A validated LC-MS/MS method for the quantitation of cefazolin in human adipose tissue: application of EMR-Lipid sorbent as an efficient sample clean-up before mass spectrometric analyses

Abstract

A novel, simple, rapid, and sensitive high-performance liquid chromatography-tandem mass spectrometry method was developed to determine cefazolin concentrations in human adipose tissue. Sample preparation was performed by protein precipitation followed by using Captiva EMR-Lipid plates. The mobile phase consisted of 5 mM ammonium formate and 0.1% formic acid in water and 0.1% formic acid in ACN, and was pumped through a Synergi Fusion-RP column with a gradient elution program at a flow rate of 0.3 mL/min. The mass spectrometer was operated in a positive ion mode. Cloxacillin was used as an internal standard due to the observed cross-signal contribution between cefazolin and ¹³C₂,¹⁵N-cefazolin. The method was validated according to the FDA and EMA guidelines and passed all the acceptance criteria. The calibration range was 0.05 – 50 μg/mL in adipose tissue homogenate (0.15 – 150 μg/g in adipose tissue), precision CV <4.5%, accuracy within 93.1 – 100.4%. The carry-over was negligible, recovery of the method was high, and no significant matrix effect was present. Rat subcutaneous adipose tissue was demonstrated to be a suitable surrogate matrix for human adipose tissue. The validated method was successfully applied in a pilot pharmacokinetic study and will further be used in a large cohort of non-obese and obese patients dosed prophylactically with cefazolin before surgeries.

Introduction

Cefazolin is a first-generation cephalosporin that exhibits bactericidal activity. The drug is widely used in various bacterial infections and as antimicrobial prophylaxis before surgeries [1]. A recent study has shown that cefazolin might have better efficacy in preventing surgical site infections (SSIs) than other drug regimens [2]. Cefazolin shortage in Japan revealed that using alternative beta-lactams doubled the risk of reoperation due to deep SSIs after spine surgery [3].

As in the case of other beta-lactam antibiotics, the efficacy of cefazolin largely depends on the appropriate level of a drug at the site of action that needs to be maintained for a period of potential bacterial contamination [4]. For bactericidal effect, cefazolin concentrations must exceed the minimal inhibitory concentration, i.e., 0.5 – 4 μg/mL for most common microorganisms causing SSIs, minimum at the opening and closure of the wound [5–7]. Pharmacokinetics of drugs might change in special patient populations, e.g., in obese patients. The pharmacokinetic (PK) processes most likely to be affected in obesity are distribution and elimination [8], both of which might significantly impact the concentration of a drug at target sites.

Currently, 1 in 3 people worldwide are overweight or obese [9]. It is estimated that, by 2030, nearly half of the adult U.S. population will face obesity and 25% – severe obesity [10]. This implies a legitimate concern about the efficacy and safety of drugs in this population of patients. Among patients treated with cefazolin, there was a trend toward a higher prevalence of SSIs in obese than in non-obese patients [11], and underdosed obese patients were more prone to develop SSIs [12].

The cefazolin plasma concentrations decline in a biphasic manner after intravenous administration [5,7], which indicates that the drug does not equilibrate rapidly throughout the body, and its transfer to slowly perfused tissues occurs. Brill and colleagues [5] demonstrated that the distribution of free cefazolin from central to subcutaneous compartment was significantly limited in morbidly obese subjects. The advantage of microdialysis utilized by authors is measuring a free concentration of the drug at target site; however, the technique also requires expensive instrumentation and trained personnel, which result in only a small number of patients enrolled in the cefazolin microdialysis studies [5,7], and challenges with its application in routine clinical settings.

Several authors have tested the total concentrations of cefazolin in adipose tissue in different populations of patients, including overweight and obese patients. However, except for one study [13], the utilized high-performance liquid chromatography (HPLC) methods (Table 1) were not validated for use in adipose tissue specimens or presented only limited validation data [6,14–17]. In the literature, there are no stability data for cefazolin in adipose tissue, information which is crucial for planning the experiments and providing reliable results.

Table 1.

The HPLC methods used by other authors to determine total concentrations of cefazolin in human adipose tissue biopsies.

Amount of matrix used for the assay	Method	Sample clean-up	Calibration range	Reference
500 μL of subnatant after centrifugation	HPLC-UV (265 nm)	PPT with ACN + LLE with DCM	LLOQ – 1 μg/g	[6]
250 mg	HPLC-MS/MS	LLE with hexane and 100 mM ammonium acetate buffer (pH 7.5)	2 – 20 μg/g	[13]
200 μL a	HPLC-UV (254 nm)	PPT with ACN + LLE with DCM	0.5 – 50 μg/g	[15]
0.5 – 1 cm2 b	HPLC c	PPT with perchloric acid	no data	[14]
150 μL d	HPLC-UV (254 nm)	PPT with ACN d	2 – 100 μg/mL d	[16]
no data	HPLC-UV (254 nm)	PPT with 70% MeOH/ 30% 0.1M Na acetate (pH 5.2)	no data	[17]

^ainformation not mentioned directly, authors referred to a method validated for serum, pericardial fluid, and homogenate of atrial appendage

^binformation not mentioned directly, 0.5 – 1 cm² was collected during surgery

^cno information about a type of detection

^dinformation not mentioned directly, authors referred to a method validated for plasma

Abbreviations: ACN, acetonitrile; DCM, dichloromethane; HPLC, high-performance liquid chromatography; LLE, liquid-liquid extraction; LLOQ – lower limit of quantitation; MeOH, methanol; MS/MS, tandem mass spectrometric detection; PPT, protein precipitation; UV, ultraviolet detection.

