Exploring specific key residues binding with LXRβ
In order to analyze the differences underlying the binding mode of agonists between LXRα and LXRβ, 22 known selective or nonselective agonists were docked to LXR and the results were shown as Table 1. The binding affinity could be calculated by docking score, and we proposed that the selectivity of agonists between LXR could be reflected by ΔScore, the docking score of agonists against LXRβ minus that of LXRα. The lower value of ΔScore, the higher selectively against LXRβ. As can be seen in Table 1, BDBM27173, BDBM27174 and BDBM50300572[54] showed favorable binding affinity against LXRβ and poor binding affinity against LXRα, and ΔScore of them were − 8.0 kcal·mol− 1, -6.1 kcal·mol− 1 and − 3.8 kcal·mol− 1 respectively. These compounds are analogs of GW3965[55], a tertiary-amine LXR agonist. GW3965 used to be considered as a nonselective and nonsteroidal agonist for LXR. However, it has been demonstrated that GW3965 was a potent full agonist that possessed LXRβ specificity (EC50 values are 176 nM and 15 nM respectively)[56]. Therefore, GW3965 was classified as a selective agonist. RGX-104[57], a small molecule LXR agonist, obtained − 5.6 kcal·mol− 1 of ΔScore. WYE672 as a LXR agonist showed potent binding affinity to LXRβ (IC50 = 53 nM). It had little binding affinity for LXRα (IC50 > 1.0 µM)[58] and its ΔScore was − 3.6 kcal·mol− 1. CHEMBL4215604 is a 2-Thienyl analog exhibited potent LXRβ selective agonistic activity (EC50 = 0.559 µM)[59] and its ΔScore was − 3.1 kcal·mol− 1. N-acylthiadiazolines[60] is a selective LXRβ agonist with poor pharmacokinetics property and its ΔScore calculated here was − 2.8 kcal·mol− 1. 4-(3-biaryl) quinoline sulfones (BDBM50317731) is a high affinity LXRβ agonist with modest binding selectivity over LXRα and its ΔScore calculated here was − 2.3 kcal·mol− 1[10].
Nonselective agonist, such as AZ876[61], LXR623[62] and T0901317[63] attained ΔScore higher than − 2 kcal·mol− 1. It has been indicated that the experimental binding free energy ΔGexp approach to -RT lnIC50, thus for every difference in binding free energy of 1 kcal·mol− 1 the difference in activity is about 5.9 times. Thereby we suggested that values of -2 kcal·mol− 1 was the threshold of whether a compound possesses selectivity. An agonist would possess selectivity against LXRβ if ΔScore lower than − 2 kcal·mol− 1 and a compound with ΔScore over − 2 kcal·mol− 1 is more likely to be a nonselective agonist. The physiological ligands for LXR are oxysterols, including 24(S)-hydroxyl cholesterol (cpd.9) and 24(S),25-epoxycholesterol (cpd.8), they did not show observed selectivity between LXRα and LXRβ compared with other synthetic ligands. Meanwhile the affinity against LXR of physiological ligands calculated here were also more unfavorable than synthetic ligands. However, despite many efforts have been made to discover effective and selective LXRβ agonists, there is rarely approved drug has been reported.
