HeLa cells and HEK-293 cells (ATCC) were cultured in Dulbecco's modified Eagle's medium with Glutamax (DMEM, Invitrogen), supplemented with 10% (v/v) fetal bovine serum (FBS). C. trachomatis LGV serovar L2 strain 434 (ATCC), or GFP-expressing L2 (^GFPCtrL2) (Vromman et al., 2014) were purified on density gradients as previously described (Scidmore, 2005).

The Y2H screens were performed by Hybrigenics (Paris, France). Two baits were used corresponding to the DUF582 domain of CT619 (CT619_Cter, amino acids 480–877) and to the N-terminal domain (amino-acids 1–490). For each screen, a minimum of 7.10⁷ interactions between the bait and a human placenta library were tested.

Interactions were further tested using the Matchmaker kit (Clontech) following the manufacturer's guidelines. The DUF582 protein constructs were cloned into the yeast vector pGBKT7 carrying the GAL4 DNA binding domain while the Hrs constructs were cloned into the pGADT7 vector carrying the GAL4 activation domain. The primers used to make these constructs are listed in the Table S1. The yeast strain AH109 was transformed with the two vectors simultaneously and plated on double dropout medium (DDO; SD/-Leu/-Trp) for 48 h at 30°C. Single colonies were then cultured over-night at 30°C in a YPD medium and from this culture a serial dilution of the same number of yeast was plated on selective media DDO and QDO (SD/-Ade/-His/-Leu/-Trp). Results were analyzed 48 h later.

Cells were transfected with the indicated plasmids 24 h after seeding using JetPrime transfection kit (Polyplus transfection) following the manufacturer's instructions. For gene silencing, 10 nM (final concentration) of siRNA (Dharmacon) was mixed with Opti-MEM (Invitrogen) and Lipofectamine RNAiMAX reagents (Invitrogen) and added to 80,000 HeLa cells in suspension in complete medium. The cells were subsequently seeded in 24-well plates. Transfection was performed twice, 48 and 4 h prior to infection. siRNA efficiency was determined by immunoblot or RT-PCR (see respective sections for details). The siRNA sequences used were the following: control UGGUUUACAUGU CGACUAA, Hrs (1) GAACCCACACGUCGCCUUG, Hrs (2) GAGGUAAAC GUCCGUAACA, Tsg101 (1) CCAGUCUUCUCUCGUCCUA, Tsg101 (2) GAA GUAGCCGAGGUUGAUA, Chmp4B AGAAAGAAGAGGAGGACGA (Guizetti et al., 2011).

The genes coding the five DUF582 proteins (CT619, CT620, CT621, CT711, CT712) were amplified from C. trachomatis D/UW-3/CX genomic DNA by PCR with Phusion high-fidelity DNA polymerase (Finnzyme) according to the manufacturer's instructions and cloned into a pEGFP-derived destination vector providing a N-terminal green fluorescent protein (GFP) tag using the Gateway technology. The primers used to make these constructs are listed in the Table S1. Other constructs were generously provided by F. Gesbert (Université Paris Sud, Villejuif, France, myc-Hrs), E.O Freed (NIH Bethesda, USA, HA-Tsg101 constructs; Goila-Gaur et al., 2003), J. Martin-Serrano (King's College London School of Medicine, London, UK, VPS4 constructs).

The full length gene of ct712 and the region of ct619 truncated of the first 81 and last 625 codons were amplified by PCR, and were cloned using the Gateway system into the pDEST15 (Invitrogen) destination vector, providing a GST tag at the N-terminus. Expression of the recombinant proteins was made after transformation of BL21 E. coli strain. BL21 bacteria transformed with the GST-Δ81CT619Δ625 construct were cultured in LB media supplemented with ampicillin at 37°C until the optical density at 600 nm reached 0.6 before addition of isopropyl β-D-1-thiogalactopyranoside (IPTG) for expression induction. The GST-CT712 construct was obtained from cultures in micro-fermentors to overcome low yields. Protein purification was performed on column using glutathione-sepharose beads (GE Healthcare) following the manufacturer indications. Purified proteins were used to immunize New Zealand White rabbits for production of polyclonal antisera (AgroBio, La Ferté Saint-Aubin, France). To test these antibodies by western blot, cell lysates or purified EBs were lysed in 8 M urea, 1% SDS (v/v), 150 mM NaCl, 30 mM Tris pH 8.0 and the proteins were analyzed by SDS-PAGE as described below.

