Triple‐Helix‐Stabilizing Effects in Collagen Model Peptides Containing PPII‐Helix‐Preorganized Diproline Modules

Abstract Collagen model peptides (CMPs) serve as tools for understanding stability and function of the collagen triple helix and have a potential for biomedical applications. In the past, interstrand cross‐linking or conformational preconditioning of proline units through stereoelectronic effects have been utilized in the design of stabilized CMPs. To further study the effects determining collagen triple helix stability we investigated a series of CMPs containing synthetic diproline‐mimicking modules (ProMs), which were preorganized in a PPII‐helix‐type conformation by a functionalizable intrastrand C2 bridge. Results of CD‐based denaturation studies were correlated with calculated (DFT) conformational preferences of the ProM units, revealing that the relative helix stability is mainly governed by an interplay of main‐chain preorganization, ring‐flip preference, adaptability, and steric effects. Triple helix integrity was proven by crystal structure analysis and binding to HSP47.


Synthesis of Boc-H2-ProM2-OtBu
Boc-ProM2-OtBu (100 mg, 255 µmol) was dissolved in 4.0 ml methanol and 6.0 mg palladium on carbon (10 wt% loading) were added. Hydrogen was passed into the vigorously stirred solution for 15 min. After filtering off the catalyst (washing with MeOH), the solvent was removed in the rotary evaporator and the raw product was purified by column chromatography on silica gel (cHex/EtOAc = 3:2 to 1:1), which produced a colorless oil that solidified after prolonged storage.
The phases were separated, the aqueous phase extracted with CH2Cl2 (3 x 15 ml) and the combined organic phases dried over Na2SO4. After column chromatographic purification on silica gel (cHex/EtOAc = 2:1 to 1:1) the desired product was obtained in diastereomeric pure form as colorless oil.
Methanesulfonic acid amide (30 mg, 315 µmol, 1.03 eq) and Boc-ProM1-OtBu (120 mg, 306 µmol, 1.00 eq) were added to the yellow two-phase mixture at 0 °C. Within 3 h the reaction mixture was warmed to RT and stirred for 24 h at RT, whereupon 15 ml of an aqueous Na2SO3 solution (10 wt%) were added. Extraction was performed with EtOAc (3 x 30 ml). The combined organic phases were washed with sat. NaCl solution (10 ml), dried over Na2SO4 and the solvent was removed. The raw product was purified by column chromatography on ultrapure SiO2 (EtOAc/EtOH = 40:1), which after drying in vacuum led to the α-Diol in the form of a colorless solid (90 mg, 211 µmol, 69%) (analytical data: see below).

Synthesis of α-und β-Boc-(HO)2-ProM1-OtBu via Ru(VIII)-catalysis
Under argon, sodium periodate (240 mg, 1.12 mmol, 1.50 eq) and cerium(III)-chloride heptahydrate (45 mg, 121 µmol, 0.15 eq) were suspended in 0.75 ml water and heated at 60 °C for 1 min. At 0 °C 3.0 ml ethyl acetate and 4.5 ml acetonitrile were added to the light yellow suspension. A solution of ruthenium(III)-chloride trihydrate (9.8 mg, 37 µmol, 0.05 eq) in 0.75 ml water and a solution of Boc-ProM1-OtBu (300 mg, 764 µmol, 1.00 eq) in 1.5 ml ethyl acetate were then directly added. After vigorous stirring for 10 min at 0 °C the reaction was quenched by adding 9.0 ml sat. Na2SO3 solution. The reaction mixture was diluted with 20 ml water and extracted with ethyl acetate (3 x 50 ml). The combined organic phases were washed with 20 ml sat. NaCl solution, dried over Na2SO4 and the solvent was removed. Purification by column chromatography on ultrapure SiO2 resulted in pure α-diol as a colorless oil that slowly solidified to a colorless solid. In addition, a mixture of α-/β-diol was obtained (approx. 1:1 ratio on TLC, separation see below).

General procedure for the synthesis of Fmoc-ProM building blocks (Fmoc-procedure)
Under argon Boc-protected ProM-tert-butylester (100 to 500 µmol, 1.0 eq) was dissolved in 280 µl dry CH2Cl2 per 100 µmol starting material and the same volume of trifluoroacetic acid (280 µl per 100 µmol starting material, 36.0 eq) was added at 0 °C. The reaction mixture was stirred for 1 h at RT. After removing the volatile components in oil pump vacuum, the residue was dissolved in 0.6 ml sat.
NaHCO3 solution (per 100 µmol starting material) and 200 mg solid NaHCO3 was added to achieve pH = 8. A solution of Fmocchloride (1.5 eq) in THF (concentration: 50 mg/ml) was then added and the mixture was stirred overnight. The reaction mixture was diluted with approx. 10 ml water and the THF was quickly removed at the rotary evaporator. After washing the aqueous phase with Et2O (2 x 10 ml), it was saturated by adding Na2SO4 and adjusted to pH = 1-2 by adding drops of KHSO4 solution so that CH2Cl2 (4 x 10 ml) could be used for extraction. The combined organic phases were washed with sat. NaCl solution, dried over Na2SO4 and the solvent was removed, whereby the Fmoc-ProM building blocks were obtained as solids (Et2O addition promotes solidification).

