STATEMENT REGARDING FEDERALLY-SPONSORED R&D Not applicable.

REFERENCE TO MICROFICHE APPENDIX Not applicable.

FIELD OF THE INVENTION This invention relates to compositions of soluble recombinant endostatin and methods of producing soluble recombinant endostatin from a prokaryotic host cell.

BACKGROUND OF THE INVENTION Endostatin is a 183 amino acid C-terminal fragment of collagen XVIII with a molecular mass of approximately 20 kDa. Endostatin has been reported to have anti-angiogenic activity sufficient to induce the regression of a variety of <BR> <BR> <BR> implanted, primary tumors in mice (O'Reilly et a/., 1997; WO 97/15666). However, endostatin does not directly inhibit the proliferation of tumor cells nor does it induce apoptosis in endothelial cells in pre-existing vessel systems.

The anti-angiogenic activity of endostatin is reported to prevent the formation of new blood vessels which are recruited by fast growing tumors (O'Reilly <BR> <BR> <BR> et a/., 1997; Hanahan and Folkman, 1996). Therefore, it is believed that endostatin may exert anti-tumor effects through starvation of the tumor mass. The mechanism by which endostatin induces not only tumor starvation but also tumor regression is not known. Repeated cycles of endostatin treatment did not cause the development of endostatin-resistant tumors. That result indicated that endostatin might be a promising cure for cancer (Boehm et a/., 1997; Kerbel, 1997).

Several attempts to produce recombinant endostatin in an active, soluble form have been reported in the literature. A review of the reports indicate that the results have been inconsistent and frequently unsuccessful. For example, O'Reilly

et al., 1997, reported their attempt to make soluble murine endostatin in the bacteria Escherichia coli resulted in insoluble material. The report describes that an attempt to refold the insoluble endostatin resulted in the loss of greater than 99% of the material.

This report also describes that murine endostatin was produced as a soluble preparation when using a Baculovirus expression system. However, the yield was reported to be low so the material was not used in in vivo testing. A later report by this group, Boehm et al., 1997, also relied on the use of insoluble murine endostatin made by the method of O'Reilly et al., 1997, and delivered as a suspension.

A report by Hohenester et al., 1998, indicates that murine endostatin was produced as a soluble protein from human embryonic kidney cells. The particular form of murine endostatin produced by those authors included four additional amino acids (Ala-Pro-Leu-Ala) at the N-terminus. Hohenester et al., reported that the X-ray structure of this form of murine endostatin was determined to 1.5 A resolution and revealed a structural similarity to the C-type lectin family but lacking the characteristic Ca2+ binding site. The report proposed a three dimensional structure for murine endostatin. However, the first ten amino acids of the N-terminus and last six amino acids of the C-terminus were not visible and were presumed to be disordered. Hohenester did not report whether the murine endostatin they produced was active or inactive. Therefore, it could not be concluded whether the structure proposed in the article is related to an active form of endostatin.

Two publications describe the production of murine endostatin in the yeast, Pichia pastoris. Boehm et al, (1998) report that soluble murine endostatin was produced from the P. pastoris. However, the report indicated that several batches of murine endostatin produced in P. pastoris contained endostatin having different N- termini. The variations were reported to be due to protease activity in the batches in which cleavage occurred. Boehm et al., also produced murine endostatin from a mammalian tumor cell line, B16F10 melanoma cells. While the murine endostatin so produced was reported to have a full length N-terminus, no data was presented to indicate that the murine endostatin was active or that it was soluble.

Dhanabal et al., 1999, also report on the production of murine endostatin in the yeast P. pastoris. This report does not specifically note the cleavage of the N-terminus described by Boehm et al., 1998, as related to expression in the P. pastoris system, but Dhanabal et al., did classify their yeast produced endostatin at about 20 kDa and characterized that size estimate as being in agreement with the 22- 24 kDa size reported by others. However, while Dhanabal et al.'s P. pastoris derived

murine endostatin may be similar in size to the inactive, N-terminal truncated endostatin prepared in P. pastoris by Boehm et al., 1998, Dhanabal et al., report that their 20 kDa murine endostatin produced anti-angiogenic effects in vivo and in vitro.

Dhanabal et al., also expressed murine endostatin in bacteria according to the method of O'Reilly et al., 1997. It is noteworthy that Dhanabal et al. observed a size of 22-24 kDa for the murine endostatin they produced in bacteria and also observed dimer-sized complexes in the 44-46 kDa range. Like O'Reilly et al., Dhanabal et al., were unable to produce soluble endostatin from bacteria.

One report of the production of recombinant human endostatin, Ding et al., 1998, relied on the use of a murine myeloma cell line to express the protein as an Fc-endostatin fusion protein. Ding et al., reported that Fc-endostatin was purified on Protein A Sepharose and subsequent treatment of this form of endostatin with enterokinase and trypsin resulted in two different forms of cleaved endostatin. When trypsin was used, a cleaved form of endostatin was obtained in which four amino acids (His-Ser-His-Arg) were removed from the natural N-terminus of human endostatin. Treatment with enterokinase resulted in a cleaved form of endostatin with an additional leucine at the N-terminus. Ding et al., also reported that soluble recombinant Fc-endostatin was obtained and it suppressed tumors in mice. However, the report did not explicitly state whether either form of the cleaved endostatins were soluble or whether they showed anti-angiogenic activity.

The Ding et al. publication reported a crystal structure of human endostatin to 2.9 A resolution and indicated that the molecule contains a Zn2+ binding site formed by three Histidine1, 3,11 and one Aspartic acid76 residues. In the endostatin prepared by Ding et al., the Zn2+ was reported to be ligated at His 1 N6, His3N£ and HisllNE. Ding et al., also reported that human endostatin is expected to form zinc-dependent dimers. This statement is in agreement with the 44-46 kDa dimer-sized endostatin material observed by Dhanabal et al., 1998.

In their subsequent report on murine endostatin, Boehm et al., 1998 described that mutations of the putative zinc-chelating residues of murine endostatin to Alanine residues reduced the effect of endostatin on the regression of tumors implanted in mice. Mutant murine endostatin was reported to exhibit a reduction in activity of 50% when Histidine1 1 or Aspartic acid76 were replaced with Alanine and by 80% in a Histidine1/3 to Alanine double mutant (Boehm et al., 1998). However, these mutant constructs contained more differences than the indicated amino acid

changes. Therefore, while these experiments indicate that metal binding might be important, its requirement for the biological activity of endostatin was not shown.

A second report on the production of recombinant human endostatin relied on expression of the protein in human kidney cell culture. (Sasaki et al., 1998) The human endostatin was soluble but no information on the anti-angiogenic activity of that protein was reported. Sasaki et al., also expressed recombinant murine endostatin in soluble form using the kidney cell expression system. The authors stated that the murine endostatin was active, but did not provide data to support that assertion.

Finally, the difficulties in preparing and measuring the activity of endostatins have been reviewed in a commentary by J. Cohen in Science 283: 1250-51 (1999). Cohen reiterated that researchers had been unable to produce soluble recombinant endostatin from E. coli and that this difficulty hampered research using the protein. Cohen also describes as positive results an attempt to produce soluble human endostatin from human kidney cells. However, no details of that production of human endostatin were described by Cohen.

SUMMARY OF THE INVENTION The present invention solves an existing problem in the field of angiogenesis by providing soluble recombinant endostatin and a method of producing said endostatin from Escherichia coli. The endostatin can be from any mammal but human endostatin is preferred. Compositions containing soluble recombinant endostatin made from E. coli are also provided by this invention. Importantly, this invention provides assays conducted using soluble recombinant endostatin produced from E. coli. These assay are advantageous over those known in the art because the endostatin used is easily made by in vitro refolding of insoluble material to obtain homogeneous protein, plentiful, inexpensive and a soluble recombinant product.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1A. Plot of the number of nOe distance constraints, n, versus their range along the amino acid sequence: vertically hatched, intra-residue constraints (these entries would go off-scale; the numbers at the top indicate their actual heights) and constraints between protons in sequentially neighboring residues; horizontally hatched, constraints between residues separated by 2 to 5 positions in the sequence;

white, longer-range constraints. Along the horizontal axis li-j is the sequence separation of the two residues containing the protons considered.

FIG. 1B. Plot of the number of nOe distance constraints per residue, n, versus the amino acid sequence of endostatin. The constraints are specified as follows: black, intra-residue; dark grey, constraints between protons in sequentially neighboring residues; light grey; constraints between protons located in residues separated by 2 to 5 positions along the sequence; white, all longer-range constraints.

FIG. 2. Amino acid sequence of endostatin and survey of the sequential connectivities and additional data collected for secondary structure identification. The sequential nOe connectivities dN, dNN and dnN are indicated with thick, medium or thin black bars for strong, medium and weak nOe's, respectively. Medium-range <BR> <BR> <BR> connectivitiesdNN (iXi+2), dgN (i, i+2), dgN (i, i+3), d22 (ivi+3) andd2N (i,<BR> <BR> <BR> <BR> <BR> <BR> i + 4) are shown by lines starting and ending at the positions of the residues related by the nOe.

FIG. 3A-3B. Stereo view of the polypeptide backbone of the final 20 energy refined DYANA conformers of endostatin used to represent the solution structure of metal-free human endostatin. The conformers were superimposed for pairwise minimum r. m. s. d. of the backbone atoms N, C and C'of residues 8-178.

FIG. 4. In plots versus the amino acid sequence of the mean global backbone displacements per residue, Dbbglob, of the 20 energy-minimized DYANA coordinates relative to the mean NMR structure calculated after superposition of the backbone heavy atoms N, C and C'for minimal r. m. s. d. (solid line). The dashed line represents the mean heavy atom displacement per residue, DHglob.

FIG. 5A. 15N-chemical shift analysis of metal bound and EDTA treated human endostatin. (A) 1D-15N spectra of the imidazole nitrogen atoms of endostatin with and without 4fold molar excess of EDTA. * represent resonances that and + represent peaks that appear after EDTA treatment.

FIG. 5B. 2D-1H-15N HMQC spectrum of the imidazole nitrogen atoms. Interresidue connectivities are indicated by lines.

FIG. 6. Stereo diagram of a model of the endostatin metal binding site in solution.

FIG. 7A-7B. Two views of the polypeptide backbone of the final 20 energy refined DYANA conformers of endostatin used to represent the solution structure of zinc-bound human endostatin. The conformers were superimposed for pairwise minimum r. m. s. d. of the backbone atoms N, C and C'of residues 8-178.

DETAILED DESCRIPTION OF THE INVENTION The present invention provides soluble recombinant endostatin from Escherichia coli, a method of producing the endostatin and compositions including the endostatin.

The soluble recombinant endostatin of this invention is useful in the treatment of cancer, the inhibition of tumor growth, the inhibition of angiogenesis, the isolation of receptors for endostatin and in assays for the identification of anti- angiogenic compounds.

The soluble recombinant endostatin can be made using genetic material from any animal. In particular, soluble recombinant endostatin is produced using genetic material from primates including human, chimpanzee and ape, or from other animals including murine, pig, goat, dog, bovine, avian, whale, porpoise, etc.

Soluble recombinant endostatin made from primate genetic material is preferred and that made from human genetic material is most preferred.

