Mapping the Active Site of CD59

All DNA manipulations were carried out in the mammalian expression vector pCDNA3 (Invitrogen, San Diego, CA) containing either CD59 or Ly6E cDNA subcloned into its multiple cloning site. pCDNA3 contains a G418 resistance marker. CD59 and Ly6E cDNA were kind gifts from Drs. H. Okada (Nagoya City University, Nagoya, Japan) and U. Hämmerling (Memorial SloanKettering Cancer Center, New York), respectively. Chinese hamster ovary cells (CHO) were used for protein expression and were maintained in Dulbecco's modified essential medium (DMEM) supplemented with 10% FCS. Stably transfected CHO cell lines were selected after the cultivation of cells in the presence of G418 (GIBCO BRL, Gaithersburg, MD).

Rabbit antisera to CHO cell membranes, CD59 and to CD59-specific peptide were prepared by standard techniques (19). CHO cell membranes were prepared as described (20). CD59 antigen was prepared from human erythrocytes as described (21). The peptide antigen corresponded to the COOH-terminal sequence of mature CD59 (CKKDLCNFNEQLE) and was conjugated to KLH for immunization. Peptide synthesis and KLH conjugation was performed by Genemed (South San Francisco, CA). Anti-CD59 peptide Ab was affinity purified by means of peptide immobilized onto CNBr-activated sepharose (Pharmacia, Piscataway, NJ). Anti-CD59 mAbs YTH53.1, MEM43/5, and HC1 were kindly provided by Dr. B.P. Morgan (University of Wales, Cardiff, UK). Each recognize nonoverlapping conformational epitopes on CD59 (22). mAb 2A10 (23) is directed against (NANP)_n, a repeat domain of Plasmodium falciparum circumsporozoite protein. FITC-conjugated Abs used for flow cytometry were purchased from Sigma Chem. Co. (St. Louis, MO). Normal human serum (NHS) was obtained from the blood of healthy volunteers in the laboratory and stored at −70°C.

Based on the alignment of CD59 and Ly6E amino acid sequences (Fig. 1), cDNA encoding the chimeric constructs depicted in Fig. 2 were prepared. Segments of either CD59 or Ly6E cDNA were generated and joined using PCR. The general procedure used for the generation of chimeric constructs was as described (24). The 5′ and 3′ end primers, which matched an untranslated sequence of either CD59 or Ly6E, included a HindIII and ApaI site, respectively. Using either CD59 or Ly6E cloned into pCDNA3 as template, these primers were paired in a PCR with CD59-Ly6E chimeric primers that spanned the desired crossover point. The PCR products were purified from agarose gel after electrophoresis and used in a second amplification with the primers corresponding to 5′ or 3′ untranslated regions. The resulting full-length CD59-Ly6E chimeric cDNAs were cloned into the HindIII/ApaI sites of pCDNA3 for sequencing and expression. Five site-specific point mutations in CD59 (N18Q, F23D, Y36F, W40Y, and L54N) were prepared by PCR using similar techniques. In the first PCR amplification, 5′ and 3′ primers to CD59 untranslated region were paired with primers spanning the target site, and which contained a substituted codon at the target site. An epitope tag was engineered into CD59-Ly6E chimeric proteins between the first and second amino acids of mature CD59. The NH₂-terminal Leu was duplicated at the COOH-terminus of the tag. A 36mer oligonucleotide encoding NANPNANPNA and with appropriate overhangs was inserted at the PstI restriction site of each point-specific mutant, and of CD59 and CD-Ly constructs (chimeras with CD59 at the NH₂ terminus).

pCDNA3 constructs were transfected into CHO cells that were 50–75% confluent. Transfections were carried out with lipofectamine according to the manufacturer's instructions (GIBCO BRL). 3 d after transfection, recombinant protein expression was confirmed, and G418 was added to the medium for selection of stably transfected populations. After 2 wk in selection medium, cells were fluorescently labeled by means of appropriate Abs and sorted by flow cytometry. Stable populations of CHO cells each expressing similar relative levels of recombinant proteins on their surface were isolated by repeated rounds of cell sorting using either the tag-specific mAb 2A10 (for CD59, CD-Ly chimeras, and point mutants) or antipeptide Ab (for Ly-CD proteins). Three rounds of sorting were usually sufficient to obtain uniform cell populations that were stable for several weeks. Stably transfected CHO cell clones expressing CD59, unglycosylated CD59 and NH₂ terminus tagged CD59 were prepared according to instructions supplied by GIBCO BRL with lipofectamine reagent.

