Delivering bioactive cyclic peptides that target Hsp90 as prodrugs

Abstract The most challenging issue facing peptide drug development is producing a molecule with optimal physical properties while maintaining target binding affinity. Masking peptides with protecting groups that can be removed inside the cell, produces a cell-permeable peptide, which theoretically can maintain its biological activity. Described are series of prodrugs masked using: (a) O-alkyl, (b) N-alkyl, and (c) acetyl groups, and their binding affinity for Hsp90. Alkyl moieties increased compound permeability, Papp, from 3.3 to 5.6, however alkyls could not be removed by liver microsomes or in-vivo and their presence decreased target binding affinity (IC50 of ≥10 µM). Thus, unlike small molecules, peptide masking groups cannot be predictably removed; their removal is related to the 3-D conformation. O-acetyl groups were cleaved but are labile, increasing challenges during synthesis. Utilising acetyl groups coupled with mono-methylated amines may decrease the polarity of a peptide, while maintaining binding affinity.


Introduction
Peptides are evolutionarily designed by nature to inhibit proteinprotein interactions (PPIs) [1][2][3][4][5][6] . Peptide scaffolds are densely packed with stereochemistry, they have rich structural diversity, and their synthesis can be accomplished in a timely manner. Peptide drugs have proven to be highly successful at regulating PPIs and drug development has increasingly relied upon the peptide field to produce new PPI modulators as therapeutics 2,7,8 . The global peptide drug market is expected to double by 2024, and reach $46.6 billion 1,8 with $140 peptides being evaluated in clinical trials each year 1 .
The most challenging issue facing peptide drug development is producing a molecule with optimal physical properties while maintaining target binding affinity. The most challenging property to improve is cell permeability because structural modifications typically impact the molecule's efficacy and aqueous solubility [9][10][11][12][13][14][15] . Pioneering research has been accomplished by multiple leaders in this field 12,13,16,17 and three strategies have emerged: (1) masking polar side chains with hydrophobic moieties that can be removed within the cell, (2) incorporating D-amino acids in order to modify the conformation, and (3) N-methylating the backbone in order to alter the conformation and promote hydrophobicity 9,11,12,15 . Biological studies revealed that while including D-amino acids and N-methyl groups was favourable for membrane permeability, the irreversible chemical modifications negatively impacted the bioactivity of these molecules 9,10 . In contrast, masking peptides with protecting groups that can be removed inside the cell, produces a cell-permeable peptide, which theoretically can maintain its biological activity. Proof that a prodrug or masking strategy can succeed is evidenced by the fact that 20% of all small molecule drugs approved between 2000 and 2008 were prodrugs 18 .
We had previously reported molecules that targeted heat shock protein 90 (Hsp90), with the lead structures being LB51 and LB6332 ,33 . However, these active inhibitors had poor cell permeability, and three approaches were used to make these molecules more effective at entering cells: masking groups 10 , D-amino acids 10 , and N-methylation 9 .
Backbone N-methylation increased the membrane permeability of LB51, from Papp $3 to between 5 and 6 9 , but this increase in permeability resulted a reduction in binding affinity. Inclusion of D-amino acids at each position of LB51 also improved the molecule's permeability from Papp 3 to Papp ¼ $4.2-4.6 10 . However, similar to the compounds where N-methyl was added, molecules with D-amino acids all had decreased binding affinity compared to LB51. Although others have used the strategies described above to successfully to produce compounds with higher cell permeability, their biological activity was not assessed, or their molecules were acting on targets located outside the cell. There is no effective system that allows a peptide to enter the cell, be released, and bind effectively to its intracellular target.
Herein we describe the biological activity of molecules that were designed to target Hsp90 while maintaining cell permeability. We describe the prodrug binding affinity for Hsp90, and the ability of liver microsomes, and in vivo experiments to cleave three types of masking groups: (a) O-alkyl, (b) N-alkyl, and (c) acetyl. The cleavage of these side chains and the binding affinity of the final molecules are reported.

