Designing dual inhibitors for the treatment of Alzheimer’s disease as well as Type 2 diabetes mellitus via pharmacoinformatics approach: A step towards better medication for diabetes-associated neurological disorder

Purpose: To design dual inhibitors against Alzheimer’s disease (AD) and type 2 diabetes mellitus (T2DM) via pharmacoinformatics approach. Methods: Dual Drug Candidates (DDC) were designed and explored for their molecular interaction with several AD and T2DM targets. Pterostilbene, a natural anti-T2DM compound was coupled with different cholinesterase inhibitors to design DDC. Orisis Datawarrior online property calculator tools, Autock 4.2 and Hex 5.1 were used to investigate the potency of all DDC relative to positive controls. Results: The study found that DDC2 (pterostilbene methylene linker -octa hydro amino phenothiazine), DDC3 (pterostilbene ethylene linker N-phthalimide) and DDC5 (pterostilbene carbonyl linker 2-methyl-4-aminoquinoline) were the most promising out of all the DDCs. DDC2 showed strong molecular interaction with most of the AD and T2DM targets, including acetylcholinesterase, butrylcholinesterase, β-secretase, receptor for advanced glycation end products and ATP sensitive potassium channel, dipeptidyl peptidase IV and sodium glucose transport protien 2. The findings also revealed the amyloid anti-aggregation potential of DDC. Conclusion: The results show that DDC3 and DDC5 significantly interfer with the primary nucleation process of β amyloid. Thus, DDC2, DDC3 and DDC5 have strong anti-T2DM and anti-AD potential.


INTRODUCTION
Alzheimer's disease (AD) and Type 2 diabetes mellitus (T2DM) are two of the major debilitating diseases sharing common pathophysiologies [1].
Indeed, T2DM is known to increase the risk of AD by 1.6 folds, and untreated T2DM patients develop AD a lot earlier than treated patients [2]. Unfortunately, the worldwide prevalence of both these disorders are increasing at an alarming rate. According to World Alzheimer's Report 2018, 50 million of the world population are affected with dementia and AD, with the figures expected to increase progressively.
On the other hand, International Diabetes Federation 2017 statistics showed that 425 million individuals in the world are diabetic and the number might reach 629 million in the year 2045. In addition, every seven seconds a diabetic death occurs, and around 50 % of these deaths occur in an age group of below 60 years. Brain and pancreas are targeted organs for AD and T2DM, however, being anatomically and physiologically different they share some common pathophysiology. Clinical and animal model evidence have shown that islet amyloid polypeptide of pancreas could activate neuronal degradation and amyloid β peptide misfolding in AD [3]. It has been found that islet amyloid polypeptide enters the brain, links to amyloid β plaques and enhances misfolding of amyloid β.
Occurrence of AD is often associated with T2DM, thus, dual therapy targeting both the diseases could provide a novel alternative treatment approach. The present study deals with the design of some AD-T2DM dual therapeutic compounds and predict their efficacy against different AD and T2DM targets. Numerous natural compounds have been reported to be effective against T2DM, however, ꞌpterostilbeneꞌ, a natural dimethylated resveratol analog was chosen for this study. Pterostilbene improves insulin sensitivity, reduces beta cell apoptosis and ameliorates diabetic nephropathy/retinopathy [5,6]. During this study, pterostilbene was coupled with different acetylcholinesterase inhibitors via linkers (Table  S1). The coupled compounds were screened for physicochemical and toxicity profile before targeting them against AD. Alzheimer's disease is a complex disease with multiple pathways and hypotheses, in this study cholinergic and amyloidogenic pathways were selected.by targeting acetylcholinesterase, butrylcholinesterase, beta-secretase, amyloid β aggregation and receptor for advanced glycation end products. Furthermore, to access the anti-T2DM potential the targets were ATP sensitive potassium channel, Dipeptidyl Peptidase IV, Sodium glucose transport protein 2.

EXPERIMENTAL Preparation of dual drug candidates
The structures of the pterostilbene and different acetylcholinesterase inhibitors (Table S1) were retrieved from PubChem. The structures of the compounds were drawn in ChemDraw 8.0, coupled with methylene, ethylene and carbonyl linkers, and converted to their three-dimensional coordinates in Chem3D 8.0. Then they are subjected to structural optimization to avoid the structural strain through Merck Molecular Force Field (MMFF). Finally, all the compounds were saved in .pdb format for further docking studies.

