Comparison of liquid–liquid extraction, microextraction and ultrafiltration for measuring free concentrations of testosterone and phenytoin

Abstract

Aim:

The purpose of the study was to find methods suitable for measuring the free concentrations of testosterone and phenytoin.

Materials & methods:

Sample solutions of the compounds in buffer and human albumin were processed using liquid–liquid extraction, microextraction and ultrafiltration and analyzed by LC–MS/MS.

Results:

Liquid–liquid extraction with dibutyl phthalate provided complete extraction from buffer solutions and partial extraction from albumin samples. Spintip C18 devices provided exhaustive extraction from buffer and albumin samples. Spintip C8 devices offered complete extraction from buffer and approximately 50% recovery from albumin samples. Centrifree ultrafiltration devices showed high recovery of free concentrations from all the samples, while Amicon and Nanosep devices provided partial recovery.

Conclusion:

Spintip C8 and Centrifree devices proved useful for measuring free concentrations.

Understanding drug–protein binding and its influence on the levels of unbound drugs in plasma is necessary in preclinical animal studies and is subsequently used for the prediction of efficacy and toxicity in humans. Drug–protein binding relationships also help calculate a drug’s diffusion into tissues, which is necessary to interpret bioavailability and calculate hepatic and total body clearance [1,2]. Albumin and alpha 1-acid glycoprotein are the main plasma-binding proteins, but most small molecules will also bind to other proteins [1,3].

The fraction of drugs bound to plasma proteins is often approximated to be constant, and most drug information databases report one single value instead of a range. However, it has been found that the in vivo binding constants of some drugs to serum proteins differ depending on gene polymorphism, post-translational modifications of carrier proteins, the presence of metabolites and toxins, and patient age groups [4]. Furthermore, for some compounds, administration in rapid infusions produces higher free concentrations than those reported for oral administration [5].

Coadministration of other drugs, infusion rate and patient demographics can significantly alter free concentration values [5–7]. Because of these choices that can be made in drug therapy, measurement of unbound drugs in clinical practice is critical to adequately inform dosing decisions. Sometimes, altered plasma protein levels may necessitate the use of normalizing calculations to provide accurate therapeutic drug-monitoring data for patient care [8].

Compounds that experience different binding based on sex, age and disease state have more variability in free concentrations and should be monitored more frequently in clinical practice [2,3,5,6]. Despite observations that variable free fractions can be found even in normal patient populations, reliable methods for measuring free concentrations remain limited to research labs due to time constraints and lack of specific instrumentation [3,9].

Measurement of free concentrations

The most common and easy to use methods for measuring free concentrations are based either on separating unbound small molecules across a semipermeable membrane or on selectively extracting the free drug molecules from the sample. Current analytical methods for determining free concentrations have been recently reviewed [10]. One of the most recent methods is based on electromembrane extraction [11].

For membrane-based separation techniques, free drugs will cross the membrane toward an empty compartment or toward a blank buffer solution, while binding proteins will remain in the sample. Subsequently, the solution on the side of the membrane opposite to the sample is collected and analyzed for free concentration. Of these techniques, equilibrium dialysis is considered the most reliable method, while ultrafiltration is the easiest to use and is applied most commonly in practice. It was found that good reproducibility is attainable using ultrafiltration as long as physiological conditions are maintained and low centrifugal forces are used [12]. Other studies confirmed that free concentration measurements by ultrafiltration differ greatly depending on membrane material, molecular weight cutoff and centrifugation conditions [4,13]. Nevertheless, with properly controlled experimental conditions, ultrafiltration is a time-efficient and straightforward method that is very useful in clinical laboratories.

Extraction-based techniques can only work if they minimally or partially disturb the drug–protein-binding equilibrium. Therefore, exhaustive extraction methods such as solid-phase extraction in a cartridge format cannot be used to measure free concentrations. Alternatively, partial extraction methods such as liquid–liquid extraction (LLE) or solid-phase microextraction (SPME) are suitable. LLE is the simplest and most affordable approach but is difficult to automate. SPME is a method that has been used to determine binding constants, free concentrations and plasma protein binding in whole blood, plasma and buffers where only a drug and a specific protein are interacting. The binding parameters determined by SPME were found to correlate well to values found using equilibrium dialysis. SPME is a faster method that does not require large amounts of sample solution, has the potential to be automated and can be used to determine free concentrations both in vivo and in vitro [1–3,14].

