AU650893B2 - O-glycosylated alpha-2 interferon - Google Patents

Abstract

The objects of the invention are O-glycosylated IFN-alpha, a process for producing the same and the use of the O-glycosylated proteins as medicinal drugs.

Description

OPT DATE 04/02/92 APPLN. ID 82082 91 AOJP DATE 12/03/92 PCT NUMBER PCT/EP91/01266 I NTERi,-q. IkJINP4r%"" _ER DIE INTERNATIONALE ZUSAMMENARBEIT AUF DEM GEBIET DES PATENTWESENS (PCT) (51) Internationale Patcntkiassiikation 5 (11) Internationale Veriiffentlichungsnumnmer: WVO 92/01055 C12N 15/21, C12P 21/02, 21/08 Al (43) Internationales A61K 37/66 Verkiffentlichungsdatum:. 23. Januar 1992 (23,01.92) (21) Internationales Aktenzeichen: PCT/EP9 1/01266 (74) Gemeinsamer Vertreter: BOEH RINGER INGELHE1 M INTERNATIONAL GMBH; A Patente, Postfach 200, (22) Internationales Anmeldedatum: 6. Ju li 1991 (06.07.9 1) D-6507 Ingelheimn am Rhein (DE).

Prioritfitsdaten: (81) Bestirnmungsstaaten: AT (europiiisches Patent), AU, BE P 40 21 917.8 10, Juli 1990 (10.07.90) DE (eu~ropflisehes Patent), CA, CH (europiiisches Patent),' P 40 35 877.1 12. November 1990 (12.11.90) DE CS, DE (europiiisches Patent), DK (europaisches Patent), ES (europilisehes Patent), Fl, FR (europdisches (71) Anmelder (ffir alle Bestimmungssaalen ausser US): BOEH- Patent), GB (europiiisches Patent), GR (europiiisches RINGER INGELH-EIM INTERNATIONAL GMBH Patent), HU, IT (europaiisches Patent), JP, KR, LU (eu- [DE/DE]; Postfach 200, D-6507 Ingelheim am Rhein rop~isches Patent), NL (europ~isches Patent), NO, PL, +tSE (europliisches Patent), SU, US.

(72) Erfinder; und Veroffentlicht Erinder/Anmelder (nur fir US) ADOLF, G~nther [AT/ Mit internationaetn Recherclienbericht.

AT]; Stiftgasse 15-17/10, A-1070 Wien HIMM- Vor Ablauf derjFar Andeningen der Ansprfiche zugelasse- LER, Adolf [AT/AT]; Ftirst Liechtensteinstr. 2/3, A- uien Fris. Veridffentliclhung wird wiederhlt falls Anderun- 1236 Wien AHORN, Horst, Johann [AT/AT]; Ei- genz einreffen.

senstiidterstr. 3/1, A-2484 Weigelsdorf KALS- NER, Inge [AT/All; Geusaugasse 51/20, A-1030 Wien MAURER-FOGY, Ingrid [AT/AT]; Lindauergas- f* se 35, A-1238 Wien

(AT).

(54) Title: 0-GLYCOSYLATED IFN-ALPHA (54) Bezeichnung: 0-GLYCOSYLIERTES IFN-ALPHA (5)Astat1 5 10 (57)AbstactCYS-ASP-LEU-PRO-GLt4-THR-HIS-SER- LEU-GLY-SER-ARG-ARG-THR-LEU- The objects of the invention are 0- 20 25 glycosylated IFN-alpha, a process for prod- METLEULEUAA-GLN-METARGAGILESERLEUPESERCYS-LEUucing the same and the use of the 0-glycosy- 35 40 lated proteins as medicinal drugs. LY-S-R-R-S-H-L-HEPOGNGUGUPEGYAN (57) Zusammenfassung 50 55 GLN-PHE-GLN- LYS-At.A-GLUTHR-ILE-PRO-VAL-LEU-HIS-GLU-MET-ILE- Gegenstand der vorliegenden Erfin- 6 07 dun it -glcoylerts FNa, erahenGLN-GLN-ILE-PHE-ASN-LEU-PHE-SER-THR-LYS-ASP-SER-SER-AtA-ALAzu dessen Herstellung sowie die Verwendung der 0-glycosylierten Proteine als Arz- 80 85 9o neimittel. TRP-ASP.-GLU-THFR-LEU-LEU-ASP-LYS-PHE-TYR-THR-GLU-LEU-TYR-GLN- 100 105 GLN- LEU-ASN-ASP-LEU-GLTJ-ALA-CYS-VAL- ILE-GLN-GLY-VAL-GLY-VAL- 110 115 120 THR-GLU-THR-PRO-LEU-MET-LYS-GLtI-ASP-SER- ILE-LEU-ALA-VAL-ARG 125 130 135 LYS-TYR-PHE-GLN-ARG- ILE-THR-LEU-TYR-LtEU- LYS-GLU- LYS-LYS-TYR- 140 145 150 SER-PRO-CYS-ALA-TRP-GLU-VAL-VAL-ARG-ALA-GLI- ILE-MET-ARG-SER.- 155 160 .165 PHE-SER-LEU-SER-THR-ASN- LEU-GLN-GLU-SER- LEU-ARG-SER-LYS-GLJ a2c Siehe Riickseite 1 S012777J.35 Case 12/114,121-PCT/EP.

(3722v) O-glycosylated IFN-alpha The present invention relates to O-glycosylated alphainterferons, preferably an interferon alpha, substantially having the biological and/or immunological properties of an IFN-a2, processes for the preparation thereof and the use of the O-glycosylated interferons as pharmaceutical compositions.

_Since the discovery of interferons more than thirty years ago, their biological properties as mediators of intercellular communication have been thoroughly investigated. Originally, the various types were named after the particular cell in which they were formed leukocyte-IFN, fibroblast-IFN). As the knowledge of their structure increased, a new nomenclature was introduced. At present, four different types of interferons are distinguished (IFN-a, IFN-p, IFN-y and IFN-u), IFN-a, IFN-P and IFN-u being classified as socalled "Class 1 interferons", as they have similar structures and properties.

IFN-y is formed by lymphocytes stimulated by antigens or mitogenic substances. The amino acid sequence, which has no homology with the Class 1 interferons, contains two potential N-glycosylation sites.

IFN-a, IFN-/ and IFN-u are synthesised by various cells as a reaction to virus infection or after induction with double-stranded RNA.

IFN-a is actually a whole group of proteins. Up till now, at least 14 functional genes have been discovered which code for different types of IFN-a. These proteins

IZ.

2are closely related and usually have about 80-90% homology in their amino acid sequence. With the exception of IFN-al4, no N-glycosylation site (ASN-X-SER/THR) is present in any of the other IFN-a amino acid sequences. Thus, N-glycosylation can be ruled out in every case, with the exception of (IFN-al4), but O-glycosylation of IFN-a has been discussed (Labdon et al., Arch. Biochem; Biophys. 232, 422-426 (1984)).

Many naturally occurring proteins are modified posttranslationally, with glycosylation being one of the commonest modifications. Glycoproteins occur in membrane-bound form or in soluble form both in the intra- and extracellular matrix. There are differing views on the function of glycosylation. What is certain is that glycanes can protect the proteins from proteolytic degradation or that in many cases they are responsible for cell-to-cell interactions. In addition, they affect protein folding and contribute to the stability of the configuration of the molecule. The solubility of the proteins is also the basis of the influence of the carbohydrate chains.

A distinction is drawn between N- and O-glycosylated proteins. N-glycanes are transferred solely to the ASN of the triplet -ASN-X-SER/THR-, wherein X may be any amino acid with the exception of PRO or GLU. However, this requirement of the structure of the protein may be only one of several, since not all potential glycosylation sites are occupied by a carbohydrate. For O-glycanes there are no precisely defined structural features. There are certainly indications that Oglycanes are preferably synthesised in PRO-, SER- and THR-rich regions. This leads.one to assume that it is rather the steric accessibility of the glycosylation site than any specific amino acid sequence which is S* i c\ v -3crucial to O-glycosylation. Influences of glycosylation on pharmacokinetics and on the immunological properties of the protein cannot be ruled out. Thus, it was briefly reported (Gibbon et al., Lancet, 335,,434-437 (1990) that 4 out of 16 patients who had been treated with recombinant human GM-CSF (granulocyte-macrophagecolony stimulating factor) which had been produced in yeast developed antibodies against this-protein. It was found that these antibodies reacted with epitopes which are present in protected form in endogenous GM-CSF as a result of O-glycosylation but are freely accessible in the recombinant factor.

Hitherto, it has neither been possible to detect carbohydrate contents in IFN-a2 nor has it been possible to isolate O-glycosylated IFN-a. Various preparations of natural IFN-a and of recombinant IFN-a2 are currently used as drugs against viral and cancer diseases. Since this IFN-a2 is produced in E. coli and cannot therefore be glycosylated, the carbohydrate content does not appear to be crucial to the in vivo biological activity.

Recently, however, there have been increasing numbers of reports that patients treated over lengthy periods with recombinant IFN-a2 produced in E. coli developed antibodies against it Figlin Itri, Semin.

Haematol. 25. 9-15 (1988)).

The aim of the invention was to provide a new IFN-a2.

This objective was achieved by the insertion of the DNA sequence coding for IFN-a2 into a special expression vector with which cells of multicellular organisms were transfected. After cultivation of these cells modified in this way, surprisingly, IFN-a2-like proteins were obtained which differ significantly from the known recombinant IFN-a2 in their molecular weight.

4 For preparing the new interferons according to the invention it is possible to use cultures of cells of multicellular organisms, particularly cultures of vertebrate cells or insect cells. Examples of mammalian cell lines include VERO-cells, HeLa-cells, CHO-cells, WI38-cells, BKH-cells, COS-7-cells, MDCK-cells or murine myeloma cells. Expression vectors for these cells contain, if necessary, a replication site, a promoter, and if necessary an RNA-splicing site, a polyadenylation site and transcriptional termination sequences. The control functions of such expression vectors normally come from viral material. The usual promoters are obtained from polyoma, adenovirus 2, Simian virus and preferably cytomegalovirus (CMV). The replication site required can either be provided by means of a suitable vector construction, e.g. the replication site from SV40, polyoma, adeno, VSV or PBV or it may be provided by the chromosomal replication mechanisms of the host cell. If the vector is integrated in the host cell chromosome this latter measure is sufficient.

According to the invention, it is preferred to use expression vectors which are newly constructed from parts of plasmids. These expression vectors according to the invention have a multicloning site for the targeted insertion of heterologous DNA sequences and can preferably be replicated in high copy numbers in E. coli by means of ampicillin resistance. In order to permit the expression of heterologous genes in mammalian cells, the expression plasmids according to the invention preferably contains the cytomegalovirus (CMV) promoter/enhancer Boshart et al., Cell 41, (1985), 521-530). In order to permit autonomous replication of the expression plasmids according to the invention in high copy numbers and thereby allow high rates in transient expression in suitable cell lines such as for I;.2rT\ 5 example in COS-7 or in cell line 293 transformed with adenovirus (ATCC CRL 1573), the SV40 replication origin was used. In order to prepare permanently transformed cell lines and obtain subsequent amplification of the expression cassette by means of methotrexate a modified hamster minigene was used (promoter with coding region and the first intron) for dihydrofolate reductase (DHFR) as selection marker. In order to permit the preparation of single-stranded plasmid-DNA after superinfection of the transformed bacteria with a helper phage, e.g. with R408 or M13K07, for easier sequencing and mutagenesis of the plasmid DNA, embodiments of the plasmids according to the invention preferably contained the intergenic region of M13. If in another preferred embodiment the T7 promoter is placed before the multicloning site, this enables RNA transcripts to be produced in vitro.

