On-demand modular assembly for expedited PROTAC development

KEYWORDS:

• PROTAC
• click chemistry
• library synthesis
• parallel synthesis
• direct-to-biology
• split linker
• modular assembly

1. Introduction

The concepts of ‘proteolysis targeting chimera (PROTAC)’ [1] and ‘click chemistry’ [2] emerged at nearly the same time in the early 2000s. Both ideas have since been widely embraced by the fields of chemical biology and drug discovery, and they have converged: the bifunctional nature of PROTACs calls for connecting modules into a cohesive structure, and click chemistry provides one way to assemble a PROTAC molecule. In recent years, ‘on-demand modular assembly’ in PROTAC discovery and development has become an intriguing subject. Comprehensive overviews of broad case studies are available in previous reviews [3,4]. Here, we briefly discuss two applications of on-demand PROTAC assembly: rapid synthesis of a PROTAC library for screening, and in-cell PROTAC self-assembly to enhance cellular drug exposure.

Amgen was among the first to apply click chemistry for parallel PROTAC synthesis, utilizing a copper-catalyzed reaction that couples an azide and an alkyne to regio-selectively form a new triazole ring at 100 µmol scale [5]. Click ligation, akin to Lego bricks or the mortise-and-tenon joints of traditional Chinese architecture, is renowned for its reliability in mild reaction condition, broad scope, high yields, and stable products. The reaction allows the rapid assembly of PROTACs with varying lengths of linkers for screening, which would otherwise require increased synthetic efforts. More recently, the Jin and Wei groups showcased the feasibility of utilizing Bertozzi’s copper-free strain-promoted azide-alkyne cycloaddition reaction (SPAAC, colloquially referred to as ‘copper-free click chemistry’) to produce DNA oligonucleotide-based PROTACs [6]. Despite the straightforward nature of the click reaction, both studies included a product purification step, a significant speed bottleneck in the ‘synthesis-purification-sample handling-screening’ cycle, underscoring the need for devising a ‘direct-to-biology’ (D2B) paradigm. With respect to preparing non-purified PROTACs, the Tang group set to explore alternative high yielding condensation chemistry where one end of the PROTAC, containing a preselected functional group 1 (FG1), efficiently couples to the other end of the PROTAC, which bears a preassembled FG2 with various linkers, in a 1:1 reactant mixture [7–9]. The FG1/FG2 condensation pair can consist of hydrazide/aldehyde, resulting in an acylhydrazone linkage, or of ortho-phthalaldehyde/amine, resulting in a phthalimidine linkage. A notable aspect of these reactions is that they can be conducted in DMSO solution within individual wells of a plate at 1 µmol scale, producing a PROTAC product of high purity along with water as a sole side product, thereby facilitating the direct-to-assay use. However, a laborious aspect of all the aforementioned work, and of work seeking to apply oxime formation chemistry [10], lies in the necessity to pre-install specific reactive functionalities in the linker. This requirement may hinder the creation of linkers with a wide range of structures due to the unavailability of certain reactive handle building blocks. To address this, a Janssen research team embarked on assembling a library of 91 PROTACs through a three-step sequential process involving amide coupling, deprotection, and amide coupling at 5 µmol scale. In this approach, readily available N-Boc diamine building blocks function as diamide linkers. Notably, any unreacted reactants during the two amide coupling steps were efficiently removed using resin-bound scavengers, thus enabling D2B use [11]. This streamlined approach facilitates rapid investigation into ‘linkerology’ for PROTAC development, a process that is predominantly empirical. To further achieve miniaturization and automation in the library synthesis of PROTACs, an AstraZeneca research team developed a cost-effective automated synthesis model system, conducting each reaction in a 384-well plate in DMSO at 120 nM scale. This approach, also known as ultra-high-throughput experimentation (uHTE), was initially reported by Merck [12]. The AstraZeneca multi-parallel synthesis was accomplished using the workhorse amide coupling reaction. In this approach, a collection of 34 custom-made E3 ubiquitin ligase ligands, connected to free amine linkers, served as diverse reactive handles for conjugation with two protein-of-interest ligands pre-modified with a reactive free acid FG. After automatically quantifying the conversion of all 68 products and handling the samples, each reaction crude was subjected to downstream D2B use [13]. Owing to the miniaturization capability, the material consumption of the precious custom-made intermediates can be greatly reduced. Additionally, the automation feature allows the synthesis timeline to be compressed from weeks to days. Concurrent with this effort, a GSK research team established a similar uHTE platform for synthesizing 650 PROTACs in a 1536-well plate, employing amide coupling for assembly at 150 nM scale. As per their assertion, a single scientist can complete both the synthesis and subsequent D2B evaluation in less than one month, significantly accelerating the optimization of a PROTAC hit. Their data quality hinges on the discovery of a combination of the amide coupling reagent EDC, the additive OxymaPure, and the base N-methylmorpholine. These reagents not only mitigate cytotoxicity but also suppress the formation of undesired byproducts [14]. Taken together, the marriage of a rapid PROTAC synthesis platform with a D2B screening paradigm of reaction mixtures showcases the exceptional efficiency that is possible in PROTAC evaluation and optimization, marking a milestone in the widespread adoption of the on-demand PROTAC assembly principle.

