De novo macrocyclic peptides that specifically modulate Lys48-linked ubiquitin chains

Abstract

A promising approach in cancer therapy is to find ligands that directly bind ubiquitin (Ub) chains. However, finding molecules capable of tightly and specifically binding Ub chains is challenging given the range of Ub polymer lengths and linkages and their subtle structural differences. Here, we use total chemical synthesis of proteins to generate highly homogenous Ub chains for screening against trillion-member macrocyclic peptide libraries (RaPID system). De novo cyclic peptides were found that can bind tightly and specifically to K48-linked Ub chains, confirmed by NMR studies. These cyclic peptides protected K48-linked Ub chains from deubiquitinating enzymes and prevented proteasomal degradation of Ub-tagged proteins. The cyclic peptides could enter cells, inhibit growth and induce programmed cell death, opening new opportunities for therapeutic intervention. This highly synthetic approach, with both protein target generation and cyclic peptide discovery performed in vitro, will make other elaborate post-translationally modified targets accessible for drug discovery.

Graphical Abstract

Introduction

The initial step in ubiquitin (Ub) signaling is the modification of a protein substrate by Ub or polyUb chains^1,2. This is achieved by the collaborative action of three enzymes, the E1, E2 and E3 ligases³. In ubiquitination, an isopeptide bond is formed between the C-terminus of a Ub molecule and a lysine residue from the protein substrate. PolyUb chains are then made by repeated isopeptide or amide bond formation, linking any of the seven lysines, or the N-terminus, of the last Ub in the chain and to the C-terminus of a new Ub^4,5. Interestingly, Ub chains of different linkage types signal for different biological functions, likely due to their unique conformations which allows for distinct recognition by cellular proteins. The most studied polyUbs are the K48-linked chains, which are known to adopt a compact structure and label their protein substrate for proteasomal degradation^6,7. Ubiquitination is a reversible process and a family of enzymes known as deubiquitinases (DUBs) cleave Ub from polyUb chains, or remove polyUb entirely^5,8.

Ubiquitination is a major post-translational modification affecting many aspects of cell biology^9,10. Several tumor suppressors or oncogenes are involved in ubiquitination or deubiquitination pathways, making the Ub system an excellent target for cancer therapy¹¹. For example, the ubiquitin-proteasome system (UPS) has proven to be a valid target for anticancer drug development, as exemplified by the approval of the small molecule Bortezomib for the treatment of multiple myeloma and mantle cell lymphoma^12,13,14. Drugs such as Bortezomib interfere with protein degradation by inhibiting the UPS proteolysis machinery^15,16, whereas other strategies target different components of Ub signaling, such ubiquitination itself or the action of DUBs, with success yet to be translated to useful therapies^5,17.

Another promising approach is targeting the Ub chain itself. In this regard, the small molecules ‘ubistatins’ fortuitously emerged from a chemical genetic screen, and were found to bind monomeric Ub as well as K48-linked chains, and thereby impair recognition by the proteasome and inhibit protein degradation¹⁸. However, ubistatins have weak (sub-μM to μM) binding and poor chain specificity¹⁹.

Targeting Ub chains of particular lengths and linkages could control different cellular processes. However, this represents a significant challenge for molecular recognition, given the number of possible Ub chains⁴ and the often subtle differences in their structure and dynamics^20,21. Yet, we have envisioned that this could be achieved using macrocyclic peptides, which can tightly interact with proteins using extended interfaces²², similar to natural protein-protein interactions, but at the same time are small enough to emulate drug-like in vivo properties of small molecules²³.

There are powerful methods to discover de novo binding peptide sequences, such as the RaPID system (Random non-standard Peptides Integrated Discovery). In the RaPID system, in vitro translation is modified to use reprogrammed genetic codes²⁴, including nonproteinogenic amino acids that allow for spontaneous peptide macrocyclization²⁵. RaPID can generate huge, trillion-member libraries of DNA-tagged cyclic peptides^{26, 27}. Panning these libraries for target engagement, followed by DNA sequencing, allows for selection of highly specific, tight binding de novo cyclic peptide sequences.

We reasoned that the RaPID system could produce cyclic peptide ligands capable of discriminating between Ub chains. However, the success of such a selection would depend critically on the homogeneity, in length and linkage, of these target Ub polymers. This is difficult to achieve using biochemical and biological approaches for all the Ub chains. However, recently it has become possible to assemble the protein Ub and polyUb by total chemical synthesis. This allows for the construction of high-purity Ub chains, with defined lengths²⁸ and linkage types, including any required modifications, suitable for use as targets in these challenging selections.

