2-Cyanopyrimidine-Containing Molecules for N-Terminal Selective Cyclization of Phage-Displayed Peptides

Abstract

Current methods for the macrocyclization of phage-displayed peptides often rely on small molecule linkers that nonspecifically react with targeted amino acid residues. To expand tool kits for more regioselective macrocyclization of phage-displayed peptides, this study explores the unique condensation reaction between an N-terminal cysteine and nitrile along with the reactivity of an internal cysteine. Five 2-cyanopyrimidine derivatives were synthesized for this purpose and evaluated for their selective macrocyclization of a protein-fused model peptide. Among these, two novel linkers, 2-chloro-N-(2-cyanopyrimidin-5-yl)acetamide (pCAmCP) and 2-chloro-N-(2-cyanopyrimidin-4-yl)acetamide (mCAmCP), emerged as efficient molecules and were demonstrated to macrocyclize phage-displayed peptide libraries flanked by an N-terminal and an internal cysteine. Using these linkers to generate macrocyclic peptide libraries displayed on phages, peptide ligands for the ZNRF3 extracellular domain were successfully identified. One of the identified peptides, Z27S1, exhibited potent binding to ZNRF3 with a K_D value of 360 nM. Notably, the selection results revealed distinct peptide enrichment patterns depending on whether mCAmCP or pCAmCP was used, underscoring the significant impact of linker choice on macrocyclic peptide identification. Overall, this study validates the development of two novel regioselective, small molecule linkers for phage display of macrocyclic peptides and highlights the benefits of employing multiple linkers during phage selections.

Figure 1

Overview of approaches for phage display of macrocyclic peptides. (A) Initially developed nonselective linkers reacted with two cysteines via activated bis-electrophiles. Two subsequent N-terminal serine and cysteine approaches have been developed that react via hydroxylamines and nitriles to form stable iminoxy and thiazolidine adducts, respectively. (B) N-terminal cysteine selective bis-electrophiles are colored according to reactivity with N-terminal cysteine (blue) or internal cysteine (red). Currently available N-terminal cysteine linkers are limited by either low hydrophilicity or highly flexible scaffolds. The novel linkers reported in this paper give additional diversity of highly constrained scaffolds with good hydrophilicity and regioselectivity.

Figure 2

Identification of modified 2-cyanopyrimidines for the cyclization of an N-terminal cysteine-containing peptide fused to the N-terminus of sfGFP. (A) Five different potential N-terminal selective bis-electrophiles were screened for reactivity with the CA₅C-sfGFP protein and analyzed via LC-ESI-MS after reacting for 3 h at RT. Percent yields were calculated by the ratio between desired product and remaining unreacted protein. N.r. = no observed reaction; TCEP = the only observed adducts included a TCEP modification of acrylamide. (B) Two possible reaction paths were observed for the linkers dependent on the secondary electrophile being an acrylamide or 2-chloroacetamide moiety. Linkers containing acrylamides reacted with TCEP via a phospha-Michael addition, while the 2-chloroacetamide compounds gave the desired cyclic product. (C) CAmCP linkers were optimized for their reaction with CA₅C-sfGFP. The percent cyclized under each condition was calculated by the ratio of the peaks in the deconvoluted mass spectra (right). Masses corresponding to the unreacted product (28,255 Da) and cyclized product (28,399 Da) are labeled.

Figure 3

Kinetic characterization of the CAmCP linkers reacting with a model peptide. (A) Peptide CGK-5 was reacted with the CAmCP linkers to afford the cyclized product CGK-5cyc. Representative HPLC traces and integrated peaks for the kinetic experiments with pCAmCP (B) and mCAmCP (C) are shown. (D) For studying the chloroacetamide addition, CGK-5Abu was reacted with the CAmCP linkers. Representative HPLC traces and integrated peaks are shown for the reactions with pCAmCP (E) and mCAmCP (F). Each data point is shown as the mean ± s.d. of at least three independent experiments. Half-lives are reported as the mean ± 95% CI of the fitted curves.

Figure 4

Validation of linkers pCAmCP and mCAmCP for phage display. (A) No observed phage toxicity occurred for either linker when reacted with up to 200 μM of compound and at varying reaction temperatures. Pfu/mL = total infectious phage units per mL quantified via colony-forming unit assays. All assays are reported as mean ± s.d. for n = 3 replicates. (B) Biotin pulldown assays indicated high reactivity of the CAmCP linkers with phages that contain N-terminal cysteines. Phages were quantified for their ability to be captured on streptavidin beads using Biotin-CBT to modify free N-terminal cysteines. After reaction with CAmCP linkers, a CA₅C-modified phage library had no significant differences compared to an unmodified AAKAA library. % Captured calculated as (phages input – phages in supernatant)/phages input. All experiments are reported as mean ± s.d. for n = 3 replicates.

Figure 5

Selection of macrocyclic peptides for binding to ZNRF3. (A) Phages were selected to bind to the ZNRF3 extracellular domain through four rounds of selection with alternating streptavidin and NHS beads to capture the protein target. (B) Illumina next-generation sequencing demonstrated strong enrichment of unique sequences using both pCAmCP (blue) and mCAmCP (red). Comparison of round 4 sequences for each (Venn diagram) demonstrated that the peptide sequences were dependent upon the specific linker for enrichment, with only 6 overlap sequences being found in each selection.

Figure 6

Characterization of peptides for binding to the ZNRF3 extracellular domain. (A) Top peptides from the pCAmCP selection were investigated for their binding to ZNRF3 through biolayer interferometry studies. K_D values are reported as the mean ± s.d. of three independent experiments (n = 3). (B) Representative traces for the biolayer interferometry experiments summarized in panel (A). Raw traces (multicolored by concentration) were fitted (black lines) to a 1:1 protein/ligand binding model to calculate kinetic parameters.