The method of making the soluble recombinant endostatin from Escherichia coli is described in the examples below. However, it will be understood by those who practice protein expression and purification that aspects of the procedure can be modified from the particular details given below. It is intended that the changes and modifications that are within the skill of the art are within the scope of this invention.

In the method of the present invention a prokaryotic expression vector with an inducible promoter is preferred. Many prokaryotic vectors and many inducible promoter systems are known in the art and can be used interchangeably.

One can simply test any given vector or inducible promoter construct against the performance of the particular expression construct used in the examples to assess its suitability.

When practicing this invention, the preferred prokaryotic host is Escherichia coli. Many strains of E. coli are known and used in the art. Any of those strains commonly used for the expression of recombinant proteins can be used in the method of this invention. The most preferred strain of E. coli is BL21 (DE3). An advantage of production of endostatin in E. coli is that one does not typically observe degradation of the protein due to protease activity.

It is preferred that the prokaryotic host cells are grown in a minimal medium so that the induction of the expression construct can be more readily

controlled. A preferred minimal medium contains the following ingredients per 1 L: 5 g D-glucose, 1 mg (+) biotin, 2 mg thiamine, 1 g (NH4) 2S04,0.5% casamino acids, 100 mM potassium phosphate buffer pH 7.2,2 mg MoNa204x2H20,1 mg CoCl2, 0.3 mg CuS04, 4 mg MnS04 x H20, 35 mg MgS04, 5 mg ZnS04 x 7 H20), S mg FeCl2 x 4 H20), 10 mg CaCl2 x 2 H20,4 mg H3BO3, and appropriate antibiotics in sterile water. Induction of the expression construct is achieved by appropriate measures known in the art.

By operably linked, it is meant that nucleotide sequences are aligned in a vector such that they operate together to produce a protein of a particular sequence or regulate the expression of one nucleic acid sequence due to it's proximity or alignment with another nucleic acid sequence. In reference to amino acid sequences, operably linked means that amino acid sequences are aligned in a polypeptide in a fashion appropriate to the operation of the amino acid sequence. For example, an isolation tag is operably linked to an amino acid sequence if it appears in the polypeptide sequence in a position that allows one to isolate the polypeptide through some aspect of the isolation tag. Preferably, an isolation tag appears at a terminus of a polypeptide and is separated therefrom by a cleavable linker, that is, by an amino acid sequence that contains an endopeptidase cleavage site.

In the method of this invention it is preferred that the cells in which endostatin has been expressed are lysed by mechanical methods used in the art. For example. one can lyse cells mechanically using agitation in the presence of glass beads or pressure such as in a French Press. The use of a French Press is the preferred method of lysing the cells.

Once lysed, the cellular debris is pelleted in a centrifuge. Typically, a force of at least about 100,000 g should be generated during centrifugation. However, 80,000g up to 140,000 g can be used. In essence, this step is performed to form a pellet containing the endostatin material produced in the cells. Any force sufficient to achieve that endpoint is appropriate for this step.

Once pelleted, the endostatin material can be washed by resuspending the pellet in any suitable solution. Cold water is appropriate as is any mild buffer such as those based on Tris. A mechanical mixing device such as a pestle or a dounce homogenizer can be used to assist in the resuspension of the pellet.

A variety of isolation tags are known in the art and can be used to assist in the purification of recombinant endostatin from the cellular material. A Histidine tag is preferred or any other tag that allows an affinity purification under

denaturing conditions, e. g., in 6 M guanidine hydrochloride. Endostatin synthesized with such a tag can be isolated on a Ni2+ resin as practiced routinely in the art.

Commercially available affinity columns, including Ni-Ta resin from QUIAGEN or TALON resin from CLONTECH, can be employed.

A number of cleavable linker sequences are known in the art. These linker sequences are useful for removing secretory leader sequences and/or isolation tags from recombinantly expressed protein. For example, factor Xa, caspases, enterokinase and thrombin cleavage sites are used in this manner. In the examples below, the endostatin was made using a linker cleavable by thrombin. That linker leaves two additional amino acids at the N-terminus of the endostatin, Gly-Ser. It is preferred that after cleavage no more than five amino acids are added to the N- terminus. The linker can be cleaved at any appropriate time after the use of the affinity purification step.

After the endostatin is purified using the isolation tag, the concentration of the protein in the solution containing the endostatin is determined.

The endostatin is then diluted into a pre-folding buffer to a final concentration of about 80 pg/ml protein. A suitable pre-folding solution is a reduction/oxidation (redox) buffer. The preferred redox buffer is 2.3 M GuHCI, 100 mM Tris/HCI pH 8.0,20% glycerol, 4 mM reduced glutathione, 0.4 mM oxidized glutathione at 4°C.

The reduced and oxidized forms of glutathione are preferred. However, other redox pairs can be substituted. One can simply test a particular redox pair in the above buffer using the method of producing endostatin taught herein and compare the result to using the buffer with the glutathione pair. In addition to changing the redox pair, the basic components of the above buffer solution can be varied in the range of NaCl about 0-150 mM, GuHCI about 1.5-2.5 M, GSH/GSSH ratio about 10: 1 and concentration from about 1-10 mM, Tris/HCI about 10-150 mM, pH about 7.5-8.5, glycerol about 10-20%.

It is preferred that the endostatin is diluted into the pre-folding solution slowly, even dropwise, while stirring, to ensure even mixing and distribution throughout the solution. The solution should be stirred for a period of time sufficient to ensure that the endostatin has reached equilibrium in its interactions with the components of the pre-folding solution. This may take place as rapidly as within a few hours. However, it is preferred that the solution be stirred at least overnight and preferably at least 36 hours or even as long as two days.

Once the endostatin has been treated with the pre-folding solution, the endostatin is dialyzed against a folding buffer. The folding buffer does not have the redox components of the pre-folding buffer but otherwise is similar. It is also preferred to omit the glycerol present in the pre-folding buffer. During this step, the redox pair is dialyzed out of the endostatin preparation. Adequate volumes buffer are used over a period sufficient to dialyze the level of redox pair below detectable levels.

For example, 4 changes of 6 to 20 volumes of buffer as compared to the volume of the pre-folding solution are typically sufficient.

The actions of proteases have the potential to destabilize a protein of interest, especially during cell lysis and when the protein is in a state of incomplete refolding. One skilled in the art will recognize that as in any method of protein purification, the use of commercially available protease inhibitors can be employed in the purification of recombinant endostatin. For example, during the cell lysis step of Example 1 the addition of EDTA is preferred as protease inhibitor. Also, one can optionally add 1 mM PEFABLOC (BOEHRINGER MANNHEIM) to the lysis buffer and to the solution used to wash of the pellet containing the recombinant endostatin after the lysis. When performing the procedure of Example 1, the addition of protease inhibitors to the solubilization buffer, during the metal affinity purification or the initial pre-fold in 2.3 M GuHCI did not yield any additional stability. However, like many proteins, the incompletely folded recombinant endostatin can be stabilized by protease inhibitors during the dialysis following the pre-fold. For example, as is commonly practiced in the art, one can add a commercially available tablet of a protease inhibitor cocktail (e. g., COMPLETE from BOEHRINGER MANNHEIM, ingredients: pancreas extract, pronase, thermolysin, chymotrypsin, trypsin and papain) and 0.2 mM PEFABLOC per 20 L of dialysis buffer. Higher concentrations of inhibitors can be employed if desired. Once the endostatin is purified to homogeneity the fully folded material is usually very stable and protease inhibitors are not typically needed in the samples used for assays and NMR studies.

For consistency herein, the numbering of the amino acids of endostatin is +1 for the native N-terminal amino acid and +183 for the native C-terminus.

Where additional amino acids are present at the N-terminus, the additional amino acid adjacent to the native N-terminal amino acid is at position-1, the next additional amino acid is-2, an so on. We prefer endostatin having a native N-terminus, however, endostatin with additional amino acids are also useful. In particular,

endostatin having the N-terminal sequence Gly2Ser-lHis+lSer+2His+3 (SEQ ID NO: 1) is a most preferred endostatin of this invention.

Assays of this invention use soluble recombinant endostatin produced from E. coli. These assay are advantageous over those known in the art because one employs a plentiful, readily available, inexpensive, soluble recombinant endostatin produced from E. coli. Therefore, the assays of the present invention can be performed at lower cost and higher throughput than present assays.

The soluble endostatin is used as a standard in the ex vivo rat aortic ring assay. In this assay, endothelial cell tube outgrowth is regulated by autocrine, paracrine, and junxtacrine interactions among endothelial cells, pericytes and fibroblasts. This ex vivo model comprises the necessary sequence of events leading to the formation of new vessels including proliferation, migration and canalization.

The soluble endostatin is used as a standard in a high throughput screening (HTPS) assay based on the endothelial cell proliferation assays. In this assay endothelial and non-endothelial control cells are seeded into 96-well plates and the rate of their proliferation is measured either by 3H-thymidine incorporation or conversion of (3- (4, 5-dimethylthiazol-2-yl)-5- (3-carboxymethoxyphenyl)-2- (4- sulfophenyl)-2H-tetrazolium, inner salt (MTS reagent) into a soluble formazan dye by metabolically active cells. The MTS assay can be used not only to measure relative proliferation but also compound toxicity by comparing treated cells with untreated controls.

The soluble endostatin is used as a standard in the semi-HTPS endothelial cell migration assay. In this assay endothelial and non-endothelial control cells are evaluated for their ability to migrate from the top layer of a Sum membrane to the bottom layer in response to a chemoattractant such as a growth factor or tumor cell conditioned medium.

Soluble endostatin was made by the method taught herein and the structure of that endostatin was determined using solution Nuclear Magnetic Resonance. Both the structures of endostatin with bound zinc and metal-free endostatin were determined. The data for both metal-free and metal bound forms of endostatin indicated that the endostatin of the present invention exists in monomeric form in solution. The endostatin of the present invention was also determined to be monomeric in a sedimentation experiment described below. This characteristic of the present soluble endostatin is in contrast to the indications in the literature that

endostatin associated to form dimers (Ding et al., 1998; Dhanabal et al., 1999 Hohenester et al., 1998, Hohenester, E., Sasaki, T., Olsen, B. R. and Timpl, R. (1998) EMBO J. 17,1656-1664.). Therefore, soluble preparations of endostatin of the present invention are provided in which the endostatin is substantially monomeric. It is preferred that an endostatin preparation has at least about 80% monomeric endostatin, more preferable at least about 85-90% monomeric and most preferably at least about 95% to 100% monomeric endostatin. If the preparation is at least about 80% monomeric endostatin, the preparation is referred to as"substantially" monomeric. If the preparation is at least about 95% monomeric the preparation is referred to as"essentially"monomeric.

The solution structure of the soluble recombinant human endostatin of the present invention agrees in general with the structures of human endostatin proposed by Ding et al., 1998 and that proposed for murine endostatin by Hohenester et al., 1999. However, neither previous report was able to characterize the interaction of the N and C-termini of endostatin. It has been discovered that for the endostatin of this present invention, the N and C-termini of the molecule adopt a particular orientation with respect to each other and that this alignment stabilizes the C-terminus of the molecule.

