The chlamydial deubiquitinase Cdu1 supports recruitment of Golgi vesicles to the inclusion

Chlamydia trachomatis is the main cause of sexually transmitted diseases worldwide. As obligate intracellular bacteria Chlamydia replicate in a membrane bound vacuole called inclusion and acquire nutrients for growth and replication from their host cells. However, like all intracellular bacteria, Chlamydia have to prevent eradication by the host's cell autonomous system. The chlamydial deubiquitinase Cdu1 is secreted into the inclusion membrane, facing the host cell cytosol where it deubiquitinates cellular proteins. Here we show that inactivation of Cdu1 causes a growth defect of C. trachomatis in primary cells. Moreover, ubiquitin and several autophagy receptors are recruited to the inclusion membrane of Cdu1‐deficient Chlamydia. Interestingly, the growth defect of cdu1 mutants is not rescued when autophagy is prevented. We find reduced recruitment of Golgi vesicles to the inclusion of Cdu1 mutants indicating that vesicular trafficking is altered in bacteria without active deubiquitinase (DUB). Our work elucidates an important role of Cdu1 in the functional preservation of the chlamydial inclusion surface.


| INTRODUCTION
Chlamydia trachomatis is an obligate intracellular human pathogen and the most common transmitted sexually disease worldwide (Newman et al., 2015). Infection with the pathogen can manifest in a spectrum of diseases like ectopic pregnancy, infertility but also in the blinding eye disease trachoma (Bartlett, Levison, & Munday, 2013;Bébéar & de Barbeyrac, 2009;Wright, Turner, & Taylor, 2008). Chlamydia replicate in a biphasic developmental cycle. Chlamydial elementary bodies (EBs) infect epithelial cells via the endocytotic pathway and transform into replicative reticulate bodies (RBs) within a vacuole, the chlamydial inclusion (Engel, 2004). After several rounds of replication inside the inclusion, RBs complete their life cycle and transform back into infectious EBs which are released from the host cell via extrusion or cell lysis (Hybiske & Stephens, 2007). Chlamydia secrete several effectors into the host cell by the type III secretion system (Subtil et al., 2005;Valdivia, 2008) to orchestrate their developmental cycle, receive host cell metabolites and prevent damage or premature death of their host cell. The natural ribbon-like Golgi architecture is destroyed in cells infected with C. trachomatis and vesicular mini stacks originating from fragmented Golgi closely associate with the chlamydial inclusion . Host cell sphingolipids are required for the intracellular growth of C. trachomatis in epithelial cells, with sphingomyelin attained via the intersection of the chlamydial inclusion with these Golgi-derived vesicles (Hackstadt, Rockey, Heinzen, & Scidmore, 1996).
As all pathogenic bacteria, Chlamydia has to subvert the host's immune system to establish a successful infection and both, the adaptive and the innate immune response are effective to control chlamydial infections (Fischer & Rudel, 2016). The cell autonomous immune defence present in all human cells is particularly relevant for an obligate Daniela Auer and Sophie D. Hügelschäffer contributed equally to this work. intracellular pathogen like Chlamydia (Fischer & Rudel, 2016). A very efficient arm of cell autonomous defence is the removal of pathogens via autophagy. Foreign particles like intracellular bacteria in the cytosol or in vacuoles are specifically marked by ubiquitination. In eukaryotic cells poly-ubiquitination of proteins via lysine 63 (K63) is mainly involved in autophagic events like mitophagy or the removal of Salmonella from the cytoplasma via xenophagy, a form of autophagy specifically targeting bacteria (Geisler et al., 2010;Van Wijk et al., 2012).
K48 poly-ubiquitination of proteins often serves as a signal for proteasomal degradation (Hershko & Ciechanover, 1998;Vucic, Dixit, & Wertz, 2011). Ubiquitination of any kind happens via a cascade of several proteins (Kleiger & Mayor, 2014). This posttranslational modification is tightly regulated but also reversible by removal of the ubiquitin by deubiquitinating enzymes (DUBs).
Ubiquitination of the bacterial membrane or the surrounding vacuole membrane then recruits autophagy adapter proteins to the ubiquitin molecules (Kirkin, Lamark, Johansen, & Dikic, 2009;Svenning & Johansen, 2013). As a consequence, ubiquitin bound adapters recruit autophagy-related gene 8 (ATG8) family proteins like the light chain 3 (LC3) protein which mediate the formation of the autophagosome around the intruder. Finally, autophagosomes fuse with lysosomes resulting in the destruction of the engulfed bacteria in the formed autolysosomes (Dikic, 2017).
