The Chlamydia pneumoniae Tarp Ortholog CPn0572 Stabilizes Host F-Actin by Displacement of Cofilin

Pathogenic Chlamydia species force entry into human cells via specific adhesin-receptor interactions and subsequently secrete effector proteins into the host cytoplasm, which in turn modulate host-cell processes to promote infection. One such effector, the C. trachomatis Tarp factor, nucleates actin polymerization in vitro. Here we show that its C. pneumoniae ortholog, CPn0572, associates with actin patches upon bacterial invasion. GFP-CPn0572 ectopically expressed in yeast and human cells co-localizes with actin patches and distinctly aberrantly thickened and extended actin cables. A 59-aa DUF 1547 (DUF) domain, which overlaps with the minimal actin-binding and protein oligomerization fragment required for actin nucleation in other Tarp orthologs, is responsible for the aberrant actin phenotype in yeast. Interestingly, GFP-CPn0572 in human cells associated with and led to the formation of non-actin microfilaments. This phenotype is strongly enhanced in human cells expressing the GFP-tagged DUF deletion variant (GFP-ΔDUF). Finally ectopic CPn0572 expression in yeast and in-vitro actin filament binding assays, demonstrated that CPn0572 stabilizes pre-assembled F-actin by displacing and/or inhibiting binding of the actin-severing protein cofilin. Remarkably, the DUF domain suffices to displace cofilin from F actin. Thus, in addition to its actin-nucleating activities, the C. pneumoniae CPn0572 also stabilizes preformed host actin filaments.


INTRODUCTION
Chlamydiae are Gram-negative intracellular pathogens that infect a variety of organisms and tissues, and are responsible for several serious respiratory, ocular and urogenital diseases (Schachter, 1999). All Chlamydia species have a biphasic developmental cycle, alternating between the infectious but metabolically inert elementary body (EB) and the non-infectious, metabolically active reticulate body (RB). RBs replicate within a parasitophorous vacuole, termed an inclusion (Schramm et al., 1996;Belland et al., 2004). Successful uptake of EBs is crucial for Chlamydia infection, but the underlying molecular mechanisms are not well-understood.
Generally, the ability of bacterial pathogens to enter host cells depends upon cross-talk between bacterial and host factors, beginning with direct engagement of receptors on the target cell by adhesins, and/or translocation of effector proteins into the host-cell cytosol. These processes usually result in a rearrangement of the host-cell cytoskeleton, which in turn promotes a reorganization of the host plasma membrane architecture that facilitates bacterial uptake (Pizarro-Cerdá and Cossart, 2006). Initial attachment of the chlamydial EB is normally mediated by the interaction of the chlamydial surface protein OmcB with glycosaminoglycans (GAGs) on the host-cell surface, and is followed by more specific adhesin-receptor interactions (Hegemann and Moelleken, 2012). Thus, the C. pneumoniae adhesin/invasin Pmp21 binds directly to the human epidermal growth factor receptor (EGFR), activating signaling cascades that facilitate the uptake of C. pneumoniae EBs into their target cells (Mölleken et al., 2013). Moreover, the EB surface protein CPn0473 also mediates adhesion to human epithelial cells, and promotes EB uptake in a lipid-raft-dependent manner (Fechtner et al., 2016).
The C. trachomatis protein Tarp (translocated actin-recruiting phosphoprotein) is an early virulence effector protein implicated in host-cell invasion (Clifton et al., 2004;Lane et al., 2008;Jewett et al., 2010;Parrett et al., 2016). Tarp, which is assumed to be secreted by a Type-3 secretion system via Slc1 (SycE-like chaperone 1; CT043), is translocated into targeted cells within minutes of EB attachment, and associates with recruited actin at the site of bacterial attachment (Clifton et al., 2004;Brinkworth et al., 2011). This is accompanied by phosphorylation of several tyrosine residues near the N-terminus of Tarp by Src family tyrosine kinases and Ab1 kinase (Clifton et al., 2004;Jewett et al., 2008;Mehlitz et al., 2008). However, Tarp phosphorylation is not essential for chlamydia entry or actin recruitment. The protein most probably acts as a molecular scaffold to recruit host proteins that regulate actin dynamics and signaling events required for the early phase of chlamydial infection (Clifton et al., 2005;Jewett et al., 2008;Mehlitz et al., 2008;Thwaites et al., 2014).
Recruitment of actin to attached EBs early in the infection, in a pattern similar to that seen in C. trachomatis, has been observed for a number of chlamydial species (Clifton et al., 2005). However, although orthologs of the C. trachomatis Tarp gene are present in all sequenced Chlamydia species, they differ widely in amino acid sequence (displaying between 40 and 94% identity), domain structure and length (Clifton et al., 2005;Jewett et al., 2010;Jiwani et al., 2013), with the least conserved being the C. pneumoniae orthologs. For example, the C. muridarum and C. pneumoniae orthologs (but not the C. caviae ortholog) lack the tyrosine repeats (Clifton et al., 2005). In contrast, all Tarp orthologs harbor a protein oligomerization domain, and the actin-binding domains found in all examined chlamydial strains and species are able to nucleate actin polymerization in vitro. Interestingly, C. pneumoniae Tarp is the sole ortholog that has only a single actin-binding domain (Jewett et al., 2010). Recently, evidence was provided that the Tarp orthologs from serovars of C. trachomatis harbor two F-actin binding domains which seem to be absent from Tarp orthologs in other chlamydial species (Jiwani et al., 2013). Moreover, binding domains for the focal adhesion kinase (FAK) and for vinculin have been identified for Tarp proteins from various chlamydial species, but are not found in the C. pneumoniae ortholog (Thwaites et al., 2014(Thwaites et al., , 2015. To elucidate the functional consequences of these differences, we characterized CPn0572, the putative C. pneumoniae ortholog of Tarp. CPn0572 is secreted into the host-cell cytoplasm upon EB uptake and is associated with actin recruitment to the site of entry. Ectopically expressed CPn0572 stabilizes actin filaments (F-actin) both in yeast and in human HEK293T cells. Moreover, CPn0572 has the ability to generate or interact with other filamentous structures not associated with F-actin structures. Interestingly, detailed analysis revealed that CPn0572 blocks the disassembly of F-actin by displacing the F-actin destabilizing protein cofilin. Thus, in addition to the known function of Cpn0572 in nucleating F-actin, our findings provide evidence that the protein (like the prototypical Tarp from C. trachomatis) is a microbial F-actin-stabilizing protein. To our knowledge, this is the first bacterial effector protein that has been shown to directly modulate both actin nucleation/polymerization and depolymerization in host cells.