The particular orientation of the N and C-termini of the soluble human endostatin of the present invention is shown in FIG. 3A-3B for metal-free and FIG.

7A-7B for Zn-bound. The three dimensional orientation of the peptide backbone of the endostatin of this invention can also be represented by the data points presented in Table 2. It is possible that a particular orientation must be achieved in solution for endostatin to be active. Therefore, the endostatin provided by the present invention can be present in a particular three dimensional structure shown in FIG. 3A-3B, in FIG. 7A-7B or in Table 2. It is preferred that an endostatin preparation has at least about 80% endostatin of a particular three dimensional conformation, more preferable at least about 85-90% endostatin of a particular three dimensional conformation and most preferably at least about 95% to 100% endostatin of a particular three dimensional conformation.. If the preparation is at least about 80% endostatin of a particular three dimensional conformation the preparation is referred to as "substantially"in that conformation. If the preparation is at least about 95% endostatin of a particular three dimensional conformation the preparation is referred to as"essentially"in that conformation.

The solution NMR structure of human endostatin in the metal-bound form that confirms that the metal chelating residues His 1, His3, His 11 and Asp76 are in close proximity. Direct detection of 15N-data of the metal-bound form of endostatin prepared as taught herein show that the N6 imidazole nitrogen atoms of residues His1, His3 and His1 1 are chelating Zn2+. This finding for the endostatin of the present invention is in contrast to a previous report (Ding et al., 1998) which stated that for their endostatin, the Zn2+ was ligated at His1N6, His3N£ and His I I Nc.

Therefore, the endostatin of the present invention, when bound to zinc, is ligated at the Nb imidazole nitrogen atoms of Histidine residues 1,3, and 11.

The following Examples are provided to assist the understanding of the present invention and to aide in the practice of the present invention. The Examples are not limiting of the scope of the invention which is defined only by the claims below.

In the Examples, certain abbreviations are used: NMR, nuclear magnetic resonance 2D, two-dimensional 3D, three-dimensional NOESY, nOe spectroscopy COSY, correlation spectroscopy HSQC, heteronuclear single quantum coherence nOe, nuclear Overhauser effect TOCSY, total correlation spectroscopy, 3JHNHr ; PFG, pulse field gradient ct, constant time ppm, parts per million <BR> <BR> 3 JHNH ! 4 vicinal spin-spin coupling constant between the amide proton and the D- proton H/D, hydrogen/deuterium r. m. s. d., root mean square distance

EXAMPLE 1 Method of Making Soluble Recombinant Metal-Free Endostatin Human endostatin was PCR amplified from a human cDNA library (Quick Screen, Clontech) using primers 5'- TTGGATCCCACAGCCACCGCGACTTC-3' (SEQ ID NO: 2) and 5'- TTGTCGACCTACTTGGAGGCAGTCATGAA-3' (SEQ ID NO: 3). The PCR product was digested with BamHI/SalI and cloned into a pET30a (+) vector (Novagen) modified to contain coding sequences for a His6-tag and thrombin recognition sequence. The relevant sequences of the construct are MHHHHHHGLVPRGS- endostatin (SEQ ID NO: 4). After thrombin cleavage the non-native residues Gly-Ser remain on the N-terminus of the endostatin fragment.

The endostatin vector was used to transform Escherichia coli strain BL21 (DE3). The bacteria were cultured in minimal medium (Pryor and Leiting, 1997) and the expression of His-tagged endostatin was induced using 1 mM IPTG for 3-4 hours at 37°C. Where appropriate, the endostatin was isotope labeled during expression. Cells were collected, washed once with TNE (10 mM Tris pH 7.4,100 mM NaCI, 10 mM EDTA), lysed by french press (20 k cell from SML Amino, 1200 PSI, single run 4°C) and centrifuged (30 min, 120,000g, 4°C). The supernatant was discarded. The pellet was washed by resuspension in ice cold water using a dounce homogenizer followed by centrifugation (15, 000g, 15 min). The pellet was solubilized in buffer A (6 M GuHCl, 10 mM Tris pH 8.0,100 mM NaCI, 50 mM sodium phosphate pH 8.0) and the remaining cell debris was removed by centrifugation (12,000g, 15 min, 4°C).

The supernatant was loaded on a metal affinity column (Talon resin, Clontech), washed intensively (flow rate 1-2 ml/min washed with at least 10 column volumes of buffer A. The tagged endostatin was eluted with buffer A containing 50 mM imidazole. Endostatin containing fractions were pooled and slowly (20 drops per minute) diluted into a volume of pre-folding solution that would contain a final concentration of 80 pg/ml endostatin protein, 2.3 M GuHCl, 100 mM Tris/HCl pH 8.0,20% glycerol, 4 mM reduced glutathione, 0.4 mM oxidized glutathione, at 4°C.

The pre-fold solution was stirred for >36 hours at 4°C and then dialyzed against 6-20 volumes of 50 mM Tris/HCl pH 8.0 over a 2 day period exchanging the buffer 4 times (6-20 volumes each change). The dialysate was centrifuged at 10, 000g to remove a

moderate amount of precipitated protein and the supernatant was sterile filtered (0.2 pM). The supernatant was treated with thrombin at 375 U/L for 1 hour at room temperature to cleave off the His tag.-The reaction was terminated with 1 mM benzamidine.

Soluble recombinant endostatin was purified on SP Sepharose (Pharmacia, buffer: 10 mM Tris pH 7.4, gradient 0-750 mM NaCI). Endostatin containing fractions were pooled and stored after shock-freezing at-80°C.

To finally prepare samples of metal-free endostatin for assays etc., the protein solution was defrosted, dialyzed against PBS and concentrated to 1-10 mg/ml in Biomax 1 Ok device from Millipore (or other appropriate concentration devices).

Method of Making Soluble Recombinant zinc-bound Endostatin Zinc-bound endostatin was prepared from zinc-free endostatin as described above (after SP-Sepharose purification and storage at-80°C). The protein was dialyzed against PBS (for assays) pH 7.2 or 20 mM potassium phosphate buffer pH 6.8 (for NMR) containing 0.1-5 uM zinc sulfate such that the total molar amount of zinc in the dialysis bath is not more than 10-fold the total molar amount of endostatin in the dialysis bag (MWCO 3,500) for >12 hours at 4°C. The dialysis bath was exchanged once against PBS pH 7.2 or 20 mM potassium phosphate buffer pH 6.8 (for NMR) containing no zinc to remove excess zinc from the protein solution (>5 hours, 4°C). Finally, the protein was concentrated to 1-10 mg/ml in BIOMAX 1 Ok device from MILLIPORE (or other appropriate concentration devices) Alternatively, it is also possible to obtain zinc-bound endostatin by refolding in the presence of 1-10 uM zinc sulfate. Zinc-bound endostatin is after the zinc-substitution stable also in zinc-free buffers.

A preparation of soluble recombinant endostatin was prepared from cells grown in minimal medium with 13C and 15N (minimal medium as described above without casamino acids using 13C6-glucose and (15NH4) 2S04 as the sole sources for carbon and nitrogen). This doubly labeled was endostatin used to determine the three-dimensional structure of endostatin. The preparation was refolded in the presence and absence of ZnS04 to study the structure with and without zinc.

EXAMPLE 2 Mass Spectroscopy and Metal Analysis Purified samples of soluble endostatin expressed in the presence of 15N and/or 13C contained >99% 15N and/or 13C incorporation by mass spectroscopy using a Finnigan MAT TSQ7000 triple quadrupole mass spectrometer equipped with an Applied Biosystems 130 HPLC. The disulfide bonding pattern of soluble recombinant endostatin was confirmed by peptide mapping after tryptic digests.

Peptide fragments were analyzed by matrix-assisted laser desorption ionization time of flight mass spectrometry (MALDI-TOF-MS) using a PERSEPTIVE BIOSYSTEMS VOYAGER-DE STR mass spectrometer operated in reflective mode.

Inductively coupled plasma atomic emission spectroscopy (ICP-AES) was used to determine the metal composition in protein samples using a PERKIN-ELMER OPTIMA 3300 DV ICP-AE Spectrometer. N-terminal sequences of endostatin fragments were confirme by N-terminal protein sequencing (Edman degradation) followed by mass spectroscopy.

EXAMPLE 3 Nuclear Magnetic Resonance Measurements NMR Sample Preparation Purified endostatin was dialyzed against 20 mM potassium phosphate buffer, pH 6.7, and concentrated in a Biomax 1 Ok device (Millipore). The NMR measurements were made with 0.5-1.2 mM solutions of endostatin in 99.9% 2H20 or a mixture of H20/2H20 (90: 1 Ovol./vol.) at pH 6.8.

NMR Spectroscopy All NMR spectra were recorded at 30.0°C on a Varian 600MHz Inova system with triple resonance PFG (H, 15N, 13c) probes or a Broad Band probe using either uniformly l 3C/l SN doubly labeled endostatin or uniformly 15N-labeled endostatin. Quadrature detection in the indirect dimensions was achieved using States-TPPI (Marion et al., 1989). Spectra were processed with NMRPipe (Delaglio et al., 1995) on an IBM SP2 computer, using forward-backward linear prediction and zero filling in each indirect detected dimension. The processed data were analyzed

with XEASY (Bartels et al., 1995) on an IBM SP2 computer for visualization of NMR data, peak-picking and peak integration.

Resonance assignments and the input for the structure calculation were obtained from gradient-enhanced (Muhandiram and Kay, 1994) versions of the following experiments. Using 13C/l5N-labeled metal-free endostatin: 3D CBCANH (Grzesiek and Bax, 1992a). Time domain data size 60 x 24 x 1024 complex points, t1,max(13C) 5.7ms, t2, max (15N) 10.9ms, t3, max (1H) 124. 1ms.

3D CBCA (CO) NH (Grzesiek and Bax, 1992b). Time domain data size 60 x 24 x 1024 complex points, 5.7ms,t2,max(15N)10.9ms,t3,max(1H)124.1ms.(13C) 3D HNCA (Grzesiek and Bax, 1992c; Yamakazi et al., 1994). Time domain data size 40 x 24 x 1024 complex points, t1,max (13C) 9.1ms, t2,max (15N) 10.9ms, t3,max (1H)124.1ms.

3D HN (CO) CA (Grzesiek and Bax, 1992a). Time domain data size 40 x 24 x 1024 complex points, t l, max (13 C) 9.1ms, t2maX (15N) 10.9ms, t3, max (1H) 124. 1ms.

3D HBHA (CO) NH (Grzesiek and Bax, 1993). Time domain data size 64 x 24 x 1024 complex points, t1,max (1H) 12.8ms, t2, max (15N) 10.9ms, t3, max (1H) 124. 1ms.

3D (HB) CBCACO (CA) HA (Kay L. E., 1993) Time domain data size 64 x 24 x 1024 complex points, t i, max (13 C) 7. 1ms, t2, max (13 CO) 13.3ms, t3, max (1H) 124.1ms.

3D HNCO (Grzesiek and Bax, 1992a). Time domain data size 32 x 32 x 1024 complex points, t1 max (13C) 14.5ms, t2,max (15N) 14.5ms, t3SmaX (1H) 124. 1ms.

