Sage Journals: Discover world-class research

Abstract

Obesity is an important risk factor for diabetes, cardiovascular diseases, and cancer, reducing the quality of life and expectancy of millions of people. Consequently, obesity has turned into one of the most health public problems worldwide, which highlights the urgent need for new and safe treatments. Obesity is mainly related to excessive fat accumulation; therefore, proteins participating in white adipose tissue increase and dysfunction are considered pertinent and attractive targets for developing new methods that can help with body weight control. In this context, virtual screening of libraries containing a large number of molecules represents a valuable strategy for the identification of potential anti-adipogenic compounds with reduced costs and time production. Here, we review the scientific literature about the prediction of new ligands of specific proteins through molecular docking and virtual screening of chemical libraries, with the aim of proposing new potential anti-adipogenic molecules. First, we present the targets related to adipogenesis and adipocyte functions that were selected for the following studies: PPARγ, Crif1, SIRT1, ERβ, PC1, FTO, Mss51, and FABP4. Then, we describe the obtention of new ligands according to the characteristics of the virtual screening approach, i.e. a structure-based drug design (SBDD) or a ligand-based drug design (LBDD). Finally, the critical analysis of these computational strategies and the corresponding results points out the necessity of combining computational and in vitro or in vivo assays for the identification of effective new anti-adipogenic molecules for obesity control. It also evidences that translating molecular docking and virtual screening results into successful drug candidates for adipogenesis and obesity control remains a huge challenge.

Keywords

Adipogenesis drug design molecular docking virtual screening potential ligand

Introduction

Obesity has become a worldwide health problem, affecting 650 million people in both developed and developing countries. This is particularly relevant because of its impact on metabolic syndrome, diabetes, insulin resistance, cardiovascular diseases, and various types of cancer, which represent the main causes of morbidity and mortality worldwide.¹ Obesity is due to an imbalance between caloric intake and energy expenditure, which mainly depends on bad eating habits and low physical activity, although genetic factors have also been reported. To date, obesity management mainly relies on changes in lifestyle and the use of a few drugs, but treatment failures are frequently observed because people with obesity struggle to maintain a healthy lifestyle and use treatments that are expensive and with side effects.² Therefore, the search for new methods for obesity control is of high priority.

For this, many strategies are currently used, including the use of 2D and 3D in vitro cultures of murine 3T3-L1 preadipocytes, human mesenchymal stem cells (MSC), 3D bioprinting of adipose tissue, microphysiological systems (organ-on-a-chip), and animal models among other; these methods allow the evaluation of plant extracts, metabolites, newly synthesized molecules or previously approved or failed/abandoned compounds for use in a different disease.³ But these studies are expensive and time-consuming.

An alternative approach for drug discovery that has gained attention in recent years is the virtual screening of large libraries of small molecules to identify new compounds that are most likely to bind to a target protein and induce a biological effect. These libraries contain millions of chemical structures that can be purchased at an accessible cost to confirm their biological activity. Some of them, such as ZINC,⁴ offer free access for analyses, but some molecules are not synthesized yet and others are only theoretical. In other libraries, such as ChemBridge Corp. (https://chembridge.com/), all the compounds are commercially available.

One approach for the computational identification of new potentially active molecules among chemical libraries is the high-throughput virtual screening without any specific requirements for the ligand structure. This structure-based drug design (SBDD) approach ( Figure 1 ) considers the three-dimensional structure of the target protein and its complementary with the molecules of the library through molecular docking. It can evaluate various binding sites for each ligand and classify them through their binding affinity.⁵ It is also possible to focus on a specific kind of molecule, such as FDA-approved drugs. This option known as drug repurposing (drug repositioning, reprofiling or re-tasking) may allow proposing new indications for approved drugs currently recommended for specific diseases, reducing the high costs and the time required for the development of new, safe, and effective molecules.⁶

Figure 1.

General scheme for virtual screening of potential antiadipogenic molecules. New ligands for the selected target can be retrieved from molecular docking within a complete chemical database of a subset of a library. Then, ADME-Tox analysis and molecular dynamic simulation contribute to the selection of the best ligands, that should be further validated in in vitro and in vivo assays. This figure has been created by the authors.

The second virtual screening strategy considers the original structure of a known ligand of the target protein and investigates compounds with structural similarity with the aim of improving biological activity. The first step of this ligand-based drug design (LBDD) or ligand-based virtual screening (LBVS) method ( Figure 1 ) does not involve evaluating the ligand-protein interactions, but it is always followed by molecular docking analyses to better assess ligand selection.⁵ The main objective of this method is to identify derivatives of known ligands to improve their affinity for the target and therefore, their biological activity.

New applications, improved algorithms and developments in deep learning can positively impact virtual screening results.⁷ Other analyses can also be performed to obtain stronger virtual screening results. The system biology approach helps identifying the best target from the prediction of protein-protein interaction networks whose organization and topology reveal key proteins, which may represent a new valuable target for virtual screening.⁸ On the other hand, molecular dynamics simulations can propose the best conformations of the target for molecular docking assays, as well as a comprehensive description of the molecular interactions in the protein-ligand complex.⁹ Virtual screening results should also be completed by prediction of the absorption, distribution, metabolism and excretion, and toxicity (ADME-Tox) profile of select molecules with suitable pharmacological and toxicological properties for drug candidates, avoiding unnecessary waste of time and money and late-stage failures during drug evaluation and development.¹⁰ Consequently, the main advantages of virtual screening and these complementary computational methods when compared with other established strategies for drug design are that they can save time and money by reducing unnecessary experimental work. Moreover, they also contribute to reducing animal experimentation. They can predict new ligands with suitable affinity for the target, but experimental validation is always needed to confirm their biological activity and their potential as new drugs.