However, bacteria have evolved strategies to interfere with the autophagy system in order to stay alive and proliferate. In Salmonella infection one third of all bacteria are targeted for lysosomal degradation through xenophagy (Birmingham, Smith, Bakowski, Yoshimori, & Brumell, 2006). This affects free Salmonella in the host cytosol as well as bacteria engulfed in the Salmonella containing vacuole (SCV). The SCV membrane often gets punctured which leads to the recruitment of the autophagy system (Thurston, Wandel, von Muhlinen, Foeglein, & Randow, 2012). In order to prevent clearance Salmonella has evolved mechanisms to repair the damaged SCV (Kreibich et al., 2015). Moreover, Salmonella secrete SopA into the cytosol, a HECT E3 ligase with yet unknown targets (Zhang, Higashide, McCormick, Chen, & Zhou, 2006). Also Legionella, E. coli and Shigella are known to secrete effector proteins into the host cell that interfere with the ubiquitination system, often by mimicking E3 ligases (Huibregtse & Rohde, 2014). Chlamydia has also established mechanisms to evade autophagosomal degradation. Early studies revealed that C. trachomatis inclusions do not fuse with autophagosomes in human epithelial cells (Al-Younes, Brinkmann, & Meyer, 2004). The adaptation of Chlamydia species to the host specific autophagy system is a fascinating example of co-evolution. The mouse pathogen C. muridarum was able to overcome autophagic defence in mouse cells, whereas the human pathogen C. trachomatis was susceptible (Yasir, Pachikara, Bao, Pan, & Fan, 2011). In human epithelial cells, in contrast, C. muridarum inclusions attracted several components of the autophagy pathway. However, C. trachomatis inclusions were resistant to ubiquitination, which prevented the association of the autophagy adaptors p62 and nuclear dot protein 52 kDa (NDP52) and the autophagosomal marker LC3 with the inclusion (Haldar et al., 2016).
Although the chlamydial factors mediating the evasion of host autophagy remain unknown, Chlamydia appears to subvert host autophagy by at least two mechanisms: (a) keeping the inclusion intact (Weber et al., 2017b); and (b) interfering with initial ubiquitination of the inclusion. The genome of C. trachomatis encodes for two deubiquitinases ChlaDUB1 and ChlaDUB2, hereafter referred to as Cdu1 and Cdu2 (Misaghi et al., 2006). We recently showed that Cdu1 is secreted into the chlamydial inclusion facing with its active site to the host cell cytoplasm where it interacts with the host cell proteome and stabilises anti-apoptotic Mcl-1 through deubiquitination (Fischer et al., 2017).
In this paper we show that Cdu1 is essential for normal development in primary host cells. Inactivation of Cdu1 in Chlamydia resulted in the recruitment of the autophagy machinery to and phagophore formation at the chlamydial inclusion. Despite phagophore formation at the cdu1 mutant inclusion, autophagy was not effective in interfering with chlamydial development. The lack of active Cdu1 prevents Golgi vesicle recruitment and slows the transition of RB to EB resulting in decreasing progeny.

| Chlamydia require Cdu1 for growth in primary cells
We previously demonstrated that a transposon insertion mutant of cdu1 (Ctr Tn-cdu1) expressing truncated Cdu1 protein is attenuated in a mouse infection model and in cell lines challenged with interferon-γ (IFN-γ) (Fischer et al., 2017). In the absence of IFN-γ Ctr Tn-cdu1 grew without significant difference in several tumour cell lines (Fischer et al., 2017). Tumour cell lines are known to have altered cell autonomous defence including autophagy (Kimmelman, 2011). We therefore asked whether primary cells permit the development of Ctr Tn-cdu1 like previously observed for tumour cells. Primary human fallopian tube cells (Fimb) and human umbilical vein endothelial cells (HUVEC) were infected with wild-type and mutant bacteria and the relative infectivity was determined by transferring bacteria after 48 hr from these primary infections to fresh cells (here referred to as infectivity assay). Fimb and HUVEC infected with Ctr Tn-cdu1 produced less infectious chlamydial particles compared to wild-type infected cells infected cell was reduced compared to wild-type Chlamydia ( Figure 1d). Therefore, the transition of RB to EB seems to be stalled in the mutant infection. The differentiation of RB to EB is not completely understood, nevertheless, the size of RBs is discussed to be important in this process (Lee et al., 2018). We measured the average size of RBs in chlamydial inclusions after 30 hr in Fimb cells and found that RBs of wild-type Chlamydia were significantly smaller (0.5814 μm ± 0.056) than RBs of the bacteria lacking an active Cdu1 deubiquitinase (0.8236 μm ± 0.065) (Figure 1e). Thus, Chlamydia lacking active Cdu1 are developing slower, resulting in less infectious EBs compared to the wild-type bacteria.

