Focused Meeting held at New Royal Infirmary, Edinburgh, U.K., 21 November 2005. Organized and edited by J.-M. Sallenave and J. Govan (Edinburgh, U.K.).
Structural and functional studies of defensininspired peptides D.J. Clarke and D.J. Campopiano1 School of Chemistry, University of Edinburgh, West Mains Road, Edinburgh EH9 3JJ, Scotland, U.K.
Abstract Mammals have evolved complex self-defence mechanisms to protect themselves from infection. This innate immune system comprises a large family of hundreds of peptides and proteins which have potent antibiotic activity at nanomolar concentrations. The defensins are a group of small cationic peptides which contain a high proportion of positively charged and hydrophobic amino acids. Their exact mechanism of antimicrobial action is unclear, but it is thought that the defensins bind to and disrupt the outer cell membrane which ultimately causes lysis and cell death. They are characterized by six conserved cysteine residues which oxidize to form three intramolecular disulphide (S–S) bonds. The human and mouse defensins have been subdivided into classes based on their sequence, site of expression and the S–S bond connectivity of the cysteine residues. α-Defensins are connected by cysteines 1 and 6, 2 and 4, and 3 and 5, whereas β-defensins have a 1–5, 2–4 and 3–6 cysteine S–S connectivity. We present our structural and functional studies of a novel mouse β-defensin-related peptide (Defr1) which contains only five cysteine residues. Synthetic Defr1 was more active than its six-cysteine analogue against a large panel of pathogens. High-resolution MS techniques revealed that Defr1 contains an unusual defensin structure. These studies have guided the design of novel peptides to explore the roles of defensins as antibiotics and as stimulants of the immune response.
Introduction The mammalian innate immune system produces a large array of weapons to protect the host from the constant attack which it is under from micro-organisms [1]. One important family
Key words: antimicrobial peptide, defensin, mass spectrometry. Abbreviations used: AMP, antimicrobial peptide; CID, collision-induced dissociation; CRS, cryptdin-related sequence; Defr1, β-defensin-related peptide 1; DTT, dithiothreitol; ECD, electroncapture dissociation; ESI-MS, electrospray ionization MS; FT-ICR, Fourier-transform ion cyclotron resonance; HBD, human β-defensin; LPS, lipopolysaccharide; MALDI, matrix-assisted laserdesorption ionization; MBC, minimum bactericidal concentration; mBD, mouse β-defensin; RPHPLC, reverse-phase HPLC; S3, Sushi 3. 1 To whom correspondence should be addressed (email
[email protected]).
Antimicrobial Peptides Biochemical Society Focused Meeting
Antimicrobial Peptides: Mediators of Innate Immunity in the Development of Anti-Infective, Therapeutic and Vaccination Strategies
is the cationic AMPs (antimicrobial peptides), a vast arsenal of small polypeptides which kill Gram-positive, Gramnegative, fungal and viral pathogens within minutes of contact and display an extraordinary range of primary structure along their amino acid sequence [2]. The AMP can be wholly geneencoded or a fragment of a larger protein, and renewed interest in them has arisen due to the search for novel agents to fight the rise of multi-resistant bacteria [3]. The mode of killing of each particular AMP is dependent on the correct balance of hydrophobic and positively charged amino acids (‘amphipathicity’) which allows them to interact with the membrane and cause cell lysis [4,5]. However, it has been C 2006
Biochemical Society
251
252
Biochemical Society Transactions (2006) Volume 34, part 2
Figure 1 Sequences of β-defensins and related peptides Amino acids in italics are proposed and may not appear in the mature peptides. Cysteine residues are indicated in bold.
