Introduction

Chiral drugs contain at least one chiral center, resulting in 2n−1 pairs of enantiomers. Widely used chiral drugs, such as rosuvastatin, duloxetine and salbutamol, play an important role in treating human diseases1,2,3. In the environment of living systems, specific binding between molecules (eg, enzymes, receptors, transporters, and DNA) is required for their medicinal effect. Thus, the physiochemical and biochemical properties of racemic mixtures and individual stereoisomers can differ significantly4. In some cases, one enantiomer is active, while the other may produce deleterious side-effects, including toxicity5.

Numerous studies have reported the stereoselectivity of chiral drug metabolism and pharmacokinetic profiles6,7,8,9,10,11. Enantiomers commonly display pharmacokinetic processes (eg absorption, distribution, metabolism, and excretion) in a stereoselective manner5,12,13,14. Moreover, the plasma protein binding and tissue distribution of some chiral drugs also exhibit stereoselectivity15. To limit the scope of this review, the binding of drugs to tissue proteins is not covered. Drugs bind to plasma proteins with varying degrees, and these bindings are commonly reversible. A binding equilibrium exists between the bound and free molecules, but only the unbound drug exerts efficacy16. Moreover, stereoselective binding can sometimes significantly affect the amount of free drug present in the plasma, and in many cases, this can be species-dependent17. Consequently, the characterization of drug binding to plasma proteins is an important factor for determining the overall pharmacological activity of a drug18.

Blood is separated into the blood cells and plasma. Plasma contains various proteins and several function as carriers, including human serum albumin (HSA), α1-acid glycoprotein (AGP) and lipoproteins19,20. Among plasma proteins, HSA and AGP play predominant roles by binding to most drugs21. As a result, the structure, function, and pharmaceutical properties of HSA and AGP have been extensively investigated22,23. Fully characterizing the mechanism by which drugs bind to proteins such as HAS and AGP has become essential to interpret the pharmacokinetic, pharmacodynamic, and toxicological profiles of chiral drugs. In vivo binding studies using plasma samples and in vitro binding studies using plasma proteins, including natural proteins, recombinant fragments and variants, are helpful for understanding plasma protein binding properties. In addition, recent advances in determining the concentrations of enantiomers and fundamental analytical techniques are introduced. Here, we focused our attention on the enantioselective binding of chiral drugs to plasma proteins and the methods used to evaluate stereoselective binding.

Methods and models

Because the pharmacological activity of one enantiomer may differ from its antipode, it is particularly important to know the extent of binding for each enantiomer. Some methods have been proposed to assess protein-binding capabilities based on diverse analytical tools16,24,25. Gilabert et al26 established 3 steps to study the stereoselective binding between enantiomers and proteins: 1) equilibration of racemic mixture and proteins, 2) separation of the unbound fraction, and 3) determination of the concentration of the enantiomers from either the free fraction or drug-protein complexes.

To illustrate enantioselective drug-protein binding, classical methods, such as equilibrium dialysis (ED), ultrafiltration (UF) and ultracentrifugation (UC), are commonly combined with chiral separation techniques27,28,29. ED is an apparatus with two compartments separated by a semipermeable membrane, and only unbound drug molecules can permeate through the membrane. ED is carried out in solution, and true equilibrium is maintained during the whole process. Although ED has several disadvantages (eg, time-consuming, solubility, and non-specific adsorption), it is still considered the reference method for binding measurements. UF is a more rapid and simple alternative that depends on centrifugation forces and a sieve-like membrane to separate drug-protein complexes and the free drug. UC, another type of technique that avoids membrane effects, is based on the sedimentation coefficient differences of substances. For extensive reviews of these three approaches that discuss their advantages and pitfalls, refer to Vuignier16 and Howard et al21. Table 1 briefly summarizes the progress made in recent years regarding the methods used to study the enantiodifferentiation of chiral drugs with plasma proteins.

