Structure of the Chlamydia protein CADD reveals a redox enzyme that modulates host cell apoptosis.

The Chlamydia protein CADD (Chlamydia protein associating with death domains) has been implicated in the modulation of host cell apoptosis via binding to the death domains of tumor necrosis factor family receptors. Transfection of CADD into mammalian cells induces apoptosis. Here we present the CADD crystal structure, which reveals a dimer of seven-helix bundles. Each bundle contains a di-iron center adjacent to an internal cavity, forming an active site similar to that of methane mono-oxygenase hydrolase. We further show that CADD mutants lacking critical metal-coordinating residues are substantially less effective in inducing apoptosis but retain their ability to bind to death domains. We conclude that CADD is a novel redox protein toxin unique to Chlamydia species and propose that both its redox activity and death domain binding ability are required for its biological activity.

Chlamydiae are obligate intracellular bacteria and the causative agents of important sexually transmitted and disabling ocular (blinding trachoma) human diseases (1). Chlamydia engages in a unique relationship with its host. Upon entering host cells, the parasite starts a biphasic developmental cycle from the infectious form, called an elementary body, to a non-infectious, vegetative growth form, called a reticulate body, and then eventually back to the replication-incompetent infectious form (2). After the transition back to the infectious form, the host cell dies and releases its infectious load (3). To accommodate its life cycle, Chlamydia may inhibit apoptosis during the early stages of infection (4,5) and promote apoptosis at later stages (6,7).
Recently, the Chlamydia protein CADD 1 has been shown to associate with tumor necrosis factor family proteins and to induce apoptosis when transfected into a variety of mammalian cell lines (8). CADD has no close homologues but does show 18% sequence identity with coenzyme PQQ (pyrrolo-quinolinequinone) synthesis protein C (PqqC) family members, which are part of the six-step PQQ synthesis pathway in bacteria (9). However, homologues of other members of the pathway are not found in Chlamydia species for which genome information is available. Indeed, ectopic expression of PqqC from Klebsiella pneumoniae failed to cause apoptosis, demonstrating the specificity of CADD-induced cell death (8). CADD is expressed late in the infectious cycle of Chlamydia trachomatis and is secreted into the host cytoplasm, where it co-localizes with tumor necrosis factor receptors in the proximity of the inclusion body. Sequence comparisons had suggested that CADD contains a death domain.
Here we present the crystal structure of CADD, which reveals an iron-containing redox enzyme that bears no resemblance to death domains. Mutagenesis of the active site of CADD reduced but did not eliminate its apoptotic activity, suggesting that both its catalytic activity and death domain binding activities contribute to its biological activity.
Crystallization-Purified CADD was crystallized by the vapor diffusion method at room temperature using a sparse matrix screen (Hampton). Sitting and hanging drops consisting of 3 l of precipitant solution (10% (v/v) polyethylene glycol 12000, 20 mM cacodylate, pH 6.5) and 3 l of protein solution (12 mg/ml protein) yielded crystals within 3-5 days. Crystals grew as very thin plates with dimensions of 200 ϫ 200 ϫ 20 m in space group C222 1 . The crystal structure was determined by a selenium MAD experiment using a seleno-methionine substituted protein (10). For data collection, crystals were transferred into cryobuffer (crystallization buffer with 25% (v/v) glycerol) and flash-cooled in liquid nitrogen.
Data Collection, Structure Solution, and Refinement-The threewavelength MAD data set was collected from one single crystal, using synchrotron radiation at beamline X12B of the National Synchotron Light Source. Oscillation data were recorded in frames of 1°through a continuous angular range of 120°for the peak ( ϭ 0.9791 Å), the high energy remote ( ϭ 0.925 Å), and the inflection point ( ϭ 0.9794 Å). The native data set was collected at beamline X9B of National Synchotron Light Source. All data were processed with the programs DENZO and SCALEPACK (11). The CADD structure was phased and traced using the program SOLVE/RESOLVE (12). Model building and refinement were carried out in O (13) and REFMAC5 (14). The final CADD model comprises three protein monomers (residues A7-A219, B7-B219, C7-C219), 6 Fe 2ϩ ions with 3 closely bound putative water molecules, and 176 water molecules. Residues 1-6 and 220 -231 were not visible in the electron density maps and therefore were not included in the model. Statistics for data collection, refinement, and model quality are summarized in Table I. Surface calculations were carried out with the CASTP server (15) and the protein-protein-interaction server (16). Cell Culture, Transfections, and Apoptosis Measurements-HeLa cells were maintained in Dulbecco's modified Eagle's medium (Irvine Scientific) and supplemented with 10% fetal bovine serum, 1 mM Lglutamine, and antibiotics. Cells (10 6 ) were transfected with PEGFP-C2 plasmids containing CADDwt, CADD-mut1, and CADD-mut2, using LipofectAMINE (Invitrogen) following the vendor's protocol. Both floating and adherent cells were recovered 1 day later and pooled, and the percentage of transfected (green fluorescent) cells with nuclear apoptotic morphology was determined by staining with 0.1 g/ml 4Ј,6-diamidino-2-phenylindole (mean Ϯ S.D.; n ϭ 3). Cytosolic extracts from HeLa cells were subjected to immunoblotting and probed with rabbit polyclonal anti-green fluorescent protein (GFP) antibody (Invitrogen) for the presence of GFP-CADD fusion proteins.
