Method for Simultaneous Determination of Free Concentration, Total Concentration, and Plasma Binding Capacity in Clinical Samples

Abstract

Most quantitative research methods are based on measuring either the total or the free concentration of an analyte in a sample. However, this is often insufficient for the study of complex biological systems. The main objective of this research was to develop new methods for providing more information from samples: the free concentration (C_f), the total concentration (C_t), and the plasma binding capacity (PBC). Samples were processed using microextraction and ultrafiltration. For each of these techniques, two quantification procedures were used: addition of isotopically labeled standard and repeated analysis of the same sample. The new methods were validated by analyzing clinical samples and samples with known concentrations. Methods based on addition of labeled compound were found to be the fastest, and most reproducible. For analysis of clinical samples, methods based on microextraction were more sensitive and more accurate than those based on ultrafiltration. For analysis of pooled plasma samples, the overall accuracy of all approaches to determine PBC, testosterone C_f, and testosterone C_t was between 94–109%, 87–113%, and 94–122% respectively. The new approach goes beyond a simple concentration measurement, giving more information from clinical samples, with great potential for personalizing drug dosage and therapy to the needs of individual patients.

Introduction

Many scientific disciplines have been relying for centuries on measuring the total concentration of bioactive compounds in biological samples. While the measurement of concentrations is the end goal of analytical science, research in pharmacology uses drug concentrations to find relationships with drug effects.¹ Although pharmacokinetics (PK) and pharmacodynamics (PD) have been significantly refined over the years, no progress has been made regarding the type of information obtained from biological samples. Total drug concentrations, as currently measured, do not provide any information regarding the interaction of the analyte with other sample components. As an alternative to total concentrations, it has been shown for many drugs that free concentrations provide better correlations with drug effects. However, measuring only free concentrations can be insufficient for clinical samples.

Determining the amount of drug that binds to plasma proteins is important for drug discovery and development. Understanding these interactions is essential to conceptualizing the PK-PD of a drug. Thus, plasma protein binding (PPB) is an important factor in assessing drug disposition, efficacy, and safety.^2–4 Determining the PPB of drugs is done using protein-binding assays and is crucial for characterizing distribution into compartments.² PPB can change drastically given certain disease states or genetic backgrounds and nonlinear protein binding can occur even at low drug concentrations, diminishing the usefulness of using total concentrations in therapeutic drug monitoring.^4–9 Free concentration and PPB are necessary parameters in pre-clinical and clinical use. Preclinically, unbound (free) drug concentrations are important for PK-PD equations and modeling,⁴ while PPB is used for preliminary human trials to find the efficient free concentration at target tissues.³

Despite a better correlation between free concentrations and pharmacological effects, total concentrations are considered useful when the ratio of free to total drug does not differ between individuals. However, this constant free drug fraction is an assumption that disagrees with the numerous observations of inter-individual variability in protein binding. Since there are limited resources to address individual plasma protein binding and measurements of free concentration, pharmacotherapy providers mainly ignore this variability and fail to consider free concentrations when adjusting drug dosage regimens.^4,10 Furthermore, total and free concentrations alone are not enough for drug monitoring since the concentration of binding proteins influences drug disposition and the relationship between free and total concentration.^9–13 Documenting the relationship between drugs and the proteins they bind with are significant to clinical PK and PD.^4,10

Most current methods for measuring free concentrations lack precision and accuracy⁴ while also being more time and resource-consuming than methods for total concentrations. Total concentrations are usually measured in clinical practice using immunoassays on minimally processed samples, while free concentration measurements require more extensive sample preparation and sensitive instruments.¹⁰ Most methods for measuring free compounds involve the separation of free fractions using equilibrium dialysis, ultrafiltration, ultracentrifugation, or gel filtration, followed by subsequent analysis with quantitative methods (i.e. LC-MS, spectrophotometry).^2–4,14 These methods can be time-consuming, result in analyte loss to membranes, and produce errors from protein leakage or binding equilibrium shifts.^2–4 Commonly regarded as the golden standard for monitoring free drug concentrations, equilibrium dialysis is based on the diffusion of drugs across a semipermeable membrane that separates the sample and a buffer solution.^3,4,7,8 The challenges with this method come from long incubation periods (6 hours – 2 days), compounds not soluble in water, volume changes due to changes in osmotic pressure, and nonspecific adsorption to membranes.^3,4,14 Solid phase microextraction (SPME) is a technique that involves exposing a solid support containing a small amount of extracting phase to a sample until or before equilibrium is reached. SPME is a simple and reliable method that also allows for the collection of supplementary parameters such as binding constants and free concentrations.^2–4,15,16

Ultrafiltration (UF) involves the centrifugation of samples that have reached drug-protein binding equilibrium such that the hydrostatic pressure pushes fluid from samples through a filter with a certain molecular weight cut-off that prevents the movement of proteins. UF methods are the most commonly used in clinical practice for obtaining free concentrations^4,10 because they are simple and fast, but they can be susceptible to high variability due to changes in experimental conditions.^5,17,18

Methodology and Reference Compounds.

The current paper proposes a model for simultaneously measuring free and total concentrations along with plasma binding capacity, that addresses the need for simpler methods of free compound separation and quantification that could be implemented in-house while also accounting for inter-patient variability in binding protein concentration. This model will be based on the separation of free fractions using modified versions of microextraction – offering high sensitivity and better analytical signal, and ultrafiltration – fast, simple, and easy to use. To demonstrate the validity of the new approach, testosterone and phenytoin will be used as reference compounds. Testosterone is an important androgen present in all people and its levels are monitored as markers for several disease states. Normal levels, in men, range between 2.7 and 10.7 ng/mL. Testosterone binds primarily to human serum albumin (HSA) and sex-hormone binding globulin (SHBG), with a cumulative binding of 97.0 – 99.5%. Most academic and community hospitals perform in-house testing of total testosterone levels, but fewer offer free testosterone level testing. Despite equilibrium dialysis being considered a clinical gold standard for free testosterone measurements, most reference labs will not use this method unless requested due to the difficulty and extensive runtime.^{16,17,19–21} Phenytoin is an anticonvulsant drug highly bound to albumin (approximately 90%) and needs to be monitored for appropriate dosing. Due to its narrow therapeutic index, drug efficacy and safety can change with altered plasma protein binding. Dosing becomes difficult in cases where disease states result in changing albumin concentrations or altered protein binding capacity. Although measured free phenytoin concentrations are more correlated to clinical outcomes and reduced toxicity, many institutions do not offer this monitoring due to costs or time-consuming methods. Plasma protein concentrations are necessary for calculating normalized concentrations but these calculated values still do not correlate well with effects.^11,12,22

Accordingly, there is a need to develop new methods and calculations that would eliminate the requirement for protein concentrations. Instead of the concentrations of particular proteins, patient-individual plasma binding capacity (PBC) could be determined to more accurately reflect the equilibrium between the compound and its binding proteins in plasma while also accurately measuring free compound levels. The main objective of this research is to provide more information from the sample, by developing for the first time a new validated method for simultaneously measuring the free concentration, the total concentration, and the plasma binding capacity in a sample.

