Escherichia coli (E. coli) XL-1 (Stratagene) was used for plasmid amplification and BL21 (Stratagene) for protein expression. Transformed E. coli strains were cultured in LB medium containing the appropriate antibiotics. C. pneumoniae GiD and C. trachomatis L2 (434) were propagated in HEp-2 cells (ATCC No.: CCL-23) and purified as described previously (Jantos et al., 1997; Wuppermann et al., 2008). The Saccharomyces cerevisiae strain CEN.PK2 (MATa/MATα ura3-52/ura3-52 trp1-289/trp1-289 leu2-3,112/leu2-3,112 his3-Δ1/his3-Δ1) (EUROSCARF) was used for the homologous recombination cloning and the characterization of CPn0572 in yeast. The aip1Δ (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 YMR092c::kanMX4) strain and the corresponding wt strain BY4741 (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) (both from EUROSCARF) were used for cofilin (Cof1p) localization. Plasmid-bearing strains were grown on selective synthetic dextrose (SD) medium supplemented with 2% (wt/vol) glucose, raffinose or galactose (Sherman, 1991). HEK293T cells (ATCC No.: CRL-11268) were routinely cultured in IMDM medium (Invitrogen) supplemented with 10% (vol/vol) fetal calf serum (FCS; Invitrogen).

The following Cpn0572 and Tarp protein were used (accession numbers are included):

C. pneumoniae Cpn0572 CWL029 NP_224768.1

Levels of similarity and identity between Cpn0572 and Tarp were calculated by EMBOSS http://emboss.bioinformatics.nl/cgi-bin/emboss/needle. Multiple sequence alignments of DUF domains from CPn0572 and its chlamydial orthologs were performed using MultAlin: http://bioinfo.genotoul.fr/multalin/multalin.html. The prediction of α-helices in DUF was carried out with the helical wheel projection prediction program (http://rzlab.ucr.edu/scripts/wheel/wheel.cgi).

Plasmids were constructed either by homologous recombination in S. cerevisiae or by ligation as indicated. In general, the plasmids used to generate the plasmids required for this study are listed in Table S1. The bacterial DNA sequences were amplified from genomic DNA or pre-existing plasmids using the oligonucleotides listed in Table S2. For expression of CPn0572_6His in E. coli, the pAC2 vector was cut at BglII and EcoRI sites and the coding sequence of cpn0572 was integrated by homologous recombination in yeast. For expression of CPn0572 or its variants in mammalian cells, amplified DNA fragments were cloned into the pBYE vector (a modified pcDNA3.1/NT-GFP, see Table S1) at the Acc65I site. For yeast growth tests, the DNA coding sequences of cpn0572 and its derivatives were cloned into the p426MET25 vector at the SmaI site (Mumberg et al., 1994). For localization studies in yeast cpn0572 and its derivatives were also cloned into the pUG34 vector at the SmaI site via homologous recombination in yeast. Only the DNA fragment encoding GFP-C-terminal-ΔDUF was cloned into pUG34 by ligation. To do so, the DNA fragment was isolated from a p426MET25-GFP-C-terminal-ΔDUF vector by cleavage with SpeI and XhoI and cloned into the same sites in pUG34. To express GST fusion proteins in E. coli, cpn0572 or the coding sequence for DUF or C-terminal-ΔDUF was amplified and integrated into the pKM36 vector at the SmaI site by homologous recombination. For expression of CPn0572 in the aip1Δ mutant strain and in wild type BY4741, the coding region of cpn0572_6His was isolated from the p426MET25-cpn0572 vector using SpeI and XhoI, and integrated at the corresponding sites in p423GAL1 by ligation.

HEp-2 or HEK293T cells grown overnight on coverslips were transfected with 0.5–1 μg of the desired DNA plasmid using Turbofect (Thermo Scientific) as recommended by the manufacturer. Transfected cells were incubated in a CO₂ incubator at 37°C for the indicated time periods (16 to 24 h). Human HEp-2 cells expressing mCherry-actin were infected with C. pneumoniae by adding purified EBs (suspended in cold DMEM medium) to attached HEp-2 cells, followed by centrifugation at 2,800 rpm, 4°C for 20 min. The medium was replaced by pre-warmed fresh medium and the cells were shifted to 37°C for the time periods indicated in the Figure legends. Samples were then prepared for microscopy as described below.