For immunofluorescence (IF), HeLa cells fixed for 30 min at room temperature in paraformaldehyde (PFA) 4% (w/v) in PBS were permeabilized for 15 min in PBS supplemented with 1 mg/ml bovine serum albumin (BSA) and 0.05% (w/v) saponin (IF buffer). For anti-Hrs staining, cells were fixed for 30 min on ice in PFA 4%. For CT813 staining, cells were fixed 30 min at room temperature in PFA 2%. Coverslips were incubated with primary antibodies in IF buffer for 1 h at room temperature, followed with three washes in the same buffer. Primary antibodies used for IF were: mouse Anti-Hrs (clone A-5, Enzo Life Science), rabbit anti-Myc (ab9106, Abcam), rabbit anti-CT529 (Gehre et al., 2016), mouse anti-CT813 (kind gift from G. Zhong, University of Texas), rabbit anti-Gys1 (04-357, Millipore). Secondary antibodies coupled with a fluorophore were incubated for 1 h together with 0.5 μg/ml Hoechst 33342 (Molecular Probes) in IF buffer. Coverslips were then mounted in Mowiol buffer and analyzed using an Axio observer Z1 microscope equipped with an ApoTome module (Zeiss, Germany) and a 63 × Apochromat lens. Pictures were taken with a Coolsnap HQ camera (Photometrics, Tucson, AZ) using the software Axiovision.

For western blot analyses, cells were lysed in 1% (v/v) sodium dodecyl sulfate (SDS), 8 M urea, 150 mM NaCl, and 30 mM Tris-HCl pH 8. Samples normalized to protein content were analyzed by sodium dodecyl-sulfate poly-acrylamide gel electrophoresis (SDS-PAGE). Proteins were transferred on polyvinylidene difluoride (PVDF) membranes and incubated for 1 h in 0.1% Tween-20 (v/v) in PBS supplemented with 5% (w/v) skimmed milk. Primary and secondary antibodies were both sequentially incubated in 0.1% Tween-20 in PBS 1X on the membranes for 1 h separated by washes. Primary antibodies used were: rabbit anti-Hrs (Bethyl laboratories), mouse anti-Tsg101 (clone 4A10, GeneTex), anti-GFP (sc-8334, Santa-Cruz), and anti-HA.11 (clone 16B12, Covance), mouse anti-myc (clone 9E10, Santa-Cruz). For qualitative analysis, secondary antibodies were coupled with horse-radish peroxidase and were revealed by chemiluminescence (KODAK).

Entry experiments were performed as described (Vromman et al., 2014). Briefly, HeLa cells were incubated at 4°C for 15 min in DMEM 10% FCS (v/v) before adding the ^GFPCtrL2 (multiplicity of infection (MOI) = 10) for another 30 min at 4°C. Medium was replaced by pre-warmed medium at 37°C, and plates were transferred to the 37°C incubator for the indicated times before fixation in ice-cold 4% PFA in PBS for 30 min. Extracellular bacteria were stained with a mouse anti-MOMP-LPS (Argene #11-114) antibody followed with Cy5-conjugated secondary antibodies. Pictures of fields with 5-10 cells were acquired and analyzed as described (Vromman et al., 2014).

Cells were infected with density gradient purified ^GFPCtrL2 EBs at a MOI < 0.3. Twenty-four hours later, cells were washed with PBS and gently detached using 0.5 mM EDTA in PBS. Samples were fixed in PFA 2% in PBS and stored over-night at 4°C. Flow cytometry analysis was performed with a FACS Gallios (Beckton Coulter) using the FL-1 (detecting fluorescence emission between 505 and 545 nm), the FSC (relative cell size) and the side scatter detectors (cell granulometry or internal complexity) on 1/10 of the sample diluted in PBS. A minimum total of 10,000 gated events were collected for each sample. Data were analyzed using the Kaluza 1.2 software (Beckman Coulter).

