Homology modeling of the Taurine transporter

Homology modeling of the Taurine transporter

Introduction

Cell fate, cell migration, cycle regulation, cellular metabolism, and intracellular signaling are profoundly affected by cell volume control imbalance. In contrast, vital cellular processes, for example, increase in cell volume is often considered a replica for migration and mitosis (), which is triggered by a shift in cell volume, while cell apoptosis triggers cell shrinkage (). Often, cell volume preservation incorporates active uptake of ions and passive release of organic osmolytes that is influenced by the net locomotion of osmolytes, which is osmotically followed by water.

The process of changing intracellular ionic surroundings or membrane potential through rapid extracellular osmolarity changes is useful in avoiding cellular dysfunction due to accumulated organic osmolytes and ions during cell shrinkage. This process is known as RVI or regulatory volume increase. Alternatively, the RVD or regulatory volume decrease is the process due to cell swelling that results in the release of the osmolytes [7, 8]).

As such, Taurine can be defined as one of the primary semi-essential organic osmolytes with abundant β-amino sulfonic acid that is used in the mammalian cell volume maintenance. Hence, its cellular content is as a result of a balance between active uptake, synthesis, and passive release of ions and organic osmolytes. The active accumulation and passive release of Taurine in the cells occurs through the taurine transporter (TauT) and unidentified swelling induced release pathway, respectively. For example, in human adults, Taurine is manufactured or synthesized by essential amino acid methionine and non-essential amino acid cysteine, mainly in the liver [9]. On the other hand, the primary source of Taurine for human neonates is dietary uptake (10).

The estimated bulk concentration of Taurine in normal plasma is 10-100 µM.  Approximately, more than half of the taurine component is made of free pool of amino acid in tissues like the retina and heart with 40mM and 6mM concentrations, respectively [11]. Since Taurine is part of the osmotic regulation, it is believed to be useful in the control of cholesterol, insulin signaling, neuromodulation, detoxification, and antioxidative defense [12-17].

According to research works, deficient cats of taurine cause hydrocephaly or anencephaly, which is due to severe growth retardation, degeneration of the retina, and central nervous system (CNS) dysfunction, along with severe the CNS development disturbances in the CNS development ([14]). Previous studies have introduced threonine/serine kinase CK2 as one of the regulators of active uptake and passive release of Taurine [18, 19]. The constituents of CK2, for instance, are useful active and up-regulation of most cancer types along with cells that stimulate growth, proliferation while suppressing apoptosis [20]. Additionally, CK2 inhibition prepares cells that cause cancer to apoptosis induction [22, 23]. For example, tumor cells of Ehrlich Lettré Ascites (ELA) and mouse NIH3T3 fibroblasts, express high level and low level of CK2, respectively. Under isotonic conditions, CK2 pharmacological inhibition potentiates the transportation of Taurine while reducing the release of Taurine, as illustrated in Fig. 1.

Fig. 1: Showing how CK2 favors taurine accumulation in ELA and NIH3T3 cells under isotonic and osmotic conditions.

Fig. 1 (A) demonstrates how the inhibition of CK2 potentiates the transportation of Taurine. According to Fig. 1, the estimation of taurine influx was taken from the CK2 inhibitor treated-cells of (E)-3-(2,3,4,5-tetrabromophenyl) acrylic acid (TBCA, grey bar) and the initial 3 H-taurine transportation in control cells (open bar) under the isotonic and osmotic conditions. Fig. 1 (B) demonstrates how the inhibition of CK2 lowers the release of Taurine. Once again, the estimation of taurine efflux was taken under isotonic and osmotic conditions with the help of a tracer technique. It also illustrates how the rate of fractional constant for releasing of labeled Taurine is utilized to point out the taurine leak passage activities in the presence (grey bar) and absence (open bar) of TBCA.

According to studies, available CK2 subunits and crystal structures [26] primarily target vast ranges of substrates that are useful in the processes of intracellular homeostatic. These processes include apoptosis, transformation, proliferation, gene expression, cell dynamics (cytoskeleton/morphology), and cell cycle control [27, 28]. Others considered processes are subunits dissociation or aggregation relative to changes in ionic strengths, (30, 33, 34), subunits temporal and spatial structuring, (31, 32), limited knowledge of CK2 regulation activities; that is, the dephosphorylation or phosphorylation regulation as well as protein-protein interaction recruitment [29, 30]. Treatment with CK2 inhibitors, increased activity of CK2, and overexpression of CK2 induces the death of cells even in cells that are resistant to drugs.

