Toward a Molecular Basis of Cellular Nucleoside Transport in Humans

Abstract

Nucleosides play central roles in all facets of life, from metabolism to cellular signaling. Due to their physiochemical properties, nucleosides are lipid bilayer impermeable and thus rely on dedicated transport systems to cross biological membranes. In humans, two unrelated protein families mediate nucleoside membrane transport: the concentrative and equilibrative nucleoside transporter families. The objective of this review is to provide a broad outlook on the current status of nucleoside transport research. We will discuss the role played by nucleoside transporters in human health and disease, with emphasis placed on recent structural advancements that have revealed detailed molecular principles of these important cellular transport systems and exploitable pharmacological features.

Graphical Abstract

1. Introduction

Nucleosides play central roles in human metabolism, physiology and pharmacology. As nucleotide precursors, nucleoside pools are required for nucleic acid synthesis and cellular metabolism. Additionally, adenosine signals through specified cell surface receptors with heavy involvement in processes ranging from cellular response to hypoxic stress to neurotransmission and neuromodulation. Owing to the hydrophilic nature of these small biomolecules, nucleosides rely on dedicated transport systems to cross the kinetic barrier of the lipid bilayer. Therefore, nucleoside transport systems play key roles in the regulation of many aspects of nucleoside-related biology. Furthermore, nucleoside transport systems contribute significantly to nucleoside-derived antiviral and anticancer drug absorption, delivery, metabolism and excretion (ADME). Nucleoside transport systems are mediated by choreographed actions of the integral membrane proteins knows as nucleoside transporters (NTs). Due to the physiological and therapeutic importance of NTs, there have been numerous studies of the role of NTs in nucleoside-related physiology and drug efficacy. We wish to note detailed aspects of cellular nucleoside transport and its relation to human physiology and pharmacology have been excellently reviewed elsewhere^1–7. Therefore, the goal of this review is to provide a comprehensive, detailed overview of the molecular and mechanistic features of NTs and their exploitable pharmacological features. Emphasis will be placed on recent experimental NT structures that have not only aided in our understanding of the mechanism of nucleoside transport across the membrane by these proteins, but also have paved the way for future therapeutic interventions in nucleoside biology.

1.1. Nucleosides in human health and disease

Comprised of a ribose sugar linked to a nitrogenous base through a β-N-glycosidic linkage, nucleosides are hydrophilic biomolecules that are either sourced from diet or de novo biosynthesized in humans. As nucleosides can be readily converted to nucleotides via phosphorylation^{8, 9}, nucleoside pools are required for maintenance of cellular energy stores, DNA/RNA synthesis and cellular signaling. Conversely, nucleosides can be generated from nucleotide precursors through the actions of various enzymes^{9, 10}. The purine nucleoside adenosine is of particular importance to cell signaling, as it serves as a ligand for cell-surface adenosine G-protein coupled receptors (GPCRs). Basal concentrations of adenosine are typically in the nanomolar range, though certain physiological and pathophysiological states can trigger extracellular adenosine concentrations to reach millimolar levels¹¹. Under hypoxic/ischemic conditions, this occurs through the release of ATP into the extracellular matrix by pannexins, followed by subsequent hydrolysis of adenine nucleotides into adenosine by ectonucleotidases³, leading to the accumulation of adenosine that ultimately stimulates signaling through adenosine GPCRs^11–14. Downstream of adenosine signaling in these contexts are pathways that lead to vasodilation, angiogenesis and metabolic regulation, making adenosine a protective agent to tissues undergoing hypoxic stress¹². Adenosine receptor signaling is also a key component in neurotransmission and neuromodulation, involved in sleep or wakefulness, alcohol intoxication and use disorders, and implicated in numerous neurological and addictive behavioral disorders^{11, 13, 15–24}.

Due to their role as precursors in nucleic acid synthesis, intervention of cellular and viral replication through use of exogenous nucleoside-analogs (Figure 1) has proven a viable therapeutic strategy in the treatment of cancers and viral infections^25–28. Nucleoside-analog antineoplastic agents exhibit antimetabolite activities, achieved by simple competition of the endogenous nucleoside-pool, resulting in early chain termination, lesion incorporation into the genome, or aberrant DNA methylation, all of which result in general cytotoxicity²⁹. The modest selectivity of the nucleoside-analog antineoplastic agents comes from the fact that cellular replication levels are significantly higher in cancel cells than normal cells. Nucleoside-analog antiviral drugs work in a similar manner, inhibiting viral replication via inhibition of viral polymerases, chain termination or introduction of mutations into the viral genome. Notably, studies have shown that many antiviral drugs can be rationally modified to specifically inhibit viral DNA polymerases, viral RNA dependent RNA polymerases or viral reverse transcriptases, not endogenous human polymerases, therefore exhibiting suitable target specificities^{26, 27}. Some common nucleoside analog drugs include gemcitabine – a chemotherapeutic agent used in the treatment of pancreatic cancer, cytarabine – a therapeutic used in treatment of leukemia, ribavirin – a therapeutic used to treat hepatitis C, and azidothymidine (AZT) – a therapeutic used in the treatment of human immunodeficiency virus (HIV) infection. Of recent relevance to global public health, several nucleoside-analog antivirals have shown initial promise in the treatment of novel coronavirus disease 2019 (COVID-19)^30–32.

Figure 1.

Select endogenous nucleosides (top) and nucleoside/nucleotide antiviral analog drugs (bottom).

1.2. Cellular nucleoside transport systems

In humans, nucleoside and nucleoside-analog drug cellular uptake and excretion is mediated by two genetically distinct protein families: sodium-dependent concentrative nucleoside transporters (CNTs, gene family solute-carrier 28; SLC28) and sodium-independent equilibrative nucleoside transporters (ENTs, gene family solute-carrier 29; SLC29)⁴. While ENTs are in most cases facilitative uniporters, CNTs are secondary active symporters that transport nucleoside against its concentration gradient by thermodynamic coupling with electrochemical gradient of cation. The human CNT family consists of three isoforms (hCNT1-3) and the human ENT family consists of four isoforms (hENT1-4), with each isoform exhibiting varying tissue distributions and functional properties (to be discussed later in this review, also reviewed in⁴).

Human evolution has taken advantage of the distinctive functional features of these two transport protein families, with many tissue types exhibiting strategic spatial expression of CNTs and ENTs. For example, in intestine epithelia CNTs are expressed at apical membranes, acting as a “pump” to drive the uptake of solute from lumen to tissue, while ENTs are expressed at basolateral membranes to allow for trans-tissue diffusion of solute³³. Such a system drives efficient vectorial transport of nucleosides and nucleoside-analog drugs from lumen to blood, controlling nucleoside metabolite and nucleoside-analog drug intestinal absorption. Therefore nucleoside transporters are key determinants of drug oral bioavailability in many cases³⁴ (also reviewed in³⁵).

1.3. Nucleoside transporters in efficacy and disposition of nucleoside-analog therapeutics

Nucleoside transporters mediate cellular uptake and efflux of numerous nucleoside-analog anticancer and antiviral therapeutics, making them key regulators of drug ADME profiles^{36, 37}. The anticancer agent gemcitabine is arguably the most extensively established case of transporter involvement in drug disposition properties. Numerous studies report expression levels and polymorphisms in certain CNTs and ENTs are valid markers to predict cellular efficacy, cellular cytotoxicity, along with patient outcomes to gemcitabine treatment in cancers^38–50. Additionally, similar involvement of nucleoside transporters have been reported for the anticancer agents cytarabine (Ara-C)⁵⁰, decitabine and azacitidine⁵¹, and for the antiviral nucleoside-analog drug ribavirin^52–57. For example, multiple studies have shown that human ENT1 is the major transporter for the antiviral ribavirin and single nucleotide polymorphisms (SNPs) in human ENT1 affect ribavirin treatment response against chronic hepatitis C viral infection significantly^{53, 54, 57}. ENTs and CNTs have also been shown to transport the antiviral drug azidothymidine (AZT)^58–60.

