Generation of a recombinant, membrane-targeted form of the complement regulator CD59: activity in vitro and in vivo.

Inappropriate activation of complement contributes to pathology in diverse inflammatory diseases. Soluble recombinant forms of the natural cell membrane regulators of complement are effective in animal models and some human diseases. However, their use is limited for reasons related to cost, short half lives, and propensity to cause unwanted systemic effects. Some of these limitations may be overcome by use of bacterial expression systems, specific targeting moieties, and judicious choice of regulator. Here we describe the application of these strategies to the generation of a membrane-targeted form of CD59. A recombinant soluble form of rat CD59, comprising the first 71 residues of the mature protein and missing the membrane-anchoring signal, was expressed in bacteria, purified, and refolded in a fully active form. The protein was coupled through its carboxyl terminus to a short, synthetic address tag that confers membrane binding activity. Attachment of the membrane address tag markedly increased complement-inhibitory activity assessed in vitro in hemolysis assays. Intra-articular administration of the tagged agent markedly suppressed disease in a model of rheumatoid arthritis in Lewis rats. This novel type of agent, termed sCD59-APT542, offers for the first time the prospect of efficient and specific inhibition of membrane attack complex activity in vivo.

The complement (C) 1 system, a key player in innate immunity, also plays a pathological role in many inflammatory diseases (1). As a consequence, controlling inappropriate activation of C may be an appropriate therapeutic strategy in these diseases (2). Numerous approaches to the development of agents that regulate C activation have evolved, and some are now reaching the clinic. Two agents have led the way. The first, a soluble recombinant form of the membrane regulator C re-ceptor 1 (CR1; CD35), has proven beneficial in many animal models of human disease, although preliminary results in human disease have been disappointing (3,4). The second, a humanized monoclonal antibody that blocks cleavage of C5, more recent in development, is likely to be the first anti-C agent in clinical use (5,6). Although both of these agents have been shown to be effective and relatively safe in models, it must be stressed that systemic inhibition of C is not without consequence. Agents that efficiently inhibit C in vivo may cause iatrogenic infections or immune complex disease, probably limiting their use to acute situations (7).
The problems of systemic inhibition of C might be overcome either by targeting the agent to a specific site or by choosing an agent that permits the physiological actions of C to proceed unhindered. For the latter strategy, agents that specifically inhibit formation of the lytic membrane attack complex (MAC) are attractive, because MAC is implicated as an important cause of pathology, whereas most of the physiological roles of C are mediated by products of the activation pathways (8). Indeed, individuals deficient in components of the MAC are healthy apart from an increased susceptibility to meningococcal disease (9). CD59 is a small, broadly distributed, glycosylphosphatidylinositol-anchored protein that is the sole membrane regulator of MAC assembly (10). Recombinant soluble forms of CD59 (sCD59) have been generated with the aim of developing a specific inhibitor of MAC for therapy (11)(12)(13). However, sCD59 is a poor inhibitor of MAC both in vitro and in vivo. In vitro, sCD59 inhibits efficiently in "reactive lysis" systems where purified components are used to assemble the MAC on targets, but this inhibitory activity is markedly reduced in the presence of serum (11,14). In vivo, the problem is compounded by the small size of sCD59, which permits rapid clearance in the kidney. Attempts have been made to generate more active forms of sCD59 by designing chimeras or fusion proteins (15,16), but no clear evidence of MAC-inhibiting activity in vivo has yet been reported. We have taken a different approach to the problem of targeting and retaining C therapeutics by utilizing a membrane-associating "tag" that can be coupled to proteins. Several C regulators have previously been modified in this manner, particularly a small molecule derived from human sCR1 and a portion of the rat C regulator Crry (17,18). In each case, the addition of the membrane tag markedly increased the C regulatory capacity of the parent molecule in vitro and, in the case of the sCR1 fragment, also markedly enhanced activity in animal models. The targeting technology has been coupled with development of bacterial expression systems for these complex proteins that enable large amounts of active protein to be generated economically. CD59 inhibits MAC assembly by incorporating tightly into the forming complex (10,19). The molecule is held together by four disulfide bonds that maintain a rigid, compact, discoid structure. Soluble recombinant forms of CD59 have been expressed in mammalian cells, insect cells, and yeast and shown to be functional in "reactive lysis" assays in vitro (11,20,21). Despite its small size, expression of CD59 in a bacterial system has not previously been reported and represents a considerable challenge, given the complex disulfide bridging and tertiary structure. Here we describe the production in Escherichia coli of a recombinant soluble form of rat CD59, modified such that an additional Cys residue is available at the carboxyl terminus. This molecule was fully active as an inhibitor of MAC assembly when compared with soluble rat CD59 expressed in mammalian cells. The addition of a membrane address tag at the carboxyl terminus generated a membrane binding molecule that had markedly increased MAC-inhibitory activity in vitro when compared with the untailed molecule. Intra-articular administration of the membrane-tagged rat CD59 in rats efficiently inhibited disease progression and joint destruction in a rat arthritis model. These data demonstrate for the first time that specific inhibition of MAC can inhibit pathology in a C-mediated disease model.
