Conservation of the biochemical properties of IncA from Chlamydia trachomatis and C . caviae : oligomerization of IncA mediates interaction between facing membranes

The developmental cycle of Chlamydiaceae occurs in a membrane compartment called an inclusion. IncA is a member of a family of proteins synthesized and secreted onto the inclusion membrane by the bacteria. IncA proteins from different species of Chlamydiaceae show little sequence similarity. We report that the biochemical properties of C. trachomatis and C. caviae are conserved. Both proteins associate with themselves to form multimers. When artificially expressed by the host cell, they localize to the endoplasmic reticulum. Strikingly, heterologous expression of IncA in the endoplasmic reticulum completely inhibits concomitant inclusion development. Using truncated forms of IncA from C. caviae, we show that expression of the C-terminal cytoplasmic domain of the protein at the surface of the endoplasmic reticulum is sufficient to disrupt the bacterial developmental cycle. On the other hand, development of a C. trachomatis strain that does not express IncA is not inhibited by artificial IncA expression, showing that the disruptive effect observed with the wild type strain requires direct interactions between IncA molecules at the inclusion membrane and on the endoplasmic reticulum. Finally, we modeled IncA tetramers in parallel four helix bundles based on the structure of the SNARE complex, a conserved structure involved in membrane fusion in eukaryotic cells. Both C. trachomatis and C. caviae IncA tetramers were highly stable in this model. In conclusion, we show that the property of IncA proteins to assemble into multimeric structures is conserved between chlamydial species and we propose that these proteins may have co-evolved with the SNARE machinery for a role in membrane fusion. JBC Papers in Press. Published on August 16, 2004 as Manuscript M407227200


INTRODUCTION
Chlamydiaceae are obligate intracellular parasites which develop in a host cell within a membrane-bound compartment, termed an inclusion. The membrane of the inclusion is initially formed by the invagination of the plasma membrane, and pinching off of a vesicle containing the infectious form of the bacterium, the elementary body (EB 1 ). Thereafter, EBs differentiate into noninfectious but metabolically active reticulate bodies, which proliferate within the expanding inclusion, giving rise to 1000 or more progeny per host cell. The infectious cycle ends after 2 to 3 days depending on the strain, when bacteria that have differentiated back to EBs are released in the extracellular medium.
The composition of the inclusion membrane and the origin of its constituents is not yet fully understood. EB-containing vesicles seem to become unable to fuse with early endosomes soon after entry (1). They also fail to fuse with late endocytic compartments and lysosomes, thus escaping degradation by the host cell (2). The early separation between the chlamydial inclusion and the endocytic trafficking distinguish the inclusion from the parasitophorous vacuoles of several pathogens (reviewed in (3)). Acquisition of lipids necessary for the growth of the inclusion membrane, as well as for incorporation into bacteria (4), must originate from other sources. One origin of the bacterial lipid content is the Golgi apparatus, as sphingomyelin, synthesized in the Golgi apparatus from a fluorescent precursor, is transported to the inclusion and accumulates into the bacteria (5). This process is dependent on bacterial protein synthesis : inhibition of chlamydial early transcription or translation prevents the incorporation of sphingomyelin, and the inclusion eventually fuses with lysosomes (6). The bacteria also appear to acquire host cell cholesterol by the same Golgi-dependent pathway as sphingomyelin (7). Surprisingly, no host cell protein has been found inserted into the inclusion membrane (2). The only host protein known to interact directly with the inclusion is 14-3-3, for which interaction with a bacterial component of the inclusion has been demonstrated (8). Host proteins recruited to the inclusion include dynein, β-catenin and specific Rab GTPases, but the mechanisms of their association with the inclusion are not known (9)(10)(11).
