Formation of endo-[P4-(OH)BPO]. Our group has previously developed unique methods to grow functional substituents on the rims of pillar[n]arenes23. We conjectured that a pillar[n]arene-derived protein binding pocket mimetic could be realized through mounting a polar functionality on the rim of a pillar[n]arene, and then forcing it to point towards the tubular core either through steric effect or substrate-induced conformational change. We therefore started our work from adding a small polar functionality, hydroxyl group, to the rim of pillar[5]arene by creating a quaternary carbon atom through a nucleophilic addition reaction of pillar[4]arene[1]quinone (P4Q) with phenylmagnesium chloride (or phenyllithium) (Supplementary Information). The resultant product, pillar[4]arene[1]1-hydroxy-[1,1'-biphenyl]-4(1H)-one ([P4-(OH)BPO]), was characterized by 1H and 13C NMR spectroscopy, atmospheric pressure chemical ionization mass spectroscopy (APCI-MS), and single crystal X-ray diffraction analysis (Supplementary Information). The X-ray crystallography data of a single crystal obtained by slow evaporation of a solution of [P4-BPO(1-OH)] in acetone (Fig. 2a) clearly showed that ([P4-(OH)BPO]) has its phenyl group positioned outside the tubular frame, and its hydroxyl group pointing towards the core of the cavity, with an acetone molecule encapsulated inside the cavity through formation of a H-bond between the C=O group of acetone and the O–H group of [P4-(OH)BPO], which is a unambiguous evidence of the adoption of an endo conformational isomer, endo-[P4-(OH)BPO].
endo-[P4-(OH)BPO] as protein binding pocket mimetic. FBDD is an approach to develop potent small-molecule compounds starting from fragments binding weakly to target proteins9,10. Owing to its advantages such as saving experimental cost, offering diverse hits, and exhibiting multiple ways of binding, FBDD has been playing important roles in target-based drug discovery. In FBDD, fragment ligands are usually small molecules with low complexity in chemical structure, and H-bonds play an important role in stabilizing the fragment-pocket complexes9-10. For example, among the 462 unique fragment-pocket complexes investigated by Giordanetto, Shaw and co-workers, 92% of the fragments have at least one hydrogen bond formed with a protein, a structural water molecule, or a metal atom10. Given the role of H-bonds in fragment-pocket complexes, endo-[P4-(OH)BPO], with an inwardly pointing H-bond donor (-OH) at the bottom of its deep cavity, a strong H-bond acceptor (C=O) on the tubular inner wall of a predominantly hydrophobic surface, is naturally an appealing artificial protein binding pocket.
Binding of fragments to endo-[P4-(OH)BPO]. Recently, several biomimetic pockets with polar binding sites in their hydrophobic pockets have been disclosed11-18. Impressively, thermodynamics of interactions between biomimetic pockets and guests in water has been studied systematically by Jiang and co-workers16-18, which is highly valuable as the recognizations of ligands by protein binding pockets occur in aqueous environment in living systems. However, it is challenging to assess the contribution to pocket-ligand binding affinity “solely” from hydrogen bonding only through studies on the complexation in water. The reasons include: a) water molecules intend to bind the inner polar site of a synthetic receptor, and b) water molecules compete with a host in binding with a polar guest, weakening H-bond interaction between the host and guest16-18. In order to better understand both big picture and details of molecular recognition, investigation of pocket-fragment complexation in non-polar environment should be indispensably complimentary to the studies in aqueous solutions. Therefore, we initiated a thermodynamic study on the fragment-pocket binding abilities and selectivities of endo-[P4-(OH)BPO] with respect to various guest compounds in a non-polar solvent, CDCl3. In selecting guests for the study, ideal guest compounds resembling “fragment ligands” in FBDD are those not only geometrically complementary to the tubular binding pocket, but, more importantly, bearing H-bond acceptor(s) or donor(s) or both so that such a host-guest interaction could mimic a protein-ligand interaction event staged synergistically by both hydrogen bonding and hydrophobic interaction9,10. We selected an array of small-molecule fragment compounds for a “fragment library” which includes alkyl amines, alcohols, aldehydes, acids, esters, small heterocycles, and simple small molecules possessing bilogically important functional groups (Fig. 3). 1H NMR experiments were performed on all of the host-guest pairs at 1∶1 ratio in CDCl3 (Supplementary Information). The shifts of the proton signals of the complexed guests relative to the free species were used to evaluate the binding events. The binding constants (Ka) and binding free energies were determined by the 1H NMR titration method (Table 1)24. Generally, the energy of a single H-bond is somewhat between a van der Waals interaction and a fully covalent or ionic bond, but it varies depending on the nature of the donor and acceptor atoms, their geometries, and surrounding environments25,26. Thus, the “fragment-pocket” interaction of [P4-(OH)BPO] with a guest was expected to be affected by the number of H-bonds, angles of the hydrogen bond(s), nature of the donor and acceptor atoms, and the hydrophobic interaction between the inner surface of endo-[P4-(OH)BPO] and the guest25,26. In order to explore the correlations of the binding strength (Ka) and properties of guest fragments, molecular descriptors including the logarithm of the octanol:water partition coefficient (log P), dipole moment (μ), molecular volume (V), surface area (S), and asphericity Ωa) were collected16,27.