By combining the RaPID system and chemical protein synthesis of Ub chains, we discover for the first time de novo cyclic peptides that tightly bind the K48-linked Ub chains, with great selectivity towards the chain length (mono-, di-, and tetra-Ub) and linkage type (K48 vs K63 or K11). We also report the in vitro effects of the selected cyclic peptide on the behavior of these Ub chains, e.g. their interaction with DUBs and the proteasome, as well as their ability to inhibit cellular protein degradation.

Results

Choice of K48-linked target

We sought to determine whether selectivity towards different Ub oligomeric forms or a particular chain linkage is possible using cyclic peptides. We chose K48-linked Ub chains as targets for de novo cyclic peptide discovery, as chemical compounds that can bind K48-linked Ub have biological and therapeutic significance. In particular, the degradation of specific cellular proteins by the 26S proteasome, controlled using K48-linked polyUb tags, determines whether a cell proliferates or dies²⁹. Cancer cells use degradation of certain proteins e.g. p53, Bax and p27^30–32, to avoid the triggering of apoptosis. Therefore, K48-linked polyUb-binding cyclic peptides may elevate the levels of these short-lived proteins and might selectively restore apoptosis in these cells^31,33,11.

Total chemical synthesis of biotin tagged K48-linked Ub chains

We applied our chemical toolbox to construct K48-linked Ub chains suitable for RaPID selection. This consists of orthogonally protected δ-mercaptolysine³⁴, solid phase peptide synthesis (SPPS) of Ub monomers (modified as required), coupled with isopeptide chemical ligation and desulfurization to prepare mono Ub (Ub₁), di-Ub (^K48Ub₂) and tetra-Ub (^K48Ub₄) chains³⁵. The N-terminus of each distal Ub was further modified with biotin, to facilitate immobilization to a solid support (Fig. 1a). Highly homogenous Ub chains were obtained which could be reconstituted into folded proteins, and bound to streptavidin (Supplementary Figs 1–3).

Figure 1. RaPID selection of ^K48Ub_n binding cyclic peptides.

a) Chemical synthesis of biotin-K48-linked di/tetra-Ub chain. b) Flexizyme, an artificial tRNA aminoacylation ribozyme, can load nonproteinogenic amino acids onto tRNA (black), for use by the ribosome in in vitro translation. nonproteinogenic amino acids with an N-terminal ClAc group will spontaneously react with intramolecular cysteine and can be used to form macrocyclic peptides. c) Large libraries (~10¹³) of DNA-tagged cyclic peptides can be generated in the RaPID system, from which sequences can be isolated to tightly bind a protein target. d) Using ^K48Ub₂ as a target, RaPID DNA libraries became enriched in certain peptide sequences (grey lines), with Ub2i and Ub2ii (black) as the dominant peptides (selection repeat shown as dashed lines). Using ^K48Ub₄ as a target, Ub4i and Ub4ix (black) emerged amongst the enriched peptides. e) (top to bottom) Synthesized peptides Ub2i, Ub2ii, Ub4i, Ub4ix and their binding to ^K48Ub₂ (left) and ^K48Ub₄ (right) detected using SPR. Highest and lowest peptide SPR concentrations are shown in blue and grey respectively.

Discovery of de novo cyclic peptides for K48 Ub chains

The chemically synthesized Ub chains, biotin-^K48Ub₂ and biotin-^K48Ub₄, were used as targets for two separate RaPID selections (Fig. 1b). The initial cyclic peptide libraries contained ~10¹³ unique sequences, with 8–12 random amino acids per peptide. After the first round of selection an additional step was introduced, binding to biotin-Ub₁, to remove cyclic peptides that might use Ub₁ as the main recognition element. Deep sequencing of the DNA library after each round showed that a small number of peptide sequences, and their variants, came to dominate the library by round 4 (Fig. 1c, Supplementary Fig. 4). Rounds 2–4 were repeated and produced a similar pattern of enrichment (Supplementary Fig. 4). In the ^K48Ub₂ selections the libraries became dominated by the same two sequences, Ub2i and Ub2ii (Fig. 1c, Supplementary Fig. 4). In the ^K48Ub₄ selections, many of the top peptides had appreciable sequence identity to Ub2i (Supplementary Fig. 4); One of these high identity peptides (58 %), Ub4i, was chosen for further study together with a shorter, lower identity peptide (8 %), Ub4ix (Fig. 1c).