Normal human serum was obtained by venepuncture from healthy volunteers. Normal rat serum and guinea pig erythrocytes were obtained from the local animal facility. Monoclonal mouse anti-rat CD59, 6D1, was made in house (22). Rabbit polyclonal anti-reduced human CD59 antiserum (cross-reactive with rat CD59) was provided by Dr. S. Tomlinson (Charleston, SC). C8-depleted serum was made in house using standard methods (23). Soluble recombinant forms of rat CD59, comprising the amino-terminal 71 amino acids, either unmodified or mutated to delete the single N-glycosylation site, were made in CHO cells essentially as previously described (13). Antiserum against rat C9, cross-reactive with MAC in rat tissues, was made by repeated immunization of rabbits with purified rat C9. The antibody was purified from antiserum and biotinylated using standard protocols. Fluorescein isothiocyanate-labeled avidin was obtained from Sigma.

Construction of Soluble Recombinant Rat CD59 cDNA Bacterial Expression Vector
Using as template a plasmid containing the full coding sequence for rat CD59, a cDNA encoding a soluble recombinant form of rat CD59 (residues 1-71 of the mature protein sequence, missing the glycosylphosphatidylinositol anchor addition sequence) and including at the carboxyl terminus a 7-residue spacer (SGGSGGS) and a terminal cysteine, was engineered and amplified (sCD59-Cys). The primers used were as follows: 1) 5Ј-GTTCCACAGGTCATATGCTCAGATGC-3Ј, which added an NdeI restriction site (boldface type) immediately upstream of the nucleotides encoding the initiating methionine of the recombinant protein; 2) 5Ј-CGCGGATCCTTAGCAAGAACCGCCA-GAACCACCGGATTTGTTACACAAGTCCGCCTG-3Ј, which included the spacer and terminal Cys upstream of a stop codon (italic type) and a BamHI restriction site (boldface type). The PCR product (266 bp) was subcloned into the plasmid TOPO-TA® (Invitrogen) and sequenced to confirm fidelity. The plasmid was then digested with NdeI and BamHI to release the insert, which was purified and ligated into the bacterial expression vector pET26b (Invitrogen) digested at the corresponding sites.

Bacterial Expression of sCD59-Cys by Fermentation
The sCD59-Cys cDNA in pET26b was transformed into electrocompetent E. coli BL21 (DE3) bacteria (Novagen, Nottingham, UK) by electroporation (GenePulser™; Bio-Rad). Cells were then plated onto LB agar containing kanamycin (50 g/ml) to select for clones containing the plasmid. Positive colonies were picked and expanded, and the presence of insert was confirmed by PCR screening. A single positive colony was picked into 25 ml of LB broth containing kanamycin (50 g/ml) and grown overnight at 37°C in a shaking incubator. A 20-ml aliquot of this starter culture was inoculated into 2 liters of NZCYM medium containing kanamycin (50 g/ml) in a Bioflo 3000 Bioreactor (New Brunswick Scientific, Edison, NJ) with a 2-liter bioreactor culture vessel. The fermenter was prepared and run as described previously (18). Cultures were grown for 4 h until the bacteria were in their log phase of growth (A 600 ϭ 5-8). Protein expression was then induced by adding sterile filtered isopropyl ␤-D-thiogalactopyranoside to a final concentration of 1 mM. The fermentation culture was harvested at 3 h post-induction and centrifuged at 10,000 ϫ g for 10 min. Cell pellets were stored at Ϫ40°C.