Table 1. Docking score (kcal·mol-1) of 22 LXR agonists
Name※
|
LXRα
|
LXRβ
|
ΔScore
|
Descriptor
|
BDBM27173a
|
-7.2
|
-15.2
|
-8.0
|
Selective
|
BDBM27174 a
|
-9.3
|
-15.4
|
-6.1
|
Selective
|
RGX-104
|
-9.7
|
-15.2
|
-5.6
|
Selective
|
BDBM50300572 a
|
-11.4
|
-15.2
|
-3.8
|
Selective
|
WYE672
|
-6.7
|
-10.3
|
-3.6
|
Selective
|
CHEMBL4215604b
|
-7.9
|
-11.0
|
-3.1
|
Selective
|
N-Acylthiadiazolines
|
-9.1
|
-11.9
|
-2.8
|
Selective
|
BDBM50317731a
|
-7.8
|
-10.2
|
-2.3
|
Selective
|
GW3965
|
-9.7
|
-14.4
|
-4.8
|
Selective
|
LXR623
|
-10.8
|
-11.6
|
-0.7
|
Nonselective
|
WAY-254011
|
-9.8
|
-11.8
|
-2.0
|
Nonselective
|
Quinoline 16
|
-10.4
|
-11.8
|
-1.4
|
Nonselective
|
AZ876
|
-9.3
|
-10.6
|
-1.3
|
Nonselective
|
BDBM35089a
|
-11.6
|
-11.9
|
-0.2
|
Nonselective
|
cpd.8
|
-10.1
|
-10.1
|
0.0
|
Nonselective
|
T0901317
|
-10.7
|
-10.7
|
0.0
|
Nonselective
|
SCHEMBL14056665c
|
-9.4
|
-9.3
|
0.1
|
Nonselective
|
cpd.9
|
-8.6
|
-8.0
|
0.6
|
Nonselective
|
SCHEMBL1272009 c
|
-11.3
|
-10.6
|
0.7
|
Nonselective
|
SCHEMBL12847006 c
|
-11.9
|
-10.7
|
1.2
|
Nonselective
|
SCHEMBL3419782 c
|
-10.2
|
-8.4
|
1.7
|
Nonselective
|
CHEMBL427914 b
|
-8.8
|
-7.5
|
1.3
|
Nonselective
|
※ Compound with a unique database ID, which takes the form of a ‘BDBM’, ‘CHEMBL’ or ‘SCHEMBL’ prefix immediately by an integer. a The Binding Database. b CHEMBL database. c SureChEMBL database.
Furthermore, we investigated the differences of binding mode of selective and nonselective agonists targeting LXR using 2D and 3D ligand-receptor interaction diagram. Among 22 known agonists collected from literature, 9 agonists are selective agonists of LXRβ and the rests are nonselective agonists. As shown in Table 2, 29 key residues from LBD participated in ligand binding. Residues Leu274, Thr316 and Phe329 participated in LXRβ binding for all selective and nonselective agonists, indicating these residues are very important for agonist binding. Residues Phe271, Ser278, Met312 and His435 were involved in selective agonists binding, and they all participated in 9 selective agonists binding, but partially occurred in binding of nonselective agonists. It has been indicated that the hydrophobic part of agonist should be located in hydrophobic pocket constituted by Phe271, Phe340 and Phe349 to simulate the activity of LXR[64]. Here, we suggest that Phe271, Ser278, Met312 and His435 might be very important for selective agonist binding with LXRβ.
Table 2. Frequency of occurences of key residues in LBD of LXRβ binding with agonists
Residues
|
Frequency (%)
|
|
Residues
|
Frequency (%)
|
Selective
|
Nonselective※
|
|
Selective
|
Nonselective※
|
Leu274
|
100.00
|
100.00
|
|
Ile353
|
55.56
|
71.43
|
Thr316
|
100.00
|
100.00
|
|
Thr272
|
55.56
|
42.86
|
Phe329
|
100.00
|
100.00
|
|
Leu313
|
55.56
|
14.28
|
Phe271
|
100.00
|
71.43
|
|
Phe268
|
55.56
|
0.00
|
Ser278
|
100.00
|
71.43
|
|
Glu315
|
55.56
|
0.00
|
His435
|
100.00
|
71.43
|
|
Phe349
|
44.44
|
85.71
|
Met312
|
100.00
|
42.86
|
|
Asn239
|
44.44
|
0.00
|
Ala275
|
88.89
|
57.14
|
|
Ile309
|
44.44
|
0.00
|
Leu330
|
88.89
|
42.86
|
|
Ile350
|
44.44
|
0.00
|
Phe340
|
77.78
|
100.00
|
|
Phe354
|
33.33
|
0.00
|
Glu281
|
77.78
|
42.86
|
|
Leu449
|
22.22
|
85.71
|
Arg319
|
77.78
|
42.86
|
|
Gln438
|
11.11
|
28.57
|
Leu345
|
66.67
|
42.86
|
|
Ser242
|
11.11
|
0.00
|
Trp457
|
66.67
|
28.57
|
|
Met212
|
0.00
|
14.28
|
Ile277
|
66.67
|
14.29
|
|
|
|
|
※ The ΔScore of collected nonselective agonists were all over -2 kcal·mol-1 but not over 0 kcal·mol-1 for results of LXRβ binding patterns with agonists that possess LXRα specificity would not be considered in this study.