Yeast F-actin was stained with 6.6 μM rhodamine-phalloidin (Molecular Probes). To visualize cofilin and CPn0572_6His, cells were fixed and indirectly immunostained as described previously (Pringle et al., 1991), using polyclonal chicken antibodies against cofilin (Okada et al., 2006) and a monoclonal mouse anti-His antibody (Novagen).

For indirect immunostaining infected or transfected human cells growing on glass coverslips were washed with 1x PBS, fixed in 3.7% (wt/vol) paraformaldehyde and permeabilized with 0.1% Triton X-100 (vol/vol). Permeabilized infected cells were then incubated with antibodies against C. pneumoniae MOMP (anti-MOMP antibody was kindly provided by G. Zhong, University of Texas Health Science Center, San Antonio, USA) to identify EBs and CPn0572 (anti-CPn0572 antibody was produced against the recombinant central domain of CPn0572 (aa S343 to A535) in rabbits by Eurogentec SA, Seraing, Belgium), and subsequently with labeled secondary antibodies. Staining of actin in human cells with rhodamine-phalloidin was performed as recommended by the manufacturer (Molecular Probes). Microscopy was carried out using either a spinning-disk confocal instrument or an Axiovert 200 microscope (Carl Zeiss).

For detection of GFP-tagged proteins from yeast on Western blots, a rabbit anti-GFP antibody (Molecular Probes) was employed, and for detection of proteins from human cells a mouse anti-GFP antibody (Thermo Scientific) was used.

Yeast cells in mid-log phase were exposed to either Latrunculin-A (Lat-A) diluted in DMSO or to the solvent DMSO alone (final concentration 2.5 μM). At various times after addition of the agent, cells were removed, fixed immediately, and stained with rhodamine-phalloidin. The actin cytoskeleton of transfected HEK293T cells was disrupted by adding cytochalasin D (CD; Sigma) to the medium (final concentration 10 μM) for 30 min at 37°C.

F-actin was assembled in vitro as recommended by the supplier (Cytoskeleton Inc.). Equimolar amounts (2 μM) of recombinant human cofilin (Cytoskeleton) and pre-assembled F-actin were then incubated at pH 6.8 in the presence or absence of recombinant GST-CPn0572 (2 μM). The assay was then performed as described by the manufacturer.

CPn0572 is currently annotated as a hypothetical C. pneumoniae protein. When we used a polyclonal antibody raised against the central region of the predicted ORF (CPn0572^S343−A535) to probe western blots bearing protein extracts obtained from purified C. pneumoniae EBs, the antiserum (but not the corresponding preimmune serum) recognized a band of 110–115 kDa, similar in size to the recombinant full-length protein produced in E. coli (rCPn0572_6His), indicating that CPn0572 is present in infectious EBs (Figure 1A and Figure S1).

To determine whether the protein is translocated into the host-cell cytoplasm, infected HEp-2 cells were fixed at various time points after exposure to C. pneumoniae, and analyzed by immunofluorescence microscopy. At time 0, we observed no significant signal with anti-CPn0572 (Figure 1B). After 15 min, some bacteria were found to lie close to, or to partially overlap CPn0572 signals, as has previously been shown for the C. trachomatis Tarp and other secreted chlamydial proteins (Clifton et al., 2004; Hower et al., 2009). These results suggest that secretion of CPn0572 from the EBs had begun prior to this point. After 30 min and 60 min of infection, CPn0572 was detected in association with almost every EB, and the intensity of the signal increased over time (Figure 1B). To investigate whether EBs induce accumulation of actin at sites of entry, and to determine if CPn0572 is associated with actin, HEp-2 cells expressing mCherry-actin were infected with C. pneumoniae EBs. Adherent EBs were found to lie close by or immediately adjacent to CPn0572 signals and actin patches as early as 15 min p.i. (Figure 2A). These results indicate that CPn0572 is secreted from EBs into HEp-2 cells and is found in the vicinity of patches of F-actin accumulation at sites of EB entry.