In contrast with the seemingly poverty of the inclusion membrane in eukaryotic proteins, a large family of bacterial proteins, termed Inc proteins, are known to be inserted in this compartment (12). These proteins are unique to Chlamydiaceae, and members of the family share little primary sequence identity with each other within one species. Some of the members are somewhat conserved between different Chlamydiaceae species, but in that case the conservation is usually low. They share however one remarkable feature, which allows to predict that a given protein by guest on  http://www.jbc.org/ Downloaded from probably belongs to the family : they possess a very large (50-80 amino acids) bilobed hydrophobic domain. Confirmation that a protein which such a domain is located to the inclusion membrane requires specific antibodies against this protein (13). So far about 10 proteins have been shown to localize to the inclusion membrane, and between 40 to 90 proteins are predicted to belong to the family, from C. trachomatis and C. pneumoniae genome analysis, respectively (13,14). The topology of the insertion of Inc proteins in the inclusion membrane has not been directly addressed, but microinjection of antibodies against 4 of these proteins demonstrated that at least the carboxyterminal domain is exposed to the cytosol (15,16). The large hydrophobic domain is probably required for the insertion in the inclusion membrane and would be compatible with a hair-pin insertion, with both extremities of the proteins facing the cytosol, but this needs to be investigated.
The mechanism by which Inc proteins are secreted out of the Chlamydiaceae for insertion in the inclusion membrane has been identified. Chlamydiaceae possess a type III secretion apparatus, which is found in several Gram negative pathogenic bacteria, and which allows for the translocation of bacterial proteins through the bacterial membranes and across a eukaryotic membrane (17).
Using heterologous secretion, it was shown that Inc proteins are recognized by type III secretion machineries of other pathogens, strongly suggesting that it is the mechanism used by Chlamydiaceae to secrete these proteins into the inclusion membrane (18,19).
The first member of the Inc family to be identified, IncA, is also the one that attracted most of the attention. First cloned from C. caviae (CCA00550), it has homologs in a similar genetic environment in all sequenced genomes (CT119 in C. trachomatis serovar D and CPn0186 in C. pneumoniae CWL029). The level of sequence identity between homologs is low, and antibodies against IncA do not cross react between different species. Antibodies against IncA from each of these species have been obtained, and have allowed to show an important accumulation of the protein on the membrane of the inclusion of all species (13,20,21), as well as on fibers emanating from the inclusion that are particularly enriched in some species (22). IncA from C. caviae was shown to be exposed on the cytoplasmic face of the inclusion, and to be phosphorylated by the host cell (15). Finally, IncA is expressed rather late compared to most Inc proteins, the transcript being detected 12 h post infection with the C. trachomatis serovar L2 strain and 16 h post infection with the C. trachomatis serovar D strain (23,24).
IncA is the only member of the Inc family for which a function has been proposed, namely a role in the homotypic fusion of inclusions in C. trachomatis. Typical C. trachomatis isolates occupy inclusions that fuse with each other when the cells are infected at high multiplicities of infection. This fusion is inhibited at low temperature (32°C) and requires bacterial protein synthesis (25). Evidence for the involvement of IncA in this process came from two independent studies.  (16). Second, a minority (1.5%) of C. trachomatis clinical isolates form multiple non-fusogenic inclusions and do not express IncA (26). Careful analysis of this collection of variants later showed that if most of these variants (24/27) do not express IncA, three non-fusogenic strains do express a normal protein at the inclusion membrane, suggesting that other elements of the fusion machinery are missing in these strains (27). Consistent with the implication of IncA in inclusion fusion is the observation that the majority of inclusion fusions occurs between 10 and 16 h post infection with the serovar L2 (25,28), which correlates with the time when IncA can be detected in this strain (16). Moreover, the inhibition of inclusion fusion at low temperature correlated with an inhibition of IncA export to the inclusion membrane in these conditions (29).
However, the temperature block is likely to affect the export of several proteins, which could also account for the inhibition of fusion.
Altogether, these data argue for a role of IncA in the fusion of inclusions observed with C. trachomatis strains. However, several questions remain unsolved. In cells infected at low multiplicity of infection, microinjection of anti-IncA antibodies leads to the septation of the inclusion (16). One explanation is that inclusions, like other eukaryotic organelles, are dynamic entities that can fuse and septate, and that in the presence of antibodies against IncA fusion is slowed down, resulting in multiple inclusions. However, clinical isolates of non-fusogenic phenotypes contain only a single inclusion at low multiplicity of infection, implicating that the absence of IncA does not result in multiple inclusions in this case. Microinjection of whole anti-IncA antibody was shown to induce the aggregation of IncA on the surface of the inclusion, while the protein remained homogenously distributed when Fab fragments of the same antibody were injected (15). Therefore, microinjection of anti-IncA antibody may induce a more general disorganization of the inclusion proteins involved in fusion/septation and some of the consequences of microinjection may be indirect. Even if IncA plays a direct role in C. trachomatis inclusion fusion, other roles may be envisioned, especially in other species such as C. caviae and C. pneumoniae, which appear to be less fusogenic than C. trachomatis.