It was thought that other nucleoside-analog antiviral prodrugs, such as remdesivir, are hydrophobic and independently membrane permeable, and they are hydrolyzed to directly form an active monophosphate intracellularly²⁷. However, animal pharmacokinetics studies on remdesivir, sofosbuvir and a N⁴-hydroxycytidine prodrug have shown that these prodrugs are metabolized into nucleoside form in the blood after systemic or oral administration^61–65, suggesting that nucleoside transporters likely play important roles in their target availability, systemic distribution and drug excretion/clearance. It has yet to be established what roles nucleoside transporters would play in the drug ADME of these prodrug antiviral therapeutics, which would significantly impact development of improved nucleoside-analog antivirals against COVID-19.

1.4. Nucleoside transporter inhibitors

As nucleoside transport processes have numerous physiological implications, inhibitors targeting specific CNTs or ENTs hold high therapeutic potential. Due to the wide tissue distributions of human ENTs and their major role in adenosine transport, ENT inhibitors block adenosine reuptake (adenosine reuptake inhibitors, termed AdoRIs) and thus significantly potentiate purine signaling. As adenosine-GPCR agonists exhibit undesired side-effects due to poor tissue selectivity^{15, 66}, prolonged adenosine signaling achieved through its cellular reuptake inhibition would, in theory, have preferential effects on tissue microenvironments experiencing heightened extracellular adenosine accumulation. For these reasons interest has been shown in use of AdoRIs as therapeutic interventions in adenosine biology, for treatment of neurological disorders, hypertension, heart disease and renal disorders, and several of which are currently used in the clinic^{13, 67–69}. AdoRIs targeting human ENTs are broad and chemically diverse, which we will discuss in detail in a later section of this review. Although CNT inhibitors have potential to be used for therapeutic intervention, only a class of sub-micromolar inhibitors against human CNTs have been reported to date⁷⁰. Development of CNT and ENT isoform-specific inhibitors would greatly aid in the functional dissection of nucleoside transport, allowing for further insight into this cellular transport system in the context of human physiology and nucleoside-analog drug pharmacology.

2. Concentrative nucleoside transporters

2.1. Functional properties

Early studies determined cellular nucleoside transport consisted of multiple components, with each component identifiable by three general functional properties: transport dependence on extracellular sodium, transport sensitivity to S-(4-nitrobenzyl)-6-thioinosine (NBMPR), and permeant specificities¹. Sodium-independent nucleoside transport components were later determined to be mediated by the ENT family (to be expanded on later in this review), whereas the sodium-dependent components were later discovered to be mediated by the CNT family. Early nomenclature of the CNT family stratified the three sodium-dependent components of cellular nucleoside transport by permeant specificity: concentrative/NBMPR-insensitive/transports-thymidine (cit), concentrative/NBMPR-insensitive/transports-formycin (cif) and concentrative/NBMPR-insensitivie/transports-both-purines-and-pyrimidines (cib)⁴.

2.1.1. Isoform features and transporter regulation

In a series of papers by Young and colleagues, the molecular identification of individual members of the human CNT family yielded three distinct isoforms with apparent pairwise sequence identities in the range of 40–60%, with each isoform exhibiting an apparent 13 transmembrane domain (TM) topology based on sequence hydropathy analysis^{59, 71, 72}. These studies unambiguously assigned the previously designated cit, cif and cib cellular nucleoside transport components to human CNT isoforms 1–3, respectively, and also showed the three CNT isoforms exhibit diverse tissue distributions and high tissue abundance in epithelia^{59, 71, 72} (reviewed further in⁴).

A more detailed understanding of permeant specificities through human CNTs was advanced by the initial molecular identification of individual human CNT isoforms and subsequent work. It is now well established while all human CNTs transport uridine, human CNT1 preferentially transports pyrimidines, human CNT2 transports purines, and human CNT3 exhibits broad nucleoside selectivity transporting both purines and pyrimidines. It is not clear if human CNTs transport nucleobases, as transport of nucleobase by any CNT has yet to be reported. Later transport studies suggested the importance of ribose hydroxyl groups for nucleoside recognition and transport by all human CNT isoforms⁷³. Transporter interactions with other moieties of solute were determined to be different between isoforms, providing an initial molecular basis for solute preference exhibited by different CNT isoforms⁷³. Another feature of solute specificity exhibited by all three human CNT isoforms is the strong preference for North ribose ring (3’-endo) conformations, determined by transport studies conducted with nucleoside analogs featuring sterically constrained ribose moieties⁷⁴.

Concerning cation-cotransport properties, human CNT1 and CNT2 exhibit a strict 1:1 sodium: nucleoside coupling stoichiometry^{58, 75}, however human CNT3 features more complex coupling criteria. It has been shown that human CNT3 can cotransport with 2:1 sodium: nucleoside stoichiometry but proton can be utilized for transport in the absence of sodium with 1:1 proton: nucleoside stoichiometry⁷⁶. Proton-coupling is also observed in some fungal and bacterial concentrative nucleoside transporters that are sodium-independent, such as Candida albicans CNT (CaCNT) and Escherichia coli NupC^{77, 78}. These differences in solute specificities and substrate coupling ratios present between each human CNT isoform may allow for fine-tuning of concentrative cellular nucleoside uptake in response to specific physiological environments.

Regulation of human CNT expression also likely serves to control cellular nucleoside uptake. It has been reported glucocorticoid treatment leads to increased CNT expression in rat intestinal epithelia⁷⁹, and starvation or nucleotide-deficient diet leads to increased CNT1 expression in rat jejunum⁸⁰, leading Pastor-Anglada et. al. to suggest maturation of epithelial cells and nutritional regulation as two possible factors regulating CNT expression⁸¹. Furthermore, human CNT3 expression in HL-60 cells is stimulated by phorbol myristate acetate (PMA), and does contain an upstream PMA response element⁷¹, demonstrating transcriptional control of a human CNT.

2.2. Structural advancements

As stated above, the first molecular cloning of individual human CNT isoforms predicted a 13-TM transporter membrane topology based on sequence, however following accessibility studies suggested mammalian CNTs contained 13 or 15-TM^82–84, with an intracellular N-terminus and extracellular C-terminal. Work on rat CNT1 determined N-linked glycosylation sites are located within the extracellular C-terminal domain⁸². It was originally postulated the first 3-TM were dispensable, as bacterial CNT homologs lack these putative N-terminal TMs. Subsequent work confirmed this hypothesis, as truncations of the first putative 3-TM of human and rat CNT1 resulted in wild-type like transporter function, while providing insight into the overall topology of the transporter⁸². It is important to note while the CNT N-terminal domain is functionally dispensable, it likely plays important roles in mediating protein-protein interactions, controlling transporter trafficking and may contribute to transporter stability^{85, 86}. Although these initial studies were important advancements to the field of nucleoside research, it remained a challenge to accurately predict the transporter architecture. To gain a detailed understanding of the molecular requirements for nucleoside recognition, ion-coupling and the mechanism of nucleoside membrane translocation exhibited by CNTs, experimental transporter structures were in dire need.

2.2.1. Crystal structure of a sodium-dependent nucleoside transporter from Vibrio cholerae

A 2.4-Å resolution crystal structure of a sodium-dependent CNT from Vibrio cholerae (vcCNT) was reported by Johnson et. al., which marked the first structure determination of any nucleoside transporter⁸⁷. Being that vcCNT exhibits an overall 39% sequence identity to human CNT3, with higher conservation apparent in the nucleoside/sodium sites (~91% seq. identity), this study presented structural data with high relevance to human CNTs.

