|Home | About | Journals | Submit | Contact Us | Français|
Chemical inducers of dimerization (CIDs) are employed in a wide range of biological applications, to control protein localization, modulate protein-protein interactions and improve drug lifetimes. These bifunctional chemical probes are assembled from two synthetic modules, which each provide affinity for a distinct protein target. FK506 and its derivatives are often employed as modules in the syntheses of these bifunctional constructs, owing to the abundance and favorable distribution of their target, FK506-binding protein (FKBP). However, the structural complexity of FK506 necessitates multi-step syntheses and/or multiple protection-deprotection schemes prior to installation into CIDs. In this work, we describe an efficient, one-step synthesis of FK506 derivatives through a selective, microwave-accelerated, cross metathesis diversification step of the C39 terminal alkene. Using this approach, FK506 is modified with an array of functional groups, including primary amines and carboxylic acids, which make the resulting derivatives suitable for the modular assembly of CIDs. To illustrate this idea, we report the synthesis of a heterobifunctional HIV protease inhibitor.
The natural product FK506 is a powerful immunosuppressant that is used to prevent allograft rejection in transplant patients1. This compound blocks immune cell function through an intriguing mechanism in which it first interacts with the FK506-binding protein (FKBP) and then this complex binds tightly to the phosphatase calcineurin2,3. This unusual binding property has been mirrored in synthetic tools termed chemical inducers of dimerization (CIDs)4–6. Like FK506, these compounds are composed of two, non-overlapping chemical domains, which enables them to bridge two protein targets. CIDs have been employed in a broad range of applications, many focused on understanding signaling pathways, subcellular localization and protein-protein interactions7–14. In addition, this platform has been used in drug discovery applications to modulate specificity, lifetime and affinity15–19.
Despite this versatility, methods for the modular construction of CIDs have not been extensively explored. One notable exception is a study by the Verdine group that explored the systematic synthesis of compounds coupled to the synthetic ligand for FKBP (SLF)20. SLF is an attractive chemical component because it is a non-immunosuppressive analog of FK506 that can be attached to other molecules via pendant anilines18 or carboxylates11,21–23. Although SLF has been used to synthesize many useful CIDs, there are limitations to this approach, including modest aqueous solubility and decreased affinity for FKBP (~10-fold compared to FK506)24. In addition, SLF is challenging to use on a large scale because its synthesis requires ~12 steps, with modest overall yield21,23,24. One alternative is to employ FK506 as a commercially available FKBP-binding ligand. In fact, this strategy was used to create the first CID, a homodimer of FK506 termed FK10127. However, because of its structural complexity, manipulations of FK506 are challenging; for example, reported methods for preparing reactive intermediates require 5 steps with ~50% yield7.
We reasoned that an improved entry into reactive FK506 derivatives might involve selectively modifying the C39 alkene via cross metathesis (CM) using Grubbs’ second generation ruthenium catalyst (GII)25. The GII catalyst was chosen, in part, because it is tolerant of the many functional groups found in FK506, such as alcohols and ketones. In support of this idea, both the Wender and Schreiber groups have diversified complex natural products via CM at a terminal alkene26,27. Another attractive feature of this approach is that the extracyclic C39 position of FK506 is crucial for binding calcineurin. Thus, modifications at this site would be expected to block immunosuppression without impacting the affinity for FKBP28. Consistent with this model, Clemons et al. used CM to selectively append bulky groups to the C39 position on FK506 to interrupt binding to calcineurin27 and Diver and Schreiber used CM to produce FK1012-like homodimers, which are non-immunosuppressive but able to dimerize chimeric FKBP fusions29. Based on these studies, we sought to explore CM as a means to develop modular building blocks for the rapid creation of FKBP-binding CIDs.
