm, and the different proportions were evaluated as listed in Fig. 5. The loading time was refined to 2 min, within which the endogenous matrix was eluted at the extreme using acetonitrile–5 mM ammonium acetate buffer (pH 3.5) (2/98, v/v). To ensure that SF was sufficiently eluted from the SPE cartridge and entered the analysis column, the transfer time was optimized to 1 min and the transfer mobile phase used was 5 mM ammonium acetate buffer (pH 3.5)-acetonitrile (60/40, v/v).

Fig. 5 The eluting effect of endogenous matrix at different conditions. Black line: 98% water, red line: 99% water, green line: 100% water.

3.2. HRMS parameters optimization

Q-Exactive was operated in t-MS², t-SIM and F-SIM positive modes. In the t-MS² mode, Q-Exactive carried out the target MS/MS fragmentation scan using HCD. In the t-SIM and F-SIM modes, Q-Exactive performed the full MS and target MS scan, respectively, without HCD fragmentation. The data of the three modes were collected simultaneously when the sample was analyzed.

A satisfactory number of scan points per chromatographic peak was important for quantification, and it depended on the parameters of automatic gain control (AGC) target (the number of ions to fill C-trap), maximum injection time (IT) (the time of ions to reach C-trap) and RP. In the t-MS² mode, the AGC target was set to 2 × 10⁵, with a IT of 100 ms. In the t-SIM mode, AGC target was set to 5 × 10⁴, with a IT of 200 ms. In the F-SIM mode, AGC target was set to 3 × 10⁶, with a IT of 200 ms. The analyzer Orbitrap has RP of 17 500, 35 000, 70 000, and 140 000 FWHM. The selection of RP is crucial for the overall performance of the identification and quantification. In the t-MS² mode, ions of accurate parent mass were selected by the quadrupole before entering the HCD cell (Fig. 6); a 17 500 FWHM was sufficient to avoid false positives and provide an acceptable selectivity. This also contributed to gain more scan points. In both the t-SIM and F-SIM mode, the mass resolutions of 70 000 and 140 000, respectively, were sufficient for the required mass accuracy with MTW of 5 ppm. The resolving power is reversely correlated to the scanning speed, which has significant impact on the number of scan points. Compared with 140 000, the selection of a 70 000 FWHM would have a positive impact on the sensitivity by increasing the scanning speed.³⁴ Therefore, a 70 000 FWHM was selected for quantification study of both the t-SIM and F-SIM mode.

Fig. 6 The parameters of the t-MS² in inclusion list.

Another factor that influenced the sensitivity of the t-MS² mode was the normalized collision energy (NCE) value of the HCD cell. The area of parent ions was decreased, whereas the area of product ions was increased by increasing the NCE (Fig. 7). The optimal NCE for SF was 16%. In this condition, both the parent ions and product ions could coexist; moreover, the response of product ions was relative high. The structures of the main product ions (red lines) were determined and recognized by Mass Frontier 7.0 (Fig. 8).

Fig. 7 SF fragmentation with increasing HCD percentage.

Fig. 8 The retrieval result of MS² of SF by Mass Frontier.

3.3. Method validation

Under the optimum conditions for online SPE-LC-HRMS, the proposed quantitative performances of three modes were evaluated regarding selectivity, linearity, accuracy, precision, extraction recovery and matrix effect.

3.3.1 Selectivity. The chromatograms of blank plasma and spiked plasma with SF and CBZ are shown in Fig. 9. Little interference from endogenous materials was found at the retention time of the analyte, indicating that the online-SPE procedure was selective.

Fig. 9 The chromatograms of blank plasma (A) and plasma spiked with SF or CBZ (B).