Comparison of Protein Sequences
The following Cpn0572 and Tarp protein were used (accession numbers are included):

DNA Manipulations and Plasmid Construction
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-DUF was cloned into pUG34 by ligation. To do so, the DNA fragment was isolated from a p426MET25-GFP-C-terminal-DUF 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-DUF was amplified and integrated into the pKM36 vector at the SmaI site by homologous recombination. For expression of CPn0572 in the aip1 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.

Host-Cell Transfections and Infection with C. pneumoniae
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 CO 2 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.

Fluorescence Microscopy and Western Blots
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.

Disruption of the Actin Cytoskeleton
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/Cofilin Co-sedimentation Assay
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 Secreted into HEp-2 Cells Early During Infection by C. pneumoniae and Associates with F-Actin
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.
FIGURE 1 | C. pneumoniae EBs express CPn0572 and secreted CPn0572 associates with F-actin. (A) CPn0572 is present in EBs. rCPn0572 6His protein and the endogenous protein expressed in C. pneumoniae EBs (EB lysates) were detected on immunoblots probed with a monoclonal anti-His antibody (left panel) and a polyclonal rabbit anti-CPn0572 antibody (middle panel), respectively. The blot shown on the right was probed with the corresponding preimmune serum. Full-scale images of the original blots are presented in Figure S1. (B) CPn0572 secretion increases over the first 60 min p.i.. Secretion of CPn0272 was investigated at 0, 15, 30, and 60 min after HEp-2 cells had been infected with purified C. pneumoniae EBs (MOI = 50). EBs and CPn0572 were detected using anti-MOMP and anti-CPn0572 antibodies, respectively.
Frontiers in Cellular and Infection Microbiology | www.frontiersin.org 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.

Tarp and CPn0572 Show Different Patterns of Localization and Association with F-actin in Transfected HEK293T Cells
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 actincontaining 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.

