Efficient development of stable and highly functionalised peptides targeting the CK2α/CK2β protein–protein interaction

This work describes the efficient development of functionalised, cell-permeable, and stable peptide inhibitors of the protein–protein interaction of CK2.


Introduction
Proteins are an essential component of the cell. They exert physio-pathological functions in response to external and internal messengers in a highly regulated and selective manner. Many of these functions are made possible by a complicated network of protein-protein interactions (PPIs). Targeting of PPIs is seen as an attractive therapeutic strategy considering the large number of PPIs involved in pathological mechanisms. In addition, targeting of PPIs provides orthogonality with respect to conventional therapeutics, which are typically targeted against existing binding sites for small molecules. Therefore, inhibition of PPIs may result in the development of safer drugs. 1 PPIs are characterised by shallow surfaces which make targeting them with conventional small molecules (<500 Da), whilst possible, 2-7 a non-trivial and lengthy process. Synthetic peptides, on the other hand, provide a valuable alternative to small molecules: their structural properties make them amenable to mimic portions of the native proteins resulting in favorable interactions with the large and shallow interfaces of PPIs. 8 In addition, when structural information pertaining to the PPI of interest is known, potent and selective peptides can be rapidly designed based upon the sequence of targeted motifs of the native proteins. 9 However, the poor pharmacokinetic (PK) properties of synthetic peptides in the body limit their use and interest to the pharmaceutical industry. 10 Many strategies have been adopted to overcome the PK limitations of peptides; yet, very few of them are set up to provide peptides that are highly functionalised and cell-permeable in an efficient manner. [11][12][13][14][15] Taking into account the potency and selectivity of peptides, methodologies that introduce functionalities to make them simultaneously stable and cell-permeable would yield chemical probes of invaluable importance. 16 To this end, we have developed a two-component (2C) copper catalyzed azido alkyne (CuAAC) peptide stapling (PS) methodology that allows the development of highly functionalised, potent, and selective peptides targeting intracellular PPIs. 17 In 2C-CuAAC-PS only a limited number of peptides are synthesised and their design is based upon the structural information available for the PPI of interest. 18 Subsequently, peptide stapling is carried out by using amino acids with azido side chains and a di-alkyne linker as a constraint. One of the advantages of this methodology is that the linker can be modied and functionalised to improve the in vitro pharmacological properties independently from the peptide sequence: the result is a more efficient optimisation of the peptides. Biophysical and cellular assays are successively used to assess the peptides synthesised. Importantly, the 2C-CuAAC-PS chemistry proved to be compatible with all the natural amino acids, led to enhanced binding affinities for both helical and non-helical peptides and improved the overall in vitro pharmacological properties of the peptides, including stability to proteases. [19][20][21] Herein, we have applied this robust approach to efficiently develop the rst stable and highly functionalised conformationally-constrained peptide acting on the PPI of CK2.
CK2 is a protein kinase overexpressed in cancer cells and a validated oncology target; CX4945, a traditional small molecule ATP-binding site inhibitor of CK2, is currently undergoing clinical studies. 22 However, CX4945 targets the ATP-binding site, which is well conserved among the kinome. More recently, there have been increasing efforts to develop non-ATP competitive inhibitors of CK2 to reduce the off-target effects of competitive ligands. [23][24][25] Among the strategies designed to target CK2 outside its orthosteric binding site, is the inhibition of the PPI between the a and the b subunits. [26][27][28][29] Disruption of the holoenzyme assembly affects the function of CK2 by preventing phosphorylation of b-dependent substrates, the shuttling of the protein between different intracellular compartments, and by reducing the stability of the catalytic a subunit (Fig. 1). [30][31][32][33] With the exception of the Phe pocket, the CK2a/b interface is a shallow and hydrophobic surface; consequently, peptides are an ideal class of molecule to target this PPI. To this end, two cyclic peptides have been developed. However, one of these, Pc, 26,28 is a disuldelinked cyclic peptide that lacks cell permeability and stability in the reducing intracellular environment, and the other, TAT-Pc, 34 has not been assessed structurally or for stability in physiologic uids (Fig. 1). Therefore, a stable chemical probe that could be used in vitro and in vivo to study the interface of the important protein CK2 is still required.
Starting from the sequences of CK2b and Pc, we investigated alternative ways of constraining the peptide into its bioactive conformation using a stable linkage compatible with the 2C-CuAAC-PS chemistry. At a later stage, X-ray crystallography guided our investigation on sequence variation to increase the binding affinity of the peptide for CK2a. The most promising peptide was easily modied into a uorescent, cell-permeable probe via a novel highly functionalised constraint that allowed us to study the peptide's activity in cancer cells.
