The effect of helix-inducing constraints and downsizing upon a transcription block survival-derived functional cJun antagonist

Summary Inhibition of cJun is established as a promising therapeutic approach, particularly in cancer. We recently developed the “transcription block survival” (TBS) screening platform to derive functional peptide antagonists of transcription factor activity by ablating their ability to bind to cognate DNA. Using TBS, we screened a >131,000-member peptide library to select a 63-mer peptide that bound cJun and prevented 12-O-tetradecanoylphorbol-13-acetate response element (TRE) DNA binding. Iterative truncation was next combined with a systematic exploration of side-chain cyclization to derive a minimal active sequence. The resulting dual lactamized sequence was >40% smaller and retained low nM target affinity (equilibrium binding constant [KD] = 0.2 versus 9.7 nM), with 8 residues at the acidic region required for functional antagonism. However, even modest C-terminal truncation resulted in functional loss. The peptide functionally antagonizes cJun (half-maximal inhibitory concentration [IC50] = 13 versus 45 μM) and is considerably more stable in human serum relative to its non-lactamized counterpart and HingeW.


INTRODUCTION
Increasingly, intracellular protein-protein interactions (PPIs) are being targeted by peptide-based modulators for therapeutic effect in a range of diseases. [1][2][3][4][5] In particular, peptides can target broader and flatter PPI surfaces, which were previously disregarded as undruggable because they are typically intractable to smaller molecules. Thus, peptides and their mimetics represent a route to therapy in many diseases, where relevant target proteins are mutated, misfolded, upregulated, or overexpressed and therefore produce detrimental outcomes. To overcome potential shortcomings of peptide therapeutics, a range of strategies have been developed and added to the toolkit to produce more stable and bioavailable molecules. [6][7][8][9][10][11][12][13][14] Via alteration of the peptide sequence, chemical modification, cyclization, and introduction of non-natural backbone or side-chain components, a wide range of desirable features can be imparted such as improved binding, biostability, and cell penetration. A fundamental step in the development of therapeutic peptides is downsizing toward the smallest functional unit required to retain effective binding to produce molecules that can be efficiently synthesized. 15 cJun is a member of the activator protein-1 (AP-1) family of dimeric transcription factors that is implicated in a wide range of diseases including cancer, diabetes, and arthritis. [16][17][18][19] Constituent AP-1 proteins bind specific DNA recognition sites via their basic-leucine zipper (bZIP) domains. The AP-1 bZIP consists of a leucine zipper (LZ) domain for dimerization and a DNA-binding domain (DBD), which inserts into the major groove, where it recognizes the 12-O-tetradecanoylphorbol-13-acetate response element (TRE) via a specific network of electrostatic and hydrogen bond interactions ( Figure 1A 20 ). 21,22 In this work, we optimize a transcription block survival (TBS) assay-derived hit peptide, HingeW, 23 through sequential truncation combined with systematic screening of side-chain cyclisation.
During TBS, the coding region of the essential gene DHFR is mutated to incorporate TRE sites such that production of cJun inside E. coli produces a steric block (Figure 1B). RNA polymerase (RNAP) is then unable to transcribe the DHFR gene, (B) Cartoon illustrating the TBS assay whereby TRE DNA sites are incorporated into an essential gene to facilitate a cJun-induced transcriptional block, which can only be removed by cJun/TRE DNA interaction antagonists. This produces an assay with cell survival as a readout, which selects for peptide antagonists capable of binding to the target and preventing its interaction with DNA. (C) AlphaFold-Multimer (run on the Google Colab platform, v.2.1) prediction for the structure of the cJun-HW1 heterodimer, which illustrates the extended coiled-coil interaction between the antagonist with the full length of the bZIP domain, facilitated by the rationally designed N-terminal acidic extension. 20 The TBS winner peptide optimization process then involves truncation and sidechain lactamization.
