A High-Performance Liquid Chromatography—Mass Spectrometry Method for Simultaneous Determination of Vancomycin, Meropenem, and Valproate in Patients with Post-Craniotomy Infection

Vancomycin (VAN), meropenem (MER), and valproate (VPA) are commonly used to treat intracranial infection post-craniotomy and prevent associated epilepsy. To monitor their levels, we developed a novel bioassay based on liquid chromatography–tandem mass spectrometry (LC–MS/MS) for simultaneous determination of these three drugs in human serum and cerebrospinal fluid (CSF). Sample preparation by protein precipitation using acetonitrile was followed by HPLC on a Zorbax 300SB-C8 column (150 mm × 4.6 mm, 5 μm) maintained at 40 °C. The lower limit of quantification (LLOQ) was 5 ng/mL for MER, 0.1 μg/mL for VAN, and 1 μg/mL for VPA in serum and 50 ng/mL for MER, 1 μg/mL for VAN, and 2 μg/mL for VPA in CSF. This method was validated with satisfactory linearity, sensitivity, precision, accuracy, recovery, matrix effects, and stability for all analytes. The assay was then successfully applied to evaluate VPA, MER, and VAN levels in serum and CSF from patients with intracranial infection administrated by intrathecal injection. Compared with intravenous injections, an intrathecal injection can provide sufficient therapeutic effects even if the CSF levels did not reach the effective concentration reported. Our method provided a detection tool to study the effective concentrations of these three drugs in CSF from patients administered via intrathecal injection.


Introduction
Central nervous system infection is a common and serious complication after neurosurgery that affects prognosis and outcomes and may become life-threatening [1,2]. VAN and MER are the first-line drugs for the treatment of medical-related central nervous system infections [3,4]. Brain diseases can cause nerve cell damage, abnormal discharge, and increase the risk of secondary seizures [5]. As a first-line treatment for various types of epilepsy, VPA is also widely used to prevent seizures in neurosurgery patients [6]. The combination of these three drugs appears to be a reasonable treatment option. However, these three drugs are associated with high toxicity and high risks of adverse reactions [7][8][9]. Therefore, therapeutic drug monitoring is mandatory for the combination of VAN, MER, and VPA.
Antibiotic drugs rarely pass through the blood-brain barrier (BBB), which makes antiinfective therapy after neurosurgery challenging [10]. In addition, craniocerebralinjury may lead to the opening of the BBB and the degree of leaking varies with different pathological conditions, leading to varied drug absorption from blood to CSF. In recent years, the administration of broad-spectrum antibiotics by intrathecal injection has been used to treat an intracranial infection. Compared with intravenous injection, an intrathecal injection can greatly increase drug concentration in CSF and offers superior safety and therapeutic effects [11]. Thus, concentrations of these three drugs in serum and CSF are potential quantitative traits that can describe and define each patient's therapeutic characteristics. However, the analytical methods available to determine the levels of these three drugs with high selectivity and sensitivity are currently limited.
The multiple reaction monitoring (MRM) method using LC-MS/MS is a common analytical method that can detect various compounds with high sensitivity and specificity. Detection of VAN and MER in human plasma with an LLOQ of 1 µg/mL, extracted by protein precipitation, was reported by Barco et al. [12,13]. Lipska et al. achieved a much lower LLOQ of 0.075 µg/mL for VPA in plasma using high-performance liquid chromatography with ultra-violet detection (HPLC-UV) and gas chromatography-mass spectrometry (GC-MS) methods [14]. Ye et al. developed a high-throughput LC-MS/MS approach to measuring VAN in human CSF with an LLOQ of 0.1 µg/mL [15]. Lu et al. developed an LC-MS/MS approach to detect 18 antibacterial drugs in human plasma with an LLOQ of 2.02 µg/mL for MER and 0.41 µg/mL for VAN [16]. However, there are still no analytical methods available to determine MER in CSF, and the reported methods for MER in plasma or serum always possess higher LLOQ. Although these three drugs are used together in clinical practice, there is no analytical method available for detecting them in plasma, serum, or CSF simultaneously.
In this work, we present a novel LC-MS/MS method for the simultaneous detection of VAN, MER, and VPA in human serum and CSF. This assay was then applied to patients with postoperative intracranial infection treated with VAN, MER, and/or VPA via intrathecal injection. Our method provided a detection tool to study the effective concentrations of these three drugs in CSF from patients administered via intrathecal injection.