The significance of virtual screening coupled with molecular docking for the design of new therapeutic molecules is demonstrated by some successful examples of FDA-approved drugs that have been discovered using these computational methods.^11,12 Thus, the SBDD strategy allowed the discovery of several FDA-approved drugs. This list includes, but is not limited to saquinavir and amprenavir that target the HIV-1 protease and alter the formation of viral particles, dorzolamide that inhibits the enzyme carbonic anhydrase II and reduces intraocular pressure in the treatment of glaucoma, the ACE (angiotensin-converting enzyme) inhibitor captopril that is mainly used to reduce high blood pressure (hypertension), crizotinib that inhibits the overexpressed c-Met in inflammatory anaplastic lymphoma kinase-positive myofibroblastic tumors, and axitinib that targets the VEGF kinase domain to reduce angiogenesis in patients with advanced renal cell carcinoma. Moreover, luminespib is a highly potent inhibitor of the adenosine triphosphatase activity of heat shock protein 90 (HSP90) with potential antineoplastic activity, which is in a clinical trial phase. The LBDD strategy has also proved to play an essential role in the development of some drugs that are currently used, such as norfloxacin, a fluorinated quinolone carboxylic acid derivative that inhibits the bacterial enzyme DNA gyrase and is recommended for urinary tract infections, the selective serotonin receptor agonist zolmitriptan that narrows blood vessels in the treatment of headaches, or losartan, an angiotensin II receptor blocker (ARB) that treats high blood pressure and heart failure.

The field of virtual screening is rapidly expanding to support the pharmaceutical industry and research on anti-obesity drugs. Body weight gain is related to fat accumulation, i.e. growth of white adipose tissue (WAT) due to an increase in size (hypertrophy) and number (hyperplasia) of adipocytes; the other cellular components of WAT, such as adipocyte progenitors, leukocytes, and endothelial cells, are also affected, and altogether these alterations result in a dysfunctional WAT. In addition to lipid accumulation, hypoxia and inflammatory processes occur. Hence, the proteins involved in the adipogenesis process and biological functions of adipocytes are attractive molecular targets for the development of new therapeutic methods for obesity control.¹³ To the best of our knowledge, there are no existing reviews about the use of these proteins in virtual screening strategies. Therefore, the aim of this literature review is to describe the current advances in the development of new anti-adipogenic molecules through virtual screening combined with molecular docking with these specific proteins to identify new ligands. Studies that have performed virtual screening for the development of antiadipogenic drugs without considering any targets were not considered in this review. First, we present the relevant targets for the discovery of new anti-obesity compounds from virtual screening. Then, we describe the main results obtained from a structure-based drug design (SBDD) strategy based on high-throughput virtual screening of a complete chemical database, FDA-approved drugs, and metabolites of natural origin, or a ligand-based drug design (LBDD) method. Finally, we compare and discuss the advantages and limits of the different computational strategies used for the prediction of new ligands of relevant proteins participating in adipogenesis, which could represent new anti-adipogenesis drugs for obesity control.

PPARγ, Crif1, SIRT1, ERβ, PC1, FTO, Mss51, and FABP4 as relevant targets for the discovery of new anti-obesity compounds from virtual screening

Adipogenesis is a complex process that is regulated by many proteins. Some of them are directly involved in the formation of mature adipocytes, while others are more related to their function. Therefore, all the molecular targets selected for virtual screening have an impact on adipose tissue and have been identified as dysregulated in obesity ( Figure 2 ).

Figure 2.

Protein targets for inhibition of adipogenesis and adipocyte functions. This figure has been created by the authors.

PPARγ—One of the most used targets for the development of new drugs against adipogenesis, is PPARγ (peroxisome proliferator-activated receptor γ). This nuclear hormone receptor is mainly expressed in adipose tissue and is essential for adipose differentiation through the activation of C/EBPα (CCAAT/enhancer binding protein α) and other genes. Activation of PPARγ promotes fatty acid uptake, triglyceride formation, and lipid droplet storage, thereby improving insulin sensitivity and glucose metabolism. Activation of PPARγ leads to positive regulation of Glut4 (glucose transporter type 4), enhancing glucose transport into cells, increasing activation of PI3 K (phosphoinositide 3-kinase) and improving glucose utilization by cells. PPARγ positively regulates CAP (c-Cbl-associated protein), promoting glucose uptake and increasing the expression of IRS-1 and IRS-2 (insulin receptor substrates 1 and 2, respectively), activating insulin receptors. Also, PPARγ activation increases mitochondrial biogenesis, oxygen consumption, and antioxidant defenses, regulates autophagy and increases transcription factors, such as PGC1α (peroxisome proliferator-activated receptor gamma coactivator 1), NRF1-2 (nuclear respiratory factor 1 and 2), and TFAM (mitochondrial transcriptional factor A), involved in the process of mitochondrial biogenesis.¹⁴ Therefore, inhibiting PPARγ is a valuable strategy for reducing adipocyte formation and functions.

Crif1—Adipocyte dysfunction is associated with organelle dysfunction, particularly mitochondria. Several reports in rodent models and human subjects show that the mitochondrial oxidative phosphorylation (OXPHOS) capacity is reduced in obese individuals with obesity. One of the proteins involved is Crif1 (CR6-interacting factor-1), a mitochondrial protein that associates with large mitoribosomal subunits required for the synthesis and insertion of mitochondrial-encoded OXPHOS polypeptides into the mitochondrial membrane.¹⁵ Moreover, its deficiency causes spontaneous adipose inflammation, with macrophage infiltration and systemic insulin resistance.¹⁶ Therefore, the inhibition of Crif1 represents an interesting approach to control adipogenesis.

SIRT1—Another target for the treatment of obesity is Sirtuin 1 (SIRT1), an NAD + -dependent protein deacetylase that eliminates an acetyl moiety from the amino group of lysine residues. Several studies have shown the role of SIRT1 to inhibit adipocyte differentiation both in in vitro (murine 3T3-L1 adipocytes and C3H10T1/2 pluripotent cells) and in vivo models.^17–19 One of the important targets of SIRT1 is PPARγ, whose deacetylation attenuates its functions. This suggests that activators of SIRT1 can be useful for controlling metabolic disorders, including adipogenesis dysregulation in obesity.