Conditions for refolding of sCD59-Cys were chosen empirically by testing a large panel of solutions differing in their pH, presence of reducing and denaturing agents, presence of detergents, etc. as previously described (18,24). Successful refolding was initially assessed by demonstrating the acquisition of epitopes recognized by the conformationdependent mAb 6D1 in dot blots. In the chosen protocol, sCD59-Cys was refolded from solubilized inclusion bodies by rapid dilution (1:40) into 50 mM Tris, 1 M NaCl, 1 mM reduced glutathione, 3 mM oxidized glutathione, pH 8.0. The refold mixture was incubated for 1 h at 4°C and then buffer-exchanged into PBS and concentrated 10-fold by ultrafiltration (Amicon stirred cell; 5-kDa cut-off membrane; Amicon Inc., Beverley, MA). Aggregated and/or misfolded protein was removed by ammonium sulfate precipitation. Protein in PBS was diluted 1:2.5 into 3.8 M ammonium sulfate, 0.1 M sodium phosphate buffer, pH 6.5, incubated for 10 min at room temperature, and centrifuged. The superna- tant was retained and dialyzed into PBS prior to measurement of protein concentration.
The molecular mass of the final product was measured using matrixassisted laser desorption and ionization time-of-flight spectroscopy (MALDI-TOF; Bruker, Coventry, UK). The identity of the final product was confirmed by N-terminal sequencing (Applied Biosystems Procise sequencer).

Modification of sCD59-Cys with a Membrane Address Tag
A membrane address tag was added to sCD59-Cys by derivatization at the free carboxyl-terminal cysteine with the sulfydryl-reactive peptide, N-(myristoyl)GSSKSPSKKKKKKPGDC-(S-2-thiopyridyl) C-amide (termed APT542; Adprotech Ltd.) (25), to produce sCD59-APT542. The tagged protein was separated from untagged and purified to homogeneity by cation exchange chromatography on a Mono-S (Amersham Biosciences) column. Protein was loaded in PBS and eluted with a salt gradient to 1 M NaCl in PBS. Fractions containing sCD59-APT542 were identified by SDS-PAGE and pooled.

SDS-PAGE and Immunoblotting
Samples were resolved by SDS-PAGE using 4 -12% Bis-Tris gradient gels (Novex) according to the manufacturer's instructions. Protein bands were visualized by staining with Coomassie Blue R-250. For Western blotting, proteins were transferred to nitrocellulose membranes (Schleicher & Schuell) using the Bio-Rad miniblot system. Membranes were probed with the relevant antibodies and developed using the Western Breeze Kit (Novex).

Hemolysis Assays
Two types of hemolysis assay were used to assess the function of sCD59-Cys and sCD59-APT542.
Nonwash Assays-Guinea pig E (GPE) (2% in CFD) were incubated (15 min, 37°C) with C8-depleted human serum diluted 1:5 in CFD in order to generate C5b-7 sites on the cells (GPE5b-7). The GPE5b-7 were washed and resuspended to 2% in PBS plus 10 mM EDTA, and 100-l aliquots were delivered to wells of a 96-well round-bottomed plate. Dilutions of rat serum in PBS plus 10 mM EDTA (100 l) were then added to individual wells as a source of C8 and C9 and incubated for 30 min at 37°C. Zero lysis controls contained no rat serum, and 100% controls contained 1% Triton in place of serum. Plates were spun, and supernatant was removed to a fresh 96-well flat-bottomed plate. Absorbance was read in a Bio-Rad enzyme-linked immunosorbent assay reader at 410 nM, and percentage of hemolysis at each serum dilution was calculated relative to the zero and 100% controls. From this titration study, a serum dose that caused approximately 70% lysis of GPE5b-7 was chosen. GPE5b-7 aliquoted as above were incubated with the selected serum dilution together with dilutions in PBS plus 10 mM EDTA of sCD59-Cys, sCD59-APT542, or, as control, sCD59 generated in CHO cells. Supernatants were harvested, and hemolysis was measured as described above. The MAC-inhibitory effect at different concentrations of each of the CD59 constructs were assessed by comparing the hemolysis obtained in the presence of CD59 with that in the absence of CD59. All measures were made in triplicate.