Target compounds

Two compounds for which the importance of free concentration monitoring has been acknowledged were chosen for this research: testosterone and phenytoin.

Clinically, free and total concentrations of testosterone can be used as markers for significant abnormalities in androgen levels, but free testosterone is more useful in cases of hypo- or hyperthyroidism, liver cirrhosis, obesity or exogenous sex hormone use [7]. When using only total testosterone concentrations to diagnose, results may lose accuracy since binding protein concentration changes from aging disproportionately impact free testosterone levels versus total levels [7]. The total concentration of testosterone is more often monitored in clinical practice to reduce costs, while the more useful free testosterone testing is usually only done in larger reference laboratories that use equilibrium dialysis, ultracentrifugation, radioimmunoassay and calculation-based methods.

Phenytoin is one of the first compounds for which the importance of plasma protein binding in clinical practice was acknowledged. The free fraction of phenytoin changes with age, disease state and the presence of other drugs. For example, coadministration of phenytoin with valproate in patients with hypoalbuminemia was found to significantly increase the free fraction of phenytoin [6]. Since only unbound (free) phenytoin is considered pharmacologically active, free levels require close monitoring and are sometimes obtained by using total concentration measurements and calculations [15].

This present research focused on several methods and devices for measuring the free concentrations of testosterone and phenytoin with the purpose of identifying reliable methods that are easy to use.

Materials & methods

Chemicals & materials

Ammonium fluoride (mass spectrometry grade), methanoic acid, ethyl nitrile for gradient elution, methyl alcohol for gradient elution, physiological phosphate-buffered saline (PBS), testosterone, phenytoin, testosterone-D₃, phenytoin-D₁₀, human albumin, Tween 20, Tween 80, Centrifree centrifugal filters (regenerated cellulose membrane; 30,000 NMWL) and Amicon centrifugal filters (low-binding regenerated cellulose; 10,000 and 30,000 NMWL) were purchased from Millipore-Sigma (MA, USA). Nanosep centrifugal filters (polyethersulfone ultrafiltration membrane; 10,000 and 30,000 NMWL) devices were purchased from Pall Corporation (MI, USA). Microscale solid-phase extraction ‘Spintips’ (Hypersep biobasic; C18 extraction phase and C8 extraction phase), and microscale solid-phase extraction ‘Tips’ (Hypersep biobasic; C18 extraction phase) were bought from Thermo Scientific (MA, USA).

Samples & standard solutions

Human serum albumin (HSA) test solutions were prepared by mixing commercially available albumin with physiological phosphate buffer to a concentration of 44 g/l, corresponding to the average normal concentration in human serum. Sample solutions of the target compounds in HSA were prepared by spiking testosterone and phenytoin in HSA 44 g/l. Sample solutions of the target compounds in PBS were prepared by spiking testosterone and phenytoin in PBS. Care was taken to ensure the concentration of organic solvent was below 0.5% in all these solutions. Experiments were repeated in triplicates. The analytes and internal standards were obtained as 1 ng/μl solutions; the LC–MS system was calibrated using solutions of analyte ranging from 0.02 to 20 pg/μl for testosterone and 0.2 to 100 pg/μl for phenytoin. All samples and standard solutions were spiked with testosterone-D₃ at 5 pg/μl and phenytoin-D₁₀ at 50 pg/μl as internal standards.

LLE procedure

Solutions of testosterone at a concentration of 10 pg/μl and phenytoin at a concentration of 50 pg/μl in PBS or in HSA 44 g/l were extracted using three organic solvents: dibutyl phthalate, 1-octanol and methyl formate. Equal volumes of analyte solution and organic solvent (0.2 ml each) were mixed for 1 h on a shaker at 1000 rpm, followed by centrifugation for 10 min at 10,000 rcf. Finally, 75 μl aliquots of organic phase were collected, spiked with internal standard and analyzed by LC–MS.