ERβ—The association between weight gain and reduced level of estrogen, as well as the beneficial effects of hormone replacement therapy, has suggested the relevance of estrogen and estrogen receptors (ERα and ERβ) in metabolism and energy homeostasis, including adipogenesis and adipocyte functions. However, natural ligands of ERs, such as estradiol, cause severe side effects due to increased proliferation and inflammation process, which has been mainly related to binding on ERα. On the other hand, activation of ERβ produces anti-proliferative effects and inhibits differentiation.^20–22 Congruently, treatments with specific ERβ ligands, which share similarities with estradiol and other isoflavones, have shown to reduce adipogenesis, adipose tissue and lipid accumulation in in vivo and in vitro assays, respectively, without presenting side effects,^23–25 indicating that activation of ERβ is a noteworthy approach for adipogenesis control.

PC1-PC2—The adhesion-like receptor polycystin-1 (PC1) and the calcium channel polycystin-2 (PC2) form a mechanosensing complex that was initially known for its essential role in renal function.²⁶ Polycystins are also involved in other processes, such as osteoblastogenesis. Notably, deficiency of PC1 in osteoblasts stimulates PPARγ and adipogenesis, while the absence of PC2 has opposite effects, indicating the relevance of the functional PC1-PC2 complex for bone adiposity adipogenesis-osteoblastogenesis regulation.^27–29 These observations suggest that alteration of the PC1-PC2 complex may represent an effective strategy for adipogenesis control.

FTO—Several years ago, genetic analyses and genome-wide association studies revealed the existence of single nucleotide polymorphisms in the FTO (fat mass and obesity) gene in childhood and adult obesity.³⁰ Congruently, studies in human populations and murine models have suggested an important role for FTO variants in the regulation of food intake or energy expenditure, as the expression of this m⁶A RNA demethylase peaks at the hypothalamic site in brain tissue that controls food intake. Moreover, a positive correlation was observed between the level of FTO in subcutaneous adipose tissue and the body mass index.^31,32 Moreover, numerous studies have demonstrated that FTO overexpression promotes adipogenesis both in vitro and in vivo, by affecting gene expression in many biological and metabolic processes. Altogether, these data emphasize the potential of inhibiting FTO for obesity control.³³

Mss51—Besides regulating skeletal muscle development, myostatin, or growth/differentiation factor 8 (GDF-8), also impacts metabolism, mainly glucose uptake, insulin sensitivity, and adipogenesis.³⁴ One of the target genes of myostatin is the mitochondrial translational activator Mss51, whose yeast homolog is involved in the respiratory chain in mitochondria. However, it also modulates cellular metabolism in mammals. Thus, the inhibition of Mss51 in in vitro and in vivo studies promotes glycolysis, resistance to diet-induced obesity and insulin sensitivity, which indicates that inhibitors of Mss51 could be pertinent drugs for the treatment of obesity and its comorbidities.^35,36

FABP4—Obesity and its comorbidities are characterized by elevated plasma levels of fatty acid whose trafficking is carried out by a group of specific transporter proteins called fatty acid binding proteins (FABP). Notably, FABP4 (or aP2) that is mainly expressed in adipocytes and macrophages has been implicated in the development of various metabolic disorders. Thus, loss of FABP4 improves insulin resistance and cardiac function and reduces the inflammatory response.³⁷ Additionally, FABP4 interacts with the proadipogenic marker PPARγ, triggering its degradation³⁸ and demonstrating the relevance of FABP4 ligands as attractive drugs for the control of adipogenesis.

All the molecules mentioned above have a critical role in various cellular processes related to adipocyte formation and functions, which motivates the computational search for new ligands (agonists or antagonists) to alter their function and contribute to the control of adipogenesis and obesity.

SBDD and high-throughput virtual screening of chemical databases

The structure-based drug design (SBDD) strategy that focuses on the three-dimensional structure of the target protein has been successfully used to detect new potential ligands of SIRT1, PC1, Crif1, and PPARγ through virtual screening. In these cases, researchers used the available 3D structure of target proteins as the basis for identifying new molecules able to bind and affect the functions of the targets.

Considering the role of SIRT1 in energy homeostasis, Pulla et al. performed a structure-based virtual screening to design novel SIRT1 activators from the commercial Asinex database. They used the previously reported homology model of the AROS (active regulator of SIRT1) binding site to perform docking assays into the allosteric binding site with Glide 5.0 (Glide v5.0 Schrodinger, LLC, New York, NY). High-throughput virtual screening (HTVS) docking followed by (standard precision (SP), and extra precision (XP) docking in combination with the identification of hydrogen bonds and stacking interactions with important amino acid residues in the allosteric site of the protein, led to the prediction of 10 potential inhibitors (I1–I5) and activators (A1–A5). Known inhibitors (salerimide, splitomycin, EX-527, and sirtinol) and an activator (resveratrol) were utilized to corroborate docking scores and interactions with relevant amino acids. In the case of activators, they mainly interact with aromatic and charged residues in the allosteric binding region of SIRT1, which suggests that they were stabilized by the electrostatic and stacking interactions. Importantly, the complexes formed between SIRT1 and compounds A2, A3, and A4, were more stable than the complex form with the known activator resveratrol, probably due to the presence of extra hydrogen bonds (5 with A2 and A4, 6 with A3, versus 3 with resveratrol). Notably, A2 has two main advantages: it interacts with Ser-174, Asp-175, Arg-167and Glu-190 that gathered a good docking score, and it is big enough to occupy the entire groove of the active site pocket. Congruently, the in vitro enzymatic assay confirmed that A2 was the best activator of SIRT1; notably, the lowest concentration of A2 (10 μM) showed a similar effect to the known activator resveratrol at 100 μM. On the other hand, adipogenesis experiments using the 3T3-L1 murine cell line showed that all drugs reduced lipid content by about 40%, but A2 produced the highest reduction at 10 μM and 50 μM. Since SIRT1 deacetylates PPARγ and attenuates its functions, the inhibitory effect of A2 (and the other molecules) on adipogenesis, the authors concluded that A2 is a valuable SIRT1 activator that promotes SIRT-1-deacetylation of the adipogenic marker PPARγ to reduce adipogenesis.³⁹

