Albumin-Based Transport of Nonsteroidal Anti-Inflammatory Drugs in Mammalian Blood Plasma

Abstract

Every day, hundreds of millions of people worldwide take nonsteroidal anti-inflammatory drugs (NSAIDs), often in conjunction with multiple other medications. In the bloodstream, NSAIDs are mostly bound to serum albumin (SA). We report the crystal structures of equine serum albumin complexed with four NSAIDs (ibuprofen, ketoprofen, etodolac, and nabumetone) and the active metabolite of nabumetone (6-methoxy-2-naphthylacetic acid, 6-MNA). These compounds bind to seven drug-binding sites on SA. These sites are generally well-conserved between equine and human SAs, but ibuprofen binds to both SAs in two drug-binding sites, only one of which is common. We also compare the binding of ketoprofen by equine SA to binding of it by bovine and leporine SAs. Our comparative analysis of known SA complexes with FDA-approved drugs clearly shows that multiple medications compete for the same binding sites, indicating possibilities for undesirable physiological effects caused by drug–drug displacement or competition with common metabolites. We discuss the consequences of NSAID binding to SA in a broader scientific and medical context, particularly regarding achieving desired therapeutic effects based on an individual’s drug regimen.

Graphical Abstract

1. INTRODUCTION

Nonsteroidal anti-inflammatory drugs (NSAIDs) are non-opioid therapeutic agents that are used in the management of mild to moderate pain, fever, and inflammation. These represent some of the most widely used medicines in the world. It has been estimated that up to 30 million people in the United States use NSAIDs every day and that over 30 billion doses of NSAIDs are taken annually.^1-4 NSAIDs are not only used by humans, as domesticated animals are also treated with NSAIDs for inflammation and pain.^5-7 NSAIDs act via the inhibition of cyclooxygenase enzymes (COX), thereby inhibiting prostaglandin and thromboxane biosynthesis and reducing the inflammatory response and blood clotting.^8,9 NSAIDs are often administered orally and are generally well-absorbed in the gastrointestinal tract before subsequent delivery to the site of action by blood.¹⁰ They can also be administered topically, intravenously, intramuscularly, and rectally, and all of these delivery methods are followed by delivery via blood. In the bloodstream, NSAIDs are highly bound to plasma proteins (in many cases, 99% bound), mainly to serum albumin (SA).¹⁰

SA is a drug-binding, plasma protein that constitutes up to 55% of total plasma proteins. It has a typical blood concentration of 3.5–5.0 g/dL (600 μM) and acts as the major facilitator of vascular drug transport.^11-13 SA also transports hormones, metal ions, and common metabolites such as fatty acids and sugars. It also serves as the main determinant of plasma osmotic pressure by virtue of its high concentration and is a vital free radical scavenger.^14-20 Additionally, an unusually low concentration of SA in the blood is a cancer marker.¹⁷ During its month-long circulation in the bloodstream, SA makes nearly 15 000 trips around the body and travels through the lymphatic system 28 times, which enables the continuity of transport.^14,21 Together with SA’s ability to bind diverse compounds, this circulation facilitates the deep-tissue delivery of therapeutic agents.²² Because of its role as the major protein responsible for plasma protein binding (PPB), a molecule’s affinity to SA plays a crucial role in drug candidate optimization.^11,13,23

The extensive binding capacity of SA is the result of its highly flexible structure, high concentration in blood, and the presence of several binding sites that are able to accommodate a variety of small molecules. Crystal structures of mammalian SAs show three homologous domains (I, II, III), with each domain subdivided into A and B subdomains. The subdomains have six and four α-helices, respectively, and are connected by flexible loops. The arrangement of helices and loops in SA confers some degree of interdomain flexibility that results from the binding of certain ligands.¹⁶ This structural arrangement results in multiple, mostly hydrophobic, binding cavities located between the domains, subdomains, and helices. Until recently, drug sites 1 and 2 (Sudlow sites I and II) were generally expected to bind most drugs.^13,24-29 However, recent structural studies of human serum albumin (HSA) and other mammalian SAs have revealed several additional drug sites in this molecule.^30,31 With a variety of binding sites, SA has the impressive capability to bind various, often structurally different, classes of drugs such as anticancer agents, anticoagulants, antihistamines, anesthetics, anthelmintic agents, antiretrovirals, benzodiazepines, steroid- and amino acid-derived hormones, contrast media, and NSAIDs.^16,32,33

As a result of these interactions, the biodistribution and bioavailability of many drugs are dictated by binding to SA.³⁴ In mammalian circulation, drugs are bound to SA, bound to other transport proteins (e.g., lipoproteins and globulins), or unbound.¹³ The PPB of a drug is a reversible and rapid equilibrium process. According to free drug theory, only unbound drugs can enter tissues and exert their pharmaco-logical effects.^13,29 By virtue of its high blood concentration and high binding capacity, SA is one of the major factors that determines the free plasma concentration of many drugs, thereby modulating their therapeutic effects. Thus, drugs with higher affinities for SA may require higher dosages to achieve and maintain a desired therapeutic effect.³² In drug development, an overly high affinity of a lead compound to SA and other plasma proteins results in insufficient free drug concentrations, influencing the lead optimization process.^29,35 SA also provides a reservoir for drugs as they are removed from the blood by various processes, which prolongs the duration of the therapeutic action. Furthermore, SA plays an important role in the therapeutic efficiency of some drugs by decreasing the formation of drug aggregates.³² Consequently, the efficacy and toxicology of a drug are often impacted by the extent of its PPB, which significantly influences the pharmacokinetic and pharmacodynamic properties of a drug.^11,13 Binding to SA also plays an important role in drug lead studies on animal models. Generally, the PPB of compounds is expected to be similar across species, and ligands are expected to bind to the same sites in HSA and other mammalian SAs.³⁶ However, some compounds show PPB values in animal models that differ by an order of magnitude from the respective values in humans (e.g., valproate in mouse, etoposide in rat).^37-39 These factors make drug binding to SA an essential parameter for a drug’s efficacy and safety that is investigated in drug development programs.^11,13

When SA binds multiple drugs in circulation, a drug may be displaced by another strong-binding drug administered at a high concentration.^40-43 Drug–drug displacement usually occurs due to direct competition for a particular binding site but may also happen due to allosteric effects that change the affinity of one drug upon the binding of another.^11,13,44 This phenomenon likely occurs in a substantial number of individuals who take multiple drugs at the same time. According to the Centers for Disease Control, 72% of people taking any prescription drug take more than one drug at the same time (3.7 drugs on average).⁴⁵ Roughly 72 million people in the U.S. take NSAIDs at least three times a week for more than three months per year, often in conjunction with other drugs.⁴⁵ Gastrointestinal bleeding is a notable example of a detrimental drug–drug displacement effect that occurs when warfarin, an anticoagulant, is displaced from SA when coadministered with NSAIDs.^26,46,47 Some other drugs that have been shown to be displaced by NSAIDs include methotrexate (an immunosuppressant) and valproate (an anticonvulsant), and the displacement of small molecules such as SA-bound toxins have been shown to lead to adverse physiological effects (e.g., uremia).^13,48 However, the way in which SA binds commonly coadministered drugs has not been thoroughly investigated on a structural basis, precluding the molecular understanding of how commonly used drugs compete for binding to SA. For NSAIDs that have only been complexed with HSA, complexing the drug with a different mammalian SA would reveal potential differences in how SAs from different species bind drugs, thereby providing insight into how transport differs across species. Studying drug binding to SA across species helps us better understand the molecular determinants of binding and allows us to evaluate the usefulness of different model organisms. Also, small changes in the SA sequence most likely do not result in changes in the binding of most drugs, but studying drug binding in different systems (for instance, ESA vs HSA) provides cross-validation of structural results. Elucidating structures of novel SA–NSAID complexes will provide a molecular basis for the interactions of drugs with SA and lay a foundation for studies to probe these interactions in clinical contexts. Furthermore, the structure determination of SA–drug complexes contributes to new technological applications of SAs, e.g., using SA as a tool to deliver or remove drugs from a specific environment.^49,50

Here, we present crystal structures of equine serum albumin (ESA) in complexes with four NSAIDs (ibuprofen, ketoprofen, etodolac, and nabumetone) and 6-methoxy-2-naphthylacetic acid (6-MNA), the active metabolite of nabumetone (Figure 1). These structures provide novel molecular-level information about the interactions of these drugs with SA. In addition, the structures of ESA complexes with ibuprofen and ketoprofen provide crucial information about interspecies differences in binding as a result of comparisons with the structure of ibuprofen in complex with HSA and structures of ketoprofen in complex with bovine (BSA) and leporine SA (LSA) that were determined previously.^32,50,51 We describe how the compounds used in our study bind to sites that are known to bind numerous other FDA-approved drugs. Additionally, we examine the differences between the binding sites of these structures and the corresponding sites (hypothesized or identified) for the same ligands in HSA to understand possible differences in transport based on the conservation of amino acid residues in these sites. We further discuss the molecular details of NSAID binding to SA as a starting point for personalized medical treatments, particularly those that are focused on altering drug doses for patients who have specific metabolic disorders or take multiple drugs.