De novo cyclic peptides tightly bind K48-linked Ub chains

To test the binding and specificity of the Ub2i, Ub2ii, Ub4i and Ub4ix cyclic peptides, each was prepared by SPPS (Supplementary Fig. 5) for Surface Plasmon Resonance (SPR) analysis (Fig 1d). Each peptide was flowed over immobilized Ub chains: the selection targets ^K48Ub₂ and ^K48Ub₄, as well as Ub₁, and alternatively linked ^K11Ub₂, ^K63Ub₂. Ub2i and Ub2ii bound tightly to their selection target ^K48Ub₂, with low-nM K_d values (40 ± 12 nM, 33 ± 8 nM, respectively, Fig 1d). However, Ub2i and Ub2ii also strongly bound the longer chain ^K48Ub₄, with a low apparent K_d and an increased SPR response (Fig 1d). This is consistent with a higher stoichiometry of binding to the tetramer, which could be explained by the fact that ^K48Ub₂ structure can be superimposed twice on the ^K48Ub₄ crystal structure (Supplementary Fig. 6)³⁶.

Ub4i tightly bound its selection target ^K48Ub₄, but also bound ^K48Ub₂ similar to Ub2i and Ub2ii. This is perhaps unsurprising, given the high sequence identity between Ub4i and Ub2i. On the other hand, the cyclic peptide Ub4ix was highly specific for the K48-linked Ub tetramer: weakly binding ^K48Ub₂ ( > 1 μM K_d) but tightly binding to ^K48Ub₄ (K_d = 6 ± 1 nM).

None of the cyclic peptides showed detectable binding to Ub₁ (Supplementary Methods). This was expected as the RaPID selection was designed to avoid recovering Ub₁ binders. More interestingly, the cyclic peptides showed marked selectivity for the K48-linkage over alternative linkages: there was no detectable binding, for any of the cyclic peptides, to either K11-linked or K63-linked di-Ub (Supplementary Methods).

Mapping of cyclic peptide/K48-linked poly-Ub interface using NMR

We then became interested in understanding the binding mode of the discovered cyclic peptides with the K48-linked Ub chains. We used NMR to confirm physical interaction between Ub2ii and ^K48Ub₂ and to map the residues/sites involved in the peptide binding. The peptide was titrated into solution of ^K48Ub₂ where either distal or proximal Ub was ¹⁵N-enriched (Supplementary Fig. 7), and the binding was monitored by ¹H-¹⁵N NMR. The titration revealed site-specific interactions with both Ub units (Figs. 2a,b and Supplementary Fig. 8). Interestingly, the NMR spectra followed a text-book example of slow exchange, where a gradual decrease in intensity of the “unbound” Ub signals was accompanied by appearance and increase of alternate (“bound”) signals. At a 1:1 molar ratio, practically all affected unbound signals vanished (Figs. 2a and Supplementary Fig. 8). This likely reflects slow off-rates, in full agreement with the SPR data (Supplementary Table S1) and strong, nM-K_d binding (Fig. 1e). By contrast, only minor spectral perturbations were observed upon addition of Ub2ii to mono-Ub and ^K63Ub₂ (data not shown), corroborating high specificity of the peptide for ^K48Ub₂.

Figure 2. Cyclic peptides bind to residues at the hydrophobic Ub-Ub interface in K48-linked di- and tetra-Ub.