Because of the relevance of the PC1-PC2 complex for adipogenesis in osteoblastogenic differentiation, Xiao et al. aimed to identify compounds able to disrupt the coil-coil interaction domain and alter protein functions. For this, they performed a molecular docking and virtual screening in the 1000 diverse fragment-like molecules from the ZINC database using the MOE docking facility (Chemical Computing Group) version 2015, and the PC1:PC2 coiled-coil model. One molecule, Zinc01442821 (4-phenyl-1H-pyrrole-3-carboxylic acid), was selected as the best ligand of PC2. Notably, it forms hydrophobic interactions with Val880 and Leu881 residues involved in the PC1-PC2 coiled-coil stabilization, and with Arg877, Arg878, and Lys874 in the central region that is essential for the PC1:PC2 interactions. Experimental assays in HEK-293 T kidney and MC3T3-E1 osteoblastic cells strongly suggested that Zinc01442821 interacts with PC2 to disrupt PC1/PC2 functions and enhance calcium signaling. Additionally, the mRNA expression of the adipogenic marker PPARγ and its downstream gene aP2 was significantly reduced, which is consistent with an inhibition of adipogenesis.⁴⁰

By a virtual screening in the Life Chemicals database (462608 compounds) using the program Autodock_vina for molecular docking, Xiang et al. searched for new ligands of Crif1, focusing on molecules able to bind around His120 that are important for protein functions. A set of 13 compounds with binding energy lower than −12.0 kcal/mol was selected; they established multiple interactions with Crif1, mainly with hydrophobic amino acids. Five compounds named F0382-0033, F3408-0076, F1430-0134, F3408-0031, and F1430-0130 potentially affect the interaction of Crif1 with the protein kinase cyclic adenosine monophosphate-activated catalytic subunit alpha (PRKACA) complex, and therefore inhibit Crif1 functions. Interestingly, they were able to suppress adipogenic differentiation of human bone marrow mesenchymal stem cells (H-BM-MSCs), which strongly suggests that these compounds are bona fide inhibitors of Crif1.⁴¹

Recently, Ahmad et al. identified four molecules (Z1982689600, Z2235802137, Z2235801970, and Z2037275165) as potential new PPARγ ligands by using AutoDock Vina for molecular docking and virtual screening in the Enamine library. Remarkably, these compounds target the same catalytic pocket as SR1664, a known PPARγ antagonist used as the positive control, and interact with 12 of the 31 residues involved in SR1664 binding. Moreover, the four compounds have better binding affinity values toward PPARγ than SR1664, which indicates that these ligands could be more potent antagonists than the control molecule. Additionally, the ADME-Tox analyses suggest that they are safe molecules.⁴² However, no experimental assays were performed to confirm the biological effects of these molecules.

SBDD and virtual screening among FDA-approved drugs

The SBDD approach has also been used to evidence new applications for FDA-approved drugs. Namely, two groups have investigated the potential of currently used drugs as ligands of FAPB4 and ERβ with the aim of identifying new anti-adipogenic molecules.

Wang et al. used four human FABP4 tridimensional structures to perform a molecular docking screening with AutoDock Vina among 1500 compound FDA-approved drugs. The top 10 molecules with the best binding free energy were identified as potential FABP4 inhibitors and five compounds were selected considering their commercial availability and lack of secondary effects. Only three of them inhibited FABP4 activity in in vitro assays. Levofloxacin, a broad-spectrum antibiotic that interacts with the four models, was the most potent FABP4 inhibitor in vitro showing 70.01 ± 12.15% inhibition of FABP4 activity at 10 μM. Additionally, in binding assays, levofloxacin exhibited an IC50 value of 5.83 μM, which is very close to the IC50 of 7.90 μM obtained for arachidonic acid, an endogenous ligand used as control. Levofloxacin (10 μM) inhibited adipolysis in 3T3-L1 cells, as does benzbromarone (10 μM) that is used as a positive control, suggesting that levofloxacin is a true FABP4 inhibitor. However, levofloxacin did not affect lipid accumulation, indicating that it does not promote PPARγ expression, which makes it a good candidate for adipogenesis control.⁴³

Our group also used the drug repurposing approach to identify new ligands for ERβ among a set of 1615 FDA approved drugs included in the ZINC15 database using the Autodock Vina server for molecular docking assays. Using the interaction fingerprint and ΔGb values of known ER modulators (geneticin, S-Equol, tamoxifen, and raloxifene) included as control compounds, ten FDA-approved drugs were selected and the four commercially available ones were considered for further analyses. Notably, Mefloquine, Ezetimibe, Ketoprofen, and Palonosteron were able to establish predominantly hydrophobic interactions with the active site of ERβ.⁴⁴ Unfortunately, none of these drugs was able to affect 3T3-L1 adipogenic differentiation of murine 3T3-L1 cells.⁴⁵ Despite all the advantages and qualities of virtual screening, the biological effects of predicted ligands must always be confirmed in experimental assays.

SBDD and virtual screening in a specific dataset of metabolites of natural origin

When researchers have a specific interest in a certain kind of ligands, it is possible to perform the virtual screening in a restricted chemical library that contains this type of molecule. Notably, several groups have taken advantage of the biological potential of plant metabolites to identify new ligands for FTO and Mss51.

Flavonoids present mainly in foods of plant origin are a broad class of low molecular weight molecules, characterized by the flavon nucleus. They have been shown to exhibit several biological properties, including inhibitory effects on lipolysis and adipogenesis. Therefore, Mohammed et al. used the docking software AutoDock 4.2 to perform a virtual screening among flavonoid molecules contained in the Drug Bank database⁴⁶ to identify new inhibitors of FTO, an interesting target for obesity control. Although binding energy values were relatively high (−3.96 to −1.78 kcal/mol), the authors identified flavonoids that established distinct interactions (mainly hydrogen bonds) with amino acid residues located in different regions of FTO. They also suggested that flavonoids such as quercetin and kaempferol affect adipogenesis by acting as FTO antagonists, although they did not perform any experimental assay.⁴⁷