leading to loss of the ability to reduce dihydrofolate and, ultimately, cell death. Thus, the cJun-induced steric block to RNAP can only be removed upon introduction of a functionally active cJun/TRE antagonist. The growth rate of a particular cell is intrinsically linked to the ability of a peptide library member to remove the transcriptional block and therefore to restore DHFR activity, allowing direct competition between library members until a single TBS winner sequence is selected. The library design template combined a previous library-derived LZ antagonist with a rationally designed N-terminal acidic extension. [24][25][26] The latter was first proposed by Olive et al. to extend the coiled-coil PPI from the LZ into the DBD, thus preventing DNA binding. The N-terminal design incorporated Leu residues at putative heptad d positions and a significant number of negatively charged amino acids via rational replacement of Arg/Lys to form favorable electrostatic interactions with basic residues within the cJun DBD, thus outcompeting the native interaction with TRE DNA. In our study, a central tract of residues across the acidic extension and LZ in this parent peptide sequence were semi-randomized, and the >131,000-member library was screened using the TBS platform, resulting in the selection of HingeW as an optimized cJun antagonist capable of both target binding and, more importantly, ablation of DNA binding. 23 The binding epitope of HingeW is presented on one side of a single a-helix, and, as such, target binding requires the peptide to adopt this secondary structure. However, as an a-helical peptide is progressively downsized, this tends to reduce the helicity as both the interaction interface and extended internal hydrogen bonding network become reduced and water competes for these interactions, shifting the folding equilibrium toward a random coil. Peptide stapling is a common methodology utilized to increase the a-helicity of peptides by the covalent linking of amino acid sidechains, typically via i to i+4 or i to i+7 linkages. [6][7][8]11,14,27 For i/i+4 stapling in coiled-coil antagonists, linkages are typically placed on the opposing solvent-exposed face of the helix at b-to-f (i.e., within one heptad) or f-to-c positions (i.e., spanning adjacent heptads) to avoid interference with the binding epitope. [28][29][30] A range of synthetic methodologies have been utilized for this purpose, including lactamization, all hydrocarbon linkers, disulphide bridges, and various thioethers, with one study indicating b-to-f (K/D) lactam bridges as the most effective at inducing helicity in short alanine pentapeptides. [6][7][8]14,31 Furthermore, lactamization provides additional benefit in terms of biostability since proteases universally recognize b-strands, with the constraint providing a further steric block, denying access to the backbone. 32 However, sequence-specific effects influence lactamization, and therefore, potential points throughout each peptide must be tested to discover those locations that maximally improve target binding affinity. 30 The process of peptide downsizing combined with the introduction of helix-inducing lactam constraints described here has produced a range of short peptides with a >100-fold range in efficacy and an optimized peptide, which retains nanomolar affinity while removing >40% of the parent sequence ( Figure 1C).

RESULTS
Partial truncation of the acidic domain significantly reduces peptide size while retaining activity The precise nature of the interaction between the cJun DBD and the rationally designed acidic domain of HW1 is unknown. HW1 was, therefore, iteratively truncated from the N terminus (HW2-6) to investigate the effect on cJun binding and cJun/TRE DNA antagonism (Table 1). Peptide fraction helicity (fH) was determined by quantifying the circular dichroism (CD) signal at 222 nm of peptide-only samples (Figure S1). Thermal denaturation experiments were then used to determine the melting temperature (T m ) of peptide-cJun heterodimers, which serves as an approximate measure of target engagement. CD was also used to investigate the functional activity of peptides in antagonizing the cJun/TRE DNA interaction. The TRE-DNA construct used produces a positive CD peak at $281 nm, which decreases in intensity upon cJun binding in a concentration-dependent manner. 23 cJun does not absorb at this wavelength, which allows a direct measurement of DNA structure, and the shift is only observed for DNA that contains the TRE site ( Figure S2). Further, antagonist peptides were shown to have no effect on TRE DNA structure ( Figure S3). This provides a clear and direct measurement of the proportion of DNA bound to cJun. As antagonist peptide concentration was increased, the peak shifted to overlay with that of free TRE-DNA ( Figure S4). For each peptide, the data were fit to a Hill equation to determine half-maximal inhibitory concentration (IC 50 ) values, specific to these assay conditions, providing a clear comparative measure of antagonism. The same 60 amino acid cJun construct, encompassing the entire bZIP domain, was used for all experiments regardless of antagonist peptide length.