Here, we present a fully validated HPLC method coupled with tandem mass spectrometry (LC-MS/MS) for the determination of cefazolin in adipose tissue. The advantages of the method are: utilizing mass spectrometry, a small amount of matrix required for the assay, a large calibration range, and effective sample clean-up. Methods available in the literature were mostly based on the HPLC coupled with ultraviolet detection (UV) [6,15–17], which is much more prone to co-eluting interferences and, therefore, less selective than mass spectrometry. Moreover, sample clean-up proposed by other authors was often labor-intensive and time-consuming (as liquid-liquid extraction (LLE) [13] or protein precipitation (PPT) followed by LLE [6]), or only minimal (PPT) [14,16,17], which – in turn – can contribute to interferences and matrix effect, especially in LC-MS/MS. Therefore, a need for a novel sample preparation and analysis method was noticed that would overcome these limitations and enable high throughput. We demonstrated that filtration plates with a sorbent removing lipids could be a simple and fast alternative for effective sample preparation before LC-MS/MS analyses. The developed method was successfully applied in 12 patients and will further be used in a larger PK study assessing the distribution of cefazolin to subcutaneous adipose tissue in obese and non-obese patients.

1. Materials and methods

1.1. Reagents and materials

Cefazolin (Alfa Aesar, purity 99.2%) and cloxacillin (Alfa Aesar, purity 92.8%), both available as sodium salts, were purchased from Fisher Scientific (Fair Lawn, NJ, USA); stable isotope-labeled cefazolin (¹³C₂,¹⁵N-cefazolin, purity ≥95%) was purchased from Cayman Chemical (Ann Arbor, MI, USA). LCMS grade solvents and additives: acetonitrile (ACN), methanol (MeOH), water, water with 0.1% formic acid, formic acid (purity ≥99.0%), ammonium formate (Honeywell, purity ≥99.0%) were obtained from Fisher Scientific. Zirconium oxide beads for tissue homogenization (2.0 mm, ZROB20) were obtained from Next Advance (Troy, NY, USA). A phosphate-buffered saline (PBS) 10× (molecular biology grade) was purchased from Corning (Corning, NY, USA). Captiva Enhanced Matrix Removal (EMR)-lipid plates were obtained from Agilent Technologies (Santa Clara, CA, USA).

Blank subcutaneous rat adipose tissue was collected at Rutgers University (Piscataway, NJ, USA) from adult male Sprague Dawley rats. The protocol (ID999900460) was approved by the Rutgers IACUC. Subcutaneous adipose tissue from humans (approximately 0.5 – 2 g) was collected from patients hospitalized in Robert Wood Johnson University Hospital in New Brunswick (NJ, USA); samples were stored at −80°C before processing. The protocol (Pro20150001538) was approved by the Rutgers New Brunswick Health Sciences IRB. All patients gave informed consent before being included in the study.

1.2. Adipose tissue homogenization

Adipose tissue homogenate was prepared by thawing adipose tissue samples at room temperature, cutting them into smaller pieces, and accurately weighting before homogenization in 1× PBS at a ratio of 1:2 (m/v). A 1× PBS buffer was obtained by a 10-fold dilution of concentrated PBS with LCMS water. Adipose tissue samples were homogenized in 5 mL Eppendorf tubes (Eppendorf, Hamburg, Germany) at a maximum speed for 5 min using a Blender Bullet Gold homogenizer (Next Advance, Averill Park, NY, USA) and 2.0 mm zirconium oxide beads (Next Advance, Troy, NY, USA), and stored at −80°C before analysis. After thawing, the homogenate was vigorously vortexed to ensure uniform sampling.

1.3. Solutions, calibrators, and quality controls

The stock solution of cefazolin (5 mg/mL of cefazolin free base, correction factor 0.9539) was prepared by dissolving cefazolin sodium salt in water; working solutions of cefazolin (0.5 – 500 μg/mL) were obtained by further dilution of cefazolin stock solution with water. Adipose tissue calibrators and quality controls (QCs) were prepared by a 10-fold dilution of working solutions with adipose tissue homogenate (rat or human). After dilution, the concentrations of calibrators in adipose tissue homogenate were 0.05 – 50 μg/mL, which corresponded to cefazolin concentrations of 0.15 – 150 μg/g in adipose tissue.

The stock solution of IS (5 mg/mL of cloxacillin free base, correction factor 0.9519) was prepared by dissolving cloxacillin sodium salt in MeOH. The working solution of IS (50 μg/mL) was prepared by further dilution of cloxacillin stock solution with ACN.

All stock and working solutions were stored at −20°C. Calibrators and QCs were prepared on a daily basis.

1.4. Sample preparation

Fifty microliters of adipose tissue homogenate (standards or unknown samples) were placed in a 1.5 mL Eppendorf tube, and 25 μL of the IS working solution (50 μg/mL) were added. Samples were spiked with 400 μL of 0.5% formic acid in ACN, vortexed for 15 s, and shaken out for 5 min. After centrifugation at 15,700 × g and 4°C for 5 min, the supernatant was loaded onto the Captiva EMR-lipid plates, and a vacuum was applied. The filtrate was transferred to the HPLC vials with glass inserts, and 2 μL were injected into the column.

1.5. HPLC conditions

Chromatographic separation was carried out on an ExionLC AD system (AB Sciex, Framingham, MA, USA) equipped with a Synergi Fusion-RP column (50 × 2 mm, 4 μm) protected by a corresponding guard column (Phenomenex, Torrance, CA, USA). The autosampler and the column compartment were maintained at 10°C and 40°C, respectively. Mobile phase A (MPA) was composed of 5 mM ammonium formate and 0.1% formic acid in LCMS water, while mobile phase B (MPB) – 0.1% formic acid in LCMS-grade ACN. The following gradient mode was applied at a flow rate of 0.3 mL/min: 0.0 – 0.2 min, 5% MPB; 0.2 – 3.0 min, 5 – 95% MPB; 3.0 – 3.5 min, 95% MPB; 3.5 – 4.0 min, 95 – 5% MPB; 4.0 – 5.5 min, 5% MPB. The mobile phase was directed to waste for the first 1.8 and the last 2.2 min of the analysis to prevent mass spectrometer contamination with salts and other early- and late-eluting matrix components. A mixture of in-house MilliQ water (Millipore, Billerica, MA, USA) and ACN (50/50, v/v), both supplemented with 0.1% formic acid, was used for external and internal rinsing as well as injection port washing. The total run time, including gradient elution and autosampler rinsing, was about 6.7 min.