LC-MS/MS Conditions
Due to their strong hydrophily, VAN and MER are not well retained in the C18 column; thus, we selected the C8 column to improve retention efficiency. Acetonitrile and methanol were tested for the mobile phase. Acetonitrile was chosen for the mobile phase because it provided lower background interferences and column pressure than methanol. To evaluate the effects of different additives on the mobile phase, we compared the addition of ammonium acetate, formic acid (FA), and acetic acid at different concentration levels. The addition of ammonium acetate can increase the buffering capacity of the system but leads to a decreased analyte signal. The addition of acetic acid could not improve the signal intensities of analytes at 0.05% and improved slightly at 0.1%. Formic acid (FA) was added to the mobile phase to improve the ionization of the analytes, and 0.05% FA in the mobile phase instead of 0% and 0.1% were used to improve the signal intensities and peak shapes. With the appropriate mobile phase (0.05% FA), the analytes showed good chromatographic peak shapes, minimum interference, and stable retention times.
VPA in the positive ion mode was unsuitable due to strong interference from endogenous compounds; it could only be detected in the negative ion mode. Therefore, MS should be switched between positive and negative ion modes for the simultaneous determination of these three analytes.

Sample Preparation
Due to significant differences in the lipophilicity of these analytes, protein precipitation was selected for sample preparation. During the optimization of the sample preparation step, methanol and acetonitrile were tested for protein precipitation. However, the analyte peaks were strongly influenced when protein-precipitated by methanol. It is possibly caused by the different lipophilicity of mobile phase and precipitation solvents, leading to a solvent effect. However, the calibration curves of VPA in samples precipitated by acetonitrile alone exhibited non-linearity. We overcame this issue by adding a mixture of acetonitrile and acid to precipitate the protein. for 0.5% acetic acid; r = 0.9830 for 1% acetic acid. In this study, 0.5% FA in acetonitrile was selected for protein precipitation. Notably, the stability of MER in serum and plasma differed at RT, with MER in serum showing greater stability ( Figure 1). Therefore, a serum sample is chosen for clinical detection in this study.
Molecules 2023, 28, x FOR PEER REVIEW acetonitrile and acid to precipitate the protein. The addition of different FA and acet products different r values for the calibration curves of VPA: r = 0.9659 for 0.2% F 0.9960 for 0.5% FA; and r = 0.9689 for 1% FA; r = 0.9539 for 0.2% acetic acid; r = 0.96 0.5% acetic acid; r = 0.9830 for 1% acetic acid. In this study, 0.5% FA in acetonitri selected for protein precipitation. Notably, the stability of MER in serum and plasm fered at RT, with MER in serum showing greater stability ( Figure 1). Therefore, a sample is chosen for clinical detection in this study.

Assay Performance and Validation
The method was confirmed to be specific, as no peaks due to endogenous c nents were observed in serum or CSF around the retention times of t he analytes ( Figure 2). In addition, no crosstalk or carryover was observed among analytes a The calibration curves were linear over the range 5-500 ng/mL for MER, 0.1-10 μg/ VAN and 1-100 μg/mL for VPA in serum and 50-5000 ng/mL for MER, 1-100 μg/m VAN and 2-200 μg/mL for VPA in CSF with the following equations: y = 0.0044x + 0 r = 0.9970 for MER in serum; y = 0.0109 × − 0.00023, r = 0.9984 for VAN in serum; y = x − 0.002, r = 0.9955 for VPA in serum; y = 0.0006x + 0.00198, r = 0.9959 for MER in = 0.0805x − 0.00944, r = 0.9982 for VAN in CSF; and y = 0.0108x + 0.00087, r = 0.99 VPA in CSF. The LLOQs of this method were 5 ng/mL for MER, 0.1 μg/mL for VAN 1 μg/mL for VPA in serum and 50 ng/mL for MER, 1 μg/mL for VAN and 2 μg/m VPA in CSF with signal-to-noise ratio > 10:1 (Figure 2), and accuracies and prec within 11.1% (Table 1).