2.2.1.1. Novel Transporter architecture and the sodium binding site of vcCNT

vcCNT forms a homotrimer, with each protomer comprised of 8 TM (TM1-TM8), 3 interfacial helices (IH1-IH3), 2 re-entrant helix-turn-helix hairpins (HP1 & HP2) and short extracellular helices (EH) (Figure 2A). The CNT protein fold is novel and can be divided into two structural domains: the scaffold domain and transport domain. Comprised of TM1-3, IH1, EH and TM6, the scaffold domain is located at the exterior of a CNT protomer, appearing to contribute to structural integrity of the homotrimer as it forms extensive oligomeric contacts (Figure 2B). Located inside the scaffold domain are two different structural elements that comprise the transport domain - TM4, TM5, IH2 and HP1 form a single group that exhibits apparent internal two-fold symmetry with a structural element formed by TM7, TM8, IH3 and HP2. This observation of apparent internal symmetry, in addition to the presence of conserved positions previously implicated in human CNT function, led to the designation of these structural elements as the transport domain. The high-resolution crystal structure of vcCNT also unambiguously resolved the atomic coordinates of uridine and sodium in complex with the transporter, revealing the CNT substrate binding and ion coordination sites within the transport domain. Located between HP1, HP2, TM4 and TM7 of the transport domain and facing the trimer axis, uridine forms multiple interactions with transporter, many of which were previously reported by functional studies of human CNTs^{83, 88–90} (Figure 2B). Approximately 8 Å from substrate nucleoside is an octahedrally coordinated sodium ion, which was designated as the sodium binding site in vcCNT based on its coordination distances and geometry (Figure 2B). Because subsequent higher-resolution (2.1-Å resolution) structure of vcCNT in complex with uridine did not reveal any additional sodium site⁹¹, we suggest that vcCNT adopts 1:1 sodium: nucleoside stoichiometry. The domain architecture of vcCNT is vaguely similar to that of the bacterial excitatory amino acid transporter homolog Glt_Ph^92–96, which also features distinct transport and scaffold domains. Further, the Glt_Ph transport domain contains two helix-turn-helix hairpins related by internal two-fold symmetry, similar to vcCNT. However, several stark structural differences are apparent, including highly distinct locations of coupled ions and different scaffold domain architectures.

Figure 2.

A) Overview of the vcCNT trimeric structure (PDB ID 3TIJ). B) Detailed interactions of substrate nucleoside (yellow sticks) and coordinated coupled sodium ion (purple sphere) with vcCNT. Scaffold domain depicted in grey, HP1 in salmon, HP2 in green, TM4/5 in orange and TM7/8 in blue.

2.2.1.2. Mechanistic implications

The vcCNT structure adopts a conformation in which the substrate uridine and sodium are facing the intracellular side of the membrane. Interaction of TM3 and TM6 of the scaffold domain with TM4 and HP2 of the transport domain form a thick seal, completely preventing sodium and uridine access to the extracellular side. Although the sodium and uridine sites are solvent accessible to the intracellular side, HP1 appears it would sterically prevent sodium and uridine release, representing an inward-facing occluded state in the context of the alternating access model of membrane transport^{97, 98}.

A finding from this structural study with clear mechanistic implication on sodium-dependent nucleoside transport: the sodium ion did not directly interact with substrate uridine, which is distinct from the sodium-dependent amino acid transporter LeuT. In LeuT, a sodium ion is coordinated by the carboxylate moiety of leucine in LeuT, suggesting simultaneous binding of sodium ion and leucine by the transporter⁹⁹. Johnson et. al. proposed that in vcCNT sodium serves to promote nucleoside binding – as residues interacting with uracil base are located on HP1, stabilization of HP1 by sodium would elicit or stabilize a conformation posed to productively bind nucleoside. The proposed role of sodium in stabilizing the nucleoside binding site was consistent with the follow up equilibrium binding studies that mutation of sodium coordinating residues result in decreased affinity for nucleoside¹⁰⁰. These observations, in combination with structural studies of the bacterial excitatory amino acid transporter homolog Glt_Ph^92–96, led Johnson et. al. to propose a transport mechanism for CNT. In order to transition to additional conformational states, the transport domain would likely rearrange relative to the scaffold domain, which would allow nucleoside and sodium sites alternating access to each side of the membrane (discussed in detail in Section 2.2.3). The proposed role of sodium in the affinity control of nucleoside binding by vcCNT, together with the alternate access model, provides an initial glimpse into the sodium-dependent transport mechanism by vcCNT. When CNT is in an outward facing conformation where the substrate binding sites in the transport domain faces the extracellular side, the high concentration of sodium in the extracellular solution will increase sodium occupancy in the transporter, which in turn increases nucleoside binding. On the contrary, when CNT is in an inward facing conformation, the low concentration of sodium in the intercellular side lowers sodium occupancy of the transporter, which in turn decreases nucleoside binding, driving nucleoside release. This proposed role of affinity control by sodium is in line with the reported kinetic studies on hCNTs which revealed sodium binds to the transporter before nucleoside⁵⁸. Our proposed role of sodium in controlling affinity for substrate could be widely extended to other types of sodium-coupled transporters, as not all LeuT fold transporters feature a coupled-sodium directly interacting with substrate^{101, 102}.

2.2.2. Structural basis of substrate specificity

As vcCNT was determined to be a suitable tool to structural interrogate sodium-dependent nucleoside transport⁸⁷, various nucleosides and nucleoside analogs were later assessed for binding to vcCNT in parallel with co-crystal structure determination (Figure 3A), in order to interrogate general principles of nucleoside recognition and transport by CNTs⁹¹.

Figure 3.

A) Detailed interactions of nucleosides and nucleoside analog drugs with vcCNT (PDB IDs, top left to bottom right: 4PD6, 4PD9, 4PDA, 4PB2, 4PD7, 4PB1, 4PD5, 4PD8). B) Summary of chemical determinants of solute recognition by vcCNT C) Rational chemical modifications that resulted in a chemotherapeutic drug with enhanced transport properties.

2.2.2.1. Nucleobase interactions

Compared to uridine nucleoside, substitutions at the N3 and C4 positions of the uracil base were found to decrease nucleoside affinities. Furthermore, adenosine and guanosine-analog ribavirin also exhibited lower affinities to transporter. This is explained by the crystallographic data, in which uracil base makes important interactions with Q154, T155 and E156 of HP1, many of which are disrupted in the cocrystal structures of nucleoside analogs featuring said substitutions (Figure 3A). However, uridine-analogs featuring substituents on the C5 position of the uracil base exhibited differential affinities, where marked decreases in nucleoside affinities correlate with increasing atomic radii and decreasing in electronegativity of the C5 substituent. The atomic radii trend is an unlikely explanation for this effect, as crystallographic data shows C5-substituents of varying sizes could be accommodated in the nucleoside binding site, therefore substituent electronegativity is the driving factor. Located ~3.5Å from the uracil base is the highly conserved F366 of TM7, which is required for nucleoside binding. This position participates in offset pi-pi stacking interaction with nucleobase, along with CH-pi interactions with ribose. Electron withdrawing groups at the C5 position of the uracil ring would strengthen the offset the pi-pi stacking interaction, which is reflected by the increased affinities of 5-flourouridine and 5-chlorouridine for vcCNT (K_D~14–16 μM) compared to that of uridine (K_D ~36 μM).

2.2.2.2. Ribose interactions

As stated previously in this review, early functional studies highlighted the importance of the hydroxyl groups of ribose for nucleoside recognition by CNT⁷³. Crystallographic data on vcCNT supports this^{87, 91}, as conserved residues N368 of TM7, and E332 of HP2, are in close proximity and interact with the 3’-OH of nucleoside. Additionally, S371 of TM7 and E332 of HP2 form interactions with 2’ and 5’-OH groups, respectively. Consistent with previous functional studies and the crystallographic data, 3’- and 5’- deoxyuridine analogs exhibit large decreases in binding affinities to transporter, whereas a 2’-deoxyuridine analog did not exhibit appreciable binding loss. However, other perturbations at the epimeric 2’ position opposite to -OH (such as in the anticancer drugs gemcitabine and cytarabine) result in more dramatic decreases in nucleoside binding affinities – these observations are explained by the crystallographic data, as substituents at this position disrupt the CH-pi interaction between ribose and residue F366 (Figures 2B and 3A).