Our initial efforts focused on optimizing the microwave-accelerated coupling between FK506 and a representative olefin, 3,4-epoxy-1-butene (Figure 1). Recently, Chapman et al. identified a narrow window of conditions for efficient GII-mediated, microwave-accelerated, ring-closing metathesis on peptide substrates30. Guided by this effort, we systematically varied the reaction conditions and found that microwave irradiation for 5 minutes yielded good conversion rates (40–64%), while shorter (1–2 minutes) or longer (10–20 minutes) irradiation times reduced the yields (Figure 2A). Additionally, by varying temperature, we identified a clear peak of efficiency at 150 °C. Higher temperatures resulted in lower yields and greater side-product formation, while reactions at lower temperatures consisted mostly of unreacted starting materials (Figure 2B). Importantly, when a 20-fold molar excess of the olefin was used, these reactions proceeded readily on unprotected FK506, while lower ratios favored formation of FK1012-like homodimers.
As the success of microwave-accelerated reactions is dependent on absorption of microwave energy by the reaction medium, we examined a variety of solvents and monitored production of the desired CM product (1) by liquid chromatography-mass spectrometry (LC-MS). Using this screening approach, we found that 1,2-dichlorobenzene (DCB) and 1,2-dichloroethane (DCE) produced the highest yields (61–64%; Table 1). Other solvents with strong dielectric constants, such as ethanol, were poor choices because of reagent insolubility.
With these results in hand, our next goal was to evaluate a series of target olefins. Structural and chemical diversity were key in our selection criteria, as we hoped to produce a range of FK506 intermediates containing orthogonally reactive electrophiles and nucleophiles for the synthesis of CIDs. Accordingly, we prepared a series of reactions using 21 different target olefins and 10% GII catalyst irradiated in a sealed tube at 150 °C for 5 minutes in DCB (Table 2). Based on the work of Chatterjee et al., we hypothesized that target olefins bearing relatively bulky and electron-withdrawing groups might provide the highest yielding CM reactions31. In agreement with this concept, some of the compounds with carbonyls or carboxylates close to the reacting olefinic center had the highest conversion efficiency. For example, acrylate gave an excellent yield (99%), as did olefins with acrylic or benzoic acids (79–99%). Some of the halogenated olefins, such as allyl bromide, had low yields, which MS experiments confirmed was due to the propensity of the unhindered allyl halide to undergo rapid, non-productive homodimerization. Poor yields were also observed for free amines, such as allylamine and 3-butenylamine (less than 5%), while protecting with Boc anhydride significantly improved reaction yield for the allylamine reaction (50%). Finally, for a subset of olefins that were only mildly soluble in DCB, we explored isopropanol as an alternative solvent. This strategy improved the solubility of all the chosen substrates and the yield of the CM reaction involving acryloyl chloride improved from 6% to 88%; however, this solvent did not significantly improve the reaction efficiency for the other olefins.
From these studies, we selected FK506 derivatives with the most favorable combination of properties (e.g. yield, solubility and reactive functionality) for re-synthesis and expanded studies. No effort was made to separate E and Z isomers, because these have been found to be equally active in biological assays29. Following HPLC purification to remove trace amounts of unmodified FK506, a small collection of derivatives (1–6) suitable for the generation of CIDs was produced (Figure 3).
Because these molecules are intended for use in the exploration of biological systems, we were also interested in finding efficient ways to remove ruthenium byproducts. Even trace amounts of these organometallic complexes can lead to olefin isomerization, product decomposition and cell toxicity. Several relevant purification methods have already been explored, including the use of tris-(hydroxymethyl)phosphine, triphenylphosphine oxide, and dimethyl sulfoxide32–34. However, these procedures require multiple reagents, subsequent column chromatography, and relatively long exposure times for maximal efficacy. Here, we explored a quick and efficient strategy using silica-bonded equivalents of the metal scavenger, 1-propanethiol. We found that treatment of the crude reaction mixture with 5 equivalents of resin for 30 minutes reduced the levels of the ruthenium byproducts by greater than 50-fold (Figure 4). This simple method compared favorably to other approaches, especially in the minimal handling steps and short times required.