The peptide developed in this work is the rst stable, cell permeable macrocyclic peptide that disrupts the CK2a/b PPI in vitro and leads to cancer cell death and arrest of the cell cycle; as such, it will serve as a useful chemical probe in oncology. Furthermore, the structure of the peptide in complex with CK2a will act as a valuable starting point to develop novel CK2 inhibitors.

Results and discussion
In order to design stable peptides targeting the CK2a/b interaction, we used a rational-design approach based on the valuable crystal structures of CK2b and the disulde bridged Pc peptide. 34 Disulde bridges are unstable under reducing environments; therefore, our aim was to replace the labile disulde group with a stable constraint. To this end, the 2C-CuAAC macrocyclisation technique was chosen for its validated ability to constrain peptides in their binding conformation, simultaneously enhance the stability against proteolytic cleavage, introduce functionalities (cell-penetrating peptide (CPP), uorescent dyes, biotin, and PEG and other tags), and improve the poor stability in physiological uids and cell-penetration in a combinatorial manner. 19,20,[35][36][37][38] Rational design of conformationally constrained peptides mimicking CK2b Molecular modelling identied Cys2 and Gly11 of Pc, 34 corresponding to P185 and P194 of CK2b, as suitable residues to staple: they make negligible contributions to the binding and are positioned at a suitable distance from each other to accommodate a 2C-CuAAC staple (ESI, Fig. S1 †). To cyclise the peptide, azido amino acids bearing one-carbon-atom side chains (Fmoc-Aza-OH) were used in combination with aliphatic linkers of different lengths as proposed by molecular modelling ( Fig. 2 and ESI Table S1 †). The ability of the synthesised macrocycles to disrupt the a/b PPI of CK2 was then tested in a preliminary uorescent polarisation assay (FP) (Fig. 2).
Peptides constrained with linkers C2 and C3 proved to be insoluble under the assay conditions. To overcome this limitation, it was decided to incorporate a functional handle on the constraint to simultaneously improve the solubility and allow for future functionalisation of the peptides. Peptide P1-C4 was the most effective macrocyclised peptide at displacing the FP probe from the CK2a subunit (93 AE 1% inhibition at 15 mM). This suggested that constraint C4 was optimal at constraining the peptide in its binding conformation. The binding affinity of P1-C4 and P1-C5 for the CK2a subunit was measured via isothermal titration calorimetry (ITC). P1-C4 showed the highest affinity with a K d of 460 nM, showing a modest increase compared to Pc (1000 nM, ESI Table S3 †).
To understand how P1-C4 was able to bind CK2a, we determined its crystal structure in complex with CK2a (PDB: 6Q38). The crystal structure showed the peptide binding in a conformation that overlays well with CK2b and Pc (Fig. 3a-c). The backbone residues of P1-C4 are all slightly shied compared to Pc; however, these differences are greatest closer to the constraint that holds the two ends of the peptide further apart than the disulde linker does; consequently, the terminal residues adopt entirely different conformations (Fig. 3b and c).
The X-ray crystal structure of P1-C4 (Fig. 3d) shows Ile9 oriented in the right direction to be replaced by an amino acid capable of p-p stacking with the benzene ring of the C4 constraint. Similarly, chlorobenzene probes in ligand-mapping MD simulations 39,40 of P1-C4 indicated a region of high occupancy by the aromatic carbon atoms of chlorobenzene around Ile9 of the peptide (ESI Table S1 †). Therefore, the sequence of P1 was modied to replace Ile9 with a larger, non-polar Trp to create peptide P2 (Ac-GXRLYGFKWHXGG-NH 2 where X ¼ Fmoc-Aza-OH). The resulting peptide was macrocyclised using C4 as a linker to afford P2-C4.