Figure 1.

Structures of the compounds used in this study; chiral centers are labeled with an asterisk. All of the chiral compounds (etodolac, ibuprofen, and ketoprofen) are used in medicine as racemic mixtures and were used as such in this study.

2. RESULTS

2.1. Structure Determination and Overview.

All crystals of ESA grew in the P6₁ space group and contained one ESA molecule in the asymmetric unit (Table 1). The quality of electron density for all determined structures and built models can be inspected interactively at https://molstack.bioreproducibility.org/p/IR6d/. The models are complete, except for the first two or three residues that were not located in the electron density maps of these structures. The determined structures revealed these NSAIDs bound in several binding sites, which are located in all three ESA domains (Figure 2). To better describe the locations of NSAID binding sites, we use an expanded nomenclature of sites that was previously proposed by Handing et al.³⁰ Names of sites 1–6 were previously used in literature (Zsila, Handing et al.),^30,52 while drug sites 7–10 are introduced in this report. The ESA–NSAID structures were compared to each other and ligand-free ESA and HSA structures using the Dali server,⁵³ and RMSD values between the aligned Cα atoms are presented in Table S1. The ESA–NSAID structures presented in this report essentially have an identical fold to previously determined structures of ESA and show low RMSD values when compared to the ligand-free ESA and HSA structures, indicating similar structural morphology.

Table 1.

Data Collection, Structure Refinement, and Structure Quality Statistics^a

name	etodolac	ketoprofen	ibuprofen	nabumetone	6-MNA
PDB ID	5V0V	6U4R	6U4X	6CI6	6U5A
diffraction images DOI	10.18430/M35V0V	10.18430/M36U4R	10.18430/M36U4X	10.18430/M36CI6	10.18430/M36U5A
Data Collection Statistics
resolution (Å)	50.00–2.45 (2.49–2.45)	50.00–2.45 (2.49–2.45)	50.00–2.25 (2.29–2.25)	50.00–2.80 (2.85–2.80)	50.00–2.65 (2.70–2.65)
beamline	23-ID-D	21-ID-F	19-ID	23-ID-D	21-ID-F
wavelength (Å)	0.979	0.979	0.979	0.979	0.979
space group	P61	P61	P61	P61	P61
unit-cell dimensions (Å)	a = b = 94.2 c = 141.8	a = b = 95.5 c = 141.7	a = b = 95.2 c = 141.9	a = b = 93.9 c = 140.8	a = b = 94.3 c = 142.0
protein chains in the ASU	1	1	1	1	1
completeness (%)	99.9 (98.6)	99.9 (100.0)	99.7 (97.0)	100.0 (100.0)	100.0 (100.0)
no. of unique reflections	26311 (1304)	26761 (1340)	34270 (1661)	17321 (871)	20955 (1057)
redundancy	4.0 (3.5)	7.7 (6.8)	10.7 (6.6)	5.1 (4.9)	7.6 (6.5)
⟨I⟩/⟨σ(I)⟩	14.0 (1.2)	17.3 (1.4)	21.8 (1.1)	14.1 (2.1)	18.9 (1.1)
CC 1/2	(0.75)	(0.66)	(0.68)	(0.82)	(0.50)
Rmerge	0.088 (0.574)	0.119 (1.335)	0.113 (1.119)	0.115 (0.660)	0.113 (1.700)
Rmeas	0.102 (0.629)	0.128 (1.445)	0.119 (1.204)	0.129 (0.742)	0.121 (1.842)
Refinement Statistics
Rwork/Rfree	0.180/0.238	0.182/0.233	0.191/0.238	0.182/0.256	0.185/0.256
bond lengths rmsd (Å)	0.007	0.002	0.002	0.007	0.003
bond angles rmsd (°)	1.1	1.2	1.1	1.1	1.3
mean B value (Å2)	65	48	44	64	55
mean B value for ligands (Å2)	(S)-etodolac: 86.2 (DS1) 89.1 (DS7) (R)-etodolac: 85.7 (DS3) 78.3 (DS7)	(S)-ketoprofen: 64.1 (DS4) 69.5 (DS6) 36.4 (DS10)	(S)-ibuprofen: 49.1 (DS2) 79.4 (DS4)	nabumetone: 65.7 (DS2) 85.3 (DS6)	6-MNA: 46.0 (DS2) 82.7 (DS6) 74.2 (DS7)
no. of protein atoms	4512	4578	4582	4524	4548
mean B value for protein (Å2)	64	49	44	66	56
no. of water molecules	164	209	266	96	96
mean B value for water molecules (Å2)	55	38	40	48	35
Clashscore	0.33	0.98	3.25	0.77	3.09
MolProbity score	0.62	0.81	1.12	0.74	1.10
rotamer outliers (%)	0.00	0.20	0.40	0.00	0.20
Ramachandran outliers (%)	0.00	0.00	0.00	0.00	0.00
Ramachandran favored (%)	98.10	97.93	98.27	98.10	98.10

^aValues in parentheses are for the highest resolution shell. Ramachandran plot statistics are calculated by MolProbity. DS1-DS10 refers to drug-binding sites 1–10.

Figure 2.

Location of the NSAID binding sites in ESA structures reported in this study (PDB ID: 5V0V, 6U4R, 6U4X, 6CI6, 6U5A). Molecules are shown with atoms as black spheres. Each site is labeled on only one panel. Domains are labeled with Roman numerals (I, II, III) and subdomains with letters (e.g., IA), with each subdomain shown in a different color; alternative names of the binding sites are in parentheses; FA stands for the fatty acid-binding site.

2.2. NSAID Binding Sites in ESA.

2.2.1. Complex with Etodolac.

Well-defined electron density indicates four etodolac molecules bound to the ESA molecule in drug sites 1, 3, and 7 (PDB ID: 5V0V; resolution 2.45 Å; Figure 3). Residues involved in etodolac binding to ESA, including details pertaining to salt bridges and hydrogen bonds, are listed in Table 2. The etodolac molecule bound to drug site 1 (also known as Sudlow site I) was modeled as (S)-etodolac, although the possibility of a small fraction of (R)-etodolac in the same position cannot be excluded. Drug site 1 is located between subdomains IIA and IIB and is formed by the helices of subdomain IIA and a seven-residue loop from subdomain IB. This site is predominantly hydrophobic in nature but contains hydrophilic residues as well; this mixed character allows for favorable hydrophobic interactions between drug molecules and amino acid residues as well as for possible hydrogen bonds, van der Waals interactions, and charge–charge interactions (e.g., a salt bridge with a hydrogen bond between (S)-etodolac’s carboxylate group and the side-chain guanidine group of Arg256 and a hydrogen bond with the carboxylate group of Glu195). The etodolac molecule bound at drug site 3 (also known as a major oncological drug-binding site)³¹ was clearly identified as (R)-etodolac. Drug site 3 is located in subdomain IB, between α-helices IB-h2, IB-h3, and IB-h4 and a loop, and can be described as a hydrophobic groove with several amino acid residues having the possibility to form some hydrogen bonds (e.g., Gln122, Arg144, His145).³¹ At this site, (R)-etodolac forms a bifurcated hydrogen bond with His145. Drug site 7 is located in domain II, between subdomains IIA (helices IIA-h2 and h3) and IIB (helices IIB-h2 and h3); this drug site is large and can be split into two subsites, each of which binds one enantiomer of etodolac. One was modeled as (S)-etodolac (a small fraction of (R)-etodolac might be present) and the other as (R)-etodolac. Most of the interactions in these subsites are hydrophobic in nature, but Arg208 forms a salt bridge with Asp323, which helps to close this cavity and is known to contribute to fatty acid binding. The (R)-etodolac molecule in drug site 7 is situated roughly 4 Å away from the (S)-etodolac molecule in the same site. Both (R)- and (S)-etodolac are surrounded mostly by hydrophobic residues that contribute to interactions with etodolac’s rings. However, the carboxylate group of (R)-etodolac forms a salt bridge (with a hydrogen bond) with Lys350’s side-chain amino group and hydrogen bonds with Ser479’s backbone nitrogen and side-chain oxygen. The carboxylate group of (S)-etodolac forms a salt bridge (with a hydrogen bond) with Lys211’s side-chain amino group. In this structure, four Tris molecules and seven sulfate ions from the crystallization conditions were also observed.

Figure 3.

NSAIDs binding sites (PDB ID: 5V0V, 6U4R, 6U4X, 6CI6, 6U5A) with the omit electron density map (mF_o–DF_c map, calculated after 10 refinement cycles without a drug, RMSD = 2.5) presented in green and red (positive and negative contours). All NSAIDs are shown in stick representation with carbon atoms in yellow, oxygen atoms in red, and nitrogen atoms in blue. Colors of helices correspond to colors in Figure 2. Helices are labeled with the subdomain name and helix number. The electron density can be inspected interactively at https://molstack.bioreproducibility.org/c/WFY4.

Table 2.

