Protein Engineering vol.14 no.2 pp.135–140, 2001

Asn to Lys mutations at three sites which are N-glycosylated in the mammalian protein decrease the aggregation of Escherichia coli-derived erythropoietin

Linda O.Narhi1, Tsutomu Arakawa2, Kenneth Aoki, Jie Wen, Steve Elliott, Thomas Boone and Janet Cheetham Amgen Inc., Thousand Oaks, CA 91320 and 2Alliance Protein Laboratories, CSUCI, Camarillo, CA 93016, USA 1To

whom correspondence should be addressed. E-mail: [email protected]

Erythropoietin (EPO) derived from Escherichia coli is unstable to elevated temperature and tends to aggregate with time, making it unsuitable for high-resolution structure analysis. The mammalian EPO contains about 40% carbohydrate, which makes this protein more stable and less prone to aggregate than non-glycosylated E.coli-derived EPO, but makes it unsuitable for high-resolution analysis owing to its size and flexibility. In an attempt to decrease the aggregation of E.coli-derived EPO, the three asparagine residues at positions 24, 38 and 83 were mutated to lysine residues. In the native protein, these residues are the sites of N-linked glycosylation, which suggests that they should be located on the surface of the protein and should not be involved in interactions in the hydrophobic protein core. Therefore, the substitution of basic amino acids for these neutral asparagine residues is not expected to affect the protein structure, but should increase the isoelectric point of the protein and its net positive charge, decreasing its tendency to aggregate at or below neutral pH due to electrostatic interactions. No apparent alterations in receptor binding, as determined by both cell-surface receptor competition assay and in vitro receptor dimerization experiments, were observed when these mutations were introduced into the EPO sequence. However, this mutant protein displayed a significant increase in stability to heat treatment and to storage, relative to the wild-type molecule. This resulted in a greater number of observable cross peaks in the mutant EPO in 2D NOESY experiments. However, the mutant was similar to the wild-type in stability when urea was used as a denaturant. This indicates that the introduced mutations resulted in a decrease in aggregation with heating or with prolonged incubation at ambient temperature, without changing the conformational stability or the receptor binding affinity of the mutant protein. This approach of placing charged residues at sites where N-glycosylation occurs in vivo could be applied to other systems as well. Keywords: NMR/protein conformation/protein engineering/ protein stability

Introduction Naturally occurring human erythropoietin (EPO) and also EPO derived from other mammalian sources is heavily glycosylated, with carbohydrate accounting for ~40% of the molecule by weight (Davis et al., 1987; Kolvenbach et al., 1991). EPO contains 166 amino acid residues and has N-linked glycosylation at three positions, 24, 38 and 83, and O-linked glycosylation at position 126 (Lai et al., 1986; © Oxford University Press

Takeuchi et al., 1988). This extensive glycosylation makes the protein too large and complex for NMR analysis, which therefore requires a non-glycosylated form of the protein. Deglycosylated or Escherichia coli-derived non-glycosylated EPO is folded into the same conformation as the mammalianderived EPO, as determined by circular dichroism (CD) and fluorescence (Davis et al., 1987; Narhi et al., 1991). Although non-essential for in vitro receptor binding, glycosylation at these positions increases the stability and solubility of the protein (Narhi et al., 1991; Sytkowski et al., 1991; Wasley et al., 1991; Delorme et al., 1992; Higuchi et al., 1992). The non-glycosylated EPO is prone to aggregation, making it difficult to carry out high-resolution structure analysis (Narhi et al., 1991; Endo et al., 1992). The isoelectric point of the non-glycosylated EPO, determined by isoelectric focusing in the presence of urea, is about 9.2 (Davis et al., 1987). Mutation of the above three Asn residues at the N-linked glycosylation sites to Lys should make the isoelectric point more basic, resulting in an increase in the net positive charge at or below neutral pH and possibly a decrease in the aggregation of the protein. Since these Asn residues, being glycosylation sites, are not involved in protein folding and are located at the protein surface, these mutations should not interfere with the refolding of the protein. Therefore, we generated an EPO mutant with Asn24, -38 and -83 mutated to Lys, expressed it in E.coli and purified and refolded the expressed protein using a procedure similar to that used to purify the wild-type protein. The refolded mutant was comparable to the wild-type protein in a receptor binding assay, yet showed a greater stability against heating and storage, with less aggregation, as described in this paper. This was reflected in improved spectral features in high-resolution NMR experiments, opening a way to obtain the three-dimensional structure of the EPO by this technique. This approach of mutating residues at N-glycosylation sites to either negatively or positively charged amino acids or more polar ones could be generally applicable to other proteins which are prone to aggregation. Materials and methods Materials The mutant gene was obtained by site-directed mutagenesis. The wild-type (wt) and mutant (Lys) proteins were purified and refolded from the inclusion bodies of lysed E.coli cells. The proteins were solubilized with N-lauroylsarcosine and allowed to oxidize with the addition of CuSO4. The oxidized extract was fractionated by cation-exchange chromatography at pH 8.0 on CM Sepharose FF (Pharmacia). Fractions containing EPO were pooled, diluted with an equal volume of water and then subjected to hydroxyapatite chromatography on HA Ultrogel (Sepracor). The EPO was further purified by gel filtration on Sephacryl S-200 HR (Pharmacia). The EPO was then concentrated and buffer exchanged by ultrafiltration on Centricon-10 concentrators (Amicon) into appropriate buffers for various studies. 135