a) Overlay of the ¹H-¹⁵N correlation spectra of ^K48Ub₂ (proximal Ub) free (blue) and in the presence (red) of 1 molar equivalent of Ub2ii. Signal shifts for select residues are indicated. Insets on the right illustrate the behavior of the signal of Q40 (boxed) during the titration, at the indicated peptide:Ub₂ molar ratios. b) Residue-specific chemical shift perturbations (CSP) at the endpoint of titration with Ub2ii in the distal (left) and proximal (right) Ub units in ^K48Ub₂. c) The perturbed residues are mapped (CSP above the mean for each domain colored green) on the crystal structure of the closed state of ^K48Ub₂ (PDB 1AAR) d) (Left) Comparison of the intensities of the unbound signals of the most distal (unit 4, black bars) and the next to distal (unit 3, cyan bars) Ub units in ^K48Ub₄ upon addition of Ub4ix in 2:1 and 1.5:1 molar ratio, respectively. The signal intensity for each residue was normalized to the corresponding intensity in the absence of peptide. (Right) A similar comparison of the signal intensities in ^K48Ub₄ unit 3 (cyan) versus the distal Ub in ^K48Ub₃ (unit 3, black), both at the 1.5:1 molar ratio (peptide:polyUb). The drawings on top indicate which Ub units in ^K48Ub₄ and ^K48Ub₃ are analyzed. Asterisks in (b) and (d) indicate residues that could not be detected or could not be reliably quantified due to signal overlap. e) Map of the residues showing strong signal attenuations (normalized signal intensity < 0.2 colored green) in the two distal Ub units (4 and 3) of ^K48Ub₄ on the crystal structure of the closed form of ^K48Ub₄ (PDB 2O6V).

The ^K48Ub₂ sites affected by Ub2ii binding were mapped to residues in and around the L8-I44-V70 hydrophobic patch on both Ub units (Fig. 2b–c), indicating that the cyclic peptide intercalates between the two Ub units in ^K48Ub₂. Simultaneous binding of both Ub units to the same cyclic peptide is supported by the fact that the saturation was reached at the 1:1 molar ratio indicating a 1:1 stoichiometry of the resulting Ub₂/Ub2ii complex. The observed interface is similar to that involved in ^K48Ub₂ interactions with ubistatins^18,19 as well as K48-specific protein receptors³⁷. Interestingly, despite the general symmetry between the two Ub units in Ub₂, the chemical shift perturbations (CSPs) and the directions of the signal shifts are drastically different between the distal and the proximal Ubs. Combined with the fact that only a single set of “bound” NMR signals is observed for each Ub unit, this indicates that Ub2ii binds ^K48Ub₂ unidirectionally and in a single conformation.

Using NMR, we also monitored the interaction between Ub4ix and ^K48Ub₄. As with ^K48Ub₂/Ub2ii, the binding was in slow-exchange regime, and the interface was mapped primarily to the hydrophobic surface patch on each Ub unit studied (Figs. 2d,e and Supplementary Figs. 9–11). Interestingly, the most distal Ub exhibited relatively weaker spectral perturbations compared to the other units in the chain, and the disappearance of the unbound signals occurred at higher peptide concentrations than for Ub unit 3 or the proximal Ub, suggesting relatively weaker Ub4ix interaction with the most distal Ub in the chain. To examine if this was related to the length of the chain or the external (distal) position of the Ub unit, we performed similar binding studies for ^K48Ub₃ (Supplementary Figs. 12–14). The spectral perturbations in the distal Ub of this tri-Ub were comparable with those in unit 3 of ^K48Ub₄ but different from the most distal Ub in ^K48Ub₄ (Figs. 2d, Supplementary Fig. 15). Also the proximal Ub units in ^K48Ub₄ and ^K48Ub₃ exhibited similar patterns of residue-specific signal shifts (Supplementary Fig. 16). Together, these results suggest that the cyclic peptide Ub4ix primarily binds to/across the first three Ub units in ^K48Ub₄. Consistent with this conclusion and the SPR results, the addition of Ub4ix to ^K48Ub₂ caused lesser spectral perturbations than in the case of Ub2ii peptide at the same (1:1) molar ratio (data not shown), indicating weaker binding.

Cyclic peptide binding to Ub chains protects these chains against DUB cleavage in vitro.