CPn0572 Inhibits Yeast Growth and DUF Mediates Disruption of 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 6Hisexpressing 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 (M 1 -A 536 ) or C-terminal (Q 478 -K 755 ) 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 (Q 478 -A 536 , DUF, 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-DUF, Figure 4A) restored normal yeast growth ( Figure 4B; compare C-terminus vs. Cterminal-DUF). 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-DUF 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-DUF 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 Factin ( 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  Jewett et al., 2006Jewett et al., , 2010 and the DUF domain are indicated. The conserved hydrophobic amino acids in DUF are predicted to form three α-helices. Multiple sequence alignment of DUF domains from CPn0572 and its chlamydial orthologs (C. pneumoniae Cpn0572, C. trachomatis Tarp, C. muridarum TC0741 and C. caviae CCA00170) as shown by MultAlin. The first and last amino acids of each DUF are indicated on the left. The number of each amino acid residue within the CPn0572 sequence is shown by the subscripts. Conserved amino acids are boxed in green. Hydrophobic amino acids are shown in red. Aligned residues are shown and the numbers of intervening unaligned residues in each sequence are given in parentheses. (D) Schematic representation of the disposition of the residues organized into α-helices (as indicated by the lower case letters) as depicted by the helical wheel projection prediction program. The amino acids at positions a and d (in red) represent the hydrophobic face of the wheel as shown on the right. arrowhead in Z: 16). In contrast, in more than 80% of transfected cells, most GFP-DUF 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-DUF 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-DUF 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 DUFindependent ability to generate or interact with other filamentous structures.

CPn0572 Stabilizes F-actin by Inhibiting Its Depolymerization
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 Factin 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  GFP-CPn0572 cables made up of F-actin are indicated by arrowheads, while GFP-CPn0572-or GFP-DUF-bound cables not containing F-actin fibers are marked by arrows.
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.

CPn0572 Displaces Cofilin from F-actin Structures
ADF/cofilins are key regulatory proteins that bind to and sever Factin 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 aip1 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 aip1 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. Preassembled 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 FIGURE 6 | CPn0572 and its C-terminal segment stabilize F-actin and displace cofilin from F-actin in yeast in vivo. (A) GFP-(control) or GFP-CPn0572-expressing cells were treated with Lat-A (+ Lat-A) for the indicated times, and stained for F-actin. (B) Actin cytoskeleton patterns in yeast cells expressing full-length CPn0572 or its deletion variants before and after treatment with Lat-A. Yeast cells expressing GFP (control), GFP-CPn0572, GFP-C-terminus or GFP-DUF were first grown to mid-log phase in inducing medium, exposed to DMSO alone (-Lat-A) or Lat-A (final concentration 2.5 µM) in DMSO (+ Lat-A) for 1 h, then fixed and stained with rhodamine-phalloidin to visualize F-actin, and examined using an Axiovert 200 microscope (Carl Zeiss). (C) In aip1 yeast cells carrying the empty vector (control), cofilin (green) localizes to actin patches and cables (marked by a and b, respectively). In cells expressing CPn0572 6His (red), cofilin is diffusely distributed (arrowheads), while non-expressing cells retain distinct cofilin patches and cables (arrows). See also Figure S4. Binding of GST-CPn0572 and cofilin to pre-assembled F-actin was analyzed using a co-sedimentation assay. Supernatant (s) and pellet (p) fractions were analyzed by SDS-PAGE and stained with Coomassie blue. The presence (+) or absence (-) of each protein in the reaction mix is indicated. The corresponding protein band is marked. (B) The experiment was performed as described in (A), using 2 µM GST, GST-DUF (DUF) or GST-C-terminus (C-term.) or 4 µM GST-DUF (2x DUF). Supernatant (s) and pellet (p) fractions were analyzed by SDS-PAGE and stained with Coomassie blue. The presence or absence of each protein is indicated. The corresponding protein band is marked. A full-scale image of the original blot is presented in Figure S5.
into solution in vitro (Figure 7B). These results are consistent with our finding that DUF displays less colocalization with Factin, 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, GFPexpressing 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-DUF fusion with CD resulted in the complete loss of Factin structures. Moreover, no colocalization of the remaining cortical actin with GFP-DUF 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.