Physicochemical and toxicity profile screening of dual drug candidates
Physico-chemical and toxicity profile of DDCs was done by Orisis Datawarrior property explorer tool (http://www.openmolecules.org/datawarrior/down load.html). Various parameters, such as topological polar surface area (TPSA), molecular weight, cLogP, the number of hydrogen bond donors, the number of hydrogen bond acceptors, number of rotatable bonds and violations of Lipinski's rule of five [7] were calculated (Table  1). However, Zhao et al [8] method was used to calculate the percentage of absorption.

Retrieval of target proteins and positive control drugs structures from databases
The three-dimensional structure of target proteins acetylcholinesterase (

Molecular docking
Autodock 4.2 was used to dock the ligands with protein following the protocol of Rizvi et al [9]. Energy minimization of each ligand was done by MMFF94 force field and gasteiger partial charges were added. Rotatable bonds were defined after including non-polar hydrogen atoms. Kollman united atom type charges, Solvation parameters and hydrogen atoms were added with the help of AutoDock. 'Auto Grid' was used to set the dimensions of grid (60 x 60 x 60 Å with 0.375 Å point separation

RESULTS
In the present study, a natural diabetic compound ꞌpterostilbeneꞌ was coupled with different acetylcholinesterase (AChE) inhibitors to design five dual targeting compounds and named them as Dual Drug Candidates (DDC) ( Table S1). The physicochemical properties and toxicity potential of DDC1 to DDC5 and control compounds/drugs were tested. Table 1 shows all the physicochemical properties and Lipinski violations and percentage of absorption. According to Lipinski rule [7], DDC1, DDC4 and DDC5 showed only one violation in cLogP value (logarithm of compound partition coefficient between n-octanol and water) while DDC2 and DDC3 showed no violation that is similar to the control compounds (Table 1). Table 2 represents the toxicological potential of DDC, among all DDCs, DDC1, DDC4 and DDC5 showed high mutagenic effect and reproductive effect whereas DDC2 and DDC3 showed only reproductive effect.
The results for cholinesterase interaction with DDCs are presented in Table 3. All DDCs appear to be promising inhibitors of AChE when compared to Tacrine. However, DDC1, DDC2, DDC4 and DDC5 appear to be most potent against AChE with binding energy (∆G) and inhibition constant (Ki) of ꞌ-11.33kcal/mol and 4.94 nMꞌ, ꞌ-11.07kcal/mol and 7.71nMꞌ, ꞌ-11.55kcal/mol and 3.42nMꞌ and ꞌ-11.23kcal/mol and 5.84nMꞌ, respectively. Interestingly, DDC2, DDC3, DDC4 and DDC5 effectively bound to two amino acid residues, namely, S203 and H447 of the catalytic triad of AChE ( Figure 1). Concurrently, compounds DDC1 to DDC5 all showed better binding with butyrylcholinesterase (BChE) than tacrine. Among them, DDC1, DDC2 and DDC5 were the best with ∆G and ki values of ꞌ-8.77kcal/mol and 0.37µMꞌ, ꞌ-9.52kcal/mol and 0.10µMꞌ and ꞌ-8.55kcal/mol and 0.53µMꞌ, respectively. Interestingly, DDC1, DDC4 and DDC5 showed interaction with F329 of BChE peripheral site ( Figure 2).  β-secretase (BACE-1) molecular docking study showed that only two compounds DDC2 and DDC3 appear to be promising when compared with known BACE-1 inhibitor AZD3293 in terms of binding energy. The results of ∆G and Ki for DDC2, DDC3 and AZD3293 against BACE-1 were estimated to be ꞌ-7.95 kcal/mol and 1.49µMꞌ, ꞌ-8.07 kcal/mol and 1.22 µMꞌ and ꞌ-7.97 kcal/mol and 1.43 µMꞌ, respectively (Table 4).    ꞌReceptor for Advanced Glycation End products (RAGE)-DDCsꞌ molecular interaction results are presented in Table 6. Among all DDCs, DDC2, DDC3 and DDC5 showed better interaction with the domain 1 of RAGE i.e., almost equivalent to the control curcumin, whereas, P42, P46, Q47, R48, M102, N103, R104 were the common amino acid residues of RAGE that showed interaction with DDC 2, DDC 3, DDC 5 and curcumin. Hence, it could be concluded that DDC2, DDC3 and DDC5 were the best candidates for amyloid aggregation inhibition. However, to check the anti-T2DM potential, the interaction potential of test compounds were checked with ATP sensitive potassium channel (KATP Channel), dipeptidyl peptidase IV (DPP IV) and sodium glucose transport protein 2 (SGLT2).    (Figure 3) H175, I296, T297 were the commonly interacting amino acid residues of KATP channel with DDC1, DDC2, DDC3 and DDC4 as well as the positive control glimepiride. During DPPIV-DDCs interaction studies, the tested compounds were compared with a known selective DPPIV inhibitor saxagliptin. Unfortunately, none of the tested compounds, DDC1 to DDC5 showed better interactions than the positive control saxagliptin (Table 7). However, 'DDC2-DPPIV interaction' appears to be somewhat promising with ∆G and Ki values of ꞌ-6.58 kcal/mol and 15.06 µMꞌ. Figure 4 shows the interacting amino acids of DPPIV active site interacting amino acid results.  The results from SGLT2-DDCsꞌ interaction showed that DDC1, DDC2, DDC4 and DDC5 interacted with SGLT2 better than a positive control canagliflozin in terms of binding energy (Table 7). ∆G and Ki values of DDC1, DDC2, DDC4 and DDC5 interaction with SGLT2 were ꞌ-8.32 kcal/mol and 0.79 µMꞌ, ꞌ-8.29 kcal/mol and 0.83 µMꞌ, ꞌ-7.60 kcal/mol and 2.67 µMꞌ and ꞌ-8.22 kcal/mol and 0.94 µMꞌ, respectively. In addition, all the tested compounds showed interaction with the most important amino acid residue Q457 of SGLT2 ( Figure 5).

