Active human erythropoietin expressed in insect cells using a baculovirus vector: A role for N-linked oligosaccharide

Biologically active recombinant human erythropoietin has been expressed at high levels in an insect cell background. Expression involved the preparation of a human erythropoietin cDNA, the transfer of this cDNA to the Autographa californica nuclear polyhedrosis virus (AcNPV) genome under the polyhedrin gene promoter, and the subsequent infection of Spodoptera frugiperda cells with recombinant AcNPV. Erythropoietin cDNA was prepared through the expression of the human erythropoietin gene in COS cells using pSV2 and the construction of a COS cell cDNA library in bacteriophage Lambda GT10. Prior to transfer to the AcNPV genome, erythropoietin cDNA isolated from this library was modified at the 3′-terminus in order to replace genomic erythropoietin for SV40 cDNA derived from pSV2. Transfer of this cDNA to AcNPV and the infection of S. frugiperda cells with cloned recombinant virus led to the secretion of erythropoietin: based on bioassay, rates of hormone secretion (over 40 U/ml per h) were 50-fold greater than observed for COS cells. The purified recombinant product possessed full biological activity (at least 200000 U/mg), but was of lower M_r (23000) than human erythropoietin produced in COS cells (30000) or purified from urine (30000 to 38000). This difference was attributed to the glycosylation of erythropoietin in S. frugiperda cells with oligosaccharides of only limited size. Further removal of N-linked oligosac-charides from this M_r 23000 hormone using N-Glycanase yielded an apo-erythropoietin (M_r 18000) which possessed substantially reduced biological activity. These results indicate that glycosylation, but not the normal processing of oligosaccharides to complex types, is required for the full hormonal activity of human erythropoietin during red cell development.     

Expression of the influenza virus haemagglutinin in insect cells by a baculovirus vector.

The insect baculovirus Autographa californica nuclear polyhedrosis virus (AcNPV) has played a major role in studies on the molecular biology of insect DNA viruses. Recently, this system has been effectively adapted as a highly efficient vector in insect cells for the expression of several mammalian genes. A cDNA sequence of the influenza (fowl plague) virus haemagglutinin gene has been inserted into the BamHI site of the pAc373 polyhedrin vector. Spodoptera frugiperda cells were co-transfected with this construct, pAc-HA651, and authentic AcNPV DNA. Recombinant virus was selected by adsorption of transfected cells to erythrocytes followed by serial plaque passages on S. frugiperda cells. We have determined the site of insertion of the haemagglutinin gene into the AcNPV genome by restriction enzyme cleavage and Southern blot hybridization analyses using haemagglutinin cDNA as a probe. The influenza haemagglutinin gene is located in the polyhedrin gene of AcNPV DNA. Immunofluorescent labelling, immunoprecipitation and immunoblot analyses with specific antisera revealed that S. frugiperda cells produce immune reactive haemagglutinin after infection with the recombinant virus. The haemagglutinin is expressed at the cell surface and has haemolytic capacity that has been activated by post-translational proteolytic cleavage. When chickens were immunized with S. frugiperda cells expressing haemagglutinin, they developed haemagglutinin-inhibiting and neutralizing antibodies and were protected from infection with fowl plague virus. These observations demonstrate that the haemagglutinin is processed in insect cells in a similar fashion as in fowl plaque virus-infected vertebrate cells and that it has full biological activity.     