DISCUSSION
The actin cytoskeleton is regarded as an important target for reorganization during Chlamydia uptake. Previous studies indicated that the C. trachomatis Tarp protein binds G-actin and nucleates its polymerization in vitro (Clifton et al., 2004;Jewett et al., 2006). However, previous work had also suggested that the C. pneumoniae Tarp ortholog might be functionally distinct in some respects from other Tarp proteins (see Introduction). In this work, we show that the C. pneumoniae ortholog of Tarp, CPn0572 (previously classified as "hypothetical"), is expressed in EBs ( Figure 1A) and secreted into HEp-2 cells early during infection ( Figure 1B). During infection secretion of CPn0572 is detectable as early as 15 min after EB attachment. Thus, CPn0572 and Tarp are both translocated into the host cell, and are available at the site of bacteria entry. The fact that we could not observe CPn0572 signals at time point zero ( Figure 1B) is consistent with the results obtained previously for Tarp (Clifton et al., 2004). Very likely, the antibody does not reach the bacterial cytoplasm during immunostaining.
After infection, secretion of CPn0572 continues for at least the first 60 min p.i. and the protein becomes associated with actin patches within 15 min after EB attachment. The in vitro data indicate that CPn0572 binds directly to Factin ( Figure 7A). This corroborates and extends biochemical data which demonstrated that CPn0572 is capable of binding G-actin and nucleating F-actin formation in vitro (Jewett et al., 2010). Thus, CPn0572 and Tarp share the same actin nucleation activity in vitro and show the same actin-recruiting activity early in infection in vivo; both functions are most probably essential for entry of the respective chlamydial species into their target cells (Clifton et al., 2004;Jewett et al., 2010).
Expression of CPn0572 in yeast has a very detrimental effect on growth, and converts the typical yeast actin cytoskeleton into a large aggregate very like that induced by the Tarp expressed by C. trachomatis serovar L2 (Sisko et al., 2006). The growth defect indeed results from the disruption of the actin cytoskeleton, since removal of the actin-binding domain DUF (aa 478 to aa 536) from CPn0572 restores both wild-type actin organization and growth rate. Conversely, CPn0572 binds actin in vitro via its DUF domain ( Figure 7B). Moreover, it may well oligomerize via a proline-rich domain within its N-terminal segment (aa 349 to aa 371) as shown for the prototypic C. trachomatis Tarp (Jewett et al., 2006). Indeed, expression of a CPn0572 truncation mutant that lacks only the segment C-terminal to the DUF domain in yeast induces an actin phenotype almost identical to that induced by the full-length protein ( Figure 4B). In contrast, the CPn0572 Cterminal segment including DUF does not form actin aggregates, but colocalizes with extra-long and thickened actin cables in yeast and also reduces growth rate, suggesting that the C-terminus also harbors actin-modulating capacities. Recently, two F-actinbinding domains (FAB1 and FAB2) have been described in the C-terminus of the C. trachomatis Tarp, the latter overlapping with a vinculin-binding site (VBS3) in the C. caviae GPIC Tarp (Jiwani et al., 2013;Thwaites et al., 2015). It is worth noting that FAB and VBS-like sequences have been found in Tarp orthologs from various chlamydial species, but not in CPn0572, which once again points to functional differences between Tarp proteins and CPn0572 (Jiwani et al., 2013;Thwaites et al., 2015). Interestingly, when the CPn0572 DUF domain alone (aa 478 to aa 536) is fused to GFP, the fusion protein colocalizes with actin patches and distinctly thickened and extended actin cables, indicating that the 59-aa DUF domain is sufficient for binding and bundling of Factin. This is compatible with the in vitro actin-binding activity previously reported for a longer actin-binding segment (aa 440 to aa 540) of the protein (Jewett et al., 2010). In fact, the action of the CPn0572 DUF domain on F-actin in yeast is reminiscent of the actin phenotype induced by the Salmonella F-actin stabilizing protein SipA (Lesser and Miller, 2001).
As in yeast, ectopic expression of either CPn0572 or C. trachomatis Tarp in human cells induces an actin aggregation phenotype; however, only the former shows a continuous colocalization with distinct actin fibers ( Figure 2B). Not only that, Cpn0572 was often seen to form filamentous structures that are apparently devoid of actin (e.g., Figure 5). The nature of these CPn0572-positive fibers and their relationship to other cellular structures are unknown at present.