L.O.Narhi et al.

Fig. 1. Size exclusion chromatograms (UV signal) of wt-EPO (A), sEPOR (B), sEPOR–wt-EPO complex made at a 2.4:1 molar ratio (C) and sEPOR– Lys-EPO complex made at a 2.4:1 molar ratio (D).

Preparation of NMR samples The NMR sample for native EPO was prepared in 10 mM phosphate, pH 6.8 (10% D2O) containing 6% (w/v) glycerol (99% deuterated) at a protein concentration of 0.4 mM; these conditions appear to be the only ones among those tested that are acceptable for the analysis. Unlabeled and 15N-labeled NMR samples for Lys-EPO were prepared at higher protein concentrations, i.e. 1.8 and 1.3 mM, respectively, in 10 mM deuterated acetic acid, 1 mM NaN3 (8% D2O) and the pH of the samples was adjusted to either pH 6.2 or 6.8 using sodium hydroxide. This sample did not aggregate as readily and could be analyzed under these conditions. Receptor binding Receptor binding was determined by competition with [125I]EPO (Amersham) on a human cell line OCIM1 (Broudy et al., 1988). Cells were grown to (2–5)⫻105 cells/ml, collected by centrifugation, washed twice in binding buffer (RPMI 1640, 1% BSA, 25 mM HEPES, pH 7.3) and resuspended at (1–2)⫻107 cells/ml in binding buffer containing 0.1% NaN3 and 10 µg/ml cytochalaisin B. Cells were incubated with varying amounts of an in-house EPO standard or EPO samples and [125I]EPO at 37°C. After 3 h, cells were centrifuged through phthalate oil [60:40 (v/v) dibutyl phthalate–dinonyl phthalate]. Tubes were frozen in dry ice–ethanol and the pellets clipped and then counted in a gamma counter. Receptor binding activity in units/ml was determined from a standard curve generated with the in-house EPO standard and then converted to specific activity in units/mg with protein concentrations determined spectrophotometrically. Light scattering/size exclusion chromatography On-line light scattering/size exclusion chromatography (SEC) was carried out as described previously (Takagi, 1990; Arakawa 136

Fig. 2. CD spectra of wt-EPO (solid line) and Lys-EPO (dotted line) in the far-UV region (A) and near-UV region (B).

et al., 1994; Philo et al., 1994) to determine the molecular weights of protein complexes. The results were analyzed according to Wen et al. (1996a). The complexes were made by mixing 120 µl of 2.02 mg/ml of the extracellular domain of the EPO receptor (soluble EPO receptor, sEPOR) with 15 µl of 5 mg/ml wt and Lys-EPO, respectively. The results were fitted using the previously published procedure (Wen et al., 1996a,b). The lower concentration samples were made by diluting these two mixtures 10-fold with PBS. CD CD spectra were determined on a Jasco J-720 spectropolarimeter using cylindrical cuvettes with a pathlength of 1 cm for the near-UV (340–240 nm) and 0.02 cm for the far-UV region

Asn to Lys mutations decrease EPO aggregation

Fig. 4. Time course of turbidity at 38°C and 350 nm for wt-EPO (solid line) and Lys-EPO (dotted line).