The analysis of cells transfected with myc-tagged VPS4 was performed in the following way: after fixation, the cells were centrifuged at 1500 g, washed in PBS, centrifuged again, and incubated for 1 h with rabbit anti-myc antibodies in 1 mg/ml BSA, 0.05% saponin in PBS. The cells were then washed and incubated for 1 h in the same buffer with anti-rabbit antibodies conjugated to Cy5. After another washing step the cells were resuspended in PBS and analyzed by flow cytometry in the FL-1 (green) and FL-4 (far-red) channels.

To measure progeny, cells were washed in PBS 24 h post-infection (hpi), detached and lysed with 1 mm glass beads. Fresh HeLa cells were inoculated with serial dilutions of the cell lysates and 24 h later flow cytometry was used to determine the infection rate and deduce inclusion forming units (ifu) in the inoculum.

HEK-293 cells were lysed on ice for 45 min in ice cold 150 mM NaCl, 10 mM NaF, 1 mM EDTA, 1 mM EGTA, 0.5% Triton X-100 (v/v), 50 mM Tris pH 7.5 (IP buffer) supplemented with 5 mM PMSF, 1 mM vanadate and 1:100 of proteases inhibitor cocktail (P8340, Sigma). Cells were removed with a scraper, homogenized and centrifuged at 16000 g for 20 min at 4°C. For anti-myc IP, the supernatant was incubated at 4°C for 1–2 h with anti-myc 9E10 antibody (Santa-Cruz) followed by a 1.5 h incubation with protein G coupled to sepharose-beads. For anti-HA IP, the supernatant was incubated for 2 h with the anti-HA antibody coupled to agarose beads (Sigma) at 4 C. Immunoprecipitates were then washed at 4°C for a minimum of 5 times with 1 mL of the IP buffer with gentle centrifugations of maximum 400 g. Beads were then resuspended in a Laemmli buffer supplemented with 1% ß-mercapto-ethanol (v/v).

Total RNA was isolated with the RNeasy Mini Kit with DNase treatment (Qiagen) according to the manufacturer's protocol at the indicated times after transfection with control or Chmp4B siRNA. RNA concentrations were determined by NanoDrop and the samples normalized to an equal RNA content. Reverse transcription (RT) was performed with SuperScript III Reverse Transcriptase (Life Technologies) and cDNA was used as a template for amplification of Chmp4B with PrimeStar (Clontech) using GAGTTCCTGGAGAAGAAAATCG and CCTCTTCTTACGCTTCAGTGC as primers. Equal volumes were loaded on agarose gels and bands were revealed using UV-light visualizing ethidium bromide.

A two-hybrid screen in yeast using the DUF582 domain of CT619 as bait identified 10 putative partners, with moderate to high scores of confidence. The best score was assigned to only two proteins: dynamin 2 and Hrs. Attempts to detect an interaction between CT619 and dynamin 2 by co-immunoprecipitation failed, and this direction was not pursued any further. The DUF582 domain is predicted to be mainly α-helical and to contain a segment adopting a coiled-coil conformation. It is only moderately conserved, with an amino acid identity ranging from 18% (CT711–CT712) to 39% (CT620-CT621) (Figure 1B and Muschiol et al., 2011). We reasoned that if, in spite of this low conservation, we could detect an interaction between Hrs and the DUF582 domain of other members of the family, this would considerably strengthen the hypothesis that Hrs is a true interactor of the domain. To test this, we fused the DUF582 domains of CT619, CT621, CT711, and CT712 of C. trachomatis L2 to the Gal4 binding domain coded by the pGBKT7 vector (CT620 was not tested in this assay). Full-length Hrs, or its minimal interacting domain found in the screen using the DUF582 of CT619 as bait, Δ368HrsΔ138, were fused to the Gal4 activation domain coded by the vector pGADT7 (Figure 1A). Yeast were co-transformed with each plasmid and plated on selective medium for isolation of co-transformed cells. Dilutions of the same quantity of co-transformed yeast were plated on high stringency medium (quadruple dropout) to reveal positive interactions (Figure 1C). All of the DUF582 domains tested, except that of CT621, showed an interaction with Hrs in this experimental system. For all cases where an interaction was observed, it appeared to be stronger when the shorter form of Hrs, Δ368HrsΔ138, rather than the full-length form, was tested.