Table 1. Fragment library compounds screened in fragment-pocket complexation |
compound
|
Ka (M-1)a
|
R2
|
- ΔGb
|
V (Å3)c
|
Ωad
|
S (Å2)e
|
log Pf
|
μ(D)g
|
1
|
138.0±20.7
|
0.9957
|
11.8-12.6
|
129.80
|
0.18040
|
141.48
|
0.86
|
1.159
|
2
|
92.7±11.0
|
0.9942
|
10.9-11.5
|
177.01
|
0.21016
|
183.71
|
2.06
|
1.143
|
3
|
42.0±6.3
|
0.9880
|
8.9-9.6
|
224.21
|
0.22452
|
225.94
|
2.90
|
1.139
|
4
|
67.4±13.1
|
0.9977
|
9.9-10.9
|
123.07
|
0.18229
|
136.00
|
0.84
|
2.057
|
5
|
24.0±2.1
|
0.9936
|
7.7-8.1
|
170.27
|
0.21121
|
178.24
|
2.03
|
2.038
|
6
|
28.0±2.6
|
0.9941
|
8.0-8.5
|
217.47
|
0.22514
|
220.47
|
3.07
|
2.020
|
7
|
40.8±13.0
|
0.9957
|
8.2-9.9
|
115.11
|
0.18099
|
129.54
|
0.88
|
3.490
|
8
|
24.1±3.5
|
0.9987
|
7.5-8.2
|
162.30
|
0.20842
|
171.83
|
1.78
|
3.593
|
9
|
10.6±1.6
|
0.9998
|
5.4-6.2
|
209.51
|
0.22311
|
214.08
|
2.95
|
3.640
|
10
|
11.2±1.6
|
0.9989
|
5.6-6.3
|
147.78
|
0.16638
|
159.66
|
1.30
|
5.115
|
11
|
8.8±1.8
|
0.9987
|
4.8-5.9
|
195.01
|
0.20433
|
201.95
|
2.31
|
5.080
|
12
|
1.7±0.1
|
0.9996
|
1.2-1.5
|
242.19
|
0.22186
|
244.18
|
3.33
|
4.805
|
13
|
8.7±0.3
|
0.9992
|
5.3-5.4
|
123.99
|
0.13045
|
137.19
|
0.79
|
5.097
|
14
|
5.2±0.3
|
0.9982
|
3.9-4.2
|
171.15
|
0.19348
|
179.29
|
1.92
|
4.940
|
15
|
3.3±0.2
|
0.9989
|
2.8-3.1
|
218.36
|
0.21638
|
221.53
|
3.05
|
4.949
|
16
|
321.6±17.0
|
0.9999
|
14.2-14.4
|
83.08
|
0.11042
|
100.38
|
-2.06
|
4.853
|
17
|
22.5±2.2
|
0.9992
|
7.5-7.9
|
106.69
|
0.08171
|
120.67
|
-1.01
|
4.685
|
18
|
390.5±75.9
|
0.9957
|
14.3-15.2
|
106.79
|
0.08234
|
120.75
|
-1.55
|
4.961
|
19
|
241.5±22.9
|
0.9993
|
13.4-13.8
|
248.40
|
0.21017
|
247.71
|
1.50
|
5.259
|
20
|
109.4±16.1
|
0.9981
|
11.2-12.0
|
106.44
|
0.09492
|
121.96
|
-0.70
|
4.786
|
21
|
222.9±32.0
|
0.9985
|
13.0-13.7
|
130.15
|
0.13194
|
143.33
|
-0.19
|
4.391
|
22
|
103.9±3.6
|
0.9992