3D C (CO) NH (Grzesiek et al, 1993). Time domain data size 96 x 24 x 1024 complex points, tlmax (13C) 9.3ms, t2, max (15N) 10.9ms, t3,max (1H) 124.1ms.

3D HC (CO) NH (Grzesiek et al, 1993). Time domain data size 64 x 24 x 1024 complex points, t1,max (1H) 7.7ms, t2, max (15N) 10.9ms, t3, max (1H) 124. 1ms.

3D HCCHTOCSY (Bax et al., 1990; Sattler et al., 1995) Time domain data size 64 x 64 x 1024 complex points, tl, max (IH) 7.7ms, t2, max (13C) 5. 33ms, t3, max (1H) 124.1ms.

2D [13C, 1H]-COSY (Bodenhausen and Ruben, 1980; Kay et al., 1992). Time domain data size 320 x 1024 complex points, t1,max (13C) 26.7ms, t2,max (1H) 124.1ms.

2D ct- [13C, IH]-COSY (Vuister and Bax, 1992). Time domain data size 320 x 1024 complex points, tl, max (13C) 26.7ms, t2, max (1H) 124.1ms.

3D 13C-resolved [I H, 1 H]-NOESY (Ikura et al., 1990). Time domain data size 80 x 64 x 1024 complex points, t1,max (1H) 9.7ms, t2, max (13 C) 9.7ms, t3, max (1H) 124.1ms.

3D 15N-resolved [1H, 1H]-NOESY (Fesik and Zuiderweg, 1988). Time domain data size 104 x 24 x 1024 complex points, t1,max (1H) 12.6ms, t2, max (15N) 10.9ms, t3, max (1 H) 124. lms.

3D 15N-resolved [1H, 1H]-TOCSY (Fesik and Zuiderweg, 1988). Time domain data size 96 x 24 x 1024 complex points, t 1, max ( H) 11.6ms, t2,max (15N) 10.9ms, t3, max (1 H) 124.1 ms.

Using 13C/l 5N-labeled ZN-endostatin: 3D CBCANH (Grzesiek and Bax, 1992a). Time domain data size 48 x 24 x 1024 complex points, t2,max(15N)10.0ms,t3,max(1H)124.1ms.5.3ms, 3D CBCA (CO) NH (Grzesiek and Bax, 1992b). Time domain data size 48 x 24 x 1024 complex points, t1max (13C) 5.3ms, t2, max SN) 10. Oms, t3, max I H) 124.1ms.

3D HNCA (Grzesiek and Bax, 1992c; Yamakazi et al., 1994). Time domain data size 40 x 24 x 1024 complex points, t 1, max (13 C) 9. 1 ms, t2, max (N) 10. 0ms, t3, max (1H) 124.1ms.

3D HN (CO) CA (Grzesiek and Bax, 1992a). Time domain data size 40 x 24 x 1024 complex points, 9.1ms,t2,max(15N)10.0ms,t3,max(1H)124.1ms.(13C) 3D HBHA (CO) NH (Grzesiek and Bax, 1993). Time domain data size 64 x 24 x 1024 complex points, 12.8ms,t2,max(15N)10.0ms,t3,max(1H)124.1ms.(1H) 3D HCACO (Vuister, G. W and Bax A. J. Am. Chem. Soc. 115,7772-7777) Time domain data size 24 x 32 x 1024 complex points, t1,max (13CO) 10.0ms, t2,max (13C) 6.7ms, t3, max (1H) 124.1ms.

3D HNCO (Grzesiek and Bax, 1992a). Time domain data size 32 x 32 x 1024 complex points, tlmax (13C) 14. 5ms, t2, maX (15N) 13.3ms, t3, max (1H) 124.1ms.

3D C (CO) NH (Grzesiek et al, 1993). Time domain data size 96 x 24 x 1024 complex points, tl, max (13C) 9.4ms, t2SmaX (15N) 10. 0ms, t3maX (1H) 124.1ms.

3D HC (CO) NH (Grzesiek et al, 1993). Time domain data size 96 x 24 x 1024 complex points, t1,max (1H) 12.8ms, t2, max (15N) 10.0ms, t3, max (1H) 124. 1ms.

3D HCCHTOCSY (Bax et al., 1990; Sattler et al., 1995) Time domain data size 64 x 96 x 1024 complex points, t1,max (1H) 8.0ms, t2,max (13C) 8.0ms, t3, max (1H) 124.1ms.

2D ct- [13C, I H]-COSY (Vuister and Bax, 1992). Time domain data size 320 x 1024 complex points, tl, maX (13C) 26.7ms, t2, maX (1H) 124. 1ms.

3D 13C-resolved [1H, 1H]-NOESY (Ikura et al., 1990). Time domain data size 80 x 64 x 1024 complex points, tlmax (1H) 9.7ms, t2,max (13C) 9.7ms, t3,max (1H) 124.1ms.

3D 15N-resolved [1H, 1H]-NOESY (Fesik and Zuiderweg, 1988). Time domain data size 80 x 24 x 1024 complex points, t1max (1H) 9.6ms, t2maX (15N) 10. 0ms, 124.1ms.t3,max(1H) 3D 15N-resolved [1H, 1H]-TOCSY (Fesik and Zuiderweg, 1988). Time domain data size 96 x 24 x 1024 complex points, t1,max (1H) 11.6ms, t2, max (15N) 10. 0ms, t3, max (1 H) 124.1 ms.

Using 15N-labeled ZN-endostatin: 2D [15N, 1H]-COSY (Bodenhausen and Ruben, 1980). Time domain data size 256 x 1024 complex points, tl, max (15N) 116.3ms, t2, max (1H) 124.1ms.

2D 1H 15N HMQC-PS (Summers et al., 1986; Boyd et al. 1992) Time domain data size 256 x 2048 complex points, tl, max (15N) 20.5ms, t2, max (1H) 163. 3ms.

The carrier position was set to 4.74 ppm for 1H, 117.1 ppm for 15N, 200.6 ppm for 15N HMQC, 55.5 ppm, 43.1 ppm or 174.9 ppm for 13 C 13Caliph or 13C'carbon atoms. The 1H chemical shifts are relative to internal 2,2-dimethyl-2- silapentane-5-sulfonate sodium salt (DSS). The 15N and 13C chemical shifts are relative to DSS using the conversion factors that have been reported by Wishart et al.

(1995).

Vicinal 3JHNHn coupling constants were determined by inverse Fourier transformation of in-phase multiplets (Szyperski et al., 1992) from a 2D [15N, 1H]-COSY spectrum recorded with a data size of 256 x 1024 complex points, ti rnax t2,max(1H)(15N)10.0ms, 128.0ms.

For the estimation of the amide proton exchange rates, a 1.0 mM solution of fully protonated, uniformly 13C/l5N-labeled, endostatin was lyophilized and redissolved in 2H20. A 2D [15N, 1H]-COSY spectrum was recorded with a time domain data size of 128 x 1024 complex points, t1smax (15N) 10-0ms, t2, max (1H) 128. 0ms at 30°C. Those amide protons that remained after 72 hours were considered to have exchange rates <<1s-1.

Assignent of resonances and collection of conformational constraints.

The complete assignment of 1H, 15N and 13C chemical shifts and the collection of nOe's were done simultaneously using the program GARANT (Bartels et al., 1996; Bartels et al., 1997) and-a method of application (O'Connell et al., 1999).

All GARANT assignments were done using the complete spectral data set described in the methods section. It should be noted that in each application of GARANT all spectra were assigned over again according to their consistence within the spectral data itself and additionally provided information such as fixed disulfide bonds, secondary structure as determined by chemical shift indices, three-dimensional structures etc. Initially, about 450 intraresidue and sequential nOe's were assigned based on NOESY probabilities derived from the sequence and used together with the upper and lower distance constraints for the two disulfide bonds and consecutive secondary structure information obtained from the C D chemical shifts as input for DYANA structure calculations. The low resolution structures obtained from this calculation were used as an input for a second run of GARANT from which about 1200 assignments were obtained. These assignments were cycled back into DYANA to result in a second generation of low resolution structures which essentially contained already the fold of the protein at r. m. s. d's of about 12. Interestingly, the r. m. s. d. analysis of this structure set revealed two groups of structures which were mirror images of each other. This is not surprising for a protein structure such as the one of endostatin that contains only 34% total defined secondary structure composed of short elements with an average length of 4.5 residues. The correct set of structures was identified using a combined approach of violation analysis, r. m. s. d cluster analysis and molecular dynamics calculations. In deviation from the originally published assignment method (O'Connell et al., 1999), the application of unsolvated restrained molecular dynamics using the program GROMOS96 helped to solve the symmetry problems and to prevent inherent problems associated with"lonely"long range upper distance constraints. DYANA uses a repel type function for non-bonded contacts and does not treat electrostatic interactions. Using DYANA in stages of incomplete assignments as for these low resolution DYANA structures, incorrect "lonely"long range nOe's often distort secondary structure elements resulting in bending of helices or separation of _-sheets. Using such structures as an input for GARANT may drive the assignments into local minima that prohibit the correction of false nOe's as well as the complete assignment of all resonances. The unrestrained,

solvated GROMOS96 calculations allowed for examining the early structures with a full force field and investigated the overall structural stability and optimal geometry.

About 10-12 cycles of GARANT/DYANA/interactive violation analysis were performed to complete the assignments at the same time while the structures improved. At further advanced stages of the process those assignments that were reproducibly obtained after each GARANT application were fixed. The process was continued until nearly all NOESY cross peaks were assigned and almost complete chemical shift assignments were obtained for all residues.

Collection of additional constraints The input for the DYANA calculations consisted of upper distance limits derived from NOESY cross-peak intensities with the program CALIBA (Giintert et al., 1991 a; Guntert et al., 1991b), dihedral angle constraints obtained from the program HABAS (Guntert et al., 1991 a; Guntert et al., 1991b; Guntert et al., 1990) and spin-spin coupling constants 3JHNHn together with backbone dihedral angle constraints derived from conformation-dependent 13C 7 chemical shifts (Luginbuhl et al., 1995). Upper and lower distance constraints were included for the disulfide bonds within DYANA. No additional constraints were used to enforce hydrogen bonds implicated by nOe's or by amide proton exchange data at any time during the DYANA structure calculations or energy minimizations.

EXAMPLE 4 Structure Determinations Assignment Strategy Chemical shifts and nOe's were assigned simultaneously using GARANT version 2.0 (Bartels et al., 1996; Bartels et al., 1997) on an IBM SP2 with eight 591 processors with an input of all XEASY (Bartels et al., 1995) peak-picked 3D-spectra and a 2D [15N, IH]-COSY as described (O'Connell et al., 1999). The NOESY probabilities were set to 1.0,0.5 and 0.25 for distances less than 3.0 A in 20/20 structures, 4.0 A in 15/20 structures and 5.0 A in 10/20 structures, respectively.

Chemical shift tolerances were defined to 0.3 ppm for heteronuclear dimensions, 0.05 ppm and 0.03 ppm for all proton dimensions. The secondary shifts for the C'. C and C carbons for residues in well defined secondary structure were included. The population size of GARANT was typically 100 to 150 using the standard optimization

macro and the calculation times were 100 to 200 hours per run. Multiple GARANT assignments were run with different random seeds and compared until they were consistent between calculations. Some characteristic automated assignments were cross-validated with manual assignments using the standard strategy (Wuthrich, 1986; Powers et al., 1992).