General
AlphaLISA V R binding assay The binding assays were performed using an Hsp90 (C-terminal) Inhibitor Screening Kits (cat. 50314 or 50317) purchased from BPS Bioscience. The assay was performed according to the manufacturer's protocol and utilised AlphaLISAV R technology (PerkinElmer). The test compounds were dissolved in 100% DMSO and diluted with water to the desired concentration, so that the final dilution was dissolved in 5% DMSO with water. Two microlitres of the dilution was added to a 10 mL reaction so that the final concentration of DMSO was 1% in all reactions. The reactions were conducted at room temperature for 30 min in a 10 mL mixture containing assay buffer, 6 ng (24 nM) of a C-terminal Hsp90b (UniProt P08238, a.a. 527-724) or C-terminal Hsp90a (UniProt P07900, a.a. 535-732), 40 ng (60 nM) Cyp40, and the test compound. Note: for the full-length Hsp90 binding assays, C-terminal Hsp90 was replaced with 24 nM of full-length Hsp90a (Abcam cat # ab48801). After the 30 min incubation, 10 mL of buffer containing 20 mg/mL glutathione acceptor beads (Perkin Elmer) were added to the reaction mix and incubated for 30 min in the dark. Ten microlitres of 40 mg/mL streptavidin donor beads (Perkin Elmer) were then added and the final 30 mL mixture was incubated for 1 h in the dark. The AlphaLISA V R signal was measured using a Tecan F200 Pro multimode plate reader.
Metabolism assays using human liver microsomes Human liver microsomes (Life Technologies Gibco cat. # HMMPCL) were thawed on ice and suspended in 100 mM phosphate buffer (pH 7.4) at 0.5 mg/mL. Compounds dissolved in DMSO were then added to the reaction mixture at a final concentration of 5 mM and a final DMSO concentration of 0.1%. Metabolism was initiated by adding 5 mM NADPH and the reactions were incubated at 37 C in a water bath. Fifty microlitre aliquots of the reaction mixture were taken at various time points and added to 100 mL of cold methanol to terminate the reaction. The time points were centrifuged at 1000 Â g for 5 min. Samples were stored at À20 C until LC/MS analysis.

Metabolism assay in a mouse
Mice were purchased from Australian BioResources (ABR) (Moss Vale, NSW, Australia) at 7 weeks of age and shipped to a specific pathogen-free facility at UNSW Biological Resources Centre (Sydney, Australia) with 12 h light/dark cycles. Following 1 week of acclimatisation, mice were injected in the intraperitoneal space with LB55 (1.2 mg in 120 mL Milli-Q water). One hour later, mice were euthanised by CO 2 asphyxiation and the liver was collected for drug extraction. All mouse experiments were approved by the UNSW Animal Care and Ethics Committee. To extract the compound for LC/MS analysis, the liver tissue was powdered in a mortar and pestle that had been pre-cooled with liquid nitrogen. The tissue was then homogenised in phosphate buffered saline (pH   .4) by sonication for 2 h. The homogenate was then centrifuged at 800 Â g for 10 min at 4 C before collecting the supernatant. The drug was extracted, and the proteins were precipitated by adding 1 ml of supernatant to 9 ml of a mixture of methanol and acetonitrile (1:9 v/v). The extraction was then centrifuged at 1000 Â g for 10 min at 4 C and subsequently analysed by LC/MS.