Drug-Binding Sites, Residues That Participate in Binding the Particular Drug, and Hydrophilic Interactions Involved in ESA-NSAIDs Complexes Reported Herein^a

DS	subdomains	drug	residues	salt bridges and hydrogen bonds
1	IIA and IIB	(S)-etodolac*	Tyr149, Glu152, Lys194, Glu195, Lys198, Trp213, Leu218, Lys221, Phe222, Ile233, Leu237, Val240, His241, Arg256, Leu259, Ala260, Ile263, Ser286, Ile289, Ala290, Glu291	(S)-Etodolac’s carboxylate group forms a salt bridge (with a hydrogen bond to NE atom) with the side-chain guanidine group of Arg256 and a hydrogen bond with the carboxylate group of Glu195 (OE2 atom). A water molecule acts as a hydrogen bond mediator between (S)-etodolac’s nitrogen and a nitrogen atom (NE) of Arg256’s side-chain.
2	IIIA	(S)-ibuprofen	Leu386, Asn390, Leu406, Arg409, Tyr410, Lys413, Ala414, Val417, Leu422, Ile425, Leu429, Leu452, Leu456, Leu459, Ile472, Arg484, Phe487, Ser488, Leu490	(S)-Ibuprofen’s carboxylate group forms a salt bridge (with a hydrogen bond to the NH1 atom) with a side-chain guanidine group of Arg409 and a hydrogen bond with Tyr410’s hydroxyl group (OH).
		6-MNA	Leu386, Val387, Asn390, Cys391, Phe402, Leu406, Arg409, Tyr410, Lys413, Leu429, Val432, Gly433, Cys436, Cys437, Ser448, Leu452, Arg484, Ser488	6-MNA’s carboxylate group forms a salt bridge with the side-chain guanidine group of Arg409 and hydrogen bonds with hydroxyl groups of Tyr410 (OH) and Ser488 (OG).
		nabumetone	the same as for 6-MNA at DS2	Nabumetone’s carbonyl group (atom O1) forms a hydrogen bond with the hydroxyl group of Tyr410 (atom OH).
3	IB	(R)-etodolac	Lys114, Leu115, Pro117, Gln122, Tyr137, Glu140, Val141, Arg144, His145, Tyr160, Leu181, Leu184, Lys185, Ile188, Ile189	(R)-Etodolac’s ring oxygen and carboxylate group form a bifurcated hydrogen bond with the side-chain nitrogen atom (NE2) of His145.
4	IA and IB	(S)-ibuprofen*	Lys17, His18, Lys20, Gly21, Leu24, Asp131, Leu134, Gly135, Leu138, Leu154, Ala157, Glu158, Leu283	none
		(S)-ketoprofen	Lys17, Lys20, Gly21, Leu24, Phe36, Val40, Val43, Asn44, Asp129, Asp131, Lys132, Leu134, Gly135, Leu138	(S)-Ketoprofen’s carboxylate group forms a salt bridge (with a hydrogen bond) with Lys20’s side-chain amino group (NZ atom).
6	IIIA and IIIB	(S)-ketoprofen	Leu393, Val397, Asp401, Asn404, Ala405, Val408, Leu528, Lys540, Glu541, Leu543, Lys544, Leu547	(S)-Ketoprofen’s carboxylate group forms a salt bridge with Lys540’s side-chain amino group.
		6-MNA	the same as for (S)-ketoprofen at DS6	6-MNA’s carboxylate group forms a salt bridge (with a hydrogen bond) with Lys540’s side-chain amino group (NZ atom).
		nabumetone	the same as for (S)-ketoprofen at DS6	The oxygen atom (O2′) of nabumetone’s methoxy functional group forms a hydrogen bond with Asn404’s side-chain nitrogen atom (ND2).
7	IIA and IIB	(R)-etodolac	Phe205, Arg208, Ala209, Ala212, Asp323, Leu326, Gly327, Leu330, Leu346, Ala349, Lys350, Ser479, Leu480, Ala481, Glu482	(R)-Etodolac’s carboxylate group forms a salt bridge (with a hydrogen bond) with Lys350’s side-chain amino group (NZ atom), a hydrogen bond with Ser479’s backbone nitrogen (N), and a hydrogen bond with Ser479’s side-chain oxygen (OG).
		(S)-etodolac*	Arg208, Lys211, Ala212, Val215, Phe227, Ser231, Thr235, Asp323, Gly327, Leu330	(S)-Etodolac’s carboxylate group forms a salt bridge (with a hydrogen bond) with Lys211’s side-chain amino group (NZ atom).
		6-MNA	Arg208, Ala209, Ala212, Asp323, Leu326, Gly327, Leu330, Leu346, Ala349, Lys350, Ser479, Leu480, Ala481	6-MNA’s carboxylate group forms a salt bridge with Lys350’s side-chain amino group and hydrogen bonds with the hydroxyl group of Ser479 (OG) and with the backbone nitrogen atoms (N) of Leu480 and Ala481.
10	IA	(S)-ketoprofen	Ile7, Phe19, Val23, Ala26, Phe27, Val46, Phe49, Leu66, His67, Leu69, Phe70, Lys73, Gly247, Asp248, Leu249, Leu250, Glu251	(S)-Ketoprofen’s carboxylate group forms hydrogen bonds with Glu251’s side-chain and the backbone nitrogen atoms (N) of Leu249 and Leu250.

^aAn asterisk next to the drug name indicates that a small fraction of the opposite enantiomer may be present in this location.

2.2.2. Complex with Nabumetone.

In the crystal structure of ESA in complex with nabumetone (PDB ID: 6CI6; resolution 2.80 Å), the electron density clearly indicates the binding of two nabumetone molecules to drug sites 2 and 6 (Figure 3). Regions with a density that could not be unambiguously interpreted (observed in drug sites 3, 4, and 9) were modeled as unknown atoms or ions (UNX) and are likely due to a loosely bound drug, remnants of ESA purification from horse serum, or crystallization components. All of the residues involved in nabumetone binding to ESA are listed in Table 2. Nabumetone binds to drug site 2 (also known as Sudlow site II), which is located in subdomain IIIA and composed of six α-helices. Most of the interactions between the ligand and amino acid residues in this site are hydrophobic in nature (only one hydrogen bond between nabumetone’s carbonyl group and hydroxyl group of Tyr410 was observed). Nabumetone also binds to the predominantly hydrophobic drug site 6, which is situated in domain III, between subdomains IIIA (helices IIIA-h1 and h2) and IIIB (helix IIIB-h1). The electron density in this site is not as strong as that observed at drug site 2 in this structure but is still well-defined and supports the presence of a nabumetone molecule, which is stabilized by hydrophobic interactions and one hydrogen bond (between nabumetone’s methoxy functional group and Asn404’s side-chain nitrogen). The structure of ESA in complex with nabumetone also contains several other small molecules: one Tris molecule, five sulfate ions, and one molecule of fatty acid, which was modeled as nonanoic acid.

2.2.3. Complex with 6-MNA.

The active metabolite of nabumetone, 6-MNA, binds to drug sites 2, 6, and 7 in the ESA-6-MNA structure (PDB ID: 6U5A; resolution 2.65 Å; Figure 3). The 6-MNA molecule located in drug site 2 has the same orientation as the nabumetone molecule and is surrounded by the same residues (Table 2). The main difference between the binding mode of both molecules is that 6-MNA’s carboxylate group forms a hydrogen bond not only with the hydroxyl group of Tyr410 but also with the hydroxyl group of Ser488 and forms a salt bridge with the side-chain guanidine group of Arg409. The 6-MNA molecule located in drug site 6 has its position and interactions with surrounding residues nearly identical (except for a hydrogen bond with Asn440 replaced by a salt bridge with Lys540) to those observed for the nabumetone molecule in the same site (PDB ID: 6CI6, see section 2.2.2). The 6-MNA molecule bound to drug site 7 partially overlaps with the position of the (R)-etodolac molecule described above (PDB ID: 5V0V, see section 2.2.1). The subsite of drug site 7 to which 6-MNA binds has a mixed character consisting of both hydrophobic and hydrophilic residues (Table 2). At this subsite, the 6-MNA molecule’s carboxylate group forms a salt bridge with Lys350’s side-chain amino group and hydrogen bonds with the hydroxyl group of Ser479 and with the backbone nitrogen atoms of Leu480 and Ala481. Nabumetone was not observed to bind to site 7. The ESA-6-MNA structure also contains four sulfate ions, which were observed on the surface of the protein.