3.3.2 Linearity and LLOQ. The LLOQ was determined by the criteria that the analyte response is at least 10 times the baseline noise (S/N ≥ 10), and it was defined with acceptable accuracy (within ±20%) and precision (within ±20%). In the concentrations of 0.5–2000 ng mL⁻¹, a linear relationship between the chromatographic peak area ratio and the different concentrations of SF was investigated by Xcalibur 3.0. A concentration of 2000 ng mL⁻¹ should be able to satisfy the bioanalysis demand of C_max (1823.7 ± 639.1 ng mL⁻¹) in human plasma.³² At the concentration of 0.5 ng mL⁻¹, the signal-to-noise ratio (S/N) of each of the three modes is shown in Fig. 10. The ratios of the t-MS² mode (17) and t-SIM mode (21) were higher than that of the F-SIM (Not able to calculate). At the concentration of 0.5 ng mL⁻¹, the result indicated that the noise was so serious that S/N could not be recognized by Xcalibur. Due to the selection of ions by the quadrupole, the t-MS² mode and t-SIM modes improved the sensitivity and reduced the quantification limit. To meet the requirement of bioanalysis, the linear range of SF was eventually achieved between 2 ng mL⁻¹ and 2000 ng mL⁻¹ when the F-SIM mode was used for quantification. As a result, F-SIM was only used for quantification performance comparison purposes and not used for method development. In the following experiment, only the quantification ability at 200 ng mL⁻¹ (MQC) and 2000 ng mL⁻¹ (HQC) for the F-SIM mode were compared with those of the other two modes. Data for the linear regression equation of the analytes is listed in Table 3. There was good linearity in all three quantification methods (R² > 0.99).

Fig. 10 The S/N of three modes at the concentration of 0.5 ng mL⁻¹.

Table 3 Data for linear regression equation of SF

Mode	Linear range (ng mL^−1)	LLOQ (ng mL^−1)	Slope	Intercept	R^2
t-MS^2	0.5–2000	0.5	0.418886	−0.0227852	0.9988
t-SIM	0.5–2000	0.5	0.0067732	0.05638435	0.9971
F-SIM	2–2000	2	0.00742484	−0.00246331	0.9966

---

3.3.3 Accuracy and precision. The accuracy of SF quantification was expressed as the ratio of the nominal concentration to the mean value of the detected concentration. The precision was calculated by RSD%. The intraday and interday accuracy and precision for the QC samples are shown in Table 4.

Table 4 Intraday and interday precision and accuracy

Mode	Nominal concentration (ng mL^−1)	Intraday				Interday
		N	Mean concentration found (ng mL^−1)	Accuracy (%)	RSD (%)	N	Mean concentration found (ng mL^−1)	Accuracy (%)	RSD (%)
t-MS^2	0.5	5	0.535	107.00	6.93	15	0.530	105.92	7.20
	200	5	194.7368	97.37	1.38	15	194.7964	97.40	5.24
	2000	5	1965.951	98.30	4.72	15	2065.84	103.29	3.92
t-SIM	0.5	5	0.524	104.72	4.26	15	0.526	105.12	8.07
	200	5	198.68	99.33	1.20	15	196.75	98.27	4.01
	2000	5	2225.92	111.20	1.20	15	2118.58	105.92	4.47
F-SIM	200	5	193.62	96.81	3.1	15	186.58	93.29	5.42
	200	5	2190.9	109.54	1.24	15	2135.29	106.76	3.32

---

For the t-MS² mode, intraday precision was calculated for 3 concentration levels (n = 5) and ranged from 97.37% to 107.00%, whereas the RSD% of QC samples was in the range of 1.38–6.93%. Interday precision calculated for 3 levels (n = 5) ranged from 97.40% to 105.92%, whereas the RSD% of QC samples was in the range of 3.92–7.20%.

For the t-SIM mode, intraday precision was calculated for 3 concentration levels (n = 5) and ranged from 99.33% to 104.72%, whereas the RSD% of QC samples was in the range of 1.20–4.26%. Interday precision calculated for 3 concentrations levels (n = 5) ranged from 98.27% to 105.92%, whereas the RSD% of QC samples was in the range of 4.01–8.07%.

For the F-SIM mode, intraday precision calculated for MQC and HQC levels (n = 5) ranged from 96.81% to 109.54%, whereas the RSD% was in the range of 1.24–3.1%. Interday precision calculated for MQC and HQC levels (n = 5) ranged from 93.29% to 106.76%, whereas the RSD% of QC samples was in the range of 3.32–5.42%.