To delimit further the interaction between CT619 and Hrs, we tested truncated versions of the DUF582 domain of CT619. We generated the construct CT619_CterΔ89, in which the last 89 amino acids, that correspond to the most conserved region in the DUF582 (Muschiol et al., 2011), were deleted. We also designed a construct truncated for the last 191 amino acids (CT619_Cter Δ191), which no longer contains the coiled coil region common to all DUF582 (Muschiol et al., 2011). Surprisingly, none of these truncations had an impact on the interaction with Hrs (Figure 1D). Thus, we concluded from these experiments that the ability to interact with Hrs is conserved between several DUF582 domains and does not involve the C-terminal part of this domain.

To document this interaction with another technique we investigated the localization of GFP-tagged DUF582 proteins, in conditions where Hrs was overexpressed or not. When expressed alone, the GFP-tagged DUF582 proteins were uniformly distributed throughout the cytoplasm. When myc-Hrs was co-expressed, in the majority of the cells the distribution of the GFP-tagged proteins was no longer uniform, but accumulated in structures of various sizes (Figure 2A). The change in the distribution of the DUF582 proteins when Hrs was overexpressed, and their partial co-localization, support the hypothesis that the proteins interact. We tested this by co-immunoprecipitation experiments in HEK-293T cells overexpressing myc-Hrs together with various GFP-tagged proteins (Figure 2B). In these experiments a GFP-tagged fusion with an irrelevant chlamydial protein of similar size, GFP-CT671, was used as a negative control. No interaction was detected between this protein and myc-Hrs. In contrast, all the DUF582 proteins co-immunoprecipitated with myc-Hrs, including GFP-CT620, which had not been tested in the two-hybrid assay and colocalized with myc-Hrs. When only the DUF582 domain of CT619 was used, it was expressed at a lower level than the full-length protein and co-immunoprecipitated with myc-Hrs relatively better. Thus, it is possible that the DUF582 domain is somehow less accessible for Hrs binding when expressed in the full-length protein. Interestingly, we also detected an interaction between Hrs and CT621 or its DUF582 domain alone. It is consistent with the relocation of GFP-CT621 upon myc-Hrs co-expression reported above and indicates that, in spite of the negative result obtained in the two-hybrid screen, the DUF582 of CT621 is also able to bind the ESCRT-0 protein.

A second two-hybrid screen in yeast, this time using the N-terminal domain of CT619 as bait, generated 20 hits with moderate to high confidence scores. Eleven proteins came out with the highest score: Bin3, Dact2, Dynamin-2, HMG20B, Keratins 8 and 19, PMC1, PHF16, RCN1, SPTB1, and Tsg101. The latter is a well-studied component of the ESCRT-I complex, and binds to Hrs. From the different clones obtained, that showed an interaction with the N-terminal domain of CT619, we inferred that the domain of interaction was comprised within the amino acids 168 and 282 of Tsg101, thus including parts of the proline-rich region and the coiled-coil domain (Figure 3A).

To verify this interaction, we performed immunoprecipitation experiments using truncated HA-tagged Tsg101 constructs (Figure 3A). HEK-293T cells were transfected with HA-tagged Tsg101 constructs alone or together with GFP-tagged CT619. Full-length HA-tagged Tsg101 was expressed at a very low level and was not included in these experiments. Immunoprecipitation was performed with anti-HA antibodies coupled to agarose beads. Inputs and immunoprecipitated fractions were analyzed by western blot with anti-HA and anti-GFP antibodies. CT619 co-immunoprecipitated with the two constructs of Tsg101 that contained the coiled-coil domain (Figure 3B). Using constructs that expressed separately the amino-terminal domain of CT619 (GFP-CT619_Nter) or its DUF582 domain (GFP-CT619_Cter), we confirmed that the interaction between Tsg101 and CT619 implicates the N-terminal domain of CT619. In contrast, CT621, whose N-terminal domain shows no similarity with the one of CT619 (Muschiol et al., 2011), does not interact with the Tsg101 construct containing the coiled-coil domain, nor with the shorter construct that was used as a negative control (Figure 3C).