2.2.4. Complex with Ketoprofen.

Three molecules of ketoprofen were found to bind to ESA in drug sites 4, 6, and 10 (PDB ID: 6U4R; resolution 2.45 Å). All of them were modeled as (S)-ketoprofen (Figure 3); no (R)-ketoprofen molecules were located in this structure. Drug site 4 is located in domain I between α-helices IA-h2, IA-h3, IB-h2, and IB-h3. Most of the residues comprising this site contribute to hydrophobic interactions with the (S)-ketoprofen molecule; however, a salt bridge (with a hydrogen bond) between (S)-ketoprofen’s carboxylate group and Lys20’s side-chain amino group is observed (Table 2). Another (S)-ketoprofen molecule is bound at drug site 6 in the same mode as the molecule of nabumetone (PDB ID: 6CI6, see section 2.2.2), although the electron density is weaker for this (S)-ketoprofen molecule. The aromatic rings of (S)-ketoprofen are placed in the hydrophobic cavity, while its carboxylate group is directed toward the protein–solvent interface and forms a salt bridge with Lys540’s side-chain amino group. The electron density for the carboxylate and methyl groups is clearly weaker than for aromatic rings, which may suggest that this part of the drug is flexible and can rotate when bound at this site. The third (S)-ketoprofen molecule is bound to drug site 10. This site is located in subdomain IA, buried between α-helices h1, h2, h3, and h4 and overlaps with the previously characterized second fatty acid binding site (FA2). Ketoprofen is the second FDA-approved drug known to bind in this site (the first being halothane). Though the (S)-ketoprofen molecule in this site is located in the hydrophobic cavity, its carboxylate group forms hydrogen bonds with Glu251’s side-chain and the backbone nitrogen atoms of Leu249 and Leu250 (Table 2). The ESA–ketoprofen structure also contains three sulfate molecules, one molecule of fatty acid (modeled as nonanoic acid, bound at drug site 2), and two molecules modeled as unknown atoms or ions (UNX).

At drug site 1, weak electron density was observed and may suggest the binding of another (a fourth) ketoprofen molecule at this position, which was recently reported to be the only binding site for (R)-ketoprofen in BSA (PDB ID: 6QS9⁵⁰). However, the positive region of electron density at drug site 1 in our structure could not be unambiguously assigned to any particular enantiomer of ketoprofen (electron density for a part of the ligand is missing) and was modeled as an unknown ligand (UNL). Surprisingly, drug sites 4, 6, and 10 in the BSA–ketoprofen structure are unoccupied, which might be caused by the following: different crystallization conditions (18% PEG MME 5000, 0.2 M ammonium chloride, 0.1 M MES pH 6.5), space group (C2), and lower ketoprofen concentration. (S)-Ketoprofen was also recently reported to bind to drug sites 2 and 6 in LSA (PDB ID: 6OCK).⁵¹ Both LSA and ESA bind (S)-ketoprofen at drug site 6, but the drug molecule bound at this site has a different conformation in each structure, likely due to different residues comprising the binding sites (pairwise sequence identity and similarity for ESA/LSA are 71.0% and 85.2%, respectively)⁵⁴ and different conformations of the proteins. The structure of the ESA–ketoprofen complex reported in this study was obtained from a solution containing 0.2 M lithium sulfate, 2.0 M ammonium sulfate, and 0.1 M Tris buffer at pH 7.4 (P6₁ space group), while the LSA–ketoprofen complex was crystallized from 8% PEG 400, 16% PEG 3350, 0.2 M ammonium acetate, and 0.1 M Tris buffer at pH 8.5 (P2₁2₁2₁ space group). Moreover, LSA was defatted before crystallization, and the ketoprofen molecule was shown to bind to LSA at drug site 2 (PDB ID: 6OCK), which is occupied by a fatty acid molecule in our structure (PDB ID: 6U4R).⁵¹ Drug site 4 in the LSA structure is occupied by a molecule of polypropylene, and drug site 10 is unoccupied.

2.2.5. Complex with Ibuprofen.

In the structure of ESA complexed with ibuprofen (PDB ID: 6U4X; resolution 2.25 Å), two molecules of (S)-ibuprofen were found to bind to drug sites 2 and 4 (Figure 3); no (R)-ibuprofen molecules were located in this structure. At drug site 2, the hydrophobic part of the (S)-ibuprofen molecule is bound inside the hydrophobic cavity, while (S)-ibuprofen’s carboxylate group forms a salt bridge (with a hydrogen bond) with the guanidine moiety of Arg409’s and a hydrogen bond with the hydroxyl group of Tyr410 (Table 2). The electron density observed for the (S)-ibuprofen molecule at drug site 4 is weaker than the density observed in drug site 2 but still clearly indicates drug binding, although the possibility of a small fraction of (R)-ibuprofen in the same position cannot be excluded. The (S)-ibuprofen molecule in drug site 4 partially overlaps with the position of the (S)-ketoprofen molecule from the structure described above (PDB ID: 6U4R, see section 2.2.4) and is mostly stabilized by hydrophobic interactions with the surrounding amino acid residues (Table 2). The weak density observed for (S)-ibuprofen’s carboxylate group may suggest that this functional group is flexible. The ESA-ibuprofen structure also contains 10 sulfate ions and an unknown ligand (UNL). We believe that the unknown ligand is likely a sugar molecule, based on the shape of the observed density and its location, which is known to bind sugars in HSA (see PDB ID: 4IW2).⁵⁵ No sugars were added to ESA during purification or crystallization, so the presence of a sugar molecule in the structure is likely the result of its presence in the blood and binding to ESA under physiological conditions.

In a recently published structure of ESA in complex with ibuprofen (PDB ID: 6OCI), which was obtained from different crystallization conditions (75% Tacsimate at pH 6.0) but the same space group (P6₁), (S)-ibuprofen was found to bind to drug sites 2 and 7.⁵¹ These two sites are known to bind (S)-ibuprofen in HSA (Figure 4). We compared this structure (PDB ID: 6OCI) with our structure (PDB ID: 6U4X; crystallization conditions: 0.2 M lithium sulfate, 2.4 M ammonium sulfate, 0.1 M Tris buffer pH 7.4) and observed two different binding modes of (S)-ibuprofen at drug site 2 (Figure S1). In both structures, (S)-ibuprofen’s carboxylate group occupies roughly the same position, while its hydrophobic parts are oriented in opposite directions. The only amino acid residue that has a different conformation in these structures, is Tyr410. At drug site 4, where we observed (S)-ibuprofen binding to ESA, Zielinski et al. observed binding of succinic acid (a component of Tacsimate), which had potentially prevented binding of ibuprofen at this site.⁵¹ At drug site 7, electron density for ibuprofen was not observed in our ESA structure.

Figure 4.

Comparison of ibuprofen binding in ESA (PDB ID: 6U4X) and HSA (PDB ID: 2BXG) in drug sites 2 and 7. Carbon atoms in ESA and HSA are shown in green and gray, respectively. Oxygen atoms are red, and nitrogen atoms are blue; ibuprofen molecules are shown with carbon atoms in yellow (ESA structure) and in gray (HSA structure). Residue numbers correspond to positions in ESA; the naming scheme is as follows: residue from ESA, residue number, residue from HSA (if different). At drug site 4, the electron density for ibuprofen was observed only in the ESA structure (PDB ID: 6U4X); in the HSA–ibuprofen structure (PDB ID: 2BXG), drug site 4 is unoccupied (for comparison of the site, see Figure S2).

2.3. Conservation of NSAIDs Binding Sites in ESA/HSA.

Human and equine SAs have very high pairwise sequence identity/similarity (76.1%/86.2%, respectively) and secondary structure conservation, allowing these SAs to have a similar function in both organisms.⁵⁶ This similarity leads to the expectation of drug-binding sites being similar or identical, and drugs likely bind to the same sites in SA from these species. However, only a few drugs have been cocrystallized with both ESA and HSA, which does not provide enough evidence to confirm this hypothesis. Among the drugs whose complexes with ESA are reported in this study, only ibuprofen’s binding sites in HSA were characterized previously (PDB ID: 2BXG).³² (S)-Ibuprofen binds to drug sites 2 and 7 in HSA and to drug sites 2 and 4 in ESA (Figure 4). All of the SA drug-binding sites found to bind NSAIDs in this study were previously shown to bind at least one other drug (Table 3).

Table 3.

Summary of SA Drug-Binding Sites and FDA-Approved Drugs That Were Reported to Bind in These Sites Based on Crystal Structures of the Respective Complexes^a

