Molecular glues to stabilise protein–protein interactions

Targeting protein – protein interactions (PPIs) has become a common approach to tackle various diseases whose pathobi-ology is driven by their mis-regulation in important signalling pathways. Modulating PPIs has tremendous untapped therapeutic potential and different approaches can be used to modulate PPIs. Initially, therapeutic effects were mostly sought by inhibiting PPIs. However, by gaining insight in the mode of action of certain therapeutic compounds, it became clear that stabilising (i.e

biological processes in the cell such as signal transduction, expression and degradation of other macromolecules, transport, motility, and catalysis [1]. In the last decades, a huge effort has been put into the sequencing of the protein-coding genes that compose the human genome, settling on the order of magnitude of 20,000e25,000 genes. The proteome, however, is far more complex considering the fact that one gene can encode more than one protein and often undergo substantial post-translational modifications, an observation that resulted in the introduction of the concept of proteoforms e distinct, but related version of a protein per coded gene [2]. A recent study suggests that the number of proteoforms is at least a factor of 17 higher than that of the protein-coding genes [3]. The protein interactome has been estimated to consist of up to 650,000 protein interactions [4].

The PPI modulation challenge
From a drug discovery perspective, the involvement of PPIs in literally every disease make them attractive therapeutic targets [5]. However, PPIs have always been considered challenging compared to more conventional targets like enzymes and receptors [6]. Clearly defined features such as pockets and grooves that regulate and modulate the function of enzymes and receptors are typically lacking the flatter and broader interaction surfaces formed by most PPIs. Most often, there is also no starting point provided by a natural ligand, as is often the case for an enzyme or receptor [7,8].
Despite these challenges, the last two decades have witnessed successful stories unfold, bringing small molecules from their early discovery stage to the most advanced clinical trials [8,9]. Examples of such stories can be found within the Bcl-2 family of proteins where the PPI inhibitor venetoclax/ABT-199 has been approved by the Food and Drug Administration (FDA) for the treatment of chronic lymphocytic leukemia (CLL) [10].
Inhibition was the first approach to PPI modulation that has been addressed systematically and, to some extent, is the most intuitive. More recently, PPI stabilisation, has gained increased attention by the scientific community. Two key advantages over the more classical inhibitory approach make PPI stabilisation a promising approach. First, stabilisers can in theory afford to be less potent compared to inhibitors as they do not have to displace any natural binder to achieve function, but rather enhance an already existing interaction. Secondly, stabilisers often bind to a transient, unique interface formed by two (or more) interacting proteins that contribute to a selective pocket that would minimise off-target effects [11,12]. A study performed on protein crystal structures available in the Protein Data Bank (PDB) revealed some similarity between the cavities formed by PPIs and the ones present on more conventional targets suggesting the feasibility of the orthosteric PPI stabilisation approach [13]. Moreover, it would suggest compound libraries that historically have been designed with more conventional targets in mind could still provide starting points for PPI stabilisation.

Molecular glues and protein degraders: Rising stars in PPI modulation
Targeting PPI by means of molecular glues can unlock the so far considered undruggable pool of scaffolding/ adaptor proteins and transcriptional factors, that regulate many cellular processes [14,15]. From our perspective, the term 'molecular glues' encompasses both PPI stabilisers that require the already formed binary PPI complex to bind AND chemical inducers of dimerisation (CID), which bring together proteins irrespective of their endogenous interactions and already bind with high affinity to one of the two partner proteins. Among the first compounds identified as molecular glues are cyclosporin A (CsA), rapamycin, and FK506 ( Figure 1) [16]. These compounds all possess immunosuppressive activity acting as molecular glues and inhibiting T cell activation. CsA binds to cyclophilin and calcineurin, leading to inactivation of NFAT and a subsequent decrease in the immune response [17]. FK506 acts differently from CsA but with a similar outcome, by forming a ternary complex with the proteins FKBP12 and calcineurin inhibiting the enzymatic phosphatase activity of calcineurin [17]. Finally, rapamycin glues the two proteins FKBP12 and mTOR, inhibiting mTOR kinase activity ( Figure 1) [18].
Targeted protein degradation by means of either monovalent or bivalent molecules is a growing strategy in drug discovery. If compared to the more traditional inhibitors of protein function, protein degraders require only a transient interaction event for function in order to target the protein for degradation [19]. They are also free from the stoichiometric occupancy rule since they can be recycled immediately after the target has been degraded. Since they do not have to bind to functional sites of the protein, small-molecule degraders can target proteins so far considered undruggable and deliver effects that mimic the phenotype of a genetic knockout.

