Contributions of disulfide bonds in a nested pattern to the structure, stability, and biological functions of endostatin.

Endostatin can inhibit the proliferation and migration of endothelial cells. It contains two pairs of disulfide bonds in a nested pattern. We constructed three mutants, C33A/C173A, C135A/C165A, and all-Ala, to evaluate the contributions of individual disulfide bonds to the structure, stability, and biological functions of endostatin. Both tryptophan emission fluorescence spectrum and 1H nuclear magnetic resonance spectrum show that C135A/C165A and all-Ala, the two mutants lacking disulfide bond Cys135-Cys165, lost nearly their entire tertiary structure. Although C33A/C173A appears to retain some native-like structures, it is less stable and has a higher helical content, which confirms our earlier hypothesis that the polypeptide backbone of endostatin has a high helical propensity. C135A/C165A and all-Ala mutants lost most of their inhibitory activities both on the migration and proliferation of human microvascular endothelial cells, whereas C33A/C173A is partially active. The mutants without disulfide bond Cys135-Cys165 can hardly be internalized and localized to cytoskeleton and nucleus in the cell, which probably contributes to their loss of inhibition on the migration and proliferation of endothelial cells. Our studies provide a structural basis for the two disulfide bonds on the biological functions of endostatin.

Endostatin can inhibit the proliferation and migration of endothelial cells. It contains two pairs of disulfide bonds in a nested pattern. We constructed three mutants, C33A/C173A, C135A/C165A, and all-Ala, to evaluate the contributions of individual disulfide bonds to the structure, stability, and biological functions of endostatin. Both tryptophan emission fluorescence spectrum and 1 H nuclear magnetic resonance spectrum show that C135A/C165A and all-Ala, the two mutants lacking disulfide bond Cys 135 -Cys 165 , lost nearly their entire tertiary structure. Although C33A/C173A appears to retain some native-like structures, it is less stable and has a higher helical content, which confirms our earlier hypothesis that the polypeptide backbone of endostatin has a high helical propensity. C135A/C165A and all-Ala mutants lost most of their inhibitory activities both on the migration and proliferation of human microvascular endothelial cells, whereas C33A/C173A is partially active. The mutants without disulfide bond Cys 135 -Cys 165 can hardly be internalized and localized to cytoskeleton and nucleus in the cell, which probably contributes to their loss of inhibition on the migration and proliferation of endothelial cells. Our studies provide a structural basis for the two disulfide bonds on the biological functions of endostatin.
Endostatin (ES) 1 is an angiogenesis inhibitor that prevents vascular endothelial cells from proliferating and migrating in response to a spectrum of proangiogenic proteins (1) and can potently inhibit tumor growth without inducing toxicity and acquired drug resistance (2)(3)(4). The crystal structure of ES reveals a compact fold containing predominantly ␤-sheets and loops as well as two ␣-helices (5,6).
ES is a globular protein with two pairs of disulfide bonds in a nested pattern, Cys 33 -Cys 173 and Cys 135 -Cys 165 (5). The former disulfide bond connects helix ␣ 1 (longer ␣-helix) to the central ␤-sheet and the latter circularizes a twisted loop containing three strands (5). Our group recently reported that ES is acid-resistant with slow kinetics upon acid-induced unfolding (7). The disulfide bonds are very difficult to access in native ES, because they cannot be completely reduced in the absence of denaturants (7). These properties are proposed to be attributed to the existence of a nested pattern of disulfide bonds (7).
Disulfide bonds are one of the significant interactions for protein native conformations, stabilities, and activities (8 -12).