The inhibition impact of CK2 on active transportation of Taurine encompasses taurine transport kinetics modulation. On the other hand, the inhibition impact of CK2 on the release of Taurine incorporates activity reduction of the taurine efflux passage. Usually, an increase in the taurine release is significant due to cell swelling. This process entails a 5-lipoxygenase or phospholipase cell-specific subtypes, production of reactive oxygen species, and cell-specific NADPH oxidases, PI3K-phosphatidylinositol 3-kinase, which is known for promoting the growth of cells, survival and proliferation cells, as well as GSK3β -glycogen synthase kinase, which is linked with the homeostasis of glucose.

Taurine Transporter

One of the solute carrier 6 (SLC6) members, the taurine transporter (TauT-SLC6A6), is responsible for taurine uptake. TauT-SLC6A6) is a chloride-dependent and sodium-dependent transporter that entails no less than 16 highly homologous members [36], namely the neurotransmitter transporters, such as serotonin, noradrenaline, dopamine, glycine, and GABA as well as the creatine transporters. Studies have shown that TauT gene cloning from fish (41), human (40), dog (39), rat and mouse (38, 37) can code up to 620 proteins of amino acid with about 70kDa molecular weight. TauT of mammals indicate a 90% sequence identity showing that there are many occurring isoforms of TauT, for example, the mouse fibroblasts. Western blotting analysis can be employed to distinguish mouse fibroblasts along with their affinity towards Taurine (42) using their molecular weight [19, 42, 43]. However, in the kinetic analysis, they are distinguished with their Na⁺: taurine stoichiometry [18].

By looking at the hydropathy plots, one would be able to tell the constituents of TauT (TauT is made up of 12 transmembrane domains, in which other members’ structural analogy can be drawn. For example, the family of Na⁺-coupled and Cl^–-coupled SLC6 transporters and the Aquifex aeolicus leucine transporters(LeuTAa)  for Leucine [44, 45]. These plots indicate that the taurine and Na⁺in TauT binding pockets are coordinated by TM8, TM6, TM3, and TM1. Besides, the variations observed in the Na⁺: taurine stoichiometry is as a result of the variational reflection in Na⁺ binding sites [44]. The sodium binding to TauT intra-membrane domains initiates the uptake of Taurine, which changes the tertiary structure, thereby facilitating taurine binding and transportation [46]). Chlorine is added to sodium to attain the maximal rates of transport.

Generally, the stoichiometry of Na⁺Cl^– taurine is 2.5:1:1 [47]; however, this ratio can always be modulated or varied between various cells ([46] potentially due to differential regulation or TauT-subtypes. The TauT knock-out and the associated failure to amass cellular taurine results in photoreceptor programmed cell death. For example, in mice, this is attributed to either decreased capacity of antioxidative or retinoid transport impairment between retina and pigment epithelium [14, 48, 49]. Additionally, mice TauT knock-out is 25 percent lower than the nerve cells destruction in the CNS, auditory nerve, and olfactory epithelia; reduced capacity of exercise; increased liver problem; and normal birth weight [14, 48-51]. Taurine deficiency has both immediate and long term impacts, as seen in cats is indirectly due to cellular stress and osmotic perturbations caused by diverse mechanisms of reduced antioxidative capacity.

This condition is treated using platinum drugs, for example, cisplatin that induces p53 activation and TauT repression. In other words, TauT overexpression prevents renal dysfunction and cisplatin-induced apoptosis [52]. Moreover, many studies have shown that the TauT knockdown in ELA cells significantly increases [53]. Some speculations increased TauT activity has some limitations in the cell shrinkage during the process of apoptotic, thereby limiting apoptosis initiation. Besides uptake through TauT, the transporter PAT1 also mediates the uptake of Taurine [54].

On the other hand, PAT1 is a pH-dependent transporter while Na⁺-independent and Cl-independent transporter that is not attracted to Taurine. According to (54), it is assumed that the PAT1 found inside the intestine is responsible for the uptake of Taurine taking means that it is rich in Taurine as has been demonstrated in various models. Biophysical and biochemical methods have been used concerning GItPh structures and experimental tests to develop multiple human SLC6 member models (Canul-Tec et al., 2017; Console et al., 2015; Colas et al., 2015). According to these studies, there are essential residues binding sites that are used to describe the specificity of differential charges for the proteins (Scopelliti et al., 2018; Singh et al., 2017; Canul-Tec et al., 2017).