| Inclusions of Ctr Tn-cdu1 are ubiquitinated
Recently we showed that the deubiquitinase Cdu1 is secreted into the inclusion membrane with the active site located on the host cell F I G U R E 1 Inactive Cdu1 results in reduced bacterial growth in primary cells. (a) Primary HUVEC and Fimb and HeLa cells were infected with Ctr WT (wild-type) or Ctr Tn-cdu1 at a multiplicity of infection (MOI) of 0.3 for primary cells and a MOI of 1 for HeLa cells. Lysates of primary infected cells were prepared after 24 hr. For progeny infection lysates were taken 48 hpi and used to infect fresh cells (see Experimental procedures for details). Twenty four hours later, lysates of these cells were investigated by Western blotting. Shown is the relative amount of chlamydial HSP60 (cHSP60) or OmpA. Protein loading was normalised to actin. Significance was calculated with the Student's t test. *p < .05, HeLa, HUVEC n = 3, Fimb n = 4. (b) Fimb and HeLa cells were infected with Ctr WT or Ctr Tn-cdu1 at MOI 0.3 and MOI 1 respectively. At indicated time points post infection, infected cells were lysed and the supernatant was used to infect fresh cells. Twenty four hours later, cells were lysed in sodium dodecyl sulfate (SDS)-sample buffer and chlamydial load was determined by immunoblot for OmpA. GAPDH was used as a loading control. (c) Fimb cells were infected with Ctr WT or Ctr Tn-cdu1 with a MOI of 0.3. 30 hpi cells were fixed and prepared for EM (see Experimental procedures for details). Scale bar, 1 μm. (d) Quantification of bacterial particles, reticulate bodies (RBs) and elementary bodies (EBs) in Fimb cells infected with Ctr WT or Ctr Tn-cdu1 for 24 hr using EM pictures. Infection and samples preparation for EM was done as described in (b). Bacterial particles of 22 inclusions were analysed. (e) Fimb cells were infected with Ctr WT or Ctr Tn-cdu1 MOI 0.3 for 30 hr. Cells were fixed and prepared for EM. Transmission electron microscopy pictures were taken and RB diameter from five inclusions for each sample was measured. Significance was calculated with the Student's t test. ** p < .01 F I G U R E 2 Ctr Tn-cdu1 inclusion is ubiquitinated. HeLa (a) or Fimb cells (b) were infected with Ctr WT or Ctr Tn-cdu1 for 24 hr at a MOI of 1 or 0.3 respectively. Cells were fixed with 4% paraformaldehyde (PFA)/sucrose followed by indirect immunofluorescence. Chlamydial HSP60 (pink) as well as K48-linked ubiquitin (green, left panel) or K63-linked ubiquitin (green, right panel) was detected. Scale bar, 10 μm. (c) Hela Cdu1-Flag cells were infected with Ctr WT or Ctr Tn-cdu1 at a MOI of 1 for 24 hr with/without addition of AHT (50 ng/ml) for 4 hr. Cells were fixed with 4% PFA/sucrose and indirect immunofluorescence was performed with ubiquitin (green) and Cdu1 (pink). The asterisk shows the chlamydial inclusion. Scale bar, 10 μm. (d) Quantification of (c) showing ubiquitination of Ctr WT or Ctr Tn-cdu1 inclusions in HeLa Cdu1-Flag and HeLa Cdu1(C345A)-Flag cells upon AHT treatment. Pictures shown in (c) from three independent experiments were used to count 100 inclusions per replicate with and without ubiquitin signal around the chlamydial inclusion. ND, not detected. (e) HeLa cells were infected with Ctr WT and Ctr Tn-cdu1 at a MOI of 1 for 24 hr before cells were fixed using 4% PFA/sucrose. Indirect immunofluorescence was performed against Cdu1 (green), ubiquitin (pink) and 4 0 ,6-Diamidino-2-phenylindol (DAPI) for DNA (blue). Cell contours are marked in white. Scale bar, 10 μm F I G U R E 3 Ubiquitinated inclusions of Ctr Tn-cdu1 are recognised by the host cell autophagy system. (a) HeLa cells were infected with Ctr WT or Ctr Tn-cdu1 at a MOI of 1 for 24 hr. Cells were fixed with 4% PFA/sucrose and indirect immunofluorescence microscopy was performed with K63 ubiquitin (green), p62/SQSTM1 (red) and DAPI for DNA. Scale bar, 10 μm. (b) HeLa cells were infected with the chlamydial strains Ctr WT and Ctr Tn-cdu1 at a MOI of 1 for 24 hr followed by fixation with 4% PFA/sucrose. For indirect immunofluorescence microscopy cells were stained for cHSP60 (green) and light chain 3 (LC3) (pink). Scale bar, 10 μm. (c) Immunofluorescence images of Ctr WT and Ctr Tn-cdu1 infected HeLa cells were analysed by counting inclusions positive for p62 or LC3 recruitment. 34 pictures of three independent experiments were counted. Significance was calculated by Student's t test. * p < .05, ** p < .001. (d) HeLa cells were infected with Ctr WT or Ctr Tn-cdu1 at a MOI of 1 for the indicated time points. Cells were then lysed in SDS-sample buffer and expression of p62, LC3, cHSP60 and beta-Actin was detected by immunoblot. (e) The levels of the indicated proteins from the immunoblots shown in panel (d) were quantified for the 24 hpi time point. Shown are the mean from three independent experiments ±SD. Significance was tested using the Student's t test. * p ≤ .05, ** p ≤ .01. (f) Fimb cells infected with Ctr WT or Ctr Tn-cdu1 at a MOI of 0.3 for 30 hr were fixed for EM preparation. Representative images are shown. The white arrow appoints to a double membrane forming around the chlamydial inclusion. Scale bar, 1 μm F I G U R E 4 Growth and replication defect of Ctr Tn-cdu1 in primary cells is independent of autophagy. (a) HeLa cells were infected with Ctr WT or Ctr Tn-cdu1 at a MOI of 1 for 24 hr before cells were fixed with 4% PFA/sucrose. Cells were stained for Lamp1 (green) and cHSP60 (pink). Scale bar, 10 μm. (b) Primary Fimb cells were infected with Ctr WT or Ctr Tn-cdu1 at a MOI of 0.3 for 24 hr and treated with or without 10 nM Bafilomycin A1 for 7 hr. Cells were lysed in SDS-sample buffer and immunoblot was performed. (c) Relative growth of Ctr WT and Ctr Tn-cdu1 after 24 hr in Fimb cells treated with or without Bafilomycin A1 for 7 hr. Bacterial load was determined by quantifying chlamydial OmpA levels and normalisation to GAPDH (b). Significance was calculated using the Student's t test. * p < .05. (d) Fimb cells were treated with a control siRNA (siControl) or sip62 for 24 hr before infection with Ctr WT or Ctr Tn-cdu1 at a MOI of 0.3. Lysates were taken 24 hpi and analysed by immunoblotting for p62, OmpA and actin. (e) Infectivity assay in PentaKO and Parental cells was performed. Cells were infected with Ctr WT and Ctr Tn-cdu1 with a MOI of 1 and lysates from these primary infections were taken 24 hpi (Primary). For progeny infection cells were lysed 48 hpi and fresh HeLa cells were infected with the infections supernatant. Lysates for progeny infection were taken 24 hpi. Both, primary and progeny infection were analysed for chlamydial load by immunoblotting for cHSP60. (f) Quantification of immunoblots showing relative bacterial growth in PentaKO and Parental cells. Bacterial load was determined by quantifying chlamydial HSP60 (cHSP60) levels and normalisation to actin (e). Growth of wild-type Chlamydia was set to 100%. Results of three independent experiments are shown. Significance was calculated using the Student's t test. * p < .05, ** p < .01 side (Fischer et al., 2017). We tested whether the inactive Cdu1 protein alters the ubiquitination pattern in the host cell. Indeed the inclusion of Ctr Tn-cdu1 was decorated with K48 and K63 linked ubiquitin chains in contrast to the wild-type strain that shows no ubiquitin signal (Figure 2a