In the absence of one or all of the disulfide bonds, many proteins cannot sustain such properties (11,13). Each disulfide bond has different roles in the stability and activity of proteins (13)(14)(15)(16)(17). In order to elucidate the functions of the two disulfide bonds in ES, each of the two individual disulfide bonds was mutated by substituting cysteine residues with alanine, which is a conservative change (18). Alanine residues are widely found in the secondary structures of proteins such as ␣-helix, ␤-sheet, and turns (19,20). The mutations to Ala do not make any specific side chain interactions to affect protein stabilities (19,20). Serine is structurally similar to cysteine except that it contains a hydroxyl group instead of a thiol group (21) and can also be used to replace cysteine residues of proteins (22)(23)(24). However, serine may not be suitable for replacing cysteine residues in ES because 1) serine tends to disrupt ␣-helices (20,(25)(26)(27)(28)(29)(30)(31) and 2) the polypeptide backbone of ES appears to have a high helical propensity as reported by our group (7). A total of three mutants were constructed: C33A/C173A, C135A/C165A, and all-Ala (Fig. 1). These mutants were subjected to conformational analyses by Trp fluorescence emission, CD, 1 H NMR, stability measurements with denaturants, and biological activity evaluation on human microvascular endothelial cells (HMECs). C33A/C173A was found to retain some of the native structures, stabilities, and biological functions. In contrast, mutants C135A/C165A and all-Ala lost almost all of their tertiary structures and biological functions. When the disulfide bond Cys 33 -Cys 173 is mutated, both C33A/C173A and all-Ala gain much increased helical contents compared with wild type ES. This observation confirms our earlier hypothesis that the polypeptide backbone of ES may have an intrinsic high ␣-helical propensity (7). Without any disulfide bonds, all-Ala has the least native conformation and the lowest stability and activity but the highest helical propensity.
Dixelius et al. reported that ES can be internalized by endothelial cells (32). Moreover, ES is believed to be involved in the interaction of cytoskeleton components (32)(33)(34)(35)(36)(37), which in turn regulate the organization of cytoskeleton (32,38). However, the interaction between cytoskeleton proteins and ES has not been elucidated at the molecular level. Our results show that lacking disulfide bonds impairs the internalization and localization of ES to cytoskeleton of endothelial cells, which results in the decrease of inhibitory activities of ES both on the migration and proliferation of HMECs. These observations may help to evaluate the contributions of disulfide bonds on the structures of ES and the correlation between structures and the biological functions of ES.

EXPERIMENTAL PROCEDURES
Construction of Mutants and Protein Expression-Constructs for ES mutants were prepared by site-directed mutagenesis using a wild type ES construct in pET25b (Novagen) as the template (Fig. 1). DNA sequencing was performed to confirm the correct introduction of mutations. The recombinant wild type ES and its mutants have MGGSHH-HHH at their N termini.
Protein Purification-Plasmids encoding human wild type ES or its mutants were transformed into Escherichia coli BL21 (DE3) strain. Wild type ES and its mutants were expressed as inclusion bodies in E. coli and purified via an Ni 2ϩ -nitrilotriacetic acid column (Qiagen). Wild type ES and its mutants were refolded and purified by Protgen. Protein concentrations were determined according to Edelhoch's method (39).
Measurements of Tryptophan Emission Fluorescence-Trp emission fluorescence of ES was measured with a Hitachi F-4500 spectrophotometer using cuvettes with an optical path length of 1 cm. The temperature was maintained constantly at 20°C using an external bath circulator. Trp emission fluorescence was measured at an excitation wavelength of 288 nm. The excitation wavelength at 288 nm was chosen because ES contains four Trp residues, and excitation at 288 nm ensures fluorescence emission mainly contributed by Trp residues. The emission spectra were measured from 300 nm to 400 nm (excitation and emission slit width ϭ 5 nm). The scan speed was 240 nm/min. The concentration of wild type or mutants was 0.9 M in 5 mM Tris-HCl buffer (pH 7.4).
Circular Dichroism Measurements-Far-UV CD was recorded at 20°C on a Jasco J-715 spectropolarimeter equipped with a temperature-controlled liquid system. A stock of ES was diluted in 5 mM Tris-HCl (pH 7.4) by aliquots to a final protein concentration of 5 M. Cuvettes of 2-mm path length were used over the wavelength range between 200 and 250 nm. Spectral acquisition was taken at 0.2-nm intervals with 4-s integration time and a bandwidth of 1.0 nm. An average of four scans was obtained for all of the spectra. Photomultiplier absorbance did not exceed 600 V in the spectral region analyzed. Data were corrected for buffer contributions and smoothed using the software provided by the manufacturer. All of the measurements were performed under nitrogen flow. The results are expressed as mean residue ellipticity, [], in units of degrees cm 2 dmol Ϫ1 . A mean residue weight of 110 was used for the peptide chromophore.
