A library‐derived peptide inhibitor of the BZLF1 transcription factor

Transcription factor dysregulation is associated with many diseases, including cancer. Peptide‐based molecules are increasingly recognised as important modulators of difficult intracellular protein–protein interaction targets, with peptide library screening consequently proven to be a viable strategy in developing inhibitors against a wide range of transcription factors (TFs). However, current strategies simply select the highest affinity of binding to a target TF rather than the ability to inhibit TF function. Here, we utilise our Transcription Block Survival (TBS) screening platform to enable high‐throughput identification of peptides that inhibit TFs from binding to cognate DNA sites, hence inhibiting functionality. In this study, we explore whether the TBS can be expanded to derive a potent and functional peptide inhibitor of the BZLF1 transcription factor. The library‐derived peptide, AcidicW, is shown to form a more stable dimer with BZLF1 than the BZLF1 homodimer, with a thermal denaturation temperature exceeding 80°C. AcidicW can also functionally inhibit the BZLF1:TRE DNA interaction with high potency and an IC50 of 612 nM.


| INTRODUCTION 1.| Peptide antagonists
Transcription factors (TFs) can become dysregulated in many diseases such as cancer, cardiovascular disease and autoimmune disease. 1,2hibiting TFs is an established therapeutic strategy with the potential for potent antagonism of function.4][5][6] In order to target such interfaces, some peptide inhibitors have been designed to dimerise with the target transcription factor, preventing the target from either homodimerising or heterodimerising into its active state. 7other approach has focused on designing mimics of the transcription factor dimers that sit on the cognate DNA sites, obscuring the target transcription factor from itself binding and affecting transcription. 7,8For example, the hybrid protein ME47 targets the E-box DNA consensus sequence, blocking the c-Myc:Max dimer from itself binding and has been shown to inhibit tumour growth in a xenograft mouse model. 9,10[13] In addition to rational design approaches, peptide inhibitors are often generated using library screening approaches, using up to 10 13 members, to facilitate the selection of high affinity (nM-pM) inhibitors.
5][16][17][18][19][20] Owing to the need to transform libraries into their screening environment, sizes are limited to around 10 7 members, however, PCA has the advantage over other library methods of being an in-cell screening platform and consequently moves beyond target affinity to select for soluble, nontoxic, target selective peptides that can resist breakdown by proteases.
2][23][24][25] However, a potential drawback is that screening is a function of target binding, rather than the ability to antagonise function. 20TFs can only impact upon transcription by binding to their cognate DNA recognition sites.Therefore, an improved TF screening platform is one that functionally antagonises the target by preventing DNA site binding.We previously reported our Transcription Block Survival (TBS) platform that screens for functional transcription factor antagonism e.g. of the TF cJun. 26re, we explore whether TBS can be expanded to other TF targets by the production of an effective BZLF1 inhibitor.