Structure Calculation and Refinement The structures were calculated using multiple cycles of GARANT, DYANA (Giintert et al., 1997) and an interactive violation analysis using the structure analysis program XAM (Xia 1992). For visual comparison of the structures, stereo views were produced with either the molecular graphics program MOLMOL (Koradi et al., 1996) or with the structure analysis program XAM (Xia 1992) as modified by O'Connell on an IBM RS6000 workstation. Modifications include a Tcl/Tk graphical interface and additional analysis features. The superposition algorithm and graphics were used as originally developed (Xia 1992). For pairs of conformers, global superposition's and r. m. s. d. values for various subsets of atoms were computed (Mclachlan 1979). The mean solution conformation was obtained by first superimposing the 20 energy minimized DYANA conformers so as to minimize the r. m. s. d. for the backbone atoms N, C, and C'of residues 8-178, and then averaging the Cartesian coordinates of the corresponding atoms in the 20 globally superimposed conformers. Displacements (D) as defined by Billeter et al. (1989) were used to quantify the local precision of the solution structure (FIG. 6). In the group of conformers used to represent the solution structure, D was obtained as the mean standard deviation for the atom positions of N, C 2 and C'after global superposition of the individual conformers with the average coordinates for the backbone residues.

In early stages the structures were also subjected to unrestrained, solvated molecular dynamics using GROMOS96 (BIOMOS) for geometry optimization and to check the stability of local structure. The molecular dynamics calculations were run at 300K using a time step of 2 fs with a temperature bath coupling time steps of 0.1 ps. Typical simulation times were 50-500ps.

The GARANT/DYANA/validation applications were continued until complete sequential assignments and a maximum number of nOe assignments with the lowest number of violations were achieved. The 20 structures with the lowest DYANA target functions were chosen from calculations with the final input data set starting with 100 random structures using the torsion-space molecular dynamics

annealing. These 20 structures were subjected to restrained energy minimization with OPAL (Luginbiihl et al., 1997). The DYANA, GROMOS96 and OPAL calculations were run on an 8 processor Cray T916-8256 computer and the XAM on an IBM RS6000 workstation. During minimization the pseudo-energy was proportional to the sixth power of the distance constraint violations, and was adjusted such that violations of 0.15 A for the distance constraints and 2.5° for the dihedral angle constraints corresponded to 1/2kBT at room temperature (Billeter et al., 1990). A cutoff of 10.0 A was used for pairs of interacting atoms in the energy evaluation. The resulting 20 OPAL energy-minimized DYANA conformers are used to represent the solution <BR> <BR> <BR> <BR> conformation of endostatin. All , y backbone torsion angles within the ordered parts of the protein lie within the allowed regions of the Ramachandran plot. The Ramachandran values of the complete protein for all conformers is 68.3% in the most favored regions, 29.1 % in the additionally allowed, 2.4% in the generously allowed and 0.2% in the disallowed regions.

Structure Analysis.

Structure Determination The input of nOe distance constraints for endostatin was derived from the 3D 15N-resolved [1H, 1H]-NOESY in H20 and 3D 13C-resolved [1H, 1H]- NOESY in 2H20. For the calibration of lH-lH upper distance limits, r, versus the NOESY cross peak intensities with the program CALIBA (Guntert et al., 1991 a; Guntert et al., l 991b) a 1/r6 dependence was used for backbone proton-backbone proton nOe's, and a 1/r4 dependence for nOe's involving side chain protons and methyl groups using the standard protein calibration macro. These automated calibration curves were prepared for DYANA input by refinement based on plots of cross peak volume versus average proton-proton distances in sets of preliminary structures. The proton-proton upper distance bounds before application of pseudoatom corrections were in all cases confined to the range 2.4-5.0 A in order to obtain reasonable upper bounds for both strong and weak nOe's. A total of 2901 NOESY cross-peaks were assigned and used for the generation of the input of upper-limit distance constraints for the structure calculation. Of these, 854 resulted from the 3D 15N-resolved [1H, 1H]-NOESY and 2047 from 3D 13C-resolved [1H, 1H]-NOESY.

Of these, 659 peaks characterized identical nuclei in both spectra and 102 constraints were found to be irrelevant on the basis that they are either independent of conformation or that there exists no conformation that would violate the constraint. In addition, a total of 109 3 JHNHm coupling constants were included. From the

HABAS interpretation of the backbone-backbone intra-residual and sequential nOe <BR> <BR> upper distance limits together with the spin-spin coupling constants and 13C D<BR> <BR> <BR> <BR> <BR> <BR> chemical shift indices, a total of 123 + and 121 D backbone dihedral angle constraints were generated. These conformational constraints comprised the final input for the structure calculations with the program DYANA, which included 2140 nOe upper distance constraints and 244 dihedral angle constraints. The range and sequence distributions of the nOe distance constraints are shown in FIG. 1. The nOe constraints obtained may be classified into four classes with lj-il = 0, lj-il = 1,2 Ij-il : < 5, lj-il > 5, where j and i are the residue positions in the sequence that contain the two protons for which the nOe is observed. The number of distance constraints in these four classes are, 228,503,393,1016, respectively. The structure was solved single- handedly in five months including greater than two months of NMR data collection.

Structure Refinement The 20 conformers with the lowest target function values resulting from the final DYANA calculation were subjected subsequently to solvated, restrained energy-minimization with the program OPAL. An analysis of the structures before and after restrained minimization is presented in Table 1. The best 20 DYANA structures before energy-minimization have low target function values, and satisfy the nOe distance constraints and dihedral angle constraints nearly perfectly. The OPAL energies range from-4933 to-4607 Kcal/mole with van der Waals energies in the range-537 to-484, which is a reduction of about 3000 Kcal relative to the unrefined DYANA conformers, respectively (Table 1). In the set of structures, all residual violation of upper distance limits were less than 0.17 A with a r. m. s. d of 0.05 A (Table 1). All dihedral angle violations were less than 2.1 ° with an r. m. s. d of 0.56° (Table 1).

A visual impression of the solution structure and the quality of the structure determination is afforded in FIG. 3A-3B, which shows the superposition of the polypeptide backbone in the final 20 representation conformers. The core structure is well defined, as indicated by the displacements in FIG. 4 and the average pair-wise <BR> <BR> <BR> r. m. s. d. values between the individual conformers and their mean of 0.27 0.03 A<BR> <BR> <BR> <BR> <BR> <BR> (range 0.22 to 0.35) for the backbone atoms, 0.58 0.05 A (range 0.52 to 0.73) for all heavy atoms of residues 8 to 178 excluding the less well defined N-and C-termini.

EXAMPLE 5 Solution Structure of Endostatin Structure of endostatin and comparison to the crystal structure A survey of the sequential nOe's and additional data used to derive the secondary structure of endostatin is shown in FIG. 2. There is a close correlation of the secondary structure with the sequential nOe's, 13Co chemical shift indices and scalar 3JHNHm coupling constants as well as slow H/D exchanging backbone amides.

About 25% of backbone amides have slow exchange rates indicating the tertiary stability of endostatin, though the overall fold contains only 34% of defined secondary structure elements and predominantly loop regions. However, a large number of hydrogen bonds were located in the final structures.

The secondary structure assignments determined by MOLMOL and the NMR data are as follows: O-helices, (A) 26-40 and (B) 64-67, (C) 83-88 and (D) 135- 138; 3-10 helices (E) 57-59 and (F) 152-154; D-stands (1.1) 9-14, (1.2) 46-48, (1.3) 71-73, (1.4) 78-80, (1.5) 101-102, (1.6.) 106-107, (1.7) 172-177, (2.1) 118-120, (2.2) 146-151, (2.3) 161-164. These D-stands correspond to sheets A, B, E, F, 1, P, J, M and O, respectively, in the X-ray structure with some differences in the length of the sheets (Hohenester et al., 1998). Sheets C, D, K, L, G and N assigned the X-ray structure which are comprised of very short elements are not designated as sheets by MOLMOL although they are in an extended conformation.

The short helices B, D, E and F in the NMR structure are not designated as such in the published X-ray structure, though they are turn residues.

Thus the NMR secondary structure is similar to the crystal structures (Hohenester et al., 1998; Ding et al., 1998) with slight variations in surface exposed regions which could be influenced by crystal packing. The r. m. s. d between 1BNL (Ding et al. 1998) and one of the NMR conformers is 1.35 A for the backbone atoms and 2.14 A for all heavy atoms of residues 8 to 178.

EXAMPLE 6 Metal Binding bv Endostatin Metal specificity of endostatin The metal analysis of endostatin samples which were refolded in the presence of zinc showed 95-100% stoichiometric zinc incorporation. Interestingly, although significant amounts of Coll were always present in the refold solution which was derived from the affinity purification resin, cobalt was never found in finally purified endostatin. Similarly, the selective labeling of endostatin with CdII to enable 113Cd NMR experiments by adding CdC12 instead of ZnS04 to the refold and dialysis solutions was attempted. However, fully zinc-labeled endostatin was obtained with no traces of cadmium. The analytical inspection of the CdC12 used for the refolding of endostatin showed that it was contaminated with 0.005% ZnII. Since the refolding and dialysis of endostatin required large volumes of buffer the total amount of ZnII provided with 5 orders of magnitude excess CdII was about equimolar. This suggests that the selectivity of endostatin for ZnII versus CdII and Coll is at least 105-fold.

NMR analysis of zinc-bound endostatin in solution An 15N-labeled endostatin sample refolded in the presence of ZnII was determined by ICP-AE spectroscopy to contain 0.95 mol% ZnII. This sample was used to measure directly the 15N-chemical shifts of the imidazole ring nitrogen atoms in the 7 His residues in endostatin by 1D-15N NMR (FIG. 5 A) and correlate them to the imidazole ring protons using an 15N-HMQC spectrum (FIG. 5 B). The goals of these experiments were to identify the number of His residues that are in complex with the metal ion and to determine whether the N6 or N imidazole nitrogen atoms are in contact with the metal ion. To investigate the number of histidine residues involved, 1D-l 5N NMR spectra of metal bound endostatin were recorded with and without EDTA (FIG. 5A). EDTA can e used to remove ZnII from endostatin when applied in 15 fold excess by dialysis. Three peaks (FIG. 5A, marked as *, 208-230 ppm) that were present in the metal bound form and occur at chemical shift resonances consistent with metal binding (Bachovchin, 1986), were significantly reduced after EDTA addition while three new peaks appeared (176-180 ppm) in the region of protonated nitrogen chemical shifts (Bachovchin, 1986; Gooley et al., 1993). This indicated the extraction of zinc from the metal binding site which made

those histidines that were metal bound before accessible for protonation at pH 6.8.

The imidazole nitrogen resonances between 158-182 ppm represent protonated nitrogen atoms (Bachovchin, 1986) which are expected for surface accessible residues at this pH.