N-terminal truncations in the series (HW2-6) sequentially reduced target binding and antagonism, indicating that the full length of the acidic extension contributes to peptide activity ( Figure 2; Table 1). However, each truncation in the series resulted in an increasingly large effect on the antagonism per residue removed. Inspection of the iterative truncations from HW1/HW2, from HW2/HW3, and from HW3/HW5, representing full heptad deletions, reveals antagonism decreases of 1-, 2-, and 2-fold, respectively. Further, the final N-terminal truncation investigated (HW5/HW6), which reduces the peptide down to its LZ only, reduced antagonism 1.7-fold despite only removing four residues. The acidic domain alone (HW7) was shown not to bind, indicating that the LZ domain is an absolute requirement for binding that operates synergistically with the acidic extension of HingeW to bind and antagonize cJun. Removal of the full acidic extension domain (HW1/HW6) reduced cJun/TRE DNA antagonism 9.7-fold. Previously, we observed a highly significant reduction in peptide activity in the TBS assay from peptides with IC 50 values similar to that observed for peptide HW6 (129.8 mM), meaning that HW6 is unlikely to be unable to fully outcompete cJun/TRE DNA binding. This supports the design rationale of the acidic extension and suggested that HW5 (IC 50 = 78.2 mM, 5.83 lower antagonism than HW1 and 1.73 better than HW6) may be considered the maximum viable truncation from the N terminus to be taken forward. Importantly, this extends beyond the LZ toward the DBD and may be important in binding and blocking the target from DNA binding.
N-terminal truncation also produced sequential increases in peptide fH, rising from $27% for HW1 to $47% for HW5, indicating that deleted regions have reduced peptide helicity ( Table 1). The LZ domain of HW1 has four point mutations from its parent sequence, FosW, which is known to homodimerize. 24 The negative charge of the acidic extension produces electrostatic repulsion and therefore decreases the propensity to homodimerize as indicated by comparing thermal denaturation curves for the peptides alone where HW5 displayed a T m of 41 C, whereas a T m could not be determined for HW1 as it was significantly lower, and thus the full sigmoidal transition was undefined. However, further truncation from HW5/HW6 reversed this trend, reducing fH to $38% while removing one acidic and three hydrophobic residues, leaving the LZ-only domain.
Characterization of HW1 HingeW (HW1) was developed to bind across the full cJun bZIP binding surface for more effective functional antagonism of TRE binding, relative to DBD-only or LZ-only cJun inhibitors. 23 The nature of the broad, shallow helical binding surface supports the use of longer peptides such as HW1. However, it was unclear whether the full length of the sequence is required to achieve functional antagonism. HW1 was recombinantly produced and biophysically characterized as an N/C-capped (MAS at the N terminus, GAP at the C terminus) and C-terminally 6xHis-tagged protein construct of 69 amino acids in length, with significant negative charge throughout. These characteristics are particularly unfavorable in a potential therapeutic molecule, even among peptides, and as such, optimization of the sequence has been considered and developed here.
Lactamization of truncated peptides increases helicity and activity HW5 was optimized by incorporating i/i+4 (K-to-D) lactam bridges through the use of orthogonal-protecting groups (Lys(Mtt) and Asp(O-2-PhiPr)), which can be selectively deprotected (2% trifluoroacetic acid in DCM) and reacted using typical solidphase chemistry while the peptide is still attached to the resin. The success of the reaction was confirmed using mass spectrometry (MS) to observe the decreased mass from the loss of a water molecule, compared with the linear unreacted peptide ( Figure S5). This intramolecular condensation reaction induces peptide a-helicity by covalently linking residues held in proximity by secondary structure despite their distance in the primary structure. In this work, lactams were only introduced at solventexposed b-to-f or f-to-c heptad positions to prevent disruption of the binding epitope of the antagonist helix. Point mutations to the sequence were required to incorporate the bridging K and D residues, with both linear (HW10, HW12) and cyclized (HW11, HW13) versions of each sequence produced by split batch synthesis.