| Inhibiting the BZLF1 transcription factor using peptides
BZLF1 is a basic leucine zipper (bZIP) TF encoded within the Epstein-Barr Virus (EBV) that is responsible for inducing a transition from latent to lytic phase.EBV infection is associated with several cancers that include Hodgkin's disease and Burkitt's lymphoma. 27BZLF1 has F I G U R E 1 A: BZLF1 contains a bZIP domain that binds to the major groove of TRE DNA as a dimer via its basic region and has a hydrophobic C-terminal region, made up of a proximal 222-231) and distal (232-245) region, which is connected to the leucine zipper via a beta-turn (PDB ID 2C9L).B: The acidic region of A-BD CC is negatively charged and binds to the basic region of BZLF1 extending the binding interface and blocking the binding of the basic region to DNA.C: Constructs and library design.The a, g and e positions of different heptads are labelled and residue numbers of BZLF1 are given.TBS-selected library residues are underlined.The interactions between residues can be clearly seen in the coiledcoil diagrams in Figure 3. also been shown to bind to TPA responsive elements (TRE: TGA G / C TCA) that are widely known as AP-1 cognate binding sites. 28crystal structure of the BZLF1 dimer bound to TRE DNA was reported by Petosa et al, which demonstrated a protein topology that is common to many human bZIPs (Figure 1A, 1B).29 BZLF1 was found to bind the major TRE DNA groove as a homodimer with one member binding in a curved conformation, and the other in a relatively straight conformation.BZLF1 binds DNA via a basic region promoted by dimerisation via a C-terminal leucine zipper motif.However, this coiled coil is also shorter than many human bZIPs, being comprised of four heptads rather than five or more (Figure 1).30 The BZLF1 coiled-coil region also deviates in sequence from typical residues found at specific positions within these domains.Finally, the structure contains a short hydrophobic C-terminal tail that interacts with the coiled-coil and is connected via a beta-turn.It is comprised of a single-turn helix followed by an extended stretch of residues that are anti-parallel to the coiled-coil region and may function to stabilise it (Figure 1: yellow).
Previously, a BZLF1 peptide lacking the basic region was shown to antagonise BZLF1 DNA binding. 31A subsequent design sought to optimise the peptide by mutating ten sites (residues 191-209 within the coiled-coil) by using the CLuster expansion and integer linear programming-based Analysis of Specificity and StabilitY (CLASSY) protein design algorithm, which sought to optimise peptide-target interactions using molecular mechanics-based and experimentally-derived terms. 30,32Subsequent BZLF1 residues (210-221) also form a coiledcoil but were not optimised using the CLASSY algorithm as they also form interactions with the proximal C-terminal region (222-231) as seen in the crystal structure. 29The designed peptide was conjugated to the C-terminal proximal and distal region (residues 222-245) of BZLF1 to create the BD CC peptide (BZLF1 Design against the coiled-coil region). 31,33,34The proximal and distal regions were included in the BD CC peptide due to the role of this region in promoting dimerisation.
[37][38][39] The intended purpose of this acidic extension was to lengthen the coiled-coil interface to improve PPI contact area and therefore affinity while simultaneously mimicking DNA to prevent the basic region of the target from accessing its cognate DNA sequence.This acidic extension incorporated an N-terminal continuation of the coiled-coil into the basic region with leucine at all putative d positions.Further, Asn was inserted at two a positions to facilitate favourable a-a' hydrophobic interactions.Finally, Glu was placed into e/g positions to guide favourable electrostatic interactions with the basic region of BZLF1.Three different variants of the acidic extension have so far been developed, including an A variant which placed Glu and Lys residues at the C-terminal heptad g and a position respectively. 40e acidic extension was conjugated to the BD CC to extend the binding interface and increase the binding affinity. 30The A variant extension was chosen since this facilitated interactions at the C-terminal heptad of the acidic extension: a Glu-Arg g-e' interaction and a Leu-Leu a-a' interaction to create acidic BD CC (A-BD CC , Figure 1B, 1C).A-BD CC was found to bind strongly to BZLF1 with a melting temperature > 80 C. Since A-BD CC is the most potent peptide antagonist of BZLF1 reported in the literature, it was used as a starting point for our peptide library screen.