The binding affinity of the P2-C4 peptide was found to be enhanced (K d 150 nM) compared to both P1-C4 (K d 460 nM) and Pc peptide (K d 1000 nM) ( Fig. 4a and ESI Table S3 †). The constrained P2-C4 peptide showed also a 300-fold improvement compared to the linear variant P2 (K d 44 mM). The crystal structure of P2-C4 bound to CK2a (PDB: 6Q4Q, Fig. 3e and f) shows the p-p stacking between the Trp and the phenyl ring of the constraint. It is likely that this p-p stacking is the main factor leading to the higher affinity of P2-C4 as no other signicant interactions were observed (ESI Fig. S2 †). Therefore, the increased binding affinity of P2-C4 for CK2a may be explained with a reduced entropic penalty upon binding due to the rigidifying interaction occurring between the constraint and the Trp residue. [41][42][43] Inhibition of the CK2a/b protein-protein interaction We then wanted to investigate whether the peptides would be able to inhibit the binding of the b subunit to the a subunit. To this end, the binding affinity of the regulatory b subunit for the catalytic domain was determined via ITC (K d 9 nM). This was then repeated in the presence of 100 mM P1-C4 and P2-C4 (ESI ,  Table S4 †). This preliminary assay showed that no binding of CK2b to CK2a was detected in the presence of either of the peptides: this result conrms that the peptides bind at the interface and prevent CK2b binding despite its high binding affinity for CK2a. As an orthogonal proof of inhibition, we used competition Bio-Layer Interferometry (BLI) experiments to demonstrate that P2-C4 was able to prevent the formation of the  (Table S2 †). Pc, P1, and P1-Cn (where n ¼ 1-5) peptides feature an amide at the C-terminus and an acetyl cap at the N-terminus. All the amino acids are the L isomers.
As one of the key roles of CK2b is to recruit substrates to the kinase, the effect of PPI inhibition on substrate phosphorylation was studied using CK2b dependent and independent substrates. It was shown that P2-C4 was able to inhibit the phosphorylation of a CK2b-dependent substratethe transcription factor eIF2bwith an IC 50 of 206 AE 29 nM (Fig. 4c). It should be noted that the kinase assay was performed using the pre-formed CK2a/b complex as the ability of CK2a to phosphorylate eIF2b requires the presence of the CK2b subunit. The assay indicates that P2-C4 can disrupt, in a dose-dependent manner, the CK2 holoenzyme thereby reducing the ability of  CK2a to phosphorylate CK2b-dependent substrates. As expected, P2-C4 did not affect the phosphorylation of a b-independent substrate peptide (RRRADDSDDDD) meaning that binding at the interface site does not displace the ATP or signicantly alter the kinase activity allosterically (ESI Fig. S3 †).

Development of a multi-functional constraint
To overcome the limitations of Pc, the main goal of this work was to efficiently develop chemical probes for the CK2 PPI that could be used in cells and ultimately in vivo. Therefore, cellpermeability and stability in serum were required. Peptide macrocyclisation has been described on several occasions as a powerful technique to enhance the stability of the peptides to proteases. 20,21,45 Although the ability of peptide macrocyclisation alone to enhance cell-permeability is highly debated, cellpenetrating peptides (CPPs) are well-established and are oen added to the cyclised peptides to gain cytosolic entry. 46,47 Preliminary cellular uptake experiments carried out with FITClabelled peptide showed that the peptide was not able to permeate the cell membrane of osteosarcoma cancer cells (U2OS) and thus, functionalisation was needed. Considering that both cell-penetrating motifs and uorescent tags were necessary for the cellular assays, we decided to develop a novel multi-functional constraint that would simultaneously: constrain the peptide in its binding conformation, enhance the stability to proteases, provide cell-permeability to the CK2 peptide, and act as a uorophore (F2C4, Fig. 5). In addition, we wanted to develop functionalised constraints that could be synthesised in an automated manner using Fmoc-SPPS so to be accessible by the wider scientic community. Exploiting one of the advantages of 2C-PS, linker C4 was elaborated independently from the CK2 peptide (ESI Scheme S1 †). The benzoic acid derivative linker C4 (pink in Fig. 5) was attached to a CPP via a spacer (blue in Fig. 5) to avoid steric clashes with the CK2 peptide and the CK2a domain, generating the full multifunctional linker F1C4. Previously, we have reported the use of an (L)-arginine tripeptide as an effective CPP to carry peptide cargos into cells. 20,36 We decided to use (D)-arginine instead to confer cell-permeability to the CK2 peptide and simultaneously provide a proteolysis resistant alternative. The CPP was in turn attached to the uorescent tag FITC via an orthogonally protected Lys to monitor the peptide entrance into the cells (full multi-functional linker F2C4, Fig. 5). A uorescent constraint without the CPP motif (F3C4) was also synthesised. The functionalised linker F1C4 was then reacted with P2 to obtain CAM7117 (Fig. 5). Importantly, CAM7117 displayed signicant stability in human serum (47% intact peptide aer 24 hours incubation, Fig. S4 †) highlighting how the covalent constraint F1C4 provided a peptide that is stable under physiological uids.