Expression of human tyrosine hydroxylase cDNA in invertebrate cells using a baculovirus vector

A human cDNA containing the complete coding sequence for a human tyrosine hydroxylase (EC 1.14.16.2, form 2) was introduced into the genome of Autographa californica nuclear polyhedrosis virus (AcNPV) downstream to the polyhedrin promoter. Infection of Spodoptera frugiperda cells (SF9) with recombinant virus resulted in the expression of human tyrosine hydroxylase in these invertebrate cells. Characterization of tyrosine hydroxylase activity in infected SF9 cells demonstrated both substrate and cofactor kinetics that were characteristic of those previously reported for the native human enzyme. Both 3-iodotyrosine and alpha-methyl-p-tyrosine competitively inhibited the recombinantly produced tyrosine hydroxylase with Ki values of 1.2 and 16 microM, respectively, similar to those previously reported for the rat and human enzymes. Western blot analysis of extracts of SF9 cells infected with the recombinant baculovirus containing human tyrosine hydroxylase cDNA revealed a major immunoreactive band with an apparent Mr of 60 kDa, identical to the size of the immunoreactive protein from rat adrenal and caudate nucleus. The use of the baculovirus expression system to produce abundant quantities of each of the multiple forms of active human tyrosine hydroxylase in eukaryotic cells should facilitate structural analysis and help clarify the physiological significance of each of the isoenzymes.     

Expression cloning of the murine erythropoietin receptor

Two independent cDNA clones encoding the erythropoietin receptor (EPO-R) were isolated from a pXM expression library made from uninduced murine erythroleukemia (MEL) cells. The clones were identified by screening COS cell transfectants for binding and uptake of radioiodinated recombinant human erythropoietin. As inferred from the cDNA sequence, the murine erythropoietin receptor is a 507 amino acid polypeptide with a single membrane-spanning domain. It shows no similarities to known proteins or nucleic acid sequences in the data bases. Although the MEL cell EPO-R has a single affinity with a dissociation constant of approximately 240 pM, the EPO-R cDNA, expressed in COS cells, generates both a high-affinity (30 pM) and a low-affinity (210 pM) receptor.     

Expression cloning of a receptor for human granulocyte-macrophage colony-stimulating factor.

Two cDNA clones encoding a receptor for human granulocyte-macrophage colony-stimulating factor (hGM-CSF-R) were isolated by expression screening of a library made from human placental mRNA. Pools of recombinant plasmid DNA were electroporated into COS cells which were then screened for their capacity to bind radioiodinated hGM-CSF using a sensitive microscopic autoradiographic approach. The cloned GM-CSF-R precursor is a 400 amino acid polypeptide (Mr 45,000) with a single transmembrane domain, a glycosylated extracellular domain and a short (54 amino acids) intracytoplasmic tail. It does not contain a tyrosine kinase domain nor show homology with members of the immunoglobulin super gene family, but does show some significant sequence homologies with receptors for several other haemopoietic growth factors, including those for interleukin-6, erythropoietin and interleukin-2 (beta-chain) and also to the prolactin receptor. When transfected into COS cells the cloned cDNA directed the expression of a GM-CSF-R showing a single class of affinity (KD = 2(-8) nM) and specificity for human GM-CSF but not interleukin-3. Messenger RNA coding for this receptor was detected in a variety of haemopoietic cells known to display hGM-CSF binding, and cross-linking experiments revealed a similar size for the glycosylated receptors in transfected COS and haemopoietic cells.     

High-level secretion of the extracellular domain of the human growth hormone receptor using a baculovirus system

A chemically synthesized gene (hGHR-ED) coding for the extracellular domain (ED) of the human growth hormone (hGH) receptor (hGHR) was inserted into the genome of Autographa californica nuclear polyhedrosis virus adjacent to the polyhedrin promoter. Spodoptera frugiperda cells infected with the recombinant virus secreted a protein with hGH-binding activity into the medium. The secreted 35-kDa protein was purified to near homogeneity. The purified protein exhibited a high binding affinity (Kd = 0.2-0.3 nM) to hGH. The highest cell production capability was estimated at more than 10-20 micrograms hGHR-ED/ml of culture. The inhibition of the hGHR-ED secretion by treatment with tunicamycin suggests that glycosylation is important for secretion.     