Hydrogen Exchange and 1 H NMR Measurements-The concentration of wild type ES or its mutants was adjusted to 200 M in 10 mM acetate buffer (pH 6.0) containing 16% D 2 O. The samples were then incubated at 20°C for 100 min before NMR measurement. 1 H NMR spectroscopy was taken at 500 MHz on a Varian Inova 500NB instrument. NMR spectra were acquired using a spectral width of 8000.0 Hz, a relaxation delay of 1.0 s, and 128 repetitions.
pH Titration-Wild type ES or its mutants were first dissolved in 30 mM acetate (pH 5.5) to make a stock solution, which was then diluted by aliquots into 4 mM acetate buffer at different pH values. 1 M HCl was used to adjust pH when pH values were lower than 3.5. The final protein concentration was 0.9 M. Trp emission fluorescence measurements were carried out immediately after the samples were incubated at 20°C for 100 min. However, the incubation time of mutants was reduced to 30 min because of their faster equilibration process.
Urea-induced Unfolding-A stock solution of ES was diluted by aliquots into different concentrations of urea to reach a final protein concentration of 0.9 M. The buffer used was 5 mM Tris-HCl at pH 7.4.
After incubation at 20°C for 120 min (wild type) or 30 min (mutants), Trp emission fluorescence measurements were carried out.
Data Analysis-Unfolding curves were analyzed by a two-state equation with linear base lines, according to the procedure of Santoro and Bolen that uses data inside as well as outside the transition zone to fix the base lines (40). Both C135A/C165A and all-Ala cannot be normalized to two-state curves because of the marginal stability and the difficulties in finding their base lines.
Proteolysis Assay-Wild type ES or its mutants were diluted in PBS to 14 M. Trypsin (0.25 units/ml) was added and incubated with ES at 0 or 37°C. The reactions were terminated at different time intervals by adding Laemmli SDS-polyacrylamide gel electrophoresis loading buffer. The digested wild type ES or its mutants were then analyzed on 12% SDS gels followed by Coomassie Brilliant Blue staining.
Cell Migration-The motility of HMECs was measured by using a modified Boyden chamber (Neuro Probe, Inc.) assay. Briefly, to inhibit the migration of HMECs, lower chambers were filled with DMEM supplemented with 0.2% FBS containing wild type ES or its mutants at concentrations ranging from 4 mg/ml to 0.4 ng/ml or saline as a control. HMECs (1 ϫ 10 4 ) were added to the upper chambers in DMEM, supplemented with 0.2% FBS and 10 ng/ml bFGF. The cells were incubated at 37°C, 5% CO 2 for 4 h, after which the number of cells migrated to the lower surface was counted. The migration results were expressed as the average number of cells per high magnification optical microscopic (ϫ400) field. Five fields were counted for each assay. Each sample was assayed in tetrad, and the assays were repeated twice.
Cell Proliferation-Wild type ES or its mutants were assayed for their inhibitory activities on the proliferation of HMECs stimulated by bFGF (5 ng/ml) with the Colorimetric Assay for Cell Survival and Proliferation Kit (Chemicon International) according to the manufacturer's instructions. Briefly, HMECs were inoculated into DMEM supplemented with 10% FBS and incubated for 24 h at 37°C, 5% CO 2 . HMECs grown to 70% confluence were serum-deprived in DMEM for 12 h and then treated for 48 h with 0.6 -60 g/ml wild type ES or its mutants supplemented with 2% FBS and 5 ng/ml bFGF at 37°C, 5% CO 2 . 3-(4,5-Dimethylthiazol-2-yl)2,5-diphenyl tetrazolium bromide (50 g/ml) was added into the culture and mixed by tapping gently on the side of the tray. After incubating at 37°C for 4 h, 100 l of color development solution (isopropyl alcohol with 0.04 M HCl) was added into each well and mixed thoroughly by repeated pipetting with a multichannel pipettor. The cells were incubated at 37°C for 30 min, and the absorbance at 570 nm was measured on an ELISA plate reader with a reference wavelength of 630 nm. Each sample was assayed in hexad, and the assays were repeated twice.
Statistical Analysis-The results are expressed as mean values Ϯ S.E. Multiple comparisons were performed by one-way analysis of variance, and differences with p Ͻ 0.05 were considered significant.