Assay Performance and Validation
The method was confirmed to be specific, as no peaks due to endogenous components were observed in serum or CSF around the retention times of t he analytes and IS ( Figure 2). In addition, no crosstalk or carryover was observed among analytes and IS. The calibration curves were linear over the range 5-500 ng/mL for MER, 0.1-10 µg/mL for VAN and 1-100 µg/mL for VPA in serum and 50-5000 ng/mL for MER, 1-100 µg/mL for VAN and 2-200 µg/mL for VPA in CSF with the following equations: y = 0.0044x + 0.00033, r = 0.9970 for MER in serum; y = 0.0109 × − 0.00023, r = 0.9984 for VAN in serum; y = 0.0195 x − 0.002, r = 0.9955 for VPA in serum; y = 0.0006x + 0.00198, r = 0.9959 for MER in CSF; y = 0.0805x − 0.00944, r = 0.9982 for VAN in CSF; and y = 0.0108x + 0.00087, r = 0.9947 for VPA in CSF. The LLOQs of this method were 5 ng/mL for MER, 0.1 µg/mL for VAN, and 1 µg/mL for VPA in serum and 50 ng/mL for MER, 1 µg/mL for VAN and 2 µg/mL for VPA in CSF with signal-to-noise ratio > 10:1 (Figure 2), and accuracies and precisions within 11.1% (Table 1).   Table 1 presents the accuracy and precision data for each analyte in serum and CSF. The intra-and inter-day accuracies (as relative error, RE) ranged from −6.2 to 6.3% with precision (as relative standard deviation, RSD) ≤7.4% (intra-day) and 14.1% (inter-day), respectively, for each analyte in serum and CSF.
The matrix effects and recovery efficiency results showed that the recoveries were all reproducible in low, medium, and high QC samples and consistent across the concentration range studied. In terms of matrix effects, Table 2 shows the actual concentrations  Table 1 presents the accuracy and precision data for each analyte in serum and CSF. The intra-and inter-day accuracies (as relative error, RE) ranged from −6.2 to 6.3% with Molecules 2023, 28, 2439 5 of 11 precision (as relative standard deviation, RSD) ≤ 7.4% (intra-day) and 14.1% (inter-day), respectively, for each analyte in serum and CSF.
The matrix effects and recovery efficiency results showed that the recoveries were all reproducible in low, medium, and high QC samples and consistent across the concentration range studied. In terms of matrix effects, Table 2 shows the actual concentrations (mean ± SD) as the percentage of nominal concentration. Matrix effects of VAN, MER, and VPA in serum and CSF were low based on Table 2. Therefore, the detection of these three drugs in serum and CSF were not affected by the sample from different patients. In terms of stability (Table 3), the concentrations under the various test conditions were all within ±12.8% of the nominal concentrations, indicating no significant degradation of the analytes under any of the storage conditions tested. Stock solutions of MER, VAN, and VPA in acetonitrile: water (1:1, v:v) are stable at −20 • C for 1 month and stable at RT for 24 h with a range from 94.1 to 107.7% and RSD% ≤ 5.44%. The evaluation of MER concentration in six 5000 ng/mL of serum samples and 50,000 ng/mL of CSF samples after 10-fold dilution with blank serum and CSF resulted in an accuracy of 101.6% and RSD of 3.8% for serum and accuracy of 97.4% and RSD of 4.1% for CSF. Thus, the developed assay can be used to quantify MER in human serum and CSF samples that are diluted to a maximum of 10 folds.

Clinical Application
Based on prior studies [17], the CSF to serum albumin ratio (CSAR; calculated as CSF albumin (mg/L)/serum albumin (g/L)) was used to evaluate BBB function. The definition of BBB damage was derived from age-adjusted reibergrams (normal if < 6.5 in patients aged < 40 years or < 8 in patients > 40 years). All patients selected in this study experienced intracranial infection after craniotomy and suffered from varying degrees of BBB damage. The effective concentrations of MER, VAN, and VPA in plasma were 4-10 µg/mL [18], 5-10 µg/mL [19], and 50-100 µg/mL [20], respectively. The CSAR and drug concentrations for the selected patients in this study are shown in Table 4. The developed LC-MS/MS method was able to capture all the serum and CSF concentrations of VAN and VPA, and MER with 10-fold dilution in all patients. In this study, due to administrate via intrathecal injection CSF concentrations were higher than serum concentrations. Compared with intravenous injections, an intrathecal injection can offer superior therapeutic effects. Although the CSF levels did not reach the effective concentration reported, these three drugs still provide sufficient therapeutic effects because the effective concentrations reported were obtained from clinical studies administrated via intravenous injections. When administrated by intravenous injection, the actual concentrations in CSF were significantly lower than in serum. It indicated the effective concentrations of these three drugs in CSF administrated by intrathecal injection may be lower than the reported effective concentrations based on the clinical study ad-Molecules 2023, 28, 2439 7 of 11 ministrated via intravenous injections. Our method provided a detection tool to study the effective concentrations of these three drugs in CSF from patients administered via intrathecal injection. To prevent and reduce adverse drug reactions while ensuring therapeutic effects, the effective concentrations of these three drugs in CSF administrated by intrathecal injections are highly essential. In addition, even when the same dose is administered, the drug concentration varies greatly across different patients. Such variation is affected by many factors such as the physiological and pathological states [21][22][23]. For this reason, it is necessary to monitor drug concentrations in CSF and serum.

MS Conditions
Detection employed a Qtrap 5500 mass spectrometer (AB Sciex, Concord, ON, Canada) equipped with a TurboIon Spray™ source. Data acquisition and integration were performed using Applied Biosystems Analyst software version 1.

Sample Preparation
A 50 µL sample (or calibration standard or QC sample) was added to 50 µL IS working solution and 400 µL 0.5% FA in acetonitrile. After vortexing for 30 s and centrifuging at 9000× g for 10 min, the supernatant (5 µL) was injected into the LC-MS/MS system. A matrix comparison between human serum and plasma (EDTA) was also performed to investigate the stability of the assay in human matrices ( Figure S1).