drug site 1 (Sudlow site I, FA7)	drug site 2 (Sudlow Site II, FA3, FA4)	drug site 3 (FA1)	drug site 4	drug site 5	drug site 6	drug site 7 (FA6)	drug site 8 (FA5)	drug site 9 (Cleft, FA8, FA9)	drug site 10 (FA2)
amantadine (HSA: 3UIV86)	aripiprazole (HSA: 6A7P87)	azapropazone (HSA: 2BXI,32 2BX832)	cetirizine (ESA: 5DQF30)	etoposide (HSA: 4LB931)	diclofenac (ESA: 4ZBQ,65 4ZBR,65 5DBY;65 OSA: 6HN0; CSA: 6HN1)	cetirizine (ESA: 5DQF30)	fusidic acid (HSA: 2VUF88)	diclofenac (OSA: 6HN0)	halothane (HSA: 1E7C89)
aspirin/salicylic acid (HSA: 3B9M,63 2130,62 2I2Z62)	diazepam (HSA: 2BXF32)	bicalutamide (HSA: 4LA031)	diclofenac (OSA: 6HN0; CSA: 6HN1)		ketoprofen (ESA: 6U4R; LSA: 6OCK51)	diclofenac (HSA: 4Z69;64 OSA: 6HN0; CSA: 6HN1)	propofol (HSA: 1E7A89)	iodipamine (HSA: 2BXN32)	ketoprofen (ESA: 6U4R)
azapropazone (HSA: 2BXI,32 2BX8,32 2BXK32)	diclofenac (ESA: 4ZBQ;65 OSA: 6HN0; CSA: 6HN1)	diclofenac (HSA: 4Z69,64 OSA: 6HN0; CSA: 6HN1)	ibuprofen (ESA: 6U4X)		nabumetone/6-MNA (ESA: 6CI6, 6U5A)	diflunisal (HSA: 2BXE32)	thyroxine (HSA: 1HK1,42 1HK2,42 1HK342)	thyroxine (HSA: 1HK4,42 1HK542)
diclofenac (HSA: 4Z6964)	diflunisal (HSA: 2BXE32)	etodolac (ESA: 5V0V)	ketoprofen (ESA: 6U4R)		naproxen (LSA: 4PO067)	etodolac (ESA: 5V0V)
diflunisal (HSA: 2BXE32)	halothane (HSA: 1E7B89)	fusidic acid (HSA: 2VUF88)	testosterone (ESA: 6MDQ57)		oxyphenbutazone (HSA: 2BXO32)	halothane (HSA: 1E7B,89 1E7C89)
etodolac (ESA: 5V0V)	ibuprofen (HSA: 2BXG;32 ESA: 6U4X, 6OCI51)	idarubicin (HSA: 4LB231)				ibuprofen (HSA: 2BXG;32ESA: 6OCI51)
halothane (HSA: 1E7C89)	ketoprofen (LSA: 6OCK51)	indomethacin (HSA: 2BXM,32 2BXQ32)				6-MNA (ESA: 6U5A)
Indomethacin (HSA: 2BXK,32 2BXM,32 2BXQ32)	nabumetone/6-MNA (ESA: 6CI6, 6U5A)	lidocaine (HSA: 3JQZ90)				naproxen (ESA: 4ZBR;654OT2;67 BSA: 4OR0;67 LSA: 4PO067)
iodipamine (HSA: 2BXN32)	naproxen (ESA: 4ZBR,65 5DBY;65 4OT2;67 BSA: 4OR0;67 LSA: 4PO067)	naproxen (HSA: 2VDB66)				testosterone (ESA: 6MDQ57)
ketoprofen (BSA: 6QS950)	phenylbutyric acid (HSA: 5YOQ91)	salicylic acid (HSA: 3B9M,63 2I3062)
naproxen (BSA: 4OR067)	propofol (HSA: 1E7A89)	teniposide (HSA: 4L9Q31)
oxyphenbutazone (HSA: 2BXB,32 2BXO32)	suprofen (ESA: 6OCJ;51 LSA: 6OCL51)	zidovudine (HSA: 3B9L63)
phenylbutazone (HSA: 2BXC,32 2BXP,32 2BXQ32)	thyroxine (HSA: 1HK1,42 1HK2,42 1HK342)
thyroxine (HSA: 1HK1,42 1HK2,42 1HK342)
warfarin (HSA: 2BXD,32 1H9Z,92 1HA292)
zidovudine (HSA: 3B9L,63 3B9M63)

^aDrugs whose complexes with SA are presented in this publication are in bold. PDB IDs for structures of HSA, ESA, BSA, and LSA complexes with drugs available in the PDB are shown in parentheses. Alternative names for binding sites are also indicated in parentheses in headings.

2.3.1. Drug Site 1.

Eighty-six percent of the residues involved in the binding of (S)-etodolac to drug site 1 (Sudlow site I, also known as FA7) in ESA are conserved in HSA (Figure S3). Eighteen residues are conserved (Tyr149, Glu152, Lys194, Lys198, Trp213, Leu218, Phe222, Leu237, Val240, His241, Arg256, Leu259, Ala260, Ile263, Ser286, Ile289, Ala290, Glu291), and three are different (Glu195Gln, Lys221Arg, Ile233Leu; the naming scheme is as follows: residue from ESA, residue number, residue from HSA). Two of these differences involve the exchange of a hydrophilic residue for another hydrophilic residue, which causes a change in the overall charge of the cavity but does not directly affect observed interactions between the (S)-etodolac molecule and ESA. The third difference is hydrophobic-for-hydrophobic. Despite these small changes in drug site 1, the residues involved in the binding of (S)-etodolac exhibit no significant conformational differences when the ESA–etodolac complex is compared to structures of ligand-free ESA (PDB ID: 3V08) and HSA (PDB ID: 4K2C). Drug site 1 has been characterized for HSA, ESA, and BSA and has been shown to bind amantadine, aspirin/salicylic acid, azapropazone, diclofenac, diflunisal, etodolac, halothane, indomethacin, iodipamide, ketoprofen, naproxen, oxyphenbutazone, phenylbutazone, thyroxine, warfarin, and zidovudine (Table 3). All of these compounds, excluding halothane and amantadine, contain at least one aromatic ring and are highly hydrophobic.

2.3.2. Drug Site 2.

Drug site 2 (Sudlow site II, also known as FA3/FA4) of both HSA and ESA binds (S)-ibuprofen, but the binding mode is very different between species. In these structures, (S)-ibuprofen’s carboxylate group roughly occupies the same position and forms hydrogen bonds with Arg409 (Arg410 in HSA) and Tyr410 (Tyr 411); in HSA, there is also a third potential hydrogen bond with Lys414. However, the hydrophobic parts of the (S)-ibuprofen molecules are oriented in opposite directions. The whole binding site (consisting of residues involved in (S)-ibuprofen binding in either HSA or ESA) is 81% conserved between ESA and HSA; 22 residues are conserved (Leu386, Asn390, Cys391, Phe402, Leu406, Arg409, Tyr410, Lys413, Va417, Leu422, Leu429, Val432, Gly433, Cys437, Glu449, Leu452, Leu456, Leu459, Arg484, Phe487, Ser488, Leu490), and five are different (Val387Ile, Ala414Val, Ile425Val, Ser448Ala, Ile472Val). All of these modifications, excluding Ser448Ala, are hydrophobic-for-hydrophobic and do not cause any significant conformational changes in the structure of the binding site (Figure 4). Sixteen amino acid residues involved in the binding of nabumetone and its metabolite 6-MNA to drug site 2 in ESA are conserved in HSA (Leu386, Asn390, Cys391, Phe402, Leu406, Arg409, Tyr410, Lys413, Leu429, Val432, Gly433, Cys436, Cys437, Leu452, Arg484, Ser488), and two are altered (Val387Ile, Ser448Ala, Figure S2), resulting in 89% conservation of the residues comprising this site. The positions of nabumetone and 6-MNA in drug site 2 of ESA are nearly identical (Figure S4) and overlap with the position of the (S)-ibuprofen molecule in HSA; as a result of the high degree of amino acid residue conservation and overlapping positions of the three-drug molecules, we expect that nabumetone, 6-MNA, and (S)-ibuprofen will mostly exhibit the same interactions with the residues comprising drug site 2 in HSA. Drug site 2 has been characterized in HSA, ESA, BSA, ovine SA (OSA), caprine SA (CSA), and LSA and has been shown to bind the following drugs: aripiprazole, diazepam, diclofenac, diflunisal, halothane, ibuprofen, ketoprofen, nabumetone/6-MNA, naproxen, phenylbutyric acid, propofol, suprofen, and thyroxine (Table 3). With the exception of halothane, all of these drugs are aromatic and predominantly hydrophobic. Drug site 2 also can bind two fatty acid molecules (e.g., myristic acid) at the same time and can be divided into two subsites known as FA3 and FA4 as a result.

2.3.3. Drug Site 3.

Drug site 3 (also known as FA1 and the major oncological site) in ESA, which binds (R)-etodolac, significantly differs from the analogous site in HSA (53% of residues are conserved; see Figure S3). Eight of the amino acid residues comprising this site are conserved between ESA and HSA (Leu115, Tyr137, Glu140, Arg144, His145, Tyr160, Leu181, Leu184), and seven are different (Lys114Arg, Pro117Arg, Gln122Met, Val141Ile, Lys185Arg, Ile188Gly, Leu189Lys). Three of these differences (Pro117Arg, Gln122Met, Leu189Lys) significantly change the structure, character, and charge of one side of the binding site. The rest of the differences involve exchanges of hydrophobic residues to other hydrophobic residues (Val141Ile, Ile188Gly) or hydrophilic to hydrophilic (Lys114Arg, Lys185Arg). Drug site 3 has been characterized in HSA, ESA, OSA, and CSA and has been reported to bind the following drugs: azapropazone, bicalutamide, diclofenac, etodolac, fusidic acid, idarubicin, indomethacin, lidocaine, naproxen, salicylic acid, teniposide, and zidovudine (Table 3).