Synthesis of Fmoc-ProM1-OH
The reaction was carried out with Boc-ProM1-OtBu (100 mg, 255 µmol, 1.0 eq) in accordance with the Fmoc procedure (see above), which resulted in a colorless foam after drying in vacuo.

Synthesis of Fmoc-epoxy-ProM1-OH
Under Argon Boc-ep-ProM1-OtBu (109 mg, 267 µmol, 1.0 eq) was dissolved in 1.3 ml dry CH2Cl2 and successively mixed with water (32 µl, 1.8 mmol, 6.7 eq), triisopropylsilane (32 µl, 156 µmol, 0.6 eq) and trifluoroacetic acid (1.24 ml, 16.1 mmol, 60 eq) at 0 °C. The reaction mixture was stirred at this temperature for 5 min and then for 1 h at 15 °C. After removal of the volatile components in vacuo, 500 mg NaHCO3 and 2.0 ml sat. NaHCO3 solution were added. At 18 °C, Fmoc chloride (100 mg, 387 µmol, 1.4 eq) in 3.0 ml THF was added to the reaction mixture, which was stirred for 2 h at the same temperature. It was diluted with 8.0 ml water and the THF was rapidly removed at the rotary evaporator. The aqueous solution was washed with Et2O (3 x 10 ml), saturated with Na2SO4 and treated with KHSO4 solution to adjust the pH to 1-2. Extraction was performed with CH2Cl2 (3 x 20 ml). Washing the combined organic phases with semi-sat. NaCl solution, drying over Na2SO4 and solvent removal resulted in a colorless foamy solid.
NaHCO3 solution. Solid NaHCO3 (100 mg) and a solution of Fmoc chloride (18 mg, 70 µmol, 1.5 eq) in 1.0 ml THF were added, followed by stirring for 2.5 h at RT. After removal of the THF at the rotary evaporator, the reaction mixture was diluted with 5 ml water, washed with Et2O (10 ml) and saturated with Na2SO4. The pH value was adjusted to 1 by adding drops of KHSO4 solution (10 wt%) so that CH2Cl2 (2 x 10 ml) could be used for extraction. The combined organic phases were washed with 10 ml sat. NaCl solution, dried over Na2SO4 and the solvent was removed resulting in a yellow oil.
This oil was dissolved together with para-toluenesulfonic acid (8 mg, 47 µmol, 1.0 eq) in 3.0 ml 2,2-dimethoxypropane /acetone/CH2Cl2 (1:1:1) and the mixture was stirred for 1 h at RT. After removal of the solvent at the rotary evaporator, the residue was dissolved in 6 ml sat. NaHCO3 solution and the above described extraction procedure was repeated. A light yellow solid was obtained.

Synthesis of Fmoc-(EtO)2-ProM1-OH
Boc-(EtO)2-ProM1-OtBu (128 mg, 265 µmol 1.0 eq) was used as a starting material for the general Fmoc-protocol while reducing the reaction time with the Fmoc-Cl solution to 2.5 h. As a result, a colorless solid was obtained after extraction and vacuum drying.

Solid Phase Peptide Syntheses
General conditions -All peptides were prepared by an automated peptide synthesizer (Multisyntech Syro I). Solid phase peptide syntheses were performed with NovaPEG low loaded Rink amide resin (Merck, loading: 0.18 -0.23 mmol/g) to avoid chain aggregation effects. [4] Swelling of resins and all standard coupling reactions were performed in DMF as solvent using equimolar amounts (8 -16 eq) of Fmoc-protected amino acids (Fmoc-Gly, Fmoc-Pro, Fmoc-Arg(Pbf), IRIS Biotech), diisopropylcarbodiimide (DIC) and oxyma following the Fmoc/tBu strategy. The Fmoc protecting group was cleaved using 30% v/v piperidine in DMF at the end of each coupling cycle. For sequences containing non-standard amino acids (Hyp or ProMs) a three-step procedure was carried out. Firstly, the C-terminal part of the sequence including Pro, Arg or Gly was synthesized automatically. Secondly, the Fmoc-protected ProMs or Fmoc-Hyp were coupled manually to the free N-termini of the previously installed peptide chains (vide infra) and thirdly, the remaining N-terminal part of the sequence was produced by automated synthesis again.
Cleavage from Resin and Purification -Generally, the N-termini of peptides were either acetylated or capped with Biotin-Ebes (vide infra) before the peptides were cleaved from the resin by shaking with conc. trifluoroacetic acid/triisopropylsilane/water Protocol for manual coupling -The Fmoc-protected ProM (2 eq) and HATU (2 eq) were dissolved in DMF/CH2Cl2 (9:1). After 5 min incubation at RT the active ester solution was pipetted on the pre-swollen resin and the coupling reaction was initiated by addition of DIPEA (4 eq). The resin was shaken at RT for 2 h, washed (DMF, CH2Cl2, MeOH, then Et2O) and dried in vacuo.
N-terminal Acetylation -To cap the N-terminus, an excess of acetic anhydride (20 eq) and DIPEA (20 eq) in CH2Cl2 was added to the resin, which was shaken 30 min at RT. After removing the solution the resin was washed and dried.
N-terminal capping with Biotin-Ebes -A solution of Fmoc-Ebes (2 eq) and Oxyma (2 eq) in DMF was pipetted to the resin and the reaction was started by addition of DIC (2 eq). After shaking at RT overnight the supernatant solution was removed, the resin was washed and dried. In the following, the Fmoc group was removed (30% v/v piperidine in DMF) and the free N-terminus capped with (+)-biotin (3 eq) under standard coupling conditions.