Assay Validation
Assay validation was performed in accordance with the US FDA Guidance for Industry on Bioanalytical Method Validation [24]. Specificity was determined by analysis of blank human serum and CSF samples from six different individuals. Linearity was evaluated by fewest squares weighted (1/×2) linear regression of duplicate calibration curves prepared using three separate batches of serum and CSF. The lowest concentration on the calibration curve that could be determined with 80.0-120.0% accuracy and had RSDs not exceeding 20.0% was chosen as the LLOQ. Accuracy and precision were determined by analyzing QC samples on three separate days. Carryover was assessed by analysis of blank serum and CSF immediately following injection of an upper limit of quantitation (ULOQ) sample. Carryover was considered negligible if the peak area at the retention time of the analyte in the blank sample was <20% of the analyte peak area in the LLOQ sample. Recovery was determined by comparing the mean peak areas of analytes in six replicate QC samples with those of post-extraction blank serum and CSF samples spiked at the same concentrations. Matrix effects were determined by comparing the mean peak areas of post-extraction blank serum and CSF samples spiked at QC concentrations with those of corresponding standard solutions. Stability in serum and cerebrospinal fluid was assessed by assay of triplicate QC samples after storage at −20 • C for 30 days, at room temperature (RT) for 6 h, and after three freeze-thaw cycles (−20 • C to RT). Stability in processed samples was assessed by assay of triplicate QC samples after storage in the autosampler at 15 • C for 6 h and at RT for 6 h. The stability of stock solution in acetonitrile: water (1:1, v:v) was evaluated at −20 • C for 1 months and at RT for 24 h. MER, VAN, and VPA with final concentrations of 100 ng/mL, 2 µg/mL and 20 µg/mL in mobile phase were prepared in triplicate from old and freshly made stock solutions. The stability of stock solution was tested by comparing the instrument response of samples prepared from old stock solution with that of samples prepared from new stock solution. To evaluate dilution integrity of MER samples, serum and CSF samples with final MER concentrations of 5000 ng/mL and 50,000 ng/mL (10 times ULOQ) were prepared in six replicates and diluted with blank serum and CSF to give the final concentration of 500 ng/mL and 5000 ng/mL (10-fold dilution). Diluted samples were processed following the method described in Section 3.5. and analyzed with calibration standards prepared on the same day. Accuracy and precision within ±15.0% were set as acceptance criteria.

Clinical Assessment
To demonstrate the clinical value of the proposed assay, 62 plasma and CSF samples obtained from patients (age range: 16−69 years old) at the First Hospital of Jilin University in 2022 were analyzed. All enrolled patients had experienced intracranial infection and were treated with VAN, MER, and/or VPA at 0.5 g, 2 g, and 1.5 g via intrathecal injection after craniotomy. This study was approved by the Human Research Ethics Committee of the First Hospital, Jilin University (protocol code is 2022-037 and date of approval is 22 February 2022).
Samples were analyzed to quantify MER, VPA, and VAN. When the level of MER in the serum or CSF samples was over 0.5 µg/mL or 5 µg/mL, the samples were 10-fold diluted by blank serum or CSF. Serum samples were obtained after centrifugation of blood in coagulation-promoting vacuum tubes. Isolated serum and CSF were immediately frozen (-20 • C) until sample preparation. To observe the stability of MER at RT, we used samples from six healthy subjects whose blood was collected with K2-EDTA vacuum tubes and coagulation-promoting vacuum tubes to obtain plasma and serum samples after centrifugation, respectively.

Conclusions
VAN, MER, and VPA are commonly used to treat intracranial infection after craniotomy and to prevent epilepsy. Monitoring these drugs in the patients' serum and CSF could be used to adjust the dosages. This approach could help realize individualized medication and prevent and reduce drug adverse reactions while ensuring the therapeutic effect. In this study, an LC-MS/MS method for simultaneous determination of MER, VAN, and VPA in CSF and serum was developed and validated. The assay has several favorable characteristics including rapid sample preparation, a small sample volume of 50 µL, and a short cycle time (8 min), allowing a high-throughput analysis for clinical therapeutic drug monitoring. Although the CSF levels in this study did not reach the effective concentration reported, these three drugs still provide sufficient therapeutic effects. It indicated the effective concentrations of these three drugs in CSF administrated by intrathecal injection may be lower than those administrated via intravenous injections. It is very essential to investigate the effective concentrations of these three drugs in CSF administrated by intrathecal injection. Our method provided a detection tool to investigate the effective concentrations of these three drugs in CSF from patients administered via intrathecal injection. For this reason, our method is an effective analysis approach for clinical purposes. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.