Binding pattern between agonists and LXRβ was predicted by Ligplot+, the results of eight representative ligands were illustrated as Figure 4 and Figure 5. First of all, as mentioned above, Leu274, Thr316 and Phe329 which were considered as important residues for agonists, were found in all agonists. Glu315, a residue located on the helix region centered on Arg319, was the specific residue binding with selective agonists. It was worth noting that only selective agonist
s could bind with Trp457 while Trp457 was not a key residue in nonselective agonists-LXRβ binding. Trp457 located on helix 12 which was agonist conformation found in some developed LXRβ agonists. Trp457 had been also proved to form a close contact to His435 to help it form polar interaction with ligands easier[52].
Therefore, a standard that we verify whether compounds possesses LXRβ selectivity were constructed. Firstly, Trp457 was selected to identify a compound whether possessed selectivity due to its unique properties. Then, His435 was indispensable for it was the bridge of interaction between Trp457 and ligands. In addition, Phe271, Ser278, Met312 were involved in all selective agonists binding. Taken together, a compound which targets LXRβ selectively should bind to all these residues in the protein: Phe271, Ser278, Met312, His435, Trp457.
Identifying of selective agonists targeting LXRβ
In order to evaluate pharmacokinetics properties and binding affinity of agonists from 112 terpenes of Alismatis Rhizoma, ADMET analysis and molecular docking were adopted to screen compounds. Binding free energy and MD simulations were introduced to further eliminate ligands whose binding affinity are unfavorable.
Pharmacokinetics and physicochemical analysis of collected terpenes
In order to understand physiochemical and pharmacokinetics properties of 112 collected terpenes, the ADMET (absorption, distribution, metabolism, excretion and toxicity) properties were predicted. RO5 which signifies the number beyond Lipinski’s rule of five regions were calculated based on the result. Values of RO5 of all compounds were less than 3 and that of 97 terpenes were located on the recommend range. 52 compounds showed RO5 were 0, which means the compounds possessed potential druggability. Log S, log D and log Kp stand for aqueous solubility, distribution coefficient D at PH=7.4 and skin permeability, respectively. All of them are important properties to evaluate the physicochemical properties of drugs. Caco-2 permeability was represented by log Papp. In our results, 97 of 112 terpenes showed that values of log Papp were located on the recommend range. HIA means human intestinal absorption. Compounds whose HIA more than or equal 30% were marked as 1 while HIA less than 30% were marked as 0. 69 of 112 terpenes showed that values of HIA were 1, which means that the body may have a high absorption for these compounds. BBB stands for brain blood barrier and too polar compounds will not cross through brain blood barrier [65]. 0 means that a compound can’t pass through blood brain barrier while 1 represents a compound can go through blood brain barrier. Results showed that none of compounds can pass through the barrier. However, LXRβ was mainly distributed in livers, and therefore BBB has little effect on the treatment effect of the compounds we screened. DILI means drug induced liver injury, and 0 represents that the drug will cause little damage to liver function. Compounds we screened must make sure values of DILI were 0 for the drugs will mainly affect in the liver. CYP1A2, CYP2C19, CYP2C9, CYP2D6 and CYP3A4 inhibitor were the probability of a molecule being the inhibitor of Cytochrome P450 families. The ADMET properties of the compounds involved in this study are presented in the Table 3. The cpd.1 to 8 indicate 16-hydroxy-alisol B 23-acetate, 16-oxo-alisol A 23-acetate, 25-O-ethylaliso A, Alismanol C, Alisol G, Alisol M 23-acetate, Alisol N 23-acetate, 24(S),25-epoxy cholesterol, respectively.
Table 3. The ADMET properties of compounds involved in the study.