1.6. Mass spectrometer settings

The AB Sciex QTRAP 6500+ mass spectrometer (Framingham, MA, USA) was equipped with an IonDrive Turbo V Source and was operated in positive ionization mode (ESI+). Detection and quantification were performed using the multiple reaction monitoring modes (MRMs) with a dwell time of 100 ms per transition. The source parameters were as follows: source temperature, 450°C; ion spray voltage, 5 kV; collision gas (nitrogen), medium level; ion source gas 1 and 2 (zero air), 60 psi; curtain gas (nitrogen), 30 psi. Nitrogen and zero air were provided by Genius 1024 generator (Peak Scientific; Billerica, MA, USA). Two mass transitions were monitored for cefazolin [454.92>323.00 and 454.92>155.90] and for IS [436.77>278.00 and 436.77>160.00]. Other analyte-specific parameters are listed in Table 2.

Table 2.

MRM settings for cefazolin and its internal standard, cloxacillin.

	MW [g/mol]	Form of ion	Precursor ion [m/z]	Product ions [m/z]*	DP [V]	EP [V]	CE [V]	CXP [V]
cefazolin	454.51	[M+H]+	454.92	323.00	16	10	15	18
				155.90	16	10	21	18
cloxacillin	435.88	[M+H]+	436.77	278.00	65	10	25	16
				160.00	65	10	17	20

^*quantifier, qualifier

Abbreviations: CE, collision energy; CXP, collision cell exit potential; DP, declustering potential; EP, entrance potential; m/z – mass to charge ratio; MRM, multiple reaction monitoring; MW, molecular weight.

1.7. Data collection and processing

Data were acquired using Analyst Software (version 1.7.1), and quantification was performed with SCIEX OS Software (version 1.6.1) (both from AB Sciex, Framingham, MA, USA).

2. Method validation

The developed method was validated according to the current bioanalytical guidelines from the U.S. Food and Drug Administration (FDA) [18] and European Medicines Agency (EMA) [19]. The following parameters were validated: selectivity, lower limit of quantitation (LLOQ), accuracy, precision, recovery, matrix effect, carry-over, and cefazolin stability.

2.1. Selectivity

The selectivity of the method was investigated by analyzing blank samples. The method was considered selective if potential interferences contributed to <20% of the peak area of the LLOQ and <5% of the area of IS.

2.2. Calibration range

Calibration curves included eight concentration levels and were constructed in the range of 0.05 – 50 μg/mL of cefazolin in adipose tissue homogenate. Each time, a blank sample (no cefazolin and no IS) and a zero sample (spiked with IS only) were also prepared along with the adipose tissue calibrators to confirm the validity of the run, no interferences, and lack of contamination. A peak area ratio between cefazolin and IS (cloxacillin) was plotted vs. nominal concentration of cefazolin, and an unweighted power function was implemented. The LLOQ was determined based on the signal-to-noise ratio ≥10 and the sufficient precision (≤20%) and accuracy (80 – 120%).

2.3. Accuracy and precision

Precision and accuracy (within-run and between-run) were examined by spiking rat adipose tissue homogenate with a known concentration of cefazolin. Four concentration levels were tested, corresponding to the method’s lower (LLOQ, 0.05 μg/mL) and upper limit of quantitation (ULOQ, 50 μg/mL), as well as two times LLOQ (low quality control (LQC), 0.1 μg/mL) and half of the ULOQ (medium quality control (MQC), 25 μg/mL). Five replicates per concentration level were analyzed during the same day for within-run experiments; for between-run experiments, cefazolin was measured in QCs prepared over six different days. The accuracy of the method was reported as a percentage of the mean observed concentration compared to the nominal concentration while precision as a coefficient of variation (CV%).

2.4. Recovery and matrix effect

Recovery and matrix effect were assessed at two concentration levels, LQC and ULOQ. Adipose tissues from several individual rats were used for recovery (n = 3) and matrix effect experiments (n = 6). One series of adipose tissue homogenates was spiked before extraction (before PPT, series A), the other one after extraction (after filtration through Captiva EMR-lipid plates, series B). The third series was prepared with water instead of adipose tissue homogenate and was spiked post-extraction (series C). Samples were prepared in duplicate, and the mean concentration was used for calculations. Parameters were calculated for each individual adipose tissue, and their mean and CV were reported.

Recovery was calculated as a percentage ratio of the mean peak areas observed in adipose tissue homogenate before and after extraction (Recovery [%] = A/B * 100%). Ion suppression or enhancement was calculated by matrix factors (MFs), i.e., percentage ratios of the mean instrument response observed in the presence of matrix and without matrix in samples spiked post-extraction (MF [%] = B/C * 100%). Moreover, IS-normalized MFs were calculated by dividing MFs for cefazolin and IS. The CV of the IS-normalized MFs from six lots of matrix should not be higher than 15%.

2.5. Stability

The stability of cefazolin was evaluated to ensure that sample storage and preparation do not affect the results. In each experiment, human and/or rat adipose tissue homogenate was spiked with low and high concentrations of cefazolin that covered the calibration range. Freeze-thaw and long-term stability in the matrix was only assessed for human samples as spiked rat homogenate was not stored in the freezer. Bench-top in-matrix stability and short-term in-filtrate stability tests were performed for both human and rat samples to confirm that cefazolin does not degrade during sample processing (neither in samples from humans nor in the surrogate matrix used to prepare calibration curves). For rat adipose tissue, long-term filtrate stability was also studied to test whether prepared calibration curves can be aliquot and used during subsequent days. In each stability experiment, QCs were prepared in triplicate and analyzed against fresh rat calibration curves. Freshly prepared stock and working solutions were used for stability tests.