Analytical Approaches for Comprehensive Analysis

The determination of free concentrations, total concentrations, and plasma binding capacity can be achieved using any analytical method that is sensitive to free concentrations. In order to obtain several parameters from the same sample, multiple data points are needed. These supplementary data points can be obtained either by spiking the sample with an isotopically labeled analyte or by processing the same sample several times at different analyte concentrations; the analyte concentration in the sample can be changed either by processing the sample or by adding more unlabeled analyte to it.

The proposed calculations and practical approaches work with all small molecules that bind to plasma proteins. Equations for comprehensive sample analysis were developed for two sample preparation techniques (microextraction and ultrafiltration) and two experimental procedures (addition of labeled analyte and sequential analysis) for a total of four approaches. These approaches should also work with any analytical procedure that can measure free concentrations. To measure the concentration of both analyte and isotopically labeled analyte, mass spectrometry-based approaches are best. In this case, for good quantitative results, a second isotopically labeled analyte should be employed as an internal standard to compensate for changes in injection volumes and ionization efficiency.

The binding of the test analytes, testosterone and phenytoin, to plasma proteins matches a 1:1 linear binding model – which was used for this study. The values of the binding constants were obtained from Hackbarth et al.⁹ The equilibrium between bound and unbound compound is described by Equation 1.^1,6,23,24 The association constant, K, is the ratio between bound compound in complex with protein [CP] and free compound [C] and free protein concentration [P] (Equation 2):

$$[C]+[P]\rightleftharpoons [CP]$$ Equation 1

$$K=\frac{[CP]}{[C][P]}$$ Equation 2

Since the concentration of bound compound is equivalent to the difference between total compound concentration (C_t) and free compound concentration (C_f) after equilibrium is reached, Equation 2 can be rewritten as:

$$K=\frac{C_t-C_f}{C_f[P]}$$ Equation 3

In the case of linear protein binding, the free concentration of protein can be considered almost the same as the total concentration of protein. Subsequently, Equation 3 can be rearranged to calculate the free fraction of analyte as

$$\frac{C_f}{C_t}=\frac{1}{1+K[P]}=\frac{1}{1+PBC}$$ Equation 4

The plasma binding capacity (PBC) is then defined as the product of the association constant (K) and the concentration of protein in the sample. Equation 4 describes the relationship between the three parameters to be measured in plasma samples: the free concentration, total concentration, and plasma binding capacity.

Comprehensive Analysis using Microextraction

In solutions at equilibrium with a microextraction device, the amount m_e (moles) of compound extracted will be proportional with the free compound concentration. The free concentration of compound after extraction (C_fe) in solution is then described by:²

$$C_{fe}=\frac{m_e}{e_c}$$ Equation 5

Where e_c is the extraction constant, which depends on the amount of extraction phase and its affinity for the analyte.² The e_c can be found experimentally by extracting the investigated compound from a standard solution with known concentration; if this solution does not contain any binding proteins, then the free concentration of analyte is equal to the total concentration. If C₀ is the initial analyte concentration before extraction from a standard solution with volume V_s, then the concentration of compound after extraction can be found with:

$$C_{fe}=C_0-\frac{m_e}{V_s}$$ Equation 6

The extraction constant can then be obtained by measuring the amount of analyte extracted with a suitable method and combining Equation 5 and Equation 6:

$$e_c=\frac{m_e}{C_{fe}}=\frac{m_e{V}_s}{C_0{V}_s-m_e}$$ Equation 7

Approach 1: Addition of Labeled Analyte

In this case, an isotopically labeled analogue of the target compound needs to be added to the unknown sample. Subsequently, the sample containing the label is equilibrated with a microextraction device and the masses of analyte (m_e) and labeled analyte (m_eL) extracted are measured. If the volume of desorption solution used for the last step of the microextraction approach is V_d, then the masses extracted can be obtained as

$$m_e=C_{LCMS}{V}_d$$ Equation 8

$$m_{eL}=C_{LCMS-L}{V}_d$$ Equation 9

Where C_LCMS and C_LCMS-L are the concentrations of analyte and labeled analyte determined by LCMS in the desorption solution.

The free concentrations of unlabeled compound, C_fe, and labeled compound, C_feL, remaining in the sample after microextraction can be calculated with Equation 5 as

$$C_{fe}=\frac{m_e}{e_c}$$ Equation 10

$$C_{feL}=\frac{m_{eL}}{e_c}$$ Equation 11

The total concentrations of unlabeled compound, C_te, and labeled compound, C_teL, remaining in the sample after microextraction can be calculated as

$$C_{te}=C_{t1}-\frac{m_e}{V_s}$$ Equation 12

$$C_{teL}=C_{tL}-\frac{m_{eL}}{V_s}$$ Equation 13

Where C_t1 is the initial total concentration of analyte and C_tL is the initial total concentration of added labeled analyte. Finally, the relationship between the free and total concentrations of analyte and labeled analyte before extraction (subscript “1”) and after extraction (subscript “e”) can be described as

$$\frac{C_{f1}}{C_{t1}}=\frac{C_{fe}}{C_{te}}=\frac{C_{feL}}{C_{teL}}=\frac{1}{1+PBC}$$ Equation 14

All the equations for approach 1 above form a system that can be solved for the unknowns initial total concentration C_t1, plasma binding capacity PBC, and initial free concentration C_f1:

$$C_{t1}=\frac{C_{tL}{m}_e}{m_{eL}}$$ Equation 15

$$PBC=C_{tL}\frac{e_c}{m_{eL}}-\frac{e_c}{V_{sample}}-1$$ Equation 16

Using the calculated values from Equation 15 and Equation 16, C_f1 can quickly be calculated with:

$$C_{f1}=\frac{C_{t1}}{PBC+1}$$ Equation 17

Approach 2: Repeated Analysis of the Sample

With this approach, after the initial unlabeled sample is extracted, the left-over sample containing unlabeled analytes is analyzed a second time using the same extraction method and the additional compound extracted by the device is measured. This method will be referred to as “multi-analysis,” for clarity. For best accuracy, the two measurements should provide significantly different results (significantly different amounts extracted), which could be achieved by ensuring that each extraction removes a significant amount of analyte from the sample or by adding a known amount of unlabeled analyte to the sample after the first analysis. This approach is a bit more time-consuming, but it does not require isotopically labeled analytes or analysis by mass spectroscopy thus rendering it a less expensive and more versatile option for comprehensive analysis.