DISCUSSION
Alzheimer's disease (AD) is a complex disease with a strong Type 2 Diabetes Mellitus (T2DM) linkage [2,10]. Interestingly, augmented amyloid plaque accumulation has been observed in the hippocampus of T2DM patients during the autopsy [11]. In addition, T2DM untreated patients have 1.6-fold more chances of developing AD than [2,10]. Till date, available treatment of AD such as use of acetylcholinesterase inhibitors and N-methyl D-aspartate receptor antagonist provides only the symptomatic relief.
However, a total of 132 agents are currently in clinical trials for AD treatment [12]. Unfortunately, in the past, a high failure rate has been observed in AD drug development [12]. The delay in treatment results in poor clinical response in AD patients. Hence early treatment in patients even at pre-clinical stages of AD is recommended. Phase III clinical trials have so far not approved any drug for dual drug therapy thus prompting the present study to design dual targeting compounds that could be plausibly used for the treatment of both linked diseases. To achieve this, we have coupled a natural diabetic compound pterostilbene [5,6] with different acetylcholinesterase (AChE) inhibitors to design five dual targeting compounds and named them Dual Drug Candidates (DDCs).
The major objective was to predict the anti-AD and anti-T2DM potential of DDCs. AD is a complicated disease with tau phosphorylation, cholinergic and amyloidogenic mechanism hypothesis. Out of these three, the cholinergic and amyloidogenic pathways were targeted in the study. Acetylcholinesterase and butrylcholinesterase (BChE) were targets for cholinergic pathways, and β-secretase (BACE1), Beta amyloid aggregation and (RAGE were targets for amyloidogenic pathway. The amino acid residues, W86, E202 and Y337 were found to be involved in binding of acetylcholine to AChE [13]. In this study, DDC2, DDC3, DDC4 and DDC5 showed binding with all the three amino acid residues (W86, E202 and Y337). The catalytic triad of BChE is made-up of S198, E325 and H438 [14]. Out of all DDCs, DDC1, DD2, DDC4, DD5 and positive control tacrine interacted with S198 and H438. It has been reported that D70, F329 and Y332 are important amino acid residues of the peripheral site of BChE [15]. The compounds, DDC1, DDC4 and DDC5 interacted with F329 amino acid residue of BChE peripheral site. The above interaction studies concluded that DDC1 and DDC2 were the best among the five and could be explored further as dual drug candidates.
BACE-1 is involved amyloid precursor pathway and plays an important role in the generation and accumulation of amyloid β in brain. Amino acid residues, D106, K107 and F108 of BACE-1 active site were found to be commonly interacting with DDC2, DDC3 and AZD3293. Similarly, Hassan et al [16] observed strong hydrogen bonding of AZ3293 with K107. In another study, Rajasekhar et al [17] observed that BACE-1 inhibitors interacted with F108 amino acid of BACE-1.
Interestingly, it has been observed that Aβ peptide fragment of 1-42 amino acid residues (Aβ1-42) is more dominant in AD patients than 1-40 amino acid residues Aβ (Aβ1-40) peptide [18]. Lührs et al [18] have deeply studied the 3D structure of Aβ1-42 fragment and found that βturn-β-fold motif are formed from 18-42 amino acid residues while 1-17 amino acids are disordered. The protofilament is formed from parallel intermolecular β sheets of β-turn-β-fold motif of Aβ1-42 peptide. Further these interactions of β-turn-β-fold motifs have shown the sequence cooperativity and selectivity for the Aβ fibril formation. In this study, DDC3 and DDC5 when bound to Aβ17-42 markedly reduced the level of interaction with other native form of Aβ17-42. Similar inhibitory results were obtained by Bibi et al [19] for anticancer drug (bexarotene).