Recently, Ali et al. have used the SBDD strategy to identify natural compounds with affinity for Mss51. For this, they modeled the three-dimensional structure of Mss51 and used it for a virtual screening from the Herbal and Specs chemical database using AutoDock for molecular docking. The binding site of Mss51 composed of two binding pockets (P1 and P2) was predicted using DS 2021 and I-TASSER. Three compounds that approve the Lipinski filter and exhibit satisfactory ADMET properties were predicted to have a high affinity for Mss51. In addition, a 100 ns molecular dynamics simulation was performed using GROMACS to investigate interactions and complex stability. The analysis of the different parameters of the simulation such as RMSD, RMSF, radius of gyration (Rg), and solvent accessible surface area (SASA) showed that ZINC00338371 and ZINC08214878 formed a more stable complex with Mss51 than ZINC95099599. The evaluation of hydrogen bonds and Gibbs’ free energy (GFE) landscapes indicated that ZINC00338371 is most likely an Mss51 inhibitor. Congruently, binding free energy values confirmed that the ZINC00338371-Mss51 complex is the most stable (−229 kJ/mol), which suggests that it could be a new Mss51 inhibitor with anti-adipogenesis potential, but this remains to be demonstrated.⁴⁸

LBDD and virtual screening of known ligand derivatives

Instead of searching among a full chemical library, several groups have taken advantage of known ligands of PPARγ and ERβ targets and have restrained their LBDD virtual screening among molecules with structural similarity, with the aim of designing better ligands with improved biological activity.

Many natural compounds, including polyphenols, have been identified as bioactive molecules. Among them, ferulic acid inhibits adipogenesis and lipogenesis through the modulation of PPARγ. Therefore, Senthil et al. selected 34 ferulic acid derivatives in PubChem and ZINC databases and confirmed that they maintain drug-like standards by computational assessment of physicochemical, pharmacokinetics, and pharmacodynamics properties. Then, using the Glide 4.0 XP scoring function and docking protocol for molecular docking assays with PPARγ (PDB ID: 4JAZ), they showed that ferulic acid and three derivatives named isoproterenol, 3-hydroxy-4-methoxycinnamic acid, and caffeic acid, presented similar binding affinity as resveratrol and troglitazone, two known PPARγ ligands that modulate fat accumulation. Moreover, they share conserved interaction with Ser342, Lys263, Glu259, Phe264, His266, Arg250, Ile281, Phe360, and Phe244. The complexes were stabilized by a 50 ns molecular dynamic simulation, as shown by RMSF (root mean square fluctuation) and RMSD (root mean square deviation) values. Altogether, these results evidenced that a ligand-based drug design is a valuable approach for identifying molecules that help understanding the mechanism of PPARγ-ligand interaction and function.⁴⁹ However, their biological effect remains to be demonstrated.

Our group recently performed a LBDD for new ERβ activators through searches in ZINC15, PubChem, and Molport databases, taking as references, substructures of two ligands co-crystallized with ERβ and the known ERβ ligand, S-equol, as scaffolds. Using the AutoDock Vina program for docking assays, a set of 63,728 compounds was identified, of which 114 met Lipinski’s rule and had a Tanimoto coefficient (TC) > 0.7 with respect to known modulators of estrogen receptors (raloxifene, tamoxifen, and S-equol). Based on commercial availability and ADMET analysis with the ADMETlab 2.0 web server, six compounds were selected for molecular dynamic simulation using GROMACS software, which evidenced a better stabilization of ERβ by C2 interactions.⁴⁴ The anti-adipogenic potential of C1 and C2 (10 µM) was then confirmed in murine 3T3-L1 fibroblasts since both molecules decreased lipid accumulation by about 40% and significantly reduced the mRNA expression of proadipogenic markers PPARγ and C/EBPα. These results suggest that C1 and C2 are bona fide ERβ activators, although experiments must be carried out to confirm this assumption. C1 (S-dihydrodaidzein) is an intermediate metabolite in the production of S-equol from the soy isoflavone daidzein, which explains its affinity for ERβ and its antiadipogenic effect and makes it an attractive molecule for further studies. The affinity of C2 (3-(1,3-benzoxazol-2-yl)-benzamine) for ERβ may be related to the presence of benzoxazole and benzamidine groups, as it has been reported for the anti-adipogenic molecules 2-(3-fluoro-4-hydroxyphenyl)-7-vinyl-1,3-benzoxazol-5-ol and 3,5-dimethoxy-(4-methoxyphenyl)-benzamide.⁴⁵

Discussion

The discovery of new drugs for diseases that threaten human health has always been of high priority. However, there is a very long way from experimental assays in the lab to clinical use in patients. Therefore, any methods to shorten and speed up drug development are greatly welcome. Virtual screening can save time and money, by reducing unnecessary experimental work; it also contributes to reducing animal experimentation. Importantly, the experimental validation of the identified compounds revealed that at least some of them, have the expected biological activity.

The significance of virtual screening coupled to molecular docking for the design of new therapeutic molecules through SBDD and LBDD strategies is demonstrated by some successful examples of FDA-approved drugs that have been discovered using these computational methods.^11,12 Moreover, recent reports showed that several groups take advantage of these approaches to identify new inhibitors of key proteins in cancer, such as the protein kinase Akt1 and the NAD ⁺-dependent lysine deacetylase SIRT3, confirming that it is a promising strategy for cancer therapy.^50,51 In this review, we show that virtual screening combined with molecular docking is also a valuable approach for identifying new molecules as anti-adipogenic drugs, with the aim of controlling adipogenesis and therefore, obesity development.

From a search in PubMed using “adipogenesis, drug design, in silico drug design, computational drug design molecular screening, virtual screening, molecular docking, potential inhibitors” as keywords, we selected 10 original research articles that use structure-based drug design and ligand-based drug design techniques for the identification of new inhibitors of adipogenesis regulators through virtual screening. The existence of these reports confirms that virtual screening is being used by groups that are looking for new anti-adipogenesis molecules for obesity control. However, the limited number of articles indicates that this strategy is still in the early steps in this area; congruently, publication dates range from 2014 to 2024, i.e. only a decade. However, there is no doubt that this approach is rapidly drawing attention; the computational screening of libraries of small molecules is expanding and various researchers are adopting this methodology to identify new compounds to inhibit adipogenesis and contribute to controlling obesity.

An important point is the selection of the target; two works focused on the well-characterized pro-adipogenic marker PPARγ,^40,42 but the others used less evident objectives although they all have connections with adipogenesis^{39–41,43,44,47,48} ( Figure 2 ). System biology studies based on protein-protein interaction networks could help to increase the list of targets and therefore the discovery of new compounds. Moreover, the use of a stable protein conformation obtained from molecular dynamics simulations could contribute to a better selection of true ligands.