For the first extraction of the sample, the relationship between free concentration after extraction (C_fe1), the amount of analyte from the first extraction (m₁), initial total concentration (C_t1), total concentration remaining after extraction (C_te1), and PBC can be described using Equation 10, Equation 12, and Equation 4.

When additional unlabeled compound is added to the sample (m_add), the total concentration after the addition (C_t2) would be equal to the total concentration remaining after the first extraction plus the concentration of added compound:

$$C_{t2}=C_{te1}+\frac{m_{add}}{V_s}$$ Equation 18

The free concentration after the second extraction will be calculated from the amount of analyte extracted from the sample the second time (m₂), which is obtained from the concentration of analyte determined by LCMS (C_LCMS2) in the desorption solution (V_d):

$$C_{fe2}=\frac{m_2}{e_c}=\frac{C_{LCMS2}{V}_d}{e_c}$$ Equation 19

The total concentration remaining in the sample after the second extraction, C_te2, will be obtained by considering the removal of m₂ moles of analyte in the second extraction:

$$C_{te2}=C_{t2}-\frac{m_2}{V_s}$$ Equation 20

The relationship between the free concentration, total concentration, and the PBC after the second extraction from the sample can be described as:

$$\frac{C_{fe2}}{C_{te2}}=\frac{1}{1+PBC}$$ Equation 21

The total concentration of the initial sample (C_t1) and the PBC can then be calculated from the system of equations above as:

$$C_{t1}=\frac{m_1\left(m_1-m_{add}\right)}{V_s\left(m_1-m_2\right)}$$ Equation 22

$$PBC=\frac{e_c\left(m_2-m_{add}\right)}{V_s\left(m_1-m_2\right)}-1$$ Equation 23

These values will then be used to determine the concentration of free compound in the initial sample:

$$C_{f1}=\frac{C_{t1}}{1+PBC}$$ Equation 24

Comprehensive Analysis using Ultrafiltration

When ultrafiltration is used for sample preparation, the concentration of analyte in the ultrafiltrate is equal to the free concentration in the original sample. Compared to microextraction, this results in simpler equations as the free concentration is measured directly.

Approach 1: Addition of Labeled Analyte

For this approach, the test sample (C_t1) needs to be spiked with a known amount of isotopically labeled target analyte (C_tL) and then subjected to ultrafiltration. Subsequently, the free concentration of analyte, C_f1, and the free concentration of labeled analyte, C_fL, are both obtained by quantitative analysis of the ultrafiltrate. The relationship between the free and total fractions of labeled and unlabeled compounds can be described similarly to Equation 4, as long as the concentrations are in the linear binding range:

$$\frac{C_{f1}}{C_{t1}}=\frac{1}{1+PBC}$$ Equation 25

$$\frac{C_{fL}}{C_{tL}}=\frac{1}{1+PBC}$$ Equation 26

The system formed by Equation 25 and Equation 26 can be solved to determine both C_t1 and PBC:

$$C_{t1}=\frac{C_{f1}{C}_{tL}}{C_{fL}}$$ Equation 27

$$PBC=\frac{C_{tL}}{C_{fL}}-1$$ Equation 28

Approach 2: Repeated Analysis of the Sample

Multiple data points can be generated by subjecting the same sample containing unlabeled analyte (C_f1, C_t1) to multiple ultrafiltrations. The volume of filtrate (V_f) after each filtration can be replaced by an equal volume of analyte solution (with the known concentration C_add) to maintain initial protein concentration. The total concentration of analyte after filtration and addition is equal to:

$$C_{t2}=C_{t1}-C_{f1}\frac{V_f}{V_S}+C_{add}\frac{V_f}{V_s}$$ Equation 29

Subsequently, the new sample solution is subjected to a second filtration to measure the new free analyte concentration C_f2. The free fractions and their relationship to PBC before and after ultrafiltration can be described similarly to Equation 4 as:

$$\frac{C_{f1}}{C_{t1}}=\frac{1}{1+PBC}$$ Equation 30

$$\frac{C_{f2}}{C_{t2}}=\frac{1}{1+PBC}$$ Equation 31

The free concentrations after the first and second filtrations, C_f1 and C_f2, are obtained by quantitative analysis of the corresponding ultrafiltrates. With measured free concentrations and filtrate volumes, the initial total concentration (C_t1) and PBC can be calculated by solving the system formed by Equation 29, Equation 30, and Equation 31.

$$C_{t1}=\frac{C_{f1}\left(C_{f1}-C_{add}\right)V_f}{\left(C_{f1}-C_{f2}\right)V_s}$$ Equation 32

$$PBC=\frac{\left(C_{f1}-C_{add}\right)V_f}{\left(C_{f1}-C_{f2}\right)V_s}-1$$ Equation 33

Experimental Procedures

Reagents and Materials

Ammonium fluoride (MS grade), formic acid, HPLC grade acetonitrile, HPLC grade methanol, HPLC grade water, phosphate buffered saline (PBS, pH 7.4), testosterone, phenytoin, testosterone-¹³C₃, testosterone-D₃, phenytoin-D₁₀, human serum albumin (HSA), bromocresol purple assay kit for albumin determination, ELISA kit for human SHBG determination, and Centrifree ultrafiltration devices (30 kDa MWCO) were purchased from Millipore-Sigma (Atlanta, GA). Pooled plasma with K2-EDTA as anticoagulant and SpinTips (HyperSep; C18 and C8) were bought from VWR (Radnor, PA).