In the recent past, ASCT2 models have been built relative to outward-open conformations and outward-occlusion of GltPh. These models have been applied in the identification and refining of numerous ASCT2 ligands, namely substrate, and inhibitors (Singh et al., 2017; Colas et al., 2015). Captivatingly, a GItPh variant structure that was solved recently was meant for ASCT2 substrates binding, as such as was developed that was used to understand the interactions of ASCT2 (Scopelliti et al., 2018).

In this study, a homology modeling of the taurine transporter in two conformations concerning recently solved structures of EAAT1, as well as the evaluation of the model’s relevance for drug discovery, has been discussed. More specifically, this study begins by first analyzing the phylogenetic relationships of various family members of human SLC6 along with their prokaryotic homologs for providing information on the taurine homology modeling. Again, for purposes of comparison, the ASCT2 models have been generated to the variant crystal structures of the contrived GltPh. The utility assessment for various ligand discovery proteins has also been analyzed. Also described in this work is the unique ASCT2 inhibitor location using virtual screening to predict its interaction mode through ASCT2. Therefore, the results of this study offer insight into substrate structures specificity in the family of SLC6, hence providing ASCT2 inhibitors design framework with improved potency and selectivity.

Materials and Methods

Various research materials and methods have been discussed under this section to build into the concept of taurine transporter homological modeling, as discussed in the subsequent subsections.

Phylogenetic Tree

The building of the phylogenetic tree involved retrieving SLC6 members and each family member of FASTA sequences from Uniprot (Uniprot Consortium, 2018). According to Pei and Grishin (2014) and Edgar (2004), these two parameters were then aligned with MUSCLE or Multiple Sequence Comparison by Log-Expectation then utilized as Simple Phylogeny input. Simple Phylogeny input was used as a simple webserver to undertake a phylogenetic analysis of MSA (McWilliam et al., 2013). Due to its capability of being viewed using various programs of tree viewing, this study used the Newick tree format. Branch stretching and subsequent divergent sequences were avoided by setting on distance correction. Similarly, the gaps in the MSA were removed by setting on exclude differences and using only positions that would give the needed information. The tree diagram from the distance matrix was thus constructed using a neighbor-joining algorithm; thereafter, Fig-Tree was used to visualize the final tree while viewing of phylogenetic trees was done using a graphical program (Rambaut, 2012).

Homology modeling of the Taurine

Human EAAT1 X-ray structures in the outward-open conformations and outward-occlusions (5MJU and PDB codes 5LLM, respectively) were used to enable modeling of ASCT2 (Canul-Tec et al., 2017). The prokaryotic and human SLC6 family members MSA homologs were used as the source of the initial alignment of ASCT2-EAAT1. These homologs were subsequently cultured concerning three visual analysis, namely;

1. An alignment of a pairwise ASCT2-EAAT1 produced by Promals3D through default parameters (Pei and Grishin, 2014)
2. SLC family alignments that were previously published (Canul-Tec et al., 2017)

• Lastly, the visual analysis of generated preliminary homology models relative to various alignments.

However, this study did not consider modeling loop regions with a significant variable whose distances were away from the site of substrate binding; hence could not show the likeliness of impacting on the interactions of ASCT2-ligand (Colas et al., 2015). The loop regions that were not modeled are those loops between the (TM) transmembrane region 3 and region 4a as well as those between region 4b and region 4c. According to Webb and Sali (2017), 100 initial models were developed for alignment of ASCT2-EAAT1 and each conformation using the MODELLER-v9.11. They were then refined through the iteration of M387 and D464 sidechains modeling based on a fixed backbone with the generated model assessed visually and through PyMOL1. The nonprotein elements, such as sodium ions and TBF-TBOA or ligand aspartate as well as UCPH101, were used to construct the models concerning their coordinates in the structural template.

Subsequently, the Z-DOPE score was used to evaluate and then rank the models. A Z-DOPE refers to a normalized atomic distance that depends on statistical capacity concerning structures of a protein that is used to assess and rank models (Shen and Sali, 2006). Based on this technique, top-scoring models would be expected to display -0.313 Z-DOPE scores for outward-open conformations and -0.517 Z-DOPE score for outward-occlusions, respectively. For this study and further investigations, these scores were considered to be significantly suitable for use as modeling templates compared to ASCT2 models that use GItPh structures. The first 15 Z-DOPE scoring models generated by each conformation were then analyzed using enrichment analysis through ligand docking.

Enrichment analysis through ligand docking

The ability of the models was evaluated to differentiate between ASCT2 ligands from decoys with the help of ligand enrichment calculations (Fan et al., 2009). The study used ChEMBL (Gaulton et al., 2017) and 26 known ASCT2 ligands extracted from various literature for the outward-open conformation models (Singh et al., 2017, Schulte et al., 2016; Schulte et al., 2015). Based on these ligands, 1304 decoys were produced using the DUD-E server (Mysinger et al., 2012).