2.3.4. Drug Site 4.

There is some variation in the set of residues involved in the binding of (S)-ketoprofen and (S)-ibuprofen in drug site 4 in ESA. Fifty-seven percent of the residues involved in (S)-ketoprofen binding at drug site 4 are conserved (eight residues are conserved: Lys20, Leu24, Phe36, Val40, Val43, Asn44, Leu134, Leu138; six are different: Lys17Glu, Gly21Ala, Asp129Asn, Asp131Glu, Lys132Thr, Gly135Lys; Figure S2). Some of the altered residues are involved only in hydrophobic interactions with the drug (aliphatic part of Lys17, Gly21), and their alterations should not affect drug binding. The Asp129Asn alteration will change the overall charge of the cavity and may change the orientation of (S)-ketoprofen’s carboxylate group. Modification of Lys132 to Thr will cause a loss of the hydrophobic interactions between the aliphatic part of lysine and the drug but introduces a hydroxyl group that may be involved in a potential hydrogen bond with (S)-ketoprofen’s carboxylate group (upon changes of (S)-ketoprofen’s orientation). The loss of the hydrophobic interactions with Lys132 can be compensated by mutations, such as the introduction of an additional carbon atom through the Asp131Glu change. The most significant structural changes may be caused by steric effects resulting from the Gly135Lys modification, potentially preventing ligand binding. At the same time, residues involved in binding (S)-ibuprofen, which partially overlaps with the location of (S)-ketoprofen, are only 54% conserved (seven residues are conserved: Lys20, Leu24, Leu134, Leu138, Leu154, Ala157, Leu283; six residues are altered: Lys17Glu, His18Asn, Gly21Ala, Asp131Glu, Gly135Lys, Glu158Lys). Most of these changes are of a hydrophilic residue for another hydrophilic residue or a hydrophobic residue for another hydrophobic residue, but they change the overall charge and size of the cavity. This binding site has been structurally characterized for ESA, OSA, and CSA, and five drugs have been reported to bind there: cetirizine, diclofenac, ibuprofen, ketoprofen, and testosterone.^30,57

2.3.5. Drug Site 6.

We observed three NSAIDs to bind to drug site 6 in ESA: nabumetone, 6-MNA, and (S)-ketoprofen. All three NSAIDs occupy roughly the same positions in this site. Drug site 6 is 75% conserved between ESA and HSA; nine amino acids are conserved (Leu393, Asn404, Ala405, Val408, Leu528, Lys540, Glu541, Leu543, Lys544), and three are altered (Val397Leu, Asp401Lys, Leu547Met; Figure S2). These alterations do not change the structure or character of the binding site significantly because they involve the exchange of one hydrophobic residue for another hydrophobic residue of a similar size (Val397Leu, Leu547Met) or involve amino acids that surround the drug molecules but are not involved in any direct interactions with them (Asp401Lys). Drug site 6 has been characterized for HSA, ESA, CSA, LSA, and OSA and has been reported to bind the following drugs: diclofenac, ketoprofen, nabumetone/6-MNA, naproxen, and oxyphenbutazone.

2.3.6. Drug Site 7.

The location of the 6-MNA molecule in drug site 7 (also known as FA6) of ESA overlaps with that of the (S)-ibuprofen molecule in HSA, and the amino acids involved in 6-MNA binding are 92% conserved when the sequence of HSA is compared to ESA (twelve conserved residues: Arg208, Ala209, Ala212, Asp323, Leu326, Gly327, Leu330, Leu346, Ala349, Lys350, Ser479, Leu480; one altered residue, Ala481Val; Figure S2). Despite high sequence and secondary structure conservation, the electron density for ibuprofen was not found at this binding site in ESA. The altered amino acid residue (Ala481Val) does not significantly change the character of the binding site because Ala481 is only involved in hydrophobic interactions with the drug. Both enantiomers of etodolac bind to drug site 7 as well; 18 amino acids comprising both etodolac subsites are conserved (Phe205, Arg208, Ala209, Lys211, Ala212, Val215, Phe227, Ser231, Thr235, Asp323, Leu326, Gly327, Leu330, Leu346, Ala349, Lys350, Ser479, Leu480), while two residues are altered (Ala481Val, Glu482Asn; 90% conservation; Figure S3). Glu482 is oriented away from the binding site, and its side chain is not involved in (R)-etodolac binding. Ala481 is involved in hydrophobic interactions with the molecule of (R)-etodolac, but its modification to valine should not disrupt these interactions. Thus, the general structure and character of this binding site are conserved between ESA and HSA. Drug site 7 has been characterized for HSA, ESA, OSA, CSA, LSA, and BSA and has been shown to bind the following drugs: cetirizine, diclofenac, diflunisal, etodolac, halothane, ibuprofen, naproxen, testosterone, and 6-MNA (Table 3).

2.3.7. Drug Site 10.

The (S)-ketoprofen binding site (drug site 10, also known as FA2) is 94% conserved between ESA and HSA (16 are conserved: Phe19, Val23, Ala26, Phe27, Val46, Phe49, Leu66, His67, Leu69, Phe70, Lys73, Gly247, Asp248, Leu249, Leu250, Glu251; one residue is altered, Ile7Val; Figure S2). The structure of the binding site is conserved between ESA and HSA, and the Ile7Val modification should not significantly affect the binding of (S)-ketoprofen because this residue is only involved in hydrophobic interactions with the drug. Drug site 10 was previously only known to exist in HSA and bind fatty acids and halothane (PDB ID: 1E7C).

3. DISCUSSION AND CONCLUSIONS

We have reported crystal structures of ESA complexed with four commonly used NSAIDs and the active metabolite of nabumetone: ibuprofen, ketoprofen, etodolac, nabumetone, and 6-MNA. The binding of many NSAIDs to SA has previously been investigated through nonstructural methods such as equilibrium dialysis. These experiments showed that ibuprofen, ketoprofen, etodolac, and 6-MNA are 99% bound to plasma proteins, mainly to SA, with micromolar affinities^58-61 (Table S2). Binding studies typically show a good correlation between drug binding by SA from human plasma and SA from other species (such as a dog, rat, or mouse³⁷) but do not provide conclusive information about the residues involved in binding. The lack of binding site identification prevents understanding how different small molecules compete for binding to SA, how small molecules are transported in vivo, and how SA-facilitated drug transport differs between species.

Prior to this study, crystal structures of only 30 FDA-approved drugs in complexes with SA had been determined (Table S3), which represents only a small fraction of the hundreds of pharmaceuticals that bind to SA.^12,28,37 Crystal structures of SA–NSAID complexes were previously reported for eleven NSAIDs: aspirin/salicylic acid (PDB ID: 2I2Z,⁶² 2I30,⁶² 3B9M⁶³), azapropazone (PDB ID: 2BXI,³² 2BX8,³² 2BXK³²), diclofenac (PDB ID: 4Z69,⁶⁴ 4ZBQ,⁶⁵ 4ZBR,⁶⁵ 5DBY,⁶⁵ 6HN0, 6HN1), diflunisal (PDB ID: 2BXE³²), ibuprofen (PDB ID: 2BXG,³² 6OCI⁵¹), indomethacin (PDB ID: 2BXK,³² 2BXM,³² 2BXQ³²), ketoprofen (PDB ID: 6QS9,⁵⁰ 6OCK⁵¹), naproxen (PDB ID: 2VDB,⁶⁶ 4ZBR,⁶⁵ 5DBY,⁶⁵ 4OR0,⁶⁷ 4PO0⁶⁷), oxyphenbutazone (PDB ID: 2BXB,³² 2BXO³²), phenylbutazone (PDB ID: 2BXC,³² 2BXP,³² 2BXQ³²), and suprofen (PDB ID: 6OCJ,⁵¹ 6OCL⁵¹). Our study has added nabumetone/6-MNA and etodolac to this list, thereby extending it to thirteen NSAIDs. These NSAIDs have been shown to bind to SA not only at drug sites 1 and 2, which were previously expected to bind most drugs, but also at drug sites 3, 4, 6, 7, 9, and 10 (Table 3).

A direct structural comparison of binding of the same drug to SA from different species is only possible when structures are available for each of the complexes. Among the ESA complexes with NSAIDs reported here, only ibuprofen was previously crystallized with HSA. In both HSA and ESA (the pairwise sequence identity/similarity is 76.1%/86.2%, respectively), (S)-ibuprofen binds to drug site 2 (ESA PDB ID: 6U4X; HSA PDB ID: 2BXG³²). (S)-Ibuprofen’s carboxylate groups occupy nearly identical positions, but the hydrophobic portions of the molecule go in opposite directions, which can be caused by the substitution of certain residues (e.g., Val387Ile and Ile425Val) in the binding site between ESA and HSA (Figure 4). (S)-Ibuprofen is also bound to drug site 7 in HSA and drug site 4 in ESA. Drug site 7 is highly conserved between HSA and ESA (86% of residues involved in (S)-ibuprofen binding and 92% residues interacting with etodolac; see Figure 4), which may suggest that binding of (S)-ibuprofen at this site in ESA is possible but not observed in this structure. Binding of (S)-ibuprofen to drug site 7 in ESA has been reported in other studies (PDB ID: 6OCI).⁵¹ On the other hand, only 54% of the residues involved in (S)-ibuprofen binding in drug site 4 are conserved between ESA and HSA, and (S)-ibuprofen may bind to this site only in ESA.