Cyclized peptides HW11 and HW13 increase antagonism relative to the parent linear sequence HW5 1.8-and 1.9-fold, respectively, with both increased peptide ll OPEN ACCESS fH and more favorable target binding (Table 1). It is interesting to note that the binding indicated by T m values of $67 C and $68 C for these peptides do not produce correspondingly low IC 50 values. By plotting the T m versus IC 50 of the peptides in this study, a clear inverse correlation is observed (r 2 = 0.77); however, there is variability (Figure 3). HW2 and HW13, for example, have similar T m values, but HW2 antagonizes the cJun/TRE interaction 43 more effectively. HW13 has been truncated by a further 14 N-terminal residues than HW2. This supports the importance of the rational design principle used for HingeW whereby inhibition of both domains of the cJun bZIP produces the most effective antagonism, as the majority of the acidic extension has been removed in peptides HW5 and HW10-13, with it all removed in peptide HW6. This results in peptides that can be optimized to bind tightly to the cJun LZ but are limited in their ability to functionally antagonize cJun by also blocking the DBD. A crucial point is that any data points below the fitted line have an IC 50 better than predicted by their T m and, as such, may be considered the most important in this study, with functional antagonism the goal of this work.
C-terminal truncation has a greater impact on functional peptide activity than does N-terminal Optimization was therefore best considered by working from HW4, which has additional N-terminal negative charge and 1.73 higher antagonism than HW5, as the above optimization of antagonism may have been limited by further truncation at the N terminus. This implicates the importance of the region in interacting with the corresponding positively charged cJun DBD surface, as well as inducing helicity and potentially stabilizing the dipole of the molecule with a negatively charged residue at the N terminus. [33][34][35] Following this, truncation at the C terminus was considered by the production of HW8, which utilized the ND20 truncation of HW4 while also truncating at the C terminus. The removal of four C-terminal residues (10% of the molecule) from HW 4 to HW8 reduced antagonism 1.83, but further truncation at the C terminus to produce HW9 vastly reduced antagonism 14.83 compared with HW8 and 26.73 compared with HW4. Attempts to optimize HW9 by lactamization to produce HW15 and HW17 were effective in that they significantly improved antagonism; however, they still produce 8.83 and 10.43 reductions in antagonism relative to HW1 (Table 1). Although these lactamizations produced a much larger impact on the peptides than the other cyclizations in this study, the resulting peptide IC 50 values remained too high for further consideration.
Systematic lactonization across the peptide sequence identified amenable cyclization sites, leading to the production of the optimized double lactam peptide HW30.
Based on this understanding of the effects of N-and C-terminal truncations, we focused our efforts on HW8 as a scaffold for further optimization; this peptide is >40% smaller than HW1 while retaining a high level of functional activity (IC 50 = 82.1 mM). i/i+4 (K-to-D) lactam bridges were systematically incorporated at sites throughout the sequence to investigate which regions were most amenable to the helix constraint and which produced improvements in affinity and inhibition. Again, due to point mutations, to introduce the bridging K and D residues, both linear and cyclized peptides were produced. The cJun/antagonist heterodimer change in T m (DT m ) due to lactamization ranged from $2 C for HW20/HW21 to $9 C for HW28/HW29 (Table 1). It is also useful to compare with the parent sequence, before the substitutions for K and D residues, where the heterodimer DT m ranges from $1 C for HW8/HW19 to $8 C for HW8/HW29. cJun/TRE antagonism is the most important measure in this work, however, and only HW29 produced a significant improvement in IC 50 , shown to be 1.63 lower than for HW8 (p = 0.003). Using a lactam constraint, HW29 was able to restore the reduction in antagonism caused by truncating the four C-terminal residues from HW4/HW8. The change in fH of the peptides ranged from $1% for HW18/HW19 to $8% for HW28/ HW29, and there is a correlation between peptide helicity induced by lactamization at a particular site and the increase in cJun/TRE antagonism observed.
Although the lactamisation within HW23 did not lead to a significant increase in antagonism (p = 0.24), it did produce the second lowest IC 50 value of the lactamized peptides in this series, a $5% increase in peptide helicity, and $9 C increase in cJun heterodimer T m relative to its non-lactamized counterpart (HW22). Due to the distance between this site and the lactamization site of HW29, a double lactamized peptide was produced, HW30. The double lactamization produces a peptide that is $4% more helical, with no significant change in heterodimer T m but a significant decrease in the value (44.9 G 2.1 mM, p = 0.02 versus HW29), compared with the most effective single lactam peptide HW29 (Table 1).