Microextraction by Spintip

The same experimental steps were used for both Spintips based on C18 and C8 extraction phases.

Conditioning

Microextraction Spintips were set on vendor-supplied connecters in centrifugation tubes after being tapped lightly on a clean surface to settle all solid-phase particles. They were subsequently washed with 50 μl of solution with high organic solvent composition (release solution; 90:10 ACN:H₂O with 0.1% methanoic acid added) and centrifuged for 1 min at 20,000 rcf (used for all Spintip centrifugations) followed by 1 min centrifugation with 50 μl of PBS. The wash solutions were thrown out, and the centrifuge tubes for eluate collection were replaced with new ones.

Sample extraction

Volumes of 0.200 ml of sample (solutions of testosterone at a concentration of 10 pg/μl and phenytoin at a concentration of 50 pg/μl in PBS or HSA 44 g/l) were transferred into Spintips and centrifuged for 2 min. When using microextraction in Spintip format, there is no need to wait for equilibration, since passing the sample through the bed of packed particles results in equilibration. After centrifugation, the samples from the centrifuge tubes were aspirated, placed back in the tips and centrifuged again for 2 min for a total of 1–5 passes. After extracting the samples, the tips are washed with 0.200 ml of water and centrifuged for 2 min, with the procedure repeated for a total of three washes; water was discarded after each wash and the centrifuge tubes were replaced afterward.

Sample desorption

For the desorption of analytes, 0.100 ml of the release solution was added to the tips followed by 1 min of centrifugation. Finally, 75 μl aliquots of release solution were collected, spiked with internal standard and analyzed by LC–MS.

Microextraction by Tip

Conditioning

Tips were attached to 100 μl pipettes; 50 μl of release solution was aspirated and expelled for a total of five-times; finally, 50 μl of PBS was aspirated and expelled for a total of three times.

Sample extraction

Volumes of 0.100 ml of sample (solutions of testosterone at a concentration of 10 pg/μl and phenytoin at a concentration of 50 pg/μl in PBS or HSA 44 g/l) were aspirated and expelled from the extraction tips for a total number of either 5, 10, 15, 20, 25, 50 or 100 cycles. After extraction, the tips were washed by aspirating and expelling 0.100 ml of water for a total of five cycles (with fresh water for each cycle).

Sample desorption

Three aliquots of 25 μl of release solution were aspirated and expelled from the extraction tips. Finally, the pooled release solution was spiked with internal standard and analyzed by LC–MS.

Ultrafiltration

Five types of centrifugal filtering devices from three manufacturers were used to separate free testosterone and phenytoin from samples. For Amicon and Nanosep devices, 0.4 ml of sample was used; for Centrifree devices, 0.8 ml of sample was used. The centrifugal filtering devices were either used as supplied by the manufacturer or after they were passivated by prewashing with 0.3 ml of solution of 5% Tween 80 or Tween 20 in water. After adding the samples (or the passivation solution) to the centrifugal devices, they were spun at 2000 rcf for 30 min. Finally, 75 μl aliquots of ultrafiltrate were collected, spiked with internal standard and analyzed by LC–MS.

Analysis by LC–MS

A Shimadzu Nexera XR liquid chromatograph equipped with a binary pump, temperature-controlled autosampler (set at 4°C) and temperature-controlled column oven (set at 50°C) was coupled with an AB Sciex QTRAP 5500 triple quadrupole mass spectrometer, which was used in multiple reaction monitoring (MRM) positive ion mode to scan for analyte transitions: 289.2–97.0 and 289.2–108.9 for testosterone; 292.2–97.0 for testosterone-D₃, 253.1–182.1 and 253.1–104.0 for phenytoin; and 263.1–192.1 for phenytoin-D₁₀. The ion spray voltage was set to 5.5 kV and the interface temperature to 500°C. Nitrogen was used as nebulizing gas. Analyst 1.7.1 software from Sciex was used for data acquisition and processing.

