Facile Synthesis of Homodimeric Protein Ligands

Many proteins exist as oligomers (homodimers, homotrimers, etc.). A proven strategy for the development of high affinity ligands for such targets is to link together two modest affinity ligands that allows the formation of a 2 : 2 (or higher‐order) protein‐ligand complex. We report here the discovery of a convenient, “click‐like” reaction for the homodimerization of protein ligands that is efficient, operationally simple to carry out, and tolerant of many functional groups. This chemistry reduces the synthetic burden inherent in the creation of homodimeric ligands since only a single precursor is required. The utility of this strategy is demonstrated by the synthesis of homodimeric inhibitors, including PROTACs.


Introduction
Small molecule modulators of protein activity are important tools in basic research and represent a starting point for drug development.Highly potent molecules rarely arise from primary discovery efforts, such as high-throughput screening (HTS) campaigns. [1]Extensive optimization of these hits is usually required, which is tedious and labor-intensive.This can sometimes be short-circuited for proteins that exist as stable homodimers or higher order oligomers through the creation of bivalent ligands that bind tightly to the target thanks to avidity effects. [2]While the increase in molecular weight inherent in this approach is not optimal for drug development, it is an excellent way to generate tool compounds rapidly.
There are many ways to create bivalent ligands.For example, a common strategy is to create derivatives of the ligand with functional groups that can be linked together efficiently and orthogonally, for example using copper-catalyzed Huisgen cycloaddition of an alkyne and an azide or by oxime formation between an aldehyde and an alcoxyamine. [3]owever, this necessitates the synthesis of two compounds.Since it is usually necessary to explore various linker lengths and geometries to realize the maximal avidity effect, these approaches to bivalent ligand synthesis can be quite laborintensive.A more attractive route from an efficiency perspective is to equip the ligand with a functional group that can homodimerize such that only a single modified ligand is required.For example, olefin metathesis has been employed to make dimeric compounds of this sort. [4]Ring-opening olefin metathesis has been used to create higher-order oligomeric ligands, which have been useful in engaging cell surface targets such as carbohydrate receptors.However, there are relatively few methods for the creation of homodimers from a single starting material, particularly if it is a functionally complex ligand.Expanding the toolbox of such reactions would be worthwhile.Towards this end, we describe here the development of a convenient protocol for dimerizing protein ligands via pyridone formation from a compound containing a masked β-formyl acetamide unit.The reaction is tolerant of many functional groups, operationally simple to carry out and does not require the use of expensive reagents or catalysts.We demonstrate the utility of this approach in several systems.

Results and Discussion
Recently, we had cause to prepare an aldehyde derivative of VH 032 (see Figure 1), a ligand for the Von Hippel Landau (VHL) E3 Ubiquitin ligase.Towards this end, the compound shown at the top left of Figure 1 was acylated with commercially available 3,3-dimethoxypropionic acid.The resultant amide, a solid, was then added to neat trifluoroacetic acid (TFA) to a final concentration of 1 M, and stirred for 36 hours, with the intention of unmasking the protected aldehyde.After removal of the TFA using lyophilization, liquid chromatographic analysis showed clean and nearly quantitative conversion to a new product (Figure 1B) However, subsequent characterization revealed the product was not the expected aldehyde but rather the pyridone WJ621.
While we were initially surprised by this, a subsequent literature search revealed that β-keto amides have been shown condense to form a pyridone product when heated in the presence of an acid catalyst (see Figure 1C for the proposed mechanism). [5]This chemistry has been employed to synthesize pyridone-containing natural products and pharmaceuticals. [6]owever, to the best of our knowledge, there are no reports in the literature of the apparently far more facile dimerization of β-formyl acetamides to form pyridones.Moreover, the published reports indicate the condensation of β-ketoamides is quite sluggish.Indeed, we confirmed that exposure of Nbenzylacetoacetamide (1 M) to neat TFA does produce the expected pyridone (Figure 1C) but in only 10 % yield after 36 hours.Even after allowing the solution to sit at room temperature for two months only a 34 % conversion was achieved (Figure S1).We explored several other acid catalysts, but none were as efficient as TFA, and, in all cases, the ketone provided much poorer results than the masked aldehyde (Figure S1).
Intrigued by the efficiency and simplicity of the masked aldehyde coupling process, we proceeded to investigate the function group compatibility of the reaction by exposing the compounds in Scheme S1 to TFA under the conditions described above (1 M starting material in TFA, 36 hours at room temperature).Many of the reactions produced the pyridonedimerized product in quantitative yield, including starting materials containing halogen, hydroxyl, ester, nitrile, amine (Boc protected), sulfone, pyridine, and ketone functionalities.However, some reactions were less efficient.For example, the phenyl amide WJ891 produced only a 60 % yield of the pyridone product (WJ908), along with a variety of side products (Figure S2). [7]No dimerization product was detected for the protected aldehyde, the Trt protected thiol, or the indole-or furan-containing substrates.This is likely due to the reactivity of these moieties with TFA.Coupling was also inhibited by placing a bulky group directly adjacent to the nitrogen (Figure 2).As expected, no dimerization product was detected for the tertiary amides WJ873 or WJ859 (structure shown in Scheme S1).
We noted that at more dilute concentrations (10 mM), a few of the reactants (WJ891, WJ888, WJ828) produced uncharacterized side products (Figure S2).Therefore, we recommend using more concentrated solutions (1 M) of the masked β-formyl acetamide when possible.

