A screen for peptide agonists of the G-CSF receptor
© Schneider et al; licensee BioMed Central Ltd. 2011
Received: 16 March 2011
Accepted: 15 June 2011
Published: 15 June 2011
Skip to main content
© Schneider et al; licensee BioMed Central Ltd. 2011
Received: 16 March 2011
Accepted: 15 June 2011
Published: 15 June 2011
Granulocyte-colony stimulating factor (G-CSF) is one of the most important pharmacologically used proteins. Potential uses beyond the stimulation of neutrophilic granulocytes are the treatment of CNS disorders. Disadvantages of the G-CSF protein as a drug are its moderate plasma half-life time and considerable production costs. We therefore conducted a screen for peptide agonists derived from the sequence of human G-CSF.
Despite of the high sensitivity of our screening system we could not detect any positive hits in a single peptide approach. In a multiplex approach using a permutation of any combination of 10 different peptides we could also not detect a positive block.
We conclude that larger coherent parts of the protein or dimerising peptides may be needed to achieve activation of the receptor.
Granulocyte-colony stimulating factor (G-CSF) is one the most widely employed protein drugs. It is mostly used for counteracting neutropenia in patients receiving chemotherapy , but also for stem cell harvesting [2, 3], and as add-on to anti-infectious therapy. G-CSF is a glycoprotein that binds in 2:2 ligand: receptor stoichiometry to its cytokine like receptor (G-CSF-R) which recruits in turn Janus kinases (JAKs), a family of protein tyrosine kinases. These kinases phosphorylate the receptor and also themselves starting by that a multiple signalling cascade, which involves amongst others STAT1 and 3, PI3K/Akt and the Ras/Mek/Erk1/2 pathway [4–8].
Recently we and others have defined a novel spectrum of G-CSF activities in the central nervous system where it acts on neurons. For example, it was shown that G-CSF can reduce the infarct size in animal stroke models [9–11] and that it has significant beneficial effects on the motor performance as well as on the overall survival in a mouse model for Amyotrophic lateral sclerosis (ALS) . G-CSF is therefore clinically explored for several neurological diseases, such as stroke [13, 14].
Disadvantages of the G-CSF protein for a potential lifetime continuous therapy as in the case of chronic neurodegenerative conditions are its limited plasma half-life time (~4 h), the relatively high costs, and chronic effects on the hematopoietic system. Pegylated forms of G-CSF are available that have a much extended half-life, however, it is unclear at present if those modifications hinder passage of the blood-brain-barrier (BBB).
We therefore decided to conduct a screen searching for peptides derived from the human G-CSF peptide sequence with agonistic activity. Peptides would be considerably cheaper to produce, be potentially suitable for delivery methods other than subcutaneous injection, and might also show improved neuronal selectivity.
It was previously shown for a number of protein receptors that this is a feasible concept in principle. For example, peptides derived from the NCAM (neural cell adhesion molecule) sequence, a cell surface glycoprotein that belongs to the Ig superfamily, act agonistically on the fibroblast growth factor (FGF) receptor [15–17]. NCAM is involved in the formation of neuronal connections during development and modulates synaptic plasticity associated with regeneration and learning [18–20]. Several mimetic peptides have been derived from its protein structure, like C3 by combinatorial chemistry from the NCAM Ig1 module , the BCL motif from the second NCAM fibronectin type III module  and P2, a 12 amino acid sequence localized in the FG loop of the second Ig module of NCAM . All these peptides are potent mimetics of NCAM and therefore attractive compounds in the development of therapies in neurodegenerative disorders.
Moreover, agonistic peptides have been found for the EPO receptor (EPO-R), functionally very close to the G-CSF receptor (G-CSF-R), both derived from the EPO sequence  or fully novel [25, 26]. Indeed, one of the agonistic peptides has entered clinical development . This agonistic peptide for the EPO-R (EMP-1) was discovered by phage display and is a cyclic peptide of 20 residues, which has nothing in common with the original EPO protein structure. EMP-1 can compete with EPO at micromolar concentrations by binding as a dimer to the extracellular domain of the EPO-R and by activating the receptor by induced dimerization. This was also shown by the crystal structure of the complex of EMP-1 with recombinant soluble EPO-receptor. In two animal models for erythropoesis EMP-1 also exhibited activity [26, 28]. In further development steps EMP-1 has been chemically modified through covalent linkage of polyethylene glycol resulting in the EMP-1 related dipeptide Hematide™ . Hematide™ obtained an increased half life and has been shown to be safe in healthy volunteers by increasing haemoglobin levels for one month . A clinical trial phase III has just been completed in 2010.
In 1998 the nonpeptidyl small molecule SB 247464 was identified as a ligand of the murine G-CSF receptor in a high-throughput cell based screening approach by Tian and colleagues . Like G-CSF, SB 247464 induced tyrosine phosphorylation of multiple signaling proteins, and stimulated primary murine bone marrow cells by oligomerizing the receptor chains. Nonetheless the synthetic compound showed lower potency compared to G-CSF and bound exclusively to the murine but not the human G-CSF-R . Up to now no other agonistic small molecules or peptides for the human G-CSF-R have been identified or described.
Thus we conducted a sensitive screen for peptide agonists derived from the sequence of human G-CSF employing the murine myeloblastic cell line NFS-60. This suspension cell line expands rapidly in culture and is dependent on G-CSF of either murine or human origin, or alternatively on M-CSF (Macrophage colony-stimulating factor) or IL-3 (Interleukin-3) . Our assay read-out was viability of these cells in the absence of G-CSF.