Our analysis of drug-binding sites that have been observed to bind NSAIDs in either ESA or HSA can be classified into three categories based on the level of conservation between ESA and HSA (Figure 4, Figures S3 and S4; drug names indicate NSAIDs that were observed to bind to particular sites in ESA in our study):

1. Highly conserved sites: drug site 1 (etodolac), drug site 2 (ibuprofen, nabumetone, and 6-MNA), drug site 7 (6-MNA and etodolac), and drug site 10 (ketoprofen) are 86%, 81–89%, 90–92%, and 94% conserved, respectively. Moreover, altered residues do not significantly change the binding site’s environment, leading to the expectation that ligands that bind to these sites in ESA will bind to the analogous sites in HSA.

2. Partially conserved sites: drug site 6 (ketoprofen, nabumetone, and 6-MNA; 75% conserved).

3. Significantly different sites: drug site 3 (etodolac) and drug site 4 (ibuprofen and ketoprofen). The conservation of residues in these sites between ESA and HSA is 53% and 54–57%, respectively.

In the ESA–ketoprofen structure reported here (PDB ID: 6U4R), we observe binding of three ketoprofen molecules to drug sites 4, 6, and 10; all drug molecules were modeled as the (S)-enantiomer. Furthermore, we observed weak electron density at drug site 1 that was modeled as an unknown ligand (UNL). Ketoprofen was recently reported to bind to the analogous site in BSA (pairwise sequence identity and similarity for ESA/BSA are 74.1% and 85.9%, respectively), but the electron density in our structure does not conclusively support binding of ketoprofen at this site. In the crystal structure of BSA in complex with ketoprofen (PDB ID: 6QS9⁵⁰), the drug molecule is modeled as the (R)-enantiomer and is only found at drug site 1. Unfortunately, the authors do not mention if a racemic mixture or the (R)-enantiomer alone was used for crystallization trials, and the electron density observed for the ketoprofen molecule in their structure is weak. In our opinion, these facts do not allow for an unambiguous interpretation of which enantiomer is bound. (S)-Ketoprofen was recently reported to bind to drug sites 2 and 6 in LSA (pairwise sequence identity and similarity for ESA/LSA are 71.0% and 85.2%, respectively).^51,54 (S)-Ketoprofen occupies roughly the same position at drug site 6 in both crystal structures (ESA PDB ID: 6U4R; LSA PDB ID: 6OCK) but has different orientations. Drug site 2 contains a fatty acid molecule in the structure of the ESA–ketoprofen complex. Drug site 4 in the LSA structure is occupied by a molecule of polypropylene, and drug site 10 is unoccupied.

The presence of multiple SA binding sites provides the flexibility that permits drugs, their metabolites, and close analogues to utilize different binding sites, as is the case of nabumetone, its metabolite 6-MNA, and nabumetone’s close structural analogue naproxen. Based on the crystal structures, nabumetone binds to ESA at drug sites 2 and 6, naproxen binds to ESA at drug sites 2 and 7 (PDB ID: 4OT2⁶⁷), and 6-MNA binds to all three of these drug sites (2, 6, and 7). The differences in binding sites among these three structurally related molecules may be explained by steric considerations and the presence of specific functional groups. Naproxen contains a branch, whereas nabumetone and 6-MNA are not branched; this may explain why naproxen does not bind to ESA at drug site 6. Naproxen and 6-MNA binding at drug site 7 may be explained by the presence of a carboxyl group in naproxen and 6-MNA; they form salt bridges with Lys350’s side-chain in drug site 7, whereas nabumetone’s ketone group cannot form such interactions. However, the differences in observed binding preferences could also be the result of different crystallization conditions and methods (naproxen was cocrystallized with ESA at pH 4.6; the nabumetone complex was obtained by soaking crystals that grew at pH 7.4). Naproxen has also been cocrystallized with HSA (PDB ID: 2VDB⁶⁶), BSA (PDB ID: 4OR0⁶⁷), and LSA (PDB ID: 4PO0⁶⁷). In SAs, from all of these species, naproxen binds to drug sites 2 and 7 in a mode similar to ESA and also binds to drug site 3 in HSA. However, the crystal structure of HSA in complex with naproxen was obtained in the presence of the GA module (the protein G-related albumin-binding module) and fatty acids, which may affect the availability of SA binding sites. Additionally, naproxen has been observed to bind to drug site 1 in BSA.

The observed differences may be attributed to intrinsic differences in NSAID binding to the SAs from these species but may also be a result of differences in crystallization conditions. Even for crystallization experiments involving the same SA, an impact of variations in crystallization conditions on preferred binding sites was observed for ESA in complex with ibuprofen (see section 2.2.5). Though the presence of electron density corresponding to a drug is the best evidence that a drug can bind to a particular binding site, the absence of electron density does not rule out the possibility that a drug could bind to that site under different conditions or a higher drug concentration. Furthermore, the concentration of a ligand used for crystal soaking or protein cocrystallization must often be much higher than its physiological range to see electron density in the determined structure. For instance, the final concentration of etodolac used for soaking in this study was 10 mM, while its plasma concentration in patients is between 0.05 and 0.1 mM.⁶⁸ Although the concentrations of drugs used for soaking are much higher than those found in the blood under normal dosages, a direct comparison between crystallization and blood concentrations is not straightforward because crystallization conditions are drastically different from physiological conditions. Additionally, a key point toward understanding why drugs may not bind to the sites to which they are expected to bind is the presence of endogenous metabolites (e.g., sugars, fatty acids). These molecules may compete with drugs for binding to SA. In the ESA–ketoprofen structure, the fatty acid molecule bound to drug site 2, which was occupied by (S)-ketoprofen in the LSA–ketoprofen structure (PDB ID: 6OCK), is likely the result of its presence in the blood and ESA sample used for crystallization.

Another important phenomenon that may affect a drug’s binding to SA and its free blood concentration is nonenzymatic glycosylation (glycation), which may occur during SA’s long circulation time. HSA has multiple Lys and Arg residues that are known to be glycated.^69-72 Depending on the method, it is estimated that up to 6% of the HSA in a healthy human is glycated, while in diabetic patients these values are 2–5 times higher.^69,71,73 Glycation has been shown to affect SA’s binding affinity for various drugs, including warfarin and (R)-cetirizine.^71,74,75 The binding of NSAIDs reported in this study also may be affected by glycation due to the presence of Lys and Arg residues that may be partially glycated in the binding sites (Table S4). However, the effect of glycation on NSAID binding remains unstudied.

Opposite enantiomers of the same drug may have preferences for different binding sites on SA. All of the drugs with chiral centers used in this study for the soaking of native ESA crystals (ibuprofen, ketoprofen, and etodolac) were racemic mixtures of the (R)- and (S)-enantiomers, which is the same formulation as the marketed drug. While modeling these NSAIDs into the electron density, the choice of which enantiomer was not immediately obvious in many cases. Both enantiomers of these drugs could initially be modeled in some of the binding pockets, but we tried to resolve the ambiguities by considering the fit to the 2mF_O-DF_C and mF_O-DF_C omit maps, B-factors of the drugs and drug-binding residues, and interactions with neighboring amino acid residues. We decided to model the (S)-enantiomer of ibuprofen at drug sites 2 and 4, (S)-ketoprofen at drug sites 4, 6, and 10, (S)-etodolac at drug sites 1 and 7, and (R)-etodolac at drug sites 3 and 7. However, based on our criteria, we cannot exclude the possibility of some (R)-ibuprofen binding at drug site 4. Likewise, some (R)-etodolac may be bound at the drug site 7 subsite, where we have modeled (S)-etodolac. It is highly probable that the electron density observed for drugs at those sites is a result of binding both enantiomers with different affinities. These observations seem to confirm the hypothesis that drug sites on SA do not have structural features that allow them to exclusively bind (S)- or (R)-enantiomers of drugs but that their stereoselectivity instead depends on the structure of the particular drug and experimental conditions.⁵¹ Previously, predictions about the location of binding sites were made based on nonstructural studies of drugs’ enantiomers binding to SA. For instance, both etodolac enantiomers were thought to bind to drug site 2, and only (R)-etodolac was thought to bind to drug site 1.⁵⁸ Our structural results clearly contradict this hypothesis. To further probe the binding of those drugs’ enantiomers to SA, structures of SA with each enantiomer should be determined to discern the location of (R)- and (S)-enantiomer binding sites. This will allow comparisons between these structures and that of ESA with the racemic mixture to understand the stereoselectivity of NSAID binding to SA in greater detail, which could further our understanding of how different enantiomers of drugs exhibit different effects in circulation. NSAIDs’ ability to inhibit cyclooxygenase (COX) is typically confined to enantiomers of the (S)-stereo-configuration,⁷⁶ and (R)-enantiomers of those drugs either are inactive or exhibit a different activity. (R)-Ibuprofen is not a COX inhibitor, but a significant fraction of it undergoes metabolic conversion to the (S)-enantiomer in the body.^77,78 A similar conversion of the (R)-enantiomer to the (S)-enantiomer is observed for (R)-ketoprofen, which is more analgesic than the (S)-enantiomer when acting alone.⁷⁹ For etodolac, (R)- to (S)-enantiomer conversion has not been observed in vivo;⁸⁰ (S)-etodolac shows anti-inflammatory effects, while (R)-etodolac exhibits gastroprotective effects.⁸¹

Information about the location of a drug’s binding may be a key element to understanding phenomena associated with drug coadministration and avoid adverse reactions due to drug–drug displacement. It is especially important for drugs whose margins of safety are small (such as warfarin) to avoid toxicity, which may be caused by drug displacement leading to increases in the free fraction of the drug.⁴⁴ So far, based on all available crystal structures, we can distinguish ten SA drug-binding sites, which can bind multiple drugs and other small molecules (Table 3). Eight of these sites were shown to bind at least three different drugs, and most of them overlap with some of the nine fatty acid sites. Drug site 1 (Sudlow site I) has the highest number of drugs reported to bind to it (16 drugs), including warfarin and several NSAIDs; interactions between warfarin and NSAIDs during coadministration may lead to gastrointestinal bleeding.⁸² Phenylbutazone, an NSAID that was shown to bind to drug site 1,³² is known to cause increases in the free fraction of etodolac by roughly 80%,⁶⁸ which yields a narrow safety margin and can lead to toxic effects like adverse gastrointestinal and hepatic reactions. The mechanism of interactions between etodolac and phenylbutazone is not known, but the structural analysis of their complexes with SA (phenylbutazone PDB IDs: 2BXC, 2BXP, and 2BXQ etodolac PDB ID: 5V0V) indicates that they share a common binding site (drug site 1), which may result in drug–drug displacement.