2.2.2.3. Molecular features of nucleoside recognition by CNT are exploitable for rational drug modification

Since many anticancer and antiviral agents feature substituents on the ribose moiety of nucleoside, they exhibit low-affinity transport through human CNTs due to the strict requirements for ribose recognition by CNTs. Pyrrolo-cytidine (MePrPmR) is a human CNT inhibitor⁷⁰ that also exhibits high-affinity binding to vcCNT. A co-crystal structure revealed the base of the pyrrolo group of pyrrolo-cytidine uniquely occupies a pocket within vcCNT formed by TM4 and TM6, which may explain the high-affinity exhibited by this nucleoside analog (Figure 3A). Johnson et. al replaced the cytidine ring of the anticancer drug gemcitabine with the nucleobase of pyrrolo-cytidine to compensate for the low affinity of gemcitabine to human CNTs (Figure 3C). The new nucleoside-analog, pyrrolo-gemcitabine, exhibited a higher affinity to the bacterial CNT compared to gemcitabine. Interestingly, pyrrolo-gemcitabine exhibited subtype-selective transport for human CNT1 - whereas human CNT3 was inhibited by the nucleoside analog, transport by hCNT1 was enhanced compared to that of gemcitabine. Modelling suggests the pocket of hCNT3 is similar to vcCNT so that it is snuggly fit with the pyrrolo-cytidine base, but in human CNT1 the pocket is not compatible with pyrrolo-cytidine binding⁹¹. Therefore, increase in affinity by occupying this pocket in human CNT3 is likely the basis for apparent transport inhibition rather than solute permeation. These studies strongly suggest that by modifying nucleobase, the affinity of nucleoside analog drug for CNTs can be tuned to maximize its transportability and isoform specificity by CNTs.

Taken together, the structural and functional data presented on vcCNT by Johnson et. al. reveal that nucleoside recognition by CNTs is mediated by both nucleobase and ribose specific interactions: HP1 and TM4 serve to interact with nucleobase, whereas HP2 and TM7 form interactions with ribose. The residue F366 is central to nucleoside recognition by the transporter, as it forms pi-pi stacking interactions with nucleobase in addition to CH-pi interactions with ribose. Structural features of nucleoside recognition by vcCNT were found to be exploitable for modelling of human CNTs, as rational modification of gemcitabine not only resulted in a nucleoside-analog with higher apparent affinity to the bacterial transporter, but also yielded a permeant with high subtype specificity for transport by certain human CNT isoforms. This proves that it is possible to use structural data in the rational modification of drugs, so they exhibit more desirable transport properties through human nucleoside transporters.

2.2.3. Multiple conformations elucidate novel insights into the elevator transport mechanism

In its simplest form, the generalized mechanism of membrane transport involves solute access to either side of the cell membrane in alternating fashion^{97, 98, 103}. In such a model, transporter gates are elements which serve to fully occlude solute within the transporter substrate cavity, preceding protein conformational transitions between inward and outward facing end states. In the case of ion-coupled cotransporters, solute occlusion by orchestrated transporter gate openings and closures are also necessary to prevent non-specific ion leak. More specific mechanisms of membrane transport are categorized by structural features of conformational change between end states and include the rocking-bundle, rocker-switch and elevator models¹⁰³. The elevator mechanism is an emerging model where the substrate-binding transport domain moves a large distance across the membrane while the scaffold domain remains static in the membrane¹⁰³. As the CNT protein fold is comprised of a scaffold domain that mediates oligomeric contacts, and a transport domain that largely contains the substrate binding site, it was initially predicted to utilize an elevator-mechanism for membrane transport⁸⁷. Many initial studies concerning the elevator mechanism were from the glutamate transporter homolog Glt_Ph^92–96. The elevator model was inferred as a single rigid-body transition of the transport domain, mainly because structures were available for two end states¹⁰³. However, biophysical studies suggested that intermediate states exist^{104, 105}. This has posed interesting questions for CNT transport mechanism. Does the elevator motion occur in a single concerted step or multiple steps? How do interactions between the transport and scaffold domain change while preventing non-specific ion leak during the elevator motion? Does the transport domain move as a rigid body, or does it undergo conformational changes during the elevator motion?

To interrogate the transport mechanism exhibited by CNT, structural elucidation of multiple conformations along the transport cycle are required. As vcCNT was refractory to crystallization in conformations other than the substrate-bound inward-facing occluded state, Hirschi et. al. identified a sodium-dependent bacterial nucleoside transporter from Neisseria wadsworthii (CNT_NW) that exhibited high biochemical stability in the absence of nucleoside and sodium¹⁰⁰. In addition to obtaining crystal structures of inward-facing occluded CNT_NW in complex with sodium and uridine, multiple conformational states were captured with crystallography by utilizing mutations that abrogated sodium and nucleoside binding (CNT_NW^N149S,F336A, CNT_NW^N149S,E332A and CNT_NW^N149L). These structural studies, combined with biochemical studies, provide structural and mechanistic insights into the conformational pathway of the elevator transition of CNT and provides novel insights into the elevator mechanism.

2.2.3.1. Outward-facing and intermediate states of CNT_NW

The crystal structures of both CNT_NW^N149S,F336A and CNT_NW^N149S,E332A were determined in the absence of nucleoside and sodium¹⁰⁰. These structures feature two protomers (A and B) in similar conformations to the previously observed inward-facing occluded state (nucleoside substrate bound). However, the third protomer (C) adopts a distinct conformation. Crystal packing analyses shows that the transport domains of the protomers A and B are involved in packing interactions while that of the protomer C is unhindered for movement. While TM3 and TM6 of the scaffold domain in protomer C exhibited no significant differences when compared to the inward-facing state, the entire transport domain is displaced substantially towards the extracellular side of the membrane in a rigid-body manner. This conformational change results in a ~12 Å displacement of the nucleoside binding site across TM6 towards the extracellular environment, allowing for solvent access of extracellular solution to the nucleoside-binding site (Figure 4). The interface between the scaffold domain (TM3 and TM6) and components of the transport domain (HP1 and TM7) seal off access to the intracellular side, therefore Hirschi et. al. proposed this observed conformation represents the outward-facing state. Intriguingly, by utilizing the CNT_NW^N149L construct, additional conformers of protomer C were observed with crystallography, all of which feature the transport domain in a position somewhere between inward and outward facing states, thus representing intermediate transporter states. Moreover, X-ray data collected from different CNT_NW^N149L crystals revealed three distinct intermediate conformations – intermediate-1, intermediate-2 and intermediate-3, named based on their position in the conformational path from inward to outward facing states (Figure 4). In all three of these intermediate states, the nucleoside binding site is located directly behind TM6 of the scaffold domain and fully sealed from both extracellular and intracellular solutions. Adding biochemical relevance to these structural observations, Hirschi et. al. performed structure-guided double cysteine crosslinking to show the observed inward, intermediate and outward-facing transporter conformers exist in membranes and proteoliposomes. Importantly, while cross-linked mutants reconstituted in proteoliposomes are nearly inactive, when reduced by DTT, the mutants restore transport activity substantially, strongly suggesting that the intermediate and outward conformations are physiologically relevant states, not the states in off-pathways.

Figure 4.

Structural superposition of crystal structures of CNT_NW in different conformational states (structures aligned with reference to protomers A and B, protomers A and B colored in grey, color key for protomer C at right). Large displacements in HP1 and TM4-5 due to conformational transitions are highlighted at right (PDB IDs: 5L26, 5L27, 5L24, 5U9W, 5L2A for inward, intermediate-1, intermediate-2, intermediate-3 and outward, respectively).

2.2.3.2. Conformational changes within the transport domain

In addition to the large rigid body like movements of the transport domain, local conformational rearrangements of HP1 and TM4 underlie the transition from inward to outward facing end states of CNT_NW. Our structural snapshots represent the structural transition from the inward-occluded (substrate-bound) to outward-open state. Notably, the nucleoside binding site within the transport domain in these intermediate states are collapsed and thus cannot accommodate nucleosides – however this nucleoside binding pocket is restored in both inward and outward states. These structures illustrate that in the return path of substrate-free CNT in the transport cycle, the transporter domain appears to undergo a conformational change to a “collapsed” state during transport, which could be necessary prevent transport of unwanted molecules (e.g. nucleoside or ions or some non-specific small molecules).

