Human granulocyte/macrophage colony- stimulating factor (hGM-CSF) produced by several recombi- nant sources including Escherichia coli, yeast, and animal ...
Proc. Nati. Acad. Sci. USA Vol. 84, pp. 4428-4431, July 1987 Biochemistry
Increased biological activity of deglycosylated recombinant human granulocyte/macrophage colony-stimulating factor produced by yeast or animal cells (glycosidases/bioassay/hnmunoassay/purification/hmmunoaffnity chromatography) PETER MOONEN*, JEAN-JACQUES MERMOD, JOACHIM F. ERNST, MARLIS HIRSCHI, AND JOHN F. DELAMARTER Biogen, CH-1227 Geneva, Switzerland
Communicated by Phillip A. Sharp, March 23, 1987 (received for review December 2, 1986)
Human granulocyte/macrophage colonyABSTRACT stimulating factor (hGM-CSF) produced by several recombinant sources including Escherichia coli, yeast, and animal cells was studied. Recombinant animal cells produced hGM-CSF in low quantities and in multiple forms of varying size. Mammalian hGM-CSF was purified 200,000-fold using immunoaffnity and lectin chromatography. Partially purified proteins produced in yeast and mammalian cells were assayed for the effects of deglycosylation. Following enzymatic deglycosylation, immunoreactivity was measured by radiounmunoassay and biological activity was measured in vitro on responsive human primary cells. Removal of N-linked oligosaccharides from both proteins increased their immunoreactivities by 4- to 8-fold. Removal of these oligosaccharides also increased their specific biological activities about 20-fold, to reach approximately the specific activity of recombinant hGM-CSF from E. coil. The E. coli produced-protein-lacking any carbohydrate-had by far the highest specific activity observed for the recombinant hGM-CSFs.
from both eukaryotic sources spans a range of molecular masses. In this communication we report the influence of deglycosylation on the response in bio- and immunoassays, together with a purification procedure for the mammalian hGM-CSF (CHO-hGM-CSF).
MATERIALS AND METHODS hGM-CSF Assays. Human bone marrow cultures. Macrophage-depleted bone marrow cells (50,000 cells) were grown in 0.3% agar and 20% (vol/vol) fetal calf serum (FCS) (8, 9) and cultured in the presence of bacterial-hGM-CSF or different dilutions of samples to be tested. After 7 and 14 days, colonies larger than 40 cells were scored. The concentration stimulating the formation of 50% of maximal colony numbers was defined as 50 units (U)/ml. Chronic myelogenous leukemia (CML) assay. Thymidine incorporation was measured on mononuclear cells purified from the peripheral blood of patients with CML (10, 11). One unit/ml induces 50% of maximal [3H]thymidine incorporation by CML cells. For both bioassays using the same cell batch, the variance in results for a given sample was not more than 20% between duplicate determinations. The interassay variability with different cell batches was larger: a factor of -2 between three experiments. Radioimmunoassay. Bacterial-hGM-CSF was conjugated in 0.1 M borate buffer (pH 8.5) with mono-1251I Bolton-Hunter reagent (Amersham) as described (12) and resulted in a specific radioactivity of -300 Ci/mmol (1 Ci = 37 Gl3q). Rabbit anti-yeast-hGM-CSF (yeast-hGM-CSF derived from Saccharomyces cerevisiae) was fixed on Pansorbin cells (Calbiochem) in assay buffer [phosphate buffered saline/ 0.05% (wt/vol) Tween 20/0.05% (wt/vol) azide]. Sample, tracer (40,000 cpm), and antibody-cell suspension (10 1ul) were mixed in small tubes (3 x 50 mm). After 3 hr of incubation the tubes were centrifuged and the pellets were washed with assay buffer. A standard curve with bacterialhGM-CSF (0.5-100 ng) was prepared by plotting the pellet radioactivities against the logarithm of the protein quantities. The maximal sensitivity was about 0.5 ng and the halfmaximal binding was =6 ng of protein. Antibodies. Swine antiserum against rabbit immunoglobulins, which was conjugated with horseradish peroxidase, and rabbit antiserum against fetal calf serum (anti-FCS) were from Dakopatts (Copenhagen, Denmark). Rabbit anti-hGM-