The expression plasmids preferred according to the invention are pAD-CMV13, pAD-CMV15 and in particular pAD-CMV19. The preparation thereof is described in detail in Example 1.

In order to achieve improved expression and facilitate controlled cloning of the cDNA coding for IFN-a2, the cDNA coding for IFN-a2 was modified according to the invention by PCR in the 5'-non-coding region by replacing the sequence of this region by the sequence of the 5'-non-coding region of human /-globin mRNA (Lawn et. al., Cell 21, (1980), 647-651). At the same time, restriction enzyme cutting sites were inserted at both ends of the cDNA to facilitate controlled cloning.

Surprisingly, this change results in a significant increase in expression.

This modified cDNA for IFN-a2.was inserted into an expression plasmid according to the invention cut with the corresponding restriction enzyme, preferably the J 6 -6plasmid pAD-CMV19. Suitable mammalian cells were transfected with the resulting expression plasmids for IFN-a2 and were then cultivated in a suitable culture medium. The culture supernatant of the mammalian cells was purified in known manner under mild conditions. It is preferable to carry out purification using affinity chromatography with the aid of monoclonal antibodies against IFN-a2. Preferred monoclonal antibodies are EBI-1 or EBI-10 or the equivalent thereof. The preparation of these highly specific antibodies has been described (Adolf G.R. J. Gen. Virol. 68, 1669-1676 (1987); Adolf et al. J. Cell. Physiol. suppl. 2, 61-68 (1982)). The methods used have also been described (Secher and Burke, Nature 285, 446-450 (1980); Adolf et al., J. Biol. Chem. 265, 9290-9295 (1990); Adolf et al., Biochem. J. 276, 511-518 (1991)). It is particularly advantageous to carry out purification according to EPA 0 203 382, which avoids breaking up the cells. In order to characterise recombinant IFN-a2 prepared in mammalian cells, Reverse Phase HPLC was used. The N-terminus and C-terminus were analysed. Recombinant IFN-a2c prepared in E. coli was used for a comparison in both cases. By means of SDS-gel electrophoretic tests it was established that the recombinant IFN-a2 prepared in mammalian cells had a higher molecular weight than the IFN-a2 prepared in E. coli. After treatment of both recombinant interferons with NaOH, the molecular weight of the recombinant IFN-a2 prepared in mammalian cells was reduced to the molecular weight of the IFN-a2 prepared in E. coli. Recombinant IFN-a2 expressed in mammalian cells must therefore be glycosylated. During identification of the glycopeptides using peptide mapping and sequence analysis it was found that the threonine located at position 106 (THR-106) carries the glycosylation. When the results are compared with those obtained with natural IFN-a2 from virus-stimulated leukocytes (see below) it was found that both the 7 glycosylation site and the oligosaccharide content are largely identical.

The objective of the invention was, however, also achieved by means of a process of purification which does not have any stages which alter or eliminate any possible substitutions of the IFN-a2. The method of purification according to the invention-used highly specific monoclonal antibodies, whilst throughout the entire purification process alkaline conditions having a pH of above 8.0 were carefully avoided.

Natural human IFN-a2 was isolated using a highly specific monoclonal antibody obtained from leukocyte interferon. Two successive purification steps using an immunoaffinity column resulted in a degree of purity of the protein of Sequence analysis resulted in the expected N-terminal sequence, with CYS as the first amino acid being shown up only indirectly.

Up till now, three variants of IFN-a2 have been known, which differ in the amino acids at positions 23 and 34, namely: IFN-a2a with 2 LYS and 3 HIS (formerly known as Le IFA; Goeddel et al., Nature, 290, 20-26 (1981)), IFN-a2b with 3 ARG and 34 HIS (Streuli et al., Science, 209, 1343-1347 (1980)) and IFN-a2c with 23 ARG and 3

ARG

(formerly known as IFN-a2"Arg"; Dworkin-Rastl et al., J.

Interferon Res. 2, 575-585 (1982); Bodo Maurer-Fogy, The Biology of the Interferon System 1985 (Stewart II, W.E. Schellehus H. Eds.) 59-64 (1986). In the isolated interferon, only ARG could be detected at positibn 23, which rules out the presence of IFN-a2a.

The amino acid at position 34 was clearly histidine, which meant that the isolated interferon was IFN-a2b.

However, it is also possible to obtain the variants IFN-a2a and IFN-a2c, depending on the cell material used as starting material. It is known that in Namalwa cells ~s"r) ~s 2 cy 8 both IFN-a2b and IFN-a2c can be found. The recombinant interferon from E. coli used as comparison substance was IFN-a2c.

RP-HPLC analysis of the purified natural IFN-a2 showed that the preparation contained two peaks which both eluted earlier from the column than the recombinant E. coli IFN-a2c. SDS-PAGE also showed a strong heterogeneity in the apparent molecular mass of natural IFN-a2. All the proteins detected in natural IFN-a2 had a substantially higher apparent molecular mass than recombinant IFN-a2c from E. coli. None of the IFN-a species described hitherto, with the exception of IFN-al4, contained an N-glycosylation site (-ASN-X-THR/SER-). Thus, N-glycosylation can also be ruled out for IFN-a2. Such structural features are not known for O-glycanes. It is therefore impossible to rule out the present IFN-a2 being O-glycosylated.

Since O-glycanes can be cleaved from the protein even under slightly alkaline conditions, both peak fractions were subjected to slightly alkaline conditions. This reaction resulted in both cases in a reduction in the apparent molecular mass to that of the recombinant IFN-a2c from E. coli; a clear indication of Oglycosylation.

Experiments with neuraminidase and 0-glycanase produced for one peak (peak 2) (see Fig. 14) a reduction in the apparent molecular mass to that of E. coli IFN-a2c and thus confirmed the O-glycosylation. The results of this sequential breakdown of the glycane with neuraminidase and O-glycanase showed that the heterogeneity of peak 2 was based on the different content of N-acetylneuramino acid (NeuAc sialic acid). The three bands (Fig. trace 4) represented the di- or monosialylated (Mr 21,000 and 20,000, respectively) and the non- S 2;;70'P salialylated (Mr 19,000) form of the natural IFN-a2.

The lightest form of I.FN-a2 was capable of being broken down by reaction with O-glycanase alone. Since O- I glycanase only cleaves the unsubstituted disaccharide i Gal(31-3)GalNAc, the reaction must be regarded as evidence for the fact that in addition to the two salialylated forms there is also a asialo variant of IFN-a2.

The apparent molecular mass of peak 1, on the other hand, could not be reduced by enzyme reactions.

Incubation with neuraminidase did not lead, as expected, i to a reduction in the apparent molecular mass. The disaccharide core must therefore be substituted in some other way than with NeuAc and could not therefore be cleaved by O-glycanase.

By comparing the peptide maps after trypsin cleaving of natural and recombinant IFN-a2 expressed in E. coli the glycopeptides were able to be identified from the peaks 1 and 2. Sequencing of these glycopeptides produced 106 THR as the glycosylation site.

References to the structure of the oligosaccharides of natural IFN-a2 were provided not only by the enzyme reactions but also by mass spectrometric investigations of the glycopeptides. The interpretation of the mass spectra together with the results of the SDS-PAGE showed Sthat natural IFN-a2 contains at least four different glycane structures: in peak 2 the neutral disaccharide Gal(pl-3)GalNAc, the structure of which can be assumed with great certainty owing to the high specificity of Oglycanase, as well as the mono- and disialylated variant; in peak 1 a neutral oligcsaccharide consisting of two hexose and two N-acetylhexosamine units. The following may be proposed as the structure of this tetrasaccharide, analogously to frequently occurring 0- 10 glycanes which have already been described: Gal-(Gal-GlcNAc-)GalNAc.

Thanks to the present invention it has surprisingly been possible to prepare O-glycosylated IFN-a2 in a highly pure form for the first time. This interferon is O- K glycc 'lated at the amino acid threonine at position 106 106 THR). The oligosaccharides which may be contained at this position are the neutral disaccharide Gal(pl-3)GalNAc, the mono- and disialylated variants thereof and a neutral tetrasaccharide Gal-(Gal- GlcNAc-)GalNAc.

This O-glycosylated IFN-a2 may be formulated in a manner known per se, analogously to the recombinant IFN-a2 expressed in E. coli, and may be used for treatment in all the indications known for IFN-a.

The proteins according to the invention may be used for treating viral infections and malignant diseases, in the form of pharmaceutical preparations containing an effective quantity of the IFN, optionally together with a significant amount of an organic or inorganic, solid Sor liquid, pharmaceutically acceptable carrier.

The preferred pharmaceutical preparations are those intended for parenteral, e.g. intramuscular, subcutaneous or intravenous use in humans. Such preparations may be isotonic aqueous solutions or suspensions which contain the proteins according to the invention, optionally together with a carrier and, if desired, adjuvants such as stabilisers, emulsifiers, solubilising agents, salts for regulating the pH and osmotic pressure, preservatives and/or wetting agents.

The pharmaceutical preparations may be prepared using methods known per se, e.g. by a method in which the proteins according to the invention and the T N \cvi oZ 11 pharmaceutically acceptable carriers and adjuvants are mixed together, lyophilised if desired and dissolved before use.

The dosage of the pharmaceutical preparations depends on the disease being treated, the body weight, age and individual condition of the patient in the estimation of the doctor treating him and the method of administration.

The present invention, therefore, provides for the first time an agent containing O-glycosylated interferon-a2, which is suitable, on the basis of the antiviral and antineoplastic properties of IFN-a2, for the treatment of viral diseases and tumours, inter alia.

The Examples which follow are intended to illustrate the invention without restricting it.

Legend relating to the Figures Fig. 1: Construction of the plasmid Fig. 2: Construction of the plasmid pAD-CMV1OA Fig. 3: Construction of the plasmid Fig. 4: Construction of the plasmids pAD-CMV13 and pAD-CMV19 Fig. 5: Construction of the expression plasmid pAD19B-IFN Fig. 6: HindIII/XBaI insert.of the expression plasmid pAD19B-IFN i -i 12 Fig. 7: DNA sequence of the plasmid pAD-CMV19 Fig. Construction of the plasmid Fig. 9: Construction of the plasmid pSV2gptDHFRMut2 Fig. 10: Construction of the plasmids pAD-CMV1 and pAD-CMV2 Fig. 11: DNA sequence of the plasmid pAD-CMV1 Fig. 12: Monoclonal antibody affinity chromatography of the human leukocyte interferon Fig. 13: ELISA for human IFN-a: reference preparation of recombinant human IFN-a2c; leukocyte interferon (starting material) throughflow a fraction of eluate A; a fraction of eluate B Fig. 14: RP-HPLC of natural IFN-a2 and E. coli IFN-a2c (a) Fig. 15: Amino acid sequence of IFN-a2c Fig. 16: SDS-PAGE of natural IFN-a2 before and after reaction with neuraminidase and O-glycanase.

peak 1, untreated; peak 1, after reaction with neuraminidase; peak 1, after reaction with neuraminidase and O-glycanase; peak 2, untreated; peak 2, after reaction with neuraminidase; peak 2, after reaction with neuraminidase and O-glycanase; E. coli IFN-a2c Fig. 17: SDS-PAGE of natural IFN-a2 (peak 2 from Fig. 14b) before and after reaction I C I 13 with O-glycanase Fig. 18: Fig. 19: Fig. 20: Fig. 21: Fig. 22: SDS-PAGE of natural IFN-a2 (peak 1 and 2) and E. coli IFN-a2c after incubation with 0.1 M NaOH. E. coli IFN-a2; peak 1; peak 2; untreated comparison samples of peak 1 and of peak 2 were also recorded.