Immunofluorescence Analysis-HMECs were plated into 24-well plates in DMEM supplemented with 10% fetal bovine serum and cultured for 24 h. After they were serum-deprived in DMEM for 12 h, HMECs were supplemented with 2% FBS and 5 ng/ml bFGF. The cells were then treated with 10 g/ml wild type ES or its mutants for 8 h. The treated cells were washed with ice-cold PBS, fixed with 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, and finally incubated with 1:200 rabbit anti-ES polyclonal antibodies in PBS, 0.2% Tween 20, 1% bovine serum albumin followed by incubation with TRITC-conjugated goat anti-rabbit IgG. The fields were photographed under the TRITC wavelength with an Olympus fluorescence microscope.
Internalization Assays of Labeled Endostatin-Wild type ES and its mutants were biotinylated with sulfo-NHS-LC-Biotin by the EZ-Link TM sulfo-NHS-LC biotinylation kit (Pierce) in PBS. HMECs or ECV304 cells were plated on 24-well plates and cultured as described in the immunofluorescence assay. After 8 h of serum deprivation, the cells were incubated in DMEM medium, 2% FBS, and 5 ng/ml bFGF containing 10 g/ml biotin-labeled wild type ES or its mutants (as described in the legend to Fig. 9). At different time intervals, the treated cells were washed with ice-cold PBS, fixed with 4% paraformaldehyde for 30 min, and then permeabilized with 0.2% Triton X-100 on ice. After they were blocked with 10% normal goat serum in PBS for 1 h, the cells were incubated with 1:300 TRITC-conjugated streptavidin (Pierce) in PBS, 0.2% Tween 20, 1% bovine serum albumin for 30 min. The fields were photographed under the TRITC wavelength with an Olympus fluorescence microscope.

Effect of Disulfide Bonds on the Structure of Endostatin-
The tryptophan fluorescence emission was used to monitor the conformational changes of wild type ES or its mutants. The maximal Trp fluorescence emission wavelength of wild type ES is at 320 nm, whereas that of C33A/C173A and C135A/C165A red shift to 320 and 335 nm, respectively ( Fig. 2A). Not surprisingly, the maximal emission wavelength of all-Ala is even at about 340 nm, which is close to a completely denatured state ( Fig. 2A).
In order to have a more precise measurement of conformational changes, NMR was employed. Since all-Ala is not very stable at neutral pH, the 1 H NMR spectra of wild type and its mutants were therefore taken at pH 6.0, at which the wild type ES still retains its native conformation (7), whereas the mutants appear to have the same conformation as at neutral pH with low protein concentrations (see Fig. 3E). In the NMR spectra, the up-field chemical shifts of methyl peaks, the sharp chemical shift dispersions in the aromatic region, and spreading out of the histidine 1 H ⑀1 resonances are all characteristic properties of tightly packed native proteins. These characters are present not only in the NMR spectrum of wild type ES but also in that of C33A/C173A, suggesting that this mutant retains some native-like conformations. Although C135A/C165A also retains some native structures, the chemical shift dispersion at the aromatic region becomes broader, and the number of peaks below 0 ppm is reduced significantly. No tertiary structures were detected in all-Ala based on its NMR spectrum (Fig. 2C).
The secondary structures of ES contain a large fraction of loop structures and ␤-sheets and only a small fraction of ␣-helices (5). The helical contents of mutants C33A/C173A and C135A/C165A are about 2-and 1.5-fold, respectively, compared with that of the wild type based on far-UV CD measurements at 222 nm. All-Ala has the highest helical content among all three mutants, greater than 2-fold wild type (Fig. 2B).
pH Titration of Endostatin Wild Type and Its Mutants-The pH titration behaviors of ES mutants vary according to the locations of the disulfide bonds (Table I). The Trp fluorescence emission spectra of acid-induced unfolding of wild type ES and its mutants are red-shifted, and their fluorescence intensity decreases significantly upon unfolding (Fig. 3, A-D). As monitored by Trp fluorescence emission, wild type ES begins unfolding at about pH 3.5; on the other hand, mutants C135A/C165A and C33A/C173A cannot retain their native tertiary structures even at pH 4.0 and 5.0, respectively (Fig. 3E). These two mutants have different slopes in the unfolding transition regions. The transition region of C135A/C165A is between pH 4.0 and 2.5, whereas that of C33A/C173A is between pH 5.0 and 2.5.