CD spectra and Thermal Denaturation Experiments
Experimental setup -Vacuum-dried peptides were dissolved in phosphate-buffered saline (10 mM Na2HPO4, 10 mM NaH2PO4, 150 mM NaCl, pH = 7.0) and the solutions (60 µM peptide) were incubated at 4 °C for at least 48 h for trimerization. CD measurements were performed on a Jasco J-715 spectropolarimeter using a quartz cuvette (Hellma Analytics, 1 mm path length) and spectra were recorded at 4 °C (1 nm band with, scan from 260 to 200 nm) indicating a maximum at 225 nm, which is characteristic of triple helix formation. [5] To study thermal transitions the ellipticity at 225 nm was monitored while increasing the temperature from 4 to 70 °C at a rather slow heating rate of 12 °C h -1 which was controlled electronically (Jasco PTC-348 WI Peltier element). According to the literature this heating rate still corresponds to non-equilibrium conditions, however, the generated data can be used for internal comparison of the peptides. [6] Data interpretation -For Tm-determination a two-state-model was employed in which three peptide chains combine to a triple helix assuming a previously reported mathematical model (vide infra). [7][8] Constants were set (c = 0.00006, R = 8.31) and auxiliary variables (P, U, V, K) were used to describe the temperature dependence of the folded fraction F. Thermal transition curves were fitted by optimizing the parameters θmin, θmax, ΔH, and Tm using the Generalized Reduced Gradient algorithm implemented in the solver tool of MS Excel. [9] . An error of ±1.0 °C was estimated for the resulting Tm values by repeating denaturation experiments.

Computational Methods and Primary Results
Conformational Search and Preoptimization -To generate collagen-relevant geometries a two-step procedure was carried out using Schrödinger Suite 2017-1. Firstly, the mixed torsional/low-mode sampling method in MacroModel was employed to find lowestenergy conformations within a 30 kJmol -1 window of OPLS_2005 force field energy. [10] The conformational search was started with N-acetylated methyl esters (Ac-[dipeptide]-OMe) to simulate a peptidic environment and to avoid intramolecular H bonds, respectively. For each investigated diproline derivative 10-100 structures were obtained, of which a set of up to four collagen-relevant conformers was selected (all-trans amides, ψYyy close to 180°, endo/endo, endo/exo, exo/endo or exo/exo puckered diproline rings). Secondly, the geometry of each conformer from this set was optimized in vacuum at the level of density functional theory (DFT) using B3LYP-D3/6-311++G** containing a dispersion-corrected functional, which has been recommended for proline-derived systems. [11] [12] To avoid convergence problems and to ensure time-efficient calculations the GVB-DIIS algorithm, an SCF level shift of 0.5 and the fine DFT grid density were employed in Schrödinger's Jaguar module. [13] These parameter adjustments have been recommended by the software manufacturer. Frequency calculations were carried out on all optimized geometries resulting in no imaginary frequencies and thus in the presence of true minima on the potential energy surface. Geometry Optimization -For the correct determination of relative energies further calculations in aqueous solution were performed for each conformer using the Gaussian 16 software package. [14] The Schrödinger geometries served as input and were refined at the B3LYP-D3/6-31G* level of theory and frequency calculations again revealed no imaginary frequencies. [12] Solvent effects were taken into account by employing the polarized continuum model (PCM) and the basis set was reduced to limit the computational cost of all the DFT-calculations. Finally, the main chain torsional angles of the lowest-energy conformers and the relative energies for each set of conformers for all diproline derivatives were evaluated (see Table S3 and section Optimized Geometries and Cartesian Coordinates). Method Validation -The procedure was performed on small test molecules (Ac-Pro-OMe, Ac-Hyp-OMe, see Tables S1-2). As a result, the values for main chain torsional angles as key geometric properties and the endo-/exo-energy difference for the ring flip could be reproduced. [a] DFT-calculations (B3LYP/6-31G*, in vacuo) from reference [15]; [b] Crystal structure parameters from reference [16]. Determination of the ring flip preferences in Yyy-position -For ProM1-derivatives, the DFT energy of the exo/endo-conformer was substracted from the exo/exo-conformer; for ProM2-derivatives and diproline derivatives (ProPro, ProHyp, HypPro, HypHyp) the DFT energy difference of the endo/exo-minus the endo/endo-conformer was formed. The results are presented in Figure 3. For Ac-ProM2-OMe no energy difference could be determined because no endo ring flip at Yyy could be found in the initial conformational search.