A second scenario of adopting the on-demand PROTAC assembly principle is the self-assembly of PROTACs within cells, coined by Heightman, Kodadek, and Barany as ‘CLIPTACs’ (click-formed proteolysis targeting chimeras), ‘split PROTACs,’ ‘SAPTACs’ (self-assembled proteolysis targeting chimeras), or ‘CURE-PROs’ (combinatorial ubiquitination real-time proteolysis). In developing large molecular-weight PROTAC oral drugs with high polar surface area (PSA), overcoming the bioavailability and dosage conundrum poses a formidable goal for pharmaceutical industry. The inherent unfavorable physicochemical properties associated with PROTACs can restrict cellular permeability and impact pharmacokinetics, while the molecular chamelonicity, involving conformational dynamics that affect cell permeability and aqueous solubility, adds further complexity to this issue [15,16]. Consequently, an Astex Pharmaceuticals research team pioneered the development of in-cell CLIPTACs by bioorthogonally combining two smaller precursors: one tagged with a reactive tetrazine handle and the other with its corresponding paired handle, trans-cyclo-octene [17]. Bioorthogonal chemistry is subject to additional constraints compared to click chemistry outside cells, as reactions need to proceed rapidly in water without disrupting normal cellular chemistry. In the Astex model system for BRD4 or ERK1/2 degradation, the two smaller click precursors must be added sequentially to prevent ligation outside cells. They confirmed that the observed degradation of target proteins results from the in-cell formation of CLIPTACs, although it involves relatively slow degradation kinetics and requires concomitant high dosages for each piece. To address the pitfall of stable PROTAC assembly outside cells, the Barany and Kodadek groups turned to a reversible covalent chemistry involving the reactive handle pair phenylboronic acid and catechol [18–20]. Judging from the design, an additional advantage of this strategy is the potential avoidance of a hook effect (a ‘bell-shaped concentration dependence of activity’ signature commonly observed with PROTACs), as the K_D value of this type of reaction is approximately 1 mM [21], much weaker than the K_D value of typical ligand-protein interactions. In their model system for pVHL30 degradation in HeLa cells, a half-maximal degradation concentration (DC₅₀) value of 8.2 µM was achieved after 12-hour incubation, with pieces mixed in a 1:1 stoichiometry. A similar DC₅₀ value was observed with an alternative reversible covalent chemistry involving oxime formation between o-acetyl phenylboronic acid and alkoxyamine [19], which has a K_D value of approximately 10 µM [22,23]. The lack of improvement in permeability of each piece compared to the canonical pVHL30 PROTAC CM11 [24] may be the primary reason for these outcomes. In a separate study, the Barany group extended the ‘CURE-PRO’ approach for BRD4 degradation [20]. In this case, a DC₅₀ value of 358 nM was observed after pre-treating MCF7 cells with an optimal CURE-PRO pair for 24 hours. Considering the high degradability of BRD4, there remains space for improving the degradation activity and kinetics of CURE-PROs. A commendable aspect of these CURE-PROs is that each monomer need not be supplied in a strict 1:1 stoichiometry. In this case, the ratio can range from 3:1 to 1:3 without significantly compromising BRD4 degradation activity, providing a viable basis for further investigation in pharmacokinetics and pharmacodynamics. Taken together, the reversible self-assembly of PROTACs within cells introduces an innovative approach to overcome certain limitations associated with canonical PROTACs. However, it remains crucial to carefully select the right chemistry for reversible self-assembly and refine the cell permeability and in vivo stability of reactive handles.