The second relevant aspect is the selection of computational approach; some groups considered the 3D structure of the target in a SBDD approach and performed a high-throughput virtual screening of a complete chemical database or FDA-approved drugs to identify new ligands for the target protein,^39–43 and others preferred the LBDD strategy and restrained the search for a specific type of molecules in a subset of a library without considering the target structure. A summary of the main strategies developed to identify new potential anti-adipogenic compounds through virtual screening and described in our systematic review is shown in Figure 1 .

Using a SBDD strategy or LBDD strategy to screen a full or a shortened database represents equally valuable approaches, and the decision is generally justified by the background of the project and the target molecules. Although the search in a complete database is thought to increase the probability of identifying new ligands, other works preferred to look for compounds with structure similarity to known ligands or metabolites of natural origin, which may improve their biological activity and innocuousness, respectively. For example, the known anti-adipogenic effects of S-equol prompted us to look for structural analogues as new ligands of ERβ⁴⁴; similarly, inhibitory effects of ferulic acid on PPARγ justified the search for derivatives of this molecule.⁴⁹

In the search for new drugs, the main advantages of virtual screening combined with molecular docking are low costs and time reduction; there is no need for supercomputers, and most docking servers and libraries provide free access. Additionally, predicting the physicochemical, drug-likeness, and ADME-Tox properties of the selected molecules is a relevant filter to improve the selection of the best ligands. However, this significant filter is barely exploited and most of the works reviewed here do not perform these analyses to assess the quality of the identified compounds. This means that the identified molecules could be toxic or mutagenic, which may limit their use, although they are true ligands of the targets.

Only some of the reviewed articles include experimental data whose results validate the potential of in silico-identified compounds.^{39–41,43–45} For others,^42,47–49 we did not find more recent papers with experimental data. However, we cannot discard that their validation is currently in progress since three of these papers were published in 2023. In general, not all the identified compounds produce the expected biological effect. For example, only two of the six compounds identified as potential ERβ from our virtual screening methodology were able to inhibit adipogenesis in vitro.^44,45 This discrepancy between in silico prediction and in vitro effect could be explained by the characteristics of the computational assays. Thus, the 10 selected reports only use one server for molecular docking, which may be associated with potential biases, false positives, and limitations in docking algorithms, leading to the identification of compounds without any biological effect. Only three performed molecular dynamics simulations of the protein-compound complex to complete the selection of the ligands.^44,48,49 Using various docking programs and evaluating stable conformations of the protein target or target-compound complex by molecular dynamics simulation, among others, would certainly increase the biological potential of the compounds. Moreover, new applications and developments of deep learning will undoubtedly contribute to improving virtual screening quality. However, it is important to point out that in silico analyses only consider a restricted system, mainly based on the target protein and the ligand, while other components, such as other interacting proteins, lipids of cytoplasmic membranes, ions, etc., could affect their interaction in biological assays and even more in biological systems. Additionally, even if virtual screening identifies molecules that fulfill the five rules of Lipinski and can be taken orally as a drug, this remains a prediction that must be verified in animal models or clinical studies. Therefore, translating virtual screening results into successful drug candidates for adipogenesis and obesity control still remains a big challenge.

Another issue is that adipogenesis control may require activation or inhibition of the selected target, depending on their function. However, in most studies, the molecular docking strategy does not specifically focus on the binding site for agonists or antagonists. Some works compare the interaction profiles with those of known ligands (activators or inhibitors) to infer the potential effect of the molecules and/or use the ΔG values.^42,49 Additionally, when reports do not perform in vitro assays to confirm the biological activity of the selected ligands, it is possible that the selected molecules have an undesirable biological effect, i.e. they may promote adipogenesis.

Drug repurposing has gained a lot of attention in the past few years, proportioning an additional value to old drugs. Notably, it has been revealed as an attractive and effective approach for attending challenging diseases such as cancer.^52,53 For example, it has been experimentally demonstrated that some anti-alcoholic, anthelmintic, and antiepileptic drugs sensitize cancer cells, which could have a positive impact on cancer therapy; importantly, there are some reports about computational approaches to identify non-cancer drugs as potential ligands of relevant protein in cancer. Virtual screening of FDA-approved drugs allows the discovery of unexpected properties of known molecules for which all pharmacological and toxicity properties have been characterized, which reduces time and cost for application. Notably, in one of the two reports selected for this review, this repositioning strategy coupled to in vitro assays revealed the anti-adipogenic effects of the antibiotic molecule Levofloxacin.⁴³ In contrast, the two FDA drugs identified as ERβ ligand were not able to inhibit adipogenesis,^44,45 indicating again that the biological evaluation is a required step.

Finally, the critical analysis of these computational strategies and the corresponding results of the studies described in this review could serve as a guide for further investigation on new ligands for proteins related to adipogenesis and adipocyte functions. The selection of SBDD and LBDD techniques depends on the context of the investigation project, and both strategies have been successfully applied for the discovery of novel drug-like candidates; however, their combination using structural information of both ligand and target could improve the quality of the ligands. Additionally, a better characterization of the structural features of the target or known ligand, through molecular dynamic simulation or description of a pharmacophore model for example, could lead to the identification of compounds with a higher affinity.

Conclusion

Virtual screening coupled to molecular docking is quickly emerging as a useful approach for the prediction of new anti-adipogenic molecules, although interactions and biological effects have only been assessed for few of them; and when biological assays were performed, not all the identified ligands appeared to be effective, which points out the limitations of virtual screening. Based on this literature review, we propose some recommendations for developing new ligands of specific proteins through molecular docking and virtual screening of chemical libraries, with the aim of identifying new potential anti-adipogenic molecules. Complementary computational studies, such as the comparative utilization of various docking servers, the implementation of a more effective program with machine learning, and the realization of molecular dynamic simulation assays together with the prediction of ADME-Tox properties must be performed for identifying better ligands that truly interact with the target protein and exert the expected biological effects. The identification of more efficient molecules could be improved by focusing on binding sites for known agonists or antagonists depending on the target, with the aim of selecting ligands that exert the required effects on adipogenesis. Finally, the critical analysis of these computational strategies and their results, emphasizes the necessity of combining computational and in vitro or in vivo assays for the identification of new efficient anti-adipogenic molecules for obesity control. Translating virtual screening results into successful drug candidates for adipogenesis and obesity control is still a big challenge, and improvements in algorithms and computational approaches are still needed to predict better ligands.