Overall, both the t-MS² and t-SIM mode showed similar performances in terms of accuracy and precision at all three concentrations, whereas F-SIM showed comparable accuracy and precision at both MQC and HQC. Both the t-MS² and t-SIM mode results were within all the defined acceptance criteria of the U. S. Food and Drug Administration (FDA).³⁵

3.3.4 Extraction recovery and matrix effect. As shown in Table 5, the results of analyte extracted using online SPE gave high extraction recoveries ranging from 98.86% to 105.56%, from 97.42% to 107.20% and from 92.06% to 99.12% for the t-MS², t-SIM and F-SIM mode, respectively. An matrix effect is signal suppression or enhancement effect caused by interference and is considered tolerable if the value is in the range of 80–120%.²⁰ The methods gave acceptable values of the matrix effect, ranging from 96.5% to 98.82%, from 97.48% to 101.10% and 94.82% to 101.89%, for the t-MS², t-SIM and F-SIM mode, respectively. The results met the criteria to be acceptable for assay of the analytes in human plasma.³⁵

Table 5 Matrix effect and extraction recovery

Mode	Nominal concentration (ng mL^−1)	Extraction recovery (%)		Matrix effect (%)
		Mean	RSD (%)	Mean	RSD (%)
t-MS^2	0.5	103.02	3.49	96.5	4.34
	200	98.86	3.90	98.18	2.97
	2000	105.56	2.42	98.82	2.81
t-SIM	0.5	107.20	3.21	97.48	7.69
	200	103.61	3.49	99.78	1.24
	2000	97.42	1.85	101.10	3.67
F-SIM	200	99.12	2.23	101.89	2.64
	2000	92.06	2.26	94.82	1.72

---

3.3.5 Stability study. SF in human plasma was stable at room temperature for 6 h and at −80 °C for 30 days. Freeze–thaw stability study at −80 °C for 3 cycles was carried out. Data from the stability experiments are listed in Table 6. The results showed that the analyte was stable under the above conditions.

Table 6 Stability study for SF under different conditions

Mode	Stability	Storage conditions		Level	Mean comparison samples (ng mL^−1)	Mean stability samples (ng mL^−1)	Accuracy (%)	RSD (%)
t-MS^2	Short-term		Room temperature for 6 h	LQC	0.535	0.505	100.92	5.53
				MQC	194.74	208.85	104.43	4.08
				HQC	2025.95	2052.54	102.63	4.22
	Freeze–thaw		At −80 °C for 3 cycles	LQC	0.518	0.478	95.57	5.15
				MQC	200.22	203.66	101.83	3.16
				HQC	2163.77	2007.37	100.37	3.79
	Long-term		At −80 °C for 30 days	LQC	0.536	0.520	104.08	3.17
				MQC	185.67	187.36	93.68	3.00
				HQC	2067.79	1925.48	96.27	4.77
t-SIM	Short-term		Room temperature for 6 h	LQC	0.524	0.552	110.37	2.16
				MQC	198.68	205.92	102.95	1.93
				HQC	2225.92	2211.06	110.55	2.83
	Freeze–thaw		At −80 °C for 3 cycles	MQC	0.515	0.477	95.47	3.45
				MQC	204.24	202.21	101.11	1.27
				HQC	2027.66	2112.31	105.62	2.03
	Long-term		At −80 °C for 30 days	LQC	0.529	0.537	107.32	4.07
				MQC	187.35	203.72	101.86	2.73
				HQC	2002.13	1916.44	95.82	1.72
F-SIM	Short-term		Room temperature for 6 h	MQC	193.62	195.50	97.75	1.63
				HQC	2190.91	2178.60	108.93	1.63
	Freeze–thaw		At −80 °C for 3 cycles	MQC	184.40	187.93	93.97	1.78
				HQC	2152.96	2157.92	107.90	2.82
	Long-term		At −80 °C for 30 days	MQC	191.72	197.52	98.76	3.32
				HQC	2062.00	1972.66	98.63	2.15

---

According to the guidelines of the US FDA summarized in Table 7,³⁵ the bioanalytical method for Sofosbuvir is valid.

Table 7 The guidelines of US FDA in Bioanalytical Method Validation

	LLOQ	Other concentrations
Accuracy	80–120%	85–115%
Precision	±20%	±15%
Sensitivity	S/N > 5	—
Calibration curve	At least six non-zero samples
Selectivity	No interference in blank samples
Matrix effect	85–115%
Recovery	Consistent, precise and reproducible
Reproducibility	Intraday and interday
Stability	±15% (room temperature, long-term and freeze–thaw)

