Recombinant N-glycosylated proteins from procaryotic cells

Abstract

The present invention relates to recombinant N-glycosylated proteins, comprising one or more introduced N-glycosylated optimized amino acid sequence(s), nucleic acids encoding these proteins as well as corresponding vectors and host cells. In addition, the present invention is directed to the use of said proteins, nucleic acids, vectors and host cells for preparing medicaments. Furthermore, the present invention provides methods for producing said proteins.

Claims

1. A recombinant target protein comprising a recombinantly introduced optimized N-glycosylation consensus sequence, wherein said consensus sequence consists of the amino acid sequence of SEQ ID NO: 7, wherein said target protein has an increased or optimized efficiency of N-glycosylation compared to the same protein which does not comprise said consensus sequence, and wherein said optimized N-glycosylation consensus sequence is sufficient for N-glycosylation of said target protein in bacteria.

2. The recombinant target protein of claim 1, wherein said target protein comprises one or more N-glycans.

3. The recombinant target protein of claim 1, wherein said protein is Pseudomonas aeruginosa exotoxin, CRM197, or Cholera toxin.

4. The recombinant target protein of claim 3, wherein said target protein comprises one or more N-glycans.

5. The recombinant target protein of claim 1, wherein said target protein comprises at least two or at least three N-glycans.

6. The recombinant target protein of claim 1, wherein said protein is derived from a mammalian, viral, fungal, or plant protein.

7. The recombinant target protein of claim 1, wherein said protein is derived from a human protein.

8. A pharmaceutical composition comprising the recombinant target protein of claim 1 and a pharmaceutically acceptable excipient, diluents, and/or adjuvant.

9. A recombinant target protein comprising a recombinantly introduced optimized N-glycosylation consensus sequence, wherein said consensus sequence consists of the amino acid sequence of SEQ ID NO: 7, wherein said target protein is Pseudomonas aeruginosa exotoxin, CRM 197, or Cholera toxin.

10. The recombinant target protein of claim 9, wherein said protein is Pseudomonas aeruginosa exotoxin.

11. The recombinant target protein of claim 10, wherein said target protein comprises one or more N-glycans.

12. The recombinant target protein of claim 9, wherein said protein is CRM197.

13. The recombinant target protein of claim 12, wherein said target protein comprises one or more N-glycans.

14. The recombinant target protein of claim 9, wherein said protein is Cholera toxin.

15. The recombinant target protein of claim 14, wherein said target protein comprises one or more N-glycans.

16. The recombinant target protein of claim 9, wherein said target protein comprises one or more N-glycans.

17. The recombinant target protein of claim 9, wherein said target protein comprises three or more N-glycans.

18. A pharmaceutical composition comprising the recombinant target protein of claim 9 and a pharmaceutically acceptable excipient, diluents, and/or adjuvant.

19. A recombinant Pseudomonas aeruginosa exotoxin, CRM197, or Cholera toxin protein, comprising a recombinantly introduced optimized N-glycosylation consensus sequence consisting of SEQ ID NO: 7, wherein said recombinant Pseudomonas aeruginosa exotoxin, CRM197, or Cholera toxin protein has an increased or optimized efficiency of N-glycosylation compared to a Pseudomonas aeruginosa exotoxin, CRM197, or Cholera toxin protein which does not comprise a recombinantly introduced optimized N-glycosylation consensus sequence, and further wherein said consensus sequence is sufficient for N-glycosylation of said recombinant Pseudomonas aeruginosa exotoxin, CRM197, or Cholera toxin protein.

20. The recombinant Pseudomonas aeruginosa exotoxin, CRM197, or Cholera toxin protein of claim 19, comprising one or more N-glycans.

21. A recombinant target protein comprising a recombinantly introduced optimized N-glycosylation consensus sequence, wherein said consensus sequence consists of the amino acid sequence of SEQ ID NO: 7, wherein said target protein comprises at least three N-glycans.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 illustrates the N-glycosylation of Lip proteins derived from constructs A to C (see example 1). E. coli Top 10 cells carrying a functional pgl operon from C. jejuni (Wacker et al., 2002, supra) and a plasmid coding for constructs A (lane 2), B (lane 1), and C (lane 3) or a mutant of construct C with the mutation D121A (lane 4). Proteins were expressed and purified from periplasmic extracts. Shown is the SDS-PAGE and Coomassie brilliant blue staining of the purified protein fractions.