Competitive binding assay
All test compounds and biotin-tagged compounds were dissolved in 100% DMSO and were diluted to a final concentration of 1% DMSO in each reaction. Each reaction consisted of C-terminal Hsp90, (100 mM), LB51-tag (100 mM), and binding buffer (100 mL). The reactions were rocked at RT for 1 h. The untagged compound was added to each reaction (0-100 mM) and incubated for 1 h. Protein pull-down was achieved by incubating each reaction with NeutrAvidin V R Agarose Resins (Thermo Scientific 29201 or 29204), for an hour. The NeutrAvidin V R Agarose Resin was pre-blocked with 1.3% of Bovine Serum Albumins (Sigma, A2153) for 1 h. Following incubation with NeutrAvidin V R , the supernatant was removed, and the resin was washed 6 times. The beads were boiled with 5Â Laemmli sample buffer, to elute the protein complex. The supernatant was loaded onto a 12% Tris-Glycine gel following the Gel electrophoresis procedure. Proteins were transferred onto a PVDF membrane and imaged following Western blotting. C-terminal Hsp90 was detected using anti-6Â His-tag antibody (Abcam, ab18184, 1:1500 dilution). The relative amount of C-terminal Hsp90 bound to LB51-tag was determined, where DMSO was set to 100% binding.

General chemistry procedures
All chemicals were purchased from commercial suppliers (Chem-Impex International and Sigma Aldrich) and used without further purification. All moisture sensitive reactions were performed using anhydrous solvents under nitrogen gas. Removal of solvent was carried out under reduced pressure using a Buchi R-210 rotary evaporator. Samples were lyophilised on a CHRIST Alpha 1-4 LD Plus freeze dryer.

Solid-phase peptide synthesis
Stepwise solid-phase peptide synthesis was performed in a polypropylene solid-phase extraction cartridge fitted with a 20 mm polyethylene frit purchased from Applied Separations (Allentown, PA) Resin loading 2-chlorotrityl chloride resin was weighed, transferred to the cartridge and swelled in CH 2 Cl 2 for 30 min prior to the resin loading reaction. The appropriate Fmoc-protected amino acid was dissolved in the minimum amount of 0.4 M DIPEA in CH 2 Cl 2 . The swelled resin was then drained, and the dissolved amino acid was added, and the suspension was agitated for 4 h. The resin was then washed 3 times with CH 2 Cl 2 , 3 times with DMF, and 3 times with CH 2 Cl 2 . The resin was then dried in vacuo overnight. An $5 mg sample of resin was used to determine the amino acid loading. Twenty percent piperidine in DMF was added to the sample to cleave the Fmoc protecting group. The resin was filtered away, and the remaining solution was diluted 1 in 20 and the UV absorbance was measured at 301 nm using a Cary 50 Bio UV-Vis instrument (Sydney, Australia). DMF was used as a blank and samples were measured in a 1 ml quartz cuvette. The resin loading was then determined using the following formula: (20), e ¼ extinction coefficient of Fmoc group (7800 ml mmol À1 cm À1 ), W=cuvette width (1 cm), m ¼ mass of resin (g).

Coupling reaction
Couplings were performed in DMF at a concentration of 0.3 M. Fmoc-protected amino acid (2 equiv) and HOBt (2 equiv) were mixed with the resin. DIC (4 equiv) was then added to activate the reaction. Coupling reaction was run for a minimum of 2 h while shaking (Labquake tube shaker, Thermo Fisher Scientific) at room temperature. A negative ninhydrin test was used to confirm reaction completion. Once completed, the reaction mixture was drained and the resin was subjected to Fmoc Removal. (Note: For particularly hindered coupling reactions, HOBt was replaced with HOAt.)

Fmoc removal
The Fmoc protecting group was removed using the following washes: and DMF (3 Â 1 min). The resin was then ready for the next coupling reaction.

Cleavage
Once the desired peptide was generated, the final Fmoc protecting group was removed following Fmoc removal procedure with the following additional washes: DMF (3 Â 1 min), i-PrOH (3 Â 1 min), and MeOH (3 Â 1 min). The resin-bound peptide was then dried in vacuo overnight. The resin was then cleaved from the linear peptide using TFE and CH 2 Cl 2 (1:1 v/v) at a concentration of 10 ml/g resin. The reaction mixture was stirred at room temperature for 6 h before filtering the resin. The filtrate was concentrated and washed at least 10 times with CH 2 Cl 2 to remove residual entrapped TFE. The product was then dried in vacuo overnight to produce the linear peptide.