Wash Assays-GPE5b-7 prepared as described above were resus-pended to 2% in PBS plus 10 mM EDTA and aliquoted into the wells of a 96-well round-bottomed plate. Dilutions in PBS plus 10 mM EDTA of sCD59-Cys, sCD59-APT542, or, as control, sCD59 generated in CHO cells, were then added (100 l) and incubated for 30 min at 37°C. Cells were then washed three times by centrifugation in PBS plus 10 mM EDTA prior to the addition of the selected dilution of rat serum to develop hemolysis, and the MAC-inhibitory effect was measured as described above.

Induction and Treatment of Antigen-induced Arthritis (AIA)
Male Lewis rats (approximately 200 g) were obtained from Charles River (Margate, UK) and housed at Biomedical Services (University of Wales College of Medicine, Cardiff). Rats were allowed free access to food and water and kept in light/dark cycles of 12 h. AIA was induced following an established protocol (17). Briefly, animals were injected subcutaneously with an emulsion of equal volumes of methylated bovine serum albumin (0.5 mg/ml) and Freund's complete adjuvant (containing 0.25 mg/ml heat-killed Mycobacteria) on two occasions a week apart. Fourteen days after the second injection (day 0), an injection of methylated bovine serum albumin (0.1 mg in 100 l of saline) was administered into the right knee of each animal. The left knee served as a control joint and received an equal volume of saline.
For assessment of therapeutic effects, AIA was induced in groups of six animals, groups receiving either sCD59-Cys (250 g in 0.1 ml of PBS), sCD59-APT542 (250 g in 0.1 ml of PBS), or the same volume of PBS alone (control) in a single dose given into the joint together with the disease-initiating antigen on day 0. Disease was assessed clinically  1, 4, and 7), the pellet (lanes 2, 5, and 8), and supernatant (lanes 3, 6, and 9) were subjected to SDS-PAGE on 4 -12% Bis-Tris gradient gels (Novex) under nonreducing (A and C) or reducing (B) conditions, and a Western blot was carried out using polyclonal anti-reduced CD59 antiserum (blots A and B) and mAb 6D1 anti-rat CD59 (blot C). Molecular weight markers are shown on each blot (m), and sizes in kDa are indicated on the left. The arrow indicates the anticipated size of the sCD59-Cys protein. by measuring the knee diameters of inflamed (right) and noninflamed (left) knees with a Mitutoyo digital caliper over a 14-day period. On each occasion, three readings were taken from each joint by an independent observer blinded to the treatment regimen. The swelling attributed to the antigenic challenge was expressed as the difference (in mm) between the mean readings of the inflamed and noninflamed knee diameters. Results were statistically evaluated using a two-sample t test, and p values less than 0.05 were taken as significant.
Disease was also assessed histologically. Rats from different treatment groups were killed 14 days after arthritis induction. Knee joints were dissected intact and fixed in formalin-buffered saline prior to processing, sectioning, staining, and scoring for arthritis as described previously (17). The histological parameters assessed were synovial hyperplasia (on a scale of 0 -3), inflammatory infiltrate (scale 0 -5), intraluminal inflammatory exudate (scale 0 -3), and number of cartilagenous/bony erosions (scale 0 -3). The sum of the scored parameters gave the arthritis index for each knee. Results were statistically evaluated using the Mann-Whitney test for nonparametric variables, and p values less than 0.05 were taken as significant. MAC deposition was detected in joint sections by incubation with biotinylated anti-rat C9 (30 g/ml final concentration in PBS, 10 min) followed by avidin-fluorescein isothiocyanate (10 g/ml in PBS, 10 min). Sections were viewed under a fluorescence microscope and scored for MAC staining, and representative areas were photographed.