Table 1 Binding parameters and techniques for the enantiodifferentiative study of chiral drugs with plasma protein.
Full size table

With the development of computational models for the prediction of drug pharmacokinetics, it is important to generate models that predict drug binding affinities and stereoselectivity to plasma proteins for virtual screening. In the last decade, several models have been developed to study the binding between HSA and restricted drug families45,46,47, and a few global models have been developed based on different approaches, such as genetic function approximation, multiple linear regression, heuristic regression procedures and ant colony systems48,49,50. Monti et al combined molecular mechanisms (MM) and molecular dynamics (MD) with circular dichroism (CD) to identify the main interactions between ketoprofen enantiomers and the surrounding amino acids at short distances in bovine serum albumin51. Similarly, Yu et al took advantage of MM and MD to identify several key residues that are involved in the enantioselectivity for the binding of AGP to mexiletine enantiomers, such as Arg9043.

Despite the techniques and computational models mentioned above, important mathematical drawbacks of parameter estimation [eg, protein binding percentage (PB), number of binding sites (n), affinity constants (Ka), and enantioselectivity to the protein (ES)] have been ignored. Sandblad et al52 calculated the adsorption energy distribution (AED) to provide a narrower selection of probable models from the surface plasmon resonance (SPR) raw data. Using this method, both the R/S-propranolol-AGP and R/S-warfarin-HSA systems were heterogeneous, comprising both high-affinity chiral sites and weak nonselective sites. Recently, the novel direct equations extracted from the classical interaction model allowed for advantageous univariate mathematical data treatment, providing the first evidence of quantitative (±)-catechin-HSA stereoselectivity40. Therefore, the integration of robust in vitro information with molecular docking estimates could provide a synergistic approach for the understanding of stereoselective binding.

Stereoselectivity of plasma protein binding to chiral drugs

Plasma

Human plasma contains HSA, AGP, lipoproteins, and globulins, which are responsible for the plasma protein binding of drugs53. All of these proteins can simultaneously bind to a drug, and the overall plasma protein binding is the sum of each binding. It is known that HSA accounts for 60% of plasma protein, while the amount of AGP is only 3% of plasma protein23. Plasma protein binding (PPB) limits free drug motion and reduces the volume of distribution, renal extraction, liver metabolism and tissue penetration. In contrast, drug absorption and half-life increase with PPB54.

Chiral drugs with different pharmacological activities have been extensively explored due to their stereoselective pharmacokinetics55,56. To characterize the stereoselective pharmacokinetics involved in protein binding and/or metabolism, several studies have been conducted in vivo and in vitro57. Herein, we enumerate the cases intending to evaluate the contribution of protein binding and metabolism to stereoselective pharmacokinetics in vivo. Lansoprazole, which is extensively metabolized in the liver, is frequently prescribed for the treatment of acid-related disorders. The disposition of lansoprazole differed in extensive and poor metabolizers of CYP2C19, which showed genetic polymorphisms58. Interestingly, Kim et al also investigated that enantioselective disposition of lansoprazole in 6 extensive metabolizers and 6 poor metabolizers, but the enantioselective protein binding was more important than the effect of CYP2C19 genetic polymorphisms59. Likewise, the unbound fraction of R-oxybutynin (OXY) in human plasma was approximately two-fold higher than that of S-OXY, and the metabolic kinetics were slightly different for the enantiomers60. Therefore, the enantioselective binding to plasma proteins was a major factor that was responsible for the stereoselective pharmacokinetics of OXY.

In vitro plasma protein binding experiments can provide valuable data, but an optimized experimental design is required to reach physiological relevance. As stated previously, whole plasma proteins studies and the contribution of individual proteins to total plasma enantioselective protein binding warrant further study. The stereoselectivity of the binding of propranolol (PL) enantiomers to plasma proteins is the opposite of HSA, but acetyl salicylic acid (ASA) and salicylic acid (SA) significantly displaced the binding of R-PL to a greater extent than S-PL from both plasma protein and HSA binding sites61. These data suggest that ASA and SA do not affect the binding of PL to AGP because of the different acid-base properties of these drugs. For nomifensine enantiomer E1 (the first elute, protein binding to HSA 40%±5%), other plasma proteins were expected to contribute according to the plasma protein binding (58%±7%), but not for E2 (the second elute, PB 63%±4% and 64%±4% for HSA and plasma, respectively)32. Thus, the relative importance of HSA for binding nomifensine enantiomers was confirmed. The binding of bimoclomol to human plasma was stereoselective, and AGP was mainly responsible for the preference toward S-bimoclomol, whereas HSA did not play a role62. Another study showed that HSA and human plasma binding tended to be stereospecific in regard to S-amlodipine, whereas the opposite binding of its enantiomer was observed for AGP29.