To test short-term (bench-top) and long-term stability, the homogenate was spiked with cefazolin and left at room temperature (~20°C) or stored in the freezer (−80°C) before processing. In the freeze-thaw experiments, the spiked homogenate was kept frozen (−80°C) for at least 12 h before thawing, and three freeze-thaw cycles were studied. Short-term filtrate stability was assessed by injecting samples kept in the autosampler (10°C); for long-term stability, the filtrate was transferred to vials with glass inserts, sealed, and stored in the freezer (−20°C) before analyzing. Stability of cefazolin in aqueous solutions (−20°C) was assessed for 1 and 500 μg/mL as these two working solutions were used to prepare LQC and ULOQ samples. Cefazolin stability in water was also evaluated for 5 mg/mL (stock solution), diluted on the day of analysis to prepare appropriate working solutions, and then QCs.

2.6. Carry-over

Carry-over was inspected by injecting a blank sample following the highest calibration standard; the procedure was repeated five times (ULOQ, blank, ULOQ, blank, etc.). Carry-over was acceptable if the observed peak areas in blank samples did not exceed 20% of the analyte peak area at the LLOQ level and 5% of the peak area of IS [18,19]. Five LLOQ samples were also injected after the ULOQ samples (in the following sequence: ULOQ, LLOQ, ULOQ, LLOQ, etc.) to confirm that the observed carry-over did not compromise the accuracy of the method. Carry-over was considered negligible if back-calculated concentrations for all LLOQ samples were in the range of 80 – 120%.

3. Results and discussion

3.1. Method development

3.1.1. LC-MS/MS analysis

The precursor and product ion pairs (Q1/Q3 transitions), as well as compound-dependent parameters (DP, EP, CE, and CXP), were optimized by direct infusion of analytes into the ion source. Cefazolin solution for infusion was initially prepared in a mixture of water with 0.1% formic acid and ACN (1:1, v/v) at a concentration of 10 ng/mL, which is a recommended starting concentration for this model of the mass spectrometer. However, only sodium adducts were formed in a positive mode, probably as a result of using sodium salt of cefazolin as an analytical standard. Increasing the concentration of cefazolin to 1000 ng/mL and using ACN supplemented with 0.25% formic acid as a sample diluent helped to mitigate the problem. In the full scan Q1 mass spectrum, the most abundant ion was found for [M+H]⁺, though [M+Na]⁺ and [M+2Na-H]⁺ ions were still present (m/z ratios corresponding to M+23 and M+45, Figure 1).

Figure 1.

Q1 mass spectrum of cefazolin obtained by direct infusion in ESI+ mode.

Source-dependent parameters (IS, TEM, GS1, GS2) were optimized by the flow injection analysis by injecting a mixture of cefazolin and IS at a 10-times lower concentration than used during direct infusion.

HPLC conditions and mobile phase composition were optimized to obtain good signal intensity, peak shape, and satisfactory retention times. The LC flow had to be diverted to waste for the first 1.8 min of analysis to prevent contamination of mass spectrometer with non-volatile components of the homogenization buffer. Non-volatile salts, sodium chloride, potassium chloride, and phosphates should be avoided in mass spectrometry; they are, however, not retained on the column and can be easily eliminated from the samples by directing the first portion of the mobile phase to waste. The total run time of 5.5 min (6.7 min with the autosampler rinsing) was acceptable and enabled high throughput.

3.1.2. Cross-signal contribution between analyte and labeled IS

At an initial step of the method development, cross-signal contribution between cefazolin and its labeled IS (¹³C₂,¹⁵N-cefazolin) was evaluated. Separate solutions of cefazolin and ¹³C₂,¹⁵N-cefazolin (at different concentration levels) have been injected and analyzed. The experiment showed that labeled cefazolin did not contribute to the signal of cefazolin; however, cefazolin contributed to the signal of ¹³C₂,¹⁵N-cefazolin. Samples spiked at the ULOQ level generated a peak for ¹³C₂,¹⁵N-cefazolin with an area of approximately 1.0% of the response of cefazolin.

In the quantitative mass spectrometric assays, using IS is necessary as it controls not only for the potential analyte loss during sample preparation or injection but also for the ionization variability. Even though compound-specific, the last one might also depend on multiple factors that are changing on a daily basis and are difficult to control [20]. ISs labeled with a stable isotope(s) are recommended for LC-MS/MS analyses as they have a similar structure to a drug of interest, which results in similar physicochemical properties and, therefore, similar recovery, retention times, and behavior in the ion source. ¹³C₂,¹⁵N-cefazolin differs from cefazolin by three atoms; specifically, three atoms in cefazolin molecular structure are replaced with stable isotopes (two of them with ¹³C atoms and one with ¹⁵N atom). It accounts for a three mass unit difference in molecular weights between cefazolin and ¹³C₂,¹⁵N-cefazolin that is a recommended minimum to successfully separate them in the mass analyzer and avoid interferences [20].

Several reasons can lead to the observed analyte-to-IS signal contribution, such as standards’ impurity, the instrument drawbacks (resulting in insufficient time to clear the ions from the collision cell before the next transition is monitored), and the analyte’s isotopes impact [21–23]. The first possibility has been eliminated as cefazolin contributed to the signal of ¹³C₂,¹⁵N-cefazolin not only in samples manually spiked with cefazolin but also in those coming from the patients. The classical ‘cross-talk’, often present in older mass spectrometers, was successfully minimized in modern instruments; it occurs when two or more analytes share similar product ions, and the mass difference between precursor ions is very small [23]. In our case, similar product ions for cefazolin and ¹³C₂,¹⁵N-cefazolin were monitored only for the first transition (454.92/155.90 for cefazolin and 457.87/156.10 for ¹³C₂,¹⁵N-cefazolin); for the second transition, product ions with a three mass-to-charge ratio difference were monitored (454.92/323.0 for cefazolin and 457.87/325.90 for ¹³C₂,¹⁵N-cefazolin) while the cross-signal contribution was still present. Also, adding ‘dummy transitions’ between cefazolin and ¹³C₂,¹⁵N-cefazolin (i.e., additional MRM transitions separating cefazolin and ¹³C₂,¹⁵N-cefazolin [23]) did not improve the results, which indicated that the observed signal might not be instrument-related. Therefore, the isotopes contribution seemed to be the most likely cause of the observed phenomenon.