Generation of Soluble Recombinant CD59 in E. coli-Soluble
recombinant rat CD59, comprising the first 71 amino acids of rat CD59 modified with a 7-residue carboxyl-terminal "spacer" region and terminal Cys residue, was expressed in E. coli. The expressed protein was present almost exclusively as inclusion bodies in the bacterial pellet, with none detected in supernatants from transfectants (data not shown). Numerous bands were present in the fermentation pellet samples, but an additional band at the predicted molecular mass for sCD59-Cys (9 kDa) was present only in the postinduction pellet sample ( Fig.  1; compare lanes 1 and 2). This band was retained during several wash steps (Fig. 1, lanes 3-5), demonstrating that the protein was a component of the inclusion bodies. Upon solubilization of the bacterial pellet, it was apparent that sCD59-Cys was the major protein component (lane 6). From a 2-liter fer-mentation, the yield of sCD59-Cys was between 70 and 100 mg at this stage. Efficient refolding, assessed by acquisition of conformation-dependent mAb epitopes (data not included), was obtained under the conditions described, and misfolded/aggregated protein was removed by ammonium sulfate precipitation. The process was monitored by Western blotting using a polyclonal antiserum that recognized reduced and denatured CD59 and a mAb, 6D1, that detected only correctly folded protein (Fig. 2). In the prepurification sample, the polyclonal reagent detected a faint band at 9 kDa, heavier bands, presumably different conformers of dimer at 18 -21 kDa, and a higher aggregate band at 48 kDa ( Fig. 2A, lane 1). The ammonium sulfate pellet contained no 9-kDa monomer, but the larger bands were present (Fig. 2A, lane 2). The supernatant con- Groups of animals (n ϭ 6) were treated at the time of disease induction with a single intra-articular dose of 250 g of sCD59-APT542 (diamonds), a single dose of 250 g of sCD59-Cys (squares), or buffer control (triangles). Animals treated with either sCD59-APT542 or sCD59-Cys had significantly reduced joint swelling compared with controls from day 2 onwards (p Ͻ 0.05). tained predominantly the 9-kDa monomer with trace amounts of dimer remaining, demonstrating that all of the aggregated protein was contained within the pellet (Fig. 2A, lane 3). Under reducing conditions, only a 9-kDa band reactive with the polyclonal anti-CD59 antiserum was seen in each lane, indicating that all higher molecular mass bands were reduced to monomer (Fig. 2B). The mAb 6D1 (Fig. 2C) strongly stained a band at 9 kDa in the ammonium sulfate supernatant (Fig. 2C, lane 9) and, more weakly, the same band in the prepurification sample (Fig. 2C, lane 7), demonstrating that these contained correctly folded protein. No bands were stained in the ammonium sulfate pellet (Fig. 2C, lane 8), indicating that bands detected in the pellet by the polyclonal reagent represented denatured and misfolded protein. The final yield of pure, correctly refolded monomeric protein was ϳ5 mg, as determined by Coomassie protein assay (Pierce), with a refolding efficiency of 5%. The obtained N-terminal sequence was MLRXYNX; this was identical to the known sequence (MLRCYNC) with, as anticipated, failure to identify Cys residues. Retention of the initiating methionine has frequently been described in prokaryotic expression systems. The molecular mass of the refolded protein, measured using MALDI-TOF, was 9039 Da (theoretical mass 8728 Da).