Analysis of cell surface protein expression was performed by flow cytometry using a Beckton Dickinson FACScan^®. Detached CHO cells were incubated with appropriate Ab for 30 min at 4°C, washed by centrifugation, incubated with appropriate FITC-conjugated second Ab (30 min/4°C), and washed again before fixation with 2% paraformaldehyde in PBS. All incubations and washing were performed in DMEM/10% FCS. The following Abs and concentrations were used: rabbit antiCOOH-terminal CD59 peptide Ab (30 μg/ml); rabbit anti-CD59 antiserum (1:100); anti-CD59 mAb YTH53.1 IgG (10 μg/ml); antitag mAb 2A10 and anti-CD59 mAb MEM 43/5 ascites (1:250); anti-CD59 mAb HC1 culture supernatant (1:25). FITC-conjugated secondary Abs were used between 1:100 and 1:400 final concentration. Cells for sorting were fluorescently labeled following the procedure above but without fixation. Sorting was done in a Coulter Epics Elite with EPS sort module.

Subconfluent CHO cells were detached with versene (GIBCO BRL), washed once and resuspended to 1 × 10⁶/ml in DMEM/10% FCS. An equal volume of calcein-AM (5 μg/ml) (Molecular Probes, Eugene, OR) in DMEM was added and the cells incubated for 15 min at 37°C. The cells were then washed twice, resuspended to their original concentration in DMEM/10% FCS and incubated with rabbit anti-CHO cell membrane antiserum (20% final concentration) for 30 min on ice. The sensitized cells were then centrifuged, resuspended to their original concentration and an equal volume of NHS diluted in DMEM added. After 1 h at 37°C, the cells were centrifuged and cell lysis assessed by measuring released fluorescence in the supernatant. Measurements were performed in microtiter plates using a Labsystems fluoroscan II set to read at excitation 485 nm and emission 538 nm. Cells were lysed with 0.01% saponin for 100% lysis controls. Cell lysis assays were typically performed in 1.5-ml microfuge tubes in a final volume of 200 μl (1 × 10⁵ cells). Cell lysis assays were also performed using propidium iodide, which fluorescently stains dead cells. CHO cells were sensitized to complement as described and 100 μl cells (1 × 10⁶/ml) mixed with NHS dilutions. After 60 min at 37°C, propidium iodide was added (10 μg/ml final) and cell lysis determined by flow cytometry.

A CD59 analytical molecular surface, which is defined as the smooth envelope touching the van der Waals surface of atoms as the solvent probe rolls over the protein molecule, was built with the contour-buildup algorithm (25) as implemented in the ICM program (26). A probe sphere of 1.4 A was used. Molecular models of the point mutants were built and analyzed using the following procedures: (a) the structure of CD59 (pdb-code 1cds) solved by NMR (1) was carefully energy minimized; (b) appropriate residue modifications introduced; (c) the mutated side-chains were energy minimized using the biased probability Monte Carlo sample of the side-chain torsions (27), (d) interactions were analyzed by visual inspection using the ICM program.

Native CD59 is glycosylated at Asn18, and there are conflicting reports concerning its functional importance (28–30). We substituted Asn18 with Gln, and expressed the mutant protein in CHO cells. Fig. 3 shows that CHO cells expressing recombinant wild-type and unglycosylated CD59 are much more resistant to lysis by complement than mock transfected controls. Two stable CHO cell clones expressing unglycosylated CD59 were isolated, and each was matched with a CHO cell clone expressing a similar level of native CD59 for functional assays (Table 1). In both cases, cells expressing unglycosylated CD59 were more resistant to complement than cells expressing similar levels of native CD59. The difference was small but significant (P <0.05) over a range of serum concentrations. The expected reduction in MW of unglycosylated CD59 was confirmed by Western blotting of CHO cell membrane preparations (not shown).