| The autophagy machinery is recruited to the ubiquitinated inclusion
Ubiquitination via lysine 48 is often a signal for protein degradation by the proteasome, whereas a linkage via K63 preferably induces autophagy (Hershko & Ciechanover, 1998;Olzmann & Chin, 2008;Vucic et al., 2011). Autophagy adapters like sequestosome1 (hereafter referred to as p62) bind to ubiquitin K63 and subsequently recruit autophagy initiation proteins like LC3. Compared to inclusions of wild-type bacteria, inclusions of Ctr Tn-cdu1 contained high amounts of p62 visible as surface-associated ring structures on these inclusions (Figure 3a, c). Furthermore, p62 protein levels were upregulated in cells infected with Ctr Tn-cdu1 starting around mid-phase of the developmental cycle (Figure 3d, e). In addition to p62, the autophagy receptors next to BRCA1 gene 1 protein (NBR1) and NDP52 were also identified at the chlamydial inclusion of Ctr Tn-cdu1 infected HeLa and Fimb cells ( Figure S4 a- were identified at the inclusion of Ctr Tn-cdu1 that could compensate F I G U R E 6 Altered Golgi recruitment in Ctr Tn-cdu1 infected cells can be rescued. (a) HeLa FlpIn Cdu1Flag cells were infected with Ctr WT or Ctr Tn-cdu1 at a MOI of 1 for 20 hr before 50 ng/ml AHT was added for 4 hr. For indirect immunofluorescence cells were fixed with 4% PFA/sucrose. Chlamydia were stained for cHSP60 (pink) and the Golgi for GM130 (green). Scale bar, 10 μm. (b) Quantification of immunofluorescence pictures showing fragmented recruited and fragmented dispersed Golgi in Ctr WT or Ctr Tn-cdu1 infected HeLa cells. Phenotype of Golgi in infected cells was determined and counted in immunofluorescence pictures obtained in (a). Three independent experiments were performed and approximately 100 cells were counted. Significance was calculated using the Student's t test. n.s. not significant, *p < .05, **p < .01, ***p < .001 for reduced p62 levels. Therefore, a PentaKO cell line with deletions in the autophagy markers p62, NBR1, NDP52, Optineurin and TAX1BP1 was used to test the involvement of any known autophagy marker (Lazarou et al., 2015). The PentaKO cell line as well as the parental strain are also negative for Parkin E3 ligase. Knock out of all these autophagy genes did not rescue the growth of Ctr Tn-cdu1 in primary or progeny infection (Figure 4e, f). In line with previous results we therefore claim that the growth defect of Ctr Tn-cdu1 is not due to the autophagy and degradation by the lysosome but rather a slowed replication and transition of chlamydial RBs to EBs.