As a protein with crucial therapeutic implications, SA is at the frontier of uncharted knowledge yet to be uncovered. Specifically, obtaining more information about the location of drug-binding sites, which includes the amino acid residues involved in the binding of each particular drug, will be useful for the elucidation of phenomena observed in patients taking multiple drugs and patients with metabolic disorders. Our analysis of all known SA drug-binding sites and their potential modifications suggest that some patients may need higher or lower drug doses to achieve the desired therapeutic effect or avoid unexpected toxicity. The characterization of binding sites for the four NSAIDs and one metabolite presented in this report is but a starting point for further studies of commonly used drugs that have narrow therapeutic ranges, such as digoxin (a cardiovascular medication), theophylline (a bronchodilator), and acenocumarol (an anticoagulant), which are 25%, 40%, and 99%, respectively, bound to plasma proteins (mainly to SA).^83-85 The knowledge that we can obtain for such drugs will have tremendous implications in personalized medical treatments that require comprehensive knowledge of a number of organ systems, and a foundation for these treatments can be achieved through detailed structural studies of how these and other commonly prescribed drugs bind to SA.

4. EXPERIMENTAL SECTION

4.1. Materials.

ESA was purchased from Equitech-Bio (no. ESA62; ≥96% purity; Kerrville, TX, USA) as lyophilized powder and purified further as described in section 4.2.1. DMSO was purchased from Sigma-Aldrich (no. 276855; ≥99.9% purity; St. Louis, MO, USA), Trizma base from Sigma-Aldrich (Tris; no. T1503; ≥99.9% purity; St. Louis, MO, USA), ammonium sulfate from Sigma-Aldrich (no. A4915; ≥99.0% purity; St. Louis, MO, USA), lithium sulfate from Alfa Aesar (no. A10410; 99% purity; Ward Hill, MA, USA), etodolac (brand name: Lodine) from Santa Cruz Biotechnology (no. 204747; ≥98% purity; Dallas, TX, USA), ibuprofen (brand names: Advil, Motrin, Ebufac, Brufen) from Sigma-Aldrich (no. I4883; ≥98% purity; St. Louis, MO, USA), ketoprofen (brand names: Orudis, Oruvail) from Santa Cruz Biotechnology (no. 205359; ≥99% purity; Dallas, TX, USA), nabumetone (brand names: Relafen, Relifex) from Sigma-Aldrich (no. N6142; ≥99% purity; St. Louis, MO, USA), and 6-MNA from Sigma-Aldrich (no. CDS014591; ≥98% purity; St. Louis, MO, USA). The purity of all reagents was reported by vendors.

4.2. Structure Determination.

4.2.1. Protein Purification and Crystallization.

ESA was dissolved in a buffer containing 10 mM Tris (pH 7.5) and 150 mM NaCl. Size exclusion chromatography using a Superdex 200 column attached to an ÄKTA FPLC (GE Healthcare) was used to separate dimeric and monomeric fractions of ESA. The purification buffer was the same as the buffer in which protein being purified was dissolved. The final protein purity was above 95% as assessed by SDS-PAGE. The absorbance at 280 nm measured with a Nanodrop 2000 (Thermo Scientific) was used to estimate protein concentrations using the extinction coefficient (ε_280-ESA = 27 400 M⁻¹ cm⁻¹) and molecular weight (MW_ESA = 65 700 Da). Collected fractions of monomeric ESA were concentrated to 34 mg/mL (0.52 mM) using an Amicon Ultra Centrifugal Filter (Millipore Sigma, no. UFC903024) with a molecular weight cutoff (MWCO) of 30 kDa.

Protein crystallization was performed in 15-well hanging drop plates (EasyXtal 15-Well Tools, Qiagen). Aliquots of 1 or 2 μL of concentrated ESA were mixed with 1 or 2 μL of reservoir solution (0.2 M lithium sulfate, 1.8–2.4 M ammonium sulfate, 0.1 M Tris buffer pH 7.4) in ratios 1:1, 2:1, or 1:2. Ligands were added to the protein in different ways: etodolac, ketoprofen, nabumetone, and 6-MNA were prepared as 100 mM solutions in pure DMSO and added to crystallization drops containing ESA crystals to reach a final drug concentration of 3 mM (ketoprofen, 6-MNA) or 10 mM (nabumetone, etodolac) and then incubated for several hours before harvesting; ibuprofen powder was added directly to the crystallization drop containing crystals before incubation for 48 h. Harvested crystals were flash-cooled in liquid nitrogen using Paratone N as a cryoprotectant.

4.2.2. Data Collection and Structure Determination.

Diffraction data were collected at the 19-ID, 21-ID-F, and 23-ID-D beamlines at the Advanced Photon Source (Argonne National Laboratory), at 100 K. The collected data were processed, integrated, and scaled with HKL-3000 using corrections for radiation decay and anisotropic diffraction.^93-95 Resolution cut-offs were chosen based on values of CC1/2, ⟨I⟩/⟨σ(I)⟩, but first of all, on the quality of the maps.⁹⁶ The structures were determined by molecular replacement (PDB ID: 3V08 was used as the template) and refined with hydrogen atoms in riding positions using HKL-3000 seamlessly integrated with REFMAC,^93,94 Fitmunk,⁹⁷ and programs from the CCP4 package.^98-100 Coot^101,102 was used for manual inspection and correction of the model. If ligands, whether from the crystallization conditions, intrinsic binding to SA, or addition during ligand soaking, could not be unambiguously assigned to positive regions of electron density, these regions were modeled using unknown atoms or ions (UNX) or an unknown ligand (UNL), in keeping with previously suggested guidelines.^103,104 The ACHESYM server was used for the standardized placement of the protein models in the unit cell.¹⁰⁵ The TLS Motion Determination server was used to determine TLS groups for use in structure refinement.¹⁰⁶ The TLS parameters were kept if confirmed by a significantly improved R_free and the Hamilton R-factor ratio test¹⁰⁷ as implemented in HKL-3000. All decisions made during structure refinement adhered to recently published, state-of-the-art guidelines.¹⁰³ All enantiomeric species used for soaking of ESA crystals (ibuprofen, ketoprofen, and etodolac) were used in a racemic mixture, which is the same as commercially available formulations of these drugs. In the structures of ESA complexes with those drugs, both enantiomers were modeled, and the (S)- or (R)-enantiomers were selected by judging the best fit to the 2mF_o-DF_c omit map and B-factor values. Stereochemical restraints for ligands molecules used during the refinement were generated using the Grade Web Server¹⁰⁸ or AceDRG¹⁰⁹ (included in the CCP4 suite). Both MOLPROBITY¹¹⁰ and wwPDB validation servers¹¹¹ were used to validate the models. PyMOL (The PyMOL Molecular Graphics System, Version 1.5.0.3, Schrödinger, LLC) was used to visualize protein structures. All experimental steps were tracked using LabDB.¹¹² Molstack, a rich Internet application,^113,114 was used for interactive interpretation, verification, and comparison of the models and their respective electron density maps. Diffraction images for all structures presented herein are available at the Integrated Resource for Reproducibility in Macromolecular Crystallography at http://proteindiffraction.org.^115,116 Atomic coordinates and structure factors for all structures were deposited in the Protein Data Bank with accession codes 5V0V (complex with etodolac), 6U4X (complex with ibuprofen), 6U4R (complex with ketoprofen), 6CI6 (complex with nabumetone), and 6U5A (complex with 6-MNA). Statistics for diffraction data collection, structure refinement, and structure quality are displayed in Table 1.

ABBREVIATIONS USED

BSA

bovine serum albumin

CSA

caprine serum albumin

ESA

equine serum albumin

HSA

human serum albumin

LSA

leporine serum albumin

NSAID

non-steroidal anti-inflammatory drug

OSA

ovine serum albumin

PPB

plasma protein binding

SA

serum albumin