To identify regions of CD59 important for function, a series of CD59/Ly6E chimeric proteins containing either CD59 or Ly6E at their NH₂-terminus were expressed on CHO cells. We reasoned that the residues involved in ligand binding were probably located in surface loops. For this reason, most of the chosen switch over points between CD59 and Ly6E were at selected loop boundaries (see Fig. 2).

CHO cell populations expressing similar levels of recombinant CD59 or chimeric proteins were assayed for their susceptibility to human complement. The functional data obtained with the CD-Ly chimeras (CD59 at NH₂ terminus) revealed that the region of CD59 which lies COOHterminal to residue 57 is not required for function; both CD-Ly 57 and CD-Ly 64 possessed the same specific activity as wild type CD59 (Fig. 4). The other CD-Ly chimeras which contained shorter CD59-specific sequences had no complement inhibitory activity. Of the Ly-CD chimeras (Ly6E at NH₂ terminus), only Ly-CD 16 possessed any functional activity (Fig. 4). The specific activity of Ly-CD 16 was ∼85% that of wild-type CD59. Taken together, these data indicate that the functionally important region of CD59 is located between amino acids 16 and 57 in the primary structure.

To quantitate the relative surface expression of CD59 and CD-Ly chimeric proteins, a 10–amino acid epitope tag was engineered onto the NH₂ termini of the constructs. The tag did not affect the expression (Table 1) or the function (Fig. 5) of CD59. However, when the tag was placed at the NH₂ terminus of Ly6E in the Ly-CD constructs, it interfered with expression and function of the chimeras. Therefore, surface expression of Ly-CD chimeras was quantitated using a rabbit Ab against a peptide corresponding to a sequence from the COOH terminus of CD59 (Table 1). This Ab reacted quantitatively with CHO cell clones stably expressing different levels of CD59 (not shown). Cell populations used in these experiments were obtained by several rounds of cell sorting, and separate CD-Ly and LyCD sets were prepared, each with a corresponding CHO cell population expressing a similar level of CD59.

To reveal the hypothetical location and extent of the functional site, we built a model of the molecular surface of CD59 and analyzed: (a) the three-dimensional location of the fragments that can be replaced by the Ly6E fragments, (b) the distribution of surface patches that are conserved between CD59 of seven different species that are either closely related to human, or that have been shown to inhibit human complement (Fig. 6), and finally, (c) the surface shape. A very clear picture emerged from this analysis. Fig. 7,a displays the molecule in an orientation that shows that the regions of CD59 that can be replaced by Ly6E without loss of function (1-16 and 57-77) are located at the backside of the molecule (in yellow, Fig. 7 a). From the position of residue 77, to which the GPI membrane anchor is attached, it can be predicted that the active site is located on the exposed face of the molecule.

Comparison of CD59 sequences within the region required for function (residues 16-57, Fig. 6) reveal a number of evolutionary conserved residues (Fig. 6). However, of these residues, only five are located on the surface of the shown face of the molecule, and three (Y36, W40, and L54) are conspicuously grouped together in a deep groove (Fig. 7,b, conserved residues colored green). Furthermore, conserved residue classes (Fig. 7,b, colored blue) are located above and below the conserved residues within this groove. From top to bottom these residue positions are (refer to Fig. 7,b): no. 31: Ala or Ser, a small residue, presumably important to keep the hypothetical binding groove open; no. 59: Leu or Val, a hydrophobic residue, outside the CD59 16-57 sequence, but conserved in Ly6E; no. 36: Tyr, no. 54 Leu and no. 40 Trp, conserved residues; no. 23: Phe or Leu, a hydrophobic residue, and no. 41: Lys or Arg, the only charged residues. The only addition to this list of conserved residues at the shown face of the molecule are Val 35 and two conserved Cys-Cys bridges (Cys19Cys39, and Cys6-Cys13, shown in yellow). Val 35 and Cys6-Cys13 are located at the inner surface of a tunnel, an interesting structural feature of the molecule (these residues are obscured by the tunnel in Fig. 7 b). The conserved cysteines are most likely necessary for conservation of the topology of the molecule.