Generation and Characterization of Membrane-targeted sCD59 (sCD59-APT542)-sCD59-Cys was modified at its C terminus by the addition of the membrane address tag, APT542, using thiol interchange chemistry. The "tagged" protein was purified by cation-exchange chromatography; SDS-PAGE analysis (Fig. 3) of soluble CD59 before (lane 1) and after (lane 2) conjugation to APT542 showed that the addition of a membrane address tag increased the apparent molecular mass of sCrry-Cys from 9 to 11 kDa. Efficiency of tailing of sCD59-APT542 was between 80 and 100%, and the residual untailed material was efficiently separated by chromatography on Mono-S (not shown). The molecular mass of the tailed protein, measured using MALDI-TOF, was 10,725 Da.
The capacity of sCD59-APT542 to inhibit MAC-mediated cell lysis was compared with that of untargeted sCD59-Cys and CHO-expressed sCD59 using nonwash (Fig. 4a) and wash (Fig.  4b) hemolysis assays on GPE5b-7 cells. In a nonwash assay, sCD59-Cys expressed in E. coli and sCD59 expressed in CHO cells (either with or without N-glycosylation), when used in equimolar amounts, were similarly effective at inhibiting lysis, indicating that the bacterially expressed and refolded protein was fully active when compared with that expressed in CHO. In contrast, sCD59 containing a membrane-targeting moiety (sCD59-APT542) was 100-fold more active than the untargeted proteins. In the wash assay, the differences were even greater. The C-inhibitory profile of sCD59-APT542 was almost identical in wash and nonwash assays, indicating that the protein had bound firmly to the cells and there exerted its effect. In contrast, no residual hemolysis inhibiting activity was seen with FIG. 6. Effect of sCD59-APT542 treatment on histological changes in rat AIA. A, normal joint; B, untreated knee joint 14 days after arthritis induction. Note the diffuse, dense infiltration of synovial tissue together with extensive cartilage destruction and bone erosion (arrows). C, sCD59-APT542-treated knee joint 14 days after arthritis induction. Cartilage surface is smooth with no erosions. There is patchy mononuclear infiltration and minimal, focal increase in synovial lining thickness. D, sCD59-Cystreated joint. In small areas where cartilage breakdown and early bone erosion are evident, there is marked mononuclear infiltration and widespread increase in synovial lining thickness. E, scoring of arthritis. The histological parameters assessed were synovial hyperplasia (0 -3), inflammatory infiltrate (0 -5), intraluminal inflammatory exudate (0 -3), and number of cartilagenous/bony erosions (0 -3). The sum of the scored parameters formed the arthritis index. The bars represent the mean arthritis index in each group Ϯ S.D. Significant differences between groups calculated using the Mann Whitney test for nonparametric variables are shown. the unmodified soluble molecules (sCD59-Cys, sCD59) after three wash steps.
sCD59-APT542 Therapy in Rat AIA-In order to assess the potential of sCD59-APT542 as a therapeutic agent, its effects on the course of rat AIA were investigated. Either sCD59-Cys or sCD59-APT542 was given as a single dose of 250 g/joint with the eliciting antigen at the time of disease induction (day 0) and compared with vehicle control. Clinical outcomes in the groups (n ϭ 6 each) are shown in Figs. 5 and 6. sCD59-APT542treated animals showed a dramatic reduction in clinical disease as assessed by joint swelling, which was significant when compared with vehicle-treated animals on days 2-14 (p Ͼ 0.05) (Fig. 5). Notably, the untailed sCD59-Cys-treated animals also showed significantly less joint swelling when compared with controls and were not significantly different from the sCD59-APT542 group at any time point (Fig. 5).
