Structure-assisted discovery of the first non-retinoid ligands for Retinol-Binding Protein 4

Yingcai Wang a,⇑, Richard Connors a, Pingchen Fan a, Xiaodong Wang a, Zhongyu Wang a, Jiwen Liu a, Frank Kayser a, Julio C. Medina a, Sheree Johnstone a, Haoda Xu a, Stephen Thibault a, Nigel Walker a, Marion Conn b, Ying Zhang b, Qingxiang Liu b, Mark P. Grillo c, Alykhan Motani b, Peter Coward b, Zhulun Wang a,⇑

Abstract

Retinol-Binding Protein 4 (RBP4) is a plasma protein that transports retinol (vitamin A) from the liver to peripheral tissues. This Letter highlights our efforts in discovering the first, to our knowledge, non-retinoid small molecules that bind to RBP4 at the retinol site and reduce serum RBP4 levels in mice, by disrupting the interaction between RBP4 and transthyretin (TTR), a plasma protein that binds RBP4 and protects it from renal excretion. Potent compounds were discovered and optimized quickly from highthroughput screen (HTS) hits utilizing a structure-based approach. Inhibitor co-crystal X-ray structures revealed unique disruptions of RBP4–TTR interactions by our compounds through induced loop conformational changes instead of steric hindrance exemplified by fenretinide. When administered to mice, A1120, a representative compound in the series, showed concentration-dependent retinol and RBP4 lowering.