---

3.4. Analysis of online SPE-LC-HRMS

An online SPE-LC coupled with Q-Exactive has been established for bioanalysis of SF in human plasma. The integration of SPE with the bioanalysis could fully automate the process of sample pretreatment and matrix removal. The benefits of online-SPE are not only time saving and easiness of application, but also the overall quality of the bioanalysis such as accuracy, precision, recovery and matrix effect. Q-Exactive is a hybrid bench-top instrument that combines a quadrupole with Orbitrap technology. The high resolving power of an accurate mass provides high security of identification compared to conventional mass spectrometer. Three modes of quantification (t-MS², t-SIM and F-SIM) were used for bioanalysis of SF in human plasma. In general, all three methods gave a better LLOQ than that published using the LC-MS/MS method³¹ (0.5 ng mL⁻¹, 0.5 ng mL⁻¹, and 2 ng mL⁻¹ vs. 10 ng mL⁻¹). Among the three quantification modes, the t-MS² mode and t-SIM mode improved the sensitivity and reduced the quantification limit through the selection of ions by a quadrupole. Indeed, the t-MS² and t-SIM modes are especially suitable for known compounds. The F-SIM mode has the ability to obtain a wealth of information about ionic compound residues or metabolites in a single analysis, which could be particularly useful in the case of a bioanalysis with limited information. The quantification performance of online-SPE-LC-Q-Exactive could meet the requirements of bioanalysis of SF in human plasma well.

4. Conclusion

In this study, HRMS was explored as an identification and quantification approach for bioanalysis of drugs in human plasma. Online-SPE-LC was coupled with Q-Exactive for bioanalysis of SF to achieve automation of sample pretreatment and high efficiency of identification and quantification. The quantification performance of the t-MS,² t-SIM and F-SIM modes has been studied. Bioanalysis methods for SF in human plasma that meet the defined acceptance criteria were achieved with both t-MS² and t-SIM quantification in a linear range of 0.5–2000 ng mL⁻¹. This suggests that HRMS could provide an alternative to MS/MS for bioanalysis in the identification and quantification of drugs in biological matrices. Compared with the published method, our online-SPE-LC-HRMS is easy to use (no manual sample pretreatment process) and highly sensitive (LLOQ 0.5 ng mL⁻¹). The established online SPE-LC-HRMS could provide a sensitive, accurate and reliable method for the bioanalysis of SF in human plasma.

Acknowledgements

This study was supported by the National Basic Research Program of China (973 program, Grant Nos 2013CB911104, 2013CB911100), the National Natural Science Foundation of China (Grant No. 21572116), the Fundamental Research Funds for the Central Universities (Grant No. 20141081129) and the Tianjin Science and Technology Program (Grant Nos 13JCYBJC24300, 13JCQNJC13100).