(2) FIG. 2 shows the N-glycosylation analysis of the different proteins that were analyzed for the sequence specific N-glycosylation by the C. jejuni pgl operon (Wacker et al., 2002, supra) in CLM24 cells (Feldman et al., (2005). Engineering N-linked protein glycosylation with diverse O antigen lipopolysaccharide structures in Escherichia coli. Proc. Natl. Acad. Sci. USA 102, 3016-3021) or Top10 cells (panel E lanes 1-6) or SCM7 cells (Alaimo, C., Catrein, I., Morf, L., Marolda, C. L., Callewaert, N., Valvano, M. A., Feldman, M. F., Aebi, M. (2006). Two distinct but interchangeable mechanisms for flipping of lipid-linked oligosaccharides. EMBO Journal 25, 967-976) (panel E, lanes 7, 8) expressing said proteins from a plasmid. Shown are SDS-PAGE separated periplasmic extracts that were transferred to Nitrocellulose membrane and visualized with specific antisera. In panels A-D the top panel show immunoblots probed with anti AcrA antiserum (Wacker et al. 2002, supra; Nita-Lazar, M., Wacker, M., Schegg, B., Amber, S., and Aebi, M. (2005). The N-X-S/T consensus sequence is required but not sufficient for bacterial N-linked protein glycosylation. Glycobiology 15, 361-367), whereas the bottom panels show immunoblots probed with R12 antiserum (Wacker et al., 2002, supra). + and indicate the presence of the functional or mutant pgl operon in the cells. Panel A contains samples of the soluble wildtype AcrA with the pelB signal sequence and the hexa histag (lanes 1, 2), AcrA-N273Q (lane 3, 4), and AcrA-D121A (lane 5). Panel B: AcrA (lanes 1, 2), AcrA-T145D (lane 3), AcrA-N123Q-N273Q-T145D (lanes 4, 5). Panel C: AcrA-F115D-T145D (lanes 1, 2), AcrA-N123Q-N273Q-N272D (lanes 3, 4). Panel D: AcrA-N273Q (lanes 1, 2), AcrA-N273Q-F122P (lanes 3, 4). Panel E: CtxB (lanes 1, 2), CtxB-W88D (lanes 3, 4), CtxB-Q56/DSNIT (lanes 5, 6), and CtxB-W88D-Q56/DSNIT.

(3) FIG. 3 shows the engineering of multiple glycosylation sites in OmpH1. The waaL strain SCM6 was co-transformed with plasmid pACYCpgl (encoding entire pgl locus) and plasmids expressing wild type OmpH1 (lane 1), OmpH1.sup.N139S-myc (lane 2), OmpH1.sup.KGN.fwdarw.NIT, HFGDD.fwdarw.DSNIT-myc (lane 3), OmpH1.sup.RGD.fwdarw.NIT, HFGDD.fwdarw.DSNIT-myc (lane 4), OmpH1.sup.KGN.fwdarw.NIT, RGD.fwdarw.NIT-myc (lane 5), OmpH1.sup.KGN.fwdarw.NIT, RGD.fwdarw.NIT, HFGDD.fwdarw.DSNIT-myc (lane 6) or OmpH1.sup.RGD.fwdarw.NIT,V83T-myc (lane 7). The cells were grown aerobically, induced with 0.5% arabinose for 3 hours prior to analysis. Whole cell lysates were TCA precipitated after equalizing the optical density of the cultures as described in the materials and methods section. The proteins were separated by 15% SDS-PAGE and transferred onto a PVDF membrane. First panel, immunoblot of whole cell lysates probed with anti-myc tag antobodies. Bottom panel, immunoblot of whole cell lysates probed with glycan-specific antiserum. The positions of unglycosylated- and glycosylated OmpH1 are indicated on the right.