The addition of EDTA also shifted the nitrogen chemical shift of two of the metal bound histidines (FIG. 5A) which suggests a transition state where EDTA may be bound to the zinc ion before the metal binding site is destroyed, thereby influencing the chemical shift of two neighboring imidazoles. The NE chemical shift of one of these metal bound imidazole ring systems is also slightly impacted (181 ppm, FIG. 5A). A careful inspection of the 1D-15N spectrum of the metal bound form shows that those peaks that increase upon EDTA addition were already present as very small peaks in the bound form. This is consistent with the fact that the metal analysis indicated only 95% metal incorporation. Interestingly, at 1 mM in solution, even a 4-fold molar excess of EDTA did not completely remove the metal bound resonances from the 15N-1D spectra. This may indicate either that the metal extraction is incomplete under the conditions used (5 days at 30°C) or that an equilibrium exists.

Overall, the 1D-15N spectrum (FIG. 5A) shows only 10 clear peaks of which two pairs (254/250 ppm; 162/160 ppm) correlate in the HMQC (FIG. 5B) to the same imidazole proton resonances and represent slightly different conformers of the same side chain with an exchange rate slower than 105 min-1 (Wuthrich, 1996).

Therefore, only 8 of the 14 nitrogen atoms expected for the 7 His residues are observed. The others undergo chemical exchange at a rate faster than 103 min-l, and thus are not observed (Wiithrich, 1996). Possible candidates for these fast exchanging residues are His 11, His 166 and His 167, which are all surface exposed in the NMR and crystal structures. The most strongly downfield shifted pair of resonances (254, 250 ppm) correspond to non-protonated type D imidazole nitrogens (Bachovchin, 1986).

These resonances are likely to belong to His 121 which is the only His residue that is buried in the hydrophobic core of the endostatin structure. The remaining histidines are residues 1,3, and 11 which were observed to complex the metal atom in the crystal structure (Ding et al., 1998). His ! 1 which is embedded in sheet 1.1 forms a strong linear hydrogen bond between its N and the carboxyl group of Glu 175 that is part of sheet 1.7 as was noted by Hohenester et al. (1998).

This N is also visible as a slow exchanging proton in an 2D-15N-HSQC spectrum

at a proton frequency of 11.2 ppm which confirms its hydrogen bonded state. This hydrogen bond could be an explanation for the stability of the metal binding site since it adds electron density to imidazole 0 electron system of His 11 to increase its affinity for metal binding. The exact nitrogen chemical shifts for His1 and His3 could not be assigned based on these data.

In a similar fashion using the GARANT/DYANA/violation analysis strategy starting with the metal-free endostatin structure the 2D and 3D spectra described above for the Zn-bound 13C/l5N endostatin were assigned and collection of conformational constraints were made and structures calculated. This proceeded very rapidly since reliable GARANT NOESY probabilities could be calculated from the core residues of the metal-free form. This was done over a 6 week period for the initial structures. Additional constraints were added to enforce the Zn ligation with His 1, His3, His 11 and Asp76 as determined by the 15N NMR spectroscopy described based on ab initio calculations (GAUSSIAN94, Frisch et al., 1995. A similar number of conformational constraints were obtained for both forms with similar goodness-of- fit to the restraints, though each structure determination relied on completely separate NMR data. Each structure was based on over 5000 peak assignments in the 3D NMR spectra. The incorporation of zinc into endostatin forming the zinc cluster has a dramatic effect on the both the N-and C-terminus as shown in FIG. 7A-7B. In contrast to the metal-free form, the Zn-bound endostatin has a fairly well defined N- terminus, looping back onto the globular fold of the protein. The C-terminus is better defined as well affording a more globular fold for all residues. The core residues of protein in both forms is very similar. This was also demonstrated by the observation of similar H/D exchange rates and that both forms of the protein have approximately 25% of the residues with very stable backbone amide hydrogen protons.

Another important implication of FIG. 5B is that for all three zinc complexing His residues the Na tautomers, and not the NC, are in complex with the metal ion which is in contrast with the crystal coordinates 1BNL where for residues 3 and 11 the NE atoms were defined. However, the solution data leave no doubt that the metal ion is facing the NC atom of all three histidines. The metal binding site was modeled by including the determined zinc-histidine constraints into the NMR structure calculations together with the fourth ligand Asp76 and presented in a stereo diagram (FIG. 6).

The inclusion of Asp76 was based on its proximity in the metal-free endostatin NMR structure and its identification as a metal binding ligand in the crystal

structure (Ding et al., 1998) although no direct evidence could be generated by NMR to confirm the Zn-Asp76 interaction due to the lack of NMR active nuclei involved.

FIG. 6 also shows the strong linear hydrogen bond of His 1 N [2H to the carboxyl group of Glu175 Two-step metal binding mechanisms of endostatin The HMQC spectrum of the EDTA titrated sample showed that the set of metal bound imidazole resonances (FIG. 5B) vanished while a new set of imidazole resonances appeared: ND resonances at 215 ppm, N at 179.5 ppm and C-H resonances at 7.96 ppm and 7.08 ppm. These resonances were very narrow and intense, indicating rapid conformational averaging distinct from the protein core. It is suggested that these resonances represent Hisl and His3, which are the only histidines located on the flexible N-terminal extension of the EDTA treated endostatin. The resonances of both are likely to be degenerate due to their flexibility and they are still in tight complex with zinc via their N 0 atoms, thus forming a new metal binding state of endostatin. Hisl and His3 hold the metal atom like tweezers, possibly forming a mixed complex with EDTA.

This zinc binding state is very similar to the presentation of immobilized metal ions such as used in His-tag protein purification resins. In the absence of EDTA, metal-free endostatin might utilize this new metal binding state to bind traces of zinc from solution in a first fast kinetic step followed by a second, time- limiting step for full metal binding site formation involving His1 l and Asp76. In fact it was found that 100 uM metal-free endostatin becomes fully supplemented with zinc when dialyzed against 5 pM ZnII solution. This model might explain how endostatin can activate itself in vivo. Once it is cleaved from collagen XVIII traces of zinc are sufficient to activate endostatin in the extracellular matrix where typical helper enzymes such as chaperones (Hausinger, 1996) that are often required for metallocenter assembly in vivo are not available.

EXAMPLE 7 Two-Dimensional 15N-1 H NMR Spectroscopy The conformation of the termini of metal-free and metal-bound forms of endostatin are different. This is strongly dictated by the ligating of zinc by

histidine residues 1,3 and 11 as well as aspartic acid 76. This ligation not only folds the N-terminus back onto the protein, but also stabilizes the conformation of the C- terminus packing it against the 40's residues. This is illustrated in FIGS. 3A-3B & 7A-7B. The conformational differences and flexibility were also observed in unrestrained, solvated molecular dynamics simulations of both metal-free and metal- bound systems using GROMOS96 over a 1.2 nanosecond range.

The prepared metal-free and metal-bound endostatin was characterized by two-dimensional (2D) 15N-1H NMR spectroscopy and direct detected 15N NMR spectroscopy. The 2D 15N-HSQC spectra maps the protons directly bonded to nitrogens in the protein affording both 1 H and 15N chemical shifts of the pairs. This spectra characterizes the"fingerprint"of the protein and is very sensitive to tertiary structure. This data is presented in a peak list of the spectra. (Table 2) Each line lists the 1 H and 15N chemical shifts of the bonded H-N pairs in the protein determined from peak positions in the 2D spectrum.

EXAMPLE 8 Zinc Ligation The metal ligation of the N-terminus was characterized by direct detected 15N spectra. The chemical shifts of the histidine imidazole ring nitrogen's are very characteristic depending on whether the nitrogen is protonated, metal-ligated or free. The 15N chemical shifts of the imidazole nitrogen's clearly show metal ligation for three residues. The use of l H-15N HMQC spectra characterized the ligation as through the delta nitrogen for each of the histidines at positions 1,3 and 11. Titration with the metal chelating agent EDTA clearly showed that the imidazole nitrogens ligating the zinc metal have strong chemical shift changes to that characteristic of protonated nitrogens.

EXAMPLE 9 Sedimentation of Soluble Endostatin Sedimentation equilibrium experiments were carried out with a Optima XL-I analytical ultracentrifuge (Beckman) in PBS buffer with either 1 FM

zinc sulfate or 1 mM EDTA at 4xC using standard 12 mm pathlength six-channel, charcoal-filled Epon cells with quartz windows. Continuous radial scanning at 280 nm was used at various speeds in 0.001 cm intervals and averaged over 10 runs.

Solvent densities of 1.00907 and 1.00917 for the zinc and EDTA samples, respectively, were calculated using the Sednterp program and the partial specific volume of 0.7305 was calculated from the weight average of the partial specific volumes of the individual amino acids. (Cohn & Edsall, 1943).

The quaternary structure of endostatin was measured using sedimentation equilibrium ultracentrifugation. The apparent molecular weights from 20 uM samples with and without I mM EDTA, as determined by a single species analysis where the Mr was treated as an adjustable parameter were 17,854 541 and 20,727403, respectively. Although both values differ which may be explained by non-idealistic behavior of endostatin in the presence of excess EDTA, these values are incompatible with dimer formation.

EXAMPLE 10 Rat Aortic Ring Assay Materials: EGM media: CLONETICS EGM bullet kit-Cat # CC-3125.

Reduced growth factor matrigel: BECTON DICKINSON/COLLABORATIVE BIOMEDICAL PRODUCTS cat# 40230C. Vascular endothelial cell growth factor (VegF): R&D SYSTEMS, Cat # 293-VE. Aortas isolated from rats with a weight range of 120-160 g.

Method: After isolation, the aorta is immediately placed in EGM media without hydrocortisone. The connective tissue is peeled back from the aorta under a dissecting microscope and each end is cut and discarded to avoid using any tissue injured with the hemostat during isolation. When possible the aortas and rings are kept on ice until being embedded.

The aortas are sliced into rings of approximately 1 mm width while bathed in media in a petri dish. Great care should be taken to discard the rings that contain small branching vessels-this eliminates an additional cut edge as a variable.

Rings from each aorta are rinsed 5 times in 5 ml each with EGM (CLONETICS)

media without hydrocortisone added. The rings are placed in individual wells of a 48 well plate that has been pre-coated with 125 uL of reduced growth factor (RDF) matrigel. The rings are randomized by mixing all aortic rings in one tube prior to placement in the 48 well plate. This helps to diminish the possibility of a single dose group being affected by rat to rat variation as well as possible distal and proximal differences. Alternatively, to diminish the possibility that a single dose group is affected by rat to rat variation, the rings from individual aortas are kept separated. A control group is included within each aorta and the dosed rings within that aorta are compared to the corresponding control group.

Pre-coating: RGF matrigel is thawed on ice and is kept on ice during the coating. The matrigel is added to each well with a pipette tip and the tip is drawn in a circular motion around the well to insure complete coating. The matrigel will solidify at room temperature after approximately 10 min. This time can be shortened by incubating the plates at 37°C.

After placing the rings in the wells, an additional 125 RL of RDF matrigel is added and the rings are stood up on their side, i. e., so that the passage through the interior cavity of each ring is oriented horizontally, not vertically. This is appropriate to help more accurately quantitate the tube outgrowth from each ring due to the difficulty of evaluating and documenting each plane of focus. well surface Correct incorrect Arrows indicate direction of tube outgrowth.