For analyte separation, a Poroshell C18 chromatographic column from InfinityLab was used with gradient elution at a total flow rate of 0.6 ml/min. Samples were eluted with an aqueous phase (A) consisting of 300 μM ammonium fluoride and 0.01% methanoic acid in water and an organic phase (B) consisting of HPLC-grade methyl alcohol. The gradient started at 10% B, was increased to 95% B over 13 min, maintained at 95% B until 17 min and restored to 10% B until 18 min; 2 more min of equilibration time at the initial gradient conditions were allowed before the next injection. The automated injection needle was washed with a mixture of 1:1 water:methyl alcohol prior to and following sample introduction into the LC–MS.

Results & discussion

The purpose of the current study was to find the best methods and devices for measuring free concentrations of testosterone and phenytoin by testing approaches such as LLE, SPME and ultrafiltration. In the case of extraction-based methods, good reproducibility and proof of equilibrium-based partial extraction from albumin solutions were sought (showing that there is still an equilibrium between analyte and binding protein after extraction). After extraction, the free concentration in the sample were determined as previously shown [16]. In the case of ultrafiltration-based methods, which separate the free concentration directly, good reproducibility and high recovery of free concentrations were sought.

Liquid-liquid extraction

Using LLE, testosterone was fully (close to 100%) recovered from PBS using dibutyl phthalate and around 50% recovered from solutions in HSA, which more closely represents biological samples. Extractions using 1-octanol and methyl formate recovered significantly less testosterone (<40%) from both PBS and HSA solutions. Phenytoin was almost fully recovered from PBS using dibutyl phthalate and 1-octanol; around 60% of the phenytoin was recovered from solutions containing HSA. Recovery of phenytoin with methyl formate was less than 40% for solutions in PBS or HSA (Figure 1). For all LLE experiments, variability was high at up to 25%.

Figure 1. . Recovery of total concentrations obtained with liquid–liquid extraction: the percentage of testosterone (10 pg/μl) and phenytoin (50 pg/μl) recovered by liquid extraction with three solvents.

Extractions are from solutions in PBS and HSA. Error bars represent standard deviations; n = 3.

HSA: Human serum albumin; PBS: Phosphate-buffered saline.

For accurate measurement of free concentrations, the recovery from PBS – where there are no proteins, should be close to 100%. On the other hand, recovery from HSA solutions should be partial, since the compounds bound to proteins should not be exhaustively extracted, allowing the method to be sensitive to free concentrations. Considering this, extraction with dibutyl phthalate would be suitable for measurement of free concentrations of both testosterone and phenytoin; 1-octanol would potentially be acceptable only for free phenytoin determination. All the other combinations tested here were found to be unsuitable for measuring free concentrations. Even for the acceptable combinations, reproducibility was poor.

Microextraction

Microextraction devices need time in contact with the sample or multiple passes of the sample through the device in order to equilibrate with the analyte. For Spintip devices, a number of passes of the sample through the device from 1 to 5 were tested; for Tip devices, a total number of 5, 10, 15, 20, 25, 50 and 100 passes were tested. For Spintip devices, equilibrium was reached after three passes (Figure 2 for C18 devices; the graph looks almost identical for C8 devices); for Tip devices, 25 passes were sufficient to reach equilibrium, but the variability was much higher than for Spintip devices (data not shown). For Spintip devices, the sample has to pass through a tall bed of particles, thus the manufacturer recommends one sample pass. For the target analytes in this study, it was found that more passes are needed to reach equilibrium; this is most likely due to the fact that the sample volume used in this study, 200 μl, was much higher than the volume of particles in the device, 20 μl. The larger sample volume was needed to achieve partial extraction from HSA solutions. In the case of Tip devices, the sample flows past a wall with extractive particles glued to the side of it, thus the manufacturer recommends 25 passes through the pipette tip. In this study, the recommended number of sample passes was found to be sufficient; a larger number of passes did not increase the amount extracted and only resulted in higher standard deviations for repeat analyses. Furthermore, the Tip devices were found to be difficult to use, since they had to be handled manually one at a time (or several at a time with a multichannel pipette), while many more Spintip devices could easily be centrifuged in parallel.