HP1 appears to function as the intracellular gate, as it interacts with substrate nucleoside in the inward-facing occluded state in both vcCNT and CNT_NW structures. In CNT_NW^N149S,F336A and CNT_NW^N149L structures, HP1 of protomers A and B adopt a variety of conformations in which HP1 is moved away from the central nucleoside binding site to varying degrees. This would allow productive nucleoside release, thus representing inward-facing open states. Additionally, Hirschi et. al. noted the presence of an additionally conformation in which HP1 is partially unwound and “tucked in” towards the center of the nucleoside binding site. This represents an inward-facing pre-translocation state, a rearrangement of HP1 required to transition from inward-apo to intermediate-apo transporter states. Such a state prevents steric clash of HP1 with the scaffold domain during such transitions. This is clear in protomer C of intermediate CNT_NW structures – HP1 and TM4 move closer together during these conformational transitions, making the transport domain more compact. In the outward-facing transporter state, TM4 is displaced upward substantially allowing the nucleoside binding site solvent accessibility to the extracellular solution (Figure 5). Therefore, TM4 likely operates as the extracellular gate. This “fixed barrier elevator with two gates” mechanism of transport exhibited by CNT is distinguished from other transporters that utilize elevator mechanisms¹⁰⁶, and has yet to be observed in other transporter families.

Figure 5.

Conformational changes in the transport domain mediates collapse of the nucleoside binding pocket during the elevator-like transitions. Surface representation of each transporter conformer shown, with the nucleoside pocket circled in red (PDB IDs: 5L26, 5L27, 5L24, 5U9W, 5L2A for inward, intermediate-1, intermediate-2, intermediate-3 and outward, respectively. Protomer C depicted for each conformer, except for inward pre-translocation which is depicted from chain A of PDB ID 5U9W).

In summary, structures of CNT_NW in the inward occluded (substrate-bound), inward open, inward-pre-translocation, intermediate, and outward open states reveal the conformational trajectory from inward substrate-bound to outward open states. Notably the new intermediate states with the transport domain halfway across the membrane provides that elevator motion by CNT_NW is composed of multiple steps including the intermediate steps. Possible reasons for the existence of these intermediate states are : 1) to provide an energetic barrier between two end states, as without the barrier switching between the end states would lead to sodium leakage, an idea suggested by smFRET studies of Glt_Ph^{104, 105}, and 2) to guide the path of large-scale motions of the transport domain, to prevent the non-specific leakage through the interface between the scaffold and the transport domains. Consistent with this idea, mutation of a glutamate residue on the domain interface in the intermediate state leads to nucleoside-uncoupled sodium leak currents in human CNT1 and CNT3^{89, 90}. It is remained to be seen whether these intermediate states exist in other transporters that exploit elevator transport mechanisms.

2.2.4. Structural mechanism of CNT mediated nucleoside transport

Taken together, structural studies performed on bacterial CNTs to date have uncovered important molecular principles of the mechanism of sodium-dependent cellular nucleoside transport. A key feature of sodium coupling in CNT is the fact there is no direct interaction between cation and solute. Observations of varying structural states highlight the likely role of sodium in affinity control of solute nucleoside – occupancy of sodium on an exofacial transporter site would enhance nucleoside affinity, whereas sodium release from an endofacial transporter site reduces nucleoside affinity, thus promoting release of solute into the intracellular environment. Indeed, in sodium-free conditions CNT_NW exhibits a wide range of HP1 conformations in protomers which adopt the inward-facing state (Figure 6), providing direct structural evidence for sodium release driving HP1 opening, followed by nucleoside release to the intracellular environment.

Figure 6.

HP1 of CNT_NW adopts a variety of conformations in the absence of sodium. Inward-facing sodium- and substrate-bound CNT_NW HP1 depicted in red, with substrate uridine shown as sticks and sodium as a purple sphere (PDB ID 5L26). HP1 conformers from protomers A and/or B from sodium free conditions shown in shades of violet to blue (PDB IDs 5L27, 5L24, 5UW9, 5L2A).

Intermediate states of the transporter on its return path from solute release to the intracellular solution back to outward-facing conformations showed the nucleoside binding pocket collapses, forming tight seals between the transport domain and scaffold domain. We propose this substrate-free return path to an outward facing transporter state serves to prevent hysteresis and non-specific transport events. Therefore, our transport mechanism illustrates the conformational dynamics of the CNT transport mechanism: starting from inward facing occluded states (solutes bound), to solute sodium and nucleoside release, and then back to the outward facing state posed for sodium and nucleoside recognition through compact intermediate states (Figure 7). The mechanism of CNT from outward-facing substrate bound to inward-facing occluded is currently not known. Additionally, it remains to be seen if the elevator mechanism of CNT is applicable to other transporters that exhibit elevator motions associated with transport.

Figure 7.

Current working model for the mechanism of nucleoside transport mediated by CNTs. Structural states in which experimental structures are available are boxed. Scaffold domain depicted in grey, transport domain in blue. The nucleoside binding pocket (light blue) is demarcated with dotted line.

3. Equilibrative nucleoside transporters

3.1. Functional properties

Use of the mercaptopurine NBMPR has allowed for numerous functional and genetic advancements in the nucleoside transporter field, which includes the identification, cloning and initial characterization of individual human ENT isoforms^107–115. The ENT family was primarily designated as equilibrium/NBMPR-sensitive (es) and equilibrium/NBMPR-insensitive (ei) components of sodium-independent cellular nucleoside transport^{4, 115}.

3.1.1. Isoform features and transporter regulation

Human ENT1 was the first sodium-independent nucleoside transporter to be cloned¹⁰⁹. Responsible for es nucleoside transport, human ENT1 is potently inhibited by NBMPR with the half maximal inhibitory concentration (IC₅₀) in the sub- to low-nanomolar range^{109, 115}. Human ENT2 was cloned next and determined to be the major contributor to ei sodium-independent nucleoside transport, as it is far less sensitive to NBMPR (micromolar or weaker inhibition)¹¹⁶. Human ENT3 was later identified as a primarily organellar transporter which exhibited virtually no sensitivity to NBMPR^{117, 118}. Human ENT4 was the final human isoform to be identified and cloned – this ENT isoform is also completely insensitive to NBMPR^119–121.

All human ENT isoforms feature an apparent 11-TM topology with sequence identities ranging from 25 to 50%, with human ENT4 exhibiting large sequence divergence from the other human ENTs. One shared feature of solute recognition among all human ENT isoforms is the preference for South (C2’-endo) ribose ring conformation for nucleoside substrate, which contrasts the preference for North (C3’-endo) ribose conformation exhibited by CNTs⁷⁴. Human ENT 1–3 exhibit broad tissue distributions and can transport both purines and pyrimidines, therefore exhibiting much broader substrate selectivity compared to CNTs. Additionally, human ENT1 and ENT2 have been shown to mediate low-affinity transport of nucleobases^{108, 122}. The permeant specificity exhibited by human ENT4 is distinct from human ENT 1–3, as the only nucleoside it can transport is adenosine at acidic pH¹¹⁹. Human ENT4 is also known as the plasma membrane monoamine transporter (PMAT) as it can also permeate monoamines, such as serotonin¹²¹. Additionally, human ENT4 exhibits distinct tissue distributions compared to the other human ENT isoforms, as it was found to be enriched in heart and brain tissue^{119, 121}.

It is well understood human ENT1 and ENT2 are facilitative nucleoside carriers, however human ENT3 and ENT4 have been shown to exhibit pH-dependent transport of nucleosides^{117, 119}. Putative titratable residues responsible for the pH-dependence of nucleoside transport for human ENT3 and ENT4 have been reported^{119, 123, 124}. Electrophysiological studies on human ENT3 failed to detect proton transport during adenosine transport, and thus suggested that proton is not a co-transporting ion but regulates adenosine transport in human ENT3^{123,, 125}. Although further studies are necessary to validate the conclusion of the role of pH in hENT3, the allosteric role of pH in adenosine transport in hENT3 is intriguing. The pH sensing exhibited by human ENT3 may reflect the environment in which the transporters is posed to operate – in acidic organellar compartments¹¹⁷. Because acidic pH is critical for maintaining lysosomal function, it is possible that hENT3 has evolved to operate only in the acidic pH to synchronize the metabolic needs of the lysosome. Adenosine transport by human ENT4 has been reported to be pH-dependent, however monoamine transport does not involve cotransport of ions or exhibit pH-dependence¹¹⁹ and was found to be membrane potential dependent¹²¹. Although it has yet to be established if pH-dependent adenosine permeation requires ion co-transport, pH-dependent purine transport mediated by human ENT4 may be physiologically relevant in the context of ischemia of the brain and heart^{119, 126}.