Stability Measurements of Endostatin Wild Type and Its Mutants-The stabilities of ES mutants vary with different disulfide bonds as expected. Urea-induced unfolding of wild type ES and its mutants was monitored by measuring the intrinsic Trp fluorescence emission. The Trp fluorescence emission spectra are red-shifted, and their fluorescence intensity increases significantly upon unfolding (Fig. 4, A-D). The unfolding curves show that unfolding of wild type and C33A/ C173A, but not the other two mutants, can be described as a two-state process (Fig. 4, E and F (Table I).

Proteolysis Stability of Endostatin Wild Type and Its
Mutants-Endostatin is relatively stable against trypsin. It requires up to 60 min to be digested completely to an 18-kDa fragment at 37°C; this fragment appears to be trypsin-resistant (Fig. 5). The proteolysis stabilities of mutants vary. C33A/ C173A is almost as resistant to trypsin as wild type, and it can also be digested to the 18-kDa fragment. Mutant C135A/C165A was completely digested by trypsin (37°C, 10 min) to smearing fragments based on SDS-PAGE analysis. Surprisingly, all-Ala, the unstable protein against denaturants, is more resistant to trypsin digestion than wild type ES (Fig. 5).

Effects of Disulfide Bonds on the Inhibitory Activity of Endostatin in the Migration of HMECs-Wild type ES inhibits
angiogenesis stimulated by vascular endothelial growth factor or bFGF (2, 4, 35-37, 41, 42). To determine the roles of disulfide bonds in the anti-angiogenesis activities of ES, the migration assay was applied for wild type ES and its mutants. Wild type ES can significantly inhibit the migration of bFGF-induced HMECs in a dose-dependent manner. The inhibitory activities of mutants vary significantly (Fig. 6). At a lower concentration (4 ng/ml), wild type ES inhibits the migration of HMECs by 15%, and the half-maximal inhibitory concentration (IC 50  C165A, and all-Ala are 0.68, 1.75, and 3.25 g/ml, respectively. C33A/C173A has about one-half the inhibitory activity of wild type ES, whereas C135A/C165A has 17% and all-Ala has only about 9% of that of the wild type.

Effects of Disulfide Bonds on the Inhibitory Activity of Endostatin in the Proliferation of HMECs-The proliferation of
HMECs stimulated by 5 ng/ml bFGF can be inhibited significantly by wild type ES. The activities of ES mutants are reduced to different extents (Fig. 7, A-E). Wild type ES inhibits the proliferation of HMECs by 13% at 6 g/ml and 25% at 60 g/ml (Fig. 7A). As expected, the inhibitory activities of C33A/ C173A and C135A/C165A measured in the proliferation assay correlate with that in the migration assay (Fig. 7, B and C). At 60 g/ml, C33A/C173A has 15% and C135A/C165A has 12% of the proliferation inhibitory activity of wild type ES. At a lower concentration of 0.6 g/ml, C135A/C165A shows almost no proliferation inhibitory activity, whereas C33A/C173A retains some activity. All-Ala is completely inactive even at 60 g/ml (Fig. 7D). The inhibitory activities of wild type ES and its mutants are compared in Fig. 7E.
Effects of Disulfide Bonds on the Internalization of Endostatin-Dixelius et al. (32) reported that ES can be internalized by murine brain endothelial cells. The immunofluorescence assay with rabbit anti-ES polyclonal antibodies was applied to detect internalized wild type ES and its mutants in endothelial cells.
The internalized ES was observed as red clusters in HMECs under the fluorescence microscopy. The amount of internalized C33A/C173A was less than that of the wild type, whereas few clusters formed in mutants C135A/C165A and all-Ala (Fig. 8).