|
11.4-11.6
|
153.76
|
0.14146
|
163.59
|
0.32
|
4.446
|
23
|
58.6±3.2
|
0.9957
|
10.0-10.2
|
177.36
|
0.15542
|
184.78
|
0.83
|
4.412
|
24
|
123.5±13.6
|
0.9961
|
11.7-12.2
|
129.62
|
0.13367
|
142.39
|
-0.19
|
4.559
|
25
|
67.3±2.8
|
0.9955
|
10.3-10.5
|
247.79
|
0.17616
|
247.68
|
2.36
|
4.149
|
26
|
4.7±0.2
|
0.9993
|
3.7-3.9
|
138.11
|
0.05565
|
146.37
|
-0.64
|
4.631
|
27
|
ndh
|
|
|
107.72
|
0.04061
|
119.30
|
0.46
|
2.470
|
28
|
ndh
|
|
|
85.86
|
0.06253
|
99.95
|
0.75
|
3.745
|
29
|
83.0±15.1
|
0.9963
|
10.5-11.4
|
95.76
|
0.06253
|
106.64
|
-0.08
|
1.918
|
30
|
6.4±0.4
|
0.9987
|
4.4-4.8
|
90.78
|
0.06253
|
106.64
|
0.05
|
4.139
|
31
|
11.2±1.6
|
0.9989
|
5.6-6.3
|
101.33
|
0.07198
|
111.92
|
0.44
|
1.517
|
32
|
13.8±2.3
|
0.9989
|
6.1-6.9
|
124.95
|
0.10538
|
133.83
|
0.97
|
1.263
|
33
|
11.2±1.6
|
0.9989
|
5.6-6.3
|
136.16
|
0.12523
|
144.96
|
1.43
|
3.910
|
34
|
11.2±1.6
|
0.9989
|
5.6-6.3
|
185.70
|
0.16779
|
189.26
|
1.43
|
4.908
|
35
|
101.5±14.6
|
0.9983
|
11.1-11.8
|
102.60
|
0.03071
|
116.00
|
-1.35
|
5.552
|
36
|
52.4±9.0
|
0.9972
|
9.3-10.2
|
118.43
|
0.13632
|
132.34
|
0.87
|
4.572
|
37
|
28.6±3.2
|
0.9990
|
8.0-8.6
|
66.25
|
0.21003
|
84.87
|
-0.34
|
4.123
|
38
|
11.2±1.6
|
0.9989
|
5.6-6.3
|
159.13
|
0.18420
|
167.76
|
1.86
|
2.828
|
39
|
750.1±31.5
|
0.9999
|
16.3-16.5
|
163.20
|
0.15823
|
173.05
|
-1.31
|
2.743
|
40
|
1190±89.5
|
0.9999
|
17.4-17.7
|
139.89
|
0.14959
|
151.57
|
-2.52
|
6.032
|
aExperiments were performed in CDCl3 at 298K (Supplementary Information). bΔG = -RTlnKa. cCalculated based on optimized structures. dCalculated by the principal moments of inertia of guests. eCalculated by using Multiwfn program28. fValues were taken from references16,29,30. gCalculated with Gaussian 09 at the RB3LYP/6-31++G level of calcualtions31. hValues were too small to be determined.