discovered that there are already pathogens which have developed resistance to these ancient highly evolved weapons by modification of components of the membrane target [6]. The defensins are gene-encoded AMPs which have interested researchers since their discovery and characterization over 20 years ago [7–9]. Mammalian defensins have been classified into α, β and θ subfamilies, based on their gene organization, cellular location, expression profile, cysteine spacing and disulphide (S–S) bond connectivity. The term defensin was first used by Ganz and co-workers over 20 years ago as the name given to peptides (of the α-class) that they extracted from human neutrophils [10,11]. In addition to potent bactericidal activity, they can also act upon T-lymphocytes and immature dendritic cells, thus playing key roles in adaptive immunity [12]. This article will concentrate on the β-defensin subclass. They are produced as prepropeptides and are processed to a mature secreted form which has six conserved cysteine residues with spacing and intramolecular S–S bridge connectivity (by cysteines 1 and 5, 2 and 4, and 3 and 6) that is distinct from the α-defensin class (which have a 1–6, 2–4 and 3–5 cysteine S–S connectivity) (Figure 1). Isolation of antimicrobial peptides from biological samples (e.g. skin, saliva, plasma, semen and vaginal fluid) still follows a similar laborious, low-yielding process involving many chromatographic steps similar to that carried out in 1985. The first human β-defensin (HBD1) was isolated from 4800 litres of human plasma where it is present in nanomolar concentrations [13]. The authors were able to isolate the peptide by cation-exchange, then sequential C4 and C18 RPHPLC (reverse-phase HPLC). Amino acid analysis and ESIMS (electrospray ionization MS) of the purified species showed it to have high sequence identity with the prev C 2006
Biochemical Society
iously isolated BNBD-1 (bovine neutrophil β-defensin 1). Degenerate DNA primers were designed based on the sequence of the isolated peptide fragments and the cDNA of the human HBD1 gene was cloned. The sequence of the encoded protein revealed a six-cysteine-containing peptide with an overall charge of +5. A second human β-defensin was isolated from human skin in 1997, and the sequence of HBD2 was determined from its cloned cDNA [14]. Basal expression of HBD2 was detected in a few tissues, but it was noted that it could be greatly up-regulated in skin, lung and trachea after exposure to pathogens. The third human β-defensin (HBD3) was isolated from psoritic skin, and the sequence revealed it to be the most cationic of the β-defensins (overall charge +11) [15]. A fourth β-defensin (HBD4) was found to be expressed in the testis, stomach and uterus [16,17]. To date, only six human α-defensins have been isolated [18]. The recent completion of the human and mouse genome sequences allowed a bioinformatics search for new β-defensin genes, and 28 human and 43 mouse genes were identified on a stretch of chromosome 8 [19]. Another defensin gene cluster was identified on chromosome locus 8p23.
Three-dimensional structures of defensins The structure of the first member of the human α-defensin family (HNP-3; human neutrophil peptide-3) was determined in 1991 [20], but the first HBD structures were resolved only recently by NMR and X-ray crystallography [21–25]. The structures of HBD1, HBD2 and HBD3 show that the peptides fold into a central three-stranded β-sheet similar to the α-defensin [21–25]. In some cases, the N-terminus forms an α-helical domain, but this appears to be dependent on the length of the peptide and the structural method used
Antimicrobial peptides
(a representative β-defensin structure, HBD3, is shown in Figure 3A). Studies have shown that specific β-defensin mutants, as well as peptide fragments of each of the defensins, have antimicrobial activity, suggesting that the three-dimensional fold is not absolutely required for antimicrobial activity [26]. The natural full-length folded forms are more resistant to proteolysis by proteases that are present in bacterial and mammalian cells. Defensins also trigger further adaptive immune responses by binding to G-protein-coupled receptors, but the specificity of protein–protein interactions involved is still unclear [8,26].