Ranking Procedure
Theoretical results (DFT-optimized geometries, alignments) of the Ac-[dipeptide]-OMe systems were assessed with respect to the ability of adopting a collagen-like conformation. The results were compared in four categories (main chain, ring flip, adaptability and sterics) and ranked on the basis of significant differences. "+" means above average, "o" average, "-" below average in terms of triple helix suitability.
(3) Adaptability: Proposed correction term to account for the ability to undergo small changes in torsional angles without developing significant ring strain. Thus, tricyclic ProMs were considered as "o" and unbridged dipeptides as "+".
The results from the ranking categories were summed up and presented in Table 1.

X-ray Crystal Structure Parameters of ProM-derivatives
Crystallization -X-ray measurements of the dipeptides were performed with a Bruker D8 Venture diffractometer. Single crystals were obtained by vapor diffusion of n-heptane into an EtOAc-solution of the sample. Figure S12. Boc-H2-ProM1-OtBu (left) and Boc-H2-ProM2-OtBu (right) in the crystalline state.

X-ray Crystal Structure Parameters of ProM2-CMP and Structural Alignment
Crystallization of ProM2-CMP and X-ray structure elucidation -Aqueous solutions of lyophilized peptide (250 µM with respect to trimer) were mixed with phosphate buffer (1.1 M (NH4)2HPO4, 0.1 M Tris, pH = 7.5) in a 1:1 (v/v) ratio. Peptide crystals were grown at RT by vapor diffusion and picked for crystal structural analysis.

HSP47 binding assay and KD-determination
Expression of HSP47 -HSP47 was essential expressed and purified as before, with some smaller adjustments (Oecal et al. JBC 2016). A canine-derived synthetic HSP47 gene coding for residues 36LSP…RDEL418 was was transformed in E. coli BL21 (DE3), cells grown to OD600 of 0.8 and protein expression induced with 0.5 mM isopropyl-ß-d-thio-galactoside (IPTG). After 3 h incubation at 37°C / 180 rpm cells were harvested and pellets stored at -20 °C. Cells were suspended in lysis buffer (TBS [50mM Tris•HCl, 150mM NaCl, pH 7.5] + 0.1 mM PMSF, 100µg/ml DNAse I, 0.5 mM DTT) and sonicated on ice. The cleared lysate was purified via Ni-NTA Superflow (Qiagen) resin with several wash steps (TBS, TBS containing 0.5M NaCl (total), TBS + 10 and 50mM Imidazol). After elution with 250mM Imidazol in TBS, the aliquots were directly adjusted to 2 mM DTT (final). To precipitate contaminants ammonium sulphate was added to a final concentration of 1.5 M and the solution cleared by centrifugation. The supernatant was concentrated and subjected to size exclusion chromatography (SEC) using a Superdex 200 increase column (GE Healthcare) in TBS + 2mM DTT.
Fractions containing the protein of interest were pooled and the solution stored at -80 °C after flash freezing in liquid N2.
Biolayer Interferometry -HSP47 binding studies were performed on a BLItz system (forte BIO) using high precision streptavidin SA biosensors (forte BIO). After 10 min re-hydration the biosensors were loaded with solutions of biotinylated peptide (10 µM with respect to trimer) for 60 s. For kinetic measurements, the biosensors were treated with fresh solutions of HSP47 (90 nM, 180 nM, 360 nM, 470 nM, 710 nM and 1.4 µM) for 60 s. Dissociation was carried out in PBS buffer (pH = 7.5, t = 120 s) and regeneration of biosensors in McIlvain buffer (pH = 6, t = 40 s). Unspecific binding and degradation effects of HSP47 could be ruled out by repeating the assay for 1.4 µM HSP47 before and after each dilution series. The spectral response was detected (in nm) and fitted for association and dissociation using a 1:1 Langmuir binding model in Origin 2018pro. [19] Data interpretation Figure [19] For clarity only every 10 th datapoint is shown; solid lines indicate fitting curves.