Keywords:
RBP4
A1120
Non-retinoid
Small molecules
Fenretinide

Summary

Retinol-Binding Protein 4 (RBP4) is a physiological carrier of retinol (vitamin A) in blood, delivering retinol from the liver to peripheral tissues.1,2 Reports have suggested that RBP4 causes insulin resistance and that lowering RBP4 levels improves insulin sensitivity.3–8 However, the relevance of RBP4 as a therapeutic target in humans remains an open question.9–12 In the circulation, the RBP4-retinol complex interacts with transthyretin (TTR—transports thyroxine and retinol), effectively increasing the molecular weight of RBP4 and preventing its loss by glomerular filtration. Disruption of the RBP4–TTR complex results in a rapid reduction in plasma RBP4 levels.13,14 In addition to retinol, other endogenous and synthetic retinoids bind to RBP4, including all-trans and 13-cis retinoic acid, retinyl acetate, N-(ethyl)retinamide, and fenretinide (Fig. 1).14–16 The co-crystal structures of RBP4 complexed with various retinoids show a similar binding mode, with the cyclohexene ring positioned within the internal cavity and the polar head group pointing towards the exterior of the protein.17–20 Two characteristics of fenretinide binding are worth noting: (1) it changes the position of the loop regions that form the RBP4–TTR interface, and (2) the large phenylamide head group sticks into the space normally occupied by TTR.14,19 As a result, fenretinide completely disrupts the RBP4– TTR interaction. Following in vivo administration of fenretinide, the protective effect of TTR is lost and RBP4 and retinol are rapidly cleared from the serum. While fenretinide has been used by many groups to probe the function of RBP4, it is difficult to distinguish the RBP4-related effects of fenretinide from its retinoid receptorrelated effects. Thus, a selective RBP4 ligand (i.e., non-retinoid) would be desired to explore RBP4 biology and hopefully to elucidate its therapeutic utility.
We previously reported in vivo studies with a non-retinoid small molecule ligand A1120 that binds with high affinity to RBP4 and disrupts the RBP4–TTR complex.21 Herein, we report the hit identification and optimization that led to the discovery of A1120. A high-throughput screen (HTS) employing a radioligand binding assay21 was conducted to identify non-retinoid ligands for RBP4. Several hits were identified, including amide 1 and urea 2 (Fig. 2). Major medicinal chemistry efforts were focused on optimizing from these two hits. The SAR was guided by the radioligand binding assay (hRBP4SPA) and a fluorescence-resonance energy transfer assay that measures the ability of compounds to disrupt the RBP4–TTR protein–protein interaction (hRPB4–TTR FRET).22
The co-crystal structure of RBP4 with HTS hit 1 was determined to a resolution of 2.4 Å (Fig. 3).23 The overall fold of RBP4 in the compound 1 co-crystal structure resembles other previously reported RBP structures.17,24,25 It adopts a core b-barrel of eight up-and-down b-strands, an N-terminal coil, and a C-terminal helix followed by a coil region (Fig. 3A). The co-crystal structure showed well-resolved electron density for compound 1 in the central cavity of the b-barrel, the same site bound by retinol. The compound makes numerous van der Waals contacts with the protein with the chlorophenyl ring binding deep in the pocket and the benzylimidazole group sitting at the opening of the cavity. In addition, compound 1 makes two direct hydrogen bonding interactions with RBP4, that is, the carbonyl accepting a hydrogen bond from the backbone amide of Leu37 and the N3 atom of the imidazole ring accepting a hydrogen bond from the side chain gNH2 of Arg121 (Fig. 3B).
Structural analysis of the protein–ligand interactions provided valuable insights for optimization of the hit. The chlorophenyl ring is located at the wide bottom of the pocket where there is more space for additional hydrophobic interactions (Fig. 3C). The middle piperazine ring is located in a narrow channel which does not tolerate larger groups or dramatic conformational changes. The benzylimidazole group is exposed to solvent, indicating great tolerance toward additional substitution and a possible preference for polar groups.
Based on these observations, we took the following approaches. For the left-hand side of the molecules, we increased the size or constrained the conformation for a better fit in the deep pocket, while trying to optimize hydrophobic interactions. For the linker, we decided to keep steric hindrance to a minimum while fine-tuning the fit to better align with the left and right sides of the pocket. For the right-hand side of the molecules, we attempted to introduce polar groups for more interactions, especially H-bonding, while keeping in mind that dramatic changes and many groups could be tolerated in this area.
Because readily available isocyanates would allow us to quickly explore the right-hand side for structural diversity, we decided to explore SAR based on compound 2. However, our initial efforts focused on the left-hand side of the molecule as shown in Table 1. A sterically-demanding group such as CF3 at the 5-position of the pyridine (3) or a smaller group such as Cl at the 3-position of pyridine (4) resulted in a ten-fold loss of activity. The significant loss of activity of compound 5 and 6 indicated that an ortho-substituent is necessary for adopting a preferred conformation. The fact that phenyl is more potent than pyridine (8 vs 2; 7 vs 4) is consistent with the crystal structure findings that hydrophobic interactions are favored in the deep pocket. Compounds 10–12 indicate that both size and conformation need to be ‘right’ to assure potency.
The good activity of compound 9 (Table 1) indicates that a larger group is tolerated for the left side as predicted, as long as it can fit the pocket. We further prepared two pairs of compounds (Table 2, 13 and 14 & 15 and 16) comparing the size effect of the left side of the molecule. It is clear that filling the deep pocket alone can dramatically increase the potency.
We were able to obtain a co-crystal structure of compound 15 with RBP4 to 2.3 Å (Fig. 4).23 Like hit 1, compound 15 occupied the retinol site in the central cavity of the protein, with its tricyclic moiety positioned deep in the pocket and the trifluoromethyl-phenyl moiety situated at the opening (Fig. 4A). The bulky cyclopenta-thieno-pyrimidine group indeed filled more space at the wide bottom of the pocket (Fig. 4A). As a result, the oxygen atom of the carbonyl moiety of 15 was positioned at the same spot as the N3 atom of the imidazole ring in hit 1. Thus, compound 15 makes one hydrogen bonding interaction through the carbonyl moiety, accepting a hydrogen bond from the gNH2 group of the Arg121 side chain (Fig. 4B). At the open end of the pocket, the trifluoromethyl group fit almost perfectly into a surface depression formed by the main and sidechainsofLeu37, andside chainsofPhe36,Val61,and Met73. The phenyl group forms a stacking interaction with Phe96.
The co-crystal structure of 15 with RBP4 revealed that the shape complementarity of 15 and the central cavity of RBP4 seem to play a key role in ligand binding to the protein. The carbonyl group, which contributes one hydrogen bonding interaction with Arg121, might not be required, as compound 14 is more potent than 15. Notably, the corresponding 3-CF3-pyridine (13, 16) compounds are much less active than compounds 14 and 15.
Comparison of the RBP4 co-crystal structures of 1 and 15 suggested that the middle piperazine ring might adopt different trajectories when the left side is modified from chloro-phenyl moiety in 1 to cyclopenta-thieno-pyrimidine in 15 (Fig. 4C). This also demonstrated some flexibility in the linker region modification, as shown in Table 3. Adding a methyl group to two different positions on the piperazine ring to constrain the conformation did not significantly impact potency (compounds 17 and 18). Piperazinone 19, however, decreased the activity significantly. Enlarging the ring to a 1,4-diazepane (20) was also tolerated. It is beneficial to remove the nitrogen atom from left side of the piperazine ring to form a piperidine (21), while similar replacement of the urea nitrogen with carbon was detrimental (22). 