Being that human ENT1 is most widely expressed and well-studied, its transcriptional regulation is best understood out of all ENT isoforms. In general, human ENT1 expression is coordinated in response to cell cycle and nutrient availability (reviewed in¹²⁷). In line with its important role in purinergic signaling, equilibrative nucleoside transporter expression has been reported to respond to stress. Specifically, it was determined human ENT1 expression during hypoxia is repressed by HIF-1α and the human ENT1 promoter contains two putative HIF-1α recognition sites¹²⁸. Additionally, activation of the JNK-cJun pathway has been reported to negatively regulate ENT1 expression in rat¹²⁹. It is important to note sequence elements for other regulatory factors are also present within the human ENT1 promoter¹³⁰.

3.1.2. Genetic disease and genome-wide associations in the human ENT family

Currently, intracellularly localized human ENT3 is the only nucleoside transporter identified to be associated with human genetic disease. Rare mutations in the slc29a3 gene have been identified as the cause of recessive autosomal disorders including H-syndrome¹³¹, pigmented hypertrichosis with insulin-dependent diabetes mellitus syndrome¹³², Faisalabad histiocytosis, familial Rosai-Dorfman disease¹³³ and dysosteosclerosis¹³⁴. These slc29a3 associated disorders are autoinflammatory in nature, with lysosomal and mitochondrial instability as a likely driver of inflammation¹³⁵. In further support of this disease etiology, human ENT3 has been shown to play roles in bone physiology¹³⁶ immune function, macrophage homeostasis¹³⁷ and T-cell survival¹³⁸, through maintenance of organellar integrity and function. Although some mutations of slc29a3 that lead to such disorders are nonsense or frameshift^{139, 140}, many are missense that diminish transporter stability, localization and transport activity¹⁴¹. Further, many of these mutations map to conserved positions of potential structural and functional importance in the recently determined human ENT1 crystal structure¹⁴² (to be discussed later in this review).

Human ENT1 is a key controller of adenosine signaling, and as adenosine is implicated in alcoholism^19–24, human ENT1 has been shown to regulate ethanol preference and sensitivity^143–146. A human polymorphism in slc29a1, 647T/C, is significantly associated with alcohol dependence with withdrawal seizures¹⁴⁵. This polymorphism results in the I216T mutation on TM6 of human ENT1, and is a common variant prevalent in the human population¹⁴⁷. A current working hypothesis suggests human ENT1 is inhibited by ethanol, and that the I216T variant alters the transporters sensitivity to ethanol¹⁴⁵ (to be discussed later in this review). It is important to note other polymorphisms of the ENT family are prevalent in the human population and are discussed further by Schaller and Lauschke¹⁴⁷.

3.1.3. ENT inhibitors

As mentioned in the introduction, human ENTs are sensitive to inhibition by a chemically diverse class of compounds termed adenosine-reuptake inhibitors (AdoRIs, Figure 8). In general, human ENT1 exhibits the highest sensitivity to many AdoRIs. Beyond the hallmark AdoRI NBMPR, other high-affinity inhibitors include dipyridamole – an FDA approved coronary vasodilator^{148, 149}, dilazep – a low nanomolar hENT1 inhibitor used as a therapeutic in renal IgA nephropathy^{69, 150}, draflazine/soluflazine (and derivatives) – highly potent AdoRIs with varying subtype and species selectivities¹⁵¹, and rapadocin – a recently discovered novel rapamycin-derivative which exhibits highly potent and subtype-specific inhibition of human ENT1 in a FKBP12-dependent manner⁶⁸. Adding to this list are other putative transporter inhibitors likely targeting human ENTs. T1-11 is an adenosine-mimetic chemically similar to NBMPR with lower apparent toxicity, which inhibits adenosine transport in the low micromolar range and has shown promise as an adenosinergic modulator in a mouse model of Huntington’s disease⁶⁷. Ticagrelor is a P2Y₁₂ receptor inverse agonist and FDA approved antiplatelet drug which was shown to inhibit human ENT1^{152, 153} – however, the extent to which inhibition of human ENT1 by ticagrelor contributes to its pharmacological properties remains unclear^{154, 155}. It is also important to note human ENTs may be inhibited by ethanol^{6, 143, 156–158}, certain JNK inhibitors, serine/threonine kinase inhibitors^{3, 159}, calcium channel antagonists¹⁶⁰, and cannabinoids¹⁶¹ (Figure 8) – however it has yet to be established if these compounds directly act on human ENTs, or act on upstream pathways.

Figure 8.

Chemical structures of select Adenosine Reuptake Inhibitors (AdoRIs) and other putative ENT inhibitors.

3.2. Structural advancements

As stated earlier, upon the molecular cloning of individual human ENT isoforms and subsequent topology studies, evidence for an 11-TM protein fold emerged^{109, 112, 116–121}. Featuring a cytosolic facing N-terminal and extracellular facing C-terminal, the ENT fold also contains two large disordered loops which are located between TM1-2 and TM6-7. The TM1-2 loop contains N-linked glycosylation sites and has been implicated in membrane trafficking and protein folding^{112, 162}. The TM6-7 loop contains multiple phosphorylation sites and has been shown to be important in the regulation of transporter function^{163, 164}. Both the TM1-2 and TM6-7 loops have been shown to be dispensable in human ENT1, without affecting general functional properties of transport and inhibition¹⁶⁵. Studies using chimeric constructs elucidated the role of various TM regions in sensitivity to NBMPR and other AdoRIs, along with substrate recognition¹⁰⁷. Mutagenesis and functional characterization of transporter interaction with substrate and inhibitors were conducted, shedding considerable light onto the molecular determinants of such interactions^{110, 111, 166–169}. Functional studies utilizing the Leishmania donovani nucleoside transporters, members of the ENT family, also contributed greatly to early efforts to functionally interrogate ENTs^{170, 171}.

Sequence analysis of the ENT family provides evidence they share a common ancestor with the ubiquitous Major Facilitator Superfamily (MFS) of transporters^{172, 173} – however, extremely low sequence identities (<15%) are apparent between members of the ENT family and MFS members. Using this as a framework for initial structural studies, experimental structures of MFS members E. coli lactose permease (LacY)¹⁷⁴ and fucose permease (FucP)¹⁷⁵ were used as a templates for ab initio homolog modeling. These studies resulted in the construction of a structural models of L. donovani nucleoside transporter 1.1 (LdNT1.1) in inward and outward facing conformations, which combined with mutagenesis highlighted general features of the ENT fold^{176, 177}. Furthermore, ENT structure predictions have been performed with de novo prediction pipelines^{7, 178–180}.

These initial structural studies were important advancements to the field of ENT research. However, constructed homology models of ENTs are generally of low fidelity, due to low sequence homology to template or the challenges of membrane protein structure prediction using de novo methods. To comprehensively characterize nucleoside recognition, inhibition by AdoRIs and the transport mechanism exhibited by human ENTs, experimental transporter structures were determined¹⁴².

3.2.1. Crystal structures of human ENT1 in complex with adenosine reuptake inhibitors

Work on bacterial CNTs highlighted the feasibility of utilizing biochemically stable prokaryotic orthologs for structural studies^{87, 91, 100}. Human ENTs lack any significant evolutionary relation to prokaryotic transporters and purified human transporter exhibits biochemical instability^180–182. Furthermore, the requirement of milligram quantities of purified transporter underlies any crystallographic effort. To circumvent these issues, Wright and Lee engineered a functional variant of human ENT1 that is biochemically stable and crystallizable¹⁴². This enabled structure determination of the human transporter in complex with the adenosine reuptake inhibitors dilazep and NBMPR, marking the first available experimental structures of any ENT family member.

3.2.1.1. Transporter architecture and evolutionary insights

Crystal structures of human ENT1 confirmed the transporter adopts a 11-TM topology¹¹², with the N-terminal in the cytoplasmic side and C-terminal in the extracellular side. The structure can be divided into two domains referred to as the N-domain and C-domain, where the first 6 TMs comprise the N-domain and the final 5 TMs comprise the C-domain (Figure 9). These two domains exhibit pseudo two-fold symmetry, a structural arrangement similar to that exhibited by MFS transporters¹⁸³. Superposition of human ENT1 with various MFS transporter structures reveals structural homology between the two protein folds. The ENT fold exhibits several notable differences from the MFS fold, including slight displacement of TM9 resulting from the lack of the canonical TM12. This results in a lower apparent internal two-fold symmetry between the N- and C- domains within the ENT fold compared to the level of internal two-fold symmetry apparent in MFS¹⁴². We speculate that the lower internal two-fold symmetry provides a gating mechanism that deviates from that of MFS (rocker switch – see 3.2.1.3).