Secretion of phytohemagglutinin by monkey COS cells

The entire coding region of a gene, which encodes a polypeptide of phytohemagglutinin (PHA-L), obtained from a library of genomic DNA of the common bean Phaseolus vulgaris cv. Greensleeves, was introduced into the SV40 expression vector pJC119. Monkey COS1 cells were transfected with the recombinant clone and the synthesis, glycosylation, and transport of PHA-L studied and compared with the normal processes in bean cotyledons. In the bean, phytohemagglutinin is synthesized on the rough endoplasmic reticulum and transported via the Golgi complex to protein bodies, vacuole-like organelles. Phytohemagglutinin was synthesized and glycosylated at the ER and processed in the Golgi apparatus of the transfected COS1 cells. After passing the Golgi apparatus, PHA-L was slowly secreted into the culture medium (half-time of 3-6 h), a result indicating that the signals for targeting proteins beyond the Golgi apparatus in plant cells are different from those in animal cells. PHA, which is stored in protein bodies in the plant cells, is secreted by animal cells. Tunicamycin inhibited both glycosylation and secretion of PHA by the COS1 cells, a finding indicating an essential role of the oligosaccharides for transport of PHA in these cells in contrast to the situation found in bean cotyledons. PHA, secreted into the culture medium, was partially sensitive to endo H, a result indicating the presence of one high-mannose and one complex oligosaccharide chain, a situation identical to that in beans.     

Glycosylation and secretion of human tissue plasminogen activator in recombinant baculovirus-infected insect cells.

Cell lines established from the lepidopteran insect Spodoptera frugiperda (fall armyworm; Sf9) are used routinely as hosts for the expression of foreign proteins by recombinant baculovirus vectors. We have examined the pathway of protein glycosylation and secretion in these cells, using human tissue plasminogen activator (t-PA) as a model. t-PA expressed in Sf9 cells was both N glycosylated and secreted. At least a subset of the N-linked oligosaccharides in extracellular t-PA was resistant to endo-beta-N-acetyl-D-glucosaminidase H, which removes immature, high-mannose-type oligosaccharides. This refutes the general conclusion from previous studies that Sf9 cells cannot process immature N-linked oligosaccharides to an endo-beta-N-acetyl-D-glucosaminidase H-resistant form. A nonglycosylated t-PA precursor was not detected in Sf9 cells, even with very short pulse-labeling times. This suggests that the mammalian signal sequence of t-PA is efficiently recognized in Sf9 cells and that it can mediate rapid translocation across the membrane of the rough endoplasmic reticulum, where cotranslational N glycosylation takes place. However, t-PA was secreted rather slowly, with a half-time of about 1.6 h. Thus, a rate-limiting step(s) in secretion occurs subsequent to translocation and N glycosylation of the t-PA polypeptide. Treatment of Sf9 cells with tunicamycin, but not with inhibitors of oligosaccharide processing, prevented the appearance of t-PA in the extracellular medium. This suggests that N glycosylation per se, but not processing of the N-linked oligosaccharides, is required directly or indirectly in baculovirus-infected Sf9 cells for the secretion of t-PA. Finally, the relative efficiency of secretion decreased dramatically with time of infection, suggesting that the Sf9 host cell secretory pathway is compromised during the later stages of baculovirus infection.     

Site-specific antibodies to human erythropoietin: immunoaffinity purification of urinary and recombinant hormone