Dotted line represents response to fenretinide treatment.

We then also obtained the co-crystal structure of compound 33 (A1120) with RBP4 at 2.9 Å (Fig. 5).21 In addition to the numerous hydrophobic interactions and van der Waals contacts, A1120 reveals the following polar interactions in the solvent exposed region: the carbonyl group of the urea forms a hydrogen bond interaction with the backbone amide of Leu37, and the carboxylic group of the benzoic acid participates in a salt bridge with Arg121 as well as in a hydrogen bond with the hydroxyl group of Tyr90. While A1120 competes the same binding site as retinol, it does not directly prevent the TTR binding. Instead, A1120 modulates the conformation of RBP4 loop b5–b6 (residues 92 to 99) through steric clashes with Gln98 (Fig. 5B), disrupting the surface complementarity between RBP4 and TTR that is required for recognition.21
In an effort to improve the pharmacokinetic (PK) profiles of the lead compounds, two metabolic soft spots were identified by studying the metabolic ID of compound 2 in rat and human liver microsomes, namely, (1) hydrolysis to give the primary urea; and (2) oxidation (M+16) in the pyridinyl piperazine region. Changing the urea phenyl to a 2-benzoic acid (34) prevented hydrolysis, which was reflected in an improvement in rat microsomal stability and PK profiles (Table 5). Replacing the pyridinyl piperazine (2, 34) with a phenyl piperidine (A1120) reduced oxidation and also contributed to enhanced rat microsomal stability and rat and mouse PK. Comparing A1120 with fenretinide, A1120 demonstrated both improved potency and increased exposure. A1120 was administered to chow-fed B6D2F1 male mice and found to lower RBP4 and retinol levels dose-dependently, with maximal efficacy greater than fenretinide (Fig. 6).
In summary, we utilized structure-based drug design to optimize a series of non-retinoid small molecules to disrupt the RBP4–TTR interaction. Administration of a representative compound A1120 to mice reduced serum RBP4 and retinol levels but did not improve the diabetic state.21 A recent study of A1120 in an Abca4/ animal model showed significant reduction of lipofuscin bisretinoid accumulation, proving a rationale for considering A1120 and its derivatives as potential treatments for atrophic age-related macular degeneration (AMD) and Stargardt disease.26