In order to detect the patterns of internalization and localization of internalized wild type and mutant proteins, ES was biotinylated with the EZ-Link TM sulfo-NHS-LC biotinylation kit (Pierce). The localization and quantity of internalized ES were observed by TRITC-conjugated streptavidin that binds to the biotin-labeled ES. HMECs internalized a mass of wild type ES and C33A/C173A after 3 h (Fig. 9A) but could barely internalize C135A/C165A and all-Ala even after 5 h. After entering cells, most of wild type ES was enriched into nuclei. C33A/ C173A, on the other hand, was not found in nuclei after it was internalized into cytoplasm.
The internalization assay was also performed in ECV304 (human vascular endothelial cell), another endothelial cell line. In this case, three groups of ES were observed; most of ES was concentrated into nuclei, some localized in cytoplasm, which was then distributed onto cytoskeleton, and others formed clusters on the surface of cells. After 8 h, most of the wild type ES was found in nuclei and on cytoskeleton (Fig. 9B). The nativelike mutant, C33A/C173A, was internalized after 8 h but only formed a few dots in the cytoplasm of cells. The other two mutants, however, could not be internalized, and were not found in nuclei or on the cytoskeleton. DISCUSSION Endostatin, a 20-kDa C-terminal fragment digested from collagen XVIII, is a specific inhibitor of endothelial cell migration, proliferation, and angiogenesis. Although the crystal structure of ES has been resolved (5,43), it is unknown which structural components are critical for the biological functions of this protein. Based on the crystal structure, ES contains two pairs of disulfide bonds in a nested pattern, which link the central core and the peripheral structures (5,6). Therefore, the two pairs of disulfide bonds can be critical for the biological functions, stabilities, and structures of ES. In the present study, a His tag was attached to the N-termini of both wild type ES and its mutants. This is a fairly mild modification widely used to facilitate the expression and purification of proteins (44 -49). It has been well documented that ES attached with a His tag perfectly retains its anti-angiogenesis activities, which indicates that the His tag does not interfere with the behavior of ES (2,34,50,51).
Effects of Disulfide Bonds on the Biological Functions of Endostatin-Wild type ES has significant inhibitory activities on the migration and proliferation of endothelial cells when stimulated by bFGF. However, the inhibitory activities of wild type ES and its mutants vary significantly. C33A/C173A has more inhibitory activities than the other two mutants. The reason can be that the poor tertiary structures of both C135A/ C165A and all-Ala impair their inhibitory activities. Our studies show that C33A/C173A retains half of the inhibitory activities of wild type ES both on the migration and proliferation of HMECs (Figs. 6 and 7). The domain containing the disulfide bond Cys 135 -Cys 165 may be responsible for the anti-angiogenesis functions of ES. Morbidelli et al. (43) reported that fragment IV ox (containing the disulfide bond Cys 135 -Cys 165 ) from mouse ES exhibits inhibitory activities on the migration and proliferation of endothelial cells. Full-length ES appears to be 4 -5 times more potent than fragment IV ox on the inhibitory activities. However, fragment IV (the disulfide bond Cys 135 -Cys 165 is reduced) showed inhibitory activities on the migration and proliferation of endothelial cells only at very high concentrations (43). In contrast with the ES fragment IV ox , C33A/ C173A has 2-fold more inhibitory activities on the migration FIG. 6. Migration assays of wild type (WT) endostatin and its mutants. ES inhibits the migration of HMECs in a dose-dependent manner. HMECs were allowed to migrate for 4 h in DMEM at 37°C and 0.2% FBS, 5% CO 2 with 10 ng/ml bFGF. Gradient concentrations of wild type ES and its mutants were added, respectively, at the beginning of the test. After incubation, the number of migrated cells was counted in five different areas, and the average number of migrated cells was obtained. The same volume of physiological saline was added in the control. IC 50 means the concentration of ES at 50% inhibitory activity. Each sample was assayed in tetrad, and the assays were repeated twice. p Ͻ 0.05. and proliferation of endothelial cells. Higher inhibitory activities of C33A/C173A can recur to other sequences besides fragment IV ox . In contrast with C33A/C173A, C135A/C165A has inhibitory activities on the migration and proliferation of HMECs only at high concentrations, and all-Ala has no inhibitory activities (Figs. 6 and 7). Therefore, the inhibitory activities of ES on the migration and proliferation of endothelial cells may be closely correlated with the structures retained by disulfide bonds, especially the disulfide bond Cys 135 -Cys 165 .