Colony-stimulating factors are glycoproteins that regulate the growth and differentiation of all six blood cell types from progenitor cells. Granulocyte/macrophage colony-stimulating factor (GM-CSF) is defined by its ability to stimulate bone marrow precursor cells to proliferate and differentiate into granulocyte and macrophage colonies (1). Human GM-CSF (hGM-CSF) also stimulates the activity of neutrophils, eosinophils, and macrophages (2-4). hGM-CSF has been partially purified from placental conditioned medium (5) and purified to homogeneity from medium conditioned by the hairy T-cell leukemia line Mo (3). The glycoprotein isolated from the Mo cell line had a molecular mass of 22 kDa. Recombinant hGM-CSF purified from monkey cells had a molecular mass of 19 kDa (6). As predicted from the nucleotide sequence, hGM-CSF has 127 amino acid residues and a calculated molecular mass of 14,459 Da. This size agrees with the value determined experimentally for the homogenous recombinant hGM-CSF produced by Escherichia coli (bacterial-hGM-CSF) (2). The apparent molecular mass variability of hGM-CSF from animal cells has been attributed to different glycosylation patterns. Native GM-CSF from mouse lung has been reported to be secreted into conditioned medium in multiple molecular mass forms (7). Despite its lack of carbohydrate, bacterial-hGM-CSF was found to exhibit the in vitro biological activities of native hGM-CSF except erythroid burstpromoting activity (2). To test the effects of glycosylation on the activity of the protein, recombinant hGM-CSF was produced in yeast and Chinese hamster ovary (CHO) cells. Recombinant hGM-CSF
Abbreviations: CHO, chinese hamster ovary, bacterial hGM-CSF, CHO-hGM-CSF, and yeast-hGM-CSF, recombinant human granulocyte/macrophage colony-stimulating factor derived from E. coli, CHO cells, and S. cerevisiae, respectively; CML, chronic myelogenous leukemia; FCS, fetal calf serum; endo H, endoglycosidase H; endo F, endoglycosidase F; U, units. *To whom reprint requests should be addressed at: Biogen S.A., 46 Route des Acacias, 1227 Geneva, Switzerland.
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.
4428
Biochemistry:
Moonen
et
al.
CSF was obtained by immunizing rabbits with partially purified recombinant GM-CSF from yeast or E. coli (13). The rabbit anti-bacterial-hGM-CSF was specific as determined by immunoblotting. No cross-reaction was observed with proteins present in medium of nontransfected yeast cells and, except for a protein with an apparent molecular mass of 5000, cross-reaction also was not observed with the proteins in conditioned medium of nontransfected CHO cells. Calf anti-bacterial-hGM-CSF was obtained from H. W. Reid (Moredun Institute, Edinburgh). To prepare an immunoaffinity column the immunoglobulin fraction of the calf antiserum was isolated by sodium sulfate precipitation and DEAESepharose chromatography (14). Dialysis against 0.2 M borate buffer (pH 8.6) was followed by coupling to vinylsulfone agarose (Mini leak, Kem-En-Tec, Hellerup, Denmark) according to the procedure of the manufacturer. A second immunoaffinity column (anti-FCS-matrix) was prepared in the same way, using anti-FCS. hGM-CSF Sources. Purified bacterial-hGM-CSF was a gift ofJ. Schrimsher and had a specific activity of 108 U/mg in the CML assay. High levels of yeast-hGM-CSF were secreted from S. cerevisiae using a fusion to the yeast a mating factor signal peptide (J.F.E., unpublished data). The protein was partially purified by Con A and molecular sieve chromatography. A high molecular mass fraction with an apparent mass of 50-70 kDa was used for immunization and deglycosylation experiments. Production of hGM-CSF by transfected CHO cells was in medium containing 10% (vol/vol) FCS. The CHO cell clone used was derived from transfection with a vector that contained the hGM-CSF gene. Transcription was promoted by a simian virus 40 early promoter (J.