Interestingly, the CPn0572-induced actin structures found upon expression of the protein in yeast and human cells are resistant to the action of the actin-depolymerizing drugs Lat-A and CD (Figures 6, 8), which suggests that CPn0572 might stabilize F-actin against depolymerization. This phenotype is comparable to those described previously for yeast cells exposed to the actin-nucleating and stabilizing toxin jasplakinolide (Jpk), which inhibits actin depolymerization both in vitro and in vivo (Bubb et al., 1994;Ayscough, 2000;Vallotton et al., 2004;Lázaro-Diéguez et al., 2008). Jpk-induced F-actin fibers are resistant to the actin-depolymerizing drug Lat-A in yeast and to Lat-B in human cells (Ayscough, 2000;Lázaro-Diéguez et al., 2008).
Cofilin plays a central role in promoting actin turnover by severing/depolymerizing F-actin in all eukaryotic cells, and we therefore tested its potential role in CPn0572induced actin stabilization (Elam et al., 2013). Indeed, upon CPn0572 expression in yeast, cofilin is excluded from the CPn0572-induced actin aggregates in vivo ( Figure 6C). Thus, CPn0572 likely stabilizes F-actin via a mechanism that involves displacement of cofilin from F-actin. That Cpn0572 is indeed capable of displacing cofilin was confirmed biochemically using purified mammalian F-actin, recombinant human cofilin and recombinant CPn0572 in vitro ( Figure 7A). Remarkably, the DUF domain suffices to displace cofilin from pre-assembled Factin ( Figure 7B). Interestingly, α-helix 3 of cofilin binds in the hydrophobic cleft located between subdomains 1 and 3 of actin (reviewed in Dominguez, 2004), and the α-helical structure found in DUF is compatible with the hypothesis that CPn0572 and cofilin compete for binding to the same cleft in actin. The ability of CPn0572 and Tarp to render actin aggregates and filaments resistant to CD in transfected HEK293T cells provides evidence that both proteins can stabilize F-actin not only against the action of cofilin but against other, perhaps even all, actindepolymerizing and severing factors.
F-actin recruitment and stabilization by CPn0572 is very likely integrated with other entry-related processes. Adhesion of C. pneumoniae EBs to human cells occurs via the Pmp proteins, which bind and activate the EGF receptor, and signaling by the EGFR via Erk1/2 and the LIM kinase may lead to phosphorylation of cofilin, thereby inhibiting its actin-binding, filament-severing and depolymerizing activities, all of which may contribute to cytoskeletal changes during endocytosis of EBs.
ADF/cofilin is the central F-actin severing factor in human cells, and thus it is not surprising that modulation of its activity is also exploited for host-cell invasion by other intracellular bacteria and by viruses (Zheng et al., 2016). For example, Listeria activates the LIM kinase, which disables cofilin via phosphorylation by an unknown factor, thus preventing excessive depolymerization of the F-actin network upon entry of Listeria (Bierne et al., 2001). In Salmonella, SipA stabilizes F-actin by inhibiting the severing and depolymerizing activities of ADF/cofilin and gelsolin (McGhie et al., 2004), while a second effector, SipC, induces actin polymerization (Hayward and Koronakis, 1999). SipA appears to augment the activity of SipC (McGhie et al., 2004). Thus, F-actin nucleation and F-actin stabilization by displacement of cofilin are induced by two different Salmonella effectors, while in Chlamydiae the two functions have been coopted into one protein. Actin assembly requires a flux of G-actin, which is normally provided by active actin depolymerization elsewhere in the cell. Cofilin displacement by CPn0572 should therefore enhance depolymerization of F-actin structures not associated with CPn0572, thus increasing the flux of G-actin required for actin nucleation by CPn0572. F-actin stabilization by cofilin displacement and the nucleation/polymerization activity of CPn0572/Tarp together would therefore be expected to significantly promote formation of highly polymerized and stable F-actin at sites of EB entry. That both functions reside in the CPn0572 protein is unprecedented and may represent an example of the concentration of functional units during genome reduction (Nunes and Gomes, 2014).
In conclusion, our data reveal that the important human pathogen C. pneumoniae has evolved an essential effector protein, which (i) re-models the actin cytoskeleton, and (ii) stabilizes F-actin by excluding cofilin. Further studies are required to fully decipher the CPn0572-mediated crosstalk between C. pneumoniae and actin dynamics.

AUTHOR CONTRIBUTIONS
RZ and JHH designed the experiments. RZ conducted the experiments and collected the data. CB performed additional transfection experiments of human cells and did Westernblot analyses of human HEK293 cells expressing GFP-CPn0572 fusion proteins. RZ prepared all figures except Figure S3B which was prepared by CB. RZ and JHH analyzed the data and wrote the manuscript.