Comparative peptide map of E. coli IFN-a2c and natural IFN-a2. peak 1 from Fig. 14b; (2) peak 2 from Fig. 14b; these peaks come from unglycosylated peptides the retention time of which was always the same.

SDS-PAGE of natural IFNa2 and E. coli IFNa2c.

molecular weight marker; E. coli IFNa2c; natural IFN-a2, peak 1 from Fig. 14b; natural IFN-a2, peak 2 from Fig. 14b; staining: Coomassie blue.

Reverse phase HPLC (RP-HPLC) of the CHO-IFN-a2c and of E. coli IFN-a2c (b) Comparative peptide maps of peak 1 and peak 2 from CHO-IFN-a2c and of E. coli IFN-a2c (c) Fig. 23: SDS gel electrophoresis (SDS-PAGE) of CHO-IFN-a2c and E. coli IFN-a2c. Traces 1 and 8: molecular weight markers (scale in kD); traces 2-4: non-reducing conditions, traces 5-7: reducing conditions; traces 2 and 5: peak 1 from CHO-IFN-a2c; traces 3 and 6: peak 2 from CHO-IFN-a2c; .1 l.T.

14 traces 4 and 7: E. coli IFN-a2c; upper gel: all IFN traces each with 4 jg; lower gel: all IFN traces each with 1 pg; staining: Coomassie blue SDS-gel electrophoresis (SDS-PAGE) of CHO-IFN-a2c and E. coli IFN-a2c before and after incubation with 0.1 M NaOH.

Fig. 24: Traces 1 and 8: molecular weight marker (scale in kD); traces 2, 4, 6: untreated samples, traces 3, 5, 7: samples incubated with 0.1 M NaOH; traces 2, 3: E. coli IFN-a2c, trace 4, 5: peak 1 from CHO-IFN-a2c, trace 6, 7: peak 2 from CHO-IFN-a2c; about 1.5 jg were applied to all the IFN traces under reducing conditions staining: Coomassie blue Example 1 Construction of the expression plasmids pAD-CMV13, and pAD-CMV19 From parts of expression plasmids (pCDM8, Seed Aruffo, Proc. Natl.Acad.Sci.USA 84 (19.87) 8573-8577; B. Seed, Nature 329 (1987) 840-842); Invitrogen, Inc., San Diego, CA; pSV2gptDHFR20, EP-Al 0321842) and the plasmid

'NI

C'

ii 15 pBluescript KS- (Short et al., Nucleic Acids Res., 11 (1988) 5521-5540; Stratagene, La Jolla, CA)) new plasmids were constructed having a multicloning site for the controlled insertion of heterologous DNA sequences and can be replicated in a high copy number in E. coli by means of ampicillin resistance. The intergenic region of M13 makes it possible to produce singlestranded plasmid DNA after superinfection of the transformed bacteria with a helper phage R408 or M13K07), for easier sequencing and mutagenesis of the plasmid DNA. The T7 promoter which nrecedes the multicloning site makes it possible to produce RNA transcripts in vitro. In mammalian cells the expression of heterologous genes takes place, driven by the cytomegalovirus (CMV) promoter/enhancer Boshart et al., Cell 41 (1985) 521-530). The SV40 replication origin makes it possible to achieve autonomous replication of the expression plasmid in high copy numbers and hence with high rates in transient expression in suitable cell lines (e.g SV40 transformed cells such as COS-7, adenovirus transformed cell line 293 (ATCC CRL1573)). To prepare permanently transformed cell lines and subsequently amplify the expression cassette using methotrexate a modified hamster minigene is used (promoter with coding region and the first intron) for dihydrofolate reductase (DHFR) as the selection marker.

Preparation of the vector and promoter parts by polymerase chain reaction (PCR).

The plasmid pBluescript KS- was linearised with HindIII and 100 ng DNA in a 100 gl PCR (Saiki et al., Science 239 (1988) 487-491) mixture (reaction medium: 50 mM KC1, mM Tris-Cl pH 8.3, 1.5 mM MgCl 2 0.01% gelatine, 0.2 mM of each of the four deoxynucleoside triphosphates 16 (dATP, dGTP, dCTP, dTTP), 2.5 units of Taq polymerase per 100 il). The primers used were 50 pmol each of the synthetic oligonucleotides EBI-1786 TGGCGAATGGG-3') and EBI-2134 TGCGTTGCTGGCGTTTTTCC-3'). After 5 minutes denaturing at 94"C PCR was carried out over 10 cycles (cycle conditions: 40 seconds at 94"C, 45 sec at 55"C, 5 min at 72*C, Perkin Elmer Cetus Thermal Cycler). The oligonucleotides flank the intergenic region of M13 or the replication origin (ori) with the intervening gene for the /-lactamase. At the same time an XhoI- and a PvuII cutting site is produced at the end of the ori and an EcoRI cutting site at the other end. The reaction mixture was freed from protein by extraction with phenol-chloroform and the DNA was precipitated with ethanol. The DNA obtained was cut with XhoI and EcoRI and, after electrophoresis in an agarose gel, a 2.3 kb fragment was isolated. 50 ng of plasmid pCDM8 linearised with SacII were amplified by PCR using the oligonucleotides EBI-2133 TGATTATTGACTAG-3') and EBI-1734 AGGAATACAGCGG-3') under identical conditions to those described above. The oligonucleotides bind at the beginning of the CMV-promoter/enhancer sequence and produce a SalI cutting site (EBI-2133) or bind to the end of the SV40 poly-adenylation site and produce an EcoRI cutting site (EBI-1734). The PCR product was cut with SalI and EcoRI and a DNA fragment of 1.8 kb was isolated from an agarose gel.

The two subsequently cut PCR products were ligated with T4 DNA-ligase and transformed with E. coli HB101. A plasmid of the desired structure (see Fig. 1) was Sdesignated pCMV+M13.

The SV40 replication origin (SV40 ori) was isolated from the plasmid pSV2gptDHFR20 (EP-A1 0321842). For this 17\ SrI 17 purpose this plasmid was doubly cut with HindIII and PvuII and the DNA ends.were blunted by subsequent treatment with the large fragment of the E. coli DNA polymerase (Klenow enzyme) in the presence of.the four deoxynucleotide triphosphates. A 0.36 kb DNA fragment obtained was isolated from an agarose gel and ligated in the plasmid vector pCMV+M13 linearised with EcoRI. A plasmid obtained after transformation of E. coli HB101, containing the SV40 ori in the same orientation as Plactamase gene and CMV-promoter, was designated pCMV- (Fig. Plasmid pCMV+SV40 was doubly cut with EcoRI and BamHI and the DNA ends were then blunted with Klenow enzyme. The DNA was purified by extraction with phenol chloroform and ethanol precipitation. Some of the DNA was circularised with T4 DNA ligase and a plasmid obtained after transformation of E. coli was designated pAD-CMV10 (Fig. The remainder of the DNA was dephosphorylated by incubating with alkaline phosphatase and the 4.4 kb long vector was isolated from an agarose gel.

1 Plasmid pSV2gptDHFR-Mut2 (see Example 4, Fig. which contains a modified hamster dihydrofolate reductase (DNFR) minigene, from which the restriction enzyme cutting sites for EcoRI, PstI, BglII, BamHI and KpnI were removed by controlled mutagenesis, was doubly cut with EcoRI and PstI and the DNA ends were blunted by minutes' incubation at 11°C with 5 units of T4 DNApolymerase (reaction medium: 50 mM Tris-Cl pH 8.0, 5 mM MgC1 2 5 mM dithiothreitol, 0.1 mM of each of the four deoxynucleotide triphosphate, 50 ug/ml bovine serum albumin). The 2.4 kb long fragment with the mutated DHFR gene was isolated from an agarose gel and ligated with the pCMV+SV40 prepared as described above. A plasmid obtained after transformation of E. coli, in which the DHFR gene was contained in the same orientation as the CMV-promoter, was designated 18 pAD-CMV1OA (Fig. 2).

Starting from the expression plasmid pAD-CMVl (see Example 4, Fig. 10), which contains an intron-sequence between the multicloning site and the polyadenylation signal, numerous variants were produced which differ in their number and position of introns relative to the multicloning site. In pAD-CMV13 (Fig. 4) the SV40 t antigen intron between the multicloning site and the polyadenylation site was deleted; pAD-CMV15 (Fig. 3) contains a synthetic intron between the CMV promoter and the multicloning site and the SV40 t antigen intron between the multicloning site and the polyadenylation signal; pAD-CMV19 (Fig. 4) contains only one intron between the CMV promoter and multicloning site.

Starting from 100 ng of the plasmid pAD-CMVl linearised with HindIII, a 1.26 kb long DNA fragment was amplified with 50 pmol each of the oligonucleotides EBI-2625 (5'-CACTGATCTAGAGATATCTTGTTTATTGCAGCTTATAATGG-3') and EBI-1857 (5'-GGCAAGGGCAGCAGCCGG-3') in 100 gl of PCR mixture (see above) in 10 PCR cycles (40 seconds at 94°C, 45 seconds at 55°C, 90 seconds at 72C). EBI-2625 binds shortly before the SV40 poly-adenylation signal (position 1280 in pAD-CMVl) and contains additional restriction sites for XbaI and EcoRV. EBI-1857 binds to the complementary DNA strand in the first intron of the following DHFR minigene (position 2525 in pAD-CMV1).

The PCR product was freed from protein by extraction with phenol and chloroform and the DNA was precipitated with ethanol. The DNA was doubly cut with Xbal and BglII, a 0.32 kb long DNA fragment was isolated from an agarose gel and ligated into plasmid vector (5.8 kb) pAD-CMV1 doubly cut with the same enzymes. A plasmid of the desired nature obtained after transformation of E. coli HB101 (see Fig. 4) was designated pAD-CMV13.

i 19 The splice-donor sequence following the CMV promoter (M.

Boshart et al., Cell 41 (1985) 521-530) was joined to the splice acceptor site of the first intron of the human P-globin gene (Lawn et al., Cell 21 (1980) 647-651) followed by the multicloning site of plasmid pAD-CMV1 by SOE-PCR (splicing by overlap extension; S.N.

Ho et al., Gene 77 (1989) 51-59). In order to do this, 100 ng of plasmid pGJ7 Jahn et Virology 49 (1984) 363-370) containing the promoter and enhancer sequence of human cytomegalovirus strain AD169 ,Boshart et el., Cell 41 (1985) 521-530) were amplified with pmol each of the oligonucleotides EBI-2133 (see above) and EBI-2586 AAGAGTCTTCTCTATAGGCGGTACTTACCTGACTCTTG-3') in 100 p1 pf PCR reaction mixture over 30 cycles (cycle conditions: seconds at 94°C, 45 seconds at 45°C, 90 seconds at 72'C). The last 24 bases of EBI-2586 fit perfectly to the CMV-sequence (in the antisense orientation) and the preceding bases correspond to the P-globin intron sequence, with 18 bases perfectly fitting the reverse complementary sequence of oligonucleotide EBI-2585 and forming the overlapping DNA sequence for the SOE-PCR.