In this report, we investigated the biochemical properties of IncA from C. caviae GPIC strain (CcaIncA) and from C. trachomatis serovar L2 (CtrIncA). We showed that IncA from both species can interact with itself via coiled-coil interactions in its C-terminal cytoplasmic domain.
Dynamic modelisation on membrane-proximal domains of CcaIncA and CtrIncA showed that tetramers of IncA are compatible with a structure similar to the SNARE complex, which is a conserved complex involved in the fusion of vesicles with their target membrane. We used 6 heterologous expression of IncA by HeLa cells to further investigate the biochemical properties of IncA.

Cells, bacteria, antibodies and other reagents
The human cervical adenocarcinoma cell line, HeLa 229, was from the American Type Culture Collection (ATCC) and was grown in Dulbecco's modified Eagle's medium with glutamax (Life Technologies) supplemented with 10% fetal calf serum (complete medium). The GPIC strain of C. caviae was obtained from Roger Rank (University of Arkansas). C. trachomatis serovar L2 strain 434 (VR-902B) and C. trachomatis serotype D (27F0734) were from ATCC. C. trachomatis serotype D(s)5058 is a clinical isolate which does not express IncA and was kinly given by Drs Dan Rockey and Walter Stamm (27). Chlamydiaceae were prepared as described (30) alkaline phosphatase-linked secondary antibodies were obtained from Pierce, goat anti-rabbit or mouse Alexa Fluor-488 antibodies were from Molecular Probes and goat anti-rabbit TRITC antibodies were from Immunotech (France). Rabbit anti-C. caviae IncA antibodies were prepared as described (30).

Plasmids
Genomic DNA from C. caviae strain GPIC and C. trachomatis serovar L2 were prepared from bacteria using the RapidPrep Micro Genomic DNA isolation kit (Amersham Pharmacia Biotech).
The incA genes were amplified by PCR, and cloned in NcoI-KpnI sites of pQE-TriSystem vector for His-tagged proteins (Qiagen) and in EcoRI-KpnI sites in pEGFP-C1 vector for GFP-tagged proteins (Clontech). Sequences of the primers used are listed in Table 1 After electophoresis, proteins were transferred to a polyvinylidene fluoride (PVDF) membrane (Millipore), and the membrane was used for blotting with anti-CcaIncA or anti-Histidine antibodies followed with HRP-or alkaline phosphatase-conjugated antibodies, and revelation were performed by enhanced chemiluminescence (ECL) or chemifluorescence, respectively, according to the manufacturer's instructions (Amersham Pharmacia Biotech). sample buffer and proteins bound to the beads were analyzed by SDS-PAGE followed with transfer to a PVDF membrane and western blotting. The same membrane was first probed with anti-Histidine and HRP-conjugated anti-rabbit antibodies, and revealed by ECL. The membrane was then stripped by a 30 min incubation at 50°C in 0.7% β-mercaptoethanol, 2% SDS, 62.5 mM Tris pH 6.8, and probed again, using anti-GFP and HRP-conjugated anti-rabbit antibodies, and revealed by ECL.  (32). In this method, missing side-chain atoms are initially placed in random positions and the resulting structure is then minimized in a three-stage protocol: first, simulated annealing with covalent and packing interactions only; second, a short molecular dynamics run with a full molecular dynamics force field in vacuo, followed by conjugate gradient minimization; and third, a molecular dynamics run in explicit solvent. We used the program X-plor (33) for stages 1 and 2, and GROMACS 3.2 (34) for stage 3. We extended the third stage to a molecular dynamics trajectory in explicit solvent of one ns. The structures were embedded in a box of SPC water molecules with minimum distance between the solute and the box boundary of 1 nm. The system consisting of protein, water and Na ions to neutralize the total charge was slowly heated from 50 K to the simulation temperature (300 or 350 K) with positional restraints on the solute during 300 ps.

Modelisation of IncA structure
The electrostatic interactions were treated with the Particle-Mesh-Ewald method (35) for interactions beyond 1 nm; weak temperature coupling with a relaxation time of 5 ps, using the Berendsen method, was employed (36).