The 19-amino acid domain Ala111----Pro129 of human erythropoietin was identified as an accessible surface antigen based on the binding of radio-iodinated and of unmodified hormone to antibodies prepared against a synthetic peptide of homologous sequence. The specificity and affinity of this binding was sufficient to provide for the use of anti-peptide antibodies in the preparation of an immunosorbent for the purification of urinary, and of recombinant human erythropoietin. Immobilization of anti-peptide antibodies using agarose activated either with CNBr or with N-hydroxysuccinimido groups largely inactivated binding sites for erythropoietin. In contrast, antibodies crosslinked to N-acetyl-DL-homocysteine agarose through the hetero-bifunctional reagent succinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate retained their antigen-binding capacity virtually completely and provided a superior immunosorbent for hormone. Urinary erythropoietin with a specific bioactivity of 100 U/A280 was prepared initially by chromatography on phenyl-Sepharose. Subsequent immunoaffinity chromatography resulted in a 350-fold purification with 46.2% recovery yielding erythropoietin with a specific bioactivity of 35,200 U/A280 (44,300 U/mg). Radioiodination of this purified protein and subsequent SDS-polyacrylamide gel electrophoresis indicated that this preparation contained a single major component (Mr 30,000) which co-migrated in gels with unmodified biologically active hormone. Recombinant erythropoietin, which was prepared by the cloning of the human erythropoietin gene and its expression in COS cells using the SV40-derived vector pSV2, was purified by the same scheme. Chromatography on phenyl-Sepharose of medium derived from transfected cells (400 U/ml, 170 U/A280) provided for a 3.6-fold purification of recombinant hormone with an apparent recovery of 122%. This erythropoietin bound to the anti-peptide antibody gel and was purified to a specific bioactivity of 10,370 U/A280 with 55% recovery. The procedure described here for attaching antibodies to a solid support maximizes their antigen-binding capacity and is generally applicable. The development of an anti-peptide immunosorbant for human erythropoietin provides a valuable means for isolating hormone for use in studies of its receptor and its presently unresolved mechanism of action.     

Dilution enhancement of COS cell expression cloning

To search for an efficient expression cloning method, we mixed plasmid pmDATsv, which contains the mouse dopamine transporter (mDAT) cDNA, with a large amount of another plasmid prGlyTsv to mimic the situation of a cDNA library and examined COS cell expression. Both plasmids have an SV40 replication origin and thus will be replicated to high copy numbers in COS cells. After transfecting COS-7 cells with pmDATsv/prGlyTsv mixture at 1/1000 ratio, we could not detect any cells expressing strong mDAT activity. In contrast, when prGlyTsv was replaced by prSERTsk (no SV40 origin) in the transfection mixture, we observed hundreds of cells expressing strong mDAT activity. The results suggested that in many cells low mDAT expression was not due to the lack of pmDATsv plasmid but due to the presence of large numbers of replicable prGlyTsv. Analysis with a mathematical model suggests that diluting cDNA libraries with other plasmids without the SV40 origin should improve the detection of COS cells expressing target cDNAs. We tested this conclusion with pmDATsv/prGlyTsv mixture. When the mixture at 1/1000 ratio was diluted with prSERTsk and used for transfection, we could now easily detect cells expressing strong mDAT activity.     

Expression of prostaglandin endoperoxide synthase-1 in a baculovirus system.

A cDNA coding for ovine prostaglandin endoperoxide (PGH) synthase-1 was used to construct a recombinant baculovirus which was expressed in Spodoptera frugiperda (Sf9) insect cells. Two proteins reactive with anti-PGH synthase antibody were produced. A larger protein (Mr = 72,000) coelectrophoresed with native enzyme; a smaller, more abundant protein (Mr = 66,000) was unglycosylated enzyme. About 90% of both the immunoreactivity and the cyclooxygenase activity were present in a low speed (10(5) g x min) pellet; variable but low peroxidase activities were observed in this fraction. The specific cyclooxygenase activity of solubilized PGH synthase-1 from Sf9 cells was 56 units/mg versus 112 units/mg for the same cDNA expressed in cos-1 cells. The baculovirus-insect cell system is not ideal for generating large amounts of active PGH synthase-1 apparently because of inefficient N-glycosylation.     

The chromosomal gene structure and two mRNAs for human granulocyte colony-stimulating factor.