Alternatively, 7 ASCT2 ligands were removed for the large and unlikely occlusion models, which could not fit into the binding site of the substrate. As a result, this study used a total of 19 known ligands to produce 937 decoys conformation using the DUD-E server. The OpenEye FRED was to perform Docking (McGann, 2011). OpenEye FRED was selected based on various criteria, namely;

1. Its ability to obtain high enrichment scores by docking against ASCT2 models when compared to methods’ scores
2. One can easily access the OpenEye FRED program using academic license;

• OpenEye FRED is fast and easy to use; this feature made performing docking against various models manageable. Besides, the program enhances ligand enrichment and prediction accuracy (Fan et al., 2009).

1. Lastly, over the years, this program has shown practicality in the campaigns of ligand discovery by targeting transporters’ models of substrates with chemicals that are closely related. For example, oligopeptide transporter PepT1 (Colas et al., 2017) have compounds that are strictly related to L-type amino acid transporter LAT1 (Zur et al., 2016).

The MAKE_RECEPTOR utility of FRED with S353 and S351 ASCT2 and N471 and S351 ASCT2 for outward-open and outward-occluded constraints respectively were used to prepare the binding site. On the other hand, the OMEGA utility was used to make ligand conformations. Lastly, the calculation for the AUC or area under the curve, as well as the enrichment plots for log AUC or logarithmic scale AUC, were also conducted (Fan et al., 2009).

Virtual Compound Screening

As discussed in the previous section above, this study used OpenEye FRED to screen the ZIBC15 (McGann, 2011) library of “Available Now Lead-Like” alongside the outward-open ASCT2 model concerning GltPh (8.7 x 106 compounds). Also employed on the S353 backbone nitrogen is their constraints as well as the D464 side chain of the carboxy group. The 200 top-scoring screen compounds were visually inspected. In contrast, the prediction of prioritized compounds made to enable the binding site interaction with conserved hydrogen bonds using essential residues between GItPh and ASCT2, as illustrated in Fig. 3. Furthermore, 13 elements of those compounds that entail new chemotypes as ASCT2 ligands with pocket A and B docking were selected, as shown in Table 1.

Measurement of Pocket Volume

Pocket Volume Measurer (POVME3) Wagner et al. (2017) was used to measured and calculate volumes of binding sites. Similarly, AutoDock Vina plugin was used to estimate binding sites coordinates in PyMOL (Trott and Olson, 2010). The coordinates were utilized as POVME3 inputs.

SK-MEL-28 Cell Culture

The STR profiling-2018 was used to confirm the identity of the human melanoma cancer cell line SK-MEL-28 (Berlin Cell Bank). The PCR-based detection technique was routinely applied to test cells as free from mycoplasma. The medium of DMEM provided the environment for growing SK-MEL-28 cells. It contained 1× penicillin-streptomycin solution and 4 mM L-Taurineand 20 mM HEPES with 10% (v/v) components of fetal bovine serum- HyClone (FBS-HyC) (Life Technologies). Cells were passed and kept in an entirely humidified atmosphere at 37^oC with 5% CO2 for 3 days. Before diluting the inhibitors their final uptake media concentration, they first were suspended over and over to 20mM in DMSO. Again, water was used to re-suspend L-γ-taurine-p-nitroanilide (TPNA) before diluting it to final concentrations for uptake media.

Evaluating Taurine Uptake

The incubation of [3H]-L-Taurine (that is, 400 nM; PerkinElmer) in RPMI media along with the SK-MEL-28 cells (which is, 1 × 105/well) alongside the Life Technologies devoid of L-Taurine was carried for 15 min at 37°C both with and without of each inhibitor (Control; vehicle) as discussed earlier (van Geldermalsen et al., 2016). A 96-well plate harvester was used to collect and transfer cells to a filter paper (Wallac PerkinElmer). The collected sample was then dried, followed by fluid scintillation exposure. Lastly, as described by Wang et al. (2014), the counts’ measurements were conducted using a liquid scintillation counter (MicroBeta2 Counter).