Figure 2. . Extraction profile for testosterone (10 pg/μl) and phenytoin (50 pg/μl) using Spintip C18 devices; the amount extracted (n) divided by the amount extracted at equilibrium (n0) was plotted against the number of passes of the sample through the Spintip C18.

Extraction of testosterone and phenytoin from PBS and HSA solutions using Spintip C18 devices was above 90% with very small deviations between repeat samples (<5%) (Figure 3). Using Tip C18 devices, only testosterone extraction from PBS was achieved with reasonable recovery (∼95%), albeit, with large variability between repeat samples; the Tip devices only recovered less than 5% of the total testosterone from HSA solutions. Extraction of phenytoin was less than 40% from PBS and around 2% from HSA solutions when using Tip devices. When using Spintip C8 devices, testosterone and phenytoin were around 95% recovered from PBS solutions with good reproducibility; extractions from HSA solutions were also very reproducible with about 50% recovery (Figure 3).

Figure 3. . Recovery of total concentrations obtained with microextraction devices: the percentage of testosterone (10 pg/μl) and phenytoin (50 pg/μl) recovered from phosphate-buffered saline and human serum albumin solutions using Spintip C18, Tip C18 and Spintip C8 extraction devices.

Error bars represent standard deviations; n = 3.

HSA: Human serum albumin; PBS: Phosphate-buffered saline.

Extractions with Spintip C18, although highly reproducible, proved impractical for measuring free concentrations, since the recovery from HSA solutions was very high. Extraction of testosterone with Tip C18 would potentially be useful for measuring free concentrations, but the variability was quite high. The free concentration of phenytoin cannot be measured with Tip C18, since the recovery from PBS solutions was too low, below 40%. Conversely, sample preparation with Spintip C8 devices proved well suited for measuring the free concentrations of both compounds, since recovery from PBS was above 90% and recovery from HSA solutions was around 50% – all with good reproducibility.

Overall, in terms of higher recovery of both analytes and better reproducibility, Spintip devices eclipsed Tips in performance. It is likely that the better reproducibility of Spintip experiments was due to the sample flowing in a consistent way through the bed with extractive particles.

Ultrafiltration

Ultrafiltration devices separate the free fraction of a compound as a solution that is removed from the bulk of the sample across a semipermeable membrane with a specific molecular-weight cutoff, which should be lower than the molecular weight of the proteins binding the target compound. With the molecular weight of HSA being around 67 kDa, all the ultrafiltration devices that were used in this study were suitable: Amicon (10 and 30 k), Nanosep (10 and 30 k) and Centrifee (30 k). All the devices were used according to manufacturer instructions.

The recovery of the target analytes from PBS solutions was obtained by comparison to the prepared total concentration (which, in PBS solutions, is equal to the free concentration). For HSA solutions, the recovery was obtained by comparison to the calculated free concentration [16].

Centrifree filters were found to be the most expensive, most difficult to fill with sample, and required the most sample volume compared with the other ultrafiltration devices. On the other hand, all the other devices showed poor recovery. Furthermore, since the same sample solutions were tested with all the devices, the filtrates obtained with Centrifree clearly had the highest concentrations of target analytes, which also corresponded to the free concentrations (Figure 4).

Figure 4. . Recovery of free concentrations with ultrafiltration devices: the percent of testosterone (10 pg/μl) and phenytoin (50 pg/μl) recovered from phosphate-buffered saline and human serum albumin solutions after centrifugation through five filters with various membranes and molecular weight cutoffs (k = kDa).

Error bars represent standard deviations; n = 3.

HSA: Human serum albumin; PBS: Phosphate-buffered saline.

As Figure 4 shows, the recovery of free testosterone and phenytoin using Centrifree devices was above 90% from PBS solutions and around 100% from HSA solutions. Recoveries from HSA solutions were expected to be higher since the presence of proteins prevents some of the losses by nonspecific adsorption to the device. In contrast, all the other devices showed very low recoveries from both PBS and HSA samples with most recoveries below 50%. When using PBS solutions, Amicon devices adsorbed almost all of the testosterone, while Nanosep devices allowed less than 20% of the testosterone to pass through the membrane (Figure 4).