Samples and Standard Solutions

HSA samples for method development were prepared by diluting HSA in PBS to make standard solutions with concentrations of 11, 22, 44, and 53.2 g/L of HSA, corresponding to the typical in vivo range. Pooled plasma samples from healthy donors were used either unchanged, spiked with testosterone (2.5 to 10 ng/mL, additional to levels in the sample), spiked with phenytoin (10 to 100 ng/mL), or diluted 1:1 and 1:3 with PBS to generate samples with lower plasma binding capacity. All patient blood samples were collected in the morning from consenting adults who were receiving hormone replacement therapy (with testosterone) at a local ambulatory care practice (Albany Medical Center, Division of Community Endocrinology). The blood was collected in vacutainer tubes coated with solid K₂-EDTA to avoid sample dilution and was centrifuged at 2,000 rcf and 4 °C for 10 min to separate the plasma. This study was approved by the Institutional Review Board at the Albany College of Pharmacy and Health Sciences. The study was conducted in accordance with the ethical principles that have their origin in the Declaration of Helsinki. Written informed consent was obtained from all participants prior to conducting any study procedures.

The analytes and internal standards were purchased as 1 mg/mL solutions. Calibration curve solutions were subsequently prepared via serial dilutions to cover a range from 0.027 to 12 ng/mL for testosterone and 0.39 to 100 ng/mL for phenytoin. Internal standard for quantitative analysis was added to all samples before analysis to reach concentrations of 5 ng/mL for testosterone-D₃ and 50 ng/mL for phenytoin-D₁₀.

Quantitative Analysis by LC-MS/MS

A Shimadzu Nexera XR liquid chromatograph equipped with a binary pump, temperature-controlled autosampler (4 °C), and temperature-controlled column oven (50 °C) coupled with an AB Sciex QTRAP 5500 triple quadrupole mass spectrometer was used in MRM mode to analyze compound concentrations. A diverter valve was used to discard undesirable co-eluents (salts and phospholipids) and electrospray ionization in positive mode was scheduled during analyte elution. The interface was maintained at 500 °C, with an ionization voltage of 5,500 V and nitrogen used as nebulization and curtain gas at 40 psi. A Poroshell 120 EC-C18 (Agilent Technologies) column was used for chromatographic separation with gradient elution at a total flow rate of 0.6 mL/min. Samples were eluted with a mobile aqueous phase (A) consisting of 0.3 mM ammonium fluoride and 0.01% HCOOH in HPLC-grade water and a mobile organic phase (B) consisting of HPLC-grade methanol. The gradient was started from 10% mobile phase B at time zero, which was increased to 60% B over 12 minutes and followed by wash with 95% B until 18 minutes. The autosampler was rinsed with a solution of 50:50 methanol in water before and after each injection. The target compounds and the internal standards were quantified using the transitions 289.2→97.0 (quantitation) and 289.2→108.9 (confirmaRon) for testosterone, 292.2→97.0 for testosterone-D₃, 292.3→100 for testosterone-¹³C₃, 253.1→182.1 (quantitation) and 253.1→104.0 (confirmaRon) for phenytoin, and 263.1→192.1 for phenytoin-D₁₀. Analyst 1.7 and MultiQuant 3.0.2 were used for automatic data acquisition and processing.

Analysis of Plasma and HSA Samples

The total concentration of testosterone and phenytoin in plasma and HSA samples was determined using protein precipitation with acetonitrile. A volume of 50 μL of sample was combined with 150 μL acetonitrile containing internal standard. After dilution with the sample, the internal standard concentrations were 5 ng/mL for testosterone-D₃ and 50 ng/mL for phenytoin-D₁₀. The mixture was placed on a vortex for 5 minutes and then placed in a freezer at −20°C for 30 minutes. After the samples were removed from the freezer, the tubes were briefly shaken and then centrifuged for 30 minutes at 20,000 rcf. While avoiding the protein pellet, 75 μL of the resulting supernatant was collected and submitted for quantitative analysis.

Spintip Microextraction Procedure

Before use, all spintips were conditioned. Spintips were placed on tube adaptors in centrifugation tubes after being tapped lightly on a clean surface to settle all solid phase particles. They were then treated with 50 μL of release solution (90:10 ACN:H₂O with 0.1% formic acid) and centrifuged for 1 minute at 20,000 rcf (used for all spintip centrifugations) followed by 1-minute centrifugation with 50 μL of PBS. The solutions that passed through the spintips were discarded and the centrifuge tube replaced.

Addition of Labeled Analyte

For this protocol, all HSA and plasma samples were spiked with testosterone-¹³C₃ to a concentration of 1 ng/mL. Volumes of 200 μL of sample (or calibration level) were transferred into pre-conditioned spintips and centrifuged for 2 minutes. After centrifugation, the liquid in the centrifuge tubes was aspirated, placed back in the tips, and centrifuged for 2 more minutes for a total of three passes. The spintips were washed three times by adding 200 μL water and centrifuging for 2 minutes; the water was discarded after each wash and centrifuge tubes were replaced at the end. For the desorption of analytes, 100 μL of the release solution was added to the spintips followed by 1 minute of centrifugation. 75 μL of the resulting liquid was then placed in vials containing IS and subjected to quantitative analysis.

Repeated Analysis of the Sample

For this procedure, there is an optional step of adding a known amount of analyte to the sample before the second analysis; to achieve this, an additional solution of analyte (V_add) was prepared in PBS with concentrations of 1 ng/mL testosterone and 10 ng/mL phenytoin. After conditioning the spintips, the sample solutions were transferred to spintips in aliquots of 200 μL and centrifuged for 2 minutes. The solution that passed through the spintips was transferred back into the spintips and centrifuged again for 2 minutes for a total of three passes. The remaining sample solutions are set aside for the second analysis. Spintips were washed with 200 μL water and centrifuged for 2 minutes for a total of three washes; the water was discarded after each centrifugation. Desorption was done with 100 μL of the release solution and a 1-minute centrifugation. The extracted solutions are analyzed.

Optional step: if more analyte is to be added to the samples before the second analysis, the V_add solution can be added to all spintips in 200 μL aliquots and centrifuged for 2 minutes. The solutions in the tubes are then transferred back into the tips for a total of three pass-throughs and remaining liquid is discarded.