Defensin-related peptide Defr1 is a covalent dimer A previous study of Defr1 found in the genome of a C57BL6 mouse revealed that it contained only five cysteine residues (Figure 1), but retained potent antimicrobial activity [27]. This peptide is coded by a variant allele of Defb8 that encodes six cysteines and is found in all other inbred murine strains tested. In the present work, we undertook a full comparative structural and functional study of Defr1 against its six-cysteine homologue, Defr1 Y5C [28]. Synthetic Defr1 and Defr1 Y5C were tested for antimicrobial activity against a panel of Gram-positive, Gram-negative and fungal pathogens. We found the MBCs (minimum bactericidal concentrations) of Defr1 ranged between 3 and 10 µg/ml. Upon reduction of the S–S bonds by incubation with excess DTT (dithiothreitol), the activity of reduced Defr1 was lowered to the range 25–100 µg/ml. In contrast, the six-cysteine analogue Defr1 Y5C displayed MBCs in the range 50–100 µg/ml, both in the reduced and oxidized forms. Our conclusion was that the presence of S–S bonds did not influence the activity of the six-cysteine Defr1 Y5C, in agreement with the results of Wu et al. [26] from their studies with isoforms of HBD3. However, the MBCs of reduced Defr1 were significantly higher than that of its oxidized form, so we concluded that, in this peptide, the S–S bonds did play a role in the activity. Furthermore, we also analysed the effect of NaCl on the activity of Defr1 and Defr1 Y5C specifically against Pseudomonas aeruginosa PAO1 cells, since many AMPs display salt sensitivity due to disruption of the ionic interactions between the peptide and the membrane target. The activity of the sixcysteine analogue was diminished >90% in the presence of 25 mM NaCl and was completely abolished in 50 mM NaCl. In stark contrast, the Defr1 peptide was fully active in 25 and 50 mM NaCl and was ∼80% and ∼30% active in 150 and 300 mM NaCl respectively. To rationalize the striking differences in activity between Defr1 and Defr1 Y5C, we investigated the structure of the peptides by high-resolution MS, native gel electrophoresis and chromatography. The defensins are particularly amenable to ESI-MS since they ionize well because of their high lysine and arginine content. The isotopic resolution and accuracy of our FT-ICR (Fourier-transform ion cyclotron resonance) MS instrument allowed us to measure the mass of the peptides and compare that with the mass predicted by the amino acid sequence. This effectively allowed us to distinguish be-
tween peptides with or without S–S bonds, i.e. between two species differing by 2 Da. Analysis of the ion envelope and deconvolution (Figure 2) for Defr1 Y5C suggested the presence of two species: a monomer containing three S–S bonds (elemental composition C157 H254 N50 O43 S6 , average mass 3722.4490 Da) and a dimeric Defr1 Y5C with six S–S bonds (C314 H508 N100 O86 S12 , average mass 7444.898 Da) (Figures 2A and 2B). The monomer/dimer ratio was 6:1, suggesting that the dimer is not stable to the electrospray process and is held together by weak non-covalent interactions. FT-ICR analysis of oxidized Defr1 produced a mass spectrum which, upon deconvolution, gave rise to one major species whose isotopic distribution matched that expected for a fully oxidized dimer containing five S–S bonds (Defr1 dimer C326 H518 N100 O88 S10 , average mass 7566.9761 Da) (Figure 2C). Interestingly, there was no peak corresponding to the Defr1 monomer. In contrast with Defr1 Y5C, the Defr1 dimer remains intact under electrospray conditions, which implies a strong interaction between monomers. Since a Defr1 dimer contains ten cysteine residues, and a fully oxidized isoform has five S–S bonds, dimerization can only occur through formation of at least one intermolecular S–S bridge.
Characterization of the Defr1 and Defr1 Y5C dimers To explore further the structural differences in the two defensin dimers, we used two fragmentation techniques [CID (collision-induced dissociation) and ECD (electron-capture dissociation)] to characterize the protein–protein interactions that are involved [29]. For CID analysis of the Defr1 Y5C dimer, a peak corresponding exclusively to a dimer ion was isolated and subjected to dissociation by gas. The selected ion readily dissociated into two monomers. Dissociation of the dimer occurs without fragmentation of the peptide backbone, demonstrating that the Defr1 Y5C dimer is unstable, thereby supporting our hypothesis that dimerization is mediated through non-covalent interactions. In stark contrast, the Defr1 dimer was stable to the same CID conditions used for Defr1 Y5C. When a stable dimeric ion was isolated and subjected to dissociation, no significant monomeric ‘daughter’ fragments were observed. Increasing the amount of gas in the collision cell still did not dissociate the dimer, but did allow us to partially sequence the peptide. Such stability under CID conditions indicates that the Defr1 dimer is held together by covalent bonding. These observations are complemented by ECD, where cleavage of Cys–Cys S–S bridges is known to be a favoured process. An isolated Defr1 dimer ion was subjected to ECD, and the molecule readily dissociated into monomers. On closer inspection, we observed low-intensity species with masses 16 Da either side of the monomer peak, which indicate the loss or gain of a sulphur atom. We can explain the appearance of these species only if the Defr1 dimer is formed by a covalent intermolecular S–S bond. Here, the ECD process has cleaved the dimer into monomers by two mechanisms. The first is symmetrical cleavage of the S–S bond C 2006
Biochemical Society
253
254
Biochemical Society Transactions (2006) Volume 34, part 2
Figure 2 FT-ICR ESI-MS of deconvoluted isotopic envelope of Defr1 Y5C monomer (A) and dimer (B) and Defr1 dimer (C) Black triangles are the predicted isotopic envelope pattern.