Immunofluorescence microscopy
HeLa cells grown in 6-well plates were transfected with the indicated plasmids using Fugene reagent (Roche Applied Science). Twenty-four hours later, the cells were washed twice in PBS and fixed with 4% paraformaldehyde, 120 mM sucrose in PBS for 30 min at room temperature. The cells were washed in PBS, incubated for 10 min in 50 mM NH 4 Cl in PBS at room temperature, saturated in 1 mg/ml BSA in PBS and permeabilized in 0.05% saponin, 1 mg/ml BSA in PBS. To observe Histidine-tagged proteins, the cells were first labeled with anti-Histidine antibody before being incubated with anti-rabbit Alexa Fluor-488 antibody. The endoplasmic reticulum and the chlamydial inclusion were labeled with anti-calnexin and anti-Chlamydia antibodies, respectively, followed by incubation with Cy TM -3-conjugated goat anti-mouse antibodies. To quantify bacterial entry in cells transfected with CcaIncA-His, the cells were fixed 3 h after infection. Extracellular and intracellular bacteria and transfected cells were labeled as described (40), except that transfected cells were identified using anti-His antibodies. Coverslips were mounted in mowiol with

IncA proteins are found in high molecular weight complexes
IncA proteins share the same general organization: a short amino-terminal domain, a large bilobed hydrophobic domain and a C-terminal cytoplasmic domain ( Figure 1). This domain was shown to be exposed on the cytosolic face of the inclusion in the case of CcaIncA and CtrIncA (15,16). The level of identity between IncA of different species is low (around 20 % of identical residues, which are distributed along the whole molecule). Using tools designed to calculate the propensity of a given sequence to form coils, we noticed that all IncA proteins are predicted to engage in coiled-coil structures over the whole length of their C-terminal domain with a high probability. This prediction is strengthened by the presence of one or more leucine-zipper domains in these regions.
The potential of IncA to interact with other molecules via coiled-coil interactions prompted us to look for its partner(s). First, we performed cross-linking experiments on cells infected for 24 h with C. caviae. Infected cells were incubated for 30 min at 4°C with the cross-linking reagent DSP before cell lysis and analysis of whole-cell extracts by SDS-PAGE, followed by western blot analysis using anti-CcaIncA antibodies (Figure 2A). In control cells, IncA migrated around 37 kD, which corresponds to its expected molecular weight. In cells incubated with the cross-linker, IncA also migrated as a higher molecular weight complex, around 75 kD. This band disappeared when the sample was treated with β-mercaptoethanol, which cleaves the disulfide bridge present in the cross-linker and separates IncA from its partner. A faint upper band (around 150 kD) was also observed in cells treated with DSP, suggesting that IncA may participate in the formation of complexes of molecular weight even higher than 75 kD, that may consist of more than 2 molecules.
Since IncA is found on the cytosolic face of the membrane of the inclusion, which contains several chlamydial proteins, its partners may be of eukaryote or prokaryote origin. To assess if the presence of other chlamydial proteins were necessary for the association of IncA with its partners, we performed cross-linking experiments in cells heterologously expressing IncA. For this purpose, incA was cloned in a vector that provided a promoter for expression in eukaryotic cells as well as a carboxy-terminal Histidine tag.
Cells transfected with CcaIncA-His were treated with DSP, lysed, and the affinity of the Histidine tag for Ni 2+ was used to purify the His-tagged protein from cell extracts using Ni-  Figure 2B). Therefore, the property of IncA to participate in high molecular weight complexes is common to at least two species of Chlamydiaceae.
Altogether, these experiments show that CcaIncA-His and CtrIncA-His participate in high molecular weight complexes. Considering the fact that cross-linking is never complete, the relative abundance of intermediate-and even high-molecular weight complexes suggests that most of the IncA molecules are engaged in these complexes.