ADMET properties
|
Molecules
|
Recommend range※
|
cpd.1
|
cpd.2
|
cpd.3
|
cpd.4
|
cpd.5
|
cpd.6
|
cpd.7
|
cpd.8
|
RO5
|
1
|
1
|
1
|
1
|
0
|
1
|
1
|
1
|
≤1
|
Molecular weight
|
528.8
|
546.7
|
518.8
|
530.7
|
472.7
|
542.8
|
528.8
|
400.6
|
150–500
|
Rotatable bonds
|
6
|
7
|
7
|
7
|
5
|
6
|
6
|
4
|
≤5
|
H-bond acceptors
|
5
|
7
|
5
|
6
|
4
|
6
|
5
|
2
|
0–10
|
H-bond donors
|
2
|
3
|
3
|
2
|
3
|
2
|
2
|
1
|
0–5
|
log P a
|
5.37
|
3.94
|
5.04
|
4.8
|
4.97
|
4.68
|
5.42
|
5.86
|
-2–5
|
log S b
|
-5.99
|
-4.97
|
-5.58
|
-5.6
|
-5.68
|
-5.61
|
-6.11
|
-6.22
|
-6.5–0.5
|
log D c
|
2.90
|
3.75
|
4.49
|
3.84
|
4.29
|
2.83
|
2.95
|
4.52
|
1–5
|
log Papp d
|
-5.07
|
-5.04
|
-4.97
|
-5.01
|
-5.03
|
-5.03
|
-5.06
|
-4.79
|
> -5.15
|
Pgp substrate e
|
0
|
1
|
0
|
1
|
0
|
1
|
0
|
0
|
0
|
HIA f
|
1
|
0
|
1
|
1
|
1
|
1
|
1
|
1
|
1
|
log Kp g
|
-5.84
|
-7.16
|
-6.1
|
-6.24
|
-5.53
|
-6.45
|
-5.71
|
-4.05
|
-8.0–1.0
|
BBB h
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
PPB i
|
86.52
|
90.02
|
84.62
|
92.44
|
90.01
|
86.77
|
86.31
|
81.12
|
90
|
CYP1A2 inhibitor j
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
CYP2C19 inhibitor j
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
CYP2C9 inhibitor j
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
CYP2D6 inhibitor j
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
CYP3A4 inhibitor j
|
0
|
1
|
1
|
1
|
1
|
0
|
0
|
0
|
0
|
hERG k
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
1
|
1
|
DILI l
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
0
|
※ The Recommend range of properties referred from ADMETlab[31] and SwissADME[30]. a average of octanol/water partition coefficient by different prediction, which consists of iLOGP[66], XLOGP3[67], WLOGP[68], MLOGP and SILICOS-IT. b aqueous solubility, which was implemented from ESOL[69]. c Distribution Coefficient D at PH=7.4. d Caco-2 Permeability. e 0: Non-substrate; 1: Substrate. f Human Intestinal Absorption. g skin permeability, unit is cm/s. h brain-blood barrier. i Plasma Protein Binding, significant with drugs that are highly protein-bound and have a low therapeutic index. j 0: Non-inhibitor; 1: Inhibitor. k hERG (human ether-a-go-go related gene) Blockers, 0: Non-blockers; 1: Blockers. l Drug Induced Liver Injury, 0: DILI negative; 1: DILI positive.
In order to verify reliability of docking models, ROC curves were used to evaluate performance of docking functions that distinguish active compounds and inactive compounds from a given dataset. A set of decoy molecules for ligands which were obtained from the ChEMBL database were constructed while they were input to dock with LXRβ with different score functions. Then, the ROC curves of each function were plotted in Figure 6A and the AUC values of Ledock, Grid, Amber, Vina are 0.710, 0.729, 0.786, 0.887 respectively. The range of AUC values is from 0 to 1. In addition, 0.5 is deemed as the threshold of whether a function has ability to distinguish[70]. AUC<0.5 suggests that the test results are worse than random results while AUC>0.5 shows that test is effective[71]. Based on this, the study chose AUC value of 0.5 as reference line, which was named Random in Figure 6A. To compare the AUC values of different functions with reference line, all of them demonstrated acceptable discrimination and the Vina score function was the most outstanding of them. Then, redock with LXRβ was executed by Autodock Vina for the sake of identifying the accuracy of Vina function in the most straightforward way[72]. RMSD is thought to be an essential standard of evaluating how much structure is different from the original structure[73]. Furthermore, if RMSD value between pre- and post-docking structure is below 2 Å, the docking model is considered to be great and can be applied for docking[74]. Herein, the RMSD value between redock structure and its initial structure of LXRβ using Vina function in the study was 1.9 Å, which met the requirement mentioned above. The result of redock and its initial structure is shown as Figure 6B and the binding pocket of LXRβ used in the study is displayed in Figure 6C.