One of the potential advantages of using FK506 as a module in the synthesis of bifunctional molecules is its superior affinity for FKBP, especially when compared to SLF24. We recently reported that appending SLF to an HIV protease inhibitor partitions the resulting molecule into blood cells and prolongs its lifetime 20-fold in vivo35. Evidence suggests that this improved persistence results from the modified compound being sequestered into the cytoplasm of the blood cells, which is an environment that contains high levels of FKBP but is nearly devoid of the P450 enzymes that metabolize xenobiotics. Based on this model, we reasoned that cellular partitioning and drug lifetime might be sensitive to the affinity for FKBP19. Therefore, as a proof-of-concept for the present work, we sought to generate a fusion between FK506 and an HIV protease inhibitor, which would allow us to directly compare SLF- and FK506-conjugated molecules. Towards that goal, we coupled the reactive FK506 derivative 2 in a single step to an amprenavir-like core to yield the bifunctional molecule 7 (Figure 5A). Coupling of the core inhibitor to the FK506 derivative occurred through a pendant amine, which is known to be a well tolerated attachment point (Figure 5B)35,36.
To test the influence of FKBP binding on cellular partitioning, we first examined the distribution of 4-methoxy amprenavir in whole blood and, consistent with previous studies, we found that it distributed evenly between the plasma and cellular compartments (Figure 6). Next, we confirmed that the corresponding SLF-modified derivative 8 showed a strong (~9-fold) preference for the cellular fraction (Figure 6)35. Remarkably, the FK506-modified derivative 7 was concentrated by more than ~22-fold in the blood cell compartment. Thus, replacing SLF with FK506 exaggerated cellular partitioning by more than 2-fold, consistent with an important role for FKBP-binding affinity. These results clearly illustrate one advantage of using orthogonally reactive FK506 derivatives to synthesize new CIDs.
The goal of this effort was to generate reactive building blocks for CID applications. Accordingly, we systematically investigated several reaction variables and identified conditions suitable for microwave-accelerated CM and diversification at the C39 position of FK506. These efforts afforded reactive analogs (1–6), which present either electrophiles or nucleophiles. These derivatives are designed for facile installation into parent compounds bearing the corresponding functionality. One key feature of this strategy is that the reactive intermediates are accessed in a single step using unprotected FK506. During the course of these studies, we also explored a scavenger resin-based method for removing ruthenium byproducts, which may be particularly advantageous for parallel syntheses of CIDs. Finally, as demonstrated by the synthesis of the bifunctional protease inhibitor 7, these methods are expected to provide a convenient and modular platform for creating FKBP-binding CIDs with interesting new properties.
Solvents and olefins were purchased from Sigma or ThermoFisher and used without further purification. FK506 was purchased from LC Labs. All mass spectrometry results collected on a Shimadzu LC-MS using an APCI probe in negative mode. All NMR spectra collected on a Varian 600 MHz system, with the samples dissolved in d-DMSO.
To a solution of FK506 (2.0 mg; 0.00249 mmol) in dichlorobenzene (2.0 mL) was added the target olefin (20 equiv; 0.0498 mmol) followed by catalyst (0.1 equiv; 0.211 mg; 0.249 µmol) in a microwave vial. The vial was then sparged with nitrogen, sealed, and placed in a Biotage Initiator EXP microwave reactor. The mixture was stirred for 30 seconds, followed by irradiation at the desired temperature and time. To achieve the desired temperature, the microwave power was typically ~20W. To quantify reaction progress, crude mixtures were placed in 1.5 ml vials and analyzed with a Shimazdu (Columbia, MD) series 2010EV Liquid Chromatograph Mass Spectrometer equipped with an APCI probe to minimize ion suppression. Samples (5.0 µL) were injected onto 1.8 micron ZORBAX Eclipse XDB-C18 column (Agilent), and elution was performed with a stepwise 10–100% gradient of acetonitrile over 11 minutes. Products eluted between 5.5 and 6 min, while the peak for unreacted FK506 was found at 6.7 min. Quantification was performed using LCMSolution Version 2.05 using external standards. For the CM reactions that were selected for further analysis, the scale was increased 10-fold and the resulting crude reactions were treated with silica-immobilized, 1-propanethiol (5 eq.; Biotage; Isolute Si-Thiol) for 30 minutes, filtered and then the product was purified by preparative chromatography on a Waters Spherisorb S10 ODS2 column (10 X 250 mm). For compound 6, the crude product was additionally treated with 5 equivalents of 1:1 TFA:DCM, followed by dilution into DCM, neutralization with 5% sodium bicarbonate and concentration by roto-evaporation prior to HPLC. See the Supporting Information for characterization (1H NMR and mass spectrometry).