IncA interacts with itself
Since the size of the intermediate-molecular weight complexes were compatible with the formation of dimers of IncA, we investigated further whether this protein was able to form multimers. There is no useful gene manipulation technique to manipulate the genome of Chlamydiaceae, and to address reporter molecules to the membrane of the inclusion. Therefore, to determine whether CcaIncA was able to interact with itself, we expressed CcaIncA in HeLa cells with two different tags : a C-terminal Histidine tag (CcaIncA-His), and a N-terminal Green Fluorescent Protein tag (GFP-CcaIncA). The affinity of the Histidine tag for Ni 2+ was used to purify the His-tagged protein from cell extracts using Ni-NTA beads. The beads were washed and Ni-NTA associated proteins were analyzed by SDS-PAGE and western blot. Probing of the membrane with anti-Histidine antibody showed that, as expected, CcaIncA-His was retained on the beads ( Figure   2C, left, lane 1). The membrane was then stripped and re-used for probing with anti-GFP antibody, 13 which revealed the presence of GFP-CcaIncA in the Ni-NTA associated fraction ( Figure 2C, right, lane 1). In control experiments, when the cells had been transfected with GFP and CcaIncA-His, or with GFP-CcaIncA alone, no signal was seen after the anti-GFP blotting, showing that neither GFP nor CcaIncA interacts directly with the NiNTA beads (data not shown).
An identical experiment was performed using cells co-expressing CtrIncA-His and GFP-CtrIncA. Probing with anti-GFP antibody showed that GFP-CtrIncA was retained with CtrIncA-His on the Ni-NTA beads, demonstrating that CtrIncA is also able to form dimers ( Figure 2C This similarity with SNARE domains, together with the fact that we showed that IncA dimerizes and probably tetramerizes, suggested to us a parallel organization of four CcaIncA helices similar to that found in the SNARE complex. We modeled a 39 residue stretch of the CcaIncA and CtrIncA sequences, 23 residues after the predicted transmembrane domain, as a parallel tetrameric coiled coil. These sequences are shown in Figure 3A, aligned with the four helices of the SNARE domain of the endosomal SNARE complex (31). Figure 3B shows helical wheel representation of the by guest on September 1, 2017 http://www.jbc.org/ Downloaded from CcaIncA tetramer, with some key interactions indicated by lines. Fig. 3C and D show the tridimensional structures of CcaIncA and CtrIncA models, respectively. In these models, the helices are connected by layers of hydrophobic amino acids with the exception of the hydrophilic central layer, like in the SNARE complex. The association of neighbouring helices is favored by numerous polar interactions. We next analyzed the stability of the modeled tetramer over time in aqueous solution. The simulations were performed at two temperatures, 27°C and 77°C. Both simulations showed that the modeled tetramers were very stable, presenting only small fluctuations around the average structure. Indeed, during the whole simulation, we observed neither separation of the monomers nor modification of the secondary structure. For comparison, we performed the same simulation on the SNARE complex. We observed that the molecular dynamics trajectory of the IncA models were very similar to those of the SNARE complex itself. The only difference was that the four central polar residues of the IncA tetramers exchanged conformations several times during the simulations while those of the SNARE complex did not. It is noteworthy that such exchanges of conformations were observed experimentally in the case of a dimeric leucine zipper (38). Movies of the simulations are available as supplementary material (Fig. 3Csup, 3Dsup and 3Esup). Altogether, our modelisation shows that CcaIncA and CtrIncA sequences are compatible with the formation of very stable four parallel helix bundles, resulting in a structure similar to the SNARE complex.

IncA-His expressed by HeLa cells localises to the endoplasmic reticulum
Using biochemical approaches, we showed that IncA molecules interact similarly when they localize in the inclusion, during normal infection, or when they are artificially expressed by a cell in the absence of inclusion. To determine the localisation of IncA in the latter case, we labelled cells transfected with IncA-His with anti-His antibody. Both CcaIncAHis and CtrIncA-His were found in a reticulate compartment extending throughout the cell. This compartment was identified as the endoplasmic reticulum (ER) using antibodies to calnexin, a well characterized ER marker (39) (Figure 4A). This distribution was not due to the His tag as it was also observed in cells transfected with GFP-tagged IncA constructs (see Figure 7A).
To determine the domains of IncA that are responsible for the cellular localisation of the protein, we constructed several truncated forms of CcaIncA-His ( Figure 4B). Deletion of the amino-terminal domain (∆53CcaIncA-His) had no effect on the distribution of the protein. which was enriched at the plasma membrane, suggesting that this mutant was able to exit from the ER. Altogether, these experiments show that the large hydrophobic domain of CcaIncA is necessary to anchor the protein in a membrane, but is not sufficient for ER retention, for which part of the carboxy-terminal domain is necessary.