to give a monomer signal. The alternative pathway involves asymmetric cleavage of the C–S bond of the intermolecular S–S bridge to give rise to two monomers: one with a persulphide thiol and its corresponding partner that has lost a sulphur atom. We probed further the structures of Defr1 and Defr1 Y5C using conventional mass mapping techniques of proteolytic digest/MALDI (matrix-assisted laser-desorption ionization) mass fingerprinting to determine the S–S connectivity of the two peptides. Samples of oxidized Defr1 Y5C were eluted as a sharp band by RP-HPLC and were cleaved readily by chymotrypsin and trypsin into fragments which were isolated by RP-HPLC and analysed further by MALDI-MS and by CID sequencing on a Q-Tof (quadrupole time-of-flight) instrument. The fragmentation pattern and masses were consistent with Defr1 Y5C having β-defensin connectivity. This was not surprising since Defr1 Y5C shows high sequence identity with the mouse β-defensin mBD-8, the structure of which has been determined by NMR, and so we suggest mBD-8 as a good model for the three-dimensional structure of Defr1 Y5C (PDB code 1E4R). In contrast, proteolytic cleavage of the oxidized Defr1 dimer produced a complex mixture of peptides that we were unable to separate by HPLC. We analysed the Defr1 by RP-HPLC and found that it was eluted as a broad over 10 min. Combining these data, C 2006
Biochemical Society
we concluded that Defr1 is a mixture of dimeric isoforms with a range of S–S connectivities both within and between each monomer. A schematic model of a single isoform is shown in Figure 3(B). During the synthesis and oxidation of Defr1 and Defr1 Y5C, the S–S connectivities were not controlled – under equilibrium conditions, Defr1 Y5C oxidized to form a single β-defensin isoform, whereas the odd-numbered Defr1 folded into a mixture. A similar observation was made by Wu et al. [26] during folding studies of HBD1, HBD2 and HBD3. They noted that both reduced HBD1 and HBD2 were oxidized into single β-defensin isoforms with high yield; however, HBD3 formed a complex mixture of isoforms of undetermined S–S connectivity. To control the S–S bond formation of HBD3, Wu et al. [26] orthogonally protected pairs of cysteines and introduced the S–S bonds sequentially. Interestingly, this S–S scrambling was not reported by others who have folded HBD3. The exact mechanism and coenzymes involved in folding of β-defensins in vivo are not known, but could presumably involve various S–S bond isomerases [30].
Other covalent AMP dimers At the same time as we discovered that the Defr1 dimer contained a novel intramolecular S–S bond, a number of other defensin-related peptides with this structural feature were reported and we discuss these below.
Antimicrobial peptides
Figure 3 Three-dimensional structure of a β-defensin representative and model of Defr1 (A) X-ray structure of HBD3 showing three β-sheets, α-helix and three S–S bonds. (B) Schematic representation of Defr1 dimer with five S–S bonds (four intramolecular and one intermolecular).
Rat 2D6 A rich source of AMPs has been the mammalian reproductive system, although, as well as acting simply as defence molecules in maintaining a disease-free environment, their full roles in these organs are unclear. An interesting protein, named 2D6 (Figure 1), was isolated by Jones and co-workers from rat epididymis [31]. This 24 kDa protein, so-called since it reacted with a monoclonal antibody 2D6, was found to interact with the plasma membrane of rat spermatozoa during sperm maturation. Sequencing of fragments of the isolated protein enabled the cloning of the gene and prediction of the encoded peptide. The protein (111 amino acids) showed 100% sequence identity with rat epididymal protein E3 and high identity with human protein ESP13.2. They all show weak similarity to the β-defensin family, especially in the core region of six cysteines, and of interest to us was that 2D6 contains a seventh cysteine residue at the C-terminus. During their biochemical studies, Zanich et al. [31] noted that the isolated oxidized 2D6 behaved as a mixture of a two forms by electrophoresis, a 250 kDa complex and a 48 kDa dimer, and as a monomer upon reduction with DTT. They concluded that the extra cysteine mediated intermolecular dimer formation. 2D6 is also glycosylated, which explains why it migrates with a much larger apparent mass than predicted for the mature form. Future studies of this and related proteins will hopefully reveal what roles these S–S-linked defensin-like dimers play in controlling sperm morphology.