The effect of sCD59-APT542 on disease progression in AIA was assessed histologically in animals sacrificed on day 14 postinduction. Representative sections from untreated and treated rats are shown in Fig. 6, A-D. Untreated diseased joints displayed diffuse, dense infiltration of synovial tissue together with extensive cartilage destruction and bone erosion. Animals treated with sCD59-Cys also displayed marked inflammatory changes similar to those seen in untreated joints. In contrast, sCD59-APT542-treated knee joints retained a smooth cartilage surface with no erosions and only patchy mononuclear infiltration with minimal and focal synovial lining thickening. Joints were scored for multiple parameters of arthritis, and an overall index of severity was calculated. A single administration of sCD59-APT542 at day 0 reduced the severity score at day 14 from a mean Ϯ S.D. of 10.7 Ϯ 3.0 for the untreated group to means of 2.0 Ϯ 1.26 for the sCD59-APT542treated group and 6.0 Ϯ 3.09 for the sCD59-Cys-treated group. Statistical analysis revealed differences between groups of sCD59-APT542 versus control, p Ͻ 0.01; sCD59-APT542 versus sCD59-Cys, p Ͻ 0.05; and sCD59-Cys versus control, p Ͼ 0.05 (not significant) by the Mann-Whitney Test (Fig. 6E). Joints were stained for MAC using a well characterized anti-rat C9 antibody (Fig. 7). MAC deposits were found in abundance in control joints along the hypertrophic synovium and in areas of infiltration. Weak and inconsistent MAC deposits were found in sCD59-Cys-treated joints, and no MAC was found in joints treated with sCD59-APT542. DISCUSSION The overall aim of this program of work is to develop novel anti-C therapies that can be utilized in both acute and chronic conditions where C activation is a driving force. The potential therapeutic value of inhibiting C activation is well illustrated by the successful use of human C regulatory proteins such as sCR1 in animal models (26,27). However, studies to date have been limited to acute diseases such as experimental nephritis, AIA, and demyelination. This is due to numerous considerations, including the short half-lives and immunogenic nature of human C regulators in rodents, the high cost of recombinant protein expression in mammalian cells, and the potential for causing harm by chronic inhibition of C activity. In order to develop anti-C therapies that can be used in the treatment of chronic conditions, these problems must be overcome. To circumvent the problem of antigenicity, we have adopted the strategy of using rodent C regulators in rodent models. To address the short half-life, we have utilized novel membrane targeting strategies that retain the C regulator on cells either in the circulation or in tissue sites (2,18). The issue of cost has been addressed by the development of efficient bacterial expression systems for C regulators. We have recently described the expression in bacteria of a truncated form of the rodent regulator of C activation, Crry (18). The expressed protein was fully active as a C regulator, and the addition at the carboxylterminus of a membrane-targeting moiety, described below, markedly increased C-inhibitory activity in vitro and half-life in vivo. However, this agent, termed sCrry-APT542, acts early in the C activation pathway, blocking formation of C opsonins and chemotactic peptides and as such will restrict the essential physiological roles of C in opsonization and bacterial killing. An agent that inhibited later in the C pathway to prevent formation of the profoundly pathogenic MAC while permitting opsonic activity would be advantageous for chronic treatment.
To this end, we have undertaken the bacterial expression of a soluble recombinant form of the rat analogue of CD59, the sole membrane regulator of MAC assembly. Soluble forms of CD59 have been expressed in a variety of eukaryotic systems and shown to have MAC-inhibitory activity in vitro (11,20,21). To date, activity in vivo of recombinant forms of CD59 has not been reported, and no reports of expression of CD59 in prokaryotic systems have been published. Here we describe the overexpression of a soluble form of rat CD59 in E. coli. Expression was optimized to obtain yields of around 50 mg/liter of bacterial culture in the fermenter. As expected with these high expression levels, the protein accumulated as dense, insoluble protein aggregates within inclusion bodies in the cells (24,28). Isolation, solubilization, and renaturation of proteins from these aggregates in active form represents a significant hurdle that increases with increased complexity of secondary and tertiary structure in the protein. CD59 contains five disulfide bonds that lock the protein in a compact structure and are essential for C regulatory activity (29). Despite this complexity, the expressed protein (sCD59-Cys) was successfully solubilized and refolded from inclusion bodies and shown to be fully active FIG. 7. Staining for MAC deposition in joints. Serial sections from those used in Fig. 6 were stained for MAC deposits using biotinylated anti-rat C9 and avidin-fluorescein isothiocyanate as detailed under "Experimental Procedures." For orientation, the femur is indicated (f). A, normal joint, no staining for MAC. B, PBS-treated control joint; dense deposits of MAC are evident in the damaged synovium and in areas of infiltration (arrows). C, sCD59-APT542-treated joint; no MAC deposition is present in any areas. D, sCD59-Cys-treated joint; traces of MAC deposition are present in synovium.