Heterologous expression of IncA disrupts the infectious cycle
We have shown that IncA molecules interact with each other when they localize in a cellular membrane, either the inclusion, during normal infection, or the ER, when experimentally expressed in eukaryotic cells. In both cases, IncA can form dimers and higher molecular complexes.
We also showed that IncA could form a tetrameric structure similar to that of the four helix bundle of the SNARE complex. In the latter case, the complex involves SNARE proteins localized on two different membrane compartments. To test the relevance of our structural model in vivo, we asked whether IncA molecules expressed on two different compartments were able to interact with each other. To that end, cells were first transfected with CcaIncA-His for 24 h, or with GFP as a control, then infected with C. caviae GPIC for another 24 h before fixation and labeling of the inclusion and

Expression of IncA C-terminal cytoplasmic domain on the ER is sufficient to inhibit the inclusion development.
We tested the CcaIncA-His mutants for their ability to disrupt C. caviae infection, in order to determine the regions of IncA which are responsible for this effect. Cells expressing ∆53CcaIncA-His could not be infected by C. caviae, indicating that even in the absence of IncA Nterminal domain the disruptive effect was total ( Figure 6A). To know whether C c aIncA hydrophobic domain played a role in this effect, we created a chimera between hepatitis C virus nonstructural protein NS5B and CcaIncA C-terminal domain. The hydrophobic carboxy-terminal domain of NS5B forms a transmembrane α-helix that contains a signal for retention in the ER. It has been shown that this domain was sufficient to target a protein on the cytosolic side of the ER (41). Indeed, fusion of 26 amino acids from NS5B transmembrane domain with CcaIncA Cterminal cytoplasmic domain resulted in a chimera that had a distribution characteristic of the ER ( Figure 6B). HeLa cells expressing this chimera presented the same defect in the development of bacteria as HeLa cells expressing CcaIncaA-His, showing that IncA hydrophobic domain is not required for the inhibition of chlamydial development ( Figure 6A). Finally, transfection of CcaIncA C-terminal cytoplasmic domain (∆118CcaIncA-His), which has a cytosolic distribution, had no effect on subsequent infection by C. caviae, which developed into similar inclusions in transfected as in non-transfected cells. Altogether, these data show that, when anchored to the ER, CcaIncA Cterminal cytoplasmic domain is sufficient to disrupt the infection. IncA interacts with itself, shows that the disruption of the inclusion development is mediated by interactions between IncA proteins on the inclusion and on the ER. Therefore, we concluded that IncA molecules anchored in facing membranes associate to form multimeric structures.

DISCUSSION
In this paper we show that IncA from two species, C. trachomatis and C. caviae, associate with themselves to form multimers. First, in infected cells, we showed using a cross-linker that CcaIncA participated in high molecular weight complexes whose sizes were compatible with the cross-linking of two or more CcaIncA molecules. High molecular weight complexes were also present in cells transfected with CcaIncA and CtrIncA treated with a cross-linker, showing that no other bacterial protein than IncA was necessary for their formation. Finally, we showed, using cells transfected with two IncA constructs linked to different tags, that IncA molecules interacted with each other in cell extracts. Our results are in partial agreement with a previous report in which a yeast two-hybrid system was used to show interactions between CtrIncA molecules (16). However, interactions between CcaIncA molecules were hardly detected with this technique, supporting the authors' conclusion that the function of IncA proteins may differ between species. Here we show that the ability of IncA to form multimers is common to the two species, and we propose that this property underlies similar function(s) during infection.
When artificially expressed in HeLa cells, IncA was localized in the ER, as shown by colocalization with an ER marker, calnexin. The same distribution was observed for other Inc proteins heterologously expressed by the host cell, although some were detected on the plasma membrane (data not shown). Inc proteins are normally synthesized by the bacteria and are transported through the bacterial and inclusion membranes via a type III mechanism. Their insertion in the inclusion membrane may be coupled to the translocation process. In the case of heterologous expression, Inc proteins may be recognized by the cellular machinery as a substrate for ERassociated ribosomes and may be inserted into the ER membrane co-translationally. Alternatively, the association might be post-translational, due to a better affinity for the ER membrane lipid composition than for other cellular membranes, or via interactions with ER membrane proteins. The large bilobed hydrophobic domain which is common to all Inc proteins is probably essential for their interaction with membranes, and deletion of this domain in CcaIncA (∆118CcaIncA-His) resulted in a cytosolic distribution. However the hydrophobic domain is not sufficient to ensure ER location as deletion of part of CcaIncA C-terminal cytoplasmic domain (CcaIncA∆135His), preserving the entire hydrophobic domain, resulted in preferential localization to the plasma membrane. We did not directly address the topology of IncA insertion into the ER membrane.