Screening potential LXRβ ligands using molecular docking
The Vina scores were applied to compute binding affinity between ligands and LXR using Autodock Vina program. In order to screen ligands targeting LXRβ selectively, ΔScore was calculated which means that values of Vina score of LXRβ minus that of LXRα. As mentioned above, we thought that compounds whose ΔScore values below -2 kcal·mol-1 possessed preferential binding affinity with LXRβ. Seven screened compound candidates which met the requirements were shown as Table 4. cpd.8 and cpd.9 which were thought to activate LXR endogenously in liver and brain respectively[75] were also presented as reference ligands in Table 4. Compared with cpd.8, all screened compound candidates showed higher binding affinities of LXRβ and most of them except cpd.4 attained lower binding affinities of LXRα, suggesting they are more likely to bind with LXRβ. Moreover, values of HIA of most the candidates except cpd.2 were 1, indicating most candidates possessed potential to enter into internal environment and activate target protein. Among all candidates, cpd.5 was the most potential ligand for its performance in pharmacokinetics analysis and great gap in docking affinities with cpd.8. cpd.1 and cpd.4 were also considerable compounds for LXRβ agonists. The former attained high value of ΔScore, which means that it highly selectively targets LXRβ. The latter possessed high docking score of LXRβ. Most of all, values of DILI of all screened compound candidates were 0, which means they all possessed essential properties of drugs targeting LXRβ that is low toxicity. The discussion about cpd.9 will not proceed because it was a ligand found in brain.
20ns MD simulations for LXR-ligand were performed via Gromacs program in order to evaluate dynamic interactions between LXRβ and ligand candidates. The calculated results of MD simulations were output in the form of MD trajectory. Then, assessment of backbone RMSD values for each MD trajectory was executed to provide a holistic sight to observe stability of LXR-ligand system. As illustrated in Figure 7, the backbone values of LXRα-ligand system were between 0.10 nm and 0.30 nm while that of LXRβ-ligand system were from 0.075 nm to 0.275 nm. The reference ligand, cpd.8, rapidly reach platform at 2 ns in both LXRα-ligand system and LXRβ-ligand system. cpd.1 and cpd.6 showed drastic fluctuation of around 0.20 nm and 0.225 nm in LXRα-ligand system and LXRα-cpd.7 system appeared stable and minimum equilibrium at 0.18nm. LXRβ-cpd.6 system fluctuated steadily while LXRβ-cpd.1 system showed fluctuation as drastic as LXRα-ligand system. LXRβ-cpd.2 system was the most stable LXRβ-ligand system, which fluctuated at 0.125 nm. Moreover, fluctuations of LXRα-cpd.2 system were on mean level among all LXRα-ligand systems. LXRβ-cpd.4 was the most fluctuant LXRβ-ligand system, which kept fluctuating during the 20 ns simulation experiment. However, LXRα-cpd.4 system showed stable equilibrium. Cpd.3 and cpd.5 presented increasingly large unstable fluctuation of RMSD values in LXRβ-ligand system whereas the level of RMSD was moderately low in LXRα-ligand system. In order to screen ligands selectively targeting LXRβ, compounds which fluctuated steadily in LXRβ-ligand system but waved drastically in LXRα-ligand system were selected. It demonstrated that cpd.6 was the molecule that most fits the requirement, and cpd.1 and cpd.2 can also be candidates.