References and notes

1. Goodman, D. S. Plasma Retinol-Binding Protein In The Retinoids; Sporn, M. B., Roberts, A. B., Goodman, D. S., Eds.; Academic Press: New York, 1984; Vol. 2, pp 41–82.
2. Makover, A.; Soprano, D. R.; Wyatt, M. L.; Goodman, D. S. J. Lipid Res. 1989, 30, 171.
3. Yang, Q.; Graham, T. E.; Mody, N.; Preitner, F.; Peroni, O. D.; Zabolotny, J. M.; Kotani, K.; Quadro, L.; Kahn, B. B. Nature 2005, 436, 356.
4. Graham, T. E.; Yang, Q.; Bluher, M.; Hammarstedt, A.; Ciaraldi, T. P.; Henry, R. R.; Wason, C. J.; Oberbach, A.; Jansson, P.-A.; Smith, U.; Kahn, B. B. N. Engl. J. Med. 2006, 354, 2552.
5. Balagopal, P.; Graham, T. E.; Kahn, B. B.; Altomare, A.; Funanage, V.; George, D. J. Clin. Endocrinol. Metab. 2007, 92, 1971.
6. Cho, Y. M.; Youn, B.-S.; Lee, H.; Lee, N.; Min, S.-S.; Kwak, S. H.; Lee, H. K.; Park, K.S. Diabetes Care 2006, 29, 2457.
7. Stefan, N.; Hennige, A. M.; Staiger, H.; Machann, J.; Schick, F.; Schleicher, E.; Fritsche, A.; Haring, H.-U. Diabetes Care 2007, 30, 1173.
8. Qi, Q.; Yu, Z.; Ye, X.; Zhao, F.; Huang, P.; Hu, F. B.; Franco, O. H.; Wang, J.; Li, H.; Liu, Y.; Lin, X. J. Clin. Endocrinol. Metab. 2007, 92, 4827.
9. Janke, J.; Engeli, S.; Boschmann, M.; Adams, F.; Bohnke, J.; Luft, F. C.; Sharma, A.M.; Jordan, J. Diabetes 2006, 55, 2805.
10. Lewis, J. G.; Shand, B. I.; Frampton, C. M.; Elder, P. A.; Scott, R. S. Clin. Biochem. 2008, 41, 1034.
11. Petersen, K. F.; Dufour, S.; Savage, D. B.; Bilz, S.; Solomon, G.; Yonemitsu, S.; Cline, G. W.; Befroy, D.; Zemany, L.; Kahn, B. B.; Papademetris, X.; Rothman, D. L.; Shulman, G. I. PNAS 2007, 104, 12587.
12. Promintzer, M.; Krebs, M.; Todoric, J.; Luger, A.; Bischof, M. G.; Nowotny, P.; Wagner, O.; Esterbauer, H.; Anderwald, C. J. Clin. Endocrinol. Metab. 2007, 92, 4306.
13. Peng, Y.-M.; Dalton, W. S.; Alberts, D. S.; Xu, M.-J.; Lim, H.; Meyskens, F. L., Jr.Int. J. Cancer 1989, 43, 22.
14. Berni, R.; Formelli, F. FEBS Lett. 1992, 308, 43.
15. Berni, R.; Clerici, M.; Malpeli, G.; Cleris, L.; Formelli, F. FASEB J. 1993, 7, 1179.
16. Cogan, U.; Kopelman, M.; Mokady, S.; Shinitzky, M. Eur. J. Biochem. 1976, 65, 71.
17. Cowan, S. W.; Newcomer, M. E.; Jones, T. A. Proteins 1990, 8, 44.
18. Zanotti, G.; Malpeli, G.; Berni, R. J. Bio. Chem. 1993, 268, 24873.
19. Zanotti, G.; Marcello, M.; Malpeli, G.; Folli, C.; Sartori, G.; Berni, R. J. Bio. Chem. 1994, 269, 29613.
20. Monaco, H. L.; Rizzi, M.; Coda, A. Science 1995, 268, 1039.
21. Motani, A.; Wang, Z.; Conn, M.; Siegler, K.; Zhang, Y.; Liu, Q.; Johnstone, S.; Xu,H.; Thibault, S.; Wang, Y.; Fan, P.; Connors, R.; Le, H.; Xu, G.; Walker, N.; Shan,B.; Coward, P. J. Biol. Chem. 2009, 284, 7673.
22. Coward, P.; Conn, M.; Tang, J.; Xiong, F.; Menjares, A.; Reagan, J. D. Anal. Biochem. 2009, 384, 312.
23. The atomic coordinates and structure factors are deposited in the RCSB protein data bank under the accession codes of 4PSQ and 4O9S for compounds 1 and15, respectively.
24. Newcomer, M. E.; Jones, T. A.; Aqvist, J.; Sundelin, J.; Eriksson, U.; Rask, L.; Peterson, P. A. EMBO J. 1984, 3, 1451.
25. Zanotti, G.; Ottonello, S.; Berni, R.; Monaco, H. L. J. Mol. Biol. 1993, 230, 613.
26. Dobri, N.; Qin, Q.; Kong, J.; Yamamoto, K.; Liu, Z.; Moiseyev, G.; Ma, J. X.; Allikmets, R.; Sparrow, J. R.; Petrukhin, K. Invest. Ophthalmol. Vis. Sci. 2013, 54, 85.