It is possible that the binding of these cyclic peptides to Ub chains prevents their recognition by interacting proteins such as DUBs. We chose to examine the effect of these cyclic peptides on two DUBs: OTUB1³⁸, a K48-linkage chain specific DUB, and USP2, which cleaves most Ub chains without specificity³⁹, using ^K48Ub₂ and ^K48Ub₄ chains as substrates (Fig. 3). All peptides were able to inhibit OTUB1 with ^K48Ub₂ as the substrate (Fig. 3a), at ratio of 1:1 cyclic peptide: substrate, even the weakest binding peptide Ub4ix was able to fully inhibit OTUB1 (Fig. 3c). This suggests that binding to Ub chains by cyclic peptides can interfere with recognition by the DUB. However, with USP2 using ^K48Ub₂ as the substrate, we observed only ~40% inhibition in the presence of Ub2i, Ub2ii or Ub4i, and only minimal inhibition by Ub4ix (Fig. 3b). Excess Ub4ix over substrate concentration was required to completely inhibit the activity of USP2 (Fig. 3d). Using a ^K48Ub₄ chain as the substrate instead, both OTUB or USP2 could cleave down to monomer Ub units, and this cleavage was inhibited in the presence of the cyclic peptides (Fig. 3e, f). However, USP2 in the presence of the cyclic peptides Ub2i, Ub4ix, and Ub4i could only remove the distal Ub of ^K48Ub₄⁴⁰, shortening the chain to the trimer ^K48Ub₃, the peptides show strong inhibition of further disassembly of the chain (Fig. 3e, f). For Ub4ix this can be explained by the NMR result that suggested that the cyclic peptide only weakly interacts with the distal Ub, which may be sterically accessible for cleavage by USP2. Interestingly, the same cyclic peptides did not protect ^K63Ub₂ chains from USP2 cleavage, nor the structurally similar ^K11Ub₂, from Cezanne cleavage⁴¹ (Supplementary Fig. 17), highlighting the specificity of the cyclic peptides towards the K48-linkage, supporting by the NMR and SPR studies (see above) .

Figure 3. Cyclic peptides inhibit the DUB cleavage of K48-linked di/tetra-Ub.

a) ^K48Ub₂ incubated with the K48-linked specific DUB OTUB1 for 1 hour resulted in cleavage to Ub₁. This reaction was inhibited by all cyclic peptides. b) ^K48Ub₂ incubated with the non-linkage-specific DUB USP2 resulted in cleavage to Ub₁, cleavage could be inhibited by all peptides. c) Ub4ix inhibits OTUB1 in a dose dependent manner, with an IC₅₀ approximately the same as the substrate concentration (^K48Ub₂ initially at 2 μM). d) Excess of Ub4ix (relative to ^K48Ub₂ substrate) was required to fully inhibit the activity of USP2 over a 1 hour incubation. e) ^K48Ub₄ incubated for 2 hours with OTUB1 resulted in cleavage to ^K48Ub₃, ^K48Ub₂ and Ub₁. In the presence of cyclic peptides, cleavage to ^K48Ub₃ occurred, but further cleavage was inhibited. f) Similar for OTUB1, ^K48Ub₄ incubated for 1 hour with USP2 could cleave to ^K48Ub₃, ^K48Ub₂ and Ub₁, and this cleavage was inhibited by the cyclic peptides. In all experiments, equal amounts were loaded on SDS-PAGE gel, electro-blotted to nitrocellulose membrane, and probed with an anti-Ub antibody. Initial concentrations of ^K48Ub₂ or ^K48Ub₄ substrate were 2 μM in each case.

Cyclic peptide binding to Ub chains prevents ubiquitinated proteins from proteasomal degradation

It is possible that the binding of cyclic peptides to K48-linked Ub chains could prevent the recognition by the proteasome, and inhibit the degradation of proteins tagged with these Ub chains. To test this we incubated synthetic α-globin-^K48Ub₄ with 26S proteasome, with and without Ub4ix. Without Ub4ix, the α-globin-^K48Ub₄ substrate was degraded by the proteasome. However, with equimolar amounts of Ub4ix relative to α-globin-^K48Ub₄, degradation was inhibited, a similar effect to that of the direct proteasome inhibitor, MG132 (Fig. 4). Uncyclized linear Ub4ix showed no inhibition, highlighting the requirement of the cyclic topology for binding to Ub.

Figure 4. Ub4ix cyclic peptide inhibits 26S proteasomal activity in vitro.

HA-α-globin-^K48Ub₄ (5 μM) and 26S proteasome (150 nM) were combined in the presence and absence of cyclic peptide Ub4ix, linear-Ub4ix or MG132. Before and after incubation, products were separated on SDS-PAGE gel, electro-blotted and probed with anti-HA antibody. Linear-Ub4ix offers no protection to HA-α-globin-^K48Ub₄ from the activity of the proteasome, resulting in degradation. Cyclic peptide Ub4ix, at a similar concentration to the HA-α-globin-^K48Ub₄ substrate (5 μM) can protect it from degradation by the proteasome, a similar result to the adding the direct proteasome inhibitor MG132.