Calculation of MM/PBSA binding free energy
Because Vina score only compute the binding free energy of lowest-scoring conformation and exist empirical component, MM/PBSA binding energy calculation was introduced to the work in order to assess binding affinity accurately between ligands and proteins[36,49]. The results of calculation of LXRα and LXRβ were listed as Table 5 and Table 6, respectively. Here, ΔGbind means comprehensive evaluation of binding free energy while ΔGvdw, ΔGele, ΔGpolar and ΔGapolar are key elements of binding free energy. The lower binding free energy suggests the binding mode between ligand and receptor is more stable. As shown in Table 5, cpd.2 and reference ligand cpd.8 presented similar level in ΔGbind, which were -45.84 kcal·mol-1 and -45.46 kcal·mol-1, respectively. ΔGbind of cpd.4 was -40.67 kcal·mol-1, which means that interactions between cpd.4 and LXRα are the most unstable. cpd.7 was the most stable ligand that binds to LXRα. The rests were at the same and moderately stable level. As listed in Table 6, the reference compound, cpd.8, showed considerably high ΔGbind of -30.29 kcal·mol-1, which indicated that ΔGbind was detrimental to binding to LXRβ. ΔGbind of cpd.6 was -56.17 kcal·mol-1, which was the lowest value among all screened compound candidates, indicating cpd.6 possessed the highest binding affinity to LXRβ. The binding affinity with LXRβ of cpd.1 and cpd.7 were at similar and low level. ΔGbind of cpd.4 and cpd.5 is the highest among all screened compound candidates. ΔGbind of cpd.2 and cpd.3 were -50.33 kcal·mol-1 and -48.99 kcal·mol-1 respectively, which were at mean level of all candidates. In order to screen compounds selectively targeting LXRβ, compounds which ΔGbind of LXRβ was higher than that of LXRα were eliminated. The result of the filter is that the cpd.1, cpd.2, cpd.4, cpd.6 are remained.
Table 5. Summary of binding free energy components for the LXRα complexes(kcal·mol-1)
Molecules
|
ΔGvdw
|
ΔGele
|
ΔGpolar
|
ΔGapolar
|
ΔGbind
|
cpd.1
|
-71.82±3.32
|
-9.40±1.50
|
37.32±1.65
|
-6.35±0.20
|
-50.26±2.96
|
cpd.2
|
-71.25±2.47
|
-10.60±1.58
|
42.45±2.41
|
-6.44±0.19
|
-45.84±2.73
|
cpd.3
|
-68.91±2.30
|
-10.13±1.83
|
35.01±2.71
|
-6.86±0.19
|
-50.89±2.87
|
cpd.4
|
-61.92±2.80
|
-0.58±1.64
|
27.93±1.64
|
-6.11±0.19
|
-40.67±3.64
|
cpd.5
|
-63.90±2.70
|
-11.00±2.15
|
31.24±1.16
|
-6.10±0.22
|
-49.75±2.22
|
cpd.6
|
-71.78±3.53
|
-6.81±1.40
|
34.54±2.93
|
-6.73±0.19
|
-50.78±3.32
|
cpd.7
|
-72.48±3.23
|
-2.68±1.18
|
30.11±1.94
|
-6.69±0.13
|
-51.73±3.28
|
cpd.8
|
-62.86±2.83
|
-4.98±1.21
|
28.02±1.96
|
-5.66±0.14
|
-45.46±2.35
|
Table 6. Summary of binding free energy components for the LXRβ complexes(kcal·mol-1)
Molecules
|
ΔGvdw
|
ΔGele
|
ΔGpolar
|
ΔGapolar
|
ΔGbind
|
cpd.1
|
-72.30±3.02
|
-11.73±2.03
|
38.88±1.66
|
-6.62±0.20
|
-51.77±3.06
|
cpd.2
|
-72.07±2.85
|
-17.66±2.79
|
45.96±3.78
|
-6.56±0.18
|
-50.33±3.81
|
cpd.3
|
-74.17±2.73
|
-6.28±1.14
|
38.15±1.89
|
-6.69±0.20
|
-48.99±2.90
|
cpd.4
|
-67.67±3.20
|
-5.52±1.47
|
34.42±1.98
|
-6.54±0.22
|
-45.31±2.32
|
cpd.5
|
-62.79±3.25
|
-6.92±1.49
|
31.25±3.07
|
-6.00±0.17
|
-44.46±3.34
|
cpd.6
|
-72.95±3.04
|
-8.05±1.51
|
31.13±1.71
|
-6.30±0.24
|
-56.17±2.81
|
cpd.7
|
-71.28±2.72
|
-9.83±2.02
|
36.65±1.97
|
-6.64±0.17
|
-51.10±3.17
|
cpd.8
|
-58.01±2.75
|
-9.63±1.56
|
43.03±2.53
|
-5.68±0.14
|
-30.29±2.85
|
Verifying of the LXRβ selectivity of screened compounds
cpd.1, cpd.2, cpd.6 were selected as screened compounds for further analysis. cpd.4 was discarded for it fluctuated drastically in LXRβ-cpd.4 system. The RMSF values of MD simulations, decomposition of binding free energy and binding model analysis were used to explore the binding mode of screened compounds and LXRβ. Then, we applied the validation standard to verify whether screened molecules possessed selectivity.