Peptide lactamization increases target affinity by increasingly improved enthalpy and an increasingly unfavorable entropic component
A selection of peptides from this work were investigated by isothermal titration calorimetry (ITC) to quantify the thermodynamic parameters of their interaction with target cJun (Figure 4 and S6; Table 2). The determined equilibrium binding constant (K D ) values follow from the CD data above and can be rationalized in terms of the truncation or lactamization they represent. The HW4/cJun affinity is 4333 lower than HW1/cJun due to the removal of 20 residues from the N terminus. Further, removal of four residues from the C terminus to give HW8 produces a further 123 reduction in affinity. From the maximally truncated HW8 peptide, we observe 163 and 133 improvements in affinity for HW23 and HW29, respectively, due to the introduction of lactam bridges. Crucially, the double lactamized HW30 affinity has ll OPEN ACCESS increased 1133 compared with HW8 (K D = 9.69 G 3.95 nM). All interactions investigated were dominated by the enthalpic component, with a smaller unfavorable entropic contribution, other than for HW8/cJun, which is the weakest interaction. The unfavorable entropic component was largest for the HW30/cJun interaction and smallest for HW23/cJun. This indicates the introduction of lactams produces a more unfavorable entropic component, perhaps a surprising result as helix-inducing lactamization is typically considered to increase affinity through entropic pre-organization. However, the larger increase in the favorable enthalpic component produces an increase in binding affinity due to lactamization. This is likely to be caused by a combination of enhanced intermolecular contacts and ordering that are opposed by unfavorable desolvation effects.

Lactamization increases peptide serum stability
To further explore the effect of truncation and lactamization, HW1, HW8, HW29, and HW30 were tested for stability in human serum ( Figure 5). cFos was included as a linear control peptide, which was the wild-type parental design template on which HW1 was based prior to optimization. [23][24][25] The linear peptides (HW1, HW8, and cFos) were fully degraded over the 3-day experimental time course. In contrast, at this endpoint, the peak intensities of HW29 and HW30 were still 31% and 56% of their starting values, respectively. These values demonstrate considerable resistance to protease activity over significant time courses relative to the linear HW8  counterpart. Moreover, the effect is cumulative, with HW30 more stable than HW29 by virtue of the second lactam constraint (p = 0.039).

DISCUSSION
We have rationally designed and synthesized a series of peptides based upon the TBS-derived, cJun antagonist HingeW (HW1). 23 Our overarching aim was to systematically study the effect of truncation and lactamization on this peptide to enable the production of an optimized antagonist that retains the parent peptides activity while improving its drug-like characteristics. We separated and individually produced the acidic (HW7) and LZ (HW6) domains of the parent peptide to illustrate their individual contributions to peptide efficacy. HW6 was shown to antagonize the ability of cJun to bind to TRE DNA (IC 50 = 129.8 G 13.0 mM), while the acidic peptide HW7 did not. However, combining the two regions in HW1 resulted in enhanced antagonism (IC 50 = 13.4 G 0.6 mM). This suggests a mechanism whereby the antagonist LZ binds to the cJun LZ, holding the acidic domain in proximity to the cJun DBD to block the contacts between cJun residues and DNA bases and facilitate higher affinity binding. Further, it is not simply a steric block, since the peptide is highly negatively charged in this region, which will electrostatically repel the like-charged DNA.