Ub4ix is able to enter into HeLa cells.

In order to assess the feasibility of using Ub4ix in cultured cells, we next tested whether Ub4ix is able to cross the cellular membrane of human cells. We synthesized Ub4ix labeled with fluorescein (Fig. 5a, Supplementary Fig. 18), and used live cell imaging to monitor its entrance into HeLa cells. The uptake of labeled Ub4ix was apparent as early as 4 hours (Fig. 5b). Following its addition to the culture medium, with further intracellular marked accumulation, throughout the following 48 hours (Fig. 5b). Comparing the uptake of fluorescein-Ub4ix with fluorescein alone in three cell lines (HeLa, U2OS osteosarcoma and U87 primary glioblastoma cells), fluorescein-Ub4ix exhibited considerable cell permeability (Fig. 5c). In addition, we screened the cellular uptake fluorescein-Ub4ix compared to free fluorescein, after incubation for 16 hr (Fig. 5d). Ub4ix was quickly and efficiently distributed within cells, allowing tests in more representative biological models without further modification or special delivery systems.

Figure 5. Uptake of Ub4ix by living cells and effect on ubiquitination.

a) Fluorescein-Ub4ix structure. b) Ub4ix can enter cells. HeLa cells were incubated with fluorescein-Ub4ix cyclic peptide (20 μM). Internalization of the peptide into the cells was monitored at the indicated time points. c) In three cell lines, HeLa, U2OS osteosarcoma and U87 primary glioblastoma cells, fluorescein-Ub4ix is retained in cells after washing, whereas fluorescein alone is lost. d) Screening of the cellular uptake fluorescein-Ub4ix after incubation for 16 hr shows significant retention inside cells. e) Ub4ix induced the accumulation of Ub-conjugates. HeLa cells were incubated in the presence of Ub4ix, Fluorescein-Ub4ix, or MG132 for the indicated time-periods. f) Ub4ix induced the accumulation of Ub-conjugates in a dose dependent manner. Cells were lysed, resolved using SDS-PAGE, proteins were transferred to a nitrocellulose membrane, and the membrane was blotted using an antibody against Ub-conjugates and Actin. g) Ub4ix inhibits protein breakdown inside cells. HeLa cells were labeled with radioactive ³⁵S-Met and ³⁵S-Cys, and were then chased for 4 hr in “cold” media. Cells were either non-treated, or incubated in the presence of DMSO, Ub4ix, or MG132. Radioactivity levels (counts per min, CPMA) from protein breakdown during the chase phase, show that Ub4ix and MG132 inhibit. Bars represent mean values. h) Ub4ix causes the accumulation of proteins that are normally turned over by the UPS. HeLa cells were incubated in the presence of 10 μM Ub4ix or 10 μM MG132 for the indicated times. Cells were lysed, resolved using SDS-PAGE, proteins were transferred to a nitrocellulose membrane, and the membrane was blotted using an antibody against p53, p27 and Actin.

Ub4ix activity causes accumulation of Ub-conjugates

Under basal conditions, the levels of Ub-conjugates reflect a dynamic steady state between the enzymatic tagging of Ub to proteins on the one hand, with the removal of conjugated Ub moieties by DUBs, and subsequent degradation by the proteasome, on the other. We have shown in vitro that the cyclic peptide Ub4ix can protect Ub-chains from cleavage by DUBs, and also prevent recognition by the proteasome. Both effects would be predicted to prevent degradation, and lead to the accumulation, of proteins tagged with K48-Ub chains in cells. First we confirmed that Ub4ix does not affect protein synthesis (Supplementary Fig. 19). Then, we monitored the cellular level of Ub-conjugates in cells upon treatment with Ub4ix. We were able to demonstrate a time and dose dependent elevation in Ub-conjugates following treatment with Ub4ix, both in its free and fluorescein labeled forms (Fig. 5e and f), a similar result to treating with the direct proteasome inhibitor MG132 (Fig. 5e and f).

In order to directly measure the effect of Ub4ix on proteolysis, we radiolabeled cellular proteins, followed by a chase experiment in the presence of either Ub4ix or the proteasome inhibitor MG132. When compared to non-treated cells, or ones treated with DMSO, protein breakdown in cells treated with Ub4ix was markedly reduced, a similar result to those observed in the presence of MG132 (Fig. 5g). Both agents show a dose-dependent effect on cellular protein degradation. We validated this inhibitory effect on the breakdown of proteins known to be proteasomal substrates; the proteins p27⁴² and p53⁴³. In the presence of the Ub4ix, p53 and p27 were accumulated over time, up to a similar level to those observed in the presence of MG132 (Fig. 5h).