Fig. 3. Thermal scans of wt-EPO (solid line) and Lys-EPO (dotted line) followed by far-UV CD. Changes in far-UV CD signal at 222 nm were followed as the temperature was increased, using a scan rate of 10°C/h.

(240–190 nm). The thermal stability was determined using the same instrument, a JTC-345 Peltier thermal unit and a rectangular thermal cuvette with a 0.1 cm pathlength. The changes in the ellipticity at 222 nm were monitored as the protein solutions at 0.5 mg/ml were heated from 20 to 70°C at 10 and 100°C/h. The stability of the proteins to urea-induced denaturation was assessed using a Jasco J-500C spectropolarimeter and cylindrical cuvettes with a 0.02 cm pathlength. Stock protein solutions in PBS were mixed with PBS and 8 M urea, resulting in a final solution of 0.5 mg/ml EPO in PBS, at the desired urea concentration. The spectra of the protein solutions were recorded from 300 to 240 nm and the percentage of the total change in ellipticity at 282 nm was plotted versus urea concentration. ANS binding 1,8-Anilinonaphthalenesulfonate (ANS) binding was determined by titrating EPO in PBS with a stock ANS solution and measuring the fluorescence of the solution at 470 nm upon excitation at 380 nm. Fluorescence spectra were determined on an SLM-Aminco SPF-500 spectrofluorimeter, using a cuvette with a 0.5 cm pathlength and protein solutions of 0.25 mg/ml in PBS. Turbidity Turbidity at 287 or 350 nm was monitored at different temperatures using a Response II spectrophotometer. The wildtype and mutant proteins at 0.2 or 1 mg/ml were delivered into small cuvettes with a 1 cm pathlength. A reference solvent and five samples were placed in a Peltier cell holder and their absorbance at 287 or 350 nm was monitored at constant temperature or as the temperature was increased at a constant rate. NMR data collection All NMR spectra were recorded at 20°C using Bruker AMX500 and AMX600 spectrometers and a 1H–13C–15N triple resonance probe. Carrier frequencies were set to 4.70 p.p.m. for 1H and

118 p.p.m. for 15N. Solvent suppression was achieved by lowpower irradiation during the relaxation delay (1.2 s) or by the use of spin-lock pulses (Messerle et al., 1989). The data were processed and analyzed using Felix 230 (BIOSYM). 2D 1H– 1H NOESY experiments (Kumar et al., 1980) were collected at 298 and 293 K, with a mixing time of 140 ms. The data were acquired as 128 complex t1 increments with 2048 complex points in t2. Spectral widths in F2 and F1 were 16.05 and 12.0 p.p.m., respectively. 1H–15H HSQC experiments (Bodenhausen and Ruben, 1980) were recorded as 128 complex t1 increments (64 scans per increment) with 4096 complex points in t2. Quadrature detection was States-TPPI. Results Dimerization of EPO receptor by EPO Binding of EPO to sEPOR was examined using on-line SEC with light scattering detection. Samples of 100 µl of each complex and their controls were injected on to a Superdex 200 column and the chromatogram (UV absorbance detector) of each sample is shown in Figure 1. Both the wt-EPO and Lys-EPO clearly showed similar complex formation with the sEPOR. It has been established previously that the peak eluting at 12.8 ml under these conditions represents the complex 2sEPOR–1EPO (Philo et al., 1996). This conclusion is supported by the calculated molecular weight of the complex eluting from the column. Partial dissociation of both EPO complexes was observed (i.e. the peak of the 2sEPOR–1EPO complex shifts toward the position of the 1sEPOR–1EPO complex) when the high-concentration complexes were diluted 10-fold (data not shown). Both the wt- and Lys-EPO behaved identically upon dilution in this analysis. Cell-surface receptor binding The cell-surface receptor competition assay on OCIM1 cells showed that Lys-EPO has about a 2-fold higher affinity for the EPOR than wt-EPO; the average of four independent experiments showed a specific activity of (1.1⫾0.3)⫻106 and (2.6⫾1.0)⫻106 units/mg for the wt- and Lys-EPO, respectively,. This may simply be due to decreased aggregation of the Lys-EPO during the incubation period, resulting in a higher effective concentration being present throughout the competition assay (see Discussion). 137

L.O.Narhi et al.