In its chemical structure, cefazolin contains three atoms of sulfur (Figure 1), known to exist in nature as several stable isotopes. In total, almost 5% of all naturally occurring sulfur atoms (³⁴S – 4.21%, ³³S – 0.75%, and ³⁶S – 0.02%) are 1 – 4 Da heavier than the most common isotope (³²S – 95.02%) [24], which might cause interferences in the mass analyzer. The solution to overcome the cross-signal contribution between cefazolin and its labeled IS could be increasing the concentration of labeled IS to the level when the cross-signal contribution is not meaningful [22]. Cefazolin signal at the ULOQ level generated a signal for ¹³C₂,¹⁵N-cefazolin equal to 7.7% of the peak area of ¹³C₂,¹⁵N-cefazolin at the concentration of 5 μg/mL in tissue homogenate and 2.3% at the concentration of 15 μg/mL. However, we noticed that a higher concentration of labeled IS started to generate carry-over. Thus, to maintain good accuracy and reproducibility of the method, we decided on using the structural analog of cefazolin as an IS.

Several authors have used ¹³C₂,¹⁵N-cefazolin in their LC-MS/MS analyses (e.g., [13,25,26]); however, the cross-signal contribution between cefazolin and ¹³C₂,¹⁵N-cefazolin was tested and reported in this work for the first time. This finding is important for developing reliable analytical methods: a high enough concentration of ¹³C₂,¹⁵N-cefazolin has to be used to keep good accuracy of the method and linearity of the calibration curves. Interestingly, even though the cross-signal contribution between analytes and their IS might significantly affect the results, as presented elegantly by Tan and colleagues [22], this aspect has not been directly mentioned in the validation guidelines [18,19]. FDA recommends assessing methods selectivity in blank and zero samples; this approach can indeed indicate issues with a cross-signal contribution between labeled-IS and analyte (by inspecting ion chromatograms of zero samples) but not vice-versa (i.e., cross-signal contribution between analyte and labeled-IS). Similarly, EMA suggests assessing selectivity only in blank samples, but a possible interference between IS and analyte is mentioned in the ‘reference standards’ section when referring to the purity of the IS. On the opposite, cross-talk between the analyte and labeled IS is not directly addressed in the guidelines, and is rarely mentioned in the literature when reviewing LC-MS/MS methods. Updating the bioanalytical guidelines and highlighting the necessity of assessing selectivity of the LC-MS/MS methods not only by evaluating blank and zero samples but also by injecting a standard at the ULOQ level without IS (to monitor cross-signal contribution between analyte and labeled IS) could facilitate the validation process.

Interference for another labeled cefazolin has previously been reported by Bellouard and colleagues: cefotaxime contributed to the signal of cefazolin-D₃ [27]. The chromatographic separation between these analytes was, however, satisfactory, and quantification was not affected (authors did not mention cross-signal contribution between cefazolin and cefazolin-D₃). In our work, good enough separation was impossible – due to similar physicochemical properties, analyte and its labeled IS will always elute at very similar retention times. This highlights the importance of evaluating the cross-signal contribution between analyte and its labeled IS at an early stage of method development. A cross-signal contribution between different analytes can be easily noticed on the ion chromatograms (provided that separation on the column is achieved; in such a case, a peak at the retention time of analyte B will be present in the MRM transition monitored for analyte A). However, a cross-signal contribution between analyte and its labeled IS could be easily missed (as retention times of analyte and its labeled IS will be almost the same).

3.1.3. Using surrogate matrix

Cefazolin is a common beta-lactam antibiotic widely used as a part of surgical prophylaxis in patients with no history of beta-lactam allergy [2,3]. It makes it challenging to obtain cefazolin-free adipose tissue from humans that could be used for method validation and preparing calibration curves. At a pre-validation step, we noticed that all human adipose tissues from the external source (bought from a vendor) that we planned to use as a blank matrix generated a signal for cefazolin; therefore, this human matrix could not be directly used for the assays without affecting the method’s sensitivity and accuracy.

Several approaches have been considered to solve this problem, including the standard addition method, background subtraction method, and using a surrogate matrix, as recommended for the analysis of endogenous compounds [28]. The advantage of the standard addition and background subtraction methods is using the same type of matrix as for the study samples (thus, matrix effect and recovery remain unaffected). However, the first method requires a large volume of the study sample and is time-consuming (as a separate calibration curve is prepared for each study sample by spiking this matrix with different concentrations of the analyte), while the second method might result in inconsistent LLOQ (as it highly depends on the amount of analyte present in the matrix used for the calibration curve, which can differ across various batches) [28]. Therefore, the third method has been implemented, and subcutaneous rat adipose tissue was used as a surrogate matrix to human adipose tissue. The suitability of this matrix has been thoroughly tested during the validation process, as explained in detail below.

3.2. Method validation

3.2.1. Selectivity

The selectivity of the method was confirmed by the absence of interfering peaks at the retention times of cefazolin and IS (cloxacillin). Blank adipose tissues from eight different sources were evaluated, including two human (patients that did not receive cefazolin during surgery) and six rat matrices. A minimum number of six individual matrices is recommended for evaluation [18,19]; though, a fewer number is acceptable in the case of rare matrices [19]. Due to difficulties with possessing adipose tissue from patients not dosed with cefazolin as prophylaxis, the method’s selectivity in humans was evaluated for two individuals. Figure 2 presents representative LC-MS/MS chromatograms for blank adipose tissue samples (rat and human), rat adipose tissue samples spiked with cefazolin at the LLOQ and ULOQ level, and an adipose tissue sample obtained from a patient dosed with cefazolin.

Figure 2.