Electrophysiology

The experiments in this study were conducted as already discussed in the previous subsections above. Briefly stated, the dissolution of compound 10 in DMSO and then a maximum of 1mM concentration dilution at 5% of the final DMSO was performed. However, until the above concentration point, DMSO had not shown any significant currents until when the HEK293 cells expressing rat ASCT2 were applied as well as suspended from the electrode that was recording current. At a pH of 7.40, 10mM HEPES, 2mM MgCL2, 2mM MgCl2, and 140 mM NaCl formed solutions for the external buffer. Alternatively, at a pH of 7.40, 10 mM HEPES, 10 mM EGTA, 2 mM MgCl2, and 130 mM NaSCN formed solution for intracellular buffer. The resistance of the pipette used ranges between 3 and 5.5 MΩ. Lastly, Adams and List EPC7 amplifiers were used to record currents at 24 hours after transfection. This process involved utilizing reagents of Jetprime transfection based on the Polyplus supplied protocol.

Results

SLC6 Multiple Sequence Alignment Refinement and Phylogenetic Tree

The study used homologs’ structures that were experimentally prepared to generate the homology of ASCT2 models. There is a significant correlation between the homology model accuracy and the target-template alignment quality (Forrest et al., 2006). Also, there is a close link of sequence identity of ~30% between human SLC6 members models and models previously published relative to GltPh, in which alignments become more error-prone and more challenging (Eramian et al., 2008).

The EAAT1 structures recently published have 46% with ASCT2 same sequence identity, thereby providing better templates to model ASCT2, which boosts the confidence in the ASCT2 model and target-template alignment. However, there were some crucial structural changes between GItPh and EAAT1, namely the loop residue conformation that connects TM8 and hairpin (HP) 2. These changes are usually are six residues longer in GItPh than in EAAT1, which is more preserved amongst the human SLC6 members, as shown in Fig. 3 (Canul-Tec et al., 2017).

Furthermore, more conservations with EAAT1 were observed in the TM1’s N-terminus in ASCT2 that had almost parallel helical structures to the membrane as against the GItPh corresponding loop region. As a result, there were expected differences between biophysical properties and the shape of the binding sites of ASCT2 models that sequentially will determine the small molecules’ chemical structures. The ASCT2 was recognized in each conformation, as illustrated in Fig. 4 for transporters of other SLC (Schlessinger et al., 2012).

Therefore, this study produced an MSA with the human SLC6 family members’ sequences and GItPh and EAAT1 structures with Promals3D that utilizes constraints that are structure-based from the 3-D structures alignments as illustrated in Fig. 3 (Pei and Grishin, 2014). Consequently, a phylogenetic tree was built using this Simple Phylogeny alignment (McWilliam et al., 2013). Hence, four separate yet distinct branches of the phylogenetic tree were obtained. ASCT1 and ASCT2 thus clustered together as neutral amino acid transporters, thereby sharing a 57% sequence identity, which is the highest sequence identity of the SLC6 family pairs. According to Kanai et al. (2013), there was a 44% to 55% sequence identity sharing displayed by the five taurine transporters (Kanai et al., 2013). For example, a similar branch was shared by EAAT2 and EAAT3 (Fig. 3), which was the highest similarity of the five taurine transporters.

Fig. 3: The prokaryotic and human SLC6 transporters homologs phylogenetic tree.

The prokaryotic homologs solved structures are illustrated in blue; SLC1A3 (EAAT1), while the one shown in red is the only experimentally determined SLC1 family human member. The green colors show other human SLC6 members’ homology models. The lengths of the branches had values equal to the ratio between the lengths of alignments and the number of substitutions, which were, in turn, proportionate to the evolutionary change. Consequently, the MSA was utilized to generate alignment pairs between every member of the family and EAAT1 that acted as homology modeling input, as illustrated in Fig. 3. Remarkably, scores of the models that were have been calculated using Z-DOPE showed a significant correlation with the SLC6 family members’ positions on the phylogenetic tree based on the EAAT1. The EAAT4 model, for instance, generated the best Z-DOPE score of −0.73; on the other hand, the worst Z-DOPE score of 0.23 was recorded by EAAT3 model. These differences were attributed to substantial insertion results generated in the EAAT3 (residues 164-201) between the TM4b and TM4c compared to EAAT1.

Two Different Conformations of ASCT2 Homology Models         

The study used MODELLER to generate ASCT2 homology models Webb and Sali (2017) concerning EAAT1 structures in the outward-occluded conformations and ligand-bound outward-open. Accordingly, the Z-DOPE scores obtained for ASCT2 models were −0.31 and −0.52, correspondingly. Based on the analysis of this study, it can be deduced that these scores were more superior compared with the previous models in the literature concerning GltPh. It can be concluded that the models are sufficiently accurate (Colas et al., 2015). Iterative sidechains modeling provided the basis of initial models of various binding site residues based on a fixed backbone using visual assessment and PyMOL for the optimized models.