(4) FIG. 4. Fluorescence microscopy of cells expressing various OmpH1 variants. Cultures of E. coli strains CLM24 or SCM6 containing the expression plasmid for the wild type OmpH1 and its variants were equalized to OD.sub.600 of 0.25/ml. Cells were washed two times with phosphate-buffered saline (PBS), pH 7.4 and 100 l cell suspensions was dropped onto gelatinized glass slides and incubated at room temperature (RT) for 30 min inside a humidified chamber. All subsequent steps in the whole-cell immunofluorescence labeling were done at room temperature inside a humidified chamber. The unbound cells were removed and rest was fixed with 4% paraformaldehyde containing PBS for 30 min at RT. Importantly, paraformaldehyde is considered not to permeabilize cells but keeping the compartimentalization by membranes intact. Fixed cells were washed two times with PBS and resuspended blocking buffer containing 5% BSA in PBS. After blocking, the cells were incubated with anti-myc monoclonal mouse IgG (1:50, Calbiochem) and/or anti-glycan antiserum (1:4000) for 1 h in 100 l of PBS containing 5% BSA. The cells were washed three times with 100 l of PBS for 5 min each and incubated with secondary anti-rabbit antibody conjugated to FITC (1:250, Jackson Immunoresearch Laboratories) and/or anti-mouse antibody conjugated to Cy3 (1:250, Jackson Immunoresearch Laboratories) for 1 h in 100 l of PBS containing 5% BSA. If required, 4,6-diamino-2-phenylindole (DAPI) (Sigma) (0.5 g/ml) was added at the time of secondary antibody incubation to stain for bacterial DNA. The secondary antibody was rinsed from the cells PBS, and coverslips were mounted on slides by using vectashield (Vector Laboratories) mounting medium and sealed with nail polish. Fluorescence microscopy was performed by the using an Axioplan2 microscope (Carl Zeiss). Images were combined by using Adobe Photoshop, version CS2. SCM6 cells expressing OmpH1 (panel A), OmpH1.sup.N139S (panel. B), OmpH1.sup.C20S (panel C), OmpH1.sup.KGN.fwdarw.NIT,HFGDD.fwdarw.DSNIT (panel D), OmpH1.sup.RGD.fwdarw.NIT,HFGDD.fwdarw.DSNIT (panel E), OmpH1.sup.KGN.fwdarw.NIT,RGD.fwdarw.NIT (panel F), OmpH1.sup.V83T,KGN.fwdarw.NIT (panel G), and OmpH1.sup.KGN.fwdarw.NIT,RGD.fwdarw.NIT,HFGDD.fwdarw.DSNIT (panel H). The first column is a merge of the pictures in columns 2, 3, and 4 represented in greytones on black background. Column 2: blue fluorescence in greytones from DAPI stain, column 3: green fluorescence from glycan specific fluorescence, column 4: red fluorescence from anti-myc staining.

(5) The following examples serve to illustrate further the present invention and are not intended to limits its scope in any way.

EXAMPLES

(6) Selection of AcrA as Model Protein for Optimizing N-Glycosylation

(7) To optimize the acceptor protein requirements for N-glycosylation detailed studies were performed on the C. jejuni glycoprotein AcrA (Cj0367c). AcrA is a periplasmic lipoprotein of 350 amino acid residues. It has been shown that secretion to the periplasm but not lipid-anchoring is a prerequisite for glycosylation (Nita-Lazar et al., 2005, supra). The signal for export can either be the native AcrA signal sequence or the heterologous PelB signal when expressed in E. coli. Of the five potential N-linked glycosylation sequons (N117, N123, N147, N273, N274) the same two ones are used in C. jejuni and E. coli (N123 and N273 (Nita-Lazar et al., 2005, supra)). AcrA was chosen as model because it is the only periplasmic N-glycoprotein of C. jejuni for which detailed structural information is available. Recently, the crystal structure of an AcrA homologue, the MexA protein from the Gram-negative bacterium P. aeruginosa, was published (Higgins et al., (2004). Structure of the periplasmic component of a bacterial drug efflux pump. Proc. Natl. Acad. Sci. USA 101, 9994-9999). Both proteins, are members of the so-called periplasmic efflux pump proteins (PEP, (Johnson, J. M. and Church, G. M. (1999). Alignment and structure prediction of divergent protein families: periplasmic and outer membrane proteins of bacterial efflux pumps. J. Mol. Biol. 287, 695-715)). The elongated molecule contains three linearly arranged subdomains: an -helical, anti-parallel coiled-coil which is held together at the base by a lipoyl domain, which is followed by a six-stranded -barrel domain. The 23-28 residues at the N-terminus and 95-101 residues in the C-terminus are unstructured in the crystals. MexA and AcrA protein sequences are 29.3% identical and 50% similar. Thus, the two proteins likely exhibit a similar overall fold.

Example 1

Elucidation of the Primary Peptide Sequence that Triggers Glycosylation

(8) It is known that lipoyl domains similar to MexA of P. aeruginosa and accordingly also in AcrA of C. jejuni form a compact protein that can be, individually expressed in E. coli (reviewed by Berg, A., and de Kok, A. (1997). 2-Oxo acid dehydrogenase multienzyme complexes. The central role of the lipoyl domain. Biol. Chem. 378, 617-634). To check which acceptor peptide sequence was required for N-glycosylation by the pgl machinery in E. coli the lipoyl domain of AcrA was taken. It was used as a molecular scaffold to transport peptides of different lengths to the periplasm and present them to the pgl machinery in vivo.