2. Expert opinion

On-demand PROTAC assembly fosters the discovery and development of PROTACs across multiple dimensions. When applied to combinatorial parallel PROTAC synthesis for D2B, this approach can substantially save resources in synthesis and purification, and deliver biological data in short turnaround times. The ease of adaptation to miniaturization, automation, and utilization in diverse assays from a single well of plates contributes to a more efficient workflow. More importantly, such an integrated platform enables rapid exploration of linkerology, including linker vector, length, rigidity, and polarity, surpassing empirical iterative exploration. The throughput is beneficial not only for PROTAC hit finding but also for hit-to-lead optimization. Judging from the published results, extensive in-depth validation studies have confirmed the reliability of the D2B workflow. To ensure high-quality assay datasets, a robust linking chemistry with broad substrate tolerance is essential, ideally featuring quantitative yield, easy cleaning procedures for unconverted reactants (e.g., resin-bound scavengers), non-cytotoxic reagents, and complete elimination of assay-interfering byproduct formation. However, as can be seen in practice, click chemistry or amide coupling in this context may not consistently provide high yields for every designed reaction pair. Lastly, the reliable datasets generated by this platform can be utilized to train machine learning models for predictive analysis. For instance, training datasets can be built from a comprehensive analysis of pDC₅₀ and maximum achievable degradation level (D_max) against descriptors such as relative target binding affinity (RBA) in live and permeabilized cells, LogD, PSA, radius of gyration, molecular weight, H-bond donor/acceptor counts of the linker, rotatable bonds of the linker, and topological diameter of the linker. This is expected to enable accurate predictive analysis, leading to a more efficient method for PROTAC linker optimization.

When applied to self-assembly of PROTACs within cells, the irreversible assembly of CLIPTACs enabled by bioorthogonal, catalyst-free, inverse electron demand Diels-Alder (IEDDA) reaction-type click chemistry [25,26] first showcased the feasibility and effectiveness of a split approach in driving targeted protein degradation, contrasting to the inactive non-split approach in certain cases. This practice also holds promise for addressing the conundrum of PROTAC exposure in the central nervous system. Still, there are several aspects that require careful consideration. First, the two click precursors must be dosed sequentially to prevent the formation of stable CLIPTACs outside cells. This introduces drug development challenges such as potential drug-drug interactions or unmatched pharmacokinetic properties. In this context, the successful advancement of the click-activated drug SQ3370 to the Phase 1/2a by the biotechnology company Shasqi [27–29] sets a precedent for CLIPTAC drug development. Second, the apparent cell permeability (regardless of transporter involvement) of each of the two click precursors must be optimized to a sufficiently high level. Third, achieving potent CLIPTACs also necessitates optimization of the linker. Last, the click reaction rate must be fine-tuned, especially in an in vivo setting. Each type of click reaction has its own reaction kinetics, e.g., the second-order reaction rate constant of IEDDA between a tetrazine and a strained alkene can range from 1 to 10⁴ L/mol·s [30]. A click rate that is too slow leads to unfavorable degradation kinetics, while a click rate that is too fast may result in undesired CLIPTAC formation before reaching the intended tissue. While the reversible assembly of CURE-PROs addresses the sequential dosing issue seen in CLIPTACs, it inherits all other issues present in CLIPTACs. In addition, as seen in protease inhibitor drugs [31,32], the potential electrophilic reactivity linked with boronic acid has to be considered, not to mention the unfavorable toxicophoric group catechol. Despite potential pitfalls, we believe that the self-assembly of PROTACs within cells could represent a unique direction in future PROTAC drug development, and we expect to see more of this practice in the future.

Declaration of interest

The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

The authors are supported by the Cancer Prevention and Research Institute of Texas through grants RR220012 and RR220039.