CPn0572 and its C. trachomatis ortholog Tarp show 22.9% sequence identity and 32.5% similarity; moreover, Tarp harbors an N-terminal extension of 276 amino acids (aa), while CPn0572 carries a unique 16-aa C-terminal extension. Interestingly, all Tarp orthologs harbor at least one copy of DUF, which overlaps with the minimal actin-binding and protein oligomerization region required for actin nucleation (Jewett et al., 2010). To determine whether CPn0572 expression in human cells results in an actin phenotype similar to that seen for the expression of C. trachomatis Tarp (Clifton et al., 2004; Jiwani et al., 2013), we expressed GFP, GFP-Tarp and GFP-CPn0572 in human HEK293T cells. In each cell expressing GFP-Tarp, the fusion protein was found exclusively in one or more discrete actin-containing structures (Figure 2B), similar to what has been described for HeLa cells expressing GFP-Tarp (Clifton et al., 2004; Jiwani et al., 2013). Interestingly, in all cells transfected with GFP-CPn0572, the fusion protein was detected not only in such actin-containing aggregates, but also in distinctly filamentous structures emanating from them (Figure 2B, merge, upper panel). Moreover, in more than 80% of the transfected cells, not all GFP-CPn0572-positive filaments were coincident with phalloidin-positive signals (Figure 2B; lower panel; compare the Cpn0572-positive fibers marked by an arrow and an arrowhead, respectively, and note the pair of Cpn0572-labeled non-actin filaments seen above the latter). As expected, GFP-CPn0572 expressed in human epithelial HEp-2 cells associated with actin aggregates and actin filaments but also formed fibers not associated with F-actin (Figure S2). These results suggest that CPn0572, like Tarp, associates with and alters the organization of the actin cytoskeleton independently of other chlamydial effectors. However, CPn0572 differs from Tarp in that it can also form fibers that apparently do not contain F-actin.

Because many of the processes and proteins in host cells that are altered or otherwise subverted during bacterial infections are conserved among eukaryotes (Valdivia, 2004; Siggers and Lesser, 2008; Zrieq et al., 2015), we used the yeast S. cerevisiae as a model system to investigate the role of CPn0572 in modulating cytoskeletal organization in greater detail. First, we expressed CPn0572_6His from an inducible promoter, and studied its effects on growth rate using serial-dilution patch tests. CPn0572_6His-expressing cells displayed a severe growth defect, implying that CPn0572 disrupts an essential process in yeast (Figure 3A). We therefore examined the actin cytoskeleton in cells expressing GFP-CPn0572, and found the fusion protein in one large structure which colocalized with F-actin (Figure 3B). Moreover, a few of these aggregates were found to be associated with small numbers of short actin cables.

Collectively, these findings demonstrate that CPn0572 severely perturbs the yeast actin cytoskeleton by transforming actin structures into aggregates, as it does in human cells expressing GFP-CPn0572. Expression of the C. trachomatis Tarp in yeast causes a similar phenotype (Sisko et al., 2006). These data confirm the utility of yeast as a model for further studies of the role of CPn0572 in remodeling the actin cytoskeleton.

The DUF domain (59 amino acids) is the most striking element common to CPn0572 and its orthologs in other Chlamydia species (Figure 3C). It is predicted to consist of three α-helices (Figure 3D) and it forms part of the shortest fragment of CPn0572 (101 amino acids) found to pull down actin in vitro, while it overlaps partially with the potential actin-binding domain (ABD) (Jewett et al., 2010). Therefore, we suspected that DUF might be linked to the actin phenotype. Indeed, yeast cells expressing either an N-terminal (M¹-A⁵³⁶) or C-terminal (Q⁴⁷⁸-K⁷⁵⁵) segment of CPn0572 including DUF displayed a growth phenotype like that seen with the full-length CPn0572 (Figure 4B), while deletion of the DUF domain alone (Q⁴⁷⁸-A⁵³⁶, ΔDUF, Figure 4A) restored normal growth of yeast cells expressing this variant (Figure 4B). In contrast, expression of the DUF domain (Figure 4A) on its own had no effect on growth (Figure 4B). Interestingly, deletion of DUF domain from the C-terminal fragment (C-terminal-ΔDUF, Figure 4A) restored normal yeast growth (Figure 4B; compare C-terminus vs. C-terminal-ΔDUF). Taken together, these results imply that DUF is essential, but not sufficient, for the severe growth phenotype.