as an inhibitor of MAC assembly when compared with soluble rat CD59 expressed in mammalian cells. The final yield of refolded, active protein averaged only 5 mg/liter of bacterial culture over several production runs. This is lower than yields obtained in our previous reports on bacterial expression of the short consensus repeat-containing C regulatory proteins, where nearly complete refolding and final yields of about 100 mg/liter were obtained (18,24). Attempts were made to improve the efficiency of refolding by empirical adjustments in the buffers and conditions used in this step, but no significant improvement was achieved in this study. This probably reflects the ease with which this multicysteine polypeptide can be diverted down unproductive and irreversible aggregation pathways. Further work is in progress to address this issue with both rat and human proteins. Nevertheless, sufficient active protein was produced for further studies. The protein was expressed with a seven-residue Ser-Gly "spacer" and final Cys residue at the carboxyl terminus to permit the addition of groups at this site. The Cys-tagged protein was fully active in hemolysis inhibition assays when compared with sCD59 from CHO cells. Some spontaneous dimerization of sCD59-Cys occurred upon storage, but dimer was easily separated from monomer by gel filtration.
Membrane-targeted recombinant rat CD59 was generated from sCD59-Cys by coupling a membrane addressin peptide (APT542), comprising a lipid moiety that interacts with the hydrophobic interior of the plasma membrane and a short positively charged peptide that interacts with negatively charged phospholipid head groups (25). Coupling to the carboxyl-terminal Cys by standard thiol interchange chemistry was highly efficient and yielded almost 100% sCD59-APT542. Residual untagged sCD59-Cys was removed by cation exchange chromatography, and the tagged and untagged agents were compared in hemolysis assays. In standard nonwash assays where agent was present throughout, the C-inhibitory activity of the sCD59-APT542 was increased over 100-fold when compared mole-for-mole with sCD59-Cys or sCD59 expressed in CHO cells. In wash assays, C-inhibitory activity of the untagged forms of sCD59 was completely lost, whereas sCD59-APT542 retained strong C-inhibitory activity even after multiple washes of the target cells, demonstrating that it had incorporated into the cell membrane.
Soluble forms of CD59 have been little tested as C therapeutics primarily because of their low specific activity in the presence of serum (11,14) and small size, probably leading to rapid elimination in urine. sCD59-APT542 overcomes these problems. As a first investigation of the capacity of this agent to inhibit C-mediated pathology in vivo, we chose to study its effect on the course of AIA in the Lewis rat. AIA is an acute monoarticular arthritis that has previously been shown to be profoundly C-dependent and susceptible to treatment with anti-C agents (17,30). A single 250-g dose of either sCD59-APT542 or sCD59-Cys given intra-articularly at the time of disease induction markedly inhibited disease as assessed by measurement of joint swelling, indicating that both agents influenced the acute inflammation associated with this model. However, when joint pathology was assessed histologically at end point, only sCD59-APT542 significantly influenced the disease process, indicating that retention of agent in the joint improved outcome. In previous unpublished work, we have been unable to demonstrate an effect of sCD59 expressed in CHO cells in this model and were surprised to find an effect, albeit limited to acute inflammation, in this study. It is possible that dimerization of sCD59-Cys in the joint increased half-life, although this is not proven. Nevertheless, the key finding was that sCD59-APT542 markedly inhibited the progression of disease as assessed by histology. MAC deposition was also blocked in joints treated with this agent. The degree of inhibition of disease was at least equivalent to that obtained using a membrane-targeted truncated form of sCR1 (APT-070) (17), strongly implicating the MAC as the major drive to pathology in this model and demonstrating the therapeutic potential of a powerful inhibitor of MAC function.
These encouraging studies in vivo will now be extended to situations requiring systemic administration of agent and to chronic disease models to test the prediction that sCD59-APT542 will provide prolonged and targeted delivery of MAC inhibition in C-mediated pathologies.