The iterative truncation of the acidic domain in peptides HW1-HW6 showed that the full domain contributed to the peptide's inhibitory effect but that the more N-terminal regions contributed less relative to further truncations in this region. An important consideration in the maximum acceptable truncation is whether a peptide binding to cJun will be able to outcompete TRE DNA. This cJun homodimer/TRE DNA tertiary interaction has been variously established as having a K D of $100-200 nM, so maintaining a stronger interaction than this was considered an important benchmark. 22 HW4 was shown to have a K D of 91.1 G 24.8 nM, which places it on the threshold and means that this ND20 truncation was considered the maximum acceptable truncation that retained sufficient functional activity. The degree of

OPEN ACCESS
truncation at the N terminus that is possible while retaining functional antagonism requires considerations beyond binding affinity, as the acidic domain is crucial for blocking the cJun DBD from binding DNA. Effective antagonism might not require binding to the full length of the cJun DBD; however, peptides HW5, HW6, and HW10-13 are no longer able to effectively prevent DNA binding to the cJun DBD, i.e., cJun may still bind to TRE sites in monomeric form, while these antagonist peptides are bound to its LZ domain. [36][37][38] Specific contacts between particular cJun DBD amino acid side chains and TRE DNA bases are known, so it can be assumed that an antagonist that blocks any of these cJun residues will significantly reduce cJun/TRE binding. 21,22 The four residues added to HW5 to make HW4 extend far enough into the cJun DBD to directly block the C-terminal cJun DBD residues known to specifically interact with the TRE site.
Study of C-terminal truncation was less detailed, as the removal of residues here produced larger effects from smaller changes, so fewer stepwise truncations were required. Peptide efficacy was substantially reduced by the CD10 truncation, and helix-inducing lactamization was unable to restore the loss to an acceptable degree, as the benefits of truncating fewer residues do not outweigh the activity loss. The CD7 truncation, on the other hand, produced losses that could be restored by lactamization and thus were considered a useful optimization, which was further systematically studied.
The thermodynamic parameters of binding observed by ITC show what may be considered an unexpected result in terms of the entropic component. It is usually asserted that introduction of lactam bridges increases binding affinity by pre-organizing the peptide molecule into its helical structure, which can bind to the target with a reduced entropic penalty. The lactamized peptides investigated by ITC have an unfavorable entropic component, with the dual lactam HW30 having the largest. Further, the length-matched linear peptide HW8 has a favorable entropic component, so the cyclization-induced shift in entropy is clear. This illustrates that increased target binding affinity from side-chain cyclization can also occur due to improved enthalpic interactions.
Having shown that every truncation of HW1 produces a peptide with less in vitro efficacy, it is important to consider the crucial aspects of a peptide that may make it more suitable for further consideration as a therapeutic. The removal of 27 (in addition to the 6xHis tag) residues to produce HW30 along with the addition of two lactam bridges produce a peptide that is $22% more helical and persists in human serum at 56% after 3 days. This is a vast improvement compared with its linear counterparts, which were completely degraded, indicating the potential for favorable pharmacokinetics. Downsizing HW1 also presented significant practical utility as, due to its length, HW1 proved to be inaccessible via solid-phase peptide synthesis (SPPS) using standard techniques, with HW2 and HW3 producing poor yields. HW1 was produced as a 6xHistagged peptide by recombinant expression and purification from cell lysate, which is a more demanding production process than SPPS. The smaller peptides (HW4-30) were all produced in high yield via SPPS, illustrating the time and money that can be saved by downsizing peptides with therapeutic potential.
HW30 is also readily synthesized by SPPS, making it cheaper and quicker to produce. Cyclization of peptides produced recombinantly is either significantly more challenging or impossible, preventing the study of HW1 lactamization here. Although truncation of HW1 comes at the cost of a reduction in binding affinity with cJun (0.21 nM), the low nanomolar affinity of HW30 (9.7 nM versus 1 mM for the linear counterpart) is indicative of an extremely high-affinity interaction, especially when ll OPEN ACCESS Cell Reports Physical Science 3, 101077, October 19, 2022 considered in the context of small-molecule antagonists, which, in inhibiting PPIs of this type, typically display low micromolar affinities at best. 18 HW30 is a highly functionally active inhibitor of cJun/TRE DNA interaction that can be readily produced (facilitating further modifications such as cell penetrance tags) and is highly serum stable. Peptide-based therapeutics are increasingly being developed for the clinic, due to their compelling ability to inhibit PPIs, and as such, downsizing and lactamization are important components in the toolkit to realize these goals.