Fig. 5. 2D 1H NOESY spectra for EPO and Lys-EPO. (A) EPO H2O–D2O–glycerol-d6 (85:10:6) at 293 K, pH 6.8 in 10 mM phosphate, 0.4 mM. (B) Lys-EPO in H2O–D2O (9:1) at 293 K, pH 6.2 in 10 mM phosphate, 1.8 mM.

Fig. 6. 1H–15N HSQC spectra of 15N-enriched EPO and Lys-EPO. (A) 15N EPO H2O–D2O–glycerol-d6 (84:10:6) at 293 K, pH 6.8 in 10 mM phosphate, 0.4 mM. (B) 15N Lys-EPO in H2O–D2O (9:1) at 293 K, pH 6.2 in 10 mM phosphate, 1.3 mM. The side-chain resonances of Trp51, -64 and -88 (shown boxed) are folded into the spectrum from 130 p.p.m.

Far- and near-UV CD The far- and near-UV CD spectra for the wt- and Lys-EPO were determined in PBS and are shown in Figure 2A and B, respectively. The spectra of the two molecules are identical, within experimental error. The far-UV CD spectra indicate that these two proteins are highly α-helical (~60%) and are identical with the spectra published previously (Davis et al., 1987; Narhi et al., 1991), with the minimum at 208 nm and a shoulder at 222 nm. The near-UV CD spectra of both proteins show maxima at 290, 282, 258 and 266 nm as reported previously (Davis et al., 1987; Narhi et al., 1991). It therefore appears that the Lys-EPO has the same overall protein fold as the wt-EPO. Heat and storage stability The wt- and Lys-EPO at 1 mg/ml in PBS were heated from 25 to 85°C and the absorbance was monitored at 287 nm. Both samples show an abrupt increase in absorbance due to aggregation and hence turbidity as soon as heat-induced 138

unfolding commences. The temperature at which this increase occurs is ~35°C for the wt-EPO and ~40°C for the Lys-EPO. The effect of heat on the secondary structure of EPO was then determined. The wt- and Lys-EPO were heated from 20 to 70°C at 0.5 mg/ml and various heating rates while the ellipticity at 222 nm was monitored (Figure 3). Both proteins again showed a change in ellipticity concurrent with an increase in turbidity, as indicated by an increase in the photomultiplier voltage. The reaction is irreversible, with no recovery of structure observed following cooling and incubation at lower temperatures. The onset of unfolding and beginning of aggregation are dependent on the scan rate used; however, the wt-EPO always unfolds before the Lys-EPO. Using a scan rate of 10°C/h, this occurs at 42°C for the Lys-EPO and 40°C for the wt-EPO, whereas at 100°C/h this occurs at 43°C for the native sequence EPO and 46–48°C for the Lys-EPO. The Lys-EPO also melted later at a 50°C/h scan rate. The fact that the reaction is affected by the scan rate demonstrates that the aggregation of the protein affects the apparent temperature

Asn to Lys mutations decrease EPO aggregation

Fig. 7. Titration of EPO with urea. The native sequence EPO (s) and Lys-EPO (⫻) in PBS were mixed with PBS–urea for a final solution of 0.5 mg/ml EPO at the indicated urea concentration and the far-UV CD spectrum from 240 to 200 nm was recorded. The ellipticity at 222 nm was determined, the signal corresponding to folded and unfolded protein determined and then the percentage of folded protein plotted versus urea concentration.