Footnotes

Acknowledgements

MFTR, ERM and LAM are also supported by SIP-IPN and COFAA-IPN, Mexico. All authors also hold fellowships from CONAHCyT, Mexico.

Author contributions

María F. Torres-Rojas: investigation, formal analysis, and writing—original draft. Gilberto Mandujano-Lazaro: investigation, formal analysis, and writing—review & editing. Esther Ramirez-Moreno: investigation, formal analysis, and writing—review & editing. Laurence A. Marchat: conceptualization, investigation, writing—original draft, and writing—review & editing.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Laurence A Marchat

References

WHO W.H.O. Obesity and Overweight. [(accessed on 15 April 2024)]. Available online: https://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight.

Fruh

. Obesity: risk factors, complications, and strategies for sustainable long-term weight management. J Am Assoc Nurse Pract 2017; 29: S3–S14.

Pamplona

Zoehler

Shigunov

, et al. Alternative methods as tools for obesity research. In Vitro and In Silico Approaches. Life (Basel) 2022; 13: 08.

Sterling

Irwin

. ZINC 15–Ligand Discovery for everyone. J Chem Inf Model 2015; 55: 2324–2337.

Lavecchia

Di Giovanni

. Virtual screening strategies in drug discovery: a critical review. Curr Med Chem 2013; 20: 2839–2860.

Jourdan

Bureau

Rochais

, et al. Drug repositioning: a brief overview. J Pharm Pharmacol 2020; 72: 1145–1151.

Kimber

Chen

Volkamer

. Deep learning in virtual screening: recent applications and developments. Int J Mol Sci 2021; 22: 4435.

Murakami

Tripathi

Prathipati

, et al. Network analysis and in silico prediction of protein-protein interactions with applications in drug discovery. Curr Opin Struct Biol 2017; 44: 134–142.

Zhao

Chen

. Insights into the molecular mechanisms of protein-ligand interactions by molecular docking and molecular dynamics simulation: a case of oligopeptide binding protein. Comput Math Methods Med 2018; 2018:3502514. https://doi.org/10.1155/2018/3502514

10.

Alqahtani

. In silico ADME-tox modeling: progress and prospects. Expert Opin Drug Metab Toxicol 2017; 13: 1147–1158.

11.

Leelananda

Lindert

. Computational methods in drug discovery. Beilstein J Org Chem 2016; 12: 2694–2718.

12.

Cui

Aouidate

Wang

, et al. Discovering anti-cancer drugs via computational methods. Front Pharmacol 2020; 11: 33.

13.

Longo

Zatterale

Naderi

, et al. Adipose tissue dysfunction as determinant of obesity-associated metabolic complications. Int J Mol Sci 2019; 20: 2358. Therefore.

14.

Qiu

Gan

Wang

, et al. The global perspective on peroxisome proliferator-activated receptor γ (PPARγ) in ectopic fat deposition: a review. Int J Biol Macromol 2023; 253: 127042.

15.

Greber

Bieri

Leibundgut

, et al. Ribosome. The complete structure of the 55S mammalian mitochondrial ribosome. Science 2015; 348: 303–308.

16.

Ryu

Kim

, et al. Crif1 deficiency reduces adipose OXPHOS capacity and triggers inflammation and insulin resistance in mice. PLoS Genet 2013; 9: e1003356.

17.

Picard

Kurtev

Chung

, et al. Sirt1 promotes fat mobilization in white adipocytes by repressing PPAR-gamma [published correction appears in Nature. 2004 Aug 19;430(7002):921]. Nature 2004; 429: 771–776.

18.

Zhou

Zhang

, et al. SIRT1 Inhibits adipogenesis and promotes myogenic differentiation in C3H10T1/2 pluripotent cells by regulating Wnt signaling. Cell Biosci 2015; 5: 61. Published 2015 Nov 12.

19.

Zhou

Song

Peng

, et al. SIRT1 Suppresses adipogenesis by activating Wnt/β-catenin signaling in vivo and in vitro. Oncotarget 2016; 7: 77707–77720.

20.

Santen

Allred

Ardoin

, et al. Postmenopausal hormone therapy: an Endocrine Society scientific statement. J Clin Endocrinol Metab 2010; 95: s1–s66.

21.

Song

Indukuri

, et al. ERα and ERβ homodimers in the same cellular context regulate distinct transcriptomes and functions. Front Endocrinol (Lausanne) 2022; 13: 930227. Published 2022 Jul 6.

22.

Warner

Huang

Gustafsson

. Estrogen receptor β as a pharmaceutical target. Trends Pharmacol Sci 2017; 38: 92–99.

23.

Yepuru

Eswaraka

Kearbey

, et al. Estrogen receptor-{beta}-selective ligands alleviate high-fat diet- and ovariectomy-induced obesity in mice. J Biol Chem 2010; 285: 31292–31303.

24.

Ponnusamy

Tran

Harvey

, et al. Pharmacologic activation of estrogen receptor β increases mitochondrial function, energy expenditure, and brown adipose tissue. FASEB J 2017; 31: 266–281.

25.

Mandujano-Lázaro

Galaviz-Hernández

Reyes-López

, et al. A short S-equol exposure has a long-term inhibitory effect on adipogenesis in mouse 3T3-L1 cells. Appl Sci 2021; 11: 9657.

26.

Zhou

. Polycystins and primary cilia: primers for cell cycle progression. Annu Rev Physiol 2009; 71: 83–113.

27.

Xiao

Zhang

Cao

, et al. Conditional disruption of Pkd1 in osteoblasts results in osteopenia due to direct impairment of bone formation. J Biol Chem 2010; 285: 1177–1187.

28.

Xiao

Dallas

Qiu

, et al. Conditional deletion of Pkd1 in osteocytes disrupts skeletal mechanosensing in mice. FASEB J 2011; 25: 2418–2432.