For the second analysis, the previously set-aside sample solutions were placed back in their corresponding spintip and centrifuged for 2 minutes. The solutions were collected and placed back into the spintips and centrifuged again for 2 minutes for a total of three passes. The spintips were then washed and desorbed as previously described and the desorption solutions were placed into vials for quantitative analysis.

Ultrafiltration Procedure

All ultrafiltration experiments were performed using Centrifree centrifugal filtering devices with a MWCO of 30 kDa, which were centrifuged at 2,000 rcf; the duration of centrifugation was adapted for each sample to obtain enough filtrate for analysis; the free concentration in the ultrafiltrate remains the same regardless of the volume collected; collecting larger volumes of ultrafiltrate is recommended in order to minimize losses to the device.

Addition of Labeled Analyte

For this protocol, all HSA and plasma samples were spiked with testosterone-¹³C₃ to a concentration of 2 ng/mL. Volumes of 600 μL of sample (or calibration level) were added to Centrifree devices which were then centrifuged for between 5–30 minutes at room temperature, until between 200 to 300 mg of ultrafiltrate was collected in the bottom cups of the devices. After filtration, 75 μL of the ultrafiltrate was then placed in vials containing internal standard and subjected to quantitative analysis.

Repeated Analysis of the Sample

For this procedure, a known amount of analyte was added to the sample before the second analysis; this was achieved by using an additional solution (V_add) containing 10 ng/mL testosterone and 100 μg/mL phenytoin in PBS. The collection cups of the Centrifree devices were weighed (with an accuracy of ± 0.1 mg) before and after centrifugation and the difference in values was used as a direct estimation of the volume of ultrafiltrate.

Volumes of 600 μL of sample (or calibration level) were added to Centrifree devices which were then centrifuged for between 5–30 minutes at room temperature, until between 200 to 300 mg of ultrafiltrate was collected in the bottom cups of the devices. After filtration, 75 μL of the ultrafiltrate was then placed in vials containing internal standard and subjected to the first quantitative analysis. The remaining ultrafiltrate in the cups was discarded and a volume of additional solution (V_add) equivalent to the volume of filtrate lost was added back into the Centrifree devices. The devices were capped and mixed by inversion (~10 times) and then placed on a shaker for 30 minutes at 750 rpm for equilibration. Subsequently, the devices were centrifuged one more time and 75 μL of the ultrafiltrate was mixed with internal standard and subjected to the second quantitative analysis.

Nonspecific Binding and Losses

Nonspecific binding and losses due to the sample preparation procedures can be significant at the low parts-per-billion concentrations investigated. These losses were determined by performing mass-balance calculations based on comparing the amounts of compounds in the initial standard solutions to those that were extracted (for spintips) or filtered (for ultrafiltration) and those that were left in the standard solutions after the sample preparation procedures.

Statistical Analysis

Data for repeated experiments were reported as mean ± standard deviation. Data for comparing different measurement approaches were reported as slope, intercept, correlation coefficient, and p-value. For samples prepared in the laboratory, the accuracy of total concentrations was calculated versus the prepared value; the accuracy of free concentrations measured by microextraction was calculated by comparison to free concentrations measured by ultrafiltration; since ultrafiltration is the standard method for measuring free concentrations, its reproducibility was reported instead of accuracy. For patient samples, the accuracy of total concentrations was calculated versus the standard sample preparation based on protein precipitation; for free concentrations, the correlation between microextraction and ultrafiltration was reported. All calculations were performed using Microsoft Excel v.16.

Results

The four newly proposed comprehensive analysis approaches were first tested on standard solutions prepared with known concentrations of testosterone, phenytoin, and human serum albumin. Results obtained with the final equations for calculating C_t, C_f, and PBC were compared to the actual concentrations of the standard solutions and the results are shown in the supplementary data file, supplementary Figures 1 to 6. Briefly, for each testosterone and phenytoin concentration, HSA levels varied between 11 and 53 g/L; for each HSA concentration, testosterone levels varied between 0.027 and 10 ng/mL and phenytoin levels varied between 0.39 and 100 ng/mL. The investigated range for phenytoin concentrations was chosen below its therapeutic window in order to avoid interference with the binding of testosterone to plasma proteins, since both analytes were added to the samples prepared in the laboratory; this was not a problem for the analysis of patient samples, which did not contain phenytoin; nevertheless, if the proposed methods are to be applied to patient samples containing phenytoin, concentrations in the therapeutic range should be investigated.

For all the investigated concentration combinations, the accuracy of determined PBC varied between 90.8% and 110%. For testosterone, the accuracy of total concentration determinations varied between 87.9% and 109%; the accuracy of free concentrations determined by spintip microextraction varied between 84.7% and 115%, while the reproducibility of free concentrations measured by ultrafiltration varied between 4.1% and 14.1%. For phenytoin, the accuracy of total concentration determinations varied between 88.0% and 115%; the accuracy of free concentrations determined by spintip microextraction varied between 88.2% and 124%, while the reproducibility of free concentrations measured by ultrafiltration varied between 1.7% and 13.5%.

Prior to the analysis of endocrine patient samples, all four methods were further validated using pooled plasma samples from healthy donors. Different levels of PBC were simulated by diluting the plasma samples with PBS, which also resulted in lower testosterone concentrations to be tested. Results for the accuracy of PBC determinations for the four methods and two test compounds are shown in Table 1. Higher testosterone and phenytoin concentrations were simulated by spiking the plasma samples with additional compound. The validation results for the four methods applied for testosterone are shown in Table 2. For phenytoin, only the repeated analysis approaches were used, and the results are shown in Table 3. For both types of samples, HSA and pooled plasma, nonspecific losses were found to be 5.5% for analysis by ultrafiltration and 9.4% for analysis by spintip microextraction.

Table 1.

Accuracy of PBC determinations for pooled plasma samples using the four proposed methods and the two test compounds; for each PBC level, testosterone levels varied between 0.45 and 11.8 ng/mL and phenytoin levels varied between 10 and 100 ng/mL. All values represent mean ± SD.

PBC Level	ST+Label, Testosterone	UF+Label, Testosterone	ST+Repeated, Testosterone	UF+Repeated, Testosterone	ST+Repeated, Phenytoin	UF+Repeated, Phenytoin
	Accuracy%	Accuracy%	Accuracy%	Accuracy%	Accuracy%	Accuracy%
25%	96.8±8.5	94.9±7.3	105.1±11.8	97.7±7.9	97.2±9.3	101.0±10.0
50%	96.6±5.9	97.1±7.1	102.5±9.0	98.0±8.4	94.3±11.7	105.7±4.5
100%	101.9±7.4	101.6±3.4	100.9±7.7	104.8±6.5	98.3±10.9	109.5±6.3

Table 2.