| Developing the TBS for BZLF1
We developed the TBS for BZLF1 by first probing the blocking of mDHFR transcription by BZLF1 and the restoration of cell growth by A-BD CC.BL21 gold E. coli cells (Stratagene) were transformed with 100 ng of each of the following plasmids; p300d WT-mDHFR or p300d TRE-mDHFR, the p230d BZLF1 bZIP plasmid or p230d BZLF1 ZIP plasmid and the pET24a plasmid or pET24a A-BD CC as a source of LacI.All the plasmid combinations used in this study are given in Supplementary Table 1.The p230d BZLF1 ZIP plasmid lacks the basic domain so is unable to bind TRE sites and was used as a control plasmid in place of the p230d BZLF1 bZIP plasmid.p300d TRE-mDHFR and WT-mDHFR constructs were used as previously reported. 26The p230d bZIP BZLF1 construct encoded residues of 175-245 of BZLF1 and the ZIP BZLF1 construct encoded residues 190-245.pET24a construct sequences were identical to the expression constructs reported below.
Colonies from these transformations were picked and grown overnight at 37 C and 200 RPM.The overnight cultures were pelleted and resuspended in M9 media and diluted to an OD 600 of 0.5.
1,000 μl of this dilution was then plated on M9 plates with 8 μM TMP and 1mM IPTG.Sealed plates were incubated at 37 C for 72 hours.These dilutions were also plated on 8 μM TMP-only plates with no IPTG and gave no colonies under the same conditions to demonstrate that cell survival was dependent on the expression of TRE-mDHFR as previously reported (Supplementary Figure 1).Each plasmid combination was plated in triplicate with the colonies counted and the mean colony number per plate calculated.Escherichia coli competent cells (Agilent Technologies).Cells were grown in LB media at 37 C and 250 RPM until an OD 600 of 0.6-0.8 was reached, and then expression was induced using 1mM IPTG and grown for 4 hours at 250 RPM.Cells were then pelleted using centrifugation at 5000 RPM at 4 C for 20 minutes ready for purification as described below. 30otein constructs used in this study are:
For the purification of A-BD CC and AcidicW proteins, pellets derived from 500 ml culture were lysed in 35 ml 12.5mM sodium phosphate, 1 M NaCl and 3mM DTT pH 7.4 buffer, with five cOm-plete™ Mini EDTA-free protease inhibitor cocktail tablets.Lysis was performed via sonication using a power of 14 at 30 seconds on, 30 seconds off intervals.The lysates were then cleared using centrifugation at 12000 RPM for 40 minutes.This cleared lysate was run on a 5 ml HisTrap FF Ni-NTA column.Eluted fractions containing the protein of interest were further purified using size exclusion chromatography with a Hiload 16/600 Superdex 75 g column, in a gel filtration buffer of 12.5 mM sodium phosphate, 150 mM NaCl, 1 mM DTT and 0.25 mM EDTA pH 7.4.SDS-PAGE and mass spectrometry analysis confirmed protein purity and identity.
For the purification of BZLF1 protein, an identical method was used, except 6 M guanidine hydrochloride was added to all buffers aside from the buffer used for the gel filtration step, and the cleared lysates were instead incubated with 4 ml Ni-NTA beads (Amintra ab270549) per 500 ml bacterial culture for 90 minutes at room temperature.Ni-NTA beads were washed three times using 50 ml of 12.5 mM sodium phosphate, 1 M NaCl, 1 mM DTT, 30 mM imidazole pH 7.4, and eluted in 12 ml of 12.5 mM sodium phosphate, 1 M NaCl, 1 mM DTT and 300 mM imidazole pH 7.4.Following elution from the Ni-NTA beads, BZLF1 was dialysed into gel filtration buffer overnight at 4 C in order to remove the guanidine hydrochloride and refold the protein.SDS-PAGE and mass spectrometry analysis confirmed protein purity and identity.

| Library construction
The libraries were constructed and cloned using approaches that had been previously reported. 17,18Briefly, libraries were cloned using overlapping extension PCR with degenerate primers (forward: The PCR product was digested using NheI and SacI restriction sites shown above in bold and cloned into a cut pET24a A-BD CC plasmid incorporating an internal SacI restriction site.The library was harvested using LB media and grown for 45 minutes at 37 C and combined with a 50% glycerol solution in a 1:1 ratio and flash frozen.The glycerol stock was then grown for 2 hours at 37 C and miniprepped.Library DNA quality was assessed by sequencing the whole DNA pool to show correct placement of degenerate codons, as well as a number of single colonies to show a high diversity of library members from single colonies.pET24a plasmid encoding the library was then transformed into electrocompetent BL21 gold cells already containing the p300d TRE-mDHFR and p230d BZLF1 plasmids.To ensure the full acidic peptide library was covered in these electrocompetent BL21 gold cells, 50 μl of the transformation was plated and colonies counted to ensure that in the remaining 950 μl of the library transformation, the library was fully covered using Equation (1).A total of 369,808 colonies were counted for a library size of 36,864, ensuring 99.9% coverage.
Equation ( 1) was used to calculate the percentage of missing library members where E is the percentage of the library missing, m is the number of colonies collected and n is the library size.

| Single-step selection
The remaining 950 μl of the library transformation was grown overnight at 37 C in 20 ml of LB media with chloramphenicol, ampicillin and kanamycin antibiotics added.Cells were then pelleted and resuspended in M9 media.Cells were then diluted to OD 600 of 0.5. 2 ml of the diluted library was plated onto each M9 agar plate with 8 μM TMP and 1 mM IPTG as previously optimised to return colonies.