Altogether, these data demonstrate specific interactions of the ESCRT-I protein Tsg101 and of the ESCRT-0 protein Hrs with the N-terminal and C-terminal domains of CT619, respectively, strengthening the hypothesis that CT619 targets an ESCRT-driven pathway.

To gain further insight into the function of the DUF582 proteins we completed our toolbox by generating polyclonal antibodies against CT619 and CT712 (Figure 4A). Like CT620 and CT711 (Muschiol et al., 2011), CT619 runs in denaturing gels as a double band, with the higher, minor, band corresponding to the expected molecular weight, and a 10 kD-shorter, major, species. CT712 migrates as a single species of expected size. CT619 and CT712 accumulated late in infected cells and were abundant in EBs (Figure 4B), a pattern of expression similar to that of CT711 and CT620. In contrast, CT621 was detected earlier (16 hpi) and was less abundant in EBs. Several type 3 secretion effectors are enriched in EBs (Saka et al., 2011), that are likely secreted in the early stages of infection, to facilitate entry and secure the nascent inclusion. By immunofluorescence, we did not detect any of the DUF582 proteins at an early stage of infection, possibly because the amount of effector secreted was too low for detection by this method (data not shown). At later time points we detected CT620 and CT621 in the host cytoplasm of some of the infected cells (Muschiol et al., 2011), but not the other DUF582 proteins (data not shown). Again, we cannot rule out the possibility that their secretion is below the sensitivity of this technique. Having shown that CT619 interacts in vitro with Hrs and Tsg101 we asked whether these proteins of the ESCRT machinery were required for the internalization step. Each of the two proteins was depleted for 24 h using two different siRNAs. Immunoblotting with specific antibodies revealed a depletion of at least 85% for each protein. Twenty-four h after depletion cells were infected with GFP-expressing C. trachomatis L2 (^GFPCtrL2, MOI = 10), and fixed 1 h later. Extracellular bacteria were stained using anti-MOMP antibody and internalized bacteria were quantified using a semi-automated image analysis software (Vromman et al., 2014). No difference in the efficiency of entry was observed in the cells depleted for Hrs or Tsg101 compared to control cells (Figure 5A).

By immunofluorescence we observed that Hrs was often present in the vicinity of the inclusion membrane, although it did not appear to be enriched at this location (Figure S1). It might transiently associate with the inclusion membrane and contribute to its exchanges with the host cytoplasm. We thus tested whether impairing ESCRT-driven pathways would impact bacterial growth. Control cells and cells depleted for Hrs or Tsg101 were infected with ^GFPCtr L2 at MOI = 1. Twenty h later cells were fixed and analyzed by flow cytometry (Vromman et al., 2014). Similar infection rates were observed in all samples, confirming that depleting Hrs or Tsg101 has no impact on bacterial entry (Figure 5B). In addition, GFP, a read-out of bacterial growth (Vromman et al., 2014), accumulated to the same extent in all conditions, indicating that Hrs and Tsg101 are not required for bacterial growth in vitro (Figure 5B).