Once the matrigel becomes solid EGM media without hydrocortisone containing 100 ng/ml VegF and any other inducer or inhibitors is added at a total volume of 500 uL/well. Only the internal 24 wells of the plate are used and PBS is added to the outer wells to avoid evaporation problems in the sample wells. The plates are incubated at 37°C in a 5% C02 incubator and checked daily for growth.

Tube outgrowth typically starts after 3-4 days.

Photographs are taken of each well using EKTACHROME 64T slide film with pseudo dark field microscopy using a 4x objective. Outgrowth is quantitated visually on a 4 point grading scale according to the following: massive growth---less than massive growth---little growth---no growth.

Alternatively, images of the growth from the rings are captured with a 2x objective using IMAGE-PROPLUS software (MEDIA CYBERNETICS), saved as black and white TIF files and quantitated by measuring the square microns of tube outgrowth surrounding each ring. The software is calibrated to measure square microns with a micrometer viewed with the 2x objective.

EXAMPLE 11 Cell Proliferation assay Materials EGM media: CLONETICS EGM bullet kit-Cat #CC-3125. Endothelial cell (EC) trypsin/EDTA: CLONETICS, Cat# CC-5012. Vascular endothelial cell growth factor (VegF): R&D SYSTEMS, Cat # 293-VE. Basic fibroblast growth factor (bFGF): R&D SYSTEMS, Cat # 233-FB. Dulbecco's modified minimal essential medium (DMEM): LIFE TECHNOLOGIES, Cat# 11965-092. Fetal Bovine serum (FBS): Life Technologies, Cat&num 10082147. Human umbilical vein endothelial cells (HWEC): CLONETICS, Cat# CC-2519.: [Methyl-3H] Thymidine 20 Ci/mmol: DUPONT-NEN, Cat# NET027X. CellTiter 96 Aqueous nonradioactive cell proliferation assay reagent: PROMEGA, Cat# G5421.

Methods HUVECs are grown and maintained in EGM medium. Sub-confluent cells at no greater than the 7th passage are removed from their flasks with EC trypsin/EDTA, resuspended in DMEM, 10% FBS and centrifuged at 1000 rpm for 5 minutes at 40C. The supernatant fluid is discarded and the cell pellet is resuspended in an appropriate volume of DMEM, 10% FBS. The cell number is determined with a hemacytometer using trypan blue vital staining solution. After counting the appropriate dilution is made with DMEM, 10% FBS to achieve a cell concentration of 40,000 cells/ml. 100 ul of this suspension is added to each well of a 96 well plate to achieve a final concentration of 4,000 cells/well. Cells are growth-arrested for 24 hours at 370C in a humidified atmosphere containing 5% C02. The medium is replaced with 100 al fresh DMEM, 10% FBS containing either 1 ng/ml bFGF or 50 ng/ml VegF +/-endostatin or compounds to be evaluated. Control wells are dosed with the appropriate vehicle (e. g. 0.25% [v/v] DMSO). The plates are returned to the

incubator and after an additional 24 hour incubation each well is dosed with 10 go ouf an 80 uCi/ml solution of 3H-thymidine to yield a final concentration of 8 uCi/well.

After 48 hours at 370C, 5% C02 the media is removed and each well is washed twice with 350 gi of PBS containing 1 mg/ml bovine serum albumin. After the final rinse 100 u1 of 1.5 N NaOH is added to each well and incubated for 30 min at 370C. The cell lysates are transferred to 7 ml glass scintillation vials containing 150 ul of water.

Scintillation cocktail (5 ml) is added and the cell associated radioactivity is determined by liquid scintillation spectroscopy.

If the assay is to be quantitated using MTS then 10 al of MTS reagent is added on day 3 and incubated for one to four hours at 370C in a humidified atmosphere containing 5% CO2, After the appropriate incubation time the absorbance at 490 nm is measured.

EXAMPLE 12 Cell Migration Assay Materials Human endothelial cell line ECV304: ATCC Cat# CRL-1998.

Dulbecco's minimal essential media (DMEM): LIFE TECHNOLOGIES, Cat# 11965- 092. Fetal Bovine serum (FBS): LIFE TECHNOLOGIES, Cat# 10082147. Vascular endothelial cell growth factor (VegF): R&D SYSTEMS, Cat # 293-VE. Basic fibroblast growth factor (bFGF): R&D SYSTEMS, Cat # 233-FB. Trans-well, 24 well plates with 8 lm polycarbonate membranes: CORNING/COSTAR, Cat# 3422.

LEUKOSTAT Stain kit; FISHER SCIENTIFIC, Cat# CS430D.

Methods ECV304 cells are grown and maintained in Medium 199 supplemented with 10% FBS and 2 mM L-glutamine. Sub-confluent cells are removed from their flasks with EC trypsin/EDTA, resuspended in Medium 199,10% FBS and centrifuged at 1000 rpm for 5 minutes at 4°C. The supernatant fluid is discarded and the cell pellet is resuspended in an appropriate volume of Medium 199,10% FBS. The cell number is determined with a hemacytometer using trypan blue vital staining solution.

After counting the cells are diluted to 150,000 cells/ml with Medium 199 containing 0.5% FBS. The lower chamber is filled with Medium 199 containing 25 ng/ml bFGF.

The upper chamber is seeded with 7,500 cells/well +/-endostatin or compounds to be evaluated. Control wells are dosed with the appropriate vehicle (e. g. 0.25% [v/v] DMSO). Cells are allowed to migrate for 4 hours at 37°C. After this incubation the media is removed from the upper chamber and the cells on the upper chamber side of the membrane removed with a cotton swab and the cells on the under side of the membrane (migrated) are fixed and stained with the LEUKOSAT stain kit according to the manufacturers'directions. The number of stained nuclei on the under side of the membrane are quantitated using IMAGE PRO PLUS software after capturing the image with a digital camera and saving it as black and white TIF file.

EXAMPLE 13 Isolation of the Endostatin Receptor The soluble endostatin can be used to isolate the putative receptor of endostatin. A first approach utilizes endostatin protein covalently linked to a chromatography resin and, as an alternative, endostatin protein bound to a endostatin antibody-resin (an antibody raised to endostatin). These resins are used to capture the receptor from solubilized endothelial cell membranes. Various solubilization methods can be employed. Proteins that are found to be bound to endostatin columns may be eluted with salt gradients in the case of the covalently linked endostatin columns and low pH solution in the case of the antibody-linked endostatin column.

Any proteins that are eluted are evaluated by mass spectrometry and N- terminal sequencing as well as peptide mapping after proteolytic cleavage.

Alternatively, the soluble endostatin can be immobilized on plastic plates and used to"pan"for transiently transfected cells expressing the putative receptor. The source of the DNA for these transfections is an endothelial cell cDNA library. This method allows for the identification and cloning of the putative endostatin receptor Other methods of using the soluble endostatin of the present invention to isolate a putative endostatin receptor will be apparent to those of skill in the art and are meant to be within the scope of the present invention. For example, several methods of isolating a putative endostatin receptor are described in WO/97/15666 and can be performed using the endostatin of the present invention.

EXAMPLE 14 Assays for Anti-angiogenic Compound Those of skill in the art can use the assays of endostatin activity described above and those known or developed in the art to screen for anti-angiogenic compounds. The endostatin of the present invention is plentiful and inexpensive to prepare. Therefore, the endostatin of the present invention can be advantageously employed in such screening assays as a positive control against which to judge the activity of candidate anti-angiogenic compounds. Assays in which the endostatin of this invention is used as a control is contemplated to be within the scope of the present invention.

REFERENCES Bachovchin, W. W. (1986) Biochemistry 25, 7751-7759.

Bartels, C., Xia, T. H., Billeter, M. and Wuthrich, K. (1995) J. Biomol. NMR, 5,1-10.

Bartels, C., Billeter, M., Giintert, P. and Wuthrich, K. (1996) J. Biomol. NMR, 7,207- 213.

Bartels, C., Guntert, P., Billeter, M., and Wuthrich, K. (1997) J. Comp. Chem., 18, 139-149.

Billeter, M., Braun, W. and Wiithrich, K. (1989) J. Mol. Biol., 206,677-687.

Billeter, M., Schaumann, T., Braun, W. and Wuthrich, K. (1990) Biopolymers, 29, 695-706.

Bodenhausen, G. and Ruben, D. (1980) Chem. Phys. Lett., 69,185-188.

Boehm, T., Folkman, J., Browder, T. and O'Reilly, M. S. (1997) Nature (London) 390,404-407.

Boehm, T., O'Reilly, M. S., Keough, K., Shiloach, J., Shapiro, R. and Folkman J.

(1998) Biochem. Biophys. Res. Com. 252,190-194.

Boyd, J.; Soffe, N.; John, B.; Plant, D. and Hurd, J. (1992) J. Magn. Reson. 98,660.

Braun, W. (1987) Biopolymers, 26,1691-1704.

Cohn, E. J., Edsall, J. T., (1943) In: Proteins, amino acids and peptides as ions and dipolar ions. New York: Reinhold Publishing Corp. pp 370-377.

Delaglio, F., Grzesiek, S., Vuister, G. W., Zhu, G., Pfeifer, J. and Bax, A. (1995) J.

Biomol. NMR, 6,277-293.

Ding, Y.-H., Javaherian, K., Lo, K.-M., Chopra, R., Boehm, T., Lanciotti, J., Harris, B. A., Li, Y., Shapiro, R., Hohenester, E., Timpl, R., Folkman, J. and Wiley, D. C.

(1998) Proc. Natl. Acad. Sci. USA 95,10443-10448.

Fesik, S. W. and Zuiderweg, E. R. P. (1988) J. Magn. Reson., 78,588-593.

Gooley, P. R., Johnson, B. A., Marcy, A. I., Cuca, G. C., Salowe, S. P., Hagmann, W., K., Esser, C. K. and Springer, J. P. (1993) Biochemistry 32,130098-13108.

Grzesiek, S. and Bax, A. (1992a) J. Magn. Reson., 99,201-207.

Grzesiek, S. and Bax, A. (1992b) J. Am. Chem. Soc., 114,6291-6298.

Grzesiek, S. and Bax, A. (1992c) J. Magn. Reson., 96,432-438.

Grzesiek, S. and Bax, A. (1993) J. Biomol. NMR, 3,185-204.

Grzesiek, S., Anglister, J. and Bax A. (1993) J. Magn. Reson. Ser. B 101,114-119.

Guntert, P., Braun, W., Billeter, M. and Wuthrich, K. (1989) J. Am. Chem. Soc., 111, 3997-4004.

Giintert, P., Braun, W. and Wiithrich, K. (1991a) J. Mol. Biol. 217,517-530.

Giintert, P., Qian, Y. Q., Otting, G., Miiller, M., Gehring, W. J. and Wiithrich, K (1991b) J. Mol. Biol. 217,531-540.

Guntert, P., Mumentaler, X. and Wuthrich, K. (1997) J. Mol. Biol. 273,283-298.

Hanahan, D. and Folkman, J. (1996) Cell 86,353-364.

Hausinger, R. P. General Principles and mechanisms of metallocenter assembly.

Advances in Inorganic Biochemistry, pp 1-18 (1996).

Hohenester, E., Sasaki, T., Olsen, B. R. and Timpl, R. (1998) EMBO J. 17,1656- 1664.