Cyclisation
Macrocyclisation of the linear peptide was achieved using a cocktail of 3 coupling reagents: HATU (1 eq.), TBTU (0.8 equiv), and DMTMM (0.8 equiv). The reaction was performed in dilute conditions using anhydrous solvents at a concentration of 0.001 M. The linear peptide and coupling reagents were dissolved separately in DMF, where 20% of the final volume was used to dissolve the linear peptide and the other 80% dissolved the coupling reagents. DIPEA (4 equiv) was added to each solution. The linear peptide solution was then added drop-wise to the coupling reagents solution via a syringe pump over $2 h. The reaction was stirred overnight and monitored via LC/MS. Upon completion, the reaction mixture was evaporated, and the dry solid was subjected to HPLC purification.

Catalytic hydrogenation
Anhydrous methanol (MeOH) was prepared by purging HPLC grade MeOH with N 2 gas over molecular sieves (3Å) for 40 min. A suspension of the protected cyclic pentapeptide and 10% Pd/C (0.1 equiv) in anhydrous MeOH (4 mM) was stirred under a H 2 atmosphere for 2 h. Reaction progress was monitored via LC/MS. Following reaction completion, the mixture was filtered through a sintered frit and the frit was washed with MeOH.

Results and discussion
Although other Hsp90 inhibitors have been reported to modulate the C-terminus [34][35][36][37][38][39][40] , LB51 was the first compound that binds directly to the C-terminal MEEVD region of Hsp90 and effectively inhibits the binding event between Hsp90 and its co-chaperone CYP40 32 ,33 . The apparent permeability (P app Â10 À6 cm/s) of LB51 measured in Caco-2 assays is 3.3 10 . The agreed upon threshold for considering peptides cell permeable are those with P app !1. However, our goal was to develop our lead molecules to be cell permeable with P app of !5, which places them in the range of small molecules marketed drugs, where their average P app is 4 41 .
Modifications to the side chains using the common strategy of Oand N-alkylations were the first prodrugs produced (Figure 2). Compounds LB58, LB56, LB97, and LB55 were all based on the structure of LB51 but contained polar side chains that were masked with methyl moieties (Figure 2). Compound LB55 was the only molecule that had cell permeability P app !5, (P app =5.6). However, masking groups often affect the target binding affinity, and evaluation of these molecules in binding assays was required. A biochemical assay using AlphaLISA technology was used to evaluate the efficacy of these molecules in blocking the interaction between Hsp90 and the co-chaperone, Cyp40 (Figure 2). Each compound was incubated at multiple concentrations with the C-terminal domain of Hsp90 and full-length Cyp40. Beads specific for each protein were incubated with the reaction mixture. Chemical structures of molecules based on LB51 and their corresponding cell permeability and inhibitory activity in protein binding assays. P app represents the cell permeability Â 10 À6 cm/s. The graph shows the impact of series II on the Hsp90-Cyp40 interaction, compared to LB51 and DMSO. Experiments were conducted in triplicate and data represents mean ± SEM of at least 2 independent experiments.
When the two proteins are bound to each other, the beads for one protein transfer energy to the beads of the other protein, thereby producing a signal. When the binding interaction between the two proteins is inhibited, energy transfer cannot occur, which results in no signal being produced. DMSO (1%) was used as a control, where it represents 100% binding between the two proteins.
Although LB55 was the most cell permeable, LB51 and LB97 were the most effective at inhibiting the binding event between Hsp90 and Cyp40. These data indicate that although the Oand N-alkylation on the three side chains (lysine, tyrosine, and serine) all facilitated cell permeability (LB55), target binding affinity was dramatically impacted. The data also show that although this molecule contains a free amine, LB51 is the most effective combination of cell permeability and binding affinity. This combination of data drove us to explore how well the molecules could be dealkylated in the cell. Specifically, we evaluated whether LB55 could be converted into LB51 within the cell, making it ideal for prodrug development.