-J.M., unpublished data). Isolation of CHO-hGM-CSF. All manipulations were at 4°C. Concentrated conditioned medium was dialyzed, filtered, and loaded on the anti-CSF matrix, equilibrated in phosphate-buffered saline with 0.05% (wt/vol) azide and PEG 6000. Bound proteins were eluted with 0.2 M glycine hydrochloride (pH 3.0). The acid-eluted fractions were immediately neutralized and assayed for CHO-hGM-CSF by immunoblotting. The fractions that produced immunopositive staining bands were concentrated and loaded on the anti-FCS matrix. The flow through was dialyzed against 50 mM Tris-HCl/0.15 M NaCl/0.05% (wt/vol) azide/1 mM CaCl2/1 mM ZnCl2, pH 7.2, and loaded on a wheat germlectin-Sepharose column (Pharmacia) equilibrated in the last mentioned buffer. Column-bound hGM-CSF was eluted with 10% (wt/vol) N-acetylglucosamine. The sugar-eluted pool was concentrated and dialyzed against phosphate-buffered saline/0.05% azide/0.05% PEG 6000 and stored in aliquots at -200C. Other Procedures. Glycosidases were from Boehringer Mannheim. Proteins were deglycosylated in 0.1 M sodium acetate buffer (pH 5.6). Bacterial- and yeast-hGM-CSF were digested with endoglycosidase H (endo H) (15). Bacterialand CHO-hGM-CSF were incubated first with neuraminidase; this was followed by incubation with a mixture of 3-galactosidase, ,B-N-acetylglucosaminidase, and endoglycosidase F (endo F) (16). After incubation for 24 hr at 370C, samples were directly analyzed by immunoblotting, radio-
Proc. Natl. Acad. Sci. USA 84 (1987)
4429
immunoassay, and bioassay. Prestained low molecular mass markers were from Bethesda Research Laboratories. NaDodSO4 gel electrophoresis used the discontinuous system of Laemmli (17) on slabs of 13.5% (wt/vol) acrylamide. After gel electrophoresis, immunoblots of the slabs were prepared following the protocol of Towbin et al. (18) with rabbit anti-bacterial-hGM-CSF as the first antibody. Protein concentrations were determined with the procedure and bicinchoninic acid reagent of Pierce using bovine serum albumin as the reference protein.
RESULTS Isolation of CHO-hGM-CSF. Recombinant hGM-CSF secreted by CHO cells was a very minor fraction of the protein in the FCS-containing culture medium. The purification method, summarized in Table 1, results in a 200,000-fold purification, retaining different glycosylated forms. Analysis of the final pool by gel electrophoresis and immunostaining shows three major reactive regions (as demonstrated in lanes b of Fig. 2). Deglycosylation Experiments. Proteins can be deglycosylated by either chemical or enzymic methods. Enzymes are preferred as they allow milder conditions to be used and a clearer interpretation of the moiety removed. We used endo H on our partially purified high molecular mass yeast-hGMCSF. This glycosidase specifically removes N-linked mannan carbohydrate chains. Treatment with endo H reduced the apparent molecular mass of the protein to that of a bacterialhGM-CSF reference (Fig. 1). As expected, endo H treatment of bacterial-hGM-CSF had no detectable effect on its migration in gel electrophoresis. Due to the differences in oligosaccharide processing of yeast and mammalian cells, other glycosidases were used for CHO-hGM-CSF. Endoglycosidases are only effective after removal of terminal sialic acid residues by neuraminidase (16). When partially purified CHO-hGM-CSF was treated with neuraminidase, all three major immunoreactive regions migrated with an apparently lower molecular mass (Fig. 2). In an attempt to remove most of the carbohydrate, the partially purified CHO-hGM-CSF was treated with a mixture of exoglycosidases (neuraminidase, 0-galactosidase, and ,B-Nacetylglucosaminidase) and endo F (an endoglycosidase that removes N-linked oligosaccharides, ref. 16). Much of the immunostaining protein moved from the high molecular mass region of the gel to nearly the position of the E. coli-produced protein (Fig. 2). Treatment of bacterial-hGM-CSF with these glycosidases did not significantly change the migration of the protein (data not shown). An interpretation of the results obtained with the glycosidases is that the region with the highest molecular mass contains protein with both potential N-glycosylation sites occupied, whereas the middle region has one N-glycosylation site occupied and the lower region has no N-glycosylation. Digestion of the neuraminidasetreated CHO-hGM-CSF with an enzyme that removes 0linked oligosaccharides (19) lowered the position of all three major bands, indicating the possibility of O-glycosylation (P.M., unpublished data).