Synthesis and evaluation of pyridone-linked homodimeric PROTACs
There is considerable interest in the development of PROteolysis-TArgeting Chimeras (PROTACs) as probe molecules and drugs leads. [8]These are chemical dimerizers that bring together a target protein and an E3 Ubiquitin (Ub) ligase in such a way that the target is poly-Ubiquitylated and subsequently trafficked to the proteasome for destruction. [9]It is clear that the length and geometry of the linker connecting the target protein and E3 Ub ligase ligands has a profound effect on the activity of a PROTAC, in part because Ubiquitylation requires proper alignment of the activated Ub and an acceptor lysine, as well as significant "neo-proteinÀ protein interactions" (neo-PPIs) between the ligase and the target protein. [10]As a result, extensive empirical optimization of the linker is often necessary, requiring significant synthetic effort.For the special case of homodimeric PROTACs (where the E3 Ub ligase is also the target protein), we imagined the pyridone homodimerization reaction could reduce the synthetic burden inherent in linker optimization by half.Ciulli and co-workers [9b] have shown that a polyethylene glycol (PEG)-linked dimer of VH 032 catalyzes the turnover of the long isoform of VHL.Therefore, we employed the pyridoneforming reaction to create various VH 032 dimers.
Four such dimers (Figure 3A) were generated by treating VH 032-linked caged β-formyl acetamide amides with trifluoro-  acetic acid (TFA) for 36 hours at room temperature.After lyophilization the desired product was obtained in each case in high yield and purity without the need for purification (Figure 3A and Figure S3).The chemical structures of the dimers are shown in Figure S4.
To the best of our knowledge, pyridones have not been employed as PROTAC linkers.Thus, we wanted to ensure that their physicochemical properties were suitable to support a high level of degrader activity in cellular assays.Thus, HeLa cells were exposed to the dimerized molecules at a final concentration of either 1 μM (Figure 3B) or 10 μM (Figure S5) for 12 hours.9b] The cells were lysed, and the level of pVHL30, the long isoform of VHL, in the extract were determined by Western blotting normalized to a GAPDH loading control (Figure 3B).As shown in previous studies, [9b,11] the short isoform of VHL ( pVHL19) is resistant to PROTACmediated degradation, presumably because it lacks an unstructured region that is essential for proteasome-mediated turnover.All four of the pyridone-linked dimers were active pVHL30 degraders, with WJ662 (Figure 3C) displaying the best activity (70 % degradation of pVHL30 at 1 μM; Figure 3C).The chemical structure of WJ662 is shown in Figure 3D.A titration experiment showed that WJ662 displayed a half-maximal degradation concentration (DC 50 ) of approximately 220 nM and a maximum degraded fraction (D max )of 98 % at a concentration of 40 μM (Figures 3E and F).A time course showed that maximal VHL degradation was achieved five hours post-treatment (Figures 3G and H).In the presence of the proteasome inhibitor Bortezomib, no degradation was observed (Figures 3I  and J), as expected.We conclude that the pyridone linker is suitable for the preparation of highly cell-active homodimeric PROTACs.