Hrs and Tsg101 are part of the ESCRT-0 and -I machineries, respectively, which sort cargo to the finale ESCRT-III driven step. Alternative adaptors appear to work upstream of ESCRT-III (Bissig and Gruenberg, 2014). Thus, to directly test if a functional ESCRT-III machinery was required for chlamydial growth we used a siRNA against Chmp4B, an essential component of the ESCRT-III complex (Hurley, 2010). The efficiency of the siRNA was verified by RT-PCR 24, 48, and 72 h post-transfection (Figure 6A). In addition, we examined the distribution of the epidermal growth factor (EGF) receptor using fluorescent EGF. Upon internalization with its ligand, the EGF receptor is processed by the ESCR-0, -I, -II, and -III machineries for finale sorting to multiversicular bodies (Henne et al., 2011). Depletion of ESCRT components alter the morphology of these compartments (Razi and Futter, 2006). Cells depleted for Chmp4 were loaded for 15 min with 100 ng/ml rhodamine-coupled EGF (EGF^Rho), washed and incubated further at 37°C for 45 min before fixation. After treatment with siRNA against Chmp4B EGF^Rho positive compartments were fewer and more dispersed than in cell treated with control siRNA, confirming that this siRNA is efficient at disturbing Chmp4 mediated pathways (Figure S2). Cells were infected with ^GFPCtrL2 at MOI = 1 24 h after the transfection with siRNA, and fixed and analyzed 24 h later by flow cytometry. Cells depleted for Chmp4B showed similar infection rates and bacterial load as control cells (Figure 6A), indicating that Chmp4B is not required for bacterial growth. Finally, to rule out the possibility that the levels of depletion obtained with the siRNA against Hrs, Tsg101, or Chmp4B, although clearly significant (see Figure S2 for functional validation of the siRNA against Hrs and Tsg101), were not sufficient to completely block ESCRT-mediated processes, we used a dominant negative form of VPS4, which prevents the disassembly of the ESCRT machinery at the end of the process, and is considered as the most potent method to inhibit ESCRT-mediated steps. Indeed the traffic of EGF^Rho was deeply affected by the expression of a dominant negative mutant of VPS4 (mutation K173 to Q) (Figure S2). HeLa cells were transfected with myc-tagged VPS4 constructs, wild type (WT) or dominant negative (DN), and infected 4 h later (to avoid excessive cell death upon expression of VPS4-DN). Cells were fixed 24 hpi and analyzed by flow cytometry. If anything, a slight increase in the percentage of infection was observed in the cells expressing myc-VPS4-DN compared to cells expressing the wild-type form. Bacterial loads were similar in both conditions (Figure 6B). This experiment confirms that functional ESCRT is not required for chlamydial growth.

Finally, we asked if impairing ESCRT-0, -I, or -III had an effect on the infectious progeny. We measured it on progeny collected 24 hpi, since prolonged (>72 h) depletion of the ESCRT proteins affected cellular viability in some of the experiments. Depletion of Hrs, Tsg101 or Chmp4B had no effect on the infectious progeny (Figure 6C), indicating that the whole infectious cycle proceeds normally in cells deficient for ESCRT functions.

We have recently shown that host glycogen translocates into the inclusion lumen, where it contributes to glycogen storage within the inclusion lumen. The host glycogen synthase Gys1 was also observed inside the inclusion lumen. This enzyme is known to associate with glycogen particles (Stapleton et al., 2010). When cytoplasmic glycogen levels were reduced by depriving the cells of glucose Gys1 translocation was also strongly reduced, supporting the hypothesis that Gys1 makes its way to the inclusion lumen bound to cytoplasmic glycogen (Gehre et al., 2016). The mechanism by which host glycogen/Gys1 reaches the inclusion lumen is unknown. By electron microscopy we observed glycogen-filled vesicles in the inclusion lumen, supporting a transport through the inward budding of vesicles from the inclusion membrane. Since this would fit with the topology of inward vesicle budding from the periphery of MVBs we hypothesized that the ESCRT machinery might be involved. To test this we silenced Hrs, Tsg101, or Chmp4B expression for 2 days before infecting the cells with C. trachomatis L2. Cells were fixed 24 h later, permeabilized, and the inclusion membrane and Gys1 were labeled with mouse anti-CT813 and rabbit anti-Gys1 antibodies, respectively. Strong reduction in Hrs, Tsg101, or Chmp4B levels did not impair Gys1 translocation into the inclusion lumen when compared to control cells (Figure 7), implying that the ESCRT-0, -I, and -III machineries are dispensable for this process.

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.