Two different cDNAs for human granulocyte colony-stimulating factor (G-CSF) were isolated from a cDNA library constructed with mRNA prepared from human squamous carcinoma cells, which produce G-CSF constitutively. The nucleotide sequence analysis of both cDNAs indicated that two polypeptides coded by these cDNAs are different at one position where three amino acids are deleted/inserted. When the two cDNAs were introduced into monkey COS cells under the SV40 early promoter, both of them produced proteins having authentic G-CSF activity and some difference in the specific activity was suggested. A human gene library was then screened with the G-CSF cDNA and the DNA fragment containing the G-CSF chromosomal gene was characterized by the nucleotide sequence analysis. The human G-CSF gene is interrupted by four introns and a comparison of the structures of the two G-CSF cDNAs with that of the chromosomal gene indicated that the two mRNAs are generated by alternative use of two 5' splice donor sequences in the second intron of the G-CSF gene. When the G-CSF chromosomal gene was expressed in monkey COS cells by using the SV40 enhancer two mRNAs were detected by S1 mapping analysis.     

In vitro expression and site-specific mutagenesis of the cloned human lipoprotein lipase gene. Potential N-linked glycosylation site asparagine 43 is important for both enzyme activity and secretion

Detailed structure-function information about human lipoprotein lipase (LPL) is unavailable because it is difficult to purify large amounts of the enzyme for study. To circumvent this problem, we constructed an in vitro LPL expression vector. Human LPL cDNA was cloned and inserted into the expression vector p91023(B). After transfection of COS M-6 cells with the human LPL cDNA construct, LPL enzyme activity was detected in cell extracts and culture medium. Purified human apolipoprotein C-II caused a 5-fold stimulation of the recombinant human LPL expressed in vitro. Using site-specific mutagenesis, Ala residues were substituted for Asn residues at two potential N-linked glycosylation sites (positions 43 and 359) and at a third unrelated Asn (position 257) in the LPL cDNA. RNA blot analysis demonstrated the presence of a single mRNA species in COS cells transfected with wild-type and mutant LPL expression vectors. Intracellular and secreted LPL activity was absent in the construct containing an Ala for Asn mutation at position 43, whereas the same substitutions at positions 257 and 359 did not appreciably affect activity. LPL activity was also absent in another construct containing a Gln for Asn mutation at position 43. Quantitation of LPL protein mass concomitant with measurement of enzyme activity showed that substitution of Ala or Gln for Asn at position 43 resulted in the production of an enzymatically inactive protein which accumulated intracellularly but was not secreted into the culture medium. Our report represents an initial documentation of the expression of cloned human LPL in vitro and of the importance of Asn-43 for both enzyme activity and secretion.     

Analysis of the glycosylation and phosphorylation of the lysosomal enzyme, beta-hexosaminidase B, by site-directed mutagenesis

Lysosomal enzymes require a mannose 6-phosphate recognition marker, constructed on asparagine-linked oligosaccharide chains, for targeting to lysosomes. We have identified the glycosylation sites of human beta-hexosaminidase B and have determined the influence of individual oligosaccharides on the phosphorylation, lysosomal targeting, and catalytic activity of the enzyme. The five potential glycosylation sites of the hexosaminidase beta-chain were modified individually by site-directed mutagenesis, and the constructs were expressed in COS 1 cells. By this analysis, we determined that four of the five potential sites were glycosylated. Two of the four oligosaccharides were preferentially phosphorylated. The absence of these two preferentially phosphorylated oligosaccharides resulted in greatly reduced amounts of the lysosomal form of the enzyme with increased secretion into the medium. The catalytic activity of beta-hexosaminidase B was not significantly altered by the absence of individual oligosaccharides suggesting the folding and assembly of the enzyme was not disrupted.     