| Competition selection
After an initial single-step selection on M9 plates with 8 μM TMP and 1 mM IPTG, colonies were pooled and the selection continued via a growth competition selection in M9 media with chloramphenicol, ampicillin, kanamycin, 16 μM TMP and 1 mM IPTG used to increase the selection stringency as previously used for PCA assays. 24A 16 μM TMP-only control flask with no IPTG was used for all passaging steps to make sure no growth was observed in the absence of IPTG and to therefore ensure all growth in the library selection flask was driven by expression of the mDHFR protein.Cultures were grown at 37 C and 200 RPM and the selection occurred over two liquid passages in total.
The culture was diluted between each passage with 1,200 μl of culture with an OD 600 of 0.5 added to 50 ml fresh M9 media with chloramphenicol, ampicillin, kanamycin, 16 μM TMP and 1 mM IPTG added.
This culture was then grown until an OD 600 of 0.5 was reached, which took approximately three days before moving on to the next passage.
After each passage, 20 ml of the pool was miniprepped and sent for sequencing to confirm current library populations.A total of 10 μl of the pool was also plated on M9 agar plates with chloramphenicol, ampicillin, kanamycin, 8 μM TMP and 1 mM IPTG and individual colonies were grown, miniprepped and sent for sequencing to identify individual library members.The passaging stage finished once a single library sequence had been selected from the library.

| Fluorescence polarisation assay
For the direct fluorescence polarisation binding assay, 10 nM 5'[6FAM]GCTTGATGACTCAGCCGGA (TRE motif underlined) annealed to its reverse complement (also labelled at 5 0 ) was incubated at room temperature for 30 minutes with a serial dilution of BZLF1 in 12.5 mM sodium phosphate, 150 mM NaCl, 1 mM DTT and 0.25 mM EDTA pH 7.4 buffer prior to reading.The binding assay was based on a previously reported Fluorescence Polarisation binding assay developed for Activator Protein-1. 42The data were fitted in GraphPad Prism 9.0 to a one-site binding model using Equation (2).We recently reported TBS as a platform screening technology that screens for functional transcription factor agonism to target the oncogenic TF cJun. 26Since TFs must bind to their target DNA sites to instigate transcription, inhibitors of TF-DNA binding must be functional.TBS is therefore able to derive functional inhibitors by directly linking the ability of a peptide library member to inhibit TF-DNA binding to bacterial growth of the cell it is expressed within.The most potent inhibitors of TF activity led to the fastest bacterial growth, allowing them to outcompete other library members in direct competitive selection.
This approach works because when the library is transformed into the bacteria, each colony represents a single library member and colonies are counted to ensure library coverage so the library is not required to get into a single cell.The cell that has the greatest expression of TRE-mDHFR should have the most potent BZLF1 inhibitor from the library and will outgrow other cells containing other library members in the passaging step to become selected for the TBS assay.
This approach sits in contrast with current peptide screening methods that select the peptide that is best able to bind to the target, which may not necessarily translate into the best functional inhibitor, slowing progress and wasting valuable resources on ineffective inhibitors.
The link between TF binding and bacterial growth results from the manipulation of the exogenous mDHFR, which is essential for bacterial survival in the presence of TMP.In TBS, TRE bZIP binding sites are placed into the mDHFR gene (Figure 2A, Step 1).The target bZIP transcription factor is next added, blocking RNA polymerase transcription of mDHFR by binding to TRE sites, preventing bacterial growth (Figure 2A, Step 2).When a peptide inhibitor is introduced that targets the TF to effectively inhibit bZIP-DNA binding, the mDHFR gene is transcribed, restoring bacterial growth (Figure 2A, Step 3).This allows peptides that are best able to inhibit bZIP binding to target DNA sites to outgrow weaker inhibitors, resulting in enrichment.As TBS occurs within cells it will tend to select for peptide library members that are soluble, non-toxic and resistant to degradation by proteases. 4The assay was previously developed for cJun as a proof of principle.Here we explore the utility of TBS in the expansion of the screening technique to the TF BZLF1.