Figure 9.

High-resolution crystal structure of the dilazep human ENT1 co-crystal structure (PDB ID 6OB7). The first 6 transmembrane helices colored blue (N-domain), and last 5 transmembrane helices colored orange (C-domain).

The 11-TM topology of human ENT1 and its structural similarity with MFS has posed an interesting question regarding the design principle of functional transporter from the evolutionary standpoint. Based on the semi-SWEET and SWEET transporter structures^184–186, it was suggested that each lobe of MFS can be further broken into a 3+3 TM assembly, suggesting the 3 TM bundles in the MFS fold (TMS 1–3, 4–6, 7–9, 10–12) are the minimal building block with which functional transporter can be built. Consistent with the nomenclature presented by Nordlund and colleagues, these 3 TM bundles consists of a sequential arrangement of A-, B- and C- helices¹⁸³. Currently available structural data concerning alternating access through MFS and semi-SWEET indicates interactions between A- and B- helices contribute most extensively to transporter gating¹⁸³. These structural observations led to the hypothesis SWEET and semi-SWEET families may represent evolutionary predecessors to MFS¹⁸⁶, consistent with the hypothesis that MFS may have evolved from gene duplication events of triple-helix bundles^{187, 188}. The human ENT1 structure highlights the achievability of a functional transporter fold without following the strict 12 TM topology of MFS or the 3 TM building block strategy from SWEET transporters. It is possible that the C-helices (TMs 3, 6, 9 or 12) are less critical for transporter function, and that the ENT family simply lost a single C-helix TM during evolution. These studies pose an interesting question as to what the minimal structural requirements are to build a functional transporter.

Located in the center of the transporter between the N- and C-domains is a hydrophilic cavity lined by conserved residues previously implicated in transporter function, substrate recognition and inhibitor binding^{110, 111, 166–169}. This central cavity is sealed from the intracellular side by extensive hydrophobic contacts formed by TMs 4, 5, 10 and 11. Due to the large protein volume apparent at this intracellular region of the transporter, we term this structural feature the “thick-gate.” At the extracellular side of the transporter TMs 1 and 7 are located in close proximity at the top of this central cavity but are still far enough apart to allow solvent accessibility from the extracellular solution. We term this structural feature the transporter “thin-gate”, as the barrier does not exhibit extensive protein volume between the extracellular side and central transporter cavity but would occlude nucleoside release from the central cavity (Figure 10A). Therefore, these human ENT1 structures represent the transporter in the outward-facing occluded state.

Figure 10.

A) Dilazep and NBMPR binding sites in human ENT1 (dilazep depicted in yellow and NBMPR green, PDB IDs 6OB6 and 6OB7, respectively). B) Detailed depiction of inhibitor-transporter interactions. Protein portion of N-domain colored blue and C-domain colored orange (lighter shades of either color used for the NBMPR human ENT1 cocrystal structure). Interacting residues depicted as sticks.

3.2.1.2. Adenosine reuptake inhibitor binding sites

A large portion of the adenosine analog NBMPR occupies the central cavity of human ENT1, where its thioinosine nucleoside moiety forms interactions with several residues previously implicated in nucleoside recognition^{110, 111, 166–169} (Figure 10B). The p-nitrobenzyl moiety of NBMPR sits in a cavity within the N-domain of human ENT1 (this cavity is also present in the dilazep structure), connected and adjacent to the transporter central cavity. The N-domain cavity is lined by TM1, TM3, and TM4 and contains G154 at the proximal side – a residue responsible for the species/subtype specificity exhibited by NBMPR¹⁶⁹ and I216 at the distal side – a human polymorphism at this position is associated with ethanol addiction and withdrawal in humans^{145, 147} (discussed in section 3.1.2 and to be discussed in 3.2.1.4). Dilazep is chemically dissimilar to nucleoside and is comprised of two trimethoxybenzoic acid groups and a central diazepane ring. One of the trimethoxybenzoic acid groups occupies the central transporter cavity, while the diazepane ring and other trimethoxybenzoic acid group protrude upward towards the extracellular solution and form contacts with the extracellular thin gate (Figure 10A). Therefore, NBMPR and dilazep both feature occupancy of the transporter central cavity, which is proposed to be the nucleoside binding site. However, specific chemical moieties of NBMPR and dilazep also occupy their own distinct sites on human ENT1 – one site being a cavity within the N-domain (NBMPR), another site being the extracellular thin gate (dilazep). These structural observations provide a possible explanation for the high avidity exhibited by these transporter inhibitors to human ENT1. In addition to the favorable interactions observed within the nucleoside-binding site, there are substantial interactions observed in each of the allosteric sites. Therefore, the energetic contributions to inhibitor binding are likely, to some degree, an additive combination of interactions observed within the nucleoside binding site and allosteric site.

3.2.1.3. Mechanistic implications

Structural mechanisms of MFS transporters have been thoroughly characterized^{183, 189–191}, and fall under the rocker-switch category for membrane transport¹⁰³. This model for membrane transport involves the sequential formation and disruption of transporter protein barriers, or gates, at either side of the membrane. Conformational changes involving movement of the two symmetric halves of the transporter relative to the membrane normal occur between gate-formation and gate-disruption steps. These transitions allows the solute within the central cavity, located between two symmetric halves of the transporter, alternating access to either side of the membrane¹⁹² (summarized in Figure 11). Due to the overall similarities of the ENT and MFS folds, their transport mechanisms are likely analogous. In the structure of dilazep bound human ENT1, inhibition of transporter function is likely achieved by steric block of extracellular thin gate occlusion – this inhibitory mechanism has been reported for other transporter inhibitors^193–196. This highlights the mechanistic requirement of gate closure by rearrangements of TM1 and TM7, purposed to fully seal the central cavity from extracellular solution. Based on the current working model for membrane transport through MFS transporters, this step would precede transporter exchange to inward-facing conformational states. In the structure of NBMPR bound human ENT1, the transporter adopts a conformation similar to the dilazep-bound structure. However, no NBMPR moiety contacts the human ENT1 extracellular thin gate, ruling out steric hindrance of extracellular gate occlusion as a plausible mechanism for transport inhibition exhibited by NBMPR. Further, it is well accepted NBMPR recognizes an exofacial transporter site^{114, 197}. Therefore, inhibition appears to be solely mediated by the occupancy of the human ENT1 N-domain cavity and results in the transporter being trapped in an outward-facing state. Since the N-domain cavity is lined by TM1, TM3 and TM4, we propose that these structural elements undergo key conformational rearrangements that occur after extracellular gate closure to drive transporter to inward-facing states. The trapping of the transporter in the outward facing state by occupancy of a site adjacent to the substrate binding cavity is distinct from steric block of gate closure and is therefore a unique mechanism of transporter inhibition.

Figure 11.

Current working model for the mechanism of nucleoside transport mediated by human ENT1. Currently, only the outward occluded transporter conformer has been structurally elucidated (boxed in diagram). Adenosine reuptake inhibitors dilazep and NBMPR exhibit distinct mechanisms of inhibition by perturbing two different conformational transitions in the transport cycle.

3.2.1.4. Insights into nucleoside recognition by hENT1 and its comparison with vcCNT

Since NBMPR is an adenosine-mimetic, we can postulate structural features of purine recognition in human ENT1 using the NBMPR-bound human ENT1 structure and compare with the adenosine-bound vcCNT structure (Figure 12). Intriguingly, although these two transporter families both recognize and transport nucleosides, they exhibit highly distinct structural architectures. Further, they appear to recognize nucleoside within their substrate binding cavities in different ways. For example, critical for nucleoside recognition through CNTs is a F366, a residue located in the middle of the nucleoside binding site – this residue not only forms pi-pi stacking interactions with base, but also mediates favorable CH-pi interactions with ribose. In human ENT1, the purine ring of NBMPR appears to be recognized through hydrophobic interactions mediated by L26 and L442, along with polar interaction with Q158. However, some commonalities do appear to exist between the two transporter families, such as the presence of interactions specified for 2’ and 3’ OH, which are mediated by polar/charged transporter residues.