Table 1. Isolation of CHO-hGM-CSF
Protein, CHO-hGM-CSF,* CHO-hGM-CSF as % Purification Material of total protein factor mg Ag Conditioned medium 5320 71 0.0013 Acid-eluted pool from anti-hGM-CSF-matrix 4.9 20 0.41 30,700 Flow through from anti-FCS-matrix 0.7 18 2.6 193,600 Sugar-eluted pool from wheat germ-Sepharose 0.3 8 2.7' 200,000 *Determined by radioimmunoassay. tAfter deglycosylation the determined amount is 64 .tg of protein; the final preparation contains -20% CHO-hGM-CSF.
Yield, t 100 28 25 11
4430
Biochemistry: Moonen et al. 1 2 3 4
Proc. Natl. Acad. Sci. USA 84 a
5 67
26
a
c
a
d
a b
a
kDa
kDa 43
b
(1987)
w
43
P
26
*
I.
4
I 4
P
18
4,4
Ili.,
oi 4
OWD4 14
0-
FIG. 1. Immunostaining of protein gel transfer of bacterial- and yeast-hGM-CSF before and after treatment with endo H. Lanes 1, 4, and 7, prestained low molecular mass markers; lane 2, untreated bacterial-hGM-CSF; lane 3, endo H-incubated bacterial-hGM-CSF; lane 5, untreated yeast-hGM-CSF; lane 6, endo H-digested yeast-hGM-CSF.
Effects of Deglycosylation on Immuno- and Bioreactivity. To determine the effects of deglycosylation on biological and immunological reactivity, samples of the proteins largely stripped of carbohydrate were tested. Although the antibody used in the radioimmunoassay was raised against the glycosylated and partially purified yeast-hGM-CSF, removal of N-linked oligosaccharides increased the immunoreactivity of yeast-hGM-CSF and CHO-hGM-CSF 4-fold and 8-fold, respectively (Table 2). Thus, N-linked sugars appear to hinder the interaction of the protein with the specific immunoglobulins. Removal of sialic acid residues did not change the response in the immunoassay. Two different biological assays were used to measure the effects of deglycosylation on activity (Table 2). Since deglycosylated proteins compete equally with the E. coliderived tracer, specific activities were calculated using protein values determined by radioimmunoassay of the deglycosylated samples. The CML assay measures the proliferative stimulus of hGM-CSF, whereas the bone marrow colony assay measures both the proliferation and differentiation of precursor cells. Partially purified hGM-CSF secreted from
6
P
3
P
FIG. 2. Immunostaining of protein gel transfer of the same concentration of bacterial and CHO-hGM-CSF before and after treatment with glycosidases. Lanes a, loaded with untreated bacterial-hGM-CSF; lanes b, untreated CHO-hGM-CSF; lane c, loaded with neuraminidase-treated CHO-hGM-CSF; lane d, CHO-hGMCSF treated with neuraminidase, endo F, f-galactosidase, and (3-N-acetylglucosaminidase. Prestained low molecular mass markers were loaded in the outside slots to allow determination of the molecular mass positions after transfer.
yeast cells was deglycosylated and its activity was measured
in both assays before and after digestion (Table 2). The specific activity of the deglycosylated protein increased 20-fold (Table 2). Similarly, removal of the major N-linked carbohydrate moieties from partially purified CHO-hGMCSF significantly increased the specific activity of the protein in both assays. In contrast, treatment with neuraminidase had no effect on the specific activity of CHO-hGM-CSF. Incubation of the E. coli-derived protein with glycosidases did not significantly change its biological activity when measured by either assay.