As a test of the modular assembly of bifunctional molecules, the reactive FK506 derivative 2 (1.0 eq; 15.8 mg; 0.0186 mmol) was added to the advanced HIV protease inhibitor (1.2 eq; 9.1 mg; 0.0224 mmol) in 3.0 mL of DMF with DIC (10.0 eq; 29.2 µL; 0.1863 mmol), HOBt (1.0 eq; 2.5 mg; 0.1863 mmole) and DMAP (1.0 eq; 2.3 mg; 0.1863 mmol). This reaction was stirred at room temperature overnight and the resulting product 7 purified in 75% yield by HPLC as described above. The blood partitioning experiments were performed as described.35 The synthesis of the amprenavir core structure and compound 8 have been reported35. See the Supporting Information for characterization (1H NMR and mass spectrometry). FK506 coupled to HIV protease inhibitor (7); molecular formula C66H97N3O17S; expected mass (m/z) = 1236.55; actual mass (m/z) = 1235.95. δ 7.74(d), δ 7.63(s), δ 7.57(d), δ 7.51(s), δ 7.44(d), δ 7.36(s), δ 7.28(t), δ 7.19(s), δ 7.11(t), δ 6.9(d), δ 6.86(d), δ 5.42(s), δ 5.36(s), δ 5.27(s), δ 5.125(dd), δ 5.04(d), δ 4.8(m), δ 4.7(d), δ 4.6(s), δ 4.48(s), δ 4.24(d), δ 3.9(s), δ 3.87(d), δ 3.81(s), δ 3.78(s), δ 3.71(s), δ 3.62(d), δ 3.54(d), δ 3.41(dd), δ 3.33(d), δ 3.28(d), δ 3.22(s), δ 2.9–3.08(m), δ 2.91(d), δ 2.78(m), δ 2.72(m), δ 2.27(s), δ 2.1(s), δ 2.0(m), δ 1.92(s), δ 1.86(m), δ 1.8(d), δ 1.72(d), δ 1.65(d), δ 1.62(s), δ 1.55(s), δ 1.48(m), δ 1.39(s), δ1.26(t), δ 1.21(t), δ 1.14(t), δ 1.03(m), δ 0.86(d), δ 0.81(d), δ 0.79(d), δ 0.73(d).
To determine the inhibitory potency of compound 7, we used a commercially available, FRET-based assay (Bachem, Torrance, CA), as previously reported35. Briefly, HIV-1 protease was preincubated with inhibitor for 60 minutes at 37 °C, followed by addition of 30 nM HIV-1 protease substrate 1 (Arg-Glu(EDANS)-Ser-Gln-Asn-Tyr-Pro-Ile-Val-Gln-Lys(dabcyl)-Arg; Molecular Probes, Eugene OR). Substrate cleavage was measured for 60 min. at 37 °C by monitoring fluorescence at 490 nm on a SpectraMax M5 plate reader (Molecular Devices, Mountain View, CA).
The authors thank R. Owen and E. Vedejs for helpful comments, J. Bauer and T. Almeida for technical assistance. C.G.E. is supported by a predoctoral fellowship from the Cellular Biotechnology Training Grant (GM008353-18). J.G.E. acknowledges ThermoFisher Sci. and the McKnight Foundation for financial support.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Synthesis and 1H NMR spectra of modified FK506 derivatives and modified HIV protease inhibitors and a complete record of all the reactions.