| Altered Golgi fragmentation of Ctr Tn-cdu1 can be rescued
Reasons for impaired development in Chlamydia can be various. As obligate intracellular bacteria Chlamydia are known to consume metabolites and nutrients from the host cell. In wild-type Chlamydia infected cells the Golgi is fragmented and recruited to the inclusion surface for enhanced lipid acquisition by the bacteria . In Ctr Tn-cdu1 infected HeLa cells we find the Golgi also fragmented, however, recruitment to the chlamydial inclusion is impaired and the fragmented Golgi vesicle appeared dispersed around the inclusion (Figure 5a, b). To demonstrate that this reduced Golgi vesicle recruitment is due to a defect in Cdu1 activity in cells infected with mutant Chlamydia we expressed active Cdu1 in the cytoplasm of the cell. In cells complemented with Cdu1 Golgi stacks were recruited to the Ctr Tn-cdu1 inclusion as in wild-type infected cells (Figure 6a, b). Golgi fragmentation was hardly visible in infected Fimb cells ( Figure S5) suggesting that defects in Golgi fragmentation do not account for the growth difference of wildtype and mutant Chlamydia in these cells. Furthermore, acquisition of ceramide appeared neither affected in HeLa nor in Fimb cells ( Figure S6), ruling out a defect in ceramide uptake as a reason for the developmental defect of Ctr Tn-cdu1. All in all our data suggest that Cdu1 is required for the efficient recruitment of Golgi vesicles to the inclusion surface of Chlamydia, an effect that appears to be host cell dependent and independent of ceramide uptake into the inclusion.