CRS (cryptdin-related sequence) peptides A novel family of AMPs named CRS peptides were isolated recently from mouse intestinal tissue which had potent antibacterial activities, as well as unusual structural features [32]. The mouse cryptdin genes were initially characterized by Ouellette and Lualdi [33] at the genomic level over 15 years ago. Hornef et al. [32] initially isolated 17 mRNA species which encoded seven different CRS peptides with high sequence identity. They then purified several of these peptides by standard non-reducing RP-HPLC procedures and analysed their structure by MS. They found that all of the
isolated peptides were covalent dimers, 38 amino acids in length and containing nine invariant cysteine residues, as well as seven Cys-Pro-Xaa repeat motifs (Xaa = Gln/Arg/Ser) along their sequence length. The CRS dimers each contained four S–S intramolecular bonds within each monomer and an intermolecular S–S bond mediating dimer formation. As well as CRS homodimers, heterodimers were also formed between the seven CRS peptide monomers. An interesting evolutionary proposal for maintaining the ‘odd’ cysteine residue is that it allows intermolecular S–S dimer formation which leads to the generation of an increased library of 28 different AMPs from seven monomers. The exact structure and function of these dimeric CRS peptides awaits further study.
Sushi peptides Many AMPs are fragments that are derived from larger proteins and one such peptide is the 34-amino-acid Sushi 3 (S3) peptide derived from the S3 domain of Factor C, which is the LPS (lipopolysaccharide)-sensitive serine protease of the horseshoe crab coagulation cascade. The antimicrobial properties of this peptide were shown recently to be due to its interaction with LPS from Gram-negative bacteria, and the S3–LPS interaction was absolutely dependent on the dimeric form of the S3 peptide [34]. Since S3 contains a single cysteine residue, the function of this peptide is crucially mediated by a single intermolecular S–S bond between the monomers. This simplest S–S structural motif allows the development of a range of optimized dimers with enhanced properties.
Conclusions Our structural work has shed light on the outcome of the deletion of the first conserved cysteine residue of a β-defensinrelated peptide, Defr1. The five-cysteine Defr1 peptide was active against a panel of pathogens, and its activity was higher than that of a six-cysteine β-defensin analogue where this residue had been replaced. High-resolution structural C 2006
Biochemical Society
255
256
Biochemical Society Transactions (2006) Volume 34, part 2
studies revealed that the Defr1 peptide was a mixture of dimeric isoforms, covalently-linked by a single intermolecular S–S bond. Finding this S–S-linked structural motif in novel peptides from various sources, each with interesting functions, encourages us to explore further and exploit its role. Why is S–S-linked dimerization used by the peptides? It could be simply that it doubles the number of charged and hydrophobic residues of a peptide target. It could lead to the ability to form stable heterodimeric species with other partners (as in the CRS peptides). We noted that Defr1 is more resistant to proteolysis than its Defr1 Y5C analogue, and this may be advantageous in vivo. We are currently exploring the structures of defensin-related dimers to better understand their interaction with their cellular targets. We thank all the members of Team Defensin, past and present for their expertise and enthusiasm: Julia Dorin, Ross Langley, John Govan, Dusan Uhrin, Derek MacMillan, Karen Taylor, Nick Polfer, Bryan McCullough and Perdita Barran.