Fragment-pocket complexation. We first assessed the binding between [P4-(OH)BPO] and fragments including alkyl amines, alcohols, aldehydes, acids, and esters (Table 1). The formation of fragment-pocket complexes between [P4-(OH)BPO] and the guests was evidenced by obvious upfield shifting of the proton signals of a-carbons of the guests (1-15) in the 1H NMR spectra of the fragment-pocket pairs in a 1∶1 molar ratio in non-polar solvent. As shown in the 1H NMR spectra (Fig. 4), upon mixing n-butyric acid (13) and [P4-BPO(1-OH)] in CDCl3, obvious upfield shifts for the methylene (Δδα = -0.26 ppm and Δδβ = -0.25 ppm) and the terminal methyl (Δδγ = -0.14 ppm) proton signals of 13 were observed. In addition, NOE correlations between protons (Hα) of 13 and the 1,4-dimethoxybenzene protons of [P4-BPO(1-OH)] were observed in the 2D NOESY spectrum (Supplementary Fig. S170). The 1H NMR and 2D NOESY data, together with Job plot (Supplementary Fig. S60) and APCI mass spectrum (peak m/z = 893.4014 for (13⊂[P4-BPO(1-OH)] + Li)+ (Supplementary Fig. S18), proved the formation of a 1:1 complex 13⊂[P4-BPO(1-OH)]. Similarly, alkyl amines (1–3) and alcohols (4–6) formed fragment-pocket complexes G⊂ [P4-BPO(1-OH)] (G = 1–6) with binding constants ranging from 40 to 140 M-1 and from 30 to 70 M-1, respectively. The binding constants between [P4-BPO(1-OH)] and aldehydes (7-9), esters (10-12) and acids (13-15) were smaller compared with those of amines (1-3) and alcohols (4-6). The large binding constants between the host and fragments 1-6 reflected stronger binding interaction originated from the H-bond accepting ability of the nitrogen atom of aliphatic amines or the oxygen atom of aliphatic alcohol32-34. As previously described by Smulders, Zarra and Nitschke27, higher μ and lower log P values meant more polarity of guests. Thus, it was not surprising to see that higher binding affinities correlate to higher μ and lower log P values in the cases of complexation in non-polar solvent CDCl3. Polar compounds tend to escape from non-polar solvent, and clutch H-bond donor in the deep pockets, which could explain the correlations between experimental values (Ka and ΔG) and calculated values (μ and log P) demonstrated by the guests 1-25 in the Table 1.
With the formation of fragment-pocket complexes confirmed, we wondered whether the host was in its endo conformation where the inward-pointing OH group forms a H-bond with a guest. Unambiguous evidence of endo conformation of [P4-BPO(1-OH)] in the complex was obtained from the X-ray diffraction analysis. The single crystal structure of the complex 13⊂endo-[P4-BPO(1-OH)] (Fig. 2c) revealed that 13 is encapsulated in the host pocket with its alkyl chain in the hydrophobic tubular region of the host cavity, and, unmistakably, with its carbonyl oxygen H-bonded to the inwardly pointing OH group in the deep cavity of the host, with the CO···HO distance of 1.941 Å, perfectly in the range of a typical H-bond length35. Thus, the complex 13⊂endo-[P4-BPO(1-OH)] is synergistically stabilized by both hydrogen bonding and hydrophobic interactions. We also obtained crystal structures of a few representative fragment-pocket complexes G⊂endo-[P4-BPO(1-OH)] (G = 9, 17, 21, 24, 25, 31, and 35) in which the carbonyl oxygen is bonded to the OH group of endo-[P4-BPO(1-OH)] in the deep pocket (Fig. 2b, 2d-i). Consequently, we could deduce that in the fragment-pocket complexes 1-25⊂[P4-BPO(1-OH)], [P4-BPO(1-OH)] was in an endo conformation with its OH group inwardly pointing to interact with H-bond acceptor of a guest fragment.