Given the extensive literature precedents that describe the dealkylation of methyl groups on amino and alkoxy moieties, it was anticipated that LB55 would be demethylated by the metabolic enzymes in liver microsomes and converted to LB51 42-44 . This metabolic assay estimates drug clearance and identifies drug metabolites that are produced in the liver. Cytochrome P450 (Cyp450) proteins are located in the endoplasmic reticulum of liver cells and $60% of marketed drugs are metabolised by this class of enzymes 45,46 .
Using commercially available liver microsomes that contain the full set of endoplasmic reticulum-based metabolic enzymes, including Cyp450s, in an in vitro system the Oand N-dealkylation processes of LB55 were evaluated (Figure 3) [47][48] . Human liver microsomes were purchased from Life Technologies Australia Pty Ltd (catalogue number HMMCPL) and were prepared from a pool of 50 individual male and female donors. The microsomes were diluted into phosphate buffer at pH 7.4 and incubated with LB55 at 37 C in the presence of the enzyme cofactor NADPH, which initiates the reactions. Aliquots of the reaction mixture were taken at several time points and quenched with cold methanol, which inhibits enzyme activity. The samples were immediately centrifuged at high speed to pellet the microsomes and the supernatants were analysed via LC/MS Isotope labelling strategies and mass spectrometry 49,50 would have revealed any low abundance unknown metabolites with high efficiency and sensitivity. However, for simplicity and speed of assessment we opted to identify the removal of the pro-drug moieties using LC/MS in selected-ion mode (SIM). We scanned for specific mass-to-charge ratios (m/z) based on the molecules molecular weight. Propranolol is extensively metabolised in the liver and it produces a major metabolite (hydroxypropranolol), therefore it was used as the control compound when initially optimising the assay conditions 51,52 . After 60 min, $40% of propranolol was hydroxylated (Supplementary Figure S1). LB55 was incubated with the human liver microsomes for 180 min and samples were taken at 0, 10, 20, 30, 60, 120, and 180 min and analysed via LC/MS in SIM mode. The scanning range was set at m/z 600-800, as this range would capture both the parent LB55 and any demethylated metabolites including LB51 (Figure 3).
After 180 min, the LC/MS spectra shows that no metabolism occurred of LB55 when it was treated with liver microsomes (Figure 3(b)). Thus, despite being a common metabolic process, liver microsomes could not accomplish the de-alkylation of the methyl groups on LB55, suggesting that the methyl groups were inaccessible. While the microsome system provides good approximations of Cyp450-mediated drug metabolism, the commercially purchased liver microsomes have some limitations where some enzymes are present at higher concentrations in the commercial systems compared to the liver 53 and other enzymes (especially cytosolic enzymes) are missing 53 . Thus, it is possible that some metabolic processes that would normally occur and convert LB55 into LB51 may still occur but could not be detected using the commercial microsomal assays. To address this issue, we conducted an in vivo animal study using a mouse.
The stability of LB55 in a mouse liver was accomplished by treating the mouse with LB55 via intraperitoneal (IP) injection at 59.4 mg/kg for 1 h. IP injections allow the drug to be rapidly absorbed through the peritoneal membrane and into the portal vein 54 . Within 60 min, drug usually accumulates in the liver, whereupon it is metabolised before it enters systemic blood circulation 55 . After 1 h of treatment, the mouse was sacrificed, and the liver was excised, snap-frozen in liquid nitrogen, and stored at À80˚C. The liver was subsequently extracted in order to isolate any parent compound and metabolites. The liver tissue was powdered in a mortar and pestle that had been pre-cooled with liquid nitrogen, and the tissue powder was homogenised by sonication in phosphate buffered saline. The homogenate was centrifuged, and the supernatant was collected. Cellular proteins were then precipitated through the addition of organic solvents (acetonitrile and methanol). The mixture was then centrifuged, and the supernatant was analysed via LC/MS in SIM mode, scanning for parent LB55 and any demethylated metabolites including LB51.