of unfolding. This suggests that the difference in melting temperatures between the wt- and Lys-EPO is due to differences in the irreversible step, probably aggregation, rather than differences in stability. The higher temperatures at which changes in ellipticity occur reflect the faster scan rate and lower protein concentration used in the CD measurements. The potential of these samples to aggregate was also analyzed more directly by following turbidity changes. The wt- and Lys-EPO at 0.2 mg/ml were incubated at constant temperature and the turbidity was followed at 287 or 350 nm. The curve of turbidity versus time obtained at 38°C is shown in Figure 4. The wt-EPO showed an increase in turbidity as soon as the sample was placed at 38°C, whereas almost no change in turbidity occurred over the entire time course of incubation for the Lys-EPO. At higher temperatures of incubation, the wt-EPO showed turbidity at earlier times than the Lys-EPO. These results are consistent with those obtained by CD, suggesting that the conformational stability of the protein at pH 7.0 is unchanged by the mutations, while the aggregation propensity of the mutant was decreased. Figure 5 shows a comparison of the 2D 1H NOESY spectra of the wt-EPO and the Lys-EPO at pH 6.8 and 6.2, respectively. The most striking difference is the significant increase in the overall number of cross peaks in the Lys-EPO spectrum, relative to the spectrum of wt-EPO. The Lys-EPO 2D NOESY spectrum is characteristic of a typical α-helical protein, with poor chemical shift dispersion in the cross peaks of the amide NH-Ca region and a large number of strong amide NH(i)– NH(i ⫹ 1) cross peaks (Wuthrich, 1986). By contrast, in the case of the wt-EPO there are relatively few cross peaks in the NH–NH region and in other regions of the spectrum the cross peaks are significantly weaker or absent. Increasing the number of scans per increment, to allow for the differences in protein concentration between the wt- and Lys-EPO, showed no difference in overall spectral quality of the wt-EPO. The intensity of the 1H NMR signal was followed for both proteins for 2 weeks. The wt-EPO was not stable in solution under NMR conditions for longer than 72 h. The protein gradually precipitated out of solution, with a concomitant loss in the 1H NMR signal intensity. In contrast, the intensities in both the

1D and 2D NMR spectra of the mutant remained unchanged over a period of weeks. This is consistent with the results reported above and suggests that the Lys-EPO is more stable to long-term storage, with a decreased rate of aggregation and precipitation. A comparison of the 1H–15N HSQC spectra of the wt- and Lys-EPO (Figure 6) shows similar behavior. The locations of the cross peaks in the two spectra are broadly similar, but those in the wt-EPO spectrum are noticeably weaker. Some 124 well-resolved cross peaks can be identified in the LysEPO HSQC spectrum, but only 110 in the wt-EPO data; 157 amide NH peaks would be expected in total for EPO, but in several regions of the spectrum there is significant overlap due to limited chemical shift dispersion. These spectra also clearly show the absence of three side-chain NH2 spin systems in the Lys-EPO, as expected from the mutation of Asn24, -38 and -83 to Lys. These experiments demonstrate that the overall structures of the wt- and Lys-EPO are the same, but that the aggregation of the wt-EPO is more extensive, making it difficult to obtain usable NMR spectra. The mutations to Lys have resulted in a protein that was more amenable to NMR structural characterization. Stability to denaturants In addition to heat-induced denaturation, the conformational stability of both the wt-EPO and the Lys-EPO was examined by urea-induced unfolding. Changes in secondary structure were monitored by following changes in the ellipticity at 222 nm. The plot of percentage of total change in ellipticity versus urea concentration is shown in Figure 7. The total change in ellipticity was determined from the difference between the baseline for the native molecule, obtained at 0–1 M, and that of the unfolded molecule, obtained at 5–6 M urea. The LysEPO and wt-EPO behaved identically during the course of this experiment, both remaining in solution during denaturation, with a transition midpoint of 3.5 M urea. This is consistent with the stability of E.coli EPO published previously (Narhi et al., 1991). ANS binding The surface hydrophobicities of the Lys-EPO and wt-EPO were compared using ANS titration. Solutions of either protein in PBS were titrated with ANS and the fluorescence intensities compared. Both proteins showed identical fluorescence spectra at every ANS concentration and the curve was similar to that published previously (Narhi et al., 1991). It thus appears that the three Asn to Lys point mutations did not affect the overall surface hydrophobicity. It also indicates that ionic interactions with these mutated residues are not involved in ANS binding or the negatively charged ANS would have interacted more strongly with the Lys-EPO, with its three extra net positive charges. Discussion Recombinant mammalian-derived EPO is a stable, highly soluble protein, owing to its extensive glycosylation. However, the E.coli-derived EPO and the mammalian EPO treated with glycanases are both prone to aggregation. Since high-resolution structure analysis by crystallography or NMR was not plausible for mammalian-derived glycosylated EPO due to its size and carbohydrate content, non-glycosylated molecules which have reduced aggregation and increased solution stability were needed for these studies. We have shown here that a mutant EPO with substitutions 139