Amides are commonly seen in drug molecules36. Amide functional groups have weak basic and acidic possibilities37. The role of a H-bond between an amide guest and the hydroxyl of [P4-(OH)BPO] in stabilizing a fragment-pocket complex is significant37, which was reflected in our model fragment-pocket binding study: Complexes G⊂endo-[P4-(OH)BPO] (G = 16–25, Table 1) were found to have much larger Ka values than G⊂endo-[P4-(OH)BPO] (G = 1–15, 26–38, Table 1). Besides H-bonding, hydrophobic interaction also plays a significant role in stabilizing the fragment-pocket complexes in which the hydrophobic fragments of the ligands are energetically favored at the interface of the host pocket38,39 Thus, the binding mode of an amide fragment in a “fragment-pocket” complex is determined not only by the strength of amide carbonyl-hydroxyl (CO···HO) H-bond, but also by the hydrophobic interaction between the inner cavity of the host and the guest chain38,39. This was well reflected in complexes 21⊂endo-[P4-(OH)BPO] and 24⊂endo-[P4-(OH)BPO] where alkyl chains of guests 21 and 24 were trapped in the host pocket albeit they were connected to different points of amide functional group (either to the amino N or the carbonyl C atoms, respectively), as shown by the single crystal X-ray structures (Fig. 2e and 2f). The hydrophobic interaction between the predominantly hydrophobic inner surface of the pocket and the alkyl chain of the fragment determined the binding modes of these two complexes. As 21 was more polar (higher μ and lower log P value) than 24 (Table 1), the Ka value of 21⊂endo-[P4-(OH)BPO] doubled that of 24⊂endo-[P4-(OH)BPO] (Table 1). As amide 25 has two butyl groups connected to the amino N and the carbonyl C, respectively, two possible binding modes for 25 exist with either of the two butyl groups confined in the cavity of endo-[P4-(OH)BPO]. The 1H NMR data (Supplementary Fig. S21, ) and the crystal structure of 25⊂endo-[P4-(OH)BPO] (Fig. 2g) indicated that the amine subgroup of 25 was left outside of the binding pocket, possibly owing to the difference in charge density and polarity between the N and C atoms associated with the C=O (Mulliken charges calculated to be -0.518556e and -0.417205e for the N and C atoms, respectively) (Supplementary Information). As a consequence, the solvation interaction between the solvent-exposed N-butyl chain and the “nonpolar” solvent (CDCl3) brought down the stability of 25⊂endo-[P4-(OH)BPO]40, as evidenced by its lower Ka than those of the 21⊂endo-[P4-(OH)BPO] and 24⊂endo-[P4-(OH)BPO].
Heterocyclic rings are predominant architectural components of pharmaceuticals and allow for variable biological interactions otherwise hard to achieve41,42. The binding of five membered heterocycles (26-34), non-aromatic and aromatic, to endo-[P4-BPO(1-OH)] were thus studied. The obvious upfield shifts of the proton signals of the five membered heterocyclic guests (26-34) in the 1H NMR spectra in CDCl3 upon mixing with [P4-BPO(1-OH)] evidenced the formation of the corresponding host-guest complexes (Table 1). Unambiguous evidence of an examplar heterocyclic fragment-pocket complex was provided by the single crystal structure of 30⊂endo-[P4-BPO(1-OH)] which revealed that thiazole 30 is cramped in the tubular cavity of endo-[P4-BPO(1-OH)] with its N atom forming a H-bond with the -OH group of the host (Fig. 2h). Among “screened” heterocyclic fragments, imidazole 29 formed fragment-pocket complex with a higher bonding constant (Table 1), which implied that 29 might use both of its H-bond acceptor (=N) and donor (-NH) in the interaction with the pocket.
Sulfoxide, oxirane, nitrile and nitro group frequently appear in pharmaceuticals43-47. We therefore included fragments dimethyl sulfoxide (DMSO) (35), 1-nitropropane (36), acetonitrile (37) and 2-butyloxirane (38) in our exploration of fragment-pocket interaction study. DMSO molecule is dipolar aprotic with a strong polar sulfoxide group and two hydrophobic methyl moieties43,44, which facilitate the formation of a stable complex 35⊂endo-[P4-BPO(1-OH)] (Fig. 2f). With a nitro group that forms H-bond with a -OH group45, plus a propyl chain to reach the hydrophobic area in the host’s cavity, fragment 36 formed stable 36⊂endo-[P4-BPO(1-OH)] as expected. Nitrile group plays efficacious roles of in drug molecules46, so acetonitrile (37), with simple structure (linear N≡C−C skeleton) and small size (C−N distance 1.16 Å), was selected as a nitrile fragment. The H-bond formed between C≡N nitrogen and -OH of the host was not strong, and the hydrophobic methyl did not contribute much binding affinity, which result in a low binding constant for 37⊂endo-[P4-BPO(1-OH)]. 2-Butyloxirane (38) formed 38⊂endo-[P4-BPO(1-OH)] with a low Ka possibly owing to the weak H-bond between OH of the host and the oxirane oxygen47.