| Developing the TBS assay for BZLF1
We previously demonstrated that the TRE-mDHFR mutant was active using a NADPH turnover activity assay. 26In order to develop the TBS assay for BZLF1, the activity of the TRE-mDHFR was probed to test that it was active as previously reported. 26No colonies were able to grow on M9 plates with 8 μM Trimethoprim (TMP) but growth was demonstrated on M9 plates with 8 μM TMP and 1 mM IPTG indicating the TRE-mDHFR mutant was active (Figure 2, S1).
Next, the BZLF1 bZIP was introduced into the TBS assay to explore whether it was able to reduce bacterial growth by directly binding to TRE sites within the TRE-mDHFR gene and blocking transcription.A reduction in bacterial colony number was observed (Figure 2).No such reduction in colony number was observed in the absence of TRE sites when using WT-mDHFR.
Finally, A-BD CC was introduced into the system to probe whether it was able to restore bacterial growth through binding to BZLF1 and inhibiting its binding to TRE DNA sites.Restoration of bacterial growth was observed by a return of colonies.Confirming its ability to functionally antagonise BZLF1, A-BD CC increased the colony number in the presence of the BZLF1 bZIP construct (Figure 2).

| Library design
Following TBS validation for BZLF1, a peptide library was designed (Figure 3, Figure 1C).The library used A-BD CC as a design scaffold since this was the most potent peptide inhibitor of BZLF1 to date, and by offering alternative amino acid options to this template, we sought to explore other sequences that may offer improved functional BZLF1 inhibition.Previously, the BD CC peptide with the full C-terminal proximal and distal region (including residues 222-245) was shown to bind to BZLF1 with a T m 13 C higher compared to when only the proximal C-terminal region was present (residues 222-231). 30We therefore decided to use the full C-terminal proximal and distal region in our library, as is present in A-BD CC, to help ensure a high affinity for the target.Moreover, in the BZLF1 dimer, both the proximal and distal region of the C-terminal region were shown to be important for dimer stability using truncation mutations. 34We decided to use the full acidic extension of A-BD CC construct as reported by Chen et al, despite it being a residue longer than the target BZLF1 at the N-terminus, since A-BD CC is the highest affinity peptide for BZLF1 reported in the literature and we wanted to measure how we could improve the exact same construct using library design.
In addition, the additional N-terminal glutamine of A-BD CC sits in the f position and glutamine can form chain-backbone interactions with i + 5 residues so we were hesitant to remove this extra residue in case it supported any backbone interactions that could improve the interaction of the A-BD CC construct with BZLF1. 43The library was designed to cover the acidic region of A-BD CC , since the design of this region in A-BD CC was not optimised using CLASSY and could therefore lead to the biggest gain in inhibitor potency.A-BD CC was previously shown to display very high alpha-helicity as a homodimer, much more than BZLF1, and with a corresponding high T m value.We Finally, f1 and f4 saw Glu options offered in addition to the template Gln residue in order to potentially increase the negative charge of the acidic region and potentially increase electrostatic interaction in the instance of the acidic region being an imperfect coiled coil.This library design led to a 36,864-member library.

| Library screening
The peptide library was constructed and screened using a single-step selection on M9 agar plates under assay conditions.Colonies were then pooled and screened using liquid competition selection to reduce the peptide library size to a single winner peptide with the pool from different passages being sequenced, identifying a winning acidic peptide, Acidic Winner (AcidicW) (Figure 3, Figure 1C, Figure S2) over two passages of liquid competition.A-BD CC was observed during library screening but was outcompeted by AcidicW (Figure S3).The library selected Ile over Asn at the a3 position, removing the Asn-Asn interaction in the antagonist dimer (Figure 3B and Di).Interestingly the library exclusively selected Asp over Glu at g2, g3 and e4 positions.It also generally selected Val over other hydrophobic residues i.e. at the a2, d2 and d4 but not at the a4 position where Ala was retained.Finally, Glu was selected at the f1 position which increased the charge of the peptide and potentially the electrostatic repulsion of the acidic regions in the acidic peptide dimer.