The three conserved residues (shown in green, Fig. 7,b) and the conserved hydrophobic residue at position F23 within the putatively identified binding groove of CD59 were individually targeted for site-specific mutagenesis. The four mutant CD59s prepared were F23D (Phe at position 23 changed to Asp), Y36F, W40Y and L54N. The amino acids used for the substitutions were chosen to have similar shape but different physico-chemical properties. The W40Y mutant was not stably expressed on the surface of transfected CHO cells. Of the three expressed mutant proteins, L54N had almost no function, F23D retained some but significantly reduced function, and Y36F had normal function (Fig. 8). For unknown reasons, only relatively moderate levels of the L54N mutant protein could be expressed on CHO cells, but activity was compared with CHO cells expressing similar levels of wild-type CD59 (see Table 1). In an attempt to assess whether the mutant proteins were correctly folded, we studied their reactivity with three different mAbs that each bind to nonoverlapping conformational epitopes on CD59. All mutant proteins were recognized by these antibodies (Table 1).

The results in this report represent a first approximation at defining the active site of CD59. We have shown that: (a) glycosylation of CD59 at Asn18 is not required for function; (b) replacement of CD59 sequences 1-16 and 5777 with those of non-functional Ly6E did not affect CD59 activity. These regions are therefore not required for CD59 activity, and they map to the backside of the molecule that is closer to the cell surface (Fig. 7,a); (c) a comparison of sequences of human CD59 and CD59 homologues from other species within the identified functional region (amino acids 16-57), reveal an almost linear arrangement of conserved residues and residue classes in the three dimensional structure. These residues were predominantly hydrophobic in character, and were not contiguous in the primary structure. Together they marked a groove across the molecule (Fig. 7 b); (d) mutation of residues within this groove disrupted the function of the expressed mutant proteins.

Taken together, these data indicate that the active site of CD59 is on the front (facing away from membrane) face of the molecule and involves residues located along a hydrophobic groove defined by evolutionary conserved residues. In further support of this conclusion, it was recently shown that a correctly folded W40E mutant (residue located within hydrophobic groove, Fig. 7 b) expressed on the surface of CHO cells was inactive (Rushmere, N.K., D.L. Bodian, S.J. Davis, S. Tomlinson, and B.P. Morgan. 1996. XVIth International Complement Workshop. Boston, MA). In a previous report, Kieffer et al. (2) suggested that a relatively hydrophobic strip composed of exposed Tyr and Phe residues may constitute the functional site of CD59. Of note however, these residues are located on the opposite face and perpendicular to the hydrophobic groove identified in this report.

All but one of the Ly-CD and CD-Ly chimeric proteins expressed on CHO cells possessed either full or zero activity, making interpretation of data simple. The Ly-CD 16 chimera displayed slightly less (∼15%) activity than wild type CD59. The simplest explanation for this, taking into account the modeling and point mutation data, is that the NH₂-terminal sequence of Ly6E causes a change in the tertiary conformation of the active site of CD59. We cannot exclude however, that the NH₂-terminal sequence of CD59 may contribute to its activity.

