Purification of human erythropoietin by affinity chromatography using cyclic peptide ligands.

Prior work described the identification and characterization of erythropoietin-binding cyclic peptides SLFFLH, VVFFVH, FSLLHH and FSLLSH (all of the form cyclo[(Nα-Ac)Dap(A)-X1-X6-AE], wherein X1-X6 is the listed sequences). In this work, the peptide ligands were synthesized on Toyopearl chromatographic resins and utilized for purifying recombinant human erythropoietin (rHuEPO) from complex sources. Elution buffer pH and composition were optimized to maximize the recovery of standard rHuEPO from the peptide resins. The peptide-based adsorbents were employed for separating rHuEPO from a mixture of albumin, myoglobin, and IgG to examine their selectivity. When using FSLLHH, the inclusion of low amounts of surfactants in the wash and elution buffers facilitated the recovery of rHuEPO with high yield and purity. Specifically, FSLLSH and VVFFVH afforded the most efficient separation of rHuEPO, with yield and purity of 85% and 95-97%, respectively. The affinity resins were also utilized to purify rHuEPO from spiked CHO cell culture fluid. In particular, FSLLSH provided the most successful separation from CHO, with yield and purity above 90%, and 1.0 log10 reduction of host cell proteins. The influence of conductivity and pH in the CHO-rHuEPO load was investigated. Finally, FSLLSH-based resins were used to purify rHuEPO spiked into a Pichia pastoris cell culture fluid, resulting in product yield and purity of 96% and 84%, respectively, and 1.3 log10 reduction of host DNA. These results compare well with values obtained using wheat germ agglutinin agarose and clearly indicate the potential of the cyclic peptide resins as a viable tool for rHuEPO purification.

Design, selection, and development of cyclic peptide ligands for human erythropoietin.

This work presents the selection and characterization of erythropoietin (EPO)-binding cyclic peptide ligands. The sequences were selected by screening a focused library of cyclic depsipeptides cyclo[(Nα-Ac)Dap(A)-X1-X6-AE], whose structure and amino acid compositions were tailored to mimic the EPO receptor. The sequences identified through library screening were synthesized on chromatographic resin and characterized via binding-and-elution studies against EPO to select a pool of candidate ligands. Sequences with higher hydrophobicity consistently showed stronger binding to EPO, with the exception of FSLLSH, which was noted for its lower hydrophobicity and high EPO binding. Mutagenesis studies performed on FSLLSH with natural and non-natural amino acid substitutions led to the identification of critical EPO-binding determinants, and the discovery of new peptide ligands. In particular, histidine-scanning mutagenesis performed on three lead sequences yielded the discovery of variants whose EPO-binding is more pH-sensitive, which facilitates EPO recovery. Selected ligands were studied to correlate the elution yield to the salinity of the binding buffer and the elution pH. Elution yields were consistently higher when EPO binding was performed at low ionic strength. The crystal structures of lead cyclic peptides were docked in silico against EPO to estimate the binding affinity in solution. Isotherm adsorption studies performed on FSLLSH indicated that the cyclic version of the ligand (KD=0.46µM) has a higher affinity for EPO than its corresponding linear variant (KD=1.44µM). Collectively, these studies set the stage for use of the cyclic peptide ligands as EPO purification and detection tools.

Design of protease-resistant peptide ligands for the purification of antibodies from human plasma.

A strategy is presented for developing variants of peptide ligands with enhanced biochemical stability for the purification of antibodies from animal sera. Antibody-binding sequences HWRGWV, HYFKFD, and HFRRHL, previously discovered by our group, were modified with non-natural amino acids to gain resistance to proteolysis, while maintaining target affinity and selectivity. As trypsin and α-chymotrypsin were chosen as models of natural proteolytic enzymes, the basic (arginine and lysine) and aromatic (tryptophan, phenylalanine, and tyrosine) amino acids were replaced with non-natural analogs. Using the docking software HADDOCK, a virtual library of peptide variants was designed and screened in-silico against the known HWRGWV binding site on the pFc fragment of IgG. A pool of selected sequences with the highest predicted free energy of binding was synthesized on chromatographic resin, and the resulting adsorbents were tested for IgG binding and resistance to proteases. The ligand variants exhibited binding capacities and specificities comparable to the original sequences, yet with much higher proteolytic resistances. The sequences HWMetCitGWMetV and HFMetCitCitHL was used for purifying polyclonal IgG from IgG-rich fractions of human plasma, with yields and purity above 90%. Notably, due to electrical neutrality, the variant showed higher selectivity than the original sequence. Binding isotherms were also constructed, which confirmed the docking predictions. This method represents a general strategy for enhancing the biochemical stability as well as the affinity and selectivity of natural or synthetic peptide ligands for bioseparations.