The PCR products were separated in an agarose gel and a 0.8 kb DNA fragment was isolated (Fig. 100 ng of plasmid pAD-CMV1 were amplified by PCR in the same way with the oligonucleotides EBI-2585 GGTGCTTAACTGGCTTATCG-3') and EBI-2112 (5'-GTCCAATTATGTCACACC-3') and a 0.2 kb DNA fragment was isolated from an agarose gel. EBI-2585 contains the last 45 bases of the P-globin intron and the five succeeded bases as well as 17 bases at the 3'-end, which are able to hybridise perfectly at position 611-627 of the pAD-CMV1 sequence. EBI-2112 binds on the complementary DNA strand at position 743-760 to the pAD-CMV1 sequence. 1/10 of the isolated 0.8 kb DNA fragment and 1/30 of the 0.2 kb DNA fragment were mixed -0 7 ;n v 9 in a new 100 gl PCR mixture (SOE-PCR) and amplified with pMol each of the ol.igonucleotides EBI-2133 and EBI-2112 in 30 PCR cycles (40 seconds at 94°C, seconds at 45"C, 2 minutes at 72*C). The reaction was stopped by extraction with phenol and chloroform and the DNA was precipitated with ethanol. The 5'-ends of the PCR product were phosphorylated with T4 polynucleotide kinase (reaction buffer: 70 mM Tris-Cl pH 7.6, 10 mM MgCl 2 5 mM dithiothreitol, 1 mM ATP) and then cut with XbaI. The DNA was separated in an agarose gel and a fragment 0.98 kb long was isolated. Plasmid pAD-CMVlO was doubly cut with PvuII and XbaI and the vector component without the CMV promoter was isolated from an agarose gel. This plasmid vector was ligated with the 0.97 kb DNA fragment, containing the CMV promoter and enhancer with intron and multicloning site, and transformed with E. coli HB101. Plasmid DNA was prepared from the resulting transformants and the new DNA insert was sequenced with the oligonucleotides EBI-2112, EBI-2586 and EBI-1733 GGTCGACATTGATTATTGACTAG-3') using the dideoxy chain breaking method Sanger et al., Proc. Natl. Acad.

Sci.USA 74 (1977) 5463-5467) using modified T7 DNA polymerase Tabor and C.C. Richardson, Proc. Natl.

Acad. Sci.USA 84 (4767-4771); Sequenase, United States Biochemical Corp.). A plasmid having the expected sequence was designated pAD-CMV15 (Fig. 3).

pAD-CMVlOA was doubly cut with Spel and BglII and the vector component without CMV promoter was isolated from an agarose gel. pAD-CMV15 was doubly cut with Spel and HindIII and a 0.8 kb DNA fragment containing the CMV promoter and the synthetic intron was isolated.

pAD-CMV13 was doubly cut with HindIII and BglII and a 0.36 kb DNA fragment was isolated which contained the multicloning site, the SV40 early poly-adenylation S. signal and part of the hamster-DHFR promoter region.

i _I rfI -21 These three DNA fragments were ligated with T4 DNA ligase and transformed with E. coli HB101. Plasmid DNA was prepared from the resulting transformants and characterised by cutting with various restriction enzymes. A plasmid of the desired structure was designated pAD-CMV19 (Fig. 4, Fig. Example 2 Preparation of a modified cDNA for huIFN-a2c The cDNA of clone 1F7 coding for human IFN-a2c (E.

Dworkin-Rastl et al., J.Interferon Res. 2 (1982) 575-585; E. Dworkin-Rastl et al., Gene 21 (1983) 237-248) was modified by PCR in the region, by exchanging the latter for the sequence of the region of the human P-globin mRNA (Lawn et al., Cell 21 (1980) 647-651). Such a change in the non-coding region brings about a clear increase in expression, possibly by more efficient initiation of the translation. At the same time, restriction enzyme cutting sites were introduced at both ends of the cDNA, to facilitate subsequent controlled cloning of the cDNA in expression plasmids.

100 ng of plasmid 1F7 linearised with EcoRI were amplified with 50 pMol each of the oligonucleotides EBI-2747 TGTGTTCACTAGCAACCTCAAACAGACACCATGGCCTTGACCTTTGCTTTAC-3') and EBI-2744 in 100 Ml of PCR mixture in 20 cycles (40 seconds at 94"C),-45 seconds at 55°C, 90 seconds at 72*C).

EBI-2747 contains, after a HindIII cutting site, the non-coding region of the human P-globin mRNA followed by the first 22 bases of the sequence coding for the signal peptide of huIFN-a2c (the start codon is underlined).

EBI-2744 binds to the complementary strand at the end of

I

22 the sequence coding for huIFN-a2c (the stop codon is underlined) and contains a cutting site for XbaI. The reaction was stopped by extraction with phenol and chloroform and the DNA was precipitated with ethanol.

The PCR product was cut with HindIII and XbaI at the ends and the 0.64 kb long DNA fragment was isolated from an agarose gel (Fig. 6, Fig. Plasmid pAD-CMV19 was also doubly cut with HindIII and XbaI and then ligated iwith the cDNA fragment. Colonies obtained after transformation of E. coli HB101 were cultivated in order to prepare plasmid DNA. One of the plasmids obtained i was fully sequenced by means of the sequence of the inserted HindIII-Xbal region. With the exception of a single base exchange (CTG to TTG) in the 8th codon of the signal peptide, which did not cause any change in the coded amino acid (Leu), the expected sequence was obtained. The expression plasmid for secreted and 0glycosylated huIFN-a2c was designated pAD19B-IFN (Fig. 6).

i Description of the sequence elements of plasmid pAD-CMV19 (Fig. Bases 1 21 Binding site of oligonucleotide EBI-2133 1 590 Cytomegalovirus enhancer and promoter 722 740 Intron sequence of cytomegalovirus (splice donor) 741 805 Intron sequence of human P-globin (splice acceptor) 836 853 T7 promotor 862 922 Multicloning site 923 1055 Polyadenylation sites of 1056 1953 Promoter and 5'-non-coding region of hamster DHFR gene 1954 2039 DHFR exon 1 2040 2333 DHFR intron 1 c 1 S- 23 2151 2168 Binding site of EBI-1857 2344 2821 DHFR exons 2-6 coding region 2822 3474 DHRF 3'-non-coding region 3475 3812 SV40 replication origin (SV40 ori) 3813 6055 pBluescript component 3813 4291 M13 intergenic region (M13 ori) 4423 5283 p-lactamase, coding region 6038 6062 Binding site of EBI-2134 Example 3 Transient expression of huIFN-a2c in higher eukaryotic cells Approximately 106 cells (293, human embryonic kidney cells transformed with part of the adenovirus genome; F.L. Graham et al., J.Gen.Virol., 36 (1977) 59-77; ATCC CRL1573) per 80 mm petri dish were mixed 24 hours before trmnsfection with medium (Dulbecco's MEM/Nutrient Mix F12 with 15 mM Hepes; Gibco) containing 10% heat inactivated foetal calves' serum and incubated at 37°C in 5% CO 2 atmosphere. 3 hours before the transfection the cells were given 10 ml of fresh medium and incubated at 37 0 C. 10 gg of plasmid DNA (purified by centrifuging twice on a CsCl density gradient) pAD19B-IFN dissolved in 0.5 ml of 250 mM CaCl 2 were added dropwise to 0.5 ml of 2x HBS (16.36 g/l NaC1, 11.9 g/l Hepes, 0.40 g/l Na 2

HPO

4 pH 7.12). The precipitate obtained was added to a petri dish and the cells were incubated for a further 4 hours at 37"C. The cells were washed with PBS, shocked for 30 seconds with glycerol in ix HBS, washed again with PBS and incubated with 10 ml of fresh medium containing calves' serum at 37"C. After.72 hours the cell supernatant was harvested and used to detect the secreted IFN.

Z"

I x ^T<

*I

S- 24 Example 4: Construction of the expression plasmids pAD-CMVl and )i pAD-CMV2 From parts of the expression plasmids pCDM8 (Seed Aruffo, Proc. Natl.Acad.Sci.USA 84 (1987) 8573-8577; Seed, Nature 329 (1987) 840-842; Invitrogen Inc., San Diego, CA), pSV2gptDHFR20 (EP-Al 0321 842) and the plasmid Bluescript SK+ (Short et al., Nucleic Acids Res., 11 5521-5540; Strategene, La Jolla, CA) a new plasmid was constructed, having a multicloning site for the targeted insertion of heterologous DNA sequences and capable of replication in a high copy number in E. coli by means of ampicillin resistance. The intergenic region of M13 makes it possible to prepare single stranded plasmid-DNA by superinfection of the transformed bacteria with a helper phage R408 or M13K07) for easier sequencing and mutagenesis of the plasmid DNA. The T7 promoter, which precedes the multicloning site, makes it possible to produce RNA transcripts in vitro. In mammalian cells the expression of heterologous genes is driven by the cytomegalovirus (CMV) promoter/enhancer (Boshart et al., Cell 41 (1985) 521-530). The SV40 replication origin makes it possible, in suitable cell lines SV40 transformed cells such as COS-7, adenovirus transformed cell line 293 (ATCC CRL1573)) to achieve autonomous replication of the expression plasmid in high copy numbers and thereby obtain high rates in transient expression. In order to prepare permanently transformed cell lines and subsequently amplify the expression cassette by means of methotrexate a modified hamster minigene is used (promoter with coding region and the first intron) for dehydrofolate reductase (DHFR).as the selection marker.

t1\'

A-

'!2P ~b rC 25 a) Preparation of the vector and promoter parts by PCR The plasmid Bluescript SK+ was linearised with HindIII and 5 ng of DNA were used in a 100 Jl PCR mixture (reaction buffer: 50 mM KC1, 10 mM Tris-Cl pH=8.3, mM MgCl 2 0.01% gelatine, 0.2 mM of the four deoxynucleotide triphosphates (dATP, dGTP, dCTP, dTTP), units of Taq polymerase per 100 The primer used consisted of 50 pmol each of the synthetic oligonocleotides EBI-1786 TGGCGAATGGG-3'; binds just outside the M13 ori region in Bluescript position 475, independently of the M13 ori orientation) and EBI-1729 TGGCGTTTTTCC-3'; binds to Bluescript at position 1195 in front of ori, corresponds to the start of the Bluescript sequence in pCDM8, 6 bases 5' yield XhoI). After minutes denaturing at 94°C, PCR is carried out over cycles (40 seconds at 94"C, 45 seconds at 55"C, 5 min at 72"C, Perkin Elmer Cetus Thermal Cycler). The oligonucleotides flank the intergenic region of M13 or the replication origin (ori) with the intermediate gene for the p-lactamase. At the same time, at the end of the replication origin, an XhoI cutting site is produced and at the other end an EcoRI cutting site if produced.

The reaction mixture was freed from protein by extraction with phenol-chloroform and the DNA was precipitated with ethanol. The DNA obtained was cut with XhoI and EcoRI and after electrophoresis in an agarose gel a fragment of 2.3 kb was isolated.

ng of plasmid pCDM8 linearised with SacII were amplified by PCR with the oligonucleotides EBI-1733 GGTCGACATTGATTATTGACTAG-3'; binds to the CMV promoter region (position 1542) of pCDM8, corresponding to position 1 in pAD-CMV, SalI site for cloning) and EBI-1734 (5'-GGAATTCCCTAGGAATACAGCGG-3'; binds to polyoma origin of 3'SV40 polyA region in pCDM8 (position S- 26

I

3590)) under identical conditions to those described for Bluescript SK+. The oligonucleotides bind at the beginning of the CMV promoter/enhancer sequence and produce an Sail cutting site (EBI-1733) and, respectively, to the end of the SV40 poly-adenylation site and produce an EcoRI cutting site (EBI-1734). The PCR product was cut with SalI and EcoRI and a DNA fragment of 1.8 kb was isolated from an-agarose gel.

The two PCR products were ligated with T4 DNA ligase, E.

1 coli HB101 was transformed with the resulting ligation product and plasmid DNA was amplified and prepared by standard methods. The plasmid having the desired A properties (see Fig. 8) was designated pCMV-M13.