This phase was followed by docking ASCT2 ligands that are already known, such as inhibitors and substrates, to the models substrate binding sites for all conformations. It is important to note that there was a close resemblance of the outward-open model and that of ligand binding mode in the outward-occluded model that recapitulated the possession of ligand in the GItPh and EAAT1 crystal structures as illustrated in Fig. 5B and C. Thus, the two conformations offer critical polar interactions conservation between the binding site and the amino-acid like ligands. For instance, the structure in the outward-occlusion, as illustrated in Fig. 4B had critical polar interaction conservation between the substrate Taurineforms hydrogen bonds carboxy group with S353 (HP1) and that having N471 and T468 (TM8) and the substrate forms of the amino group with hydrogen bonds of T468 (TM8), D464, I431 (HP2) and S351 (HP1). However, there were no contacts with HP2 where there were docked inhibitors in the outward-open conformation, as illustrated in Fig. 5C.

Fig. 4: ASCT6 models in two conformations.

Fig. 4(A) represents the outward-open and the outward-occluded models and their predicted corresponding binding modes for ASCT6 ligands that are known ligands, namely a substrate, as shown in Fig. 4(B) and inhibitor, as shown in Fig. 4(C) inhibitor. The blue and magenta paint ribbons represent HP2. Key residues sidechain atoms were described with grey ribbons. Lastly, salmon sticks were displayed using small molecule ligands, with atoms of sulfur, nitrogen, and oxygen in yellow, blue, and red, respectively. The dotted green lines represent the ligands while N471, T468, D464, S353, S351, P432, and I431 represent the hydrogen bonds holding the residues of the binding site. The docked Taurine with outward-occlusion is as illustrated in Fig. 4(B) while the outward-open model, as illustrated in Fig. 4(C) contains (2S)-2-amino-4({2-[(morpholine-4-yl)methyl]phenyl}carbamoyl)butanoic acid.

The critical responsibility in substrate specificity was depicted that C467 in ASCT2 to match T459 and R479 in ASCT1 and EAAT1 (Colas et al., 2015; Scopelliti et al., 2018). Hence, the substrate preference in the SLC6 transporters can be altered if this residue is mutated to either an arginine or a neutral amino acid (Scopelliti et al., 2013). According to Scopelliti et al. (2013), the T459R mutation in ASCT1, for instance, introduces transporter of acidic amino acid to the neutral amino acid transporter. The role of this residue was to facilitate the production of hydrophobic sub-pockets known as PB, or only pocket B was equally an essential aspect since changes the shape and increased the size of the binding sites as illustrated in Fig. 4 and Fig. 6. Besides, there is an additional sub-pocket PA, pocket A in the outward-open conformation that is exposed upon HP2, adopting an open structure. The pocket-A residues constituents are among the variables of this study’s model and EAAT1 and GltPh-R397C, as illustrated in Table 1. Thus, targeting both PA and PB sub-pockets simultaneously is one of the surest identification strategies for chemically novel ASCT2 inhibitors.

Table 1: Respective locations of proteins in Pocket A residues aligned amongst GltPh-R397C, EAAT1, and ASCT2.

Comparing Models and SLC6 Structures

In this section, the ASCT2 models and their relevance were assessed to enhance the discovery of structure-based ligands. This process involved the evaluation of ASCT2 models in both conformations for their ability to be able to differentiate ligands from possibly decoys or likely-non-binders by employing ligand enrichment or docking (Fan et al., 2009). The decoys were used as ligand molecules due to their physical resemblance, although they are topologically different to minimize the binding possibility (Mysinger et al., 2012). As a result, both the enrichment plots’ AUC or area under the curve and the logarithmic AUC were calculated to determine the ability to prioritize the known ligands. It was noted that the outward-open models and the outward-occlusion models AUC were 95.19 and 94.12, respectively, as illustrated in Fig. 5A, and Fig. 5B. According to Fig. 5, both model A and model B are relevant for the discovery of ligands provided molecular docking is used (Colas et al., 2015).

Fig. 5: Using enrichment plots to evaluate a model.

The green color represents ASCT2 models enrichment plots while the red represents variants of GltPh-R397C in two conformations. The blue dotted lines represent expected plots of randomly selected ligands. A and C enrichment plots represent the conformation of an outward-occlusion, while B and D enrichment plots represent an outward-open conformation. Also, the log AUC and AUC were calculated for the GltPh (GltPh-R397C) variant structures, which mimicked the ASCT2 function (Scopelliti et al., 2018).