Representative MRM chromatograms for analysis of cefazolin (CFZ) in adipose tissue. Ion chromatograms correspond to the signal monitored either for CFZ (A, B, C) or IS (D) and represent: A) rat adipose tissue homogenate spiked at the LLOQ level (0.05 μg/mL) overlaid with blank samples (no CFZ and no IS) extracted from rat (n=6) and human (n=2) adipose tissues; B) rat adipose tissue homogenate spiked at the ULOQ level (50 μg/mL); C) human adipose tissue from a patient dosed with CFZ (determined concentration of CFZ in adipose tissue homogenate – 4.6 μg/mL, which corresponded to 13.8 μg/g in adipose tissue); D) zero sample extracted from rat adipose tissue (concentration of CLX in the matrix – 25 μg/mL) overlaid with blank samples extracted from rat (n=6) and human (n=2) adipose tissues.

3.2.2. Calibration range

Calibration curves were constructed in the range of 0.05 – 50 μg/mL of cefazolin in the adipose tissue homogenate. For simplicity, we assumed that these values corresponded to 0.15 – 150 μg/g of cefazolin in adipose tissue (due to three times dilution with PBS buffer). The actual concentrations of cefazolin in adipose tissue would be a little higher due to the density of human and rat adipose tissue being lower than the density of water (0.90 – 0.92 g/mL in 15 – 37°C [29] vs. approximately 1 g/mL). However, the deviation was considered non-significant.

A large calibration range was chosen based on literature data [6,13–17] and preliminary results from our patients. Others reported the concentrations of cefazolin in adipose tissue ranging from 3.8 ± 1.8 μg/g [13] to 40.1 ± 24.1 μg/g [16] (after 25 mg/kg and 4 g dose of cefazolin, respectively). Four different regression models were tested during the validation process, including a linear function (without weighing, with 1/x and 1/x² weighing) and a power function (implemented in SCIEX OS software). Six independent calibration curves showed that the power function provided the best accuracy and repeatability of the results (mean back-calculated concentrations of calibrators within 98.1 – 101.7% with CV within 1.9 – 3.5%), with the correlation coefficients ≥0.999. Figure 3 shows the representative calibration curves for cefazolin in human and rat adipose tissue homogenates.

Figure 3.

Log-log plots with calibration curves and their equations for cefazolin (CFZ) in human and rat adipose tissue homogenates injected within the same analytical run. In the equations, ‘x’ corresponds to the concentration of CFZ in homogenate [μg/mL] while ‘y’ corresponds to the peak area ratio between CFZ and its IS (CLX).

3.2.3. Accuracy, precision, and LLOQ

The results for accuracy and precision are summarized in Table 3. The developed method was both precise (CV ≤4.5%) and accurate (back-calculated concentrations within 93.1 – 101.5%). The LLOQ was 0.05 μg/mL in homogenate (0.15 μg/g in adipose tissue). Based on the literature data (‘Introduction’ section, Table 1), our method provided the lowest LLOQ with only a small amount of matrix required for assay (50 μL of adipose tissue homogenate, which corresponded to less than 20 mg of human adipose tissue; in practice, a larger amount was homogenized and stored at −80°C).

Table 3.

Within-run and between-run accuracy and precision.

Nominal conc. in rat adipose tissue homogenate#	Within-run (n=5)			Between-run (n=6)
	back-calculated conc. in adipose tissue [μg/g]*	accuracy [%]	CV [%]	back-calculated conc. in adipose tissue [μg/g]*	accuracy [%]	CV [%]
LLOQ (0.05 μg/mL)	0.140 ± 0.005	93.1	3.9	0.149 ± 0.003	99.2	2.3
LQC (0.1 μg/mL)	0.284 ± 0.013	94.8	4.5	0.304 ± 0.009	101.5	2.8
MQC (25 μg/mL)	74.42 ± 2.08	99.2	2.8	75.31 ± 2.50	100.4	3.3
ULOQ (50 μg/mL)	147.78 ± 4.86	98.5	3.3	149.54 ± 5.17	99.7	3.5

^#adipose tissue was diluted 3-fold for preparation of homogenate (see text for details).

^*data are presented as mean ± SD

Abbreviations: CV, coefficient of variation; LLOQ and ULOQ, lower and upper limit of quantitation; LQC, low quality control; MQC, medium quality control.

3.2.4. Matrix effect

The effect of matrix on ion suppression or ion enhancement was assessed quantitatively, as recommended by Matuszewski and colleagues [30]. The results for rat adipose tissue homogenate fortified with low and high concentrations of cefazolin are presented in Table 4.

Table 4.

Matrix effect for cefazolin in rat adipose tissue.

	Concentration level *
	LQC [0.1 μg/mL]	ULOQ [50 μg/mL]
MFCFZ [%] (n=6)	101.3 ± 4.4	100.2 ± 5.9
	(4.4%)	(5.9%)
MFIS [%] (n=6)	93.6 ± 2.8	94.8 ± 6.4
	(3.0%)	(6.8%)
IS-normalized MFCFZ (n=6)	1.08 ± 0.05	1.06 ± 0.02
	(5.0%)	(1.7%)

^*data are presented as mean ± SD (CV%)

Abbreviations: CFZ, cefazolin; IS, internal standard; LQC, low quality control; MF, matrix factor; ULOQ, upper limit of quantitation.

The mean calculated MFs (for both concentrations) were 100.8 ± 5.0% for cefazolin (CV = 5.0%) and 94.2 ± 4.8% for IS (CV = 5.0%). These results indicate a lack of significant ion suppression (MF<100%) or ion enhancement (MF>100%), therefore no absolute matrix effect [30]. The mean IS-normalized MF was 1.07 ± 0.04 (CV = 3.8%). A value close to unity confirms similar matrix effects for cefazolin and IS even though a structural analog of cefazolin and not its labeled standard was used. For each studied parameter, the calculated CV was always <15%. Thus, the impact of the different sources of adipose tissue on the obtained results was negligible, and no relative matrix effect was present.