The SV40 replication origin (SV40 ori) was isolated from the plasmid pSV2gptDHFR20 (EP-A1 0321842). For this purpose this plasmid was doubly cut with HindIII and PvuII and the DNA ends were blunted by subsequent treatment with the large fragment of the E. coli DNA polymerase (Klenow enzyme) in the presence of the deoxynucleotide triphosphates. A 0.36 kb DNA fragment thus obtained was isolated from an agarose gel and ligated in pCMV-M13 linearised with EcoRI. A plasmid obtained after transformation of E. coli HB101 with the ori in the same orientation as the p-lactamase gene and the CMV promoter was designated pCMV-SV40. The construction of this plasmid is shown in Fig. 8.

b) Mutagenesis of the DHFR gene In order to prepare an expression plasmid having a versatile multicloning site, two restriction enzyme cutting sites were removed from the DHFR minigene by controlled mutagenesis and three restriction enzyme cutting sites were removed by deletion. In order to do this, a 1.7 kb BglII fragment of the plasmid 'NT

'.A

1 I2 27 pSV2gptDHFR20, this fragment containing the entire coding region of the hamster DHFR gene, was cloned into the BglII site of the plasmid pUC219 (IBI) and the plasmid pUCDHFR was obtained. E. coli JM109 (stratagene) cells transformed with pUCDHFR were infected with an approximately 40-fold excess of the helper phage R408 (stratagene) and shaken for 16 hours at 37"C in LB medium. Single stranded plasmid DNA was isolated from the bacterial supernatant.

Controlled mutagenesis was carried out in two successive steps, using the in vitro mutagenesis system RPN1523 (Amersham). The EcoRI site located at the start of Exon 2 was destroyed by replacing one base of GAATTC to obtain GAGTTC. This base exchange does not alter the coded amino acid sequence and moreover corresponds to the nucleotide sequence in the natural murine DHFR gene (McGrogan et al., J. Biol. Chem. 260 (1985) 2307-2314; Mitchell et al., Mol. Cell. Biol. 6 (1986) 425-440).

For the mutagenesis, an oligonucleotide (antisense orientation) of the sequence 5'-GTACTTGAACTCGTTCCTG-3' (EBI-1751) was used. A plasmid having the desired mutation was prepared as single strand DNA as described above and the PstI site located in the first intron was removed by mutagenesis with the oligonucleotide EBI-1857 (antisense orientation, 5'-GGCAAGGGCAGCAGCCGG-3') from CTGCAG into CTGCTG. The mutations were confirmed by sequencing and the resulting plasmid was designated pUCDHFR-Mut2. From the plasmid pUCDHFR-Mut2, the 1.7 kb BglII fragment was isolated and ligated into plasmid pSV2gptDHFR20, doubly cut with BglII and BamHI. After transformation of E. coli, amplification and isolation of the DNA, a plasmid of the required properties was obtained, which was designated pSV2gptDHFR-Mut2. By cutting with BamHI, in the 3'-non-coding region of the DHFR gene, a 0.12 kb DNA fragment following the BglII site was eliminated which further contains a KpnI 28 cutting site. By linking the overhanging DNA ends formed with BglII and BamHI, the recognition sequences for these two enzymes were also destroyed.

The plasmid pCMV-SV40 was doubly cut with EcoRI and BamHI and the DNA ends were subsequently blunted with Klenow enzyme. The DNA was purified by extraction with phenol-chloroform and ethanol precipitation, subsequently dephosphorylated by incubation with alkaline phosphatase and the 4.4 kb long vector DNA was isolated from an agarose gel.

The plasmid pSV2gptDHFR-Mut2 (Fig.9) was doubly cut with EcoRI and PstI and the DNA ends were blunted by minutes' incubation at 11°C with 5 units of T4 DNA polymerase (50 mM Tris-HCl pH=8.0, 5 mM MgCl 2 5 mM dithiothreitol, 0.1 mM of each of the four deoxynucleotide triphosphates, 50 gg/ml bovine serum albumin). The 2.4 kb long DNA fragment with the mutated DHFR gene was isolated from an agarose gel and ligated with the pCMV-SV40 prepared as described above. A plasmid obtained after transformation of E. coli and containing the DHFR gene in the same orientation as the CMV promoter was designated pCMV-SV40DHFR. In the last step the 0.4 kb stuffer fragment after the CMV promoter, which also originated from the starting plasmid pCDM8, was exchanged for a multicloning site. In order to do this, the plasmid pCMV-SV40DHFR was doubly cut with HindIII and XbaI and the vector component was isolated from an agarose gel. The multicloning site, formed from the two oligonucleotides EBI-1823 CATCGATGGATCCGGTACCTCGAGCGGCCGCGAATTCT-3') and EBI-1829 TCGACCTGCAGA-3'), contains, including the ends which are compatible for cloning in HindIII XbaI, restriction cutting sites for the enzymes PstI, SalI, ClaI, BamHI, KpnI, XhoI, NotI and EcoRI.

SI- 29 1 gg portions of the two oligonucleotides were incubated for one hour at 37°C in 20 pl of reaction buffer (70 mM Tris-Cl pH=7.6, 10 mM MgCl2, 5 mM dithiothreitol, 0.1 mM ATP) with 5 units of T4 polynucleotide kinase, in order to phosphorylate the 5'-ends. The reaction was stopped by heating to 70"C for 10 minutes and the complementary i oligonucleotides were hybridised with one another by incubating the sample for a further 10 minutes at 56°C and then slowly cooling it to ambient temperature. 4 Al of the hybridised oligonucleotides (100 ng) were ligated with about 100 ng of plasmid vector and transformed in E. coli HB101. A plasmid which was able to be linearised with the enzymes of the multicloning site S(with the exception of NotI) was designated pAD-CMVl.

SNone of the many clones tested could be identified as having a plasmid which could be cut with NotI. The sequencing always showed the deletion of some bases inside the NotI recognition sequence. In the same way with the oligonucleotide pair EBI-1820

AATTCGCGGCCGCTCGAGGTACCGGATCCATCGATGTCGACCTGCAGAAGCTTG-

and EBI-1821 GATCCGGTACCTCGAGCGGCCGCGAATTCTCTAG-3') the expression plasmid pAD-CMV2 was prepared, which contains the restriction cutting sites within the multicloning site in the reverse order. The plasmid pAD-CMV2 was thus obtained, which was able to be linearised with all the restriction enzymes, including NotI.

-Ai The nucleotide sequence of the 6414 bp plasmid pAD-CMV1 is shown in full in Fig.1l.

The sections on the plasmid (given in the numbering of the bases) correspond to the following sequences: 1-21 EBI-1733, beginning CMV enhancer promoter (from CDM8) 632-649 T7 promoter

:I

I,

A

30 658-713 714-1412 1413-2310 2311-2396 2516 2701-3178 2707 3272-3273 3831 3832-4169 4170-4648 4780-5640 6395-6414 Multicloning site (HindIII to XbaI from EBI-1823., EBI-1829) SV40 intron and polyadenylation site (from CDM8) 5'-non-coding region and promoter of the hamster DHFR gene (from pSV2gptDHFR20) Hamster DHFR: Exon 1 A to T mutation destroys PstI site in DHFR intron 1 DHFR Exons 2-6 (coding region) A to G mutation destroys EcoRI site Deletion between BglII and BamHI in DHFR 3'-non-coding region End of DHFR gene (from pSV2gptDHFR20) SV40 ori (from pSV2gptDHFR20) M13 ori (from pBluescript SK+) /-lactamase (coding region) EBI-1729, end of the pBluescript vector sequence The preparation of plasmids pAD-CMVl and pAD-CMV2 is shown in Example Development of recombinant "Chinese hamster ovary (CHO)" cell lines which produce glycosylated human interferon-a2 a) Transfection of CHO cells and selection of stably transfected cell lines The parental cell lines CHO-DXB11 and CHO-DG44 (Proc.

Natl. Acad. Sci. USA 77, 4216-4220, 1980; Som. Cell.

Molec. Genet. 12, 555-666, 1986) were cultivated in Roswell Park Memorial Institute (RPMI) Medium 1640 supplemented with 10% foetal calves' serum, hypoxanthine 51L/S NT3 31 (100 iM) thymidine (16 iM), sodium penicillin G (100 units/ml) and streptomycin (50 units/ml). Two days before the transfection the cells were placed in 25 cc bottles; at the time of transfection the cells were virtually confluent.

The transfection experiment was carried out as follows.

pA of a solution of plasmid pAD19B-IFN (1 Ag/ml) were diluted with 125 pg of 2 M CaC1 2 and 855 -il of sterile deionised water. This solution was added dropwise to 1 ml 2 x HSB (1 x HSB contains, per litre of solution: 8.18 g NaCI, 5.94 g HEPES, 0.2 g Na 2

HPO

4 pH The culture medium of the CHO cells was removed and 0.25 ml of the suspension were added to each bottle; the cultures were incubated for 4 hours at 37°C. The suspension was then removed, the cells were detached from the surface using trypsin/EDTA solution and suspended in selection medium (the selection medium consisted of minimum essential medium, alphamodification without ribonucleotides and .i deoxyribonucleotides, supplemented with 10% dialysed foetal calves' serum, sodium penicillin G 100 units/ml, j streptomycin 50 units/ml and amphotericin B 2.5 jg/ml; ml per bottle). The cell suspension was then transferred into the wells of two cell culture microtitre plates (96 wells per plate, 0.2 ml per well) and incubated at 37*C for two weeks. The selection |I medium specified was also used for all the other Sexperiments, although without the amphotericin B.

The cell cultures were visually inspected for cell growth. Culture medium from wells which showed cell growth were tested for IFN-a2 content using an enzyme immunoassay which uses two monoclonal antibodies against IFN-a2 (Biochem. J. 276, 511-518, 1991). This test was repeated one week later with new culture supernatants.

Cells from positive cultures were detached from the 7 v^ \?AlT 0* 32 surface using trypsin/EDTA solution and transferred into culture dishes having *24 wells. Cultures which exhibited good cell growth were then repeatedly tested for IFN production. The IFN-a2 concentrations in the supernatants were typically in the range from. 2,000 to >10,000 units/ml (1 ng of IFN-a2 protein corresponds to 230 units).

b) Amplification of the IFN-a2 gene by methotrexate selection Transfected clones from the two parental cell lines, CHO-DXB11 and CHO-DG44, which had shown high concentrations of IFN in numerous tests, were selected for amplification; in addition, 3-5 other clones were combined. These cultures were then kept in 25 cc bottles in selection medium (without amphotericin to which methotrexate was added in concentrations of 20 nM or 50 nM. The cultures were given fresh medium once a week. Any surviving clones were observed after about 2 to 3 weeks. As soon as the cells had colonised about of the culture area, the supernatants were again tested for their content of IFN-a2. The cells were then i detached, diluted and transferred into new bottles. The methotrexate concentration was then increased by a factor of about 2 to 5, e.g. from 20 nM to 50 and 100 nM, or from 50 nM to 100 and 200 nM. After several such selection cycles in the presence of increasing Samounts of methotrexate, and selection of resistant cultures in accordance with their IFN production, it was finally possible to obtain cell lines which were resistant to methotrexate concentrations of up to 5,000 nM and which secreted relatively large amounts of IFN-a2. Table I which follows illustrates the increase in productivity taking as its example the cell line CHO-DXBll-IFN-a2c-3/2D4.