To improve the recovery obtained with Amicon and Nanosep devices, which are easier to use and more affordable, passivation against adsorption was attempted. It has been suggested [17] that ultrafiltration devices can be ‘passivated’ against adsorption by prewashing them with Tween 80. When trying to use Tween 80, it was found that the wash solution needed a very long centrifugation time to pass through the filter and, even then, some liquid remained, so washing was done with Tween 20 instead, which passed easily through the filters. When devices were washed with 5% Tween 20 in water prior to analysis in an attempt to reduce analyte adhesion to the devices and membranes, recoveries unexpectedly decreased for both analytes when using Amicon and Centrifree devices; only the recoveries obtained with Nanosep devices increased after washing, though not enough to make them useful for reliable separation of free concentrations (Figure 5). Nevertheless, the recovery of phenytoin with washed Nanosep 30 k reached the same level as the recovery obtained with washed Centrifree. A potential explanation of these results may be that the washing step removes the conditioning and humectant solutions left on the membranes of Amicon and Centrifree devices during manufacturing which, subsequently, increased the adsorption of analytes on the devices.

Figure 5. . Comparison between washed and unwashed ultrafiltration devices: the percent of free testosterone (10 pg/μl) and free phenytoin (50 pg/μl) recovered from phosphate-buffered saline solutions after centrifugation through unwashed and Tween 20 (5%) washed filters.

Error bars represent standard deviations; n = 3.

The data in Figures 4 & 5 support the use of unwashed Centrifree devices for determining the free concentrations of testosterone and phenytoin, since the recovery or free concentrations from both PBS and HSA samples is close to 100%.

Conclusion

These results support the use of Spintips C8 and Centerfire 30 k devices as the most accurate and reproducible methods to experimentally measure free concentrations of testosterone and phenytoin. LLE had the poorest performance with low recoveries and high variability and is not recommended for use. Nevertheless, extraction with dibutyl phthalate could potentially be used if an approach to improve reproducibility is found. All Spintip microextraction devices provided good reproducibility, but only C8 devices were sensitive to free concentrations. Tip C18 microextraction devices proved unreliable regarding both reproducibility and recovery. Overall, SPME proved to be an easy-to-use and dependable procedure that was successfully used in Spintip format. For ultrafiltration, although a wide variety of devices were tested, only the cumbersome Centrifree devices proved useful for this application.

The current preliminary research was primarily used to determine which methods for measuring free concentrations should be selected for further testing on complex biological samples. The methods identified with this research as being suitable for PBS and HSA samples can also directly be used to measure the binding constants of drugs to albumin.

Summary points.

Background

• Measurement of free concentrations is important in all life sciences.

• However, it is difficult to accomplish, with many methods being slow or unreliable.

• Free concentrations are particularly important for the biological activity of compounds such as testosterone and phenytoin.

Methodology

• Three different approaches were investigated regarding their suitability for measuring the free concentrations of testosterone and phenytoin: liquid–liquid extraction (LLE), microextraction in spintip and tip format, and ultrafiltration.

• For each approach, several devices and extraction phases were investigated.

• LLE was performed using dibutyl phthalate, 1-octanol and methyl formate.

• Microextraction was performed using Spintips with C8 and C18 stationary phases, as well as tips with C18 phase.

• Ultrafiltration was performed using Amicon (10 and 30 k), Nanosep (10 and 30 k) and Centrifree (30 k) devices.

• For this preliminary work, solutions of the compounds in human serum albumin were investigated.

Results

• LLE with dibutyl phthalate provided complete extraction from buffer solutions and partial extraction from albumin samples.

• Spintip C18 devices provided exhaustive extraction from buffer and albumin samples.

• Spintip C8 devices offered complete extraction from buffer and approximately 50% recovery from albumin samples.

• Centrifree ultrafiltration devices showed high recovery of free concentrations from all the samples.

• Amicon and Nanosep devices provided partial recovery.

Conclusion

• Spintip C8 and Centrifree devices proved the most useful for measuring free concentrations of testosterone and phenytoin.