| DISCUSSION
Obligate intracellular bacteria like Chlamydia acquire metabolites from the host cell, which is expected to induce nutrient stress and autophagy. In addition, recent research into the role of autophagy as cell autonomous defence against intracellular pathogens suggested that autophagy not only defeats bacteria residing in the cytosol of the host but also bacteria in membrane bound phagosomes if these are leaky (Birmingham et al., 2006;Gomes & Dikic, 2014;Mitchell & Isberg, 2017). It was therefore an obvious assumption that autophagy may play a role in restricting chlamydial infection either as a consequence of nutrient stress or as cell autonomous defence. Our data suggest a complex interaction of Chlamydia with cell autonomous defence mechanisms restricting metabolite acquisition and autophagic destruction.
Although Chlamydia does not reside in the cytosol of host cells for extended times, a leaky or perforated inclusion membrane during late stage infections or as a consequence of cell autonomous immune attack may offer targets for activating autophagy (Fischer & Rudel, 2018;Weber et al., 2017a). In Ctr Tn-cdu1, but not in wild-type infected cells we find the inclusion decorated with ubiquitin and recruitment of autophagy proteins already during the replicative phase of the bacteria (24 hpi) when the inclusion is still intact (data not shown). We therefore postulate that the chlamydial inclusion is constantly targeted by the autophagy machinery but the chlamydial deubiquitinase Cdu1 efficiently interferes with this process at the initial steps by removing ubiquitin from the inclusion surface. This role of Cdu1 is supported by the observation that substitution of active Cdu1 in the cell cytoplasm removes the ubiquitin moieties from the inclusion membrane, emphasising the importance of the active DUB for keeping the inclusion free of ubiquitin (Figure 2c, d). In infected cells, this process is restricted to the inclusion surface where the deubiquitinase is anchored, since wild-type Cdu1 is not able to deubiquitinate neighbouring mutant inclusions; the involvement of soluble cytosolic factors can be excluded.
In contrast to our observations, LC3 has been shown to be recruited to C. trachomatis inclusions in other studies (Al-Younes et al., 2004;Al-Younes et al., 2011). These authors used a polyclonal goat antiserum (which is no longer available) which was directed against the unmodified LC3 form. It is therefore possible that the modified active LC3 preferably detected by the antibody used in this study is recruited only to the ubiquitinated inclusion. Despite the strong ubiquitination and the recruitment of autophagy receptors and effectors for phagophore formation to the inclusion in mutant infected cells, autophagy appears not to be a major mechanism by which host cells restrict chlamydial infection. Interestingly, p62 and LC3 are increased during the mid-phase of infection, when Cdu1 protein levels are also upregulated (Fischer et al., 2017). Since we neither find ubiquitination nor the recruitment of autophagy marker during the early phase of infection before Cdu1 is expressed, the inclusion may either not be detected or detection by the autophagic system may be actively prevented independent of Cdu1. Moreover, inhibition of autophagy neither in early phases by knock down of autophagy receptors (Figure 4d, e, f) nor in late phases by Bafilomycin A1 inhibitor treatment (Figure 4b, c) could rescue the growth defect of Ctr Tn-cdu1. Also inclusions of mutant Chlamydia in HeLa cells that do not restrict growth and progeny infection are also decorated with autophagy markers (Figures 2a and 3a,   Figure S4a, b). Finally, no fusion with lysosomes was detected indicating that Ctr Tn-cdu1 is not degraded in autolysosomes (Figure 4a).
These findings therefore strongly suggest that the growth defect of Ctr Tn-cdu1 is not due to clearance of infection by autophagy.
Our electron microscopic analyses revealed a reduced number of EBs in Ctr Tn-cdu1 compared to wild-type bacteria (Figure 1c, d) in line with impaired progeny infection of the mutant. We therefore postulate that redifferentiation of RBs to EBs is affected in Ctr Tn-cdu1.
Recently redifferentiation has been correlated to the RB size which is reduced in every replication cycle until a certain threshold is reached and EBs are formed (Lee et al., 2018). Analysis of bacterial RB size demonstrated that mutant bacteria are larger (Figure 1e) consistent with a disturbed differentiation to EBs. Although the complete mechanism of RB to EB transition is still unknown, the observed phenotype could be explained by slowed growth and replication of Ctr Tn-cdu1.
Since bacteria depend on host metabolites for growth that they also take from organelles like the Golgi apparatus, the structure of the Golgi was further analysed. Indeed, Golgi membranes were less recruited to the Ctr Tn-cdu1 than the wild-type inclusions (Figure 5a, b).
Reduced availability of Golgi membranes at the chlamydial inclusion is known to limit chlamydial growth Lipinski et al., 2009). However, we could not observe a major difference in the uptake of ceramide into the inclusion of wild-type Chlamydia and Ctr Tn-cdu1. Furthermore, the growth defect of Ctr Tn-cdu1 was prominent in primary Fimb cells, which neither display major Golgi fragmentation upon infection with wild-type or mutant Chlamydia nor defects in ceramide uptake into their inclusions. Recently, Ctr Tn-cdu1 was suggested to avoid Golgi fragmentation (Pruneda et al., 2018). In contrast to their data we did see Golgi fragmentation in the cytoplasm of