N

r^j "V 0>" 33 Table II which follows shows the results with the cell line CHO-DG44-IFN-U2c- ool,'S": iul---~ I i ;n*rrrx~ A'"4~ 9A j Table I Methotrexate Concentration (nM) IFN-a2c in (units/ml) culture supernatants 0 100 200 500 1000 2000 5000 6,000 12,000 29,000 96,000 110,000 140,000 350,000 200,000 210,000 14,000 89,000 -125,000 -120,000 -190,000 -300,000 -900,000 -350,000 -960,000 t rrrrra~~i~ Li n r Table II Methotrexate Concentration (nM) IFN-a2c in (units/ml) culture supernatants 100 200 500 1000 2000 5000 29,000 58,000 69,000 71,000 190,000 170,000 200,000 190,000 83,000 -112,000 98,000 -220,000 -220,000 -570,000 -350,000 -320,000 -u" j 36 From culture supernatants of recombinant CHO cells, it was possible to purify IFN-a2 using affinity chromatography on monoclonal antibodies the antibodies EBI-1 or EBI-10) using known methods (e.g.

Nature 285, 446-450, 1980; J. Biol. Chem. 265,, 9290-9295, 1990; Biochem. J. 276, 511-518, 1991). The proteins according to the invention may, in particular, be purified by the method described in -EPA 0 203 382.

Example 6 Characterisation of recombinant, glycosylated human IFN-a2c from Chinese hamster ovary (CHO) cells a) Reverse Phase HPLC (RP-HPLC) Recombinant glycosylated IFN-a2c from CHO cells, purified by affinity chromatography, were compared by RP-HPLC with recombinant IFN-a2c from E. coli which is not glycosylated. The exact method of analysis is described in Adolf et al., J. Biol. Chem. 265, 9290-9295 (1990). Glycosylated CHO-IFN-a2c (upper part of Fig.21) consists of two main peaks (peaks 1 and 2) and two smaller IFN peaks (peaks 3 and Unglycosylated E.

coli IFN-a2c, on the other hand (lower part of Fig.21) shows a mean peak (correct disulphide bridges) and a smaller subsidiary peak which originates from a form with "scrambled" disulphide bridges. A comparison of the retention time shows that the two main peaks of CHO-IFN-a2c elute somewhat earlier than the main peak of E. coli IFN-a2c. The reason for this reduced hydrophobicity is the presence of oligosaccharides in the CHO-IFN-a2c. The two smaller IFN peaks in the CHO-IFN-a2c have about the same retention time as the main peak of the E. coli IFN-a2c and thus most probably originate from a smaller unglycosylated part of the CHO-IFN-a2c.

VT 2 i 37 i b) N-terminal sequencing The two main peaks of the CHO-IFN-a2c were isolated from the RP-HPLC and sequenced together. The sequencing conditions are described in Adolf et al., J. Biol. Chem.

265, 9290-9295 (1990). The first 15 amino acids were able to be identified in accordance with the cDNA sequence. There were no indications of heterogeneity at the N-terminus.

c) C-terminal analysis The main peak of the E. coli IFN-a2c and the two main peaks of the CHO-IFN-a2c were isolated by the RP-HPLC and cleaved with trypsin. The tryptic peptides were separated once more by RP-HPLC. The experimental conditions are described in Adolf et al., Biochem. J.

276, 511-518 (1991). Fig. 22 shows a comparison of the peptide maps obtained. Between the peptide maps of peak 1 and peak 2 of the CHO-IFN-a2c (upper and middle part) i there was only one difference. The tryptic peptide 18 from peak 1 is virtually absent from peak 2 (here as peptide 15). Instead, a new peptide (number 19) is found in the map of peak 2, which does not appear at all either in peak 1 or in the E. coli IFN-a2c (lower part of Fig. 22).

The peptides 12 (from E. coli-IFN-a2c), 18 (from peak 1 of the CHO-IFN-a2c), 15 and 19 (from peak 2 of the CHO-IFN-a2c) were analysed by plasma desorption mass spectrometry (PD-MS). The experimental details are described in Adolf et al., Biochem. J. 27-6, 511-518 (1991). For the first three of the samples mentioned a molecular weight was found which corresponds to amino acids 150 to 162 of the IFN-a2c sequence. Peptide 19 from peak 2 of the CHO-IFN-a2c, on the other hand, gave 7 i i: jo li 38

IS

r i a lower molecular weight, corresponding to amino acids 150 to 161. Part of peptide 19 was also sequenced, and it was clearly found that this peptide begins with amino acid 150 and ends with LEU-161. From these results it can be concluded that peak 1 of the CHO-IFN-a2c contains a complete IFN molecule, whereas in peak 2 the four Cterminal amino acids (162-165) are absent. Amino acids 163-165 cannot be positively identified-in a peptide map after trypsin cleavage since the resulting dipeptide (163 to 164) and the free amino acid (165) are eluted in the dead volume of the RP column. The small proportion of peptide 15 (unabbreviated tryptic peptide with amino acids 150 to 162), which was also found in peak 2 of the CHO-IFN-a2c, can indeed be put down to a contaminating content of peak 1, since peaks 1 and 2 cannot be totally separated by RP-HPLC.

In further experiments it was found that the C-terminal shortening of the O-glycosylated IFN-a2 from CHO cells can be prevented if a trypsin-free solution EDTA disodium salt, 200 mg/L with D(+)glucose monohydrate, 200 mg/L in phosphate-buffered sodium chloride solution pH 7.4) is used to detach the cells from the surface of the culture vessels instead of the usual trypsin/EDTA solution. IFN-a2 which was purified from cell cultures cultivated in this way using the methods described above, showed in reverse phase HPLC (analogously to Figure 21a) only the peak 1, corresponding to the complete protein, but not peak 2, corresponding to the abbreviated protein. Furthermore, with the aid of the tryptic peptide maps (analogously to Figure 22) it was possible to show that the peptide pattern of this protein prepared without the use of trypsin is identical to the pattern of the peptides generated from peak 1 (Figure 22a).

611- U 39 SDS-qel electrophoresis Peaks 1 and 2 of the CHO-IFN-a2c were isolated by RP-HPLC. They were analysed individually and by comparison with E. coli IFN-a2c, both under reducing conditions (aftcr boiling with dithiothreitol) and also under non-reducing conditions by means of SDS gel electrophoresis. The experimental details are described in Adolf et al., J. Biol. Chem., 265, 9290-9295 (1990).

The results are shown in Fig. 23 (traces 2 to 4 under non-reducing conditions, traces 5 to 7 under reducing conditions; upper part with 4 pg IFN in each trace, lower part with 1 gg IFN in each trace). It is particularly apparent from traces 5 to 7 in the lower part that both peak 1 and peak 2 of the CHO-IFN-a2c have a greater molecular weight than the unglycosylated E.

coli IFN-a2c. Because of the incomplete separation of peaks 1 and 2 during RP-HPLC, the peaks 1 and 2 are mutually contaminated. For the same reason peak 2 is also contaminated with a small amount of unglycosylated CHO-IFN-a2c, originating from peak 3 (see Fig. 21).

Taking account of this contamination, the main bands of peaks 1 and 2 of the CHO-IFN-a2c appear to be homogeneous. Since the peaks 1 and 2 differ in the Cterminus (see above), it can be concluded from the results of the SDS gel electrophoresis that the oligosaccharide contents of peaks 1 and 2 of the CHO-IFN-a2c are identical (see also hereinafter in Chapter f).

e) Deglycosylation of CHO-IFN-a2c Peaks 1 and 2 of the CHO-IFN-a2c were isolated from RP-HPLC and dried in a SpeedVac Concentrator. These samples and E. coli IFN-a2c were incubated in 10 gl of 0.1 M NaOH for 20 hours at ambient temperature. The SR samples deglycosylated by this P-elimination were r~i3Jn-~ I 1 40 analysed by comparison with untreated samples using SDS gel electrophoresis. .The results in Fig. 24 show that the molecular weight of peaks 1 and 2 of the CHO-IFN-a2c is significantly reduced after treatment with. NaOH and is identical to that of the E. coli IFN-a2c treated with NaOH. The diffuse appearance of the bands of all samples treated with NaOH can be put down to changes in the peptide chain under the reaction conditions used.

f) Identification of the qlycopeptides by peptide mapping The comparison of the peptide maps after trypsin cleaving of E. coli IFN-a2c (Fig. 22, lower part) and of peak 1 (complete C-terminus) of the CHO-IFN-a2c (Fig. 22, upper part) shows that two peptides have different retention times. The peptides 18 and 21 of E.

coli IFN-a2c, which contain the amino acids 84-112 and 71-112, respectively, do not appear in the peptide map of peak 1 of the CHO-IFN-a2c. Instead, there are two new peptides (numbers 26 and 31) which show the same ratio of absorptions at 280 and 214 nm as the peptides 18 and 21 of E. coli IFN-a2c. It can be concluded from this that peptides 26 and 31 are the glycosylated versions of the amino acid sequences 84-112 and 71-112, respectively. Therefore, they also elute significantly earlier from the RP column than the analogous peptides of E. coli IFN-a2c. For the two possible lengths of the tryptic peptides (amino acids 84-112 and 71-112, respectively) there is in each case a main peak (peptide 26 or 31), from which it can be concluded that the oligosaccharide content is largely homogeneous.

A comparison of the peptide maps of peaks 1 and 2 of the CHO-IFN-a2c (upper and middle part of Fig. 22) shows that the glycopeptides in question (26 and 31 from peak Zs 1 and 24 and 30 from peak 2) are identical. It thus 4.* 41follows that the four missing amino acids at the Cterminus of peak 2 are the only difference between peaks 1 and 2.

All the main peptides of the three IFN samples relating to the amino acid sequences 84-112 and 71-112 were isolated from RP-HPLC and further cleaved with Staphylococcus aureus V8 protease at the C-terminal end of glutamic acid. The exact conditions are described in Adolf et al., Biochem. J. 276, 511-518 (1991). The resulting peptide maps were compared, all the different peaks were isolated and further analysed by N-terminal sequencing and/or mass spectrometry.

One of the Staph.A. peptides occurring in.peaks 1 and 2 of the CHO-IFN-a2c but not in E. coli IFN-a2c contained the amino acids 97-112 of the IFN-a2c sequence. It was not possible to identify THR-106 in this peptide by Nterminal sequencing. From this it can be concluded that THR-106 is present in glycosylated form in this peptide.

In the Edman sequencing used here, glycosylated amino acids are derivatised and cleaved like unglycosylated amino acids, but because of their increased hydrophilicity they cannot be extracted from the reaction vessel with butyl chloride. Therefore, in this breakdown step, it is not possible to identify an amino acid of any kind but the sequence continues completely undisrupted thereafter. A further indication that the oligosaccharide is bound to THR-106 is provided by the result that the GLU-107/THR-108 bond was only partly cleaved by the Staph.A. protease. Obviously the accessibility of this peptide bond is restricted by the presence of the oligosaccharide. In the analogous peptide from E. coli IFN-a2c, this peptide bond is almost fully cleaved.

Another Staph.A peptide which occurs only in CHO-IFN-a2c was analysed by plasma desorption mass spectrometry.

The molecular weight obtained corresponded to amino acids 97-112 including an oligosaccharide, consisting of one molecule of N-acetylgalactosamine and gal-actose and two molecules of N-acetylneuramino acid.

These results show that both the glycosylation site and the oligosaccharide content of the CHO-TFN-a2c are largely identical to the proportions found in natural IFN-a2 obtained from virus-stimulated leukocytes.