However, we observed, as described previously, that heterologously expressed CcaIncA-His showed the same migration profile in SDS-PAGE as endogenous CcaIncA ( Figure 2B), with three bands corresponding to three phosphorylation states of the protein (15). This finding strongly supports the hypothesis that heterologously expressed CcaIncA-His is inserted in the ER membrane with the same topology as in the inclusion membrane, with at least the carboxy-terminal domain exposed to the cytosol (16).
Expression of GFP-IncA in infected cells had a very striking effect. Soon after the beginning of appearance of IncA, the morphology of the inclusion was modified, with a scattering of the bacteria and simultaneous distortion of the ER. Later after transfection, all infected cells had disappeared from the transfected population and many dead cells had detached from the culture plate, strongly suggesting that heterologous expression of IncA in the ER of infected cells resulted in cell death. When the experiment was performed in the reverse order, by infecting cells which had been transfected prior to infection, no inclusion developed in the transfected population. This was not due to an inhibition of bacteria entry, as we measured that bacteria entry was not affected by IncA expression. In fact, infection with C. caviae seems to proceed normally during the first 10 hours. Fifteen hours after infection, the inclusions were disrupted. This correlates with the kinetics of incA expression, as its transcript was first detected by RT-PCR at 10 hours post infection in C.

caviae infected cells (not shown)
To determine which domain of IncA was involved in the inhibition of chlamydial development, when expressed by transfection, we measured infection in cells transfected with truncated forms of IncA. Deletion of the amino-terminal domain did not suppress the deleterious effect of IncA expression showing that IncA amino-terminal domain was not required. In fact, the C-terminal cytoplasmic domain alone, targeted to the ER using a transmembrane ER targetting signal from hepatitis C virus non-structural protein NS5B, was sufficient to produce the inhibitory effect observed with full length IncA. Finally, expression of CcaIncA C-terminal cytoplasmic domain (∆118CcaIncA-His), which has a cytosolic distribution, had no effect on the infection.
These data suggest that to disrupt the infection, CcaIncA C-terminal cytoplasmic domain needs to be anchored to a cellular compartment. We do not know at this stage whether this compartment needs to be the ER, which is naturally distributed in the whole cytosol and therefore in proximity to the inclusion. It may be that expression of IncA at another intracellular compartment, such as the Golgi apparatus, also found in close proximity to the inclusion, would similarly disturb inclusion development. However this hypothesis would be difficult to test, as the only known signals for targeting proteins to the Golgi apparatus are in the cytosolic portions of these proteins, which would need to be replaced with IncA carboxy-terminal domain.

20
The deleterious effect on chlamydial development was observed only with IncA expression, not with C. caviae IncB, which is also expressed on the ER by transfection, showing that the effect was specific of this Inc protein (data not shown). Using a strain that does not express IncA on the surface of its inclusions, we were able to demonstrate that the deleterious effect of IncA expression on chlamydial development required the presence of IncA on the membrane of the inclusion. This result, together with our demonstration that IncA molecules interact with each other, show that the disruption of the inclusion development is mediated by direct interactions between IncA molecules from the two compartments in which they localize. Inhibition on chlamydial development was complete only when IncA corresponding to the strain used for the infection was expressed, showing that the effect is very specific. However partial interspecies inhibition was also observed, suggesting that, although very different in terms of sequence, C. caviae and C.
trachomatis IncA biochemical properties are sufficiently close so that they can interact with each other to some extent. It supports the idea that IncA may fulfill similar roles during infection by different species.
In infected cells, expression of IncA in the ER leads to the disruption of the inclusion, and to changes in the morphology of the ER. Our microscopy observations are consistent with the hypothesis that these interactions lead to fusion events between the inclusion and the ER.