Ikura, M., Kay, L. E., Tschudin, R. and Bax, A. (1990) J. Magn. Reson., 86,204-209.

Kay, L. E., Keifer, P. and Saarinen, T. (1992) J. Am. Chem. Soc., 114,10663-10665.

Kay, L. E. (1993) J. Am. Chem. Soc., 115,2055-2057.

Kerbel, R. S. (1997) Nature (London) 390,335-336.

Koradi, R., Billeter, M. and Wiithrich, K. (1996) J. Mol. Graph., 14,51-55.

Luginbuhl, P., Guntert, P., Billeter, M. and Wuthrich, K. (1997) J. Biomol. NMR 9, 212-222.

Marion, D., Ikura, M., Tschudin, R. and Bax, A. (1989) J. Magn. Reson., 85,393-399.

Mclachlan, A. D. (1979) J. Mol. Biol., 128,49-79.

Muhandiram, D. R. and Kay, L. E. (1994) J. Magn. Reson. Ser. B, 103,203-216.

O'Connell, J. F., Pryor, K. D., Grant, S. K. and Leiting, B. (1999) J. Biomol. NMR 13: 311-324.

O'Reilly, M. S., Boehm, T., Shing, Y., Fukai, N., Vasios, G., Lane, W. S., Flynn, E., Birkhead, J. R., Olsen B. R. and Folkman J. (1997) Cell 88,277-285.

Powers, R., Garrett, D. S., March, C. J., Frieden, E. A., Gronenborn, A. M. and Clore, G. M. (1992) Biochemistry, 31,4334-4346.

Pryor, K. D. and Leiting, B. (1997) Prot. Express. Purif. 10,309-319.

Sasaki, T., Fukai, N., Mann, K., Gohring, W., Olsen, B. R. and Timpl R. (1998) EMBO 17: 4249-4256.

Sattler, Schwendinger, Schleucher and Griesinger, C. (1995) J. Biomol. NMR, 6,11- 22.

Summers, M. F., Marzilli, L. G. and Bax, A. (1986) J. Ant. Chem. Soc. 108,4285- 4294.

Szyperski, T., Giintert, P., Otting, G. and Wiithrich, K. (1992) J. Magn. Reson., 99, 552-560.

Vuister, G. W. and Bax, A. (1992) J. Magn. Reson., 98,428-435.

Wishart, D. S., Bigam, C. G., Yao, J., Abildgaard, H. J. D., Markley, J. L. and Sykes, B. C. (1995) J. Biomol. NMR, 6,135-140.

Wuthrich, K. (1986) NMR of Proteins and Nucleic Acids, Wiley, New York.

Xia, T. H. Ph. D. Thesis ("Software for Determination and Visual Display of NMR Structures of Proteins: the Distance Geometry Program DGPLAY and the Computer Graphics Programs CONFOR and XAM", 1992, ETH-Zurich, Diss. ETH Nr. 9831, Prof. Kurt Wiithrich).

TABLE 1. Analysis of the 20 best DYANA conformers of endostatin after restrained energy minimization with the pr@<BR> avernge value # standard deviation (range)<BR> Quantity before energy minimization after energy minimization<BR> DYANA target function (Å2)a 8.04 # 0.87 (6.14 .. 9.38)<BR> OPAL physical energy (keal/mol) -1695 # 871 (2196 .. 1552) -4778 # 79 (-4933 .. @<BR> OPAL VDW energy (keal/mol) 1665 # 862 (1265 .. 493) -512 # 15 (-537 ..04@<BR> Residual nOe distance constraint violations<BR> Sum (Å) 26.5 # 1.8 (23.3 .. 29.9) 53.5 # 1.19 (51.4 .. 5.@<BR> Maximum (Å) 0.55 # 0.17 (0.31 .. 0.99) 0.16 # 0.006 (0.16 .. 0.@<BR> r.m.s.d. (Å) 0.04 # 0.004 (0.03 .. 0.05) 0.05 # 0.000 (0.05 .. 0.@<BR> Residual dihedral angle constraint violations<BR> Sum (°) 30.1 # 8.6 (16.4 .. 51.1) 59.3 # 4.3 (52.7 .. 6@<BR> Maximum (°) 4.7 # 2.9 (2.2 .. 12.5) 2.1 # 0.5 (2.0 .. 2.6@<BR> r.m.s.d. (°) 0.50 # 0.20 (0.28 .. 0.84) 0.56 # 0.14 (0.54 .. 0.@<BR> The final structure calculation was started with 100 randomized conformers. The 20 DYANA conformers with the lo@<BR> function values were refined by energy minimization and are used to represent the NMR structure.

Zn-bound Endostat to 10-150 @SQC 20 Peak @@at<BR> peak @@@ proton nitrogen<BR> 2 9.593 113.501<BR> 3 0.914 132.191<BR> 4 10.939 110.878<BR> 6 9.203 110.878<BR> 10 9.082 129.179<BR> 11 8.598 128.561<BR> 13 8.497 128.484<BR> 14 8.470 128.484<BR> 16 8.389 128.484<BR> 17 8.329 128.484<BR> 18 8.430 128.406<BR> 19 8.362 128.406<BR> 20 8.147 127.943<BR> 23 7.986 127.634<BR> 24 7.938 127.699<BR> 26 10.905 127.402<BR> 28 9.062 127.325<BR> 30 7.609 127.093<BR> 33 8.248 126.630<BR> 35 7.770 126.553<BR> 37 8.510 126.167<BR> 40 9.586 125.935<BR> 42 8.941 125.935<BR> 43 8.295 125.858<BR> 48 8.255 125.858<BR> 49 9.385 125.780<BR> 51 8.625 125.780<BR> 52 8.840 125.549<BR> 53 9.089 125.117<BR> 59 8.557 124.931<BR> 60 8.101 124.854<BR> 61 7.891 124.854<BR> 63 9.910 124.776<BR> 64 9.829 124.776<BR> 66 8.329 124.776<BR> 67 8.806 124.545<BR> 69 8.490 124.390<BR> 70 8.456 124.390<BR> 72 10.326 124.236<BR> 73 7.031 124.158<BR> 74 0.248 124.081<BR> 75 8.187 124.081<BR> 76 8.114 123.927<BR> 77 7.603 123.849<BR> 78 9.600 123.695<BR> 80 8.409 123.618<BR> 81 8.766 123.540<BR> 82 7.824 123.109<BR> 83 10.521 123.232<BR> 84 8.120 123.077<BR> 87 9.015 122.845<BR> 89 7.347 122.227<BR> 91 8.160 122.150<BR> 92 8.113 122.073<BR> 94 8.83@ 121.919<BR> 95 8.793 121.919<BR> 98 7.528 121.223<BR> 99 9.970 121.146<BR> 100 8.699 121.069<BR> 102 7.367 120.992<BR> 103 9.310 120.849<BR> 104 11.234 120.837<BR> 108 8.140 120.760<BR> 109 7.925 120.683<BR> 110 8.853 120.606<BR> 111 9.412 120.528<BR> 112 8.093 120.528<BR> 117 8.981 120.297<BR> 118 7.636 120.219<BR> 119 7.569 120.219<BR> 120 7.178 120.219<BR> 121 7.743 120.142<BR> 123 7.656 120.065<BR> 124 8.423 119.833<BR> 125 7.817 119.756<BR> 126 0.692 119.679<BR> 127 0.147 119.679<BR> 128 8.073 119.679<BR> 130 6.002 119.679<BR> 132 0.510 139.524<BR> 133 8.315 119.524<BR> 134 8.261 119.524<BR> 135 8.618 119.447<BR> 137 8.187 119.447<BR> 138 8.947 119.215<BR> 139 7.636 119.215<BR> 140 7.447 119.215<BR> 141 9.055 119.138<BR> 142 8.376 119.138<BR> 144 8.234 119.138<BR> 146 9.519 118.906<BR> 147 8.255 118.675<BR> 148 8.362 118.597<BR> 149 7.038 118.366<BR> 150 8.551 118.280<BR> 152 8.160 118.280<BR> 153 8.053 118.280<BR> 159 8.631 118.134<BR> 160 8.651 118.057<BR> 161 7.864 117.979<BR> 163 8.181 117.902<BR> 165 8.974 117.671<BR> 167 8.094 117.671<BR> 171 7.743 117.593<BR> 172 9.351 117.516<BR> 174 5.625 117.516<BR> 176 8.766 117.362<BR> 177 10.064 117.110<BR> 180 6.223 117.053<BR> 182 7.824 116.821<BR> 183 9.930 116.744<BR> 184 8.060 116.744<BR> 185 7.636 116.744<BR> 186 7.340 116.744<BR> 190 8.423 116.589<BR> 191 11.275 116.350<BR> 192 7.004 116.280<BR> 193 7.414 115.894<BR> 194 7.811 115.817<BR> 196 9.048 115.740<BR> 197 8.995 115.740<BR> 198 8.329 115.505<BR> 199 7.582 115.505<BR> 200 7.153 115.431<BR> 202 7.004 115.431<BR> 203 7.757 115.353<BR> 204 8.450 115.199<BR> 205 8.200 115.199<BR> 206 7.656 115.199<BR> 208 7.595 115.122<BR> 209 7.360 115.122<BR> 211 7.542 114.967<BR> 212 8.692 114.090<BR> 213 7.447 114.013<BR> 214 9.465 114.736<BR> 217 7.616 114.195<BR> 218 8.847 113.731<BR> 220 0.551 113.500<BR> 221 7.898 113.268<BR> 222 7.084 113.268<BR> 223 7.017 113.268<BR> 224 7.252 113.114<BR> 225 9.270 113.036<BR> 227 8.752 112.805<BR> 228 8.051 112.805<BR> 229 7.898 112.727<BR> 234 8.725 112.496<BR> 236 7.717 112.187<BR> 239 8.598 111.878<BR> 240 6.929 111.646<BR> 241 6.056 111.646<BR> 244 8.140 111.569<BR> 245 7.508 111.569<BR> 247 6.627 111.569<BR> 249 7.649 111.492<BR> 251 7.199 111.492<BR> 252 8.766 111.@@@<BR> 253 6.862 111.183<BR> 254 6.63@ 111.105<BR> 255 7.500 111.028<BR> 257 7.444 109.292<BR> 258 9.519 109.618<BR> 259 8.443 109.175<BR> 261 7.414 108.943<BR> 262 6.701 108.943<BR> 263 7.474 108.711<BR> 264 8.356 108.634<BR> 265 0.759 108.479<BR> 267 7.347 108.479<BR> 268 6.815 108.479<BR> 269 6.694 108.402<BR> 270 7.266 108.248<BR> 271 7.347 108.016<BR> 232 8.080 107 784<BR> 273 8.369 107.707<BR> 274 7.313 107.321<BR> 276 6.896 107.166<BR> 277 7.205 106.317<BR> 278 6.593 106.317<BR> 279 7.125 106.162<BR> 280 6.191 106.162<BR> 281 6.929 105.776<BR> 282 6.560 105.699<BR> 283 8.194 105.544<BR> 285 7.641 104.927<BR> 287 7.548 104.@18<BR> 288 6.950 104.309<BR> 289 6.909 104.309<BR> 292 7.945 103.614<BR> 293 7.501 102.609