Protein Binding Assays-A plasmid containing DR5 was in vitro transcribed and translated in the presence of L-[ 35 S]methionine using the TNT kit from Promega. GST-CADD, GST-CADD-mut1(data not shown), GST-CADD-mut2, and control GST-CD40 (cytosolic domain) fusion proteins were immobilized on glutathione-Sepharose at 1 g/l and incubated with in vitro translated target proteins for 2 h at 4°C. Beads were then washed four times in 1 ml of 140 mM KCl, 20 mM Hepes, pH 7.5, 5 mM MgCl 2 , 2 mM EGTA, 0.5% Nonidet P-40, and analyzed by SDS-PAGE/fluorography.
Mass Spectrometry and ICP-AAS-Matrix-assisted laser desorption/ ionization-time of flight, peptide mapping, and ICP-AAS-spectrometric analysis on the purified CADD protein were accomplished using standard techniques at the Facility for Mass Spectrometry at the Scripps Research Institute in La Jolla.
Coordinates-Coordinates and structure factors for CADD have been deposited with the Protein Data Bank (www.rcsb.org/pdb) under accession code 1RCW.

RESULTS
CADD Structure-Recombinant CADD from C. trachomatis was expressed in E. coli, purified, and crystallized. The crystal structure was determined by a selenium MAD experiment (10). CADD is a 231-residue protein, molecular mass ϭ 26,734 Da, which forms a homo-dimer in solution, as judged by gel filtration. The CADD monomer is cylindrical with approximate dimensions of 45 ϫ 29 ϫ 37 Å. CADD folds into a seven-helix FIG. 1. The overall structure of CADD. A, CADD depicted in ribbon representation, rainbow color-coded from N terminus (blue) to C terminus (red), with helices H1-H7, the two iron ions, and loop L3 labeled. B, the CADD dimer is shown normal and parallel to its long axis.
c R free ϭ same as R cryst but comprises a test set (5% of total reflections), which was not used in model refinement.
mostly parallel/anti-parallel bundle, where six ␣-helices (H1, H2, H3, H4, H5, H7) partly embrace the seventh helix (H6) (see Fig. 2A). According to the Structural Classification of Proteins Data Base (18), CADD belongs to the "heme-oxygenase" fold. Helices H1, H3, H4 are kinked and can therefore be represented as separate shorter ␣-helices denoted as A and B. This is especially true for helix H3, where a hairpin loop, residues 82-87, is inserted (Figs. 1A and 2). The CADD dimer is formed through an interaction via helices H2 and H3A, residues 59 -85 (Figs. 1B and 2). The interface-accessible surface area is 915 Å 2 /monomer, which accounts for 9.2% of the accessible surface area of the CADD dimer. The interaction is predominantly hydrophobic (55% non-polar atoms) but also includes a number of polar interactions and salt bridges. The most similar structures found using the DALI server (19) are: PqqC (20), with an r.m.s.d. of 2.8 Å for the superposition of 221 C␣ atoms and 18% sequence identity; human heme-oxygenase (21), with an r.m.s.d. of 2.9 Å for 199 C␣ atoms and 11% sequence identity; the R2 subunit of ribonucleotide reductase (R2-RNR) (22), with an r.m.s.d. of 3.2 Å for 178 C␣ atoms and 12% sequence identity; and the ␣-subunit of methane monooxygenase (MMOH) (23), with an r.m.s.d. of 3.1 Å for 174 residues and 9% sequence identity. Although none of the active sites are conserved, each of these enzymes appears to be a redox enzyme, suggesting that this fold is particularly suitable for this type of enzyme. According to sequence similarity searches with the bioinformatics server Fold and Function Assignment System (24), CADD shares distant sequence homology with transcription enhancement gene A transcription factors (25) and can be used as a template to obtain homology models for these proteins.