At the pre-validation step, different procedures were tested to prepare samples for LC-MS/MS analysis. The goal was to minimize the matrix effect and eliminate interferences that could impede the use of rat adipose tissue as a surrogate matrix for human samples. PPT followed by LLE was first investigated as used by other authors in cefazolin assays implementing UV (e.g., [6] for fat tissue samples) and MS/MS detection (e.g., [31] for plasma samples). We tested two different organic solvents (dichloromethane and methyl tert-butyl ether). However, the sample clean-up was unsatisfactory: chromatographic peaks were broad, and a significant matrix effect was suspected. Even though the matrix effect at this step of the method development was not assessed quantitatively, human samples spiked with cefazolin and analyzed against rat calibration curve provided unacceptably low accuracy. Obtained results suggested different slopes for rat and human calibration curves (at low concentrations, back-calculated concentrations for human samples were too low, at high concentrations – too high). This could be either due to significantly different absolute matrix effects for human and rat tissues (signal affected differently by various types of matrix), relative matrix effect (signal affected differently by various lots of the same matrix), or inconsistent recovery for different types of matrix. Thus, other cleaning procedures had to be investigated. Significant ion suppression was previously observed by Lillico and colleagues [13] when analyzing cefazolin in human adipose tissue by LC-MS/MS. Despite using LLE with hexane, authors reported a mean absolute matrix effect of 20.1 ± 11.7% [13].

Using Captiva EMR-lipid cartridges noticeably improved the peak shapes and reduced differences in peak areas between rat and human samples; however, we observed significant cross-contamination between samples that was identified to come from the SPE needles. To overcome this technical obstacle and improve the method’s throughput, Captiva EMR-lipid cartridges were replaced with Captiva-EMR-lipid plates, subsequently incorporated into the validation process. The Captiva-EMR-lipid plates eliminated the matrix effect and enabled using rat adipose tissue as a surrogate matrix for human samples. Human adipose tissue homogenate spiked with cefazolin and analyzed based on the rat adipose tissue calibration curve resulted in back-calculated concentrations within acceptable 85 – 115% (Table 5). In total, samples from five individuals were analyzed against three different rat calibration curves (on three occasions), and the observed CV for repeated measurements was <15%.

Table 5.

Human adipose tissue homogenate spiked with cefazolin and analyzed against rat calibration curve.

Nominal concentration [μg/mL]	Calculated concentration* [μg/mL]	Accuracy [%]	CV [%]
0.05	0.049 ± 0.006	97.1	13.1 (n = 2)
0.1	0.109 ± 0.008	109.3	7.2 (n = 2)
0.5	0.52	103.9	(n = 1)
1.0	1.04	104.5	(n = 1)
5.0	4.97	99.4	(n = 1)
10	10.61	106.1	(n = 1)
25	26.34	105.4	(n = 1)
50	46.83 ± 2.45	93.7	5.2 (n = 5)

^*for repeated measurements, data are presented as mean ± SD

Abbreviations: CV, coefficient of variation.

3.2.5. Recovery

Table 6 presents the results from the recovery experiment. The recovery was high and consistent, mean recovery (across all concentrations) was 97.9 ± 6.0 % (CV = 6.2%) for cefazolin and 101.5 ± 4.2% (CV = 4.2%) for IS. We observed similar recovery for cefazolin and cloxacillin, which further supports the decision about using cloxacillin as an IS for cefazolin. The suitability of cloxacillin as an IS for cefazolin in LC-MS/MS assays was previously reported for plasma samples (e.g., [32]). Our results showed that it could be successfully used with adipose tissue provided that the adequate sample cleaning procedure has been implemented.

Table 6.

Recovery for cefazolin from rat adipose tissue.

	Concentration level *
	LQC [0.1 μg/mL]	ULOQ [50 μg/mL]
RECFZ [%]	96.4 ± 6.4	99.3 ± 6.7
(n=3)	(6.6%)	(6.7%)
REIS[%]	101.8 ± 5.0	101.2 ± 4.4
(n=3)	(4.9%)	(4.3%)

^*data are presented as mean ± SD (CV%)

Abbreviations: CFZ, cefazolin; IS, internal standard; LQC, low quality control; RE, recovery; ULOQ, upper limit of quantitation.

3.2.6. Stability tests

The results of the stability tests are presented in Table 7. Cefazolin was stable under all tested conditions. Aqueous working solutions did not reveal significant degradation after 4 weeks (1 – 500 μg/mL), while the stock solution (5 mg/mL) was stable for at least 11 weeks. Adipose tissue homogenates (0.1 – 50 μg/mL) were stable for at least 6 h at room temperature, at least 8 weeks at −80°C, and at least 3 freeze-thaw cycles.

Table 7.

Stability of cefazolin under different storage conditions.

	Human samples		Rat samples
	mean accuracy [%]	CV [%] (n=3)	mean accuracy [%]	CV [%] (n=3)
	adipose tissue homogenate stability
• bench-top [20°C, 6 h]
LQC	98.1	9.6	100.9	1.8
ULOQ	96.6	1.0	97.8	3.0
• freeze-thaw [−80°C, 3 cycles]
LQC	102.0	0.4	-	-
ULOQ	90.0	3.4	-	-
• long-term [−80°C, 8 weeks]
LQC	92.3	5.5	-	-
ULOQ	97.7	0.9	-	-
	in-filtrate stability
• short-term [10°C, 24 h]
LQC	98.8	2.6	102.9	5.2
ULOQ	99.3	3.5	97.6	1.4
• long-term [−20°C, 12 days]
LQC	-	-	112.4	2.1
ULOQ	-	-	106.0	2.2
	aqueous solutions stability
• stock solution [−20°C, 11 weeks]
LQC	-	-	100.3	4.5
ULOQ	-	-	100.4	4.5
• working solutions [−20°C, 4 weeks]
LQC	-	-	102.7	2.0
ULOQ	-	-	95.9	4.0

LQC = 0.1 μg/mL, ULOQ = 50 μg/mL of cefazolin in adipose tissue homogenate.

According to our knowledge, this is the first report of the stability of cefazolin in adipose tissue homogenate. The only validated method for the determination of cefazolin in adipose tissue did not report the stability data [13]. Several authors have evaluated the stability of cefazolin in other human matrices, e.g., EDTA plasma [27], urine [31], CSF [27], and peritoneal dialysate [31]. The stability of a drug may be, however, different in various matrices, as presented for CDK4/6 inhibitors (unstable in liver and kidney homogenates while stable in K₂EDTA plasma and brain) [33]. It is important to highlight that tested stability in this study reflects stability in spiked adipose tissue homogenate, not in adipose tissue per se.