Effects of Disulfide Bonds on the Localization of Endostatin
in Endothelial Cells-Wild type ES and C33A/C173A can be internalized by HMECs, but the other two mutants cannot, which suggests that the conformational changes of mutants interrupt the internalization (Figs. 8 and 9A). The same assays in another type of endothelial cells, ECV304, show that the internalized wild type ES can be located on cytoskeleton and transported into nuclei of ECV304 cells after 8 h (Fig. 9B). For C33A/C173A, however, both of the two processes were not observed. Furthermore, the amount of C33A/ FIG. 7. Proliferation assays of wild type (WT) endostatin and its mutants. ES inhibits the proliferation of HMECs in a dose-dependent manner. HMECs were inoculated in a 96-well plate (1 ϫ 10 3 cells/well) and allowed to proliferate for 48 h in DMEM supplemented with 2% FBS as well as 5 ng/ml bFGF, 5% CO 2 at 37°C. The different concentrations of wild type ES and its mutants were added, respectively, at the beginning of the test. After 48 h, viable HMECs were assessed by the 3-(4,5-dimethylthiazol-2-yl)2,5-diphenyl tetrazolium bromide assay as described under "Experimental Procedures." The same volume of medium was added in the control. A, wild type; B, C33A/C173A; C, C135A/C165A; D, all-Ala; E, inhibitory activity of wild type ES and its mutants. Each sample was assayed in hexad, and the assays were repeated twice. p Ͻ 0.05.

FIG. 8. Immunofluorescence assays of wild type (WT) endostatin and its mutants in HMECs.
HMECs were inoculated into a 12-well plate (1 ϫ 10 3 cells/well). HMECs were incubated in the presence of 10 g/ml wild type ES or its mutants for 8 h. Fixed, permeabilized, and blocked cells were incubated in PBS containing rabbit anti-ES polyclonal antibodies (1:200) for 2 h on ice and followed by incubation with TRITCconjugated goat anti-rabbit IgG for 1 h on ice. The fields were photographed under the TRITC (red) excitation wavelength at ϫ400 magnification.
C173A in cytoplasm was also lower than that of wild type ES. The other two mutants almost cannot bind to the cell surface and be internalized. Only wild type ES was observed to be transported into nuclei; the mechanism is still unknown (Fig.  9, A and B).
ES was reported to bind tropomyosin to disrupt the microfilament integrity (33). Focal adhesions and actin stress fibers decrease and the actin stress fiber network dissociates in response to ES treatment in endothelial cells (34,35). In ECV304 cells, the internalized wild type ES can be localized onto cytoskeleton but not all of the mutants (Fig. 9B), which might be one of the reasons that causes significant decrease of inhibitory activities of the mutants on the migration and proliferation of endothelial cells.
The changes of structures of the mutants can lead to the decrease of binding affinity between mutants and target proteins on the cell surface. Rehn et al. (52) reported that the significance of the integrin binding function of ES might also be to target and concentrate ES at the sites of neovascularization to function as a highly efficient angiogenesis inhibitor. Although wild type ES can bind to the surface of ECV304 cells, the mutants cannot bind to the cell surface efficiently (Fig. 9B). If ES mutants cannot be identified and concentrated by some factors on the surface of endothelial cells, their biological functions should be impaired.
The degradation of ES was reported in murine brain endothelial cells (32). We also found that both C135A/C165A and all-Ala were degraded in ECV304 cells (Fig. 9B). Interestingly, all-Ala appears more stable than C135A/C165A, which is consistent with the observation that all-Ala displays higher stability against trypsin digestion than C135A/C165A (Fig. 5). We speculate that the cleavage sites by trypsin are not easily accessible in mutant all-Ala due to the formation of more helical structures (Fig. 2B). Contributions of Disulfide Bonds to the Structure and Stability of Endostatin-ES has a compact tertiary structure with two hydrophobic regions, which can be partly attributed to the two disulfide bonds (5,7). The smaller hydrophobic region is built up mainly by four strands, two of which are circulated by the disulfide bond Cys 135 -Cys 165 that locates near the surface of ES and links a short ␣-helix and a loop between two strands (5). The other disulfide bond, Cys 33 -Cys 173 , connects the long ␣-helix to the central ␤-sheets in the larger hydrophobic region of endostatin (5).