Fragment-based drug design (FBDD). Encouraged by the aforementioned results of the fragment-pocket complexes G⊂endo-[P4-BPO(1-OH)] (G = 1–38), we attempted a model FBDD aiming at “lead molecules” to bind the “target” pocket – endo-[P4-BPO(1-OH)]. Among the fragments 1-38, amides 16, 18, 19 and 21, with amine alkyl chains, gave binding constants greater than 200 M-1, so these N-alkyl substituted formamides and acetamides were designated as the screening-generated “hit molecules”. Given that all of the fragments 1-38 are monofunctional, and do not have structural handle to utilize the H-bond acceptors on the inner surface of endo-[P4-BPO(1-OH)] cavity, bifunctional amides 39 and 40 possessing an H-bond donor at the other end of the molecules were designed for the “target” endo-[P4-BPO(1-OH)]. We were delighted to see that the binding constants of designed bifunctional “lead molecules” 39 and 40 were enhanced by more than 13 and 20 folds, respectively, compared to their monofunctional analog 23 (Table 1). We were delighted to see that 40 achieved the highest binding strength among all of the fragments listed in Table 1 with a Ka of 1.2 × 103 [M-1]. The stronger binding of 40 with endo-[P4-BPO(1-OH)] than that of 39 could be partially explained by the higher diploe moment and lower log P of the former than the latter (Table 1). In the 1H NMR spectra of both complexes (Supplementary Fig. S45 and S46), significant upfield shifting of methylene proton signals of the bound guest was observed, implying that the guests were embraced inside the pocket cavity. Despite extensive efforts, crystalization of the fragment-pocket complexes 39 or 40⊂endo-[P4-BPO(1-OH)] was not successful. Nonetheless, computational docking of 39 and 40 to the pocket of endo-[P4-BPO(1-OH)] (Fig. 5) clearly demonstrated that the “lead molecules” were bound to the pocket cavity through a H-bond between amide C=O of the fragment and OH in the deep pocket with O···H distance of 1.573 Å and 1.597 Å, respectively, and, remarkably, the second H-bond between the fragment’s OH and carbonyl O on the inner wall of pocket cavity with bond length, 1.955 Å and 1.865 Å respectively (Fig. 5).
Complexation through “induced-fit”. It is generally believed that the binding of a ligand to a conformationally free protein is mechanistically via “conformational selection”, whereby a ligand selectively binds to a form of the protein pocket, or via “induced fit”, whereby a ligand binds to a predominantly free conformation of a protein, followed by a conformational change of the protein to form a preferred protein-ligand complex49,50. Theoretically, the rim-bound hydroxyl group in conformationally free [P4-(OH)BPO] should be able to park on either inside or outside the tubular wall to result in endo-[P4-BPO(OH)] or exo-[P4-BPO(OH)]23. In solution, [P4-(OH)BPO] exists in a rapid equilibrium between endo-[P4-(OH)BPO] and exo-[P4-(OH)BPO] isomers, as evidenced by the single set of proton signals in its 1H NMR spectra in various deuterated solvents, such as dichloromethane-d2, chloroform-d3, acetone-d6, acetonitrile-d3, DMF-d4 and DMSO-d6 (Supplementary Fig. S4), most probably owing to interconversion between the endo- and exo-conformational isomers in a rate faster than the NMR timescale. We therefore deduced that formation the fragment-pocket complexes G⊂endo-[P4-(OH)BPO] (G = 1–40) could be an “induced fit” process where a fragment guest passing through the host tube clutches the freely moving OH through H-bond interaction, and, as a consequence, freezes conformationally free [P4-(OH)BPO] into an endo-complex. Based on this assumption, we conjectured that an exo-[P4-(OH)BPO] conformational isomer could be also “frozen” out by non-polar fragment guest geometrically complementary. As expected, exo-[P4-(OH)BPO] conformational isomer was “frozen” out by solvent-induced crystallization in hexane, and was confirmed by 1H and 2D NOSEY NMR spectra in CDCl3 (Supplementary Fig. S47 and S175) and unambiguously single crystal X-ray structure (Fig. 6) in which the phenyl group linked to the quaternary carbon atom lands on the inner side of the tubular core, leaving the hydroxyl group pointing outside, a hexane molecule is imprisoned in the tubular cavity through solely hydrophobic interaction (Fig. 6).