Literature data show the instability of cefazolin [34] and cloxacillin [35] under acidic conditions. Therefore, it was important to test stability in the autosampler after filtration through the Captiva EMR-Lipid plates (formic acid increased recovery). Mean back-calculated concentrations were within an acceptable range of 85 – 115 % for the filtrate (Table 7). However, an important aspect should be highlighted in the case of the processed samples. Stability of stock/working solutions and stability of an analyte in matrix (or homogenate) truly reflect the stability of the compound under tested conditions because IS is added after the storage period. But, for processed samples, back-calculated concentrations express only a relative change of the area of an analyte and its IS as IS is added before storage. In fact, the back-calculated concentration for the analyte might be close to 100% of the nominal concentration if both the analyte and IS degraded to the same extent. On the other hand, comparing peak areas might be misleading when extraction or filtration are involved. Nevertheless, the filtrate stability experiment confirmed that the accuracy of the samples injected during the long 24 h batches was not compromised. Moreover, it confirmed the possibility of using the same calibration curve for several runs when aliquots of calibrators (after filtration) are stored at −20°C or at lower temperatures.

3.2.7. Carry-over

Observed carry-over was considered acceptable (6.4% for cefazolin and 0.02% for IS after a single ULOQ). It did not affect the accuracy: all LLOQ samples injected after the ULOQ samples had back-calculated concentrations within the acceptable range (mean accuracy = 103.7%, CV = 2.3%). At the initial stage of the method development, carry-over was much higher; however, it was substantially reduced by including autosampler washing (Figure 4). Carry-over has been decreased from 0.10% of the area of the injected ULOQ to 0.02% of the ULOQ by implementing external and internal step rinsing with injection port washing.

Figure 4.

Carry-over for cefazolin in a first post-blank sample following triplicate ULOQ with different procedures of autosampler rinsing. A mixture of ACN and water (1:1, v/v) + 0.1% FA was used as a rinsing solvent. Carry-over accounted for 98.0% (no rinsing), 93.8% (external rinsing), and 16.7% (internal and external rinsing) of the peak area of cefazolin at the LLOQ level in a sample injected during the same analytical run.

Different solvents have been tested for internal rinsing, including ACN/water (1:1, v/v) + 0.1% formic acid, isopropanol (IPA)/water (1:1, v/v), IPA/ACN/MeOH/water (1:1:1:1, v/v/v/v) + 0.1% formic acid, and IPA/ACN/MeOH/acetone (1:1:1:1, v/v/v/v). All tested mixtures gave comparable results; thus, ACN/water (1:1, v/v) + 0.1% formic acid was chosen for the further method validation. Another research group [32] also reported about large carry-over for cefazolin (1 – 2% of the injected high standard); the problem has been mitigated by using strong solvent during the washing step (ACN/IPA/acetone, 45:45:10, v/v/v) along with the EDTA. In our case, rinsing solvent with IPA and acetone did not improve the results; moreover, acetone negatively affected signal response even though the equilibration step was included after the washing step (0.5 – 1.0 min). In routine practice, we injected an additional blank sample after expected high concentrations of cefazolin to monitor carry-over and decrease the possibility of sample cross-contamination.

4. In vivo application

The validated method was applied to determine concentrations of cefazolin in subcutaneous adipose tissue from 12 patients (Table 8). All subjects were dosed with 2 g of intravenous cefazolin prior to surgery. The calibration range was suitable to capture the total concentrations of cefazolin in most samples (Figure 5). The observed concentrations ranged from 0.11 to 24.94 μg/g at incision (8.13 (1.32 – 13.15) μg/g, median (interquartile range)) and from 0.84 to 23.58 μg/g at closure (7.00 (4.11 – 15.96) μg/g). Only in one sample (4.2%), the observed concentration of cefazolin was below the LLOQ limit (0.15 μg/g in adipose tissue). Full results of the pharmacokinetic study will be published in the future.

Table 8.

Characteristics of study population (n = 12) included in the pilot PK study.

Age [yr]	48.4 ± 8.5 45.0 (42.0 – 55.5)
BMI [kg/m2]	32.2 ± 8.8 31.2 (24.6 – 38.0)
Females [n (%)] Time from dosing to:	6 (50%)
- incision [min]	3.9 ± 5.0 1.5 (1.0 – 6.0)
- closure [min]	95.9 ± 75.4 75 (34.5 – 147)

continuous data are presented as mean ± SD and median (interquartile range), categorical data as number of patients (%).

Figure 5.

Concentrations of cefazolin in subcutaneous adipose tissue at incision and closure in patients dosed with 2 g of cefazolin.

Worth noting is that our LC-MS/MS method provided a wider calibration range than methods previously used in the literature (‘Introduction’ section, Table 1). A wide calibration range, especially low LLOQ, was necessary considering the application of the method to the PK study. So far, the only validated LC-MS/MS method for the determination of cefazolin in adipose tissue had the LLOQ equal to 2 μg/g [13]. Our pilot PK study showed that as many as 16.7% of samples (5/24) would fall below this threshold.

5. Conclusions

The developed HPLC-MS/MS method was robust, fast, and straightforward. It passed all the validation criteria and enabled high throughput. Rat subcutaneous adipose tissue was demonstrated to be a suitable surrogate matrix for human specimens, which helped to overcome the obstacle of obtaining cefazolin-free blank adipose tissue from humans.

Highlights.

• no matrix effect due to sample clean-up by Captiva EMR-Lipid plates

• suitability of rat adipose tissue as a surrogate matrix for human adipose tissue

• first report about cross-signal contribution between cefazolin and ¹³C₂,¹⁵N-cefazolin

• stability of cefazolin in adipose tissue specimens evaluated for the first time

• method fully validated according to the European and U.S. guidelines