The disulfide bond Cys 135 -Cys 165 is more important than Cys 33 -Cys 173 in retaining the native tertiary structure of ES. Without disulfide bond Cys 135 -Cys 165 , the twisted loops may be released; thus, the smaller hydrophobic core may collapse more easily, which may in turn disrupt the native structure of ES, induce the noncooperative unfolding induced by urea, and result in low stability against proteolysis (Figs. 4E and 5). On the other hand, the disulfide bond Cys 33 -Cys 173 links the long ␣-helix to the larger hydrophobic region. Without disulfide bond Cys 33 -Cys 173 , the helix might still cover the large hydrophobic region by hydrophobic interactions and hydrogen bonds, which can hold the native-like but unstable tertiary structure of C33A/C173A (Fig. 2, A and C). However, low concentrations of denaturants may induce the exposure of the larger hydrophobic region of C33A/C173A, which can make C33A/C173A easily unfolded by denaturants (Figs. 3

(B and E) and 4 (B and E)).
Although the stability of C33A/C173A is much lower than that of the wild type, the unfolding of C33A/C173A is more cooperative than that of the wild type (Figs. 3E and 4E, Table  I). The lower C m value and the higher m value suggest that the native-like structure of C33A/C173A is fragile. In contrast with C33A/C173A, C135A/C165A unfolds completely at higher concentrations of urea, which can be attributed to the remaining disulfide bond, Cys 33 -Cys 173 (Table I). It appears that the disulfide bond Cys 33 -Cys 173 contributes to the stability of ES more than Cys 135 -Cys 165 , whereas the disulfide bond Cys 135 -Cys 165 contributes to the tertiary structure of ES more than Cys 33 -Cys 173 .
Our group reported previously that ES is very acid-resistant (7). Our present studies on the mechanism of acid-induced unfolding of wild type ES and its mutants confirm our hypothesis that the acid-resistant property of ES is mainly from the contribution of the disulfide bonds (7,53). No matter which disulfide bond is mutated, the stability of ES is decreased significantly under acid conditions (Fig. 3E).
Disulfide Bonds Restrict the Helical Propensity of Endostatin Backbone-The secondary structure of ES contains predominantly ␤-sheets and loops; it also contains two ␣-helices (5). Resnick et al. (9) reported that disulfide bonds can provide additional stabilities to the secondary structure. The helical content was reported to be decreased differently in three reduced derivatives of lysozyme (54) and in gradual cleavage of disulfide bonds of bovine serum albumin (55). However, we found that the helical content of three ES mutants is much more than that of the wild type (Fig. 2B). Both C33A/C173A and all-Ala, without the disulfide bond Cys 33 -Cys 173 , have almost an equal amount of helical structures, which is much more than that of C135A/C165A. The extra ␣-helical structures in mutants C33A/C173A and all-Ala may come from some loops transformed into helices without the restriction of disulfide bond Cys 33 -Cys 173 . Since all-Ala lacks one disulfide bond, Cys 135 -Cys 165 , compared with C33A/C173A, and all-Ala has no tertiary structure, whereas C33A/C173A appears to have some native-like structures, the disulfide bond Cys 135 -Cys 165 protects mainly the tertiary structure, whereas the disulfide bond Cys 33 -Cys 173 restricts mainly the formation of more helices in ES (Fig. 2, A-C).
Disulfide bond mutations induce significant, various, and distinct changes of the secondary structure and tertiary structure and the stabilities of ES. Moreover, these structural changes correlate with the different inhibitory activities of ES mutants on the migration and proliferation of HMECs. Taken together, our results confirm the significance of disulfide bonds in ES and may provide an explanation that the difficulty in refolding recombinant endostatin is probably mainly due to the nested pattern of the two pairs of disulfide bonds (2). We speculate that the structural changes of disulfide bond mutants impair the binding of ES to its receptors on the surface of endothelial cells, the internalization by endothelial cells, and the localization to cytoskeleton and nuclei of endothelial cells, which result in the decrease of inhibitory activities of the mutants on the migration and proliferation of HMECs.
Hao Zhou, Wei Wang and Yongzhang Luo