(9) Therefore, a plasmid coding for the lipoyl domain (Lip) was constructed and N-terminally fused to the signal sequence of OmpA (Choi, J. H., and Lee, S. Y. (2004). Secretory and extracellular production of recombinant proteins using Escherichia coli. Appl Microbiol Biotechnol 64, 625-635) and C-terminally to a hexa histag. Cloning was performed to place the gene expression under the control of the arabinose promoter. For the Lip domain borders amino acid positions were chosen that appeared at the same positions as the domain borders of the Lipoyl domain part in MexA. To test different peptides for their ability to accept an N-glycan stretches of the sequence were inserted between the two hammerhead-like parts of the Lip domain. The stretches consisted of sequences comprising the N-glycosylation site N123 of C. jejuni AcrA. The resulting open reading frames consisted of the sequences coding for the OmpA signal sequence, the N-terminal hammerhead-like part of AcrA (D60-D95, the numbering of the amino acids refers to the mature AcrA polypeptide sequence numbering), the different stretches containing the native N123 glycosylation site of AcrA (see below), the C-terminal hammerhead-like part of AcrA-Lip (L167-D210) and the C-terminal his-tag.

(10) Construction of the plasmids was achieved by standard molecular biology techniques. Three stretches containing the native N123 glycosylation site of AcrA of different lengths were inserted between the two halves of Lip resulting in three different ORFs:

(11) Construct A contains A118-S130 resulting in a protein sequence of:

(12) TABLE-US-00001 (SEQIDNO:1) MKKTAIAIAVALAGFATVAQADVIIKPQVSGVIVNKLFKAGDKVKKGQT LFIIEQDQASKDFNRSKALFSQLDHTEIKAPFDGTIGDALVNIGDYVSA STTELVRVTNLNPIYADGSHHHHHH.

(13) Construct B contains F122-E138 resulting in a protein sequence of:

(14) TABLE-US-00002 (SEQIDNO:2) MKKTAIAIAVALAGFATVAQADVIIKPQVSGVIVNKLFKAGDKVKKGQT LFIIEQDQFNRSKALFSQSAISQKELDHTEIKAPFDGTIGDALVNIGDY VSASTTELVRVINLNPIYADGSHHHHHH.

(15) Construct C contains D121-A127 resulting in a protein sequence of:

(16) TABLE-US-00003 (SEQIDNO:3) MKKTAIAIAVALAGFATVAQADVIIKPQVSGVIVNKLFKAGDKVKKGQT LFIIEQDQDFNRSKALDHTEIKAPFDGTIGDALVNIGDYVSASTTELVR VTNLNPIYADGSHHHHHH.

(17) The underlined stretches of sequence indicate the OmpA signal peptide, singly underlined residues were introduced for cloning reasons or to render the protein resistant to degradation. Bold: glycosylation site corresponding to N123 of AcrA. Italics: hexa-histag. The corresponding genes were expressed under the control of the arabinose promoter in the backbone of the plasmid pEC415 (Schulz, H., Hennecke, H., and Thony-Meyer, L. (1998). Prototype of a heme chaperone essential for cytochrome c maturation. Science 281, 1197-1200).

(18) To check which of the three stretches triggered glycosylation of the Lip proteins protein expression experiments were performed. E. coli Top10 cells (Invitrogen, Carlsbad, Calif., USA) carrying pACYCpgl or pACYCpglmut (Wacker et al., 2002, supra) and a plasmid coding constructs A, B or C were grown in LB medium containing ampicillin and chloramphenicol up to an OD of 0.5 at 37 C. For induction 1/1000 volume 20% arabinose (w/v) solution was added and the cells were grown for another 2 hrs. The cells were then harvested by centrifugation and resuspended in 20 mM Tris/HCl, pH 8.5, 20% sucrose (w/v), 1 mM EDTA, 1 mM PMSF, and 1 g/l (w/v) lysozyme and incubated at 4 C. for 1 hr. Periplasmic extracts were obtained after pelletting of the spheroblasts and diluted with 1/9 volume (v/v) of 10 buffer A (3 M NaCl, 0.5 M Tris/HCl, pH 8.0 and 0.1 M imidazole) and MgSO.sub.4 added to 2.5 mM. Ni-affinity purification was performed on 1 ml Ni-Sepharose columns from Amersham Pharmacia Biotech (Uppsala, Sweden) in buffer A. Proteins were eluted in buffer A containing 0.25 M imidazole.