Cyclic peptides induce apoptosis

Given that Ub4ix has a pronounced inhibitory effect on the UPS system, it may also be able to inhibit cell growth and induce apoptosis in cancer cells. To test this, HeLa cells were treated with each cyclic peptide, and cell growth was found to be dramatically suppressed when viability was assessed using an MTT (3-(4,5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide) assay. We noted that Ub4ix was the most effective in suppressing growth (Fig. 6a). The effects of the cyclic peptides on apoptosis were also evaluated using fluorescence-activated cell sorting (FACS) analysis. Intriguingly, Ub4ix at 10 μM exhibited clear increase in apoptosis after 24 hr and 48 hr treatment, similar to the direct proteasome inhibitor MG132 (Fig. 6b).

Figure 6. Ub4ix reduces cell viability and induces apoptosis.

a) Cyclic peptides reduce HeLa cells viability, as measured by MTT assay. Ub4ix inhibits the cell viability of HeLa cells with an IC₅₀ of 1.6 μM after 48 hr incubation. Viability was measured after 72 hr for Ub2i, Ub2ii, Ub4i and linear-Ub4ix. Error bars represent SD of technical replicates (n=16). b) Ub4ix induces apoptosis in HeLa cells. Cells were exposed to the Ub4ix cyclic or linear peptide, MG132 (all at 10 μM) for 24 and 48 hr, FACS was used to follow apoptosis and cumulative cell death, bars represent mean values.

Discussion

We have discovered de novo cyclic peptides that bind tightly and specifically to K48-linked Ub chains. Ub-binding domains with a similar function exist in biology, but typically these bind with weak affinities (μM K_d, values⁴⁴), mainly to the L8-I44-V70 hydrophobic patch on the Ub surface, and display a narrower range of affinities for different Ub chain lengths and linkages. By contrast, antibodies have been discovered that can bind strongly (nM K_d), and specifically to polyUb chains by directly interacting with the isopeptide bond and its surrounding residues⁴⁵. Surprisingly, the de novo macrocyclic peptides exhibit remarkably high affinity and linkage specificity like the antibodies. Yet, they interact with the hydrophobic surface of Ub frequently used by the weak-binding, poorly-selective natural Ub-binding domains. The affinity and selectivity profiles of these cyclic peptides are even more impressive given their significantly lower molecular masses relative to the protein domains and antibodies. Importantly, it is this smaller size that permits entry into cells⁴⁶.

These cyclic peptides show in vitro and cellular activity comparable to a potent proteasome inhibitor, and can similarly trigger apoptosis in cancer cells. However, they act through a different mechanism. By binding to Ub chains directly and strongly, they interfere with the recognition and subsequent proteasomal processing of these chains. This different mechanism may have therapeutic advantages, as it can avoid the innate and acquired resistance associated with direct proteasome inhibitors⁴⁷: it is difficult to mutate the Ub target, given its essentiality and its short, highly conserved sequence⁴⁸. Importantly, unlike generic Ub-binding small molecules¹⁹, these cyclic peptides do not bind Ub₁, K11- or K63-linked Ub chains strongly or specifically. Therefore, many Ub-directed cellular processes can continue unimpeded in the presence of these cyclic peptides. Our ability to synthesize different Ub chains with various lengths and linkage types, coupled with RaPID system, could lead to the discovery of novel cyclic peptides that interfere with specific cellular responses associated with these different Ub chains, to shed light on the full range of Ub signals in health and disease.

The combination of high-fidelity total protein synthesis and the cyclic peptide discovery RaPID system has produced de novo ligands capable of distinguishing between subtly different oligomers of the same protein domain. The control offered by this highly synthetic, in vitro approach will make other elaborate, post-translationally modified targets proteins accessible for ligand discovery. In particular, other Ub chains with different linkages and lengths, ubiquitinated proteins and Ub-like conjugates can be targeted to interfere with a wide range of cellular processes²⁸. These cyclic peptides will be valuable tools to understand Ub signaling in health and disease, and open new opportunities in drug discovery related to the Ub system.