Accuracy of C_t determinations for ST- and UF-based methods, reproducibility for C_f measurements for UF-based methods, and accuracy of C_f determinations for ST-based methods; for each testosterone concentration, PBC levels varied between 25% and 100%; results are for pooled plasma samples. Accuracy values are presented as mean ± SD.

Testosterone conc. (ng/mL)	ST+Label, Testosterone		UF+Label, Testosterone		ST+Repeated, Testosterone		UF+Repeated, Testosterone
	Accuracy% for Ct	Accuracy% for Cf	Accuracy% for Ct	RSD% for Cf	Accuracy% for Ct	Accuracy% for Cf	Accuracy% for Ct	RSD% for Cf
0.45	109.6±2.1	113.4±10.6	93.7±5.2	9.2%	98.2±4.8	95.4±15.3	117.2±2.4	9.1%
0.9	104.9±4.2	109.0±8.1	99.7±9.1	3.2%	101.8±7.1	100.6±14.7	120.4±1.2	13.7%
1.8	99.8±1.2	107.3±5.5	101.9±11.1	8.6%	99.4±6.4	99.3±12.7	122.3±1.6	10.3%
4.3	98.7±1.6	88.7±6.9	99.8±9.2	11.1%	95.9±1.9	100.7±11.9	107.0±2.8	5.7%
6.8	101.9±5.1	87.3±2.4	99.8±6.0	6.4%	97.0±2.5	101.0±13.7	98.7±2.7	8.9%
11.8	98.7±7.3	97.6±12.6	100.7±8.1	5.5%	96.4±1.8	111.2±11.3	94.3±7.4	6.9%

Table 3.

Accuracy of C_t determinations for ST- and UF-based methods, reproducibility for C_f measurements for UF-based methods, and accuracy of C_f determinations for ST-based methods; for each phenytoin concentration, PBC levels varied between 25% and 100%; results are for pooled plasma samples. Accuracy values are presented as mean ± SD.

Phenytoin conc. (ng/mL)	ST+Repeated, Phenytoin		UF+Repeated, Phenytoin
	Accuracy% for Ct	Accuracy% for Cf	Accuracy% for Ct	RSD% for Cf
10	112.7±5.0	115.6±4.2	92.3±5.3	6.6
20	104.3±7.1	110.7±8.5	97.4±2.8	2.5
35	89.3±9.2	90.6±1.2	91.6±6.8	6.0
50	83.5±2.7	79.4±1.1	92.1±8.5	5.5
100	84.5±5.4	83.5±3.2	92.5±5.2	6.5

Analysis of Clinical Samples

The clinical samples from volunteers were analyzed using ST and UF approaches with addition of labeled analyte, since these were the simplest and fastest approaches. For validation purposes, the total concentration of testosterone was also measured by protein precipitation with acetonitrile (and isotopically labeled standard); the total concentration of albumin was measured by colorimetry using the bromocresol purple method; the total concentration of SHBG was measured by ELISA.

Total testosterone concentrations determined in volunteer samples using the ST and labeled analyte approach were highly correlated (R²=0.9954) to total concentrations obtained using the standard protein precipitation approach. Furthermore, the slope was very close to one and the intercept very close to zero (Figure 1).

Figure 1.

Correlation of the spintip (ST) with labeled analyte approach for determination of total testosterone in clinical samples versus standard protein precipitation (PP) approach; each data point represents a volunteer; n = 50 volunteers; p<0.0001

The use of UF with labeled analyte for determination of total testosterone concentrations resulted in values less correlated to the standard protein precipitation approach (R²=0.8978) and with a slope much lower than one (Figure 2A). This was probably the result of binding protein leakage through the ultrafiltration membrane, although this leakage was not directly confirmed experimentally for patient samples. Once the concentrations of binding proteins were used as co-variates to correct the values of total concentrations determined by UF, accounting for 0.32% assumed leakage of binding proteins, the corrected values had a stronger correlation (R²=0.9325) with concentrations obtained by protein precipitation; furthermore, the slope improved from 0.5663 to 1.0131 (Figure 2B).

Figure 2.

(A) Correlation of the ultrafiltration (UF) with labeled analyte approach for determination of total testosterone in clinical samples versus standard protein precipitation (PP) approach; and (B) correlation of the ultrafiltration (UF) with labeled analyte approach for determination of total testosterone corrected for 0.32% binding protein leakage versus standard protein precipitation (PP) approach; each data point represents a volunteer; n = 50 volunteers; p<0.0001

The free testosterone concentration was determined by both ST and UF approaches and were found to be reasonably well correlated, with R²=0.9102; however, the slope of the correlation was lower than one, indicating that ST results are lower than those obtained with UF – in agreement with the assumption that some binding proteins leaked through the UF membrane, resulting in a slight overestimation of the free concentration (Figure 3).

Figure 3.

Correlation of the spintip (ST) with labeled analyte approach for determination of free testosterone in clinical samples versus ultrafiltration (UF) with labeled analyte approach; each data point represents a volunteer; n = 50 volunteers; p<0.0001

The plasma binding capacity (PBC) of the volunteer samples determined using the two newly proposed methods (with addition of labeled analyte) were well correlated (r = 0.7282) and achieved statistical significance, but the slope of the relationship was considerably less than one (Table 4). This indicates that the results obtained by UF were much lower than those obtained by ST, which again points to probable leakage of binding proteins through the ultrafiltration membrane. Furthermore, the PBC determined by UF did not correlate with measured total SHBG or the calculated PBC (obtained from binding constants and protein concentrations); there was a statistically significant correlation with measured total HSA, with a slope of 0.4963. On the other hand, the PBC determined by ST correlated very well with both measured total SHBG and calculated PBC, with a slope close to one. Table 4 also presents data for comparing the free and total concentrations obtained for volunteer samples with the current methods, with the results showing statistically significant correlations.

Table 4.