The Active Site-The seven helices of CADD provide the scaffold for a narrow internal cavity equipped with a di-metal center (Figs. 2 and 3A). The experimental electron density map clearly indicates the presence of two metal ions coordinated by 6 residues (Glu-81, His-88, Glu-142, His-174, Asp-178, His-181) (Figs. 2 and 3A). The di-metal site is located in the center of the molecule adjacent to the cavity, which most likely serves as the active site. Atomic absorption measurements using ICP-AAS revealed the presence of iron and small but significant amounts of zinc in the protein. This indicates the presence of a di-iron site, which, judged by difference maps and elevated B-factors, is not fully occupied in the CADD crystals. The small amounts of zinc might be due to oxidation and partial replacement of iron for zinc, which has been observed in several crystal struc-tures of di-iron-containing proteins (26). The di-metal center appears to be octahedrally coordinated and bridged by a glutamate residue (Glu-81) and a water molecule or hydroxide ion. Fe1 is coordinated by two histidines (His-174, His-88) and the glutamate (Glu-81), as well as the putative water, which it shares with Fe2. Fe2 is coordinated by histidine (His-181), two glutamates (Glu-142, Glu-81), aspartate (Asp-178), and the bridging water molecule (Fig. 3A). All 6 active site residues coordinating the metal ions are strictly conserved among CADD proteins from Chlamydia species (Fig. 2). The water molecule or hydroxide ion is coordinated by both iron atoms at a distance of 2.2 Å (Figs. 3A and 4B). It is adjacent to Asp-178, His-174, and Tyr-170 and faces the internal active site cavity. The elliptically shaped density (3 peak in 2F o Ϫ F c map) for the water molecule/hydroxide ion is obscured by the electronrich iron atoms nearby and therefore not unambiguously interpretable. The electron density and resulting B-factors are also consistent with a reactive oxygen species bound to the di-iron site. The cavity next to the di-iron site shows an overall positive charge and measures 5 ϫ 7 ϫ 14 Å, with a volume of 340 Å 3 (15). The cavity is lined with 15 conserved hydrophilic or aromatic residues (His-50, Ile-51, Phe-54, Glu-81, Ile-89, Glu-142, Asp-178, Tyr-141, Ile-145, Lys-152, Tyr-170, His-174, Pro-55, Glu-82, and Asn-87 (Fig. 2)). Below this cavity, the hydrophobic core is largely aromatic and also contains a buried lysine (Lys-152). A system of cavities spans across the core of the molecule, with two potential openings next to loop L3 and between helices H1B and H5. One opening, E1, penetrates the surface of the protein between helices H2, H3, and the unique loop L3 (Figs. 1A and 3B). It is lined by residues Ile-51, Pro-55, Ile-89, and Glu-82. An alternative access path, E2, leads from the di-iron site through a narrow opening into a second cavity lined by residues Met-21, Tyr-43, Tyr-47, Trp-92, Ile-148, Ala-149, Phe-171, Ala-149, Lys-152, and Tyr-27 and from there to the surface next to residues Trp-30 and Asp-151 (helices H1B and H5). The size of the active site cavity openings restricts the substrates to small compounds such as O 2 , H 2 O 2 , CH 4 , CH 3 OH, CO, or CO 2 . Larger molecules could only pass through by means of a conformational change.
The active site of CADD is similar to that found in RNR-R2 from E. coli (Protein Data Bank accession code 1xsm). The helices forming the core that contains the active site can be superimposed with an r.m.s.d. of 2.8 Å. The function of RNR-R2 is to generate a tyrosyl radical on an adjacent tyrosine with the help of its di-iron center. The organic free radical is transferred to the RNR-R1 subunit, which catalyzes the de novo production of deoxy nucleotides (22). Interestingly, CADD also contains a tyrosine (Tyr-170) next to the di-iron center. These similarities raise the question of whether the physiological function of CADD is the production of radicals for RNR-R1. However, no equivalents are seen for Asp-84, Asp-237, and Trp-48, which are critical residues for the radical initiation pathway proposed in RNR-R2 (Tyr-122-Asp-84-Fe1-His-118-Asp-237 to Trp-48) (26). Taken together, these findings indicate that CADD cannot function as a RNR-R2 but might use a tyrosyl-radical for catalysis.