Isolation of the O-glycosylated interferon from virusstimulated cells: Methods Interferon Bioassay: The antiviral activity of the IFN preparations was determined in microtitre plates using an assay which measures the cytopathic effect (CPE) of encephalomyocarditis virus (EMCV). The test cells used were A549 human lung cancer cells. Details of this assay have been described Adolf, G.R., J.Gen.Virol. 68, 1669-1676 (1987)). In each bioassay, all the titrations were carried out twice. A laboratory standard preparation of recombinant human IFN-a2c produced in E. coli was included in each assay: the activity of this preparation was briefly determined by comparison with the international reference preparation for human IFN-a2, Gxa 01-901-535. All the IFN activities observed were corrected with respect of the defined activity of this reference preparation.

Interferon ELISA: An ELISA was established, using two neutralising murine IgG MAbs for IFN-a and an IFN-a2c laboratory reference preparation (see above) as standard. The preparation of the antibodies and their RAi properties have been described (Adolf et al., J. Cell NT 0 S: 43 Physiol. suppl. 2, 61-68 (1982); Adolf G.R. J. Gen.

Virol. 68, 1669-1676 (1987)). The antibody EBI-1 was used to coat the assay plates; the antibody covalently coupled to horse radish peroxidase, was added with the sample to be investigated. O-phenylenediamine and sodium perborate were used as substrates for the enzyme; the reaction was stopped by the addition of sulphuric acid and the absorption of the resulting product was measured (492 nm, reference 690 nm).

Purification of natural human IFN-a2: An affinity column was prepared by coupling to 12 mg of the monoclonal antibody, e.g. the MAb EBI-10 (purified from the mouse ascites by ammonium sulphate precipitation and protein G affinity chromatography using standard methods) on 1 g of CNBr-activated sepharose 4B following the manufacturer's recommendations (Pharmacia). The final bed volume of the column was approximately 3 ml. Partially purified human leukocyte interferon (Cantell et al. Methods Enzymol. 78, 29-38 (1981); Cantell et al. ibid. 499-505) in which the IFN-o component had been removed (Adolf et al. J. Biol. Chem. 265, 9290-9295 (1990)) and which contained about 2-3 x 106 IU/ml with a total protein concentration of 2 mg/ml, was applied to the column at a throughflow rate of 1 ml/min (200 and 350 ml). The column was then washed with 0.1 M sodium phosphate buffer pH 7.5 (buffer A) and eluted with a linear gradient of buffer A and buffer B (0.1 M sodium citrate pH 2.1) in an FPLC system (Pharmacia) at a throughflow rate of 1 ml/min. The fractions obtained were tested for IFN activity using the ELISA. Corresponding fractions in the two mixtures were collected, neutralised with 1 M NaOH and reapplied to the same column, which had been re-equilibrated with buffer A.

R The same elution programme was used. (Throughflow rate 0', 44 0.25 ml/min). Corresponding fractions were again collected, neutralised and frozen in aliquots.

SDS gel electrophoresis, HPLC techniques and -amino acid sequencing: SDS polyacrylamide gel electrophoresis and reverse phase HPLC were used to analyse the purified IFN-a2; all the methods have been described in detail (Adolf et al., J. Biol. Chem. 265, 9290-9295 (1990)).

The N-terminal sequence was determined in an automatic sequenator (Applied Biosystems, Model 477A); amino acid derivatives were analysed on-line by RP-HPLC (Adolf et al., J. Biol. Chem. 265, 9290-9295 (1990)).

"Mapping" the proteolytic peptides: Affinity-purified IFN-a2 was again purified by reverse phase HPLC, denatured and desalinated as described by Adolf et al., J. Biol. Chem. 265, 9290-9295 (1990). The peak fractions were collected and dried in a SpeedVac concentrator. 29 ig (peak 1) and 66 pg (peak 2) of protein were dissolved in 0.1 ml of 1% ammonium bicarbonate solution; 0.5 and 1 gg of trypsin (Boehringer Mannheim) in 3 and 6 .l respectively, of 0.01% trifluoroacetic acid were added and the reaction mixture was incubated at 37"C. After 6 hours' incubation time the same amount of trypsin was added once again and incubation was continued for a further 18 hours. The reaction mixture was reduced before the analysis by the addition of 10 jl of 0.5 M dithiothreitol and 100 il of 6 M urea for 2 hours at ambient temperature. Reverse phase HPLC was carried out on a Delta Pak C18 column (Waters; 3.9 x 150 mm; particle size 5 im; pore diameter 100"A) at 30°C using the following solvents: Solvent A: 0.1% trifluoroacetic acid in water; Solvent B: 0.1% trifluoroacetic acid in acetonitrile. The following gradient programme was used (throughflow rate 1 ml/min): 0-55 min: 0-55% B (linear gradient); 55-70 min: 50% B. The peptides were detected 45 L by their absorption at 214 and 280 nm. The resulting patterns were compared with those of the recombinant IFN-a2c originating from E. coli. The peptides of the natural IFN-a2, which behaved differently in .their elution characteristics from their recombinant counterparts, were collected and N-terminally sequenced or they were further broken down with Staphylococcus aureus V8 protease (endopeptidase Glu-C, Boehringer Mannheim). 0.88 pg (peak 2.6 pg (peak 2/Ia) and pg (peak 2/Ib) of the peptides were dissolved in 0.1 ml of 25 mM phosphate buffer, pH 7.8. Protease dissolved in water was added (17.5 ng, 52.5 ng and 29 ng) and the reaction mixture was incubated at 37°C.

After 6 hours the same amounts of protease were added again and the mixture was incubated for 18 hours. The samples were then subjected to reverse phase HPLC analysis (see above). Corresponding fractions were collected and N-terminally sequenced.

I

Deqlycosylation of the IFN-a2: Purified, denatured and desalinated IFN-a2 were treated with Vibrio cholerae neuraminidase (Boehringer Mannheim) (50 mU/ml, 18 hours at 37"C in 20 p1 50 mM sodium acetate pH 5.5, 4 mM CaCI 2 and/or endo-a-N-acetyl-galactosaminidase the same as O-glycanase (Boehringer Mannheim) (100 mM/ml, 18 hours at 37"C in the same buffer). Chemical elimination was achieved by incubating in 0.1 M NaOH for 20 hours at ambient temperature.

Plasma desorption mass spectrometry: Mass spectra of the tryptic peptides were measured in a "BIO-ION 20 time-of-flight" mass spectrometer (BIO-ION Nordic AB, Uppsala, Sweden). The samples were dissolved in aqueous trifluoroacetic acid and applied to nitrocellulose-coated targets (BIO-ION). The spectral accumulation times ranged between 0.5 and 12 hours, depending on the yield. The spectra were measured at an acceleration voltage of 17 kV.

r -0 e r 46 Example 7: Purification of natural human IFN-a2 Human leukocyte interferon, obtained from Sendai virus induced human peripheral leukocytes and partially purified by the purification method described by Cantell et al. (Methods Enzymol 78, 29-38 and 78, 499-512 (1981)), was used as starting material for the isolation and purification of the IFN-a2. By selective affinity chromatography with anti IFN-o monoclonal antibodies, for example OMG-4, OMG-5 or OMG-7, the content of IFN-w was removed (Adolf et al., Virology 175, 410-417 (1990); EPA 262 571). The specific antiviral activity was 1-2x106 IU/mg; IFN-a, with a specific activity of 2x10 8 IU/mg, was therefore represented by only about 1% of the total protein content. In order to cleanse the IFN-a2 of contaminating foreign proteins and at the same time of other species of IFN-a, highly selective anti-IFN-a2 monoclonal antibodies were used. These antibodies have a high specificity for IFN-a2 in standardised neutralisation bioassays (Adolf G.R. J.

Gen. Virol. 68, 1669-1675 (1987)).

An immunoaffinity column was prepared by coupling a monoclonal antibody of this kind, e.g. EBI-10, prepared for example according to J. Gen. Virol. 68, 1669-1676 (1987) or DE 33 06 060.6 to CNBr-activated Sepharose 4B.

The antibody had been purified from the mouse ascites by ammonium sulphate precipitation and protein G affinity chromatography using standard procedures. For example, 12 mg of the monoclonal antibody EBI-10 were used, i coupled to 1 g of CNBr-activated Sepharose 4B, under the conditions recommended by the manufacturer (Pharmacia).

The final bed volume of the column was about 3 ml.

4RA/ The leukocyte interferon preparation was applied to the S\ column; approximately 20% of the antiviral activity was 47 bound. The column was eluted with a linear buffer gradient of 0.1 M sodium phosphate, pH 7.5, and 0.1 M sodium citrate, pH 2.1. Two protein peaks could be distinguished in the eluate (Fig. 12): fraction A and fraction B. The two fractions were analysed for their content of IFN-a, using a "two-sided ELISA", using both and EBI-1. Both antibodies show a high specificity for IFN-a2 (Adolf et al., J. Cell Physiol.suppl. 2, 61-68 (1982)). Recombinant IFN-a2c was used as the standard. The fraction which had been eluted at a low pH (Fig. 12, peak like the sample, yielded titration curves which ran parallel to the titration curve of the recombinant IFN-a2c. The passage and fractions of the first peak yielded curves *i with different inclines; therefore they could not be quantified by the ELISA (Fig. 13) but were checked in the biological assay (Table III).

The low pH necessary for eluting peak A, as well as the results of the ELISA indicated that the IFN-e2 was a main component of peak A. To ensure that all the immune-reactive IFN-a had been bound by the antibody, the material was run through the column once more and eluted for a second time as described above. The eluted material yielded less than 10% of the IFN activity which had been bound in the first running.

Both the A fractions and the B fractions were separately collected, neutralised and again subjected to chromatographic purification on the same affinity column. In both cases, more than 95% of the IFN activity was bound; elution was carried out at the same gradient position as in the first cycle. The starting product, running and the collected fractions of both chromatographies were investigated for their protein content by Coomassie blue staining assays and for their IFN activity content by an antiviral bioassay. The results are shown in Table III:

A

it 4 i ~iy Table III Purification of natural IFN-a2 Protein Antiviral activity x 10- 6 mg/ml IU/ml 1 IU total Volume ml

P-IF

2 550 1.7 1st cycle throughflow 550 1.7 1st cycle eluate A 18 0.08 1st cycle eluate B 17 0.05 2nd cycle eluate A 8 0.1 2nd cycle eluate B 4 0.1 1 Average of 5 different bioassays 2 Partially purified human leukocyte IFN 2.8 1540 Yield 100 79 2.2 9.6 4.3 12 1216 172 4.8 6.2 after removal of IFN-wl i

IL

I

U

49 Example 8: Identification of the affinity-purified protein as IFN-a2 The affinity-purified IFN-a was first analysed by reverse phase HPLC and purified. Peak A showed two incompletely dissolved peaks 1 and 2 with a mass ratio of about 1:2 (bottom of Fig. 14); peak represented a more hydrophilic protein fraction. The two peak fractions were collected, rechromatographed and subjected to N-terminal amino acid analysis. The following sequence was obtained from both fractions (the Cys groups in brackets were not identified but derived on the basis of the conserved IFN sequences).

1 5 10 CYS]-ASP-LEU-PRO- GLN-THR-HIS-SER-LEU- GLY-SER- ARG-ARG-THR-1LEU-MET-LEU-LEU-ALA-2GLN-MET-ARG- 2ARG-ILE- 25SER-LEU-PHE-SER-[CYS]- 3 0

LEU-

By comparison with published sequences, both were identified as IFN-a2.

In both peak fractions and the amino acid at position 23 was clearly identified as arginine; the variant designated LeIFA, which has lysine at position 23 (Goeddel et al., lature, 290, 20-26 (1981)) was therefore not present in detectable amounts in the leukocyte preparation used. The amino acid at position 34 was identified as histidine; the isolated IFN-a2 was therefore the variant IFN-a2b.