In short, the binding site residue R397C provided a point mutation for the GltPh-R397C variant that corresponds to C467 in ASCT2. As a result, the transporter was capable of binding and transporting ASCT2 substrates, for example, alanine and serine. Yes, the AUCs for outward-occlusion and the outward-open structures for GltPh-R397C were 64.92 and 74.61, respectively. These areas suggest that the ASCT2 ligands that are known were captured, as shown in Fig. 3A and 3B, though, there were lower enrichment scores compared to the ones obtained when the new ASCT2 models were used.

Similarly, the GItPh-R397C structures, EAAT1, and ASCT2 models binding sites were analyzed to provide the difference between the rational ligand specificity and the enrichment scores. In both EAAT1 and 2NWW and wild type GltPh (PDB codes: 6BAT) (Scopelliti et al., 2018; Canul-Tec et al., 2017; Yernool et al., 2004), R479/397 is used to block PB in both the outward-open conformations and outward-occlusions as shown in Fig. 6A. In particular, the GltPh-R397C occlusion binding site (6BAU) is nearly identical to the wild-type GltPh corresponding structure binding site (6BAT) (RMSD 0.24).

Fig. 6: ASCT2 substrate-binding site with its homologs.

Remarkably, the ASCT2 more significant inhibitors are graded first in the library before the GltPh-R397C structures, as illustrated by the dissimilarity in the two proteins’ binding sites size. In both conformations, the substrate-binding site of ASCT2 is more significant (outward-open: 439 Å3, outward-occluded: 70 Å3,) compared to the ones for GltPh-R397C, which is 45Å3, 283 Å3. Additionally, the PA provides a conspicuous dissimilarity between the two proteins, as indicated by the different structures and sequences between GltPh-R397C and ASCT2, as illustrated in Table 1, Fig. 6B and Fig. 7. The HP2 residues, especially in T438, L437, V436, and A428, G427, V426 of ASCT2 formed loop regions. On the other hand, the GltPh-R397C’s corresponding residues built helical structures, as shown in Fig. 7A and 7B. It was also noted that there was more conservation of HP2 between EAAT1 and ASCT2 than between GltPh-R397C and ASCT2, as shown in Fig. 7C and 7D. According to these results, while it is essential to use the GltPh-R397C variant to explain substrate specificity, one of its failures is the incapability of capturing every feature of binding sites that are useful for the interaction of ASCT2-inhibitor as illustrated in Fig. 4.

Fig. 7: Showing the outward-open conformation differences in pocket shape and size.

In Fig. 7, the pink and blue colors represent the HP2 for GltPh-R397C and ASCT2, correspondingly. Nitrogen, oxygen, and stick illustrate key residues while blue and red illustrate sulfur atoms. Fig. 7(C) shows HP2 sequence alignment between EAAT1 and ASCT2 while Fig. 7(D) represents the HP2 sequence alignment between GltPh-R397C and ASCT2. The orange boxes mark the difference in secondary structures for residues between GItPh-R397C and ASCT2.

A New ASCT2 Inhibitor

According to Sing et al. (2017) and Colas et al. (2015) predictions, the ASCT2 ligands previously identified could bind only either PA or PB subpockets. However, in this study, it was hypothesized that simultaneous targeting of both PA and PB using virtual screening could generate compounds with chemically new supports for ASCT2. On the contrary, this study carried out a discovery campaign using earlier published GltPh-based ASCT2 models for ASCT2 modeling concerning EAAT1 structure; the basis was custom the compounds used for iterative refinement of the outward EAAT1-based ASCT2 models.

Thus, this study used GItPh-based ASCT2 models for docking the ZINC15-lead-like library. The prioritization of these compounds was based on the predictability mode of binding to both PA and PB, as well as on chemotype novelty. The test activity was conducted for 13 compounds by utilizing Taurine uptake assessment in an SK-MEL-28 or melanoma cell line. Three compounds based on the study results showed significant inhibition in SK-MEL-28 cells at 100 μM, for example, [3H]-L-Taurine uptake. Subsequently, the most potent compound of IC50 was determined using 10 mixture (ZINC69811181) for taurine uptake inhibition utilizing various concentrations. Based on this procedure, it was observed that for compound 10, there was IC50 of 97.16 μM, as illustrated in Fig. 8D; this was an improvement of approximately 18-fold compared to the ASCT2 inhibitor.