Synthesis and characterization of pyridone-linked bivalent BRD4 inhibitors
The bromodomain and extra terminal domain (BET) family of proteins are important mediators of chromatin structure and gene expression. [12]They employ tandem bromodomains to associate with acetylated lysine residues.One of the family members, BRD4, has been shown to be highly enriched in enhancer regions that drive the expression of pro-oncogenic factors such as c-MYC, and is thus an important cancer target. [13]mall molecule-mediated disruption of the bromodomainacetyl lysine interaction is a popular strategy for BRD4 inhibition. [14]JQ1 is a tool compound used commonly for this purpose. [15]radner and co-workers reported that a dimer of JQ1, called MT1, capable engaging both bromodomains in BRD4, was 100fold more potent than the parent compound. [16]We therefore sought to apply the pyridone-forming reaction to generate bivalent ligands for BRD4.
3c] After treatment with TFA at room temperature for 36 hours, the expected pyridone-linked dimers were produced efficiently in each case (Figure 4A; also see Figure S6).The chemical structures of all the dimerized ligands are shown in Figure S7.
MV-4-11 cells, which depend on BRD4 activity to proliferate, were treated with different concentrations of each compound for 72 hours.JQ1 and MT1 were used as controls (chemical structure shown in Figure S7).Cell viability was assessed using the CellTiter-Glo assay.The best of the pyridone-linked dimers, WJ715, was more than 50 times more potent than JQ1 (IC 50 of 0.9 nM for WJ715 vs. 50.5 nM for JQ1) though slightly less potent than MT1 (IC 50 = 0.3 nM) (Figure 4B).The chemical structure of WJ715 is shown in Figure 4C.These data demonstrate that the pyridone-forming dimerization is a convenient route to produce highly potent bivalent BRD4 inhibitors.

Conclusions
In summary, we have developed a facile homodimerization strategy for the synthesis of bivalent protein ligands.Masked βformyl acetamide can dimerize efficiently to form pyridone products in excellent yield and purity when treated with TFA at room temperature.The reaction tolerates a variety of functional groups.We applied this method to generate dimeric degraders of the long isoform of the VHL E3 Ubiquitin ligase and bivalent inhibitors of BRD4, both of which showed excellent activity in living cells.This protocol reduces the synthetic burden of creating homodimeric ligands relative to most other strategies, such as the copper-catalyzed cycloaddition of azides and alkynes, since only a single precursor is required.

Figure 2 .
Figure 2. Functional group compatibility of the dimerization reaction.*Note the starting materials for WJ901, WJ899 are tert-butyloxycarbonyl (Boc) protected amines, Boc protection group was removed by TFA in the reaction process.

Figure 3 .
Figure 3. Synthesis and characterization of pyridone-linked VHL degraders.A) General scheme to generate HOMO bivalent VHL degrader candidates by dimerization of VH 032 linked masked β-formyl amides.B,C) Western blot analysis of the level of pVHL30 by adding 1 μM indicated compound to HeLa cells and incubated for 12 h.D) Chemical structure of WJ662.E,F) Western blot analysis of the level of pVHL30 by adding indicated concentration of WJ662 to HeLa cells and incubated for 12 h.G,H) Western blot analysis of the level of pVHL30 by adding 10 μM of WJ662 to HeLa cells and incubated for the time indicated.I,J) Western blot analysis of the level of pVHL30 in HeLa cells treated with 10 μM WJ662 for 12 h with and without 10 μM proteasome inhibitor bortezomib.DMSO and CM11 (1 μM) were used as vehicle and positive controls.The amount of pVHL30 (the strongest band) present normalized to the vehicle control (DMSO) is indicated as "%".All of the experiments were performed in triplicate.The Western blots shown are typical of all the results obtained.

Figure 4 .
Figure 4. Pyridone-linked, bivalent BRD4 inhibitors.A) General scheme to generate bivalent BRD4 inhibitors.B) Effect of the BRD4 inhibitors on the viability of MV-4-11 cells.The data shown are from six technical replicates.IC50 quantification, the experiments were performed in six repeats.C) Structure of WJ715, the most potent of the pyridone-linked JQ1 dimers (IC50 = 0.9 nM).