EXPERIMENTAL PROCEDURES
Resource availability Lead contact Further information and requests for resources should be directed to the lead contact, Jody Mason (j.mason@bath.ac.uk).

Materials availability
This study did not generate new unique reagents.

Data and code availability
All data underlying this study are available from the lead contact upon reasonable request.
Peptide synthesis and purification HW1 was recombinantly expressed and purified as described previously. 23 All other peptides were synthesized using a Liberty Blue microwave peptide synthesizer (CEM) at a 0.1-mmol scale on ChemMatrix Rink amid resin using standard Fmoc solid-phase methodology. Coupling was performed using 53 amino acid, 4.53 PyBOP, and 103 diisopropylethylamine in dimethylformamide (DMF; 5 mL). Deprotection was performed using 20% piperidine in DMF. Peptides were capped at the N terminus by a final reaction with 33 acetic anhydride and 4.53 diisopropylethylamine in DMF for 5 min at 90 C. For lactamized peptides, the relevant K and D positions were orthogonally protected using Lys(Mtt) and Asp(O-2-PhiPr). The side chains of these residues were selectively deprotected by washing the resin with dichloromethane (DCM) 33, 2% trifluoroacetic acid (TFA) in DCM 310, DCM 33, then DMF 33. The newly deprotected side chains were coupled in PyBOP (1 mL), diisopropylethylamine (1 mL), and DMF (3 mL) for 5 h at 60 C. The resin was dried, and the same reagents added for a second reaction for 16 h at 60 C. Incubation in a cleavage mixture (95% TFA, 2.5% triisopropylsilane, 2.5% H 2 O, 10 mL) for 4 h at room temperature cleaved the peptide from the resin and removed side-chain-protecting groups. The resin was removed by filtration, and cleaved peptides were precipitated in diethyl ether at À80 C and centrifuged. This pellet was washed a further four times with diethyl ether before it was dried overnight at room temperature. Peptides were resuspended in 1:1 water:acetonitrile before purification using reversed phase high-performance liquid chromatography (RP-HPLC) with a Jupiter Proteo column (4-mm particle size, 90-A pore size, 250 3 10 mm; Phenomenex) using a water:acetonitrile gradient (0.1% TFA). Peptide masses and purity (>95%) were verified by electrospray ionization MS.

CD
An Applied Photophysics Chirascan was used for CD measurements, with a 200-mL sample in a 1-mm path length CD cell. Protein/DNA samples were suspended in 20 mM potassium phosphate, 150 mM potassium fluoride, and 2 mM TCEP (pH 7.4). cJun was added to DNA before addition of antagonist peptide, and these samples were equilibrated for 30 min before measurement. For full spectra, three scans between 190 and 260 nm (265-320 nm for DNA-binding experiments) were ll OPEN ACCESS collected with a bandwidth of 1 nm and data sampled at a rate of 0.5 s À1 . These scans were averaged and converted to molar residue ellipticities (MREs). Thermal denaturation experiments were performed by measuring the ellipticity at 222 nm over a 1 C to 90 C gradient at 1 C increments. Post-melt scans at 20 C confirmed the transitions were reversible, as they overlaid within 10% of the pre-melt scan. The resulting thermal denaturation curves were converted to MREs and fitted to a two-state model, derived via modification of the Gibbs-Helmholtz equation to determine the T m . 39

ITC
Peptides were studied by ITC using a Microcal PEAQ-ITC (Malvern Instruments) using an ITC buffer consisting of 20 mM potassium phosphate, 150 mM potassium fluoride, and 2 mM TCEP (pH 7.4). Two-mL injections of antagonist peptide at 25-200 mM were injected into the cell containing cJun at 2.5-20 mM at 25 C. Microcal Control and Analysis software were used to record and analyze the heat change upon addition and fit the data to a one-site binding model to extract the enthalpy change of binding (DH) and the K D , from which the free energy change of binding (DG) and the entropy change of binding (DS) were calculated. 40 Control experiments involved the injection of the antagonist peptide sample into the cell containing ITC buffer alone to determine the heat of dilution, which was subtracted. Thermodynamic parameters are presented as an average of two independent experiments with errors given as one SD.