DISCUSSION As native hGM-CSF is a glycoprotein, the sugar side chains may be expected to have a significant physiological role. Glycosylation of two hematopoietic growth factors (erythropoietin and megakaryocyte colony-stimulating fac-
Table 2. Influence of deglycosylation on radioimmunoassay response and on specific activities in
bioassays Glycosidase
Immunoassay,
CML assay, SA (U/mg) x 10-7 18 36
Bone marrow assay, SA (U/mg) x 10-8 5.0 5.0 0.3 8.1 6.0 6.0
treatment ng/ml None 150 endo H 180 1.2 None 35 Yeast-hGM-CSF endo H 135 19 Bacterial-hGM-CSF None 170 8 Exo + endo* 170 15 None 20 0.45 0.5 CHO-hGM-CSF Neuraminidase 22 0.4 ND 9.0 12 Exo + endo* 160 ND, not determined; SA, specific activity. Specific activities were calculated using protein concentrations as determined by radioimmunoassay for the deglycosylated samples of yeast-hGM-CSF and CHO-hGM-CSF [endo H and exoglycosidase (Exo) plus endoglycosidase (endo) digestion,
Sample Bacterial-hGM-CSF
respectively]. *For incubation with exoglycosidases and endoglycosidases, samples were treated with neuraminidase, 3-galactosidase, 3-N-acetylglucosaminidase, and endo F.
Biochemistry:
Moonen et al.
tor) has been shown to be necessary for the physiological activity of the proteins (20, 21). Several functions have been suggested for the carbohydrate side chains of glycoproteins, including protection against rapid proteolytic digestion (22), prolongation of serum half-life (23), and improvement of stability or solubility. In bacteria no glycosylation takes place, whereas both yeast and animal cells can synthesize glycoproteins albeit with different carbohydrate moieties (24). We have produced hGM-CSF by recombinant techniques in all three cell types. Following partial purification of hGM-CSF produced in both eukaryotes, the influence of glycosylation on immuno- and bioassays was tested and compared to bacterial-hGM-CSF. The striking increase in immunoreactivity of deglycosylated hGM-CSF parallels literature reports for heteroantisera. For prolactin (25) and interferon-y (26) the glycosylated form had only a fraction of the immunoreactivity of the nonglycosylated form. With monoclonal antibodies an increase or decrease in immunoreactivity of deglycosylated interferon-y was measured, dependent on the monoclonal antibody used
(27).
Reported observations on the influence of carbohydrate on the bioactivities ofglycoproteins vary. The glycosylated form of sheep prolactin had 60% of the bioactivity of the nonglycosylated form (25), whereas glycosylated porcine prolactin had 140% of the activity of the nonglycosylated form (28). Enzymatic deglycosylation of native human interferon-y did not alter its bioactivity (29). Placental hormones have the unusual property that the deglycosylated forms are competitive inhibitors of the native hormones (30). Removal of the N-linked carbohydrate moieties from both yeast and animal cell hGM-CSF significantly increased their biological activities in two separate in vitro assays. Treatment of CHO-hGM-CSF with neuraminidase to remove sialic acid residues had no significant effect in any of the assays. These results are consistent with other observations that sialic acid residues are important for prolonging serum half-life of glycoproteins but do not effect in vitro bioactivity (23, 31). Our results suggest that N-linked carbohydrate moieties hinder the interaction of glycosylated recombinant hGMCSF with its receptor on myeloid precursor cells or the signal transduction in these cells. Although the oligosaccharides reduced the specific activity of recombinant hGM-CSF, they did not effect the functional activity of the protein on either bone marrow cells or mature granulocytes (2). Whether the animal cell-produced hGM-CSF has an erythroid burstpromoting activity not found in E. coli-produced hGM-CSF needs to be further investigated. Similarly, the dependence of the survival and stimulation of monocytes on the carbohydrate moiety and its influence on in vivo activity have to be determined. However, our results in vitro suggests that the nonglycosylated E. coli-produced protein may be the most useful for clinical application. We thank Ms. B. Burrus and Mr. R. Gaffner for technical assistance; Drs. H. W. Reid and J. Schrimsher for gifts of calf
Proc. Natl. Acad. Sci. USA 84 (1987)
4431
antibody and bacterial-hGM-CSF, respectively; and Ms. C. Meissner for the preparation of the manuscript. 1. Metcalf, D. (1986) Blood 67, 257-267. 2. Burgess, A. W., Begley, C. G., Johnson, G. R., Lopez, A. F., Williamson, D. J., Mermod, J.-J., Simpson, R. J., Schmitz, A. & DeLamarter, J. F. (1987) Blood 69, 43-51. 3. Gasson, J. C., Weisbart, R. H., Kaufman, S. E., Clark, S. C., Hewick, R. M., Wong, G. G. & Golde, D. W. (1984) Science 226, 1339-1342. 4. Vadas, M. W., Nicola, N. A. & Metcalf, D. (1983) J. Immunol. 130, 795-799. 5. Nicola, N. A., Metcalf, D., Johnson, G. R. & Burgess, A. W. (1979) Blood 54, 614-627. 6. Wong, G. G., Witek, J. S., Temple, P. A., Wilkens, K. M., Leary, A. C., Luxenberg, D. P., Jones, S. S., Brown, E. L., Kay, R. M., Orr, E. C., Shoemaker, C., Golde, D. W., Kaufmann, R. J., Hewick, R. M., Wang, E. A. & Clark, S. C. (1985) Science 228, 810-815. 7. Burgess, A. W., Metcalf, D., Sparrow, L. G., Simpson, R. J. & Nice, E. C. (1986) Biochem. J. 235, 805-814. 8. Pike, B. & Robinson, W. A. (1970) J. Cell. Physiol. 76, 77-84. 9. Iscove, N. N., Guilbert, J. & Weyman, C. (1980) Exp. Cell. Res. 126, 121-126. 10. Griffin, J. D., Sullivan, R., Beveridge, R. P., Larcom, P. & Schlossman, S. F. (1983) Blood 63, 904-911. 11. Lansdorp, P. M. (1985) Dissertation (University of Amsterdam). 12. Bolton, A. E. & Hunter, W. M. (1973) Biochem. J. 133, 529-538. 13. DeLamarter, J. F., Mermod, J.-J., Liang, C.-M., Eliason, J. F. & Thatcher, D. R. (1985) EMBO J. 4, 2575-2581. 14. Thorpe, R. & Johnstone, A. (1985) in Immunochemistry in Practice (Blackwell Scientific, Oxford), pp. 44-46. 15. Trimble, R. B. & Maley, F. (1984) Anal. Biochem. 141, 515-519. 16. Plummer, T. H., Elder, J. H., Alexander, S., Phelan, A. W. & Tarentino, A. L. (1984) J. Biol. Chem. 259, 10700-10704. 17. Laemmli, U. K. (1970) Nature (London) 227, 600-685. 18. Towbin, A., Staehlin, T. & Gordon, J. (1979) Proc. Natl. Acad. Sci. USA 76, 4350-4354. 19. Lamblin, G., Lhermitte, M., Klein, A., Roussel, P., Van Halbeek, M. & Vliegenthart, J. F. G. (1984) Biochem. Soc. Trans. 12, 599-600. 20. Dordal, M. S., Wang, F. F. & Goldwasser, E. (1985) Endocrinology 116, 2293-2299. 21. Hoffman, R., Yang, H. H., Bruno, E. & Straneva, J. E. (1985) J. Clin. Invest. 75, 1174-1182. 22. Lefkovitz, I. (1986) Proc. Natl. Acad. Sci. USA 83, 3437-3438. 23. McFarlane, I. G. (1983) Clin. Sci. 64, 127-135. 24. Kornfeld, R. & Kornfeld, S. (1985) Annu. Rev. Biochem. 54, 631-664. 25. Lewis, U. J., Singh, R. N., Lewis, L. J., Seavy, B. K. & Sinha, Y. N. (1984) Proc. Natl. Acad. Sci. USA 81, 385-389. 26. Maeger, A. & Leist, T. (1986) J. Interferon Res. 6, 225-232. 27. Le, J., Rubin, B. Y., Kelker, H. C., Feit, C., Nagler, C. & Vilcek, J. (1984) J. Immunol. 132, 1300-1304. 28. Pankov, Yu. A. & Butnev, U. Yu. (1986) Int. J. Pept. Protein Res. 28, 113-123. 29. Kelker, H. C., Yip, Y. K., Anderson, P. & Vilcek, J. (1983) J. Biol. Chem. 258, 8010-8013. 30. Sairam, M. R. & Bhargavi, G. N. (1985) Science 229, 65-67. 31. Dufau, M. L., Catt, K. J. & Tsuruhara, T. (1971) Biochem. Biophys. Res. Commun. 44, 1022-1029.