Receptor for advanced glycation end products is immunoglobulin superfamily member that could bind to variety of ligands including Aβ peptides. In fact, RAGE interaction with Aβ peptides is responsible for the influx of circulating Aβ peptides in the brain [20]. It has been observed that up-regulation of RAGE leads to higher accumulation of Aβ peptides in the brain. In another report, augmented levels of RAGE were observed in the AD hippocampus [21]. Therefore, targeting RAGE would have a positive effect on AD treatment [20]. Receptor for Advanced Glycation End products crystal structure has two immunoglobulin domains, where 23 to 118 amino acid residues form domain 1 and 121-231 amino acid residues forms domain 2 that are linked via a short linker. Bibi et al [19] observed that domain 1 of RAGE was critical for the binding of Aβ peptides and transportation across the blood brain barrier. Positively charged surface of RAGE is constructed by R29, K37, K39, K43, K44, R48, K52, R98, R104, K107, K110, R114 and R116 of domain 1 and R216 of domain 2 [22]. In this study, interestingly, DDC2, DDC3 and DDC5 showed better interaction with the domain 1 of RAGE.
ATP sensitive potassium channel (KATP Channel) channels have a major role in glucose triggered insulin secretion of pancreatic β cells. Closure of KATP channel initiates the secretion of insulin, while, opening of KATP channel results in vice versa [23]. KATP channel of pancreatic β cells is an octameric complex with 4 sulfonylurea-receptor regulatory subunits (SUR1) and 4 inward rectifying K channel subunits (Kir6). Sulfonylurea-receptor regulatory subunits 1 and Kir6 are important for functioning of KATP channels; Mg-ADP/ATP interact with SUR1 to stimulate the KATP channel activity, while ATP interact with Kir6 to close KATP channel. In this study, DDC1-4 showed better interaction than glimepiride.
Dipeptidyl Peptidase IV (DPP IV) is responsible for inactivation of incretin hormones glucosedependent insulinotrophic polypeptide and glucagon like peptide-1. These incretin hormones act as triggers for secretion of insulin and regulation of blood glucose level. Thus, targeting DPP IV provides an alternative way to curb T2DM [24]. A comparative analysis of DDCs with a known DPPIV inhibitor Saxagliptin were performed in this study, however, none of the DDCs showed better interaction than saxagliptin.
Proximal convoluted tubules of kidneys have SGLT2 proteins which is responsible for maximum reabsorption of glucose [25]. Therefore, SGLT2 is considered as a newer target for T2DM and their inhibitors have been recently approved by FDA for T2DM treatment [25]. Investigation have shown that Q457 of SGLT2 is responsible for glucose reabsorption [26] and all our compounds interacted with Q457.
Overall, DDC2, DDC3 and DDC5 were the most promising among all dual drug candidates. However, it can be safely stated that DDC2 appears to be the best with potent affinity against most of the targeted proteins. Interestingly, DDC2 showed no violation of Lipinski rule, only reproductive toxicity and strong molecular interaction with cholinesterase, BACE-1, RAGE, KATP channel, DPP IV and SGLT2. The compound DDC2 has been designed on phenothiazine scaffold and coupled with pterostilbene. Phenothiazine is known for its cholinesterase inhibition potential and pterostilbene is an amazing natural anti-diabetic compound. Thus, combining these might provide an interesting dual drug candidate for future AD and T2DM therapy. Nevertheless, some important structural insights critical for binding of dual drug candidates with different AD and T2DM targets were revealed in this study. Parallel molecular docking experiments with positive control for each target helped to get nice comparative analysis on binding behavior of each dual drug candidate.

CONCLUSION
In the present study, five dual drug candidates (DDC) have been successfully designed and their physicochemical and toxicity profile evaluated. Extensive structural and interaction analysis predicts that DDC2, DDC3 and DDC5 are the most promising candidates. However, strong molecular interaction against most AD and all T2DM targets has been observed for DDC2. Thus, DDC2 should be further studied for dual drug therapy. Nonetheless, the results obtained in the present study may reduce the time and cost for the development of drugs against T2DM associated neurological disorders.

Open Access
This is an Open Access article that uses a funding model which does not charge readers or their institutions for access and distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/ 4.0) and the Budapest Open Access Initiative (http://www.budapestopenaccessinitiative.org/rea d), which permit unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.