Activity of CAM7117 in cancer cells CAM7117 was successfully internalised by the cancer cells as observed by confocal microscopy of its FITC-labelled analogue, P2-F2C4 (Fig. 6a). Once internalised, the peptide was able to inhibit human osteosarcoma (U2OS) cell growth with a GI 50 of 32 AE 2 mM aer 4 day incubation, and induced apoptosis aer just 4 hours (ESI Fig. S5-S7 †). A marginally reduced biological effect was observed for both CAM7117 and the clinical candidate CX4945 when human colorectal cancer cells (HCT116) were used (Fig. 6b). No biological effects were observed when the functionalised constraint only (F1C4) or the negative control peptide (P3-F1C4) were used. The negative control P3-F1C4 features an F7W mutation in the sequence and did not show binding to CK2a by ITC (ESI Table S3, Fig. S5-S7 †).
Although specic biological activity was observed in cancer cells, the drop-off between the enzymatic and cellular assays was higher than expected. Therefore, we investigated the effect of intracellular localisation of the FITC-labelled CAM7117, and imaging experiments were carried out to look at co-localisation with several organelle markers. We were unable to detect any colocalisation with endosomal markers (Fig. 6c, ESI Fig. S8 and S9 †), but partial co-localisation was detected with lysosomal marker, and a signicant amount of the peptide was found to localise to the same place as the Golgi and endoplasmic reticulum (ER) markers (Fig. 6c, ESI Fig. S8 and S9 †). A signicant proportion of the peptide was able to diffuse to the cytosol and localised mainly in the nucleus (Fig. 6c).
Therefore, the drop-off in cellular activity may be attributed to trapping of the peptide in the Golgi/ER. Moreover, the activity drop-off in cells could also be due to the mechanism of inhibition: unlike the inhibition of the a catalytic domain (such as CX4945), displacing the b subunit does not inhibit the phosphorylation of all CK2 substrates. Structures of the multi-functionalised peptides. The peptides are constrained with a multifunctional linker containing a linkage core that locks the peptide in its binding conformation and enhances stability to proteases (pink), a spacer to avoid steric clashes (blue), a protease-resistant poly(D)arginine tag to gain cellular permeability (green) and a fluorescent tag to monitor intracellular localisation (yellow).
Further imaging experiments using U2OS cells were carried out to evaluate the effect of CAM7117 in subcellular localisation of the CK2 subunits.
Already aer 15 minute incubation, little or no uorescence associated with the CK2b antibody was found in the nucleus; on the contrary, untreated cells showed signicant punctate staining nuclear accumulation of the CK2b (Fig. 6d). This evidence suggests that CAM7117 could engage with CK2a and compete with CK2b in a cellular context.

Conclusions
In the present study, we have showcased an approach to efficiently develop peptides that are highly functionalised, cellpermeable and stable in serum. In particular, using the 2C-CuAAC-PS methodology, we have developed conformationally constrained peptides that act as inhibitors of the CK2 a/b PPI in vitro and in cancer cells. The lead peptide, CAM7117, presents an enhanced binding affinity for CK2a with respect to the previously developed Pc and, most importantly, is stable under conditions mimicking physiological uids. Owing to the lack of intrinsic cell-permeability of the peptides, we developed an easily-synthesised multi-functional constraint that allowed us to investigate the intracellular activity of CAM7117, which arrests cancer cell proliferation and induces apoptosis in a dose-dependent manner. To the best of our knowledge, this study describes the development of the rst inhibitory peptide of the CK2a/b PPI that is: stable in serum, cell permeable, active in cells, able to engage the target and structurally characterised. Such a peptide would act as a chemical probe that enables the study of the CK2 PPI using endogenous levels of proteins and could, therefore, be used to elucidate CK2 dependent mechanisms leading to cancer progression.
Remarkably, the multi-functional constraint developed in this work could be used to lock other peptides into their binding conformation and simultaneously functionalise them. Therefore, the strategy adopted herein could, provide a more universal approach to develop modulators of PPIs for many targets where linear sequence epitope provides the majority of the binding energy. Moreover, considering that the peptide addition to either cells or organisms can be done with rigorous temporal and quantitative control, these probes are extremely powerful tools for validating PPIs in drug discovery and dissecting biological processes.

Conflicts of interest
There are no conicts to declare.  Table S11. † Nuclei are stained with Hoechst 33342 (blue). (d) Change in intracellular localisation of CK2b (red) following treatment with CAM7117 (30 mM) for 15 and 120 minutes. Nuclei are stained with Hoechst 33342 (blue).