We then examined the actin cytoskeleton in yeast cells expressing the different Cpn0572 variants. The GFP-N-terminal segment of the fusion protein colocalized with actin structures comparable to those induced by the full-length protein (Figure 4C). Interestingly, cells expressing the GFP-C-terminal segment showed a very different actin pattern, characterized by complete colocalization of the fusion protein with long, thick actin cables and little cytosolic GFP-C-terminus staining (Figure 4C). In contrast, cells expressing the GFP-C-terminal-ΔDUF segment showed typical actin cables and patches found in wild-type cells and a diffused distribution of the fusion protein comparable to that of cells expressing GFP alone, indicating that the lack of toxicity and actin remodeling is due to the absence of the DUF domain (Figure 4C). As expected, cells expressing the GFP-ΔDUF construct showed diffuse GFP staining in the cytosol, and their actin cytoskeleton was unaffected (Figure 4C). Finally, GFP-DUF itself was found to stain distinctive, unusually shaped and mostly cortically localized actin cables against a strong and diffuse background of the protein in the cytosol (Figure 4C). These data indicate that the DUF region of CPn0572 is crucial for the association of CPn0572 with F-actin (Figure 4C). The differences in the localization of the different GFP-CPn0572 segments are unlikely to be due to major alterations in protein expression levels (Figure S3A). Taken together, these observations strongly suggest that DUF is required for the association of CPn0572 with F-actin in yeast cells, while other elements further modulate the actin phenotype, which then impairs yeast cell growth.

Next we tested these findings in human cells. As shown above, GFP-CPn0572 was found to localize to complex actin structures, and all CPn0572-expressing cells showed the fusion protein filaments colocalizing with some, but not all actin fibers (Figure 5, upper panel, Z: 29; compare arrow and arrowhead in Z: 16). In contrast, in more than 80% of transfected cells, most GFP-ΔDUF did not colocalize with actin fibers (Figure 5, lower panels; see arrow in Z: 16), although this variant was occasionally associated with F-actin aggregates (Figure 5, lower panels, MIP and Z: 29). Strikingly, however, the vast majority of GFP-ΔDUF was found in the form of filamentous structures that did not contain actin (Figure 5, lower panel). The phenotypic differences seen for GFP-CPn0572 and GFP-ΔDUF are likely not due to differences in protein expression (Figure S3B). These results confirm that DUF is required for the association of CPn0572 with F-actin in human cells. Moreover, CPn0572 has a DUF-independent ability to generate or interact with other filamentous structures.

Ectopic expression of CPn0572 in eukaryotic cells induces aggregation of F-actin into thick cables, which resemble analogous structures observed in yeast and human cells under conditions in which actin polymerization is favored over F-actin depolymerization (Ayscough, 2000; Lázaro-Diéguez et al., 2008). In part this can be explained by CPn0572's known ability to nucleate and promote the polymerization of F-actin in vitro (Jewett et al., 2010). However, the fact that in human cells CPn0572 exhibits continuous colocalization with distinct actin fibers suggested that CPn0572 might also stabilize F-actin against the action of actin-depolymerizing agents. To test this hypothesis, yeast cells expressing GFP alone or GFP-CPn0572 were exposed to Latrunculin A (Lat-A), which binds actin monomers, thus preventing their polymerization and effectively promoting depolymerization (Coué et al., 1987; Ayscough et al., 1997; Belmont and Drubin, 1998; Okada et al., 2006). Control cells expressing GFP alone were devoid of actin cables after treatment with Lat-A for 5 min, while some cortical actin patches were still visible after 1 h of exposure to Lat-A (Figure 6A). In contrast, the massive actin structures induced by GFP-CPn0572 were still detectable even after exposure to Lat-A for 1 h. Similarly, yeast cells expressing the GFP-C-terminus or GFP-DUF retained actin cables; however, in the presence of Lat-A, the GFP-C-terminus associated with more distinctively defined (compared to the GFP-DUF construct) cables and patches which were also phalloidin-positive (Figure 6B). Thus, CPn0572 stabilizes F-actin in yeast by inhibiting depolymerization in vivo.