Observation at a better resolution would be necessary to support this hypothesis. Unfortunately, the process is not easy to observe because infected cells are transfected with a low efficiency (independently of the transfected gene), and the few infected and transfected cells die rapidly after IncA expression. Our data therefore support previous reports indicating that IncA may play a role in the fusion of inclusions. The aim of the modeling and molecular dynamics studies was to get insights into possible mechanisms of IncA-induced in vesicle fusion. Amphipatic helices alone can be sufficient to induce membrane fusion in vitro (42). However, a fusion mechanism similar to the one involving the formation of the SNARE complex is attractive because of the similarity of the sequences and their properties: there is strong coiled-coil or leucine zipper propensity; the coiled coil sequence starts shortly after a putative transmembrane region; there is a conserved hydrophilic residue in the interface that may serve to organize the helices, infer specificity, or to bias the equilibrium between multiple possible oligomeric states; finally, the tetramer proved very stable in molecular dynamics calculations at 300 and 350 K.
The model is based on the capability of the IncA sequences to form stable oligomers.
Whereas the stability in molecular dynamics calculations is an indication that the model is  (43). These properties are met in our structural modelisation: (i) beta-branched residues at "a" positions, that could disfavor tetramers due to rotamer preferences of these residues, are absent in both modeled sequences. (ii) a polar residue in the interface, common in coiled coils and which may serve to organize the four helices such that the four hydrophilic residues form one layer, is found in both modeled sequences. (iii) residues on the surface of the tetramers form interactions; as an example, the polar interhelical interactions involving "b", "c", "g" and "e" positions are shown in Figure 3B. (iv) interactions involving "b" and "c" positions are more likely in tetrameric than in dimeric coiled coils because of the different relative orientations of the helices in tetramers. (v) there are no repulsions between residues of like charge in neighboring helices. For CcaIncA, an antiparallel orientation of the helices, on the other hand, would bring the residues on the "g" position on one helix into contact with those on the neighboring helix, thus leading to six Arginine residues into close proximity. We conclude that a parallel orientation of the helices is more likely.
It is generally agreed that these membrane-bridging SNARE complexes mediate membrane fusion directly (44,45). Our IncA model proposes a parallel tetramer formed from helices from two different vesicles. The transmembrane regions of the four monomers are all on the same side of this rod-like structure, similar to proteins involved in viral fusion and in membrane fusion involving the SNARE complex. The formation of the tetramer would therefore induce close approach of the membranes, and possibly destabilize the membranes locally close to the transmembrane domains (46). In that case, one could expect expression of IncA to disturb the ER structure, although this compartment has homotypic fusion abilities of its own. Indeed, we noted that expression of CtrIncA slightly modified the ER, whose network had a somewhat rougher appearance compared to that in non transfected cells (see Fig. 4A). Whether formation of IncA tetramers by itself is sufficient to promote fusion remains to be examined. The difference in the fusogenicity of C. trachomatis and C.
caviae inclusions may indicate that IncA by itself is not sufficient to trigger fusion between inclusions, and that accessory molecules, absent from C. caviae, participate in the fusion of C. trachomatis inclusions. This hypothesis is supported by the observation that some of the nonfusogenic isolates do express IncA, suggesting that other factors necessary for the fusion to occur are missing from these strains (27).
Interestingly, we found, using BLAST analyses, that the genome of the chlamydia-related symbiont of free-living amoebae (UWE25) encodes an ORF annotated pc0399 which presents some sequence similarity with IncA (48). Moreover, it shares the characteristic organization of IncA proteins, with a short amino-terminal domain, a large bilobed hydrophobic domain and a carboxyterminal domain that is predicted to engage in coiled-coils interactions. If this molecule is present on the surface of the inclusion, it suggests that IncA has coevolved with its host cells since the divergence of the pathogenic and symbiotic chlamydiae, more than 700 million years ago.
Finally, we would like to speculate that, in addition to its role in the fusion of inclusions, IncA may participate in the fusion of cellular vesicles with the inclusion membrane. The fact that IncA structure is compatible with the participation to a SNARE-like complex suggests that this molecule may have evolved to interact not only with itself but also with cellular SNAREs to form fusion-competent SNARE complexes. One possibility is that IncA mimicry with host SNARE proteins could enable the bacteria to hijack part of the cellular traffic by allowing fusion of vesicles with the inclusion.  Table 1. Primers used for plasmid construction by PCR.