| Biophysical analysis
Circular dichroism was first used to explore the secondary structure of AcidicW and its affinity for the BZLF1 bZIP target, compared to the A-BD CC peptide template (Figure 4, Table 1).The winner peptide, AcidicW homodimer was found to be considerably less helical than the A-BD CC template homodimer, displaying a significantly lower melting temperature (55 C compared to 64 C).The lower stability of AcidicW compared to A-BD CC may be in part due to the loss of the Asn-Asn interaction at the a3 position.The changes to hydrophobic residues at a and d positions may also impact homodimer stability due to changes in coiled-coil packing geometries.
A thermal denaturation CD experiment was carried out with both acidic peptides, A-BD CC and AcidicW, with BZLF1.Both the A-BD CC : BZLF1 and AcidicW:BZLF1 heterodimeric complexes were more stable than the BZLF1 homodimer, and were so highly stable that a full sigmoidal melt transition was not observed.The two heterodimers has visually very similar stabilities and the lack of a full sigmoidal melt transition prevented fitting to determine accurate T m values meaning it was not possible to decipher which heterodimeric complex was more stable.In order to decipher which complex was more stable, the very high stability of the heterodimers meant that it was necessary to (Figure 4D).This experiment suggested that the A-BD CC has a greater affinity for BZLF1 than AcidicW, despite the AcidicW homodimer destabilisation, although both heterodimers were highly stable.We can rationalise this difference in potency by carrying out further analysis of the AcidicW and A-BD CC sequences.Using the bZIP coiled-coil Interaction Prediction Algorithm (bCIPA) software, which can provide a quantitative measure of stability from primary peptide significance for the coiled-coil regions, we find that the BZLF1:A-BD CC is again predicted to have higher stability than the BZLF1:AcidicW interaction and also predicts a difference in helical propensity for the two acidic peptides in their interaction with BZLF1 of 1.36 vs. 1.30 for A-BD CC and AcidicW, respectively. 21,44,45This prediction is backed up by our circular dichroism experimental data where we can clearly see that the BZLF1:ABD CC dimer is more alpha-helical than the BZLF1:Acid-icW dimer (Figure 4A I and Figure 4B I, respectively).We, therefore, hypothesise that a major driver in the different affinities of the two acidic peptides for BZLF1 is driven by the decreased ability of Acid-icW to form a coiled-coil with BZLF1 compared to A-BD CC .At a residue level, this difference in helical propensity is exemplified by the fact that the AcidicW had replaced three Glutamic acidic residues of the A-BD CC sequence with Aspartic acid residues, which are known to have a lower alpha-helical propensity.

| Functional analysis
In order to probe the functional ability of the peptides to inhibit the binding of BZLF1 to TRE DNA, we performed a fluorescence polarisation competition assay.Here, binding of BZLF1 to a fluorescently labelled double-stranded TRE oligo was inhibited using an acidic peptide titration, which reduces the fluorescence polarisation signal and probes inhibitor potency.To develop the assay, we first showed that BZLF1 is bound to 6-FAM-labelled TRE DNA with a K D of 981 nM (Figure 5A).A fluorescence polarisation competition assay was then developed, with both peptides shown to lead to potent inhibition of BZLF1 binding to TRE DNA (Figure 5B, Table 1).A-BD CC was shown to inhibit with an IC 50 of 231 nM and AcidicW with an IC 50 of a higher potency, compared to AcidicW, for BZLF1.A-BD CC therefore has an increased ability to dimerise with BZLF1, compared to AcidicW and in turn block BZLF1 binding to TRE DNA sites.The difference in the potency of the two acidic peptides for BZLF1 is underscored by the thermal denaturation curves shown in Figure 4.
We hypothesise that no improvement in peptide inhibitor potency, and in fact a slight decrease in potency, is observed for Acid-icW over A-BD CC because both are potent inhibitors of BZLF1.This mer of the peptide antagonist with the target. 18,26,46This may suggest intracellular screening is better suited to optimising lower affinity peptide antagonists where a large improvement in inhibitor potency is possible and selection due to inhibitor potency is able to dominate in the screening process.