Monovalent protein degraders
Direct protein degradation has been achieved by selective estrogen receptor degraders, SERDs, and selective androgen receptor degraders, SARDs. The progenitor of these small molecules is the FDA approved drug fulvestrant used in breast cancer treatment (Figure 1) [20]. A considerable amount of effort has been put into the development of both SARDs and SERDs of either steroidal and non-steroidal derivation which have been extensively reviewed [21]. Another example of a monovalent degrader is ASTX660, a BIR3binding domain IAP (inhibitor of apoptosis protein) antagonist, which destabilises and induces cIAP degradation. This increases dimerisation and autoubiquitylation of the latter, restoring the apoptosis pathway ( Figure 1) [22]. Bcl6 is a transcription factor of which dysregulation leads to conditions such as diffuse large B cell lymphoma (DLBCL). Small molecules like BI-3802 have been shown to induce Bcl6 degradation, resulting in an inhibition of its action in the disease course ( Figure 1) [23].
Monovalent small molecules acting as molecular glues as well as a protein degraders are also known and are represented by the immunomodulatory imide drugs (IMiDs) and by the splicing inhibitor sulphonamides (SPLAMs) class of molecules [21]. Thalidomide, lenalidomide (Figure 1), and pomalidomide (IMiDs) have all been identified as cereblon (CRBN) binders, which are the substrate receptors of an E3 ligase complex known as CRL4CRBN [24]. IMiDs can recruit non-native substrates for the CRL4CRBN complex by acting as molecular glues, bridging interactions between CRBN and neosubstrates. Investigations on the mechanisms of action of these molecules combined with the elucidation of crystal structures in which these molecules have been co-crystallised, suggests that IMiDs induce conformational changes on CRBN making it capable of recruiting different substrates [25]. Interestingly, these conformational changes appear to be IMiDs-structure dependent, revealing a huge potential for this class of molecules. IMiDs have been associated with the degradation of CRBN neosubtrates such as the lymphoid transcription factors Ikaros and Aiolos (IKZF1 and IKZF3) resulting in IL-2 upregulation in T cells [26]. IMiDs have also been shown to induce degradation of casein kinase 1a [24]. Further examples that highlight that molecular glue degraders are more common than previously thought are compounds that degrade CDK12 [27] and cyclin K [28]. Among the most current generation of IMiDs that has been developed we find avadomide (CC-122) (Figure 1) that is currently being tested in clinical trials for DLBCL treatment [29].
The last class of protein degraders with a small molecule molecular glue mode of action are the splicing inhibitor sulphonamides (SPLAMs). Indisulam (Figure 1) has been reported to mediate the recruitment of the RBM39 splicing factor to the CUL4-DCAF15 E3 ligase protein complex causing its degradation [30].