RMSF analysis of the key amino acid residues
RMSF values for each MD trajectory were calculated in order to assess flexibility of proteins while higher values indicated better flexibility. Figure 8 are the results of RMSF values of LXR-ligand system. Low RMSF values mean that the system is stable, and therefore residues of this region may participate in the binding interactions of LXR-ligand system. It is clear that all compounds shared similar RMSF profiles in both LXRα-ligand system and LXRβ-ligand system. Compared with LXRβ-ligand system, LXRα-ligand system showed relatively higher RMSF values of average 0.1328 nm while that of LXRβ-ligand system was at mean level of 0.1145 nm, which suggested screened compounds are more likely to bind to LXRβ on the whole. In LXRβ-ligand system, region Ser254-Leu264 was the most flexible, indicating the residues contribute less to binding of the system. RMSF values of region Ala306- Arg319 and Gly369-Ile377 were lowest, suggesting the region are more responsible for binding of the system. Interestingly, the unique mutational residue Asn377 in LXRβ was located on the region, which means that Asn377 contribute a lot to the binding of LXRβ and it is a candidate screening target for LXRβ. Region Ala306- Arg319 was all situated on helix region except Ser307 and residues of protein-binding pocket Met312, Leu313, Glu315, Thr316, Arg319 also lay on the region. Met312 which has been proved to bind to LXRβ such as T-0901317[52] is a part of validation standard for selective targeting LXRβ binding residues. The RMSF values of all binding system on Met312 were at a low level, which indicates that binding between screened compounds and Met312 was stable. Arg319 was a vital residue which participated in forming hydrogen bonds with ligands. Thr316 is thought to be an active site of forming hydrogen bonds with LXRβ[76]. cpd.1, cpd.2 and cpd.6 all possessed ability to bind with Thr316 on LXRβ in the form of hydrogen bonds.
Decomposition of MM/PBSA binding free energy
Binding free energy decomposition was applied to identify contributions of each residue in the LXR-ligand binding model. The contributions of each residue of LXRα and LXRβ are depicted in Figure 9 and Figure 10, respectively. Low values of energy indicate an important role of residues in the LXR-ligand binding. As demonstrated in Figure 9, most residues showed values which were negligible to the system. Only LXRα-cpd.8 system had a moderately high positive value located on Arg305, which means that the residue contributes less to binding to LXRα. Phe257, Leu260, Ala261, Met298, Leu331, Trp443 (values of energy of selected residues are below -0.8 kcal·mol-1) showed relatively larger contributions of binding to LXRα. As illustrated in Figure 10, most residues had little effect on LXRβ-ligand binding. The reference, LXRβ-cpd.8 system, showed two extremely high values of binding free energy of 8.21 kcal·mol-1 and 3.71 kcal·mol-1, which were located on Glu281 and Glu315, respectively. Energy of Arg319 was high in LXRβ-cpd.1 system while the Phe271, Phe329 and Trp457 showed the lowest values of binding free energy. Phe271 and Phe329 have been proved an important contributor to hydrophobic interactions between LXRβ and ligand, and Trp457 can form a close contact to His435 which is able to form polar interaction with ligand[52]. All of four residues are within validation standard and indicated cpd.1 possessed LXRβ selectivity. Arg319 is an inhibitive site which can form hydrogen bonds with ligands. Compared to the residues above, smaller change in ΔGMM is direct cause of the higher energy value of Arg319. ΔGbind of Phe271, Phe329 and Trp457 were lower than the rest residues because of lower ΔGMM while ΔGpolar were even higher than the rest residues. ΔGMM indicates potential energy in the vacuum which are composed of ΔGvdw and ΔGele. It demonstrated that potential energy was a crucial factor that affects binding between cpd.1 and key residues in LXRβ. Energy of Glu281 and Phe329 were the maximum of 2.22 kcal·mol-1 and minimum of -2.80 kcal·mol-1 of LXRβ-cpd.2 system, respectively. Glu281, whose energy was highest in both LXRβ-cpd.2 and LXRβ-cpd.8 system, showed an extraordinary ΔGpolar of 2.48 kcal·mol-1 and 8.80 kcal·mol-1, respectively. Glu281 was considered as a hydrophilic site which is able to bind to ligand in the form of hydrogen