| CONCLUSION
TBS is a high throughput intracellular screening platform to derive functional transcription factor antagonists.Here we report the derivation of a potent BZLF1 TF inhibitor AcidicW using a TBS screening approach.The selected peptide, AcidicW, tightly binds its BZLF1 tar- In comparison to small molecules, peptides can struggle to enter cells, limiting their potential as therapeutics.A potential limitation of the TBS is that although it selects active compounds, peptides would still have to pass through the cell membrane.As has been used for other therapeutic peptides, we propose that TBS-derived peptides could be induced through conjugation to cell-penetrating peptide sequences such as Transactivator of Transcription (TAT) or Antennapedia peptides. 47,48Peptides could also be conjugated to nuclear localisation signals to target peptides to the cell nucleus.Some peptide antagonists, such as Omomyc, have been found to have an intrinsic cell-penetrating property without the need for conjugation, and this is believed to be due to its amphipathic helical basic region. 8nce in the BZLF1 homodimer, the C-terminal proximal region contributed more to dimer stability than the C-terminal distal region, future work could focus on truncating AcidicW to try and maintain as much potency as possible but facilitate a shorter peptide length. 34ture work could also focus on exploring the effects of inserting less than 15 TRE sites, and perhaps a single TRE site, in the mDHFR gene on inhibitor selection.
His-tagged purification constructs, as shown below, were cloned into a pET24a expression plasmid and transformed into BL21-Gold (DE3) T A B L E 1 Stability of the different protein mixtures measured by circular dichroism and the potency of acidic peptides using fluorescence polarisation competition assay.

2. 6 |
Circular dichroism (CD) studiesAll experiments were carried out using an Applied Photophysics Chirascan Circular Dichroism spectrophotometer with proteins in 12.5 mM sodium phosphate, 150 mM NaCl, 1 mM DTT and 0.25 mM EDTA pH 7.4 buffer.Samples were sealed with parafilm and heated to 65 C for 5 minutes in order to equilibrate the mixtures, and subsequently cooled and equilibrated at 20 C for ten minutes prior to reading.Before carrying out the thermal denaturation experiments, the spectra were measured for each sample between 200 nm and 280 nm at 20 C in a 1-mm cuvette.1-nm increments were used with 0.5 seconds/reading with three readings being taken and subsequently averaged.Thermal melts were recorded between 0 C to 90 C at 222 nm with heating carried out at a rate of 0.5 C/minute.After the thermal denaturation, protein samples were cooled back down to 20 C and spectra between 200 nm and 280 nm were re-recorded postdenaturation to confirm a reversible melting transition.For individual proteins, recordings were taken at a concentration of 40 μM, and the numerical average (mean) of these was calculated.1:1 mixtures were composed of 20 μM of each protein.All data was analysed in Graph-Pad Prism 9.The melting curves were fit using a two-state model as previously reported.41

Equation ( 2 4 buffer. 3 | 3 . 1 |
) is a one-site binding model where AF is the polarisation fraction, B is the maximum specific binding, [P] is the protein concentration in micromolar (μM), KD is the midpoint concentration where half the maximum signal is reached, Ns is the gradient of the non-specific binding and C is the background polarisation fraction.For the fluorescence polarisation competition assay, 10 nM annealed double-stranded 5'[6FAM]GCTTGATGACTCAGCCGGA was incubated with 400 nM BZLF1 and a serial dilution of the acidic peptide (A-BD CC or AcidicW) at room temperature for 30 minutes.The experiment was repeated three times, and the data were fitted in GraphPad Prism 9.0.All fluorescence polarisation experiments were carried out using 384 well-black opaque optiplate microplates with a total volume of 40 μl.For all experiments, the gain adjustment was set at 200 mP for 10 nM 5'[6FAM]GCTTGATGACTCAGCCGGA in 12.5 mM sodium phosphate, 150 mM NaCl, 1 mM DTT and 0.25 mM EDTA pH 7.RESULTS AND DISCUSSION Development of a Transcription Block Survival (TBS) platform to screen for a functional BZLF1 transcription factor inhibitor