Correlations between the results obtained using the newly proposed spintip (ST) and ultrafiltration (UF) approaches with addition of labeled analyte versus results obtained by specific protein analysis and protein precipitation (for total concentrations); the data is based on samples from 50 endocrine patient volunteers

Comparison	Correlation coefficient	Slope	p
Ct by ST vs Ct by PP	0.9977	1.0047	7.9E-58
Ct by UF vs Ct by PP	0.9475	0.5663	2.0E-25
Ct by UF & SHBG, HSA vs Ct by PP	0.9657	1.0131	9.3E-30
Ct by UF vs Ct by PP	0.9475	0.5663	2.0E-25
Cf by ST vs Cf by UF	0.9540	0.7174	9.1E-27
PBC by UF vs PBC by ST	0.7282	0.2050	2.1E-09
HSA conc. vs PBC by ST	0.1647	0.0680	0.25
SHBG conc. vs PBC by ST	0.4663	0.8819	6.4E-04
calculated PBC vs PBC by ST	0.5040	0.9500	1.9E-04
HSA conc. vs PBC by UF	0.3383	0.4963	0.02
SHBG conc. vs PBC by UF	0.0901	0.6053	0.53
calculated PBC vs PBC by UF	0.1646	1.1016	0.25

Discussion

The newly developed methods can, in principle, work for any small molecules that bind to proteins. To test the utility of the methods, two relevant compounds were chosen: testosterone and phenytoin. Testosterone samples for method validation were prepared to encompass the physiological range and phenytoin samples were prepared at ten times higher levels in order to explore a different concentration range. For testosterone, two different isotopically labeled standards were available, so all four proposed methods were utilized. For phenytoin, only one isotopically labeled standard was available and was used for quantitative analysis; therefore, only the repeated analysis approaches were investigated for phenytoin. Ultrafiltration is considered a standard method for measuring free concentrations and was used to assess the accuracy of free concentration measurements by spintip microextraction for HSA and pooled plasma samples, as shown in Table 2 and Table 3; accordingly, these tables show the reproducibility of free concentration measurements for ultrafiltration-based methods.

A good analytical method should have an accuracy close to 100% and an RSD close to 0%. Given the inherent difficulties in analyzing biological samples, it is often acceptable for chromatography-based methods to have deviations of up to ±15% throughout the reporting range and up to ±20% at the lower limit of quantitation.^25,26 Nevertheless, there is no general consensus on method validation, as exemplified by differences in the guidelines of U.S. Food and Drug Administration versus the European Medicines Agency.²⁷ Methods based on immunoassays are often found to have much higher deviations in clinical practice.

For validating the new proposed method, samples with precisely known composition were first prepared by using analytical standards of albumin and the two test compounds diluted in physiological PBS. For these samples, PBC values determined with equations 16, 21, 28, and 31 were very accurate, within ±10% of the nominal value. The accuracy for determining free and total concentrations of testosterone and phenytoin varied over a wider range but was still acceptable.

Next, all four new methods were tested using pooled plasma, either ‘as received,’ spiked with the test compounds, or diluted with PBS. When microextraction with a labeled analyte was used, the calculated free concentrations of testosterone were between 87 and 113% accurate, with wider variability (±SD) than calculated total concentrations which were between 99 and 110% accurate. Calculated PBCs were consistent with the prepared samples, ranging from 97 and 102% accuracy. When microextraction with repeated analysis was used, calculation of total testosterone concentrations from extracted analyte ranged between 96 and 102% accuracy with minimal variability. Free concentrations also demonstrated desirable accuracy (95–112%) but with wider variability between samples. Phenytoin displayed less desirable recovery, as accuracy for both free and total concentrations decreased with increasing phenytoin concentrations. Determinations of PBC were consistent and reliable for both compounds. When ultrafiltration with a labeled analyte was used, the determination of free testosterone concentrations was very reproducible with less than 11% RSD and calculated total concentrations offered very good accuracy (95–102%). Calculated PBCs were reliable for this approach as well, with accuracies between 95 and 105%. Lastly, when ultrafiltration with repeated analysis of testosterone and phenytoin from pooled plasma was used, free testosterone and phenytoin determinations were highly reproducible (RSD < 14%) across all concentrations in plasma. Experimental errors resulted in some total testosterone calculations near 120% accuracies for lower concentrations, which could occur if the difference between subsequent free concentrations measurements was not significant enough (C_f1-C_f2 in Equation 32). Calculated total phenytoin concentrations consistently underestimated actual values (92–97% accuracy) but were still reasonable for work with plasma samples at low parts-per-billion levels. Calculated PBCs were accurate between 98–110%. In summary, all three types of measurements were found to be in an acceptable range of ±20% error for analysis of clinical samples.

The results for phenytoin accuracy when using the repeated analysis with ST devices, as shown in Table 3, indicate a decrease in accuracy with increasing concentrations of phenytoin; a possible explanation could be that the microextraction devices become saturated at higher concentrations of phenytoin and are not able to extract an amount proportional to the concentration in the sample; if this is the case, microextraction devices with a larger amount of extractive phase would be needed to improve accuracy at higher concentrations.

When appropriate sample preparation approaches are used, the proposed methods tend to produce low uncertainty in final calculations but rely on accurate quantitation by LC-MS/MS methods. The use of spintips was found as the preferred method for the calculation of free and total testosterone concentrations, while Centrifree devices were better suited for free phenytoin determinations. Using Centrifree devices with a labeled analyte was acceptable for the calculation of free and total testosterone at all investigated concentrations, while the repeated analysis methods were more reliable at higher concentrations. Using repeated analysis, the determination of total phenytoin was underestimated at all concentrations which may be due to difficulty in producing a detectibly large difference in compound concentrations after the addition of more analyte between filtrations. These methods were all suitable for the calculation of PBC, with an overall accuracy between 94 and 110%. While for some total testosterone concentrations measured with Centrifree devices the accuracy was overestimated at 122%, and for some total phenytoin concentrations the spintip approach resulted in underestimation at 84%, the methods are still suitable for bioanalytical investigations.