29.

Xiao

Cao

Liang

, et al. Osteoblast-specific deletion of Pkd2 leads to low-turnover osteopenia and reduced bone marrow adiposity. PLoS One 2014; 9: e114198. Published 2014 Dec 2.

30.

Chang

Park

, et al. The fat mass- and obesity-associated (FTO) gene to obesity: lessons from mouse models. Obesity (Silver Spring). 2018; 26: 1674–1686.

31.

Zhao

Yang

Sun

, et al. FTO and obesity: mechanisms of association. Curr Diab Rep 2014; 14: 486.

32.

Tews

Fischer-Posovszky

Wabitsch

. Regulation of FTO and FTM expression during human preadipocyte differentiation. Horm Metab Res 2011; 43: 17–21.

33.

Huang

Chen

Wang

. Studies on the fat mass and obesity-associated (FTO) gene and its impact on obesity-associated diseases. Genes Dis 2022; 10: 2351–2365. Published 2022 May 6.

34.

Rahimov

King

Warsing

, et al. Gene expression profiling of skeletal muscles treated with a soluble activin type IIB receptor. Physiol Genomics 2011; 43: 398–407.

35.

Rovira Gonzalez

Moyer

LeTexier

, et al. Mss51 deletion enhances muscle metabolism and glucose homeostasis in mice. JCI Insight 2019; 4: e122247. Published 2019 Oct 17.

36.

Moyer

Wagner

. Mammalian Mss51 is a skeletal muscle-specific gene modulating cellular metabolism. J Neuromuscul Dis 2015; 2: 371–385.

37.

Furuhashi

Saitoh

Shimamoto

, et al. Fatty acid-binding protein 4 (FABP4): pathophysiological insights and potent clinical biomarker of metabolic and cardiovascular diseases. Clin Med Insights Cardiol 2015; 8: 23–33. Published 2015 Feb 2.

38.

Adida

Spener

. Adipocyte-type fatty acid-binding protein as inter-compartmental shuttle for peroxisome proliferator activated receptor gamma agonists in cultured cell. Biochim Biophys Acta 2006; 1761: 172–181.

39.

Pulla

Alvala

Sriram

, et al. Structure-based drug design of small molecule SIRT1 modulators to treat cancer and metabolic disorders. J Mol Graph Model 2014; 52: 46–56.

40.

Xiao

Baudry

Cao

, et al. Polycystin-1 interacts with TAZ to stimulate osteoblastogenesis and inhibit adipogenesis. J Clin Invest 2018; 128: 157–174.

41.

Xiang

Ran

Chen

, et al. CR6-interacting factor-1 contributes to osteoclastogenesis by inducing receptor activator of nuclear factor κB ligand after radiation. World J Stem Cells 2020; 12: 222–240.

42.

Ahmad

Alghamdi

. Molecular docking analysis of PPARγ antagonists for obesity associated diabetes management. Bioinformation 2023; 19: 633–637. Published 2023 May 31.

43.

Wang

Law

, et al. Discovery of FDA-approved drugs as inhibitors of fatty acid binding protein 4 using molecular docking screening. J Chem Inf Model 2014; 54: 3046–3050.

44.

Méndez-Álvarez

Torres-Rojas

L-R

, et al. Ligand-Based virtual screening, molecular docking, and molecular dynamic simulations of new β-estrogen receptor activators with potential for pharmacological obesity treatment. Molecules 2023; 28: 4389. Published 2023 May 27.

45.

Torres-Rojas

Mandujano-Lazaro

Lopez-Camarillo

, et al. S-Dihydrodaidzein and 3-(1,3-benzoxazol-2-yl)-benzamide, two new potential β-estrogen receptor ligands with anti-adipogenic activity. Med Chem 2024; 20: 434–442.

46.

Wishart

Knox

Guo

, et al. Drugbank: a comprehensive resource for in silico drug discovery and exploration. Nucleic Acids Res 2006; 34: D668–D672.

47.

Mohammed

Al-Numair

Balakrishnan

. Docking studies on the interaction of flavonoids with fat mass and obesity associated protein. Pak J Pharm Sci 2015; 28: 1647–1653.

48.

Ali

Ahmad

Shaikh

, et al. Mss51 protein inhibition serves as a novel target for type 2 diabetes: a molecular docking and simulation study. J Biomol Struct Dyn 2024; 42: 4862–4869.

49.

Senthil

Sakthivel

Usha

. Structure-based drug design of peroxisome proliferator-activated receptor gamma inhibitors: ferulic acid and derivatives. J Biomol Struct Dyn 2021; 39: 1295–1311.

50.

Etikyala

Reddyrajula

Vani

, et al. An in silico approach to identify novel and potential Akt1 (protein kinase B-alpha) inhibitors as anticancer drugs. Mol Divers 2024: 1–24. Published 2024 May 26. doi:https://doi.org/10.1007/s11030-024-10887-9

51.

Ullah

Rehman

Islam

, et al. Identification of small molecular inhibitors of SIRT3 by computational and biochemical approaches a potential target of breast cancer. Sci Rep 2024; 14: 12475. Published 2024 May 30.

52.

Malla

Viswanathan

Makena

, et al. Revitalizing cancer treatment: exploring the role of drug repurposing. Cancers (Basel) 2024; 16: 1463. Published 2024 Apr 11.

53.

Rauf

Joshi

Olatunde

, et al. Comprehensive review of the repositioning of non-oncologic drugs for cancer immunotherapy. Med Oncol 2024; 41: 22. Published 2024 Apr 23.

Virtual screening combined with molecular docking for the !identification of new anti-adipogenic compounds

Abstract

Keywords

Introduction

PPARγ, Crif1, SIRT1, ERβ, PC1, FTO, Mss51, and FABP4 as relevant targets for the discovery of new anti-obesity compounds from virtual screening

SBDD and high-throughput virtual screening of chemical databases

SBDD and virtual screening among FDA-approved drugs

SBDD and virtual screening in a specific dataset of metabolites of natural origin

LBDD and virtual screening of known ligand derivatives

Discussion

Conclusion

Footnotes

Acknowledgements

Author contributions

Declaration of conflicting interests

Funding

ORCID iD

References