The major peak from the liver extract is the starting material LB55 (Figure 3(c)). Although a small quantity of LB51 is produced, other metabolites are the primary products generated by mouse liver degradation. Thus, the human liver microsomes and mice liver analysis indicate that methylation was not a suitable prodrug moiety for effectively delivering LB51 into cells and thus, a new generation of prodrugs was designed.
Since Oand Ndealkylation processes failed to cleave during standard enzyme treatments, we investigated prodrug moieties that could be removed more easily: acetyl groups are rapidly released when esterases are present. Given the failure of the methyl groups, it was important to evaluate the acetyl removal prior to synthesis of the molecules. We elected to not only test acetyl moieties but, because removing Oand N-methylations was surprisingly challenging, we also investigated trifluoroacetyl moieties. The trifluoroacetyl group is more labile than the acetyl group given the electron-withdrawing nature of -CF 3 56 . Using the same commercially available liver microsomal assay described earlier, we evaluated the removal of acetylated and trifluoroacetylated amino acids. The studies were performed on three amino acids on LB51 that required protecting groups: serine, tyrosine, and lysine ( Figure 4). Both serine and tyrosine had their acetyl groups rapidly cleaved within a 10-min incubation period. In contrast, the acetyl and trifluoroacetyl groups on lysine were stable, even after 180 min of incubation ( Figure 4). The data showing that acetyl masking groups can only be removed from serine and tyrosine residues lead to the new design of prodrugs. Starting from two of our most effective compounds, LB51 and LB63, we designed new prodrugs using acetylated serine and tyrosine amino acids (Figure 5(a)) 32 .
Although incorporating acetyl groups is likely to improve the pharmacokinetic properties of the lead compounds, these groups are chemically labile, which makes their synthesis challenging. Acetyl groups are easily hydrolysed in both acidic and basic conditions 57,58 , and therefore standard peptide synthesis procedures needed to be modified in order to minimise acetyl hydrolysis during the synthesis of these molecules. During the synthesis of the LB51(Ac) 2 , the asparagine side chain was unprotected, which was also the case during the synthesis of LB63(Ac) 2 . The serine and tyrosine were used as purchased with acetylated groups already on the amino acid side chains. However, the amino side chain on lysine required a protecting group that could be removed under orthogonal conditions to those used for resin cleavage and leaving the acetyl moieties.
Two of the most commonly used protecting groups for lysine are tert-butyloxycarbonyl (Boc) and carboxybenzyl (Cbz) 59 . Boc is stable under basic conditions, making it most suitable to protect side chain amines when the backbone amino acid is protected using Fmoc. However, Boc is removed in acidic conditions, making it incompatible with maintaining the acetyl groups on the serine and tyrosine side chains, therefore another protecting group was used 59 . Cbz is stable under basic conditions, and it can be removed via catalytic hydrogenation, which would have no impact on the acetyl moieties 60 . Thus, Cbz was used as a protecting group on the lysine side chain (Figure 5(b)). The synthesis of LB51(Ac) 2 and LB63(Ac) 2 were produced by starting with Fmocphenylalanine that was loaded onto 2-chlorotrityl chloride resin. The Fmoc group was cleaved using piperidine and the resinbound phenylalanine underwent sequential coupling and Fmocremoval reactions with the appropriate Fmoc-protected amino acids in order to generate the resin-bound linear peptide ( Figure  5(b)). The typical procedure for removing Fmoc was modified in the presence of the acetylated amino acids, and the compound's exposure to piperidine during deprotection of the amine was reduced from 20% to 10% in order to minimise ester hydrolysis.  The linear peptide was then cleaved from the resin in mildly acidic conditions for 6 h instead of 24 h using TFE to produce the crude cleaved peptide. The cleaved peptide was then cyclised in solution under dilute conditions using anhydrous DMF, a combination of coupling reagents (HATU, TBTU, and DMTMM) 61 , and DIPEA. The reaction was monitored via LC/MS and once complete, the solvent was removed by rotary evaporation. The crude Cbzprotected cyclic peptide was then subjected to HPLC purification, which yielded the purified LB51(Ac) 2 -Cbz as a white solid (100 mg, 24% overall yield) ( Figure 5(b)). LC/MS and 2 D NMR confirmed the structure and purity of the final product. The synthesis of LB63(Ac) 2 -Cbz was accomplished in a similar manner (Supp. Scheme S1) and HPLC purified to produce the compound as a white solid (41 mg, 28% yield).