References 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Boman, H.G. (2000) Immunol. Rev. 173, 5–16 Zasloff, M. (2002) Nature (London) 415, 389–395 Walsh, C. (2000) Nature (London) 406, 775–781 Brogden, K.A. (2005) Nat. Rev. Microbiol. 3, 238–250 Shai, Y. (2002) Biopolymers 66, 236–248 Peschel, A. (2002) Trends Microbiol. 10, 179–186 Ganz, T. (2003) Nat. Rev. Immunol. 3, 710–720 Selsted, M.E. and Ouellette, A.J. (2005) Nat. Immunol. 6, 551–557 Lehrer, R.I. (2004) Nat. Rev. Microbiol. 2, 727–738 Ganz, T., Selsted, M.E., Szklarek, D., Harwig, S., Daher, K., Bainton, D.F. and Lehrer, R.I. (1985) J. Clin. Invest. 76, 1427–1435 Selsted, M.E., Harwig, S.S., Ganz, T., Schilling, J.W. and Lehrer, R.I. (1985) J. Clin. Invest. 76, 1436–1439 Yang, D., Biragyn, A., Kwak, L.W. and Oppenheim, J.J. (2002) Trends Immunol. 23, 291–296 Bensch, K.W., Raida, M., Magert, H.-J., Schulz-Knappe, P. and Forssmann, W.G. (1995) FEBS Lett. 368, 331–335 Harder, J., Bartels, J., Christophers, E. and Schroder, J.-M. (1997) Nature (London) 387, 861
C 2006
Biochemical Society
15 Harder, J., Bartels, J., Christophers, E. and Schroder, J.M. (2001) J. Biol. Chem. 276, 5707–5713 16 Garc´ıa, J.R., Krause, A., Schulz, S., Rodr´ıguez-Jimenez, ´ F.J., Kluver, ¨ E., Adermann, K., Forssmann, U., Frimpong-Boateng, A., Bals, R. and Forssmann, W.G. (2001) FASEB J. 15, 1819–1821 17 King, A.E., Fleming, D.C., Critchley, H.O. and Kelly, R.W. (2003) J. Reprod. Immunol. 59, 1–16 18 Ericksen, B., Wu, Z., Lu, W. and Lehrer, R.I. (2005) Antimicrob. Agents Chemother. 49, 269–275 19 Schutte, B.C., Mitros, J.P., Bartlett, J.A., Walters, J.D., Jia, H.P., Welsh, M.J., Casavant, T. and McCray, P.B. (2002) Proc. Natl. Acad. Sci. U.S.A. 99, 2129–2133 20 Hill, C.P., Yee, J., Selsted, M.E. and Eisenberg, D. (1991) Science 251, 1481–1485 21 Hoover, D.M., Chertov, O. and Lubkowski, J. (2001) J. Biol. Chem. 276, 39021–39026 22 Hoover, D.M., Rajashankar, K.R., Blumenthal, R., Puri, A., Oppenheim, J.J., Chertov, O. and Lubkowski, J. (2000) J. Biol. Chem. 275, 32911–32918 23 Schibli, D.J., Hunter, H.N., Aseyev, V., Starner, T.D., Wiencek, J.M., McCray, P.B., Tack, B.F. and Vogel, H.J. (2002) J. Biol. Chem. 277, 8279–8289 24 Sawai, M.V., Jia, H.P., Liu, L.D., Aseyev, V., Wiencek, J.M., McCray, P.B., Ganz, T., Kearney, W.R. and Tack, B.F. (2001) Biochemistry 40, 3810–3816 25 Bauer, F., Schweimer, K., Kluver, E., Conejo-Garcia, J.R., Forssmann, W.G., Rosch, P., Adermann, K. and Sticht, H. (2001) Protein Sci. 10, 2470–2479 26 Wu, Z., Hoover, D.M., Yang, D., Boulegue, C., Santamaria, F., Oppenheim, J.J., Lubkowski, J. and Lu, W. (2003) Proc. Natl. Acad. Sci. U.S.A. 100, 8880–8885 27 Morrison, G.M., Rolfe, M., Kilanowski, F.M., Cross, S.H. and Dorin, J.R. (2002) Mammal. Genome 13, 445–451 28 Campopiano, D.J., Clarke, D.J., Polfer, N.C., Barran, P.E., Langley, R.J., Govan, J.R., Maxwell, A. and Dorin, J.R. (2004) J. Biol. Chem. 279, 48671–48679 29 Barran, P.E., Polfer, N.C., Campopiano, D.J., Clarke, D.J., Langridge-Smith, P.R.R., Langley, R.J., Govan, J.R., Maxwell, A., Dorin, J.R., Millar, R.P. and Bowers, M.T. (2005) Int. J. Mass Spec. 240, 273–284 30 Sevier, C.S. and Kaiser, C.A. (2002) Nat. Rev. Mol. Cell Biol. 3, 836–847 31 Zanich, A., Pascall, J.C. and Jones, R. (2003) Biol. Reprod. 69, 1831–1842 32 Hornef, M.W., Putsep, K., Karlsson, K., Refai, E. and Andersson, M. (2004) Nat. Immunol. 5, 836–843 33 Ouellette, A.J. and Lualdi, J.C. (1990) J. Biol. Chem. 265, 9831–9837 34 Li, P., Wohland, T., Ho, B. and Ding, J.L. (2004) J. Biol. Chem. 279, 50150–50156 Received 13 January 2006