(19) FIG. 1 shows Coomassie brilliant blue stained SDS-PAGE gel of the peak elution fractions from the Ni-purified periplasmic extracts. The expression analysis showed that construct B produced a prominent single protein species (FIG. 1, lane 1). Constructs A and C both lead, in addition to the prominent protein, to a second protein band with slower electrophoretic mobility (FIG. 1, lanes 2 and 3). That the heavier protein species was indeed glycosylated was proven by MALDI-TOF/TOF (not shown). The only amino acid missing in construct B but present in A and C was D121, the aspartate residue 2 positions N-terminally to the glycosylated N123. This demonstrates that D121 plays an important role for glycosylation by the OTase. To verify that D121 is essential for glycosylation it was mutated to alanine in construct C. Expression analysis resulted in only one protein band (FIG. 1, lane 4), thus showing that D121 is important for glycosylation. Furthermore, the fact that an artificial peptide display protein can be glycosylated shows that a short peptide of the D/E-X-N-Z-S/T (SEQ ID NO:7) type contains all information for C. jejuni-borne N-glycosylation to occur.

Example 2

Verification of Example 1: AcrA-D121A is not Glycosylated at N123

(20) To confirm the findings from the peptide display approach an aspartate to alanine mutation was inserted at position 121 (D121A, i.e. 2 residues before the glycosylated N123) in the full length soluble version of the AcrA protein and it was tested whether the site N123 could still be glycosylated in E. coli. In order to test this AcrA-D121A was expressed and its glycosylation status was analyzed. For the analysis an engineered AcrA was used. It differed from the original C. jejuni gene in that it contains the PelB signal sequence (Choi and Lee, 2004, supra) for secretion into the periplasm and a C-terminal hexa histag for purification. It has been shown that this AcrA variant gets secreted, signal peptide-cleaved and glycosylated as the lipid anchored, native protein (Nita-Lazar et al., 2005, supra). The following is the protein sequence of the soluble AcrA protein:

(21) TABLE-US-00004 (SEQIDNO:4) MKYLLPTAAAGLLLLAAQPAMAMHMSKEEAPKIQMPPQPVTTMSAKSED LPLSFTYPAKLVSDYDVIIKPQVSGVIVNKLFKAGDKVKKGQTLFIIEQ DKFKASVDSAYGQALMAKATFENASKDFNRSKALFSKSAISQKEYDSSL ATFNNSKASLASARAQLANARIDLDHTEIKAPFDGTIGDALVNIGDWSA STTELVRVTNLNPIYADFFISDTDKLNLVRNTQSGKWDLDSIHANLNLN GETVQGKLYFIDSVIDANSGTVKAKAVFDNNNSTLLPGAFATITSEGFI QKNGFKVPQIGVKQDQNDVYVLLVKNGKVEKSSVHISYQNNEYAIIDKG LQNGDKIILDNFKKIQVGSEVKEIGAQLEHHHHHH

(22) The underlined residues are the PelB signal peptide, italics the hexa-histag, and bold the two natural glycosylation sites at N123 and N273. A plasmid containing the ORF for the above protein in the pEC415 plasmid (Schulz et al., 1998) was constructed to produce pAcrAper.

(23) The assay to test the glycosylation status of AcrA and mutants thereof (see below) was as follows: expression of AcrA was induced with 0.02% arabinose in exponentially growing E. coli CLM24 (Feldman et al., 2005, supra) cells containing the plasmid-borne pgl operon in its active or inactive form (pACYCpgl or pACYCpglmut, see (Wacker et al., 2002, supra)) and a plasmid coding for AcrA (pAcrAper). After four hours of induction, periplasmic extracts were prepared as described above and analyzed by SDS-PAGE, electrotransfer and immunodetection with either anti-AcrA antiserum or R12 antiserum. The latter is specific for C. jejuni N-glycan containing proteins (Wacker et al., 2002, supra).

(24) The first two lanes of FIG. 2A show AcrA in the absence and presence of a functional pgl operon. Only one band appears in the absence but three in the presence of the functional pgl operon (FIG. 2A, top panel). These correspond to unglycosylated AcrA (lane 1) and un-, mono- and diglycosylated AcrA (lane 2). That the two heavier proteins in lane 2 were glycosylated was confirmed by the R12 western blot (lane 2, bottom panel). When the mutant AcrA-N273Q was expressed the same way, only the monoglycosylated AcrA was detected in presence of the functional glycosylation pgl operon (lane 3). Unglycosylated AcrA was detected in absence of the functional pgl locus (lane 4). Analysis of the mutant AcrA-D121A produced only two bands, one of them glycosylated (lane 5) as observed with AcrA-N273Q in lane 3. This means that D121 is essential for efficient glycosylation at position 123-125.