Removal of the Cbz protecting group on lysine for both LB51(Ac) 2 -Cbz and LB63(Ac) 2 -Cbz was performed using catalytic hydrogenation. Catalytic palladium on carbon (Pd/C, 10% loading) was added to the compound that was dissolved in anhydrous, deoxygenated methanol. Purging the reaction flask with hydrogen gas and stirring for 2 h at room temperature under hydrogen, produced the deprotected compounds LB51(Ac) 2 and LB63(Ac) 2 , however, both molecules rapidly deacetylated after removal of the Pd catalyst (Supplementary Figure S2(a)). Several conditions were evaluated in an attempt to generate LB51(Ac) 2 and LB63(Ac) 2 including THF as an alternative solvent. However, none were successful, and we speculate that the free amine on lysine, in conjunction with trace amounts of catalytic palladium were creating conditions that removed the acetyl groups and caused decomposition. The LC/MS immediately after reaction completion (Supplementary Figure S2(b)) shows the desired product is the major peak. Note that molecules with free lysines appear as 2 peaks on the LCMS: protonated (early retention time) and free amine (late retention time). However, after filtration and removal of the catalyst, the compound deacetylated (Supplementary Figure S2(c)). Previous studies had indicated that when the lysine of LB51 had a single methyl masking group, it was still an effective inhibitor of the Hsp90-Cyp40 interaction. Given that the cleavage of the Cbz group was challenging, we opted to assess how effective LB51(Ac) 2 -Cbz and LB63(Ac) 2 -Cbz were when binding to Hsp90.
The alpha biochemical assay was used to evaluate the efficacy of these two molecules, LB51(Ac) 2 -Cbz and LB63(Ac) 2 -Cbz, in blocking the interaction between Hsp90 and the co-chaperone, Cyp40 ( Figure 6(A)). Each compound was incubated at multiple concentrations with the C-terminal domain of Hsp90 and fulllength Cyp40. DMSO (1%) was used as a negative control to represent 100% binding between the two proteins, and LB51 and LB63 were used as positive controls, inhibiting the binding event. LB51(Ac) 2 -Cbz was ineffective at inhibiting the interaction between Hsp90 and Cyp40 (Figure 6(a)). LB63(Ac) 2 -Cbz bound to Hsp90 with an IC 50 =10 mM, which was not as effective as LB51 and LB63, which have IC 50 values of 4 mM and 2 mM, respectively. Thus, although LB63 was more effective than LB63(Ac) 2 -Cbz, the prodrug was still actively inhibiting this PPI interaction.
Competitive binding assays are useful strategies for evaluating how effectively a molecule binds to a protein target. We used this assay in order to validate the data collected in our Alpha assay and to evaluate how effectively our molecule bound to Hsp90. By incubating LB51-biotin tag (LB51-Tag) with Hsp90, and then adding increasing amounts of LB51(Ac) 2 -Cbz and LB63(Ac) 2 -Cbz to the reaction, we were able to assess how much Hsp90 was removed from LB51-Tag ( Figure 6(b)). An effective inhibitor would remove a significant amount of Hsp90, and an ineffective inhibitor would not compete effectively for Hsp90 protein. Western blot analysis of how much Hsp90 remained bound to the LB51-Tag assessed the efficacy of the compounds. Similar to the Alpha binding assays (Figure 6(a)), the competitive binding assays using Hsp90 and LB51-Tag showed that LB63(Ac) 2 -Cbz is effective at binding to Hsp90 with an IC 50 of $20 mM in this assay ( Figure  6(b)). As observed in the Alpha assay, LB51(Ac) 2 -Cbz was not effective at binding to Hsp90 and competing off of LB51-Tag.