ADF/cofilins are key regulatory proteins that bind to and sever F-actin in vitro and in vivo (Elam et al., 2013). We therefore tested the hypothesis that CPn0572 might alter F-actin dynamics by interfering with cofilin function. We determined the localization of cofilin in yeast by expressing CPn0572_6His in an aip1Δ strain, in which both actin patches and cables are enriched for cofilin, and cofilin can be visualized on all actin structures (Okada et al., 2006). In aip1Δ cells carrying the empty vector, cofilin was detected in structures interpreted to represent actin cables and patches, as previously documented (Figure 6C, upper panel). Remarkably, in CPn0572_6His-expressing cells, this clear localization of cofilin was completely abolished, and only a diffuse cytosolic staining pattern was detectable (Figure 6C, lower panel; CPn0572_6His-expressing cells marked by arrowheads), while in the same micrograph, yeast cells with little or no CPn0572_6His expression retained actin cables and patches (Figure 6C, lower panel; cells marked by arrows). Similarly, CPn0572_6His also altered cofilin localization in a wild-type yeast strain (Figure S4). These results strongly suggest that CPn0572 inhibits F-actin depolymerization in yeast by displacing cofilin from F-actin.

Next we used in vitro assays with recombinant proteins to test whether the displacement of cofilin from filamentous actin by CPn0572 was a direct or indirect effect of the latter. Pre-assembled mammalian F-actin was mixed with recombinant human cofilin in the absence or presence of an equivalent amount of GST-CPn0572, and the mixture was subjected to ultracentrifugation. When F-actin was mixed with cofilin alone, 100% of the cofilin was found in the pellet, as expected given its ability to bind to F-actin (Figure 7A). However, when recombinant GST-CPn0572 was present in the mixture, cofilin was now found exclusively in the supernatant, and the fusion protein was found in the pellet with F-actin. Thus, these results show that CPn0572 binds F-actin and competes with cofilin for, and/or displaces it from, F-actin. Further analysis revealed that the GST-C-terminus displaced about 90% of the cofilin into the supernatant. Finally, the GST-DUF domain was able to displace only about 20% of the cofilin into the supernatant and a 2-fold excess of DUF was necessary to bring about 80% of the cofilin into solution in vitro (Figure 7B). These results are consistent with our finding that DUF displays less colocalization with F-actin, and is a less potent stabilizer of F-actin in the presence of Lat-A. Thus, the ability of CPn0572 to prevent binding of cofilin to F-actin largely resides in its C-terminal segment.

We also probed the ability of CPn0572 to stabilize F-actin in human cells. When we destabilized F-actin by incubating HEK293T cells with cytochalasin D (CD), which inhibits actin filament dynamics in cells via multiple mechanisms, GFP-expressing control cells showed depolymerized F-actin with some dispersed cortical actin remaining intact (Figure 8, upper and lower panels). In contrast, CD had little effect on actin structures and fibers in HEK293T cells expressing GFP-CPn0572 in comparison to untreated cells (Figure 8, upper and lower panels). Interestingly, the actin aggregates induced by the C. trachomatis Tarp were also resistant to CD (Figure 8). Most interestingly however, treatment of HEK293T cells expressing a GFP-ΔDUF fusion with CD resulted in the complete loss of F-actin structures. Moreover, no colocalization of the remaining cortical actin with GFP-ΔDUF was observed, supporting our contention that the DUF domain is crucial for stabilization of F-actin by CPn0572 (Figure 8). Overall, these data show that CPn0572 and Tarp stabilize F-actin against depolymerizing and severing factors present in human cells.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.