| Cultivation of Ctr and inhibitor treatment
For chlamydial infection cells were grown until a confluency of 70% was reached. Medium was exchanged for RPMI-1640 containing 5% FCS and chlamydial particles resembling a MOI of 1. Cells were incubated at 35 C and 5% CO 2 until further needed. For inhibitor treatment 10 nM Bafilomycin A1 (Thermo Fischer) or 2 μM MLN4924 was added to the infected cells 17 or 22 hpi respectively. For stock preparation cells were scraped and lysed with glass beads 48 hpi.
Chlamydia were separated by centrifugation at 1000g for 10 min and pelleted at 30 000g for 30 min both at 4 C. After washing the bacteria with 1× sucrose-phosphate-glutamic acid (SPG buffer (75 g/L sucrose, 0.52 g/L KH 2 PO 4 , 1.22 g/L Na 2 HPO 4 , 0.72 g/L L-glutamic acid, pH 7.4) Chlamydia were re-suspended in 1× SPG, aliquoted and stored at −80 C until needed. Bacteria were tested for mycoplasma contamination by PCR and inclusion forming unit was determined.

| Infectivity assays
Fimb, HUVEC HeLa and PentaKO cells were seeded in two 12 well microplates and grown to a confluency of 70%. Cells were infected with Ctr WT or Ctr Tn-cdu1 with an MOI of 1 for HeLa and PentaKO cells and an MOI of 0.3 for primary cells as described before. Primary lysates were taken 24 hpi by adding 2× SDS lysis buffer (100 mM Tris/HCl [pH 6.8], 20% glycerin, 4% SDS, 1.5% 2-mercaptoethanol, 0.2% bromphenolblue) to the cells. For progeny infection infected cells were lysed with glass beads 48 hpi and an aliquot of the supernatant was used to infect a fresh 12 well plate seeded with fresh cells.
Lysates of progeny infection were taken 24 hpi as described. For analysis in immunoblot lysates were heated at 94 C for 5 min.

| Indirect immunofluorescence
Cells were grown on glass cover slides and infected as described above.
For inducible cell lines 50 ng/ml AHT was added 4 hr before harvest.

| Ceramide uptake assay by live cell imaging
Cells were seeded in glass bottom dishes (ibidi, Gräfelfing, Germany) and infected with Ctr WT and Ctr Tn-cdu1 for 24 hr. 1.5 μM SiR-DNA dye (Spirochrome, Stein am Rhein, Switzerland) was added and incubated for 1 hr. Cells were washed three times with PBS, before live cell imaging media (RPMI 1640 (Gibco) + 25 mM N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (HEPES) (Gibco)) was added. Imaging was performed using a Leica TCS SP5 confocal microscope with a 63× oil immersion objective connected with a prewarmed (37 C) incubation chamber. 0.5 μM nitrobenzoxadiazole (NBD)-C 6 -ceramide (Invitrogen, Waltham, Massachusetts, USA) was added to the cells and images were acquired every minute using the LAS software. Recorded pictures were processed with ImageJ.
Blots were quantified using ImageJ software.

| Statistical analysis
Statistical analysis was performed using Excel. Statistical significance between groups was determined using two-tailed Student's t test.
Data are shown as mean ± SE.

CONFLICT OF INTEREST
The authors declare that they have no conflict of interest with the contents of this article.