bonds[77]. Hydrogen bonds are usually considered as a set of dispersion force and intermolecular or interatomic electrostatic interaction[78], whose energy is represented by ΔGMM. However, ΔGMM of the two compounds were not low and therefore ΔGbind were at a high level, suggesting the two compounds were less likely to form hydrogen bonds with LXRβ at Glu281. Nevertheless, the result was opposite for Phe329. Although ΔGpolar was at moderately high values at 2.17 kcal·mol-1 in LXRβ-cpd.2 system, ΔGMM reached an extremely low value of -4.79 kcal·mol-1, which counteracted the negative effect of ΔGpolar on binding. The ΔGapolar of Phe329 in LXRβ-cpd.2 system was also low. ΔGapolar was calculated from solvent-accessible surface area that can be affected by hydrophobic effect. It indicates that cpd.2 possesses ability to participate in hydrophobic effect of LXRβ. In LXRβ-cpd.6 system, there was no high positive value. Phe271, Leu274 and Trp457 provided key contributions of the lowest value of energy. Leu274 was considered as a hydrophobic site in the LXRβ[59] and showed comparatively higher ΔGapolar. The other hydrophobic residue, Phe271, showed high values of ΔGpolar and low ΔGMM. ΔGMM value of Trp457 were at low level of -2.57 kcal·mol-1, suggesting it has huge potential to form hydrogen bonds with cpd.6 while ΔGpolar and ΔGapolar were at moderately low level, which means Trp457 would play an important role in LXRβ-cpd.6 binding.
Binding model analysis of screened compounds
In order to investigate the differences of 2D and 3D LXR-ligands binding mode of screened compounds, Ligplot+[50] was introduced to identify key residues which interact with ligands in LXRβ. Key residues of all screened compound candidates were predicted and they were shown as Supplementary Table 2. The results showed that most of them attained plentiful binding residues at helix3 and helix12, indicating they possessed high binding affinity with LXRβ. Results of screened compounds and reference ligand was plotted in Figure 11. Circled residues were equivalent side chains, which were considered as key residues in binding free energy decomposition. The green line represented hydrogen bonds between ligands and residues and residues who formed hydrogen bonds were highlighted particularly. As depicted in Figure 11, Phe271, Thr316, Ser278 and His435 formed hydrogen bonds with screened compounds. These residues were also considered very important for selective agonist binding with LXRβ in our study. Interestingly, Thr316, a residue found in both selective and nonselective agonists, presented in all screened compounds, indicating all the compounds may be potential agonists against LXRβ. His435, which was proved an important residue that form hydrogen bonds to stabilize the protein binding pocket[52], could be found in all screened ligands. However, reference ligand can’t bind with His435. Another residue, Trp457, which was located on helix 12 and could form contact with His435[52], had the same case. Arg319 and Glu315 were located in protein binding pocket and Arg319 was an inhibitive site which can form hydrogen bonds. However, results of energy decomposition demonstrated that binding affinity between
ligands and Arg319 on LXRβ was not high. And binding affinity between ligands and LXRβ of Glu315 was average.
The interactions between LXRβ and ligands in 3D were illustrated as Figure 12 to show the hydrogen bonds intuitively. More hydrogen bonds and shorter length of hydrogen bonds indicate larger interaction intensity between ligands and receptor. It is clear that intensity of interaction between LXRβ and screened compounds was higher than that of reference ligand, which was reflected on the number and length of hydrogen bonds between LXRβ and ligands. All screened compounds could bind to Thr316 and the distance between ligands and the residues was the nearest among all residues in Figure 12, indicating ligands possessed high binding affinity with Thr316. Phe271 was the residue which binds closest with cpd.1, suggesting the residue was important in LXRβ-cpd.1 binding. Both His435 and Ser278 which were located on protein binding pocket had an average binding affinity, which means that all screened ligands could bind with the two residues. However, reference ligand, which was a kind of nonselective agonists in our body, didn’t find the characters described above.