L.O.Narhi et al.

of lysine residues for Asn24, -38 and -83 exhibits increased stability to heat and to prolonged storage, when compared with the wild-type protein, with a decrease in the rate of aggregation, but has stability identical with that of ureainduced denaturation. What is the mechanism for this improved stability and solubility during heating and storage? The net positive charge of Lys-EPO at physiological pH should be increased by three over the wt-EPO at this pH and thus the solubility of Lys-EPO would also be expected to increase. Urea denaturation demonstrated that wt-EPO and Lys-EPO have identical conformational stability, within experimental error. This suggests that the increased electrostatic free energy due to the three extra positive charges destabilized the native and denatured state of the protein to the same extent, resulting in an identical free energy of unfolding for the wt-EPO and Lys-EPO. In contrast, Lys-EPO showed a small but significant increase in melting temperature compared with the wt-EPO at all scan rates. For both proteins heat-induced denaturation is irreversible, in contrast to the glycosylated molecule, where it is a fully reversible reaction. As shown by an increase in turbidity or photomultiplier tube voltage, thermal unfolding of both proteins is irreversible owing to aggregation of heatinduced structures. We have shown (Narhi et al., 1999) that an irreversible unfolding reaction and the kinetics of the irreversible step, can alter the apparent melting temperature. The thermal unfolding process of EPO may be expressed by k

N ←→ D→aggregate with both wt-EPO and Lys-EPO exhibiting irreversible thermal unfolding due to aggregation. In this case, this most likely occurs concurrently with unfolding, meaning that as the protein unfolds it undergoes aggregation at a rate that depends on the type of protein, protein concentration and temperature. If the unfolded form (D) of the wt-EPO is more prone to aggregation than that of the Lys-EPO (i.e. the value of k is larger for the wt-EPO), the rate of aggregation would be faster for the former and this would explain the faster turbidity development observed for the wt-EPO. This is consistent with the melting temperature being scan-rate dependent. Although the model can explain the observed difference in stability between wt-EPO and Lys-EPO, actual unfolding or aggregation may be more complex. In addition to the described pathway, aggregation of the native state may be temperature dependent or there may be multiple states in the native and unfolded states (Hilser and Freire, 1997). In any state, Lys-EPO would have a decreased rate of aggregation. CD and NMR spectroscopy and ANS binding could not distinguish between the Lys-EPO and the wt-EPO and the conformational stability of the two proteins was also equivalent. This demonstrates that the mutations have not affected the global protein fold of the molecule. We have also shown here that the mutations do not decrease the binding of EPO to the sEPOR and the cell-surface receptor, demonstrating that the Lys-EPO has a conformation capable of binding the receptor. The 2-fold increase in receptor binding of the Lys-EPO could simply reflect the decreased aggregation, and thus increased availability, of the mutant. Alternatively, it could reflect a decrease in the electrostatic repulsion between receptor and ligand. The EPO binding site on the receptor is primarily negative and although residue 24 is not involved in receptor binding, it is located close enough to the binding region that a change in charge might decrease the electrostatic repulsion 140

(Syed et al., 1998). Various mutational and deglycosylation studies have shown that the Asn24, -38 and -83 are not involved in receptor binding (Delorme et al., 1992). This is consistent with these residues being sites for glycosylation and is in full agreement with the observations in this study. Determination of the three-dimensional structure of proteins by X-ray crystallography or high-resolution NMR requires a stable, homogeneous protein sample preparation. This can be achieved by two alternative means, either by modifying the solution conditions or by generating more stable mutant proteins. This study demonstrates that mutation of residues responsible for glycosylation to charged or more polar residues could be used to make a protein more amenable to these types of studies. The NMR spectra of the Lys-EPO had more cross peaks, of greater intensity, than the wt-EPO spectra, such that it was easy to discern that this was a helical protein. Various simple spectroscopic techniques such as those described here can be used to screen the conditions and mutants to find the combination which results in a protein which is stable enough to use under the conditions commonly employed to collect usable NMR spectra. This might be applicable to other glycosylated proteins. Acknowledgements We thank Joan Bennett for manuscript preparation and Doug Paulin for assisting with the figures.