As the amino acid sequence differs among animals, interspecies plasma proteins exhibit different binding abilities. It has been recognized that preclinical data from animals cannot be extrapolated to humans63. Further studies also showed that stereoselective differences are dependent on species28,29. Nevertheless, mammalian results are occasionally consistent with human disposition64. For example, the in vitro concentration ratios of R-MK0767 to S-enantiomer were similar in dog and human plasma (1.5–1.7), but the stereoselectivities in rat and rabbit plasma were inverted65. Although the results of enantioselective binding between animal and human are complicated, in vitro experiments may help explain some of these unusual discrepancies.

Human serum albumin (HSA)

HSA, a single non-glycosylated stranded protein consisting of 585 amino acids, is the most abundant protein in plasma, reaching high concentrations of approximately 0.5 to 0.7 mmol/L66. He and Carter67 characterized the atomic structure of HSA using X-ray crystallography and described it as a heart-shaped protein with three homologous domains (labeled I, II, and III), each containing two subdomains (A and B) with similar structure68,69,70.

As a carrier for endogenous ligands such as fatty acid, bilirubin and peptides71, HSA solubilizes hydrophobic compounds. It also assists in providing a homogeneous and buffered drug distribution through the body and increases the biological lifetime of a drug by preventing its metabolism49. According to recent reports, two principal hydrophobic binding sites for aromatic and heterocyclic molecules were identified in the native conformation of HSA70. Site I (warfarin-azapropazone site) and Site II (indole–benzodiazepine site) are located in subdomains IIA and IIIA, which correspond to Sudlow's Sites I and II, respectively. Another high-affinity binding site (Site III) was shown to specifically bind to digitoxin on albumin72.

In particular, HSA exhibits the highest potential stereoselectivity among all plasma proteins, and it plays a key role in the distribution, metabolism and elimination of enantiomers23. The protein binding properties of a chiral drug, including the specific binding sites and affinity constants, could differ among enantiomers, resulting in different biological properties. Therefore, it is important to study the nature of the interaction between chiral drugs and HSA. Cooperative and allosteric equilibria between different binding sites and competition between multiple drugs or between drugs and endogenous ligands make it difficult to interpret HSA binding properties in vivo73. However, alternative methodologies have been proposed extensively in vitro.

Competition interaction

Competitive binding was exploited to determine protein binding sites using drug displacement assays. The displacement of equilibrium between a racemic drug can weaken the efficacy for the more active enantiomer and take the place of their antipodes74,75. One example of this is the competitive binding between indobufen enantiomers76. R-indobufen displaced its antipode, thus increasing the steady state concentration of free S-enantiomer in patients. Consequently, the quicker elimination of the S-enantiomer is associated with its weaker binding to HSA.

Usually, displacement experiments using classical markers or probes are performed to elucidate specific binding sites in vitro77. Additionally, in the case of enantiomers sharing the same binding site, the binding mechanism may be different.

Using the well-characterized HSA ligand, rac-ibuprofen, Zsila et al suggested that leukotriene B4 bound to site II in subdomain IIIA using CD displacement experiments78. The naproxen (NPX) moiety of S-NPX bound to a cyclopentapeptide with an arginine-glycine-aspartate sequence (cRGD) bioconjugate that is farther from Trp than the R-epimer (approximately 16 and 6 Å, respectively). This finding highlights the critical role that the absolute configuration of epimeric macromolecular systems plays because the chirality of its different stereocenter can affect its binding mode to HSA, although both bioconjugates bound preferentially to site I37. In contrast, there is an increasing trend toward investigating chiral recognition in excited states. The interaction between excited carprofen (CP) and HSA shows remarkable stereodifferentiation, which is reflected by the markedly different triplet lifetimes of the two CP enantiomers in both binding sites (site I and site II), especially Trp-containing site I79.