Thus, a lead molecule, LB51, which binds to Hsp90 and inhibits co-chaperones from interacting with Hsp90, was modified using three prodrug approaches. Our first approach involved producing cell permeable analogues of LB51 with Oand N-methyl masking groups. However, these groups could not be removed in vitro with liver microsomes or in vivo with a mouse. Liver microsomes successfully removed the acetyl groups on serine and tyrosine side chains, however, they were unsuccessful on lysine, which led to the design of a new generation of acetylated prodrugs. However, generating these prodrugs proved to be synthetically more challenging than methylated prodrugs with Oand N-methyl moieties, given the instability of the acetyl groups.
Prodrugs of LB51 and LB63 with acetyl moieties on serine and tyrosine were produced. Removal of the orthogonal protecting group, Cbz, from the lysine side chain was successfully  accomplished, however, the molecules rapidly degraded upon reaction completion. Given the challenges, we opted for a third approach, which was to evaluate how effectively these prodrugs that contained Cbz, LB51(Ac) 2 -Cbz and LB63(Ac) 2 -Cbz, bound to Hsp90 and blocked the interaction with Cyp40 or competed for Hsp90 versus LB51. LB63(Ac) 2 -Cbz was the most effective molecule with IC 50 of 10 mM and $20 mM in the Hsp90-Cyp40 and Hsp90-LB51-Tag binding assay, respectively.

Conclusions
In conclusion, our findings have implications for the development of peptide prodrugs. First, the binding affinity of a prodrug is dramatically impacted by even small changes to the peptide side chains. Oand N-alkylations on the polar side chain of LB51 increased cell permeability from P app from 3.3 to 5.6, but these compounds no longer bound effectively to Hsp90 (IC 50 ! 10 mM).
Molecules that maintained the acetyl moieties on serine and tyrosine were also ineffective at binding to Hsp90 with the IC 50 > 10 mM Thus, we conclude that traditional strategies such as alkylations used to mask polar side chains on small molecules do not work well on cyclic peptide structures. Despite dealkylation occurring in a predictable manner when liver microsomes are used on small molecules, peptide-based prodrugs that contain alkylated side chains are not metabolised under the same conditions. We believe this is because the side chains must be inaccessible to the metabolic enzymes that cleave these structures where the enzymes because of their positioning in a 3-D cyclic peptide.
We also demonstrated that using acetyl moieties as masking groups was more effective than methyl moieties because they were more easily removed from peptides. However, acetyl moieties are labile, which increases synthetic difficulty. Another conclusion that can be drawn from this study is the success of monomethylated amines (LB97). The permeability of LB97 was greater than the free lead LB51, while the binding affinity was similar. Thus, the optimal prodrug would involve mono-methylated amine side chains and acetylated hydroxyl side chains, where the acetyl groups would likely be removed in cells, and the methylated amine would attenuate the polarity of the amine without blocking Figure 6. (a) The impact of LB51 and LB63 versus LB51(Ac) 2 -Cbz and LB63(Ac) 2 -Cbz on the Hsp90-Cyp40 interaction. Experiments were conducted in triplicate and data represent mean ± SEM of at least two independent experiments. (b) Competitive assays where LB51-Tag is competing for binding with LB51(Ac) 2 -Cbz, or LB63(Ac) 2 -Cbz to the C-terminal domain of Hsp90. Experiments were conducted in duplicate and data represent mean ± SEM of two independent experiments. the binding affinity. Studies examining these findings are ongoing and will be reported in due course.