The rationale in choosing the substituted residues in the CD59 point mutants was as follows: (a) F23D. Aspartate is smaller than phenylalanine, but has a consistent shape with branching at Cγ carbon; (b) Y36F. Phenylalanine is lacking the hydroxyl group of tyrosine, and based on its conservation, this group may be important; (c) W40Y. Tyrosine is smaller than tryptophane and the aromatic ring superimposes well onto the first ring of tryptophane; (d) L54N. Asparagine is very close to leucine in shape and both sidechains branch at Cγ, although the arrangement at Cγ is tetrahedral for leucine and flat for asparagine. Of the four point mutants prepared, one (Y36F) retained function, two (F23D and L54N) disrupted function but preserved protein topology, and one (W40Y) could not be stably expressed. By building an explicit model of this mutant we found that mutation of Trp40 to Tyr40 leads to a severe clash of the hydroxyl group with the backbone H alpha atom of Val50. This clash cannot easily be resolved and likely leads to the loss of tertiary structure. The near full functional activity of Y36F suggests either that this residue is not involved in ligand binding, or that ligand interaction is primarily with the phenol ring, rather than hydrogen bonding via the hydroxyl group. A further series of multiple residue mutations within and around the hydrophobic groove will better define the active site.

It is not clear what function the N-linked carbohydrate of CD59 may serve. It may be necessary for costimulation of T cell activation (31, 32); it has been reported that the ability of CD59 to enhance CD58-dependent T cell responses is dependent on the N-linked carbohydrate (33). Earlier reports are in conflict over the requirement of N-linked glycosylation for CD59 complement inhibitory function (28– 30). Here we confirm that glycosylation at Asn18 is not required for the inhibition of C5b-9 assembly on the cell surface. In fact, the absence of the carbohydrate enhances slightly, but significantly, the complement inhibitory activity of the protein. The mechanism for this increase in inhibitory activity is not clear, but it is intriguing that membrane sialic acid has recently been described as an acceptor for C5b6, thus enhancing C5b-9–mediated lysis (34). The absence of the sialylated carbohydrate of CD59 may therefore decrease, in a non-specific manner, C5b6 binding to the cell surface. Whatever the mechanisms involved, our data indicate that CD59 lacking N-linked carbohydrate yields an improved inhibitor, an important consideration in designing complement inhibitors for therapeutic applications and in engineering animal organs for human transplantation.

CD59 interferes with the final stages of MAC assembly either by binding to the α-chain of C8 in C5b-8 and blocking the binding of C9, by binding to C9 in the nascent C5b-9 complex and preventing proper membrane insertion and C9 polymerization, or by both mechanisms (3, 5, 6). If both mechanisms are operative, then the identification of a single functionally important site in CD59 would suggest that both C8α and C9 bind at this site. The putative CD59 binding site lies in the vicinity of a deep extended linear and hydrophobic groove. These characteristics are consistent with the ligand being an unfolded polypeptide chain with exposed hydrophobic residues, as is postulated to occur when C8 and C9 interact to form the MAC. Of possible relevance, the molecular surface contains a long tunnel under loop Asn8 to Cys13. There are two side chains facing the inner surface of the tunnel, the conserved Val35 as well as residues Asn, Ser, or Gln at position 72. This channel may represent a secondary binding site, but one that may not be required for its complement inhibitory function.

The intense current interest in inhibitors of complement is due to their potential use as anti-inflammatory agents and for preventing unwanted complement-mediated cytolysis. In addition to its role in hyperacute rejection of xenografts, there is evidence indicating that complement-mediated cytolysis is important in the pathogenesis of many autoimmune and inflammatory diseases (35). Understanding precisely how CD59 and other complement inhibitors function will provide a foundation for designing functional inhibitors, possibly multivalent and with enhanced inhibitory activity. The identification of the active site for CD59 may also enable the design of CD59 inhibitors, quite possibly with affinities higher than those of its natural ligands (36); CD59 and other membrane-bound inhibitors of complement are functionally expressed on tumor cells (37, 38), and inhibiting CD59 on tumor cells may enhance their lysis when targeted by complement activating tumor-specific antibodies.

We thank Dr. Michael Whitlow for critical discussion, Dr. John Hirst for performing flow cytometry and Dr. Jonathon Weider for the generation of photographs.

This work was supported by National Institutes of Health grants AI34451 and AI08499, and by a Grant in Aid from the American Heart Association.