References

1. L. Nováková and H. Vlčková, Anal. Chim. Acta, 2009, 656, 8–35.
2. X. Ding, H. Ghobarah, X. L. Zhang, A. Jaochico, X. R. Liu, G. Deshmukh, B. M. Liederer, C. E. C. A. Hop and B. Dean, Rapid Commun. Mass Spectrom., 2013, 27, 401–408.
3. H. Zhao, X. D. She, X. X. Yu, J. F. Fu, Y. Chen, Y. Tian and Z. J. Zhang, Bioanalysis, 2015, 7, 895–905.
4. L. Liu, Y. B. Wen, K. N. Liu, L. Sun, Y. X. Lu and Z. Yin, RSC Adv., 2014, 4, 19629–19639.
5. Y. R. Fan, G. H. Shen, P. Li, X. N. Xi, H. T. Wu, H. J. Tian, Y. X. Lu and Z. Yin, RSC Adv., 2015, 5, 34342–34352.
6. A. F. Li, H. Fan, F. F. Ma, P. McCarron, K. Thomas, X. H. Tang and M. A. Quilliam, Analyst, 2012, 137, 1210–1219.
7. L. Bailly-Chouriberry, F. Cormant, P. Garcia, A. Kind, M. A. Popot and Y. Bonnaire, Anal. Chem., 2013, 85, 5219–5225.
8. A. Kaufmann, P. Butcher, K. Maden, S. Walker and M. Widmer, Anal. Chim. Acta, 2010, 673, 60–72.
9. S. J. Bruce, B. Rochat, A. Béguin, B. Pesse, I. Guessous, O. Boulat and H. Henry, Rapid Commun. Mass Spectrom., 2013, 27, 200–206.
10. M. Furlong, A. Bessire, W. Song, C. Huntington and E. Groeber, Rapid Commun. Mass Spectrom., 2010, 24, 1902–1910.
11. R. Bakhtiar and T. K. Majumdar, J. Pharmacol. Toxicol. Methods, 2007, 55, 227–243.
12. M. A. Marzinke, A. Breaud, T. L. Parsons, M. S. Cohen, E. Piwowar-Manning, S. H. Eshleman and W. Clarke, Clin. Chim. Acta, 2014, 433, 157–168.
13. D. G. Rocha, F. A. Santos, J. C. C. da Silva, R. Augusti and A. F. Faria, J. Chromatogr. A., 2015, 1379, 83–91.
14. H. Henry, H. R. Sobhi, O. Scheibner, M. Bromirski, S. Nimkar and B. Rochat, Rapid Commun. Mass Spectrom., 2012, 26, 499–509.
15. M. Solliec, A. Roy-Lachapelle and S. Sauvé, Anal. Chim. Acta, 2015, 853, 415–424.
16. T. Gicquel, S. Lepage, M. Fradin, O. Tribut, B. Duretz and I. Morel, J. Anal. Toxicol., 2014, 38, 335–340.
17. A. Kaufmann, P. Butcher, K. Maden, S. Walker and M. Widmer, Anal. Chim. Acta, 2014, 820, 56–68.
18. J. Wang, W. Chow, D. Leung and J. Chang, J. Agric. Food Chem., 2012, 60, 12088–12104.
19. P. Li and M. G. Bartlett, Anal. Methods, 2014, 6, 6183–6207.
20. B. K. Matuszewski, M. L. Constanzer and C. M. Chavez-Eng, Anal. Chem., 2003, 75, 3019–3030.
21. N. Negreira, M. López de Alda and D. Barceló, J. Chromatogr. A, 2013, 1280, 64–74.
22. L. Liu, K. N. Liu, Y. B. Wen, H. W. Zhang, Y. X. Lu and Z. Yin, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2012, 893–894, 21–28.
23. L. Liu, Y. B Wen, K. N. Liu, L. Sun, M. Wu, G. F. Han, Y. X. Lu, Q. M. Wang and Z. Yin, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2013, 923–924, 8–15.
24. S. R. Batchu, C. E. Ramirez and P. R. Gardinali, Anal. Bioanal. Chem., 2015, 407, 3717–3725.
25. F. Wode, C. Reilich, P. van Baar, U. D Dünnbier, M. Jekel and T. Reemtsma, J. Chromatogr. A., 2012, 1270, 118–126.
26. N. V. Heuett, C. E. Ramirez, A. Fernandez and P. R. Gardinali, Sci. Total Environ., 2015, 511, 319–330.
27. F. D. A. government, http://www.accessdata.fda.gov/drugsatfda_docs/label/2013/204671s000lbl.pdf, 2013, December.
28. M. J. Sofia, D. H. Bao, W. Chang, J. F. Du, D. Nagarathnam, S. Rachakonda, P. G. Reddy, B. S. Ross, P. Y. Wang, H. R. Zhang, S. Bansal, C. Espiritu, M. Keilman, A. M. Lam, H. M. Micolochick Steuer, C. R. Niu, M. J. Otto and P. A. Furman, J. Med. Chem., 2010, 53, 7202–7218.
29. M. G. Ghany, D. R. Nelson, D. B. Strader, D. L. Thomas and L. B. Seeff, Hepatology, 2011, 54, 1433–1444.
30. X. Tong, S. Le Pogam, L. Li, K. Haines, K. Piso, V. Baronas, J. M. Yan, S. S. So, K. Klumpp and I. Nájera, J. Infect. Dis., 2014, 209, 668–675.
31. M. R. Rezk, E. B. Basalious and I. A. Karim, J. Pharm. Biomed. Anal., 2015, 114, 97–104.
32. M. Rodriguez-Torres, E. Lawitz, K. V. Kowdley, D. R. Nelson, E. DeJesus, J. G. McHutchison, M. T. Cornpropst, M. Mader, E. Albanis, D. Y. Jiang, C. M. Hebner, W. T. Symonds, M. M. Berrey and J. Lalezari, J. Hepatol., 2013, 58, 663–668.
33. B. J. Kirby, W. T. Symonds, B. P. Kearney and A. A. Mathias, Clin. Pharmacokinet., 2015, 54, 677–690.
34. F. Shi, C. C. Guo, L. P. Gong, J. Li, P. Dong, J. L. Zhang, P. Cui, S. Y. Jiang, Y. X. Zhao and S. Zeng, J. Chromatogr. A., 2014, 1344, 91–98.
35. U.S. Department of Health and Human Services, F. D. A., C. D. E. R., Guidance for Industry, Bioanalytical Method Validation, 2013.

---