Cooperativity and allosteric interaction

Because the conformational adaptability of HSA extends well beyond the immediate vicinity of the binding site(s), cooperativity and allosteric modulation occur among binding sites73. Cooperativity represents a synergistic effect between two ligands that are sequentially bound to the different receptor sites80. Allosteric modulation occurs when the interaction between one ligand and HSA changes the interaction of a second ligand with the same protein at a separate site. This interaction implies that the simultaneous binding of two ligands induces conformational changes of the protein, in addition to other factors such as pH, temperature and ionic strength81,82. To date, allosteric interactions have been reported to affect the binding equilibrium between HSA and endogenous or exogenous compounds, such as fatty acids, heme, carbamazepine and verapamil83,84,85,86.

It was previously reported that the simultaneous binding of S-warfarin and S-benzodiazepines demonstrated mutual and exceedingly enhanced binding87. Due to the different kinetic features of warfarin enantiomers binding to HSA, Fitos et al88 explained that S-lorazepam acetate allosterically enhanced the binding of S-warfarin by accelerating the relaxation kinetics of S-warfarin. A majority of the studies regarding allosteric interactions involved only qualitative observations. Chen and Hage89 used a biointeraction chromatography technique to provide quantitative information on both directions of an allosteric effect, especially for a multisite binding agent. Aside from the effects of secondary interactions or the nonspecific binding of phenytoin, the coupling constant for the effect of L-tryptophan on phenytoin was accurately determined to be a negative allosteric interaction. This result is quite different from the results of direct competition on the effects of phenytoin toward L-tryptophan. Therefore, examining the interaction between two ligands in both directions on a multisite binding agent is necessary. They also applied the previously described chromatography techniques to study the allosteric effects of R- and S-ibuprofen on the binding of benzodiazepines to HSA85. Additionally, detailed information on the identity of the ligand binding pocket(s) and specific amino acid(s) of HSA that are responsible for this allosteric effect is needed. Lammers et al90 showed the stereoselective binding of flurbiprofen (FBP) enantiomers and their methyl esters to HSA using time-resolved phosphorescence. Based on the phosphorescence lifetimes, R-flurbiprofen quenched Trp more effectively than S-flurbiprofen, in contrast to its methyl esters. However, the quenching constants of 3×10−7(mol/L)−1s−1 for R-FBP and 2.5×10−7(mol/L)−1s−1for its antipode were not influenced by methylation, suggesting that stereoselectivity existed in the accessibility of HSA Trp-214. Recently, single amino acid mutants and HSA conformational rearrangements were reported to elucidate their governing role in allosteric ligand binding91.

Recombinant domains of HSA

Recombinant HSA domains are a useful tool for characterizing the stereoselective binding properties of chiral drugs, and they also represent a suitable platform for the characterization of ligand binding. Based on the quasi-independence of the three HSA domains, proteolytic and chemical cleavage have been used to produce fragments of HSA to define binding sites92,93. Considering the folding of the domains and their viability as “stand alone” proteins, the cloning and expression of independent recombinant domains of human serum albumin was introduced94.

To study the accurate localization of ketoprofen and mexiletine binding sites on HSA, Shi et al95 produced three highly purified recombinant HSA domains, each of which had a specific ligand binding site. They found that HSA DOM III possessed the chiral recognition ability for the ketoprofen enantiomers, whereas HSA DOM II recognized the mexiletine enantiomers. Recombinant fragments of native proteins provide an indispensable contribution, but we doubt that they could completely displace the native functional protein. For example, measurements of ochratoxin A (OTA) complexes with recombinant proteins using fluorescence spectroscopy revealed that it bound to all domains, but the binding constants decreased in the series as follows: DOM II>>DOM III>DOM I96. Interestingly, the OTA binding constant for DOM II (7.9×105 (mol/L)−1) was smaller than the largest constant for HSA by nearly a factor of 7, whereas the binding constant for OTA with DOM III [1.1×105 (mol/L)−1] was similar to that of the secondary binding site for HSA.

Additionally, the essential structural elements required for the formation of functional ligand binding sites on HSA remain unclear. A defined set of five recombinant proteins comprising combinations of domains and/or subdomains of the N-terminus were prepared to investigate the binding mechanism of warfarin to the stand-alone protein fragments97. The primary warfarin binding site was centered in subdomain IIA and received indispensable structural contributions from subdomain IIB and domain I, but domain III was not involved in this binding site. Aside from the characterization of the warfarin binding site, the kinetic step(s) in the binding mechanism between enantiomers and albumin may also be responsible for chiral discrimination. The binding of warfarin to albumin occurred in at least two steps — a rapid diffusion-controlled step and a slower rate-limiting step98. Domain fragments of recombinant human albumin that possessed a functional warfarin binding site, corresponding to domains 1 and 2 (D12) and domains 2 and 3 (D23), were produced to demonstrate that the preference for the R-enantiomer can largely be explained by these domains, particularly the observation that D12 had a faster rate for R-warfarin binding in the second step99.