F I G U R E 2
Development of the Transcription Block Survival (TBS) assay for BZLF1.A: Transcription of the essential gene, TRE-mDHFR with 15 inserted TRE sites, in step 1 imparts cell survival in the presence of TMP.Transcription is blocked through introduction of the transcription factor (BZLF1 bZIP) in step 2 and is subsequently rescued in step 3 through an inhibitor peptide (A-BD CC ) that can inhibit binding of the target bZIP TF to engineered TRE binding sites within the mDHFR gene.B: Colony counts for the development of the TBS assay to target BZLF1.thereforesought to develop a library that would maximise the binding of the acidic region to the BZLF1 basic region but also destabilise the acidic peptide homodimer.To this end, we introduced several residue options that would offer both stabilising and destabilising hydrophobic and electrostatic interactions.Residues at a and d positions of the basic region are generally not hydrophobic unlike those seen in the acidic A-BD CC region.Therefore, different hydrophobic residue options were included at five positions (a2, d2, a3, a4 and d4) to explore whether these would destabilise the A-BD CC homodimer, whilst also potentially inducing improved helix packing at the interaction interface with the BZLF1 basic region in the heterodimer.In the case of a3, a hydrophobic Ile residue was offered in addition to the F I G U R E 3 Figure showing the library design for this study for A: the homodimer and B: the heterodimer based on the A-BD CC template.Original A-BD CC template residues are shown within the black circles of the helical wheel, with additional offered residues at that position shown proximal to this template residue and the TBS selected residue is underlined in red.Helical wheel diagrams of the region covered by the library in this study for the i) heterodimers of C: A-BD CC and D: AcidicW with BZLF1, and ii) their corresponding homodimers.Figure made using the DrawCoil 1.0 program.Asn template residue to provide an option to disrupt the Asn-Asn interaction in the A-BD CC homodimer.With regard to electrostatics, five new options were offered at the e and g positions to both destabilise the A-BD CC homodimer through electrostatic repulsion whilst also improving electrostatic interactions in the heterodimer with BZLF1.This included Asp instead of Glu at the g2, g3 and e4 positions as the use of Asp in the acidic extension has not previously been explored.It was speculated that the introduction of Asp at the g2 and g3 positions may lead to a better interaction with the a2 and a3 Arg of the BZLF1 basic domain.It also includes the option of Ala at the e2 position over Glu to explore whether this might increase the electrostatic interaction with the Arg at the d1 position in the BZLF1 basic region.It also included offering a Lys option instead of Arg to allow alternative electrostatic interactions and/or packing at the g1 position.
repeat the melting experiment under denaturing conditions in presence of 2 M guanidinium chloride to observe a full unfolding profile for the A-BD CC :BZLF1 and the AcidicW:BZLF1 heterodimer with measured T m values of 72.8 C and 66.6 C, respectively, equivalent to a ΔT m of 3 C for the dimerisation of A-BD CC with BZLF1, compared to a ΔT m of 1 C for the dimerisation of AcidicW with BZLF1

F I G U R E 4
Circular dichroism studies of A: A-BD CC and B: AcidicW as homodimers and heterodimers with BZLF1 including (i) full spectra to determine structure and (ii) melts monitored at 222 nm to explore stabilities.C: Comparing the alpha-helicities of the heterodimers of A-BD CC and AcidicW with BZLF1 and D: a repeat of the melting curves of A-BD CC :BZLF1 and AcidicW:BZLF1 but this time in the presence of 2 M GdnCl to get full denaturation curves to facilitate comparison of their binding.similar high potency means both peptides likely lead to very potent inhibition intracellularly and are therefore difficult to distinguish in the screening process and instead other cellular factors may dominate in the intracellular selection process over inhibitor potency.In fact, TBS and other PCA-based approaches have generally previously been carried out to optimise the design of lower affinity peptides with melting temperatures lower than 70 C, often 50 C or lower for the heterodi- get with the AcidicW:BZLF1 heterodimer displaying a melting temperature in excess of 80 C and is more stable than the BZLF1 homodimer with a melting temperature of 76 C. AcidicW can also inhibit BZLF1 binding to TRE DNA with high potency (IC 50 of 612 nM).TBS library screening reduced the stability of the acidic peptide homodimer, with the melting temperature reducing from $64 C for A-BD CC to $55 C for AcidicW.A-BD CC and AcidicW were shown to be both very potent inhibitors of BZLF1 but this loss of stability for the AcidicW homodimer did not translate into a more potent antagonist.We hypothesise that the very high potencies of both A-BD CC and AcidicW are difficult to distinguish during the intracellular TBS screening assay and other cellular factors are able to instead dominate, helping to inform the choice of inhibitor peptides for this type of screening in the future.