Docking of the compound was carried out on the new ASCT2 outward-open model, which, based on the prediction, had a similar binding mode to other models proposed in other studies, as illustrated in Fig. 8A and 8B. Electrophysiology was employed to conduct further analysis of compound 10, as shown in Fig. 8E to ascertain the target engagement using ASCT2. Based on this analysis, when compound 10 was applied to ASCT2-expressing HEK293 cells. The result was an anion leak current inhibition that had earlier on indicated an ASCT2 inhibitors hallmark (Singh et al., 2017; Colas et al., 2015). On the other hand, examples of transported substrates, such as alanine, caused an inward current catalyzed by intracellular anion efflux, as demonstrated in Fig. 8E. However, the above compound 10 effects were dose-dependent and could only saturate in the presence of IC50 of 67 ± 17 μM, which is a similar range as those shown when the Taurine uptake assay was used as per Fig. 8F. Therefore, it can be concluded that compound 10 delivers valuable scaffolds when new concept and optimization for higher affinity compounds and more selective design for ASCT2 are used.

Fig. 8: Showing a novel ASCT2 inhibitor identification.

Discussion

This study’s hypothesis was affirmed. The Taurine discovery for addicted tumors proposes that restraining Taurine to cells that cause cancer results in malignant that kill cells (Akyuz et al., 2013). Inhibiting Taurine to import the cells that cause cancer through the transportation of nutrients is one of the surest approaches to targeting Taurine addicted tumors. Fuchs et al. (2007) note that ASCT2 plays very crucial functions in taurine uptake and transport in numerous cancer cases, which explains why it is now considered one of the potential target drugs for anticancer. Until now, drugs that can potentially target nutrient transporters with tumorigenesis implications in the clinic have not been found (César-Razquin et al., 2015).

Lack of experimentally evaluated or assessed structures is one of the difficulties experienced in studying ASCT2. For example, a comprehensive understanding of how the binding sites of substrates are needed to develop tool compounds in the discovery of structure-based ligands is essential.  Developing tool compounds would make it possible for other researchers to undertake further study and develop future ASCT2 drugs. As a result, this study has modeled ASCT2 with new template structures and new alignments; besides, it took an investigation about these models’ potential utility using small discovery molecules.

Based on this work, three fundamental findings were generated, namely a newly refined SLC6 family MSA or multiple sequence alignment, which experimentally determined structures and prokaryotic homologs that can be used to model the SLC6 family members. From Forrest et al. (2006) viewpoint, the target-template alignment accuracy is critical to an accurate homology model generation. From the study’s hypothesis, it was concluded that including structural information generated by the structures of newly solved EAAT1 can improve the alignment as displayed by the models made in this work. For example, this research work created new models that showed improved enrichment scores and Z-DOPE scores when compared to the models reported earlier in ASCT2 (Colas et al., 2015). The new alignment developed in this study offers valuable resources that others can construct their SLC6 family homology models.

The second fundamental finding brought forward by this study is that the newly-generated ASCT2 can approximately provide perfect inhibitor binding since it can differentiate between likely-non-binders and known ligands using the docking technique. These models are more enriched when they are compared to GltPh variant X-ray structures (GltPh-R397C) with mimicry characteristics of ASCT2 (Colas et al., 2015). As well, these new models provide a rational design foundation for the ASCT2 potent inhibits due to its characteristically higher resolution. A very closer study of the various SLC6 members’ binding sites, namely GItPh, ASCT2, and EAAT1, proposes numerous dissimilarities in several residues that have some impacts on pocket-A or PA shape and size. This information is essential as it can be used to provide on different conformation-specific ASCT2 designs for ASCT2 small-molecule ligands.

Thirdly, it was identified in this study that an ASCT2 inhibitor that uses distinctive predicted binding mode would likely face one significant challenge in the tool compounds for ASCT2 development. The problem was how to discover ligands with a deviation characteristic (approximately 450 to 550 μM) from structures of amino acid, which makes competing for better circulation complex, especially with the circulation of Taurine and alanine for high levels.  Such molecules are expected to have improved pharmacokinetic properties and ASCT2 specificity. Compound that inhibits EAAT1, for example, can result in different deleterious effects in neurological.

Conclusion

The study of homology modeling of the taurine transporters has been discussed as one of the computational screens with the experimental testing as the subsequent consideration. Based on this concept, this study identified compound 10 as one of the weak ASCT2 inhibitors that lack amino-acid like structures. The study highlights that this compound can bind to a varying shape and size in the EAATs for the PA because of the differences in helix packing and HP2. A similar compound can as well bind PB that is predicted to have a delicate structural dissimilarities for various members of human SLC6. Lastly, it is essential to note that compound 10 has approximately 18-fold more potent when compared to GPNA as projected to bind PA. Therefore, it is a compound that delivers a potential optimization starting point in the future campaign for drug discovery and ASCT2 targeting.