References Arakawa,T., Wen,J. and Philo,J.S. (1994) Arch. Biochem. Biophys., 308, 267–273. Bodenhausen,G. and Ruben,D.J. (1980) Chem. Phys. Lett., 69, 185–188. Broudy,V.C., Lin,N., Egrie,J., De-Haen,C., Weiss,T., Papayannapoulou,T. and Adamson,J.W. (1988) Proc. Natl Acad. Sci. USA, 85, 6513–6517. Davis,J.M., Arakawa,T., Strickland,T.W. and Yphantis,D.A. (1987) Biochemistry, 26, 2633–2638. Delorme,E., Lorenzini,T., Giffin,J., Martin,F., Jacobsen,F., Boone,T. and Elliott,S. (1992) Biochemistry, 31, 9871–9876. Endo,Y., Nagai,H., Watanabe,Y., Ochi,K. and Takagi,T. (1992) J. Biochem., 112, 700–706. Higuchi,M., Oheda,M., Kuboniwa,H., Tomonoh,K., Shimonaka,Y. and Ochi,N. (1992) J. Biol. Chem., 267, 7703–7709. Hilser,V.J. and Freier,E. (1997) Proteins, 27, 171–183. Kolvenbach,C.G., Langley,K.E., Strickland,T.W., Kenney,W.C. and Arakawa,T. (1991) J. Biochem. Biophys. Methods, 23, 295–300. Kumar,A., Ernst,R.R. and Wuthrich,K. (1980) Biochem. Biophys. Res. Commun., 95, 1–4. Lai,P.-H., Everett,R., Wang,F.F., Arakawa,T. and Goldwasser,E. (1986) J. Biol. Chem., 261, 3116–3121. Messerle,B., Wider,G., Otting,G., Weber,C. and Wuthrich,K. (1989) J. Magn. Reson., 85, 608–613. Narhi,L.O., Arakawa,T., Aoki,K.H., Elmore,R., Rohde,M.F., Boone,T. and Strickland,T.W. (1991) J. Biol. Chem., 266, 23022–23026. Narhi,L.O., Chang,B., Sun,B., Philo,J.S. and Arakawa,T. (1999) Pharm. Res., 16, 799–807. Philo,J., Talvenheimo,J., Wen,J., Rosenfeld,R., Welcher,A. and Arakawa,T. (1994) J. Biol. Chem., 269, 27840–27846. Philo,J.S., Aoki,K.H., Arakawa,T., Narhi,L.O. and Wen,J. (1996) Biochemistry, 35, 1681–1691. Sytkowski,A.J., Feldman,L. and Zurbuch,D.J. (1991) Biochem. Biophys. Res. Commun., 176, 698–704. Syed, RS. et al. (1998) Nature, 395, 511–516. Takagi,T. (1990) J. Chromatogr., 280, 409–416. Takeuchi,M., Takasaki,S., Miyazaki,H., Takashi,D., Hoshi,S., Kochibe,N. and Kobata,A. (1988) J. Biol. Chem., 263, 3657–3663. Wasley,L.C., Timony,G., Murtha,P., Soutdmire,J., Dorner,A.J., Caro,J., Krieger,M. and Kaufman,R.J. (1991) Blood, 77, 2624–2632. Wen,J., Arakawa,T. and Philo,J.S. (1996a) Anal. Biochem., 240, 155–166. Wen,J., Arakawa,T., Talvenheimo,J., Welcher,A., Horan,T., Kita,Y., Tseng,J., Nicolson,M. and Philo,J.S. (1996b) In Marshak,D.R. (ed.), Techniques in Protein Chemistry VII. Academic Press, San Diego, pp. 23–31. Wuthrich,K. (1986) NMR of Proteins and Nucleic Acids. Wiley-Interscience, New York. Received May 23, 2000; revised October 19, 2000; accepted November 9, 2000