Human Platelet-Derived Endothelial Cell Growth Factor
is Homologous to E.coli Thymidine Phosphorylase
G.J. Barton, C.P. Ponting and G. Spraggon
Laboratory of Molecular Biophysics, University of Oxford, South Parks Road, Oxford, UK, OX1 3QU.
C. Finnis and D. Sleep
Delta Biotechnology Ltd., Castle Court, 59, Castle Boulevard, Nottingham, UK, NG7 1FD.
Monomeric human platelet-derived endothelial cell growth factor (PD-ECGF) is a single-chain protein of relative molecular mass (kD which stimulates the growth and chemotaxis of endothelial cells in vitro and possesses angiogenic activity in vivo . As angiogenesis is central to the pathological conditions of tumour growth, rheumatoid arthritis, diabetic retinopathy, psoriasis and haemangiomas a detailed understanding of the molecular action of PD-ECGF would provide clues to therapeutic strategies for these disease states.
Here we report the striking similarity between the primary sequences of PD-ECGF and thymidine phosphorylase (TP) from E. coli . Human TP catalyses the reversible phosphorolysis of thymidine and other pyrimidine 2-deoxyribosides, with the exception of 4-amino substituted compounds, and has nucleoside deoxyribosyl transferase activity. TP is one of two pyrimidine phosphorylases in the base and nucleoside salvage pathway. Under near-physiological conditions TP is a homodimer with a molecular mass of 110kD in mammals  and 90kD in E. coli . Specific inhibitors of TP are considered as potential chemotherapeutics either to reduce clearance of thymidine and other deoxyuridine analogues presently in use as antineoplastic and antiviral agents or by interfering with the salvage process . The design of such inhibitors may be aided by the 2.8Å-resolution crystal structure of E. coli TP which shows TP to consist of a small helical domain and a larger domain both of which comprise two non-continuous segments of polypeptide (residues 1-65;163-193 and 80-154;197-440 respectively) . Thymidine and phosphate moieties appear to be bound in a cleft between these two domains.
Figure 1 (part1 part2) highlights the sequence similarity between the human PD-ECGF and E. coli TP sequences. The similarity extends over all but the N-terminal 32 and the C-terminal 4 residues of PD-ECGF. Human PD-ECGF apparently undergoes post-translational maturation, whereby 10 and 4 amino acids are removed from the amino and carboxyl termini respectively . Mature PD-ECGF has a 22 amino acid N-terminal extension with respect to the E. coli TP sequence. The sequences show 40%identity calculated over the 438 common amino acid positions (see legend to Figure 1) and are therefore likely to have diverged from a common genetic ancestor and share the same overall tertiary fold . Thus, the crystal structure of E. coli TP may be used as a scaffold on which to model the structure of human PD-ECGF. The amino acids of the putative phosphate binding sites of TP are conserved in PD-ECGF (indicated by `P' in Figure 1), as are the thymidine-binding residues (Arg, Ser and Lys), (indicated by `T' in Figure 1). The small helical domain of TP has 46%identity with PD-ECGF, whilst the larger domain has 37%identity. Further subdivision of these domains shows the segments containing binding residues for thymidine (163-193, helical domain) and phosphate (80-154, domain) to be the most highly conserved regions of the molecules (74%and 60%identity respectively). In contrast, the remainder of the small and large domains (1-65 and 197-440) show 32%and 30%identity respectively. The large domain may be divided between helices 14 and 15 into two putative folding units , this division coincides with an eight residue insertion in PD-ECGF relative to TP. The C-terminal unit (330-440) is the least conserved region of the proteins (21%identity), whilst the N-terminal unit (197-320) shows 38%identity with PD-ECGF. There are two deletions of one residue in addition to the eight-residue insertion in the PD-ECGF sequence. These changes lie on the surface of the TP structure and are distant from the active site cleft. The deletion of Tyr at position 267 is spatially adjacent to the eight-residue insertion (position 329 of TP) in the structure whilst the deletion of Leu is in a region proposed as a hinge-point between the two domains . The TP dimer interface is centred on three hydrophobic residues in helix H3 (Met, Phe and Phe). Of these, only the first (Met) is identical in PD-ECGF whilst the second and third are replaced by Arg and Leu respectively.
Human TP and PD-ECGF appear to have a common tissue distribution (platelets, placenta and some tumours) , to share a cytosolic cellular location  and to have similar monomeric molecular weights . Given the strong overall sequence similarity between human PD-ECGF and E. coli TP especially around the active site, it is tempting to suggest that PD-ECGF is human TP. The principal objections to this hypothesis are that PD-ECGF has an additional 22 residues at the N-terminus relative to E. coli TP, and that there is only limited residue conservation in the core of the dimer interface. These observations may explain why active PD-ECGF is apparently found in both monomeric and dimeric forms . We currently cannot rule out the possibility that the differences in length, and in the dimer interface are simply due to inter species variation rather than functional differences.
The expression and purification of recombinant PD-ECGF from the yeast Saccharomyces cerevisiae has enabled us to establish that recombinant PD-ECGF is mitogenic toward endothelial cells . Preliminary results also suggest that recombinant PD-ECGF possesses TP activity and that TP purified from E. coli is mitogenic toward endothelial cells (unpublished observations). Moreover, the recently reported covalent modification of PD-ECGF by nucleotides in vitro and in vivo, , probably occurs as a result of the nucleoside and phosphate binding properties of PD-ECGF.
We conclude that PD-ECGF is a human TP-homologue, the mitogenic and angiogenic activities of which are a result of its TP activity. The endothelial cell specificity of PD-ECGF  might be achieved via a specific endothelial cell receptor, however, it seems more likely that endothelial cells respond specifically to a modulation in intracellular DNA precursor pools brought about by the TP activity of PD-ECGF.
Sequence alignment of human platelet-derived endothelial cell growth factor (PD-ECGF) and E. coli thymidine phosphorylase (TP). Sequence similarity was detected by scanning the NBRF-PIR database (V.28) with the PD-ECGF sequence using the Smith and Waterman  local similarity algorithm with Dayhoff's MDM78 matrix and a length-dependent penalty of 8. The alignment shown was generated with the AMPS package  and shows 40%identity over 438 aligned positions. The similarity score is 30 standard deviation units from the mean of scores for randomised sequences of the same length and composition as PD-ECGF and TP. Aligned residues identical in PD-ECGF and TP are shown in bold in the PD-ECGF sequence and not repeated in the TP sequence. Conserved: highlights residue pairs that have positive substitution values in Dayhoff's MDM78 matrix. SS: open boxes show the secondary structure of TP ; H1-H17 are helices; A1-A6 and B1-B4 are -strands in sheets A and B respectively. Notes: Residues involved in the putative phosphate binding site are identified by a `P' character; similarly, residues of the thymidine binding site are shown by a `T'. Regions which are specifically involved in the dimer interface (H1, H3 and H8-H9 loop) are labelled with `D'. Putative hinge points at 66-79, 155-162 and 193-196 are shown by `H' characters.