Example 3

Introducing Artificial Glycosylation Sites into AcrA

(25) To test if the introduction of an aspartate residue could generate a glycosylation site, AcrA mutants were generated in which the residue in the 2 position of the not used glycosylation sites in positions N117 and N147 of soluble AcrA were exchanged for aspartate (F115D, T145D). It was then tested whether the modified glycosylation sites could be glycosylated by the same assay as described in example 2. Both mutations were individually inserted either into the wildtype sequence of the soluble version of AcrA or in the double mutant in which both used glycosylation sites were deleted (N123Q and N273Q). Periplasmic extracts of cultures induced for 4 hrs were prepared, separated by SDS page and analyzed by Western blotting (FIG. 2B). As controls the samples of wildtype glycosylated and non glycosylated AcrA were run on the same gel (lanes 1 and 2). The T145D mutation affected the -2 position of the natively not used glycosylation sequon N147-S149. Upon expression of AcrA-T145D Western blotting with anti AcrA antiserum resulted in four bands, the highest of them with slower electrophoretic mobility than the doubly glycosylated protein in lane 2 (lane 3 in FIG. 2B). The R12 blot confirmed that the fourth band was a triply glycosylated AcrA. Despite the low intensity towards anti AcrA the heaviest band gave the strongest signal with the glycosylation specific R12 antiserum. When the same mutant AcrA-T145D was expressed in the absence of the native N-glycosylation sequence (AcrA-N123Q-N273Q-T145D), only monoglycosylated AcrA was detected in the presence of a functional pgl operon (FIG. 2B, lane 4), that was missing in absence of a functional pgl operon (lane 5). This demonstrates that the heavier band in lane 4 was glycosylated. Hence, by simply introducing the T145D mutation an optimized glycosylation site was generated (DFNNS, SEQ ID NO: 8).

(26) To further confirm that it is possible to introduce a glycosylation site by inserting an aspartate residue in the 2 position, the natively not used sites N117-S119 and N274-T276 were changed to optimize N-glycosylation. For this purpose further mutants were generated (FIG. 2C). Expression of AcrA-F115D-T145D in the above described system, resulted in five protein species detected with the anti AcrA antiserum (lane 2). This is indicative for four glycosylations taking place on the same AcrA molecule. When the detection was performed with the C. jejuni N-glycan-specific R12 antiserum, a ladder of five bands was detected. The lowest faint band is unglycosylated AcrA because it is also present in the absence of glycosylation (lane 1), the highest results in a strong signal probably due to the five antigenic determinants in a fourfold glycosylated AcrA. Thus, the two introduced sites (at N117 and N147) and the two natively used sites (N123 and N273) are used and glycosylated by the pgl machinery. Expression of AcrA-N123Q-N273Q-N272D with and without the pgl operon demonstrated that a third artificially introduced glycosylation site, N274 (DNNST; SEQ ID NO:9), was also recognized by the pgl operon (FIG. 2C, lanes 3 and 4).

(27) The above experiments confirm the finding that the bacterial N-glycosylation site recognized by the OTase of C. jejuni consists partly of the same consensus as the eukaryotic one (N-X-S/T, with XP) but, in addition, an aspartate in the 2 position is required for increasing efficiency. Furthermore, they demonstrate that it is possible to glycosylate a protein at a desired site by recombinantly introducing such an optimized consensus sequence.

Example 4

Verification of Position 1 in the Optimized N-Glycosylation Sequence

(28) A further experiment was performed to test whether the 1 position in the bacterial glycosylation site exhibits the same restrictions as the +1 position in eukaryotes (Imperiali, B., and Shannon, K. L. (1991). Differences between Asn-Xaa-Thr-containing peptides: a comparison of solution conformation and substrate behaviour with oligosaccharyl-transferase. Biochemistry 30, 4374-4380; Rudd, P. M., and Dwek, R. A. (1997). Glycosylation: heterogeneity and the 3D structure of proteins. Crit. Rev. Biochem. Mol. Biol. 32, 1-100). A proline residue at +1 is thought to restrict the peptide in such a way that glycosylation is inhibited. To test if a similar effect could also be observed in the 1 position a proline residue was introduced at that position of the first natively used site in a point mutant that had the second native site knocked out (AcrA-N273Q-F122P). The control expression of AcrA-N273Q showed a monoglycosylated protein in the presence of a functional pgl operon (FIG. 2D, lane 1 and 2). However, AcrA-N273Q-F122P was not glycosylated (FIG. 2D, lanes 3 and 4). This indicates that proline inhibited bacterial N-glycosylation when it constitutes the residue between the asparagine and the negatively charged residue of the 2 position.