Figure 12.

Comparison of inward-facing adenosine-bound V. cholerae CNT (left, PDB ID 4PD9) with outward-facing NBMPR-bound human ENT1 (right, PDB ID 6OB6).

3.2.1.5. Structural basis of slc29a3 genetic disorders

Rare mutations in human slc29a3 are associated with a wide spectrum of autosomal recessive genetic disorders that are often autoinflammatory in nature. The availability of an ENT experimental structure allows for preliminary understanding of the molecular basis of slc29a3 genetic disorders – disease associated mutations were mapped to their corresponding conserved positions on the human ENT1 crystal structure (Figure 13). The mutation M116R¹³² has previously been shown to disrupt protein trafficking and results in human ENT3 retention in the ER/Golgi¹⁴¹. Three additional disease mutations (G427S, G437R and T449R)^{132, 133, 198, 199} impaired human ENT3 activity but appeared to traffic normally in the same study¹⁴¹. G437R and T449R map to the TM10-11 loop, which contains one of the putative pH-sensitive residues in human ENT3^{123,, 125} (human ENT3 residue E447, conserved position in human ENT1 is E428). Other mutations identified from human patients include R134H/C¹⁹⁸, S203P¹³⁴, R363Q²⁰⁰ and R386Q²⁰¹. While the mutation R386Q likely affects protein stability or turnover, examination of the locations corresponding to R134H/C, S203P and R363Q in the available ENT crystal structure are of particular interest. Human ENT3 position R134 is highly conserved across the entire ENT family and when mapped on the human ENT1 structure, would interact with the conserved position corresponding to the pH-sensing residue E447 in human ENT3^{123, 125}. R363Q has clear implications in disrupting nucleoside binding in the central cavity of the transporter as this corresponding position in the human ENT1 crystal structure (R345) interacts with the ribose ring of the adenosine mimetic NBMPR¹⁴². Human ENT3 disease mutant S203P, a polymorphism associated with dysosteosclerosis¹³⁴, is located in close proximity to this conserved position. In summary, while various mutations or truncations of slc29a3 associated with rare genetic disease result in changes in protein fold, stability or expression, many mutations result in perturbations of functionally important positions that would diminish transport activities or alter functional properties of nucleoside transport mediated by human ENT3.

Figure 13.

Mapping of slc29a3 mutations associated with a broad spectrum of genetic disorders onto their corresponding conserved positions in the human ENT1 crystal structure (PDB ID 6OB7; mutation sites highlighted in red, pH sensing residue highlighted in grey). Mutations in human ENT3 are labeled in bold, and the corresponding conserved positions in human ENT1 are listed below in parentheses.

3.2.1.6. Potential role of human ENT1 in ethanol related physiology and disease

As discussed in Section 3.1.2, a common variant in slc29a1 (I216T) may be associated with alcohol dependence with withdrawal seizures¹⁴⁵. Located on TM6 of human ENT1, I216 faces the distal end of the cavity within the transporter N-domain in which the p-nitrobenzyl ring of NBMPR occupies in the co-crystal structure. This cavity is still present in the dilazep co-crystal structure, despite the absence of ligand occupancy at this site. Due to the high-resolution of the dilazep co-crystal structure, several well-ordered waters were resolved at the proximal end of this cavity¹⁴². It has been reported that the I216T mutation in human ENT1 results in a ~2 fold decrease in NBMPR affinity¹⁴⁵, which is consistent with this mutation changing the electrostatic properties of this otherwise apolar site (Figure 14, surface electrostatic potential map calculated with the Adaptive Poisson-Boltzmann Solver²⁰²). Therefore, ethanol could possibly occupy the distal end of this cavity and achieve transport inhibition in a similar mechanism to that observed for NBMPR, where the I216T mutation would result in an altered binding affinity of ethanol at this site. Consistent with this hypothesis, acute ethanol exposure has been shown to inhibit nucleoside transport, and in some cases potentiate adenosine signaling^156–158. However convincing these observations, the molecular basis for ethanol inhibition of human ENT1, along with the role played by I216T in altering human ENT1 sensitivity to ethanol, has yet to be rigorously investigated with biochemical and structural studies.

Figure 14.

Human ENT1 residue I216 lines the distal end of the cavity within the transporter N-domain. Surface electrostatic potential map calculated for the dilazep human ENT1 cocrystal structure (PDB ID 6OB7) using the Adaptive Poisson-Boltzmann Solver. Well-ordered waters observed in the proximal end of the N-domain cavity in the high-resolution structure are depicted as red spheres.

4. Future advancements

Although representative experimental structures are now available for both the CNT and ENT protein families, our understanding of their transport mechanisms are still limited. The transport mechanism exhibited by CNT is most extensively characterized¹⁰⁰, however the outward-facing occluded state (substrate bound) remains to be determined. Moreover, despite their high sequence similarities, would human CNTs exhibit any differences in solute recognition and transport mechanism compared to that established for bacterial sodium-dependent nucleoside transporters? For example, would human CNTs employ a multi-step elevator motion including the intermediate states that are observed in CNT_NW? Where is the second sodium site in human CNT3? A polymorphism in human CNT3 (C602R) disrupts the observed 2:1 coupling stoichiometry^{203, 204}, indicating it might be in proximity to this second sodium site - would this second sodium also contribute to nucleoside substrate affinity control that we observe in bacterial CNTs? What is the structural basis of proton involvement in transport mediated by human CNT3? What are the timescales of conformational transitions during the transport cycle, and could they be measured with biophysical approaches? Additional functional, structural and biophysical studies focused on all three human CNT isoforms are needed to address these remaining questions.

Our understanding of the transport mechanism by ENTs is at an even earlier stage with many remaining questions – what are the structural features that underlie nucleoside transport? Although many nucleoside drugs bind to hENT1, why do some drugs act as permeants while others act as inhibitors? Facilitative solute transport characterizes the ENT family, but isoforms 3 and 4 utilize pH-dependent transport mechanisms. What is the structural and mechanistic basis for these pH-dependent transport mechanisms, and what specific purposes do they serve in human physiology? Understanding the structural basis of solute recognition as well as transport dynamics would help answer these important questions. This would require structural data on human ENT1 conformations beyond the previously reported outward-facing occluded state¹⁴² and in complex with additional ligands. Structures of ENT isoforms 2–4 would also be needed to understand the molecular basis of functional discrepancies between the different ENT isoforms. Further functional and biophysical approaches would also be needed to interrogate transport dynamics exhibited by ENTs.

The availability of the current structural data on bacterial CNTs and human ENT1 sets the stage for the advancement of rational drug design. Nucleoside-analog antiviral or anticancer drugs exhibiting improved ADME properties and transporter inhibitors with desirable pharmacological activities stand to be gained from utilizing this available structural information. Future efforts would certainly impact the effectiveness of therapeutics purposed to intervene in nucleoside biology.

Biographies

Biographies

Nicholas Wright earned a B.S. in biochemistry from Roanoke College in 2016. He is currently an American Heart Association predoctoral fellow pursuing Ph.D. training at Duke University in the Department of Biochemistry, under the supervision of Dr. Seok-Yong Lee. His research has been focused on the structure and mechanism of membrane transport proteins.

Seok-Yong Lee, Ph.D., is a professor of Biochemistry at Duke University School of Medicine. Dr. Lee completed his doctoral work in 2003 at the University of California at Berkeley under the supervision of Dr. David Wemmer. He then pursued postdoctoral studies with Dr. Roderick MacKinnon at Rockefeller University, where he carried out structural and biophysical studies of potassium channels. In 2009, Dr. Lee started his independent career at Duke University and built a research program focusing on the structural biology of ion channels, transporters and membrane-embedded enzymes. He has been recognized for his contributions to the field of membrane protein structural biology by multiple awards including the NIH Director’s New Innovator Award, NIGMS award, the SER-CAT Outstanding Science Award, and McKnight Scholar Award.