Bivalent protein degraders
Bivalent protein degraders normally consist of two distinct chemical moieties required for the molecule to function. The bifunctional aspect of these tools generally provides a modular action that can be redirected on multiple targets, conferring a general applicability [14].
The first examples of bivalent protein degraders exploit the natural recognition of hydrophobicity on proteins of interest to target them for degradation. Hydrophobic tagging can be achieved by fusing small molecules with adamantane or Boc3Arg chemical moieties (Figure 1.3). For instance, an adamantyl moiety has been linked to a ligand of the kinase HER3, a therapeutic cancer target. The resulting bivalent molecule TX2-121-1 has been shown to induce HER3 degradation via the proteasome pathway [31].
The second and certainly one of the most groundbreaking technologies of the last decade is represented by the proteolysis targeting chimeras class of molecules (PROTACs). PROTACs are a class of bifunctional molecules designed and engineered for binding to an E3 ligase neo-substrate and to an E3 ligase protein complex. The proximity induced by the PROTAC agent causes the neo-substrate to be polyubiquitinated and therefore targeted for degradation [32].

14-3-3 PPI modulation
A class of proteins on which extensive small-molecule modulation work has been performed in the last decades, are the regulatory family of proteins collectively referred to as 14-3-3 14-3-3 proteins are one of the most abundant proteins in eukaryotic cells and act as hubs contributing to signal transduction by facilitating the formation of transient signalling protein complexes [33]. The 14-3-3 interactome is estimated to be around 500 binding partners [34]. Being such a ubiquitously expressed class of molecules, 14-3-3 proteins are present in most of human tissues and have been implicated in many human diseases [35].

14-3-3 molecular glues
The stabilisation of 14-3-3 PPIs has been the focus of a number of academic early stage drug discovery projects in the last decade. The inspiration came from nature in the form of the wilt-inducing toxin fusicoccin-A (FC-A), which was identified as a binder of the protein complex composed of 14-3-3 and the regulatory domain of the plasma membrane H þ -ATPase (PMA) [36]. The disclosure of the FC-A mechanism allowed it to be used as a valuable tool to stabilise other 14-3-3 PPIs, also in human cells. One example is the stabilisation of the 14-3-3/ERa complex, which results in the inhibition of ERa transcriptional activity [37]. Another example is the 14-3-3/CFTR interaction for which stabilisation results in increased CFTR trafficking to the cell membrane [38]. A structurally related compound to FC-A is cotylenin-A (CN-A). It has been reported to possess anti-tumour activity in a few human cancers (Figure 1) [39,40]. 14-3-3 has been reported to be a cofactor for Raf kinase activity and a published structure of CN-A bound to the 14-3-3/Raf complex provided useful insight as to how CN-A can provide antitumor activity [40].
Non-natural compound stabilisers of 14-3-3 PPIs were also discovered. The main example is pyrrolidone1, which was identified by an HTS conducted on the 14-3-3/PMA2 complex using a surface-based format [41,42]. Another example is the adenosine monophosphate molecule AMP that has been reported to directly stabilise the 14-3-3/ChREBP PPI [43] and inspired the use of phosphonates as stabilisers of this PPI [44]. Routes to develop more potent and specific versions of 14-3-3 PPI stabilisers have been pursued and the field is rapidly moving forward [45]. In particular, fragment-based approaches for the identification of 14-3-3 PPI stabilisers have seen a considerable amount of progress in the last few years [46] [e] [48].

Conclusions
The field of small-molecule PPI stabilisation or molecular glues has seen a dramatic development in the past few years. Whereas at the beginning of the 2000s, PPI inhibition was the predominant strategy to modulate PPIs, the landscape has been expanded in appreciation of the potential therapeutic value of stabilising regulatory protein complexes. This development was undoubtedly been spurred by the recognition that molecules like lenalidomide, rapamycin, and FK6506 already used successfully in the clinic work as stabilizers of PPIs. A main driver for the excitement around PPI stabilisation is targeted protein degradation with the coming of age of both heterobifunctional and molecular glue degraders. As many of these molecules now enter the clinic, the expectation is that the concept of PPI stabilisation and induced proximity will significantly enlarge the druggable proteome and add to our armamentarium to treat diseases like cancer and neurodegeneration. However, despite this impressive progress, the de-novo development of a molecular glue is a tremendous challenge, not the least due to the complex biophysical interplay between the components of this three-body system.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.