The experimental methods were initially challenging due to poor reproducibility caused by the separation devices that were used and by the LC-MS/MS method for quantitation. The use of spintips with hydrocarbon chains smaller than C8 did not produce a strong enough signal for good quantitative analysis, while the use of particles with higher carbon content resulted in significant co-extraction of proteins from plasma. Several types of ultrafiltration devices were investigated and Centrifree was found to result in the lowest loss of analytes. Nevertheless, at high concentrations of plasma proteins, the time needed for collecting an appropriate amount of ultrafiltrate increased greatly due to the low surface area of the membrane and increased osmotic pressure. The initial LC-MS/MS method based on 0.3 mL/min flow rate and 7 min run time proved to suffer from significant signal suppression, which resulted in poor sensitivity and reproducibility; doubling the flow rate and increasing the run time produced a much better quantitative method. Beginning experiments before allowing spiked samples to equilibrate for at least 30 minutes on a shaker at low speed produced poor results and lacked reproducibility. Nonetheless, these roadblocks were amendable to allow for dependable usage of the new methods, but they outline the need for careful consideration of all experimental variables – especially for equilibrium-based sample preparation approaches.

Analysis of Clinical Samples

The use of spintips on patient samples was found to be the most reliable method of determining total testosterone concentrations with a correlation slope of one versus the protein precipitation approach, demonstrating a 1:1 relationship between the newly proposed method and the standard approach (Figure 1). This strong correlation was further supported by a y-intercept very close to zero and an R² of 0.9954. The use of Centrifree devices with addition of labeled analyte resulted in a significant underestimation of the total testosterone concentrations in clinical samples, with a correlation slope of only 0.5663 vs protein precipitation data (Figure 2A and Table 4). This was unexpected, since the method worked very well for HSA and pooled plasma samples (supplementary Figures 1 to 6 and Tables 1 to 3). A possible explanation, supported by the results obtained for free concentrations and PBC values, is that small amounts of binding proteins leaked through the ultrafiltration membrane. When ultrafiltration results for total concentrations were corrected for 0.32% presumed binding protein leakage, the correlation with protein precipitation data was reasonably good with a slope close to one, a y-intercept close to zero, and an R² of 0.9325 (Figure 2B and Table 4). However, even with corrections for potential protein leakage, the total concentration accuracy obtained by ultrafiltration varied greatly among the samples, most likely because the fraction of protein leakage varied from sample to sample.

For the determination of free concentrations of testosterone, any available method would have its own biases, so only results obtained with the two newly proposed methods were compared. The two methods correlated with a slope of near 0.72, a y-intercept close to zero, and an R² of 0.91 (Figure 3 and Table 4). The slope indicates that ultrafiltration consistently overestimated the free concentration by comparison to spintip microextraction, which is in accord with the assumption that a small amount of binding proteins leaked through the ultrafiltration membrane (along with bound testosterone) when clinical samples were analyzed.

PBC values determined by ultrafiltration and spintip microextraction from clinical samples achieved a statistically significant correlation, but with a small slope for UF vs ST (Table 4), indicating that the UF method detected a lower amount of binding proteins – in accordance with the observation that UF reported higher free concentrations and probably suffered from protein leakage. The superiority of the ST approach was further proven by a statistically significant correlation with the PBC values calculated from measured HSA and SHBG concentrations (Table 4); in addition, the slope of this correlation was close to one. The correlation between PBC by UF and calculated PBC also had a slope close to one, but did not achieve statistical significance (Table 4).

According to our results and the current FDA guidelines for ‘Bioanalytical Method Validation,’ all four proposed methods could be validated for use with laboratory generated samples, such as the HSA and pooled plasma samples that were tested here. For clinical samples, with their higher variability and unexpected composition, only the methods based on addition of labeled analyte were tested, and of these only the ST method met the requirements for accuracies and precisions that are ±20% of actual values. The method based on ultrafiltration consistently underestimated total concentration values while the microextraction method retained reliable accuracy. Future work would involve testing for specificity with the addition or presence of other molecules that can cross-react or interfere with plasma protein binding. More finalized methods would test the effects of sample collection, handling, storage, and stability.

Conclusions

Four new experimental approaches for comprehensive analysis were proposed and validated with HSA and pooled plasma samples. The methods offer more data from samples while using the same single quantitative analysis technique. When the addition of a labeled analyte approach is used, all three parameters, the free concentration, total concentration, and plasma binding capacity were measured using a single sample preparation procedure. Spintip microextraction devices were found to offer superior signal for quantitation methods while Centrifree ultrafiltration devices offered ease of use. Currently, Centrifree devices are the most commonly used for free fraction separation in clinical settings.

The methods based on addition of labeled analyte were also validated by analysis of 50 clinical endocrine patient samples and showed strong correlations with standard analytical methods. This supports the continued use and development of the newly proposed approaches to fulfill an evident need for more functional testing that can personalize healthcare in cases of altered protein binding and patient variability while also improving the efficacy and safety of drugs currently in use.

The overall success of proposed experimental approaches with simulated samples and patient samples support further research in this area. For future studies, the correlations between the three determined parameters and therapeutic effects will need to be examined to determine if the usage is valuable in pharmacotherapy decision making. To make these new methods more easily transferable to current clinical laboratory settings, the use of Centrifree devices needs to be adjusted to produce more reliable results. Depending on the target compounds these new methods will be used for, current laboratory equipment may need updates. However, for common compounds, the currently available quantitative methods can be used on separated free fractions. These features make the development of these methods more attractive since they are easily translatable to practical use and more cost effective than the development of entirely new methods.

Abbreviations

PK

pharmacokinetics

PD

pharmacodynamics

PPB

plasma protein binding

PBC

plasma binding capacity

HSA

human serum albumin

SHBG

sex-hormone binding globulin

PBS

phosphate buffered saline

SPME

solid-phase microextraction

ST

spintip

ME

microextraction

UF

ultrafiltration

K

drug-protein binding constant

m_e

moles of compound extracted

m_eL

moles of labeled compound extracted

m_add

total moles of additional compound

m₁

moles of extracted compound from 1^st extraction

m₂

moles of extracted compound from 2^nd extraction

e_c

extraction constant

C_f1

initial free concentration before addition

C_f2

initial free concentration after addition

C_fL

initial free labeled compound

C_fe

free concentration after extraction

C_feL

free concentration of labeled compound after extraction

C_tL

initial total labeled compound concentration

C_t1

initial total concentration

C_t2

initial total concentration after addition

C_te

total concentration after extraction

C_teL

total labeled compound concentration after extraction

C_te1

total concentration after 1^st extraction

C_te2

total concentration after 2^nd extraction

C_LCMS

concentration measured with LCMS

C_LCMS-L

concentration of labeled compound measured with LCMS

C_add

concentration of added compound

V_add

added compound volume

V_s

sample volume

V_d

desorption volume

V_f

filtrate volume