Cellular Activity of CADD-To test whether Tyr-170 and the di-metal site are involved in the toxicity of CADD, we generated two active site mutants by PCR mutagenesis and tested their apoptotic activity through transfection experiments in mammalian cells. The role of Tyr-170 was tested with a Y170F (CADD-mut1) mutant. To prevent the formation of a functional di-metal center, we made the quadruple mutant of the metalcoordinating residues: E81A/H88A/H174A/Y170F (CADD-mut2). When equivalents of each plasmid DNA were transfected into HeLa cells, CADD-mut1 showed a decrease in toxicity of about 5-15% when compared with the wild-type. CADD-mut2 showed more than 60% reduction in apoptotic activity (Fig. 5A). Immunoblotting shows (Fig. 5B) that both CADD mutants are expressed at similar or higher levels to wild-type. This indicates that the mutants, especially CADD-mut2, are better tolerated by the transfected mammalian cells than the wild-type. To address the question of whether the active site mutant proteins still bind to death receptors, we carried out an in vitro DR5 binding assay (Fig. 5C), comparing GST-CADD-wt, GST-CADD-mut1 (data not shown), and GST-CADD-mut2. CADD wild type and active site mutants show comparable binding to death receptor DR5, indicating that the active site mutations do not alter the DR5 binding activity of CADD.

DISCUSSION
The crystal structure shows that CADD shares similarity to heme-oxygenase and PqqC enzymes. The sequence similarity and "PqqC-like" annotation for CADD proteins are reflected by the same fold, but the active sites are not conserved, and the two proteins are therefore functionally and most likely also evolutionarily unrelated (20). CADD is consequently an orphan unique to Chlamydia species, which further emphasizes its role as a highly specific toxin that evolved in this intracellular parasite. Comparison with the more distant structural homologues, RNR-R2 (22) and MMOH (23), reveals di-iron active sites in a strikingly similar structural context. Although the three proteins belong to different fold subclasses (CADD shows the heme-oxygenase fold, whereas RNR-R2 and MMOH belong to the ferredoxin fold), the helices forming the core containing the active site can be superimposed with an r.m.s.d. Ͻ2.8 Å. The active site of CADD is structurally similar to that in RNR-R2 but does not contain the conserved residues of the radical pathway. CADD can therefore not serve as an RNR-R2, but it is tempting to speculate that CADD, like RNR-R2, may generate and use a free tyrosyl radical on Tyr-170 to facilitate redox reactions. However, mutagenesis studies with an Y170F mutant show only a 5-15% decrease in toxic activity, indicating that Tyr-170 is not essential for CADD function. The central cavity of CADD contains several tyrosines, and it is possible that another one (Tyr-47, Tyr-141) may substitute for the loss of Tyr-170.
A structural comparison with the di-iron center in MMOH from Methylococcus capsulatus (Protein Data Bank accession code 1mhyD) (23) reveals strong conservation of the metalcoordinating residues, except for a difference in the coordination of Fe1 in CADD, where Glu-114 is replaced on the other side of Fe1 with His-174 (Fig. 4, A and B). A detailed analysis of the active sites further reveals that in contrast to RNR-R2, MMOH and CADD contain an internal cavity next to the diiron center (Fig. 4C). In MMOH, the cavity functions as the site of catalysis, where substrate and product access the di-iron center through the tunnel-like cavity from the bottom of the molecule. CADD contains a similar tunnel when the entrance next to Trp-92, between H1B and H5, is used (Fig. 3B). On the other hand, the opening next to the loop L3 is a potential region for a conformational change that could open the cavity to the outside for the exchange of substrate and product. Thus, CADD is most likely an enzyme similar to MMOH (23), which uses an internal active site equipped with a di-iron center to catalyze redox reactions on small molecule substrates. Further biochemical studies are needed to determine the reaction catalyzed by CADD.
Transfection assays with a CADD mutant lacking critical metal-coordinating residues establish a direct connection between the di-iron site and the apoptotic activity of CADD. Alterations at the active site, which is buried within the molecule, do not abolish interaction with death receptors, which suggests that the optimal induction of apoptosis by CADD requires both the intracytoplasmic cross-linking of death receptors as well as its redox activity.