α1-acid glycoprotein (AGP)

AGP, also called orosomucoid, is a major binding protein for basic drugs and a diversity array of ligands42,57,100. Because AGP has only one drug-binding site, its binding to each molecule differs from albumin. The drug-binding properties of AGP are saturable and displaceable101. Human AGP is present in the plasma of healthy subjects at concentrations between 0.6 and 1.2 mg/mL, accounting for approximately 1% to 3% of the total protein102.

AGP is an acute phase reactant protein, and its serum concentration may increase up to three- or four-fold as a result of inflammation or immunological response103,104. The serum level of AGP significantly increases in renal disease patients, and the concomitant reduction in the free concentration of the S-alprenolol, with a large binding constant, was higher than that of the R-isomer, which has a small binding constant105. To study the impact of plasma protein binding on pharmacodynamics, Steeg et al106 indicated that the plasma protein binding of S-propranolol was restricting its effects on heart rate due to the elevated AGP concentration.

The AGP molecule consists of a single polypeptide chain of 183 amino acids with up to five asparaginyl linked glycans107. In addition to the high heterogeneity of glycans, polymorphisms have also been identified in the protein portion or AGP108. F1 S variants are encoded by the AGP A gene, and the A variant is encoded by the AGP B gene109. Although the binding activities of many racemic drugs to AGP are known to be stereoselective, the effect of the AGP subfractions has not been fully elucidated.

The stereoselective binding of coumarin-type anticoagulants to AGP F1S variants was the same as native AGP binding with the S-enantiomers of warfarin and acenocoumarol110. Similarly, Zsila et al reported that both AGP and the F1S variant preferred binding to (−)-mefloquine, whereas the stereoselectivity was reversed for the A variant111. Warfarin enantiomers had a higher binding affinity for the F1S variant compared to the A variant, and the dissociation constants for the F1S and A variants differed by 12.6-fold for the S-enantiomer and 8.3-fold for the R-enantiomer109. The affinity of AGP for (+)-tetrahydropalmatine (THP) was notably higher than (−)-THP, and the F1S variants of AGP proved to be the key variants for (−)- and (+)-THP binding38. Deramciclane effectively displaced acridine orange-10-dodecyl bromide, a high-affinity fluorescent probe of native AGP, binding to variant A, whereas it was less effective displacing the same probes bound to the F1S variant; this phenomenon could not be caused by its enantiomer112. Apart from the different enantioselective binding ability of AGP genetic variants, their relative concentrations also influenced the stereoselective binding results113. These data prompted us to further study the AGP phenotype that affects drug disposition in humans and its possible influence on pharmacologically relevant variables.

Lipoprotein

Plasma lipoproteins are a group of binding agents that are known to interact with solutes in serum, and they can bind several basic and neutral hydrophobic drugs114. Because apolipoprotein and lipid constituents of lipoprotein are chiral compounds, their enantioselective binding should be considered.

Recently, the interactions between R/S-propranolol and low-density lipoprotein (LDL) were studied by using HPAC44. Two types of interactions occurred between R-propranolol and LDL, whereas only the second interaction was observed for R-propranolol, which involved saturable binding with an association equilibrium constant (Ka) of (5.2±2.3)×105(mol/L)−1 at 37 °C. This study was the to provide information regarding LDL chiral selectivity. In another study, halofantrine enantiomers showed some stereoselectivity for lipoprotein binding in vitro, but they did not show stereoselectivity for plasma protein binding115.

Perspectives

Given the importance of stereoselective binding to plasma proteins and that approximately 50% of marketed drugs are racemates, it would be extremely useful to develop in vitro models to evaluate and predict binding affinities and relevant sites. Recently, the recombinant domains of HSA and variants of AGP have become powerful tools to elucidate the stereoselective binding properties of chiral drugs in vitro. However, further studies are needed to determine whether the binding results are consistent in vivo.