For the design of the peptide library the protein sequence of human G-CSF [NCBI: NP_000750] was used. The first 29 amino acids of the G-CSF protein (the signal sequence) were not considered for the library. For the remaining 178 amino acids of the G-CSF sequence we designed an overlapping peptide library for G-CSF with the help of the PepScreen® Library Design Tool . We chose an overlapping 20 mer peptide library with a shift of 2 amino acids per peptide, resulting in a total of 80 peptides. The peptide library was ordered in crude quality with a minimum content of 30% intact peptide that was analyzed by MS and HPLC by the provider (Genscript USA Inc., Piscataway, NJ, USA). Because of the on average high hydrophobicity of the peptides they were dissolved in DMSO at a concentration of 1 mM and stored at -80°C.
The murine myeloblastic cell line NFS-60 expands rapidly in culture and is dependent on G-CSF of either murine or human origin, or alternatively on M-CSF (Macrophage colony-stimulating factor) or IL-3 (Interleukin-3) . The cell line was cultured as suspension culture in RPMI 1640 medium supplemented with 5.1 ml L-glutamine (200 mM), 10% fetal bovine serum, 1x P/S, 10 μM β-Mercaptoethanol and 2 ng/ml G-CSF at 37°C and with a CO2 fraction of 5%.
NFS-60 cells were washed twice with cell culture medium containing no G-CSF and were then plated at a total density of 1 × 104/well in 96-well plates. Subsequently human recombinant G-CSF (AX200, Dr. Reddy's Laboratories, Hyderabad, India; 300 μg/ml) was added to the cells at increasing concentrations, and cells were incubated for 48 h at 37°C. Viability of the cells was detected using the MultiTox-Fluor Multiplexicity Assay (Promega, Mannheim, Germany) as described by the provider. Fluorescence was measured using the plate reader FLUOstar (BMG Labtech, Offenburg, Germany) at 390 nm/520 nm.
96-well plates were prepared with 8 peptides with 8 replicates/plate, 2 negative controls (G-CSF buffer (10 mM Acetic Acid, 250 mM Sorbitol, 0.004% Tween 80)) at 8 replicates/plate, and two positive controls, one at the lower end of sensitivity, and one at a 100% viability level (each at 8 replicates). In order to avoid evaporation of the edging wells and to have the same conditions for all probes the samples with their eight replicates were pipetted in a diagonal pattern on a 96-well plate automatically by the MultiProbeII EX from Packard [Additional file 1].
NFS-60 cells were washed twice with cell culture medium containing no G-CSF and were then plated in a 96-well plate at a total cell number of 1 × 104/well. The peptide plate was then transferred by the Liquidator96 (Steinbrenner Laborsysteme GmbH) to this plate and incubated at 37°C for 48 hours. Final DMSO concentrations were 0.2% for the 1 μM screen, and 2% for the 10 μM screen. For detection of viability the MultiTox-Fluor Multiplexicity Assay from Promega was used as described by the provider. The fluorescence of the plate was measured in a fluorescence reader (FLUOstar) at 390 nm and 520 nm.
Up to 20 peptides per well were tested in a multiplex approach. The 80 peptides of the library were arranged in eight groups per ten peptides. Permutations of all combinations of two peptide groups were tested.
All results were evaluated in EXCEL. For analysis of the data a student's t-test against both negative controls was performed. A p-value <0.05 was considered significant.
As both screening approaches of single peptides with molarities of 1 μM and 10 μM were unsuccessful, we finally tried a multiplex approach, based on possibly necessary cooperative effects of different protein epitopes in order to activate the G-CSF receptor. As the cell line NFS-60 showed high sensitivity towards higher DMSO concentrations the maximum number of peptides to be tested per well were 20 at a total molarity of 10 μM (single concentration 0.5 μM). This resulted in a DMSO content of 2%.
In the present study we have utilized the G-CSF dependence of the cell line NFS-60 to screen for agonistic peptides derived from the G-CSF human sequence. Although the screening system was highly sensitive and would have allowed the detection of an agonistic peptide at least 3.4 × 10 7-fold less potent than the parent protein, we could not detect any positive hits. Also in a multiplex approach using a permutation of any combination of 10 different peptides we could not detect a positive block. A post-hoc power analysis of our screen revealed that it should have been possible to detect an increase in viability down to 10% with a power of 80%.
Reasons for this failure may either be technical problems, or principal in nature. We believe that the latter is true. The coverage of the G-CSF sequence was certainly dense with a frame shifted by 2 amino acids over the entire sequence. The purity of the synthesized peptides was reasonable with a minimum content of pure and intact peptide of 30%, and the peptides were all fully dissolved at the assayed concentrations.
It is possible that the length of the peptides was not sufficient. However, other agonistic peptides described for the EPO (Erythropoietin) receptor or NCAM (neural cell adhesion molecule) have 20 or less amino acids. Screening much longer peptides would also make less sense in view of their pharmaceutical advantage over the full-length protein.
An interesting point is the potential need for dimerising peptides. G-CSF itself dimerizes when binding to its receptor, and induces receptor dimerization [4–6]. Indeed, an engineered G-CSF dimer (F-627) appears more potent in vivo than the monomer form . It has also been speculated that the nonpeptidyl small molecule SB 247464 acts because of its twofold rotational symmetry as a dimerization inducer of the G-CSF receptor . It may therefore be advantageous in future approaches to utilize dimerizing peptides for such a screen.
Granulocyte-colony stimulating factor
Granulocyte-colony stimulating factor receptor
Amyotrophic lateral sclerosis
neural cell adhesion molecule
fibroblast growth factor, (EPO-R): Erythropoietin receptor
Macrophage colony-stimulating factor
We thank Frank Herzog, Uli Bolz, Simone Hoppe, and Gisela Eisenhardt for excellent technical help. This work was supported by a grant from the BioProfile für die Region Braunschweig, Göttingen, Hannover (Forschungsvorhaben 0315180C).
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.