Isolation of a cDNA encoding the human CD38 (T10) molecule, a cell surface glycoprotein with an unusual discontinuous pattern of expression during lymphocyte differentiation

A cDNA clone encoding the human lymphocyte differentiation Ag CD38 was isolated from a mixture of four different lymphocyte CDNA libraries expressed transiently in COS cells and screened by panning with mAb. Transfected COS cells expressed a surface protein of Mr 46,000 that was similar to the native CD38 molecule expressed on the B cell line Daudi and the T cell leukemia HPB-ALL and which was recognized by each of the CD38 specific mAb HIT-2, T16, T168, HB7, 5D2, ICO-18, and ICO-20. The CD38 cDNA sequence predicts an unusual 30-kDa polypeptide with a short N-terminal cytoplasmic tail, and a carboxyl-terminal extracellular domain carrying the four potential N-linked glycosylation sites. The absence of significant homology with other known surface Ag including members of the Ig superfamily ruled out the possibility that CD38 was the human homologue of the murine Qa2 molecule as has been suggested previously. PvuII digests of human genomic DNA revealed a polymorphism linked to the CD38 gene.     

Rat liver beta-glucuronidase. cDNA cloning, sequence comparisons and expression of a chimeric protein in COS cells.

A cDNA for rat liver beta-glucuronidase was isolated, its sequence determined and its expression after transfection into COS cells studied. The deduced amino acid sequence of the rat liver clone showed 77% homology with that from the cDNA for human placental beta-glucuronidase and 47% homology with that deduced from the cDNA for Escherichia coli beta-glucuronidase. Several differences were found between the cDNA from rat liver and that previously reported from rat preputial gland. Only one change leads to an amino acid difference in the mature enzyme. A chimeric clone was constructed by using a fragment encoding the first 18 amino acid residues of the signal sequence from the human placental cDNA clone and a fragment from the rat clone encoding four amino acid residues of the signal sequence, all 626 amino acid residues of the mature rat enzyme, and all of the 3' untranslated region. After transfection into COS cells the chimeric clone expressed beta-glucuronidase activity that was specifically immunoprecipitated by antibody to rat beta-glucuronidase. The Mr value of 76,000 of the expressed gene product was characteristic of the glycosylated rat enzyme. It was proteolytically processed in COS cells to Mr 75,000 6 h after metabolic labelling. At least 50% of the expressed enzyme was secreted at 60 h post-transfection, but the secreted enzyme did not undergo proteolytic processing. These results provide evidence that the partial cDNA isolated from a rat liver library contains the complete coding sequence for the mature rat liver enzyme and that the chimeric signal sequence allows normal biosynthesis and processing of the transfected rat liver enzyme in COS cells.     

Identification of bluetongue virus VP6 protein as a nucleic acid-binding protein and the localization of VP6 in virus-infected vertebrate cells.

Recently the insect baculovirus Autographa californica nuclear polyhedrosis virus (AcNPV) has been effectively adapted as a highly efficient vector in insect cells for the expression of various genes. A cDNA sequence of RNA segment 9 of bluetongue virus serotype 10 (BTV-10, an orbivirus member of the Reoviridae family) encoding a minor core protein (VP6) has been inserted into the BamHI site of the pAcYM1 transfer vector derived from AcNPV. Spodoptera frugiperda cells were cotransfected with the derived vector in the presence of authentic AcNPV DNA to produce recombinant viruses. These synthesized significant amounts of a protein (representing ca. 50% of the stained cellular protein) similar in size and antigenicity to the authentic BTV VP6. The expressed protein was identified as a nucleic acid-binding protein by using an RNA overlay-protein blot assay. A polyclonal anti-VP6 serum prepared by using the expressed VP6 protein has been used in an immunogold procedure to locate VP6 in BTV-infected mammalian cells. Gold was found to be associated with the matrix of virus inclusion bodies (VIB), with viruslike particles in the VIB, as well as with mature virion particles that were in close proximity to the VIB or were released from cells and adsorbed to cell surfaces. The recombinant virus antigen has also been used to identify antibodies to different BTV serotypes in infected sheep sera, indicating the potential of the expressed protein as a group-reactive antigen for the diagnosis of BTV infections.