Facile saccharide-free mimetics that recapitulate key features of glycosaminoglycan sulfation patterns† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c8sc02303d

We report a new class of saccharide-free glycosaminoglycan (GAG) mimetics where polyproline imparts facilely-made sulfation patterns with GAG-like structure, function and tunability.

General procedure for the synthesis of PGM precursors.
All PGM precursors were prepared via standard Boc chemistry in the solution phase as reported previously. 1  Purification of crude compounds by reverse phase HPLC afforded desired final PGM precursors as a white solid.

Circular dichroism (CD) analysis.
CD spectra were obtained using a Jasco-815 CD spectrometer equipped with a Peltier [θ] = θ/(10·N·c·l) θ represents the ellipticity in millidegrees, N the number of amino acid residues, c the molar concentration in mol·L -1 , and l the cell path length in cm ( Figure 2c; Figure S21, S31).

MD simulations.
Modified proline residues were constructed using PyMol. 2 Subsequent two-step quantum mechanical minimization at the theory levels PM6/aug-cc-pVDZ and wB97XD/aug-cc-pVDZ using Gaussian 09 Rev. C.01 3 yielded starting geometries. Subsequently, partial charges were derived using the AM1-BCC methodology 4 as implemented in Amber14 5 tleap. Charges were derived with an acetyl-cap on the N-terminal side and a N-methyl cap on the C-terminal side respectively. Subsequent removal of these caps yielded the final residue topology. Bonded terms were assigned according to the Amber force field ff14SB. 6 ff14SB parameters were chosen over GAFF 7 in order to seamlessly integrate the new residue with arbitrary default amino acids. Ligands were then constructed using the sequence command in tleap. In the case of the PGMs, the N-terminus consists of an acetyl-capped glycine residue followed by each of the four sequences of proline and modified proline residues and are capped with a methoxy-group on the C-terminus.
Parameters for chondroitin sulfate-E (CS-E) were derived following the same protocol.
PGMs and CS-E were solvated as an extended conformation in cubic boxes with a wall separation of 0.6 nm using TIP3P water 8 and charges were neutralized using sodium counter ions parameterized according to Joung and Cheatham 9 . Electrostatics were described using the particle mesh Ewald procedure 10 with a long-range cutoff of 0.8 nm coinciding with truncation for Lennard-Jones interactions. The systems were optimized for 2000 steps of steepest descent followed by 2000 steps of conjugate gradient minimization. Subsequently, the systems were equilibrated for 1 ns to constant density at 1 bar using a Berendsen barostat 11 . Langevin thermostatization 12 was used to achieve a constant temperature of 300 K. Bonds to hydrogen were constrained to allow a simulation time step of 2 fs. All calculations were performed using the GPU implementation of pmemd. 13 (Table S7).
In order to visualize the dynamics of the different PGMs and gauge their relative conformational flexibility, we aligned the ligands on their respective backbones. In the case of PGMs, this is the polyproline backbone. In the case of CS-E, the alignment was performed on the 6 heavy atoms within the pyranose substructure of the individual sugars. Subsequent to alignment, frames were extracted at 10 ns intervals and superimposed to visualize the conformational space explored by each PGM or CS-E.
Additionally we calculated the distances of the center of mass of the charged sulfate groups with those of neighboring residues. The distance plots include measurements between sulfates on selected residues and the first sulfate-bearing proline residue.

Unbiased docking simulations.
The structure of human P-selectin was extracted from the PDB structure 1G1S. 16 In the absence of experimental structural information on either CS-E or PGM complexes with P-selectin, we conducted an extensive series of molecular dynamics simulations inspired by an dynamic molecular docking (DMD) approach reported previously for protein-GAG systems. 17 Unlike the protein-GAG systems described in the reported DMD approach, Pselectin does not have an apparent binding pocket or groove and the selection of an entry direction for the DMD pulling process may be biased. To avoid such bias, we instead created 100 independent starting geometries by placing human P-selectin at the center of a 7x7x7 nm cubic box and randomly inserting the respective ligands using the 'insertmolecules' tool of the Gromacs suite 18
Fluorescence measurements and IC 50 were then obtained as per competitive ELISA. After 20 min, the mice were injected with 200 μl of the prepared tumor cell suspension.
All injections were performed by an experimenter blinded to the treatment assigned to each mouse. Whole body weight of mice was measured on day 0 (before injection), 1, 3,