The specific antiviral activity of natural IFN-a2 based on the international reference preparation for IFN-a2, Gxa01-901-535, based on determining the protein content 1.K> ~4hl :22;, 50 of the sample by its absorption at 214 nm (Adolf et al., Virology 175, 410-417 .(1990), was determined as 1.5x10 8 IU/mg (average from five independent bioassays).

In a comparison of the retention times of natural IFN-a2 on reverse phase HPLC with that of recombinant E. coli IFN-a2c, it became obvious that the recombinant protein was eluted significantly later (Fig. 14). The increase hydrophilicity of the natural protein as well as its heterogeneity must therefore be connected with posttranslational modifications.

Reverse phase HPLC of the elution peak yielded a complicated pattern of five incompletely dissolved peaks. Sequence analyses showed that all the peaks represented IFN-a species but none of them represented IFN-a2.

The IFN-a2 purified by HPLC was further analysed by SDS- PAGE after reduction with dithiothreitol (Fig. Under the conditions selected, the recombinant IFN-a2c of E. coli showed an apparent molecular weight of 17,500 D (molecular weight starting from the amino acid sequence: 19,287 HPLC peak fraction gave a single, relatively broad band (apparent molecular weight 20,000 D) whereas peak fraction was split into two main components (20,000 and 19,000 D) and into one subsidiary component (21,000 These differences in molecular weight compared with the recombinant protein from E. coli, the heterogeneity of size as well as the increased hydrophilicity indicate that the natural IFN-a2 is glycosylated. Since there is no recognition site for N-glycosylation in the IFN-a2 structure, there must be O-glycosylation.

-51 I 51 Example 9: Reaction of natural IFN-a2 with endo- and exoqlycosidases The following experiments were each carried out with two peaks after separation by RP-HPLC (peaks 1 and 2 from Fig. 14b). Both samples were incubated with neuraminidase and then with O-glycanase. After each enzyme reaction, one aliquot was examined by SDS-PAGE.

As can be seen in Fig. 16, peak 1 did not react with either neuraminidase nor with O-glycanase. The apparent molecular mass remained constant at 20,000. The three bands of peak 2, on the other hand, reacted both with neuraminidase and also, subsequently, with O-glycanase.

The reaction with neuraminidase brought about a reduction in the apparent molecular mass of the two heavier bands (Mr 21,000 and 20,000) to 19,000.

However, traces of the band with the apparent molecular mass of 20,000 lagged behind. Subsequent incubation of the protein with O-glycanase resulted in a further reduction in the apparent molecular mass from 19,000 to 17,500 apparent molecular mass of E. coli IFN-a2c).

The component with Mr 19,000 was totally broken down.

As before, small amounts of the band with the apparent molecular mass of 20,000 could still be detected. Since the separation of the two peaks 1 and 2 from Fig. 14b by means of RP-HPLC was incomplete, the uncleavable proportion of the band with Mr 20,000 can presumably be put down to a contamination of peak 2 with peak 1.

In another experiment, peak 2 was incubated with Oglycanase, without having previously been treated with neuraminidase (O-glycanase will only cleave the disaccharide Gal(pl-3)GalNAc from the protein when the latter is not substituted by any other compounds). The h<V o f 52 reaction product was once again separated using SDS-PAGE (Fig. 17). It is clearly apparent that only the lightest component of peak 2 undergoes a reduction in its molecular mass (reduction from Mr 19,000 to Mr 17,500). The apparent molecular masses of the two heavier components (Mr 21,000 and 20,000) remained unchanged.

Example Reaction of natural IFN-a2 with 0.1 M NaOH Since O-glycosylations can be broken down even under mild alkaline conditions, attempts were made to deglycosylate the O-glycanase-resistant component (peak 1 from Fig. 14b) by incubation with 0.1 M NaOH. The reaction took place as described above. At the same time, as a control, E. coli IFN-a2c and peak 2 were incubated under the same conditions. The reaction products were analysed by SDS-PAGE. As can be seen in Fig. 18, the molecular masses of all components of natural IFN-a2 were reduced to the apparent molecular mass of E. coli IFN-e2. The lack of definition of the protein bands can be attributed to the slight destruction of the protein under the conditions described. The bands in the higher molecular range (Mr >30,000) also occurred as a consequence of alkaline treatment.

Example 11: Identification of the glycopeptides by means of peptide mapping The two peaks of natural IFN-a2 (Fig. 14b) and E. coli IFN-a2c were cleaved with trypsin, reduced and separated Xg by RP-HPLC. Fig. 19 shows sections of the r- 1. 53 chromatograms. Two peaks from the peptide map of E. coli IFN-a2c are striking for their hydrophobicity (and therefore relatively delayed elution) in relation to the analogous peaks from the natural IFN-a2: peak I and peak II (in the peptide map of E. coli IFN-a2c) were I eluted significantly later than their corresponding peaks 1/I and 1/II of peak 1 in Fig. 14b and peaks 2/Ia 2/IIb, 2/IIa and 2/IIb of peak 2 in Fig; 14b.

N-terminal sequencing of the above-mentioned peaks of natural IFN-a2 and the two E. coli peaks produced the sequence of the peptide of amino acid (AS) 84-112 for j the peaks I, 1/I, 2/Ia, 2/Ib (Fig. 19) and the sequence i of AS 71-112 for the peaks II, 1/II, 2/IIa, 2/IIb (the amino acid sequence of IFN-a2c is shown in Fig. The different retention times must therefore be attributable to the glycosylation of the peptides from natural IFN-a2.

Example 12: Plasma desorption mass spectrometry of the qlycopeptides Sof natural IFN-a2 The peaks 1/II, 2/Ia, 2/IIa and 2/IIb were further characterised by PD-MS. The results of the measurements are shown in Table IV. The difference in the molecular masses calculated from the amino acid sequence and the i molecular masses of the individual peptides actually obtained can be explained by different glycane structures: the molecular mass of the peptide 1/II, which was obtained from the O-glycanase-resistant form of IFN-a2, corresponds to the molecular mass of the peptide (AS 71-112), which is substituted with a tetrasaccharide consisting of two N-acetylhexosamine units and two hexose units. Analogously to structures Sky l' S of O-glycanes of this kind already described, this ~I

J

-J 54 should be an oligosaccharide with the following Sstructure: Gall-3(Gall-4GlcNAcl-6)GalNAc-.

Peptide 2/Ia had a molecular mass of 3,975 amu, which can be explained by the substitution of the peptide with I Ithe trisaccharide NeuAc-Gal-GalNAc. The same glycane j structure can be derived from the molecular mass oi the Speptide 2/IIa (5,448 amu). For peptide 2/IIb, a value of 5,132 amu was measured, corresponding to a Sglycosylation with the disaccharide Gal-GalNAc.

In principle, all the peaks analysed had a molecular mass which was approximately 23 amu higher. This can be explained by the accumulation of Na -ions on the peptide. These impurities could have been avoided by intensive washing of the targets before measurement but in this particular instance they were taken into account in order to minimise the losses of glycopeptides. The glycane structures listed in Table IV can be derived from the results of the glycosidase breakdown (see above) and the measurements by mass spectrometer. The small peaks which can be seen in the region of the glycopeptides in the peptide map may originate from other glycosylation variants.

ii

H

I>

~LC-L~r

-I

'n I Table IV Peak' Peptide (amino acid number) measured Molecular mass (amu) 2 calculated (from ASsequence) 4.736 difference Proposed glycane structure Structure Mass 3 1/II 71-112 5.485 749 -GalNAc-Gal GlcNAc-Gal 2/Ia 2/IIa 2/IIb 84-112 71-112 71-112 3.975 5.448 5.132 3.304 4.736 4.736 671 712 396 -GalNAc-Gal-NeuAc 678 -GalNAc-Gai-NeuAc 678 -GalNAc-Gal 387 Table IV: Molecular masses of some glycopeptides of natural IFN-a2 proposed glycane structures. Peak numbers corresponding to Fig.

unit of mass; 3) Calculated mass including an Na -ion.

with the corresponding 19; amu, atomic i I 1 i;

I

i iP i 56 Example 13: Identification of the O-glycosylated amino acid by gas phase sequencing Since the glycopeptides obtained by cleaving with trypsin were too long for their entire sequence to be determined, these peptides were cleaved again by means of Staphylococcus aureus protease V8 and separated by RP-HPLC. A similar procedure was used with the corresponding peptides from E. coli IFN-a2c. After comparison of the peptide maps, all the peptides with different retention times were isolated and sequenced.

All the glycopeptides from natural IFN-a2 contained the amino acids 97-112. Whereas in E. coli IFN-a2c peptide 06 THR could be detected, it could not be found in the peptides obtained from natural IFN-a2. Thus, 1 06 THR was identified as the glycosylation site.

~CI"

i-/1 'ir

I

X

C

Claims (15)

1. Recombinant interferon alpha, O-glycosylated and having essentially the biological activity or immunological properties of an IFN-a2.

2. Recombinant interferon alpha according to claim 1, selected from 0-glycosylated human IFN-a2a, IFN-a2b and IFN-a2c.

3. Interferon alpha according to claim 1 or claim 2, wherein the threonine-106 (THR-106) is O-glycosylated.

4. Interferon alpha according to any one of claims 1 to 3, wherein the oligosaccharide is selected from the neutral disaccharide Gal-GalNAc, the mono- or disialylated variant thereof or the neutral tetrasaccharide Gal-(Gal-GlcNAc-)GalNAc. Recombinant human IFN-a2c according to any preceding claim.

6. A process for preparing recombinant O-glycosylated interferon alpha according to claim 1, which comprises the steps of: a) introducing a DNA coding for IFN-a into an expression plasmid which is suitable for the transfection of cells of multicellular organisms; b) transfecting cells of a multicellular organism with r the expression plasmid obtained in steps ti c) cultivating the transfected organisms obtained in step in a suitable medium; ~I 58 d) harvesting a cell supernatant; and e) isolating and purifying the O-glycosylated IFN-a in a manner known per se.

7. A process according to claim 6, wherein the expression plasmid used in step a) is pAD-CMV13, 15 or 19, according to Figures 3 and 4, and the DNA inserted at step a) codes for a protein which essentially has the biological activity or immunological properties of an IFN-a2.

8. A process according to claim 7, wherein the expression plasmid used in step a) is pAD-CMV19.

9. A process according to claim 7 or claim 8, wherein the IFN-a2 is selected from human IFN-a2a, IFN-a2b and IFN-a2c. A process according to any one of claims 6 to 9, wherein the expression plasmid used is pAD19B-IFN according to Figure

11. A process according to any one of claims 6 to wherein the cells of a multicellular organism are vertebrate cells.

12. A process according to claim 11, wherein the cells are CHO cells.

13. An expression plasmid for transfection of multicellular organisms, selected from pAD-CMV13, r« pAD-CMV15 and pAD-CMV19 according to Figures 3 and 4.

14. O-glycosylated recombinant interferon alpha, I t prepared according to any one of claims 6 to 12. -59 An agent for treating viral or tumoural diseases, containing a recombinant interferon alpha according to any one of claims 1 to 5 or 14.

16. An agent according to claim 15, comprising a mixture of at least two recombinant O-glycosylated proteins selected from IFN-a2a, IFN-a2b and IFN-a2c.

17. Recombinant interferon alpha according to claim 1 substantially as hereinbefore defined. j 18. A process as claimed in claim 6 substantially as hereinbefore defined.

19. An expression plasmid as claimed in claim 13 substantially as hereinbefore defined. An agent as claimed in claim 15 substantially as p hereinbefore defined. Dated this 9th day of March, 1993 BOEHRINGER INGELHEIM INTERNATIONAL GmbH By their Patent Attorneys DAVIES COLLISON CAVE i i C t t