(29) Sequence alignments of all the sites known to be glycosylated by the C. jejuni pgl machinery indicate that they all comprise a D or E in the 2 position (Nita-Lazar et al., 2005, supra; Wacker et al., 2002, supra; Young et al., (2002). Structure of the N-linked glycan present on multiple glycoproteins in the Gram-negative bacterium, Campylobacter jejuni. J. Biol. Chem. 277, 42530-42539). Thus, it was established that the glycosylation consensus sequence for bacteria can be optimized by a negatively charged amino acid in the 2 position, resulting in D/E-X-N-Z-SIT (SEQ ID NO:7), wherein X & ZP.

Example 5

N-Glycosylation of a Non-C. jejuni Protein

(30) To demonstrate that the primary sequence requirement (optimized consensus sequence) is sufficient for N-glycosylation in bacteria, it was tested whether a non-C. jejuni protein could be glycosylated by applying the above strategy. Cholera toxin B subunit (CtxB) was employed as a glycosylation target. The corresponding gene was amplified from Vibrio cholerae in such a way that it contained the coding sequence of the OmpA signal sequence on the N-terminus and a hexahistag at the C-terminus, just the same as constructs A through C in example 1. The resulting DNA was cloned to replace construct A in the plasmids employed in example 1. A point mutation of W88 to D or a D insertion after W88 generated an optimized glycosylation site (DNNKT; SEQ ID NO:10). The wildtype and W88D CtxB proteins containing the signal sequence and his-tag were expressed in E. coli Top 10 and other cell types in the presence and absence of the functional pgl locus from C. jejuni. When periplasmic extracts from Top 10 cells were analyzed by SDS-PAGE, electrotransfer and consecutive immunoblotting with a CtxB antiserum, only CtxB W88D produced a higher and thus glycosylated band in the pgl locus background (FIG. 2E, compare lanes 3 and 4). A consensus sequence (DSNIT; SEQ ID NO:11) was also inserted by replacing G54 or Q56 of CtxB (the latter is denoted CtxB-Q56/DSNIT (SEQ ID NO:11)), i.e. in one of the loops that was reported to contribute to the ganglioside GM1 binding activity of CtxB. Lanes 5 and 6 of FIG. 2E demonstrate that the engineered protein (exemplified by the construct which contains the peptide sequence DSNIT (SEQ ID NO:11) instead of Q56 expressed in Top10 cells) produced a lower mobility and thus glycosylated band in glycosylation competent but not glycosylation-deficient cells when analyzed in the same way as described above. It was also demonstrated that a CtxB containing two manipulations, i.e. the insertion of D after W88 as well as DSNIT (SEQ ID NO:11) replacing Q56, was double-glycosylated in SCM7 cells (Alaimo et al., EMBO Journal 25: 967-976 (2006)) (panel E, lanes 7 and 8). The double-glycosylated protein CtxB shown in lane 7 was Ni.sup.2+ affinity-purified and analyzed by ESI-MS/MS after in-gel trypsinization according to standard protocols. The expected glycopeptides were detected confirming that bacterial N-glycosylation can also be directed to a non-C. jejuni protein by mutating or inserting the optimized consensus sequence according to the invention for bacterial N-glycosylation (not shown). Examples of other suitable exemplary E. coli strains for practicing the present invention are W3110, CLM24, BL21 (Stratagene, La Jolla, Calif., USA), SCM6 and SCM7.

(31) The amino acid sequence of the CtxB protein used here is indicated below (recombinant OmpA signal sequence underlined, hexa-histag italics, W88 bold):

(32) TABLE-US-00005 (SEQIDNO:5) MKKTAIAIAVALAGFATVAQATPQNITDLCAEYHNTQIHTLNDKIFSYT ESLAGKREMAIITFKNGATFQVEVPGSQHIDSQKKAIERMKDTLRIAYL TEAKVEKLCVWNNKTPHAIAAISMANGSHHHHHH

Example 6

Introduction of Artificial N-Glycosylation Sites into the C. jejuni Outer Membrane Protein, OmpH1

(33) A potential application of the N-glycosylation in bacteria is the display of the glycan on the surface of a bacterial host cell in order to link the pheno- to the genotype and thereby select for specific genetic mutations. To demonstrate that N-glycans can be presented on outer membrane proteins the OmpH1 protein was engineered in a way that it contained multiple optimized consensus sites according to the invention. The sites were engineered into loop regions of the protein as deduced from the known crystal structure (Muller, A., Thomas, G. H., Horler, R., Brannigan, J. A., Blagova, E., Levdikov, V. M., Fogg, M. J., Wilson, K. S., and Wilkinson, A. J. 2005. An ATP-binding cassette-type cysteine transporter in Campylobacter jejuni inferred from the structure of an extracytoplasmic solute receptor protein. Mol. Microbiol. 57: 143-155). Previous experiments showed that the best glycosylation sequons were generated by the mutations V83T, K59N-G60I-N61T, R190N-G191I-D192T and H263D-F264S-G265N-D266I-D267T. For surface display it was desired to evaluate different combinations of those introduced sites in order to establish the most. N-glycan-specific sample. The combinations were generated in a wild type OmpH1 encoding plasmid construct and tested in a similar manner as described for AcrA. FIG. 3 shows the analysis of various OmpH1 variants harboring multiple glycosylation sequons in addition to the existing wild type sequon. OmpH1 variants were generated with three (lane 3, 4, 5 and 7) and four glycosylation sequons (lane 6). A wild type OmpH1 with only one glycosylation sequon and a mutant lacking the critical asparagine for glycosylation were also included in the experiment. All variants tested here did not only demonstrate a high level of glycosylation efficiency but also that every glycosylation sequon was utilized. The results were confirmed with Campylobacter N-glycan specific immuneserum (FIG. 3 lower panel).

(34) The following is the protein sequence of the OmpH1 protein of Campylobacter jejuni (strain 81-176) with attached myc tag in italics:

(35) TABLE-US-00006 (SEQIDNO:6) MKKILLSVLTTFVAVVLAACGGNSDSKTLNSLDKIKQNGVVRIGVFGDK PPFGYVDEKGNNQGYDIALAKRIAKELFGDENKVQFVLVEAANRVEFLK SNKVDIILANFTQTPERAEQVDFCLPYMKVALGVAVPKDSNITSVEDLK DKTLLLNKGTTADAYFTQDYPNIKTLKYDQNTETFAALMDKRGDALSHD NTLLFAWVKDHPDFKMGIKELGNKDVIAPAVKKGDKELKEFIDNLIIKL GQEQFFHKAYDETLKAHFGDDVKADDVVIEGGKILEQKLISEEDL

(36) The native glycosylation site in the protein is bold, the signal sequence underlined.

Example 7

Surface Display of N-Glycans from C. jejuni on OmpH1 on the Outer Membrane of E. coli Cells

(37) In order to answer the question whether multiple glycosylated OmpH1 variants can be displayed on the surface of bacterial cells, immunofluorescence was performed on bacterial CLM24 or SCM6 (which is SCM7 waaL) cells expressing various OmpH1 variants. A wild type OmpH1 and a mutant lacking the critical asparagine for glycosylation were included in the experiment. In addition, a C20S mutant was constructed in order to retain the protein in the periplasm, thus serving as a control in the experiment. Immunostaining was carried out on the cells treated with paraformaldehyde. Paraformaldehyde fixes cells without destroying the cell structure or compartmentalization. The c-Myc- and N-glycan-specific immuneserum in combination with corresponding secondary antibodies conjugated to FITC and Cy3 were used to detect the protein (red fluorescence) and N-glycan (green) on the bacterial cell surface, respectively. Additionally, 4,6-diamino-2-phenylindole (DAPI, blue) was employed to, stain for bacterial DNA to unambiguously differentiate between bacterial cells and cellular debris. When the cells expressing wild type OmpH1 were stained, immunofluorescence specific to the protein as well as the N-glycan was detected (FIG. 4A). When a mutant lacking the critical asparagine N139S was stained with both anti-Myc- and N-glycan-specific immuneserum only the protein but not glycan specific signals were obtained (panel 4 B) indicating specificity of the N-glycan-specific immune serum. When the protein was retained within the periplasm as in the C20S mutant, no protein specific, red immunofluorescence was detected indicating that the antibodies were unable to diffuse within the cell and were competent enough to detect any surface phenomenon (panel 4 C). Next, cells expressing multiple OmpH1 variants different in glycosylation were stained: OmpH1.sup.KGN.fwdarw.NIT,HFGDD.fwdarw.DSNIT (panel 4 D), OmpH1.sup.RGD.fwdarw.NIT,HFGDD.fwdarw.DSNIT (panel 4 E), OmpH1.sup.KGN.fwdarw.NIT,RGD.fwdarw.NIT (panel 4 F), OmpH1.sup.V83T,KGN.fwdarw.NIT (panel 4 G) and OmpH1.sup.KGN.fwdarw.NIT,RGD.fwdarw.NIT,HFGDD.fwdarw.DSNIT (panel 4 H). All the OmpH1 variants were double-stained indicating the presence of glycosylated protein on the bacterial surface. FIG. 4 is represented in grayscale, the first column is a merge picture of the other pictures of the same row.