GBF1 but not BIG1/2 function is partially required for SM acquisition but dispensable for C. trachomatis replication

In an RNA interference (RNAi) screen for host factors required for C. trachomatis L2 infection of Drosophila S2 cells, homologs of Arf1 and its associated GEFs, GBF1 (localized to the cis-Golgi) and BIG1 (localized to the trans-Golgi) were identified [53]. Since Chlamydia is thought to occupy a post Golgi compartment and since C. trachomatis induced Golgi fragmentation enhances replication [6], [21], [22], we tested the hypothesis that spatially-localized Arf1 function within the Golgi might be important for SM acquisition and/or for C. trachomatis intracellular growth.

We first examined the localization of Arf1-GFP, GBF1, and BIG1 during C. trachomatis infection of HeLa cells. We found that GBF1 and BIG1 were localized to the fragmented Golgi around the inclusion in C. trachomatis-infected cells and appeared to maintain their cis and trans polarity, as observed in uninfected cells (Figure 1A). GBF1 (Figure 1B) and BIG1 (data not shown) were primarily localized adjacent to but not on the inclusion membrane. Similar results were obtained in cells expressing GFP-tagged GBF1 (data not shown). In contrast, Arf1-GFP was found not only on Golgi vesicles adjacent to the inclusion but was also localized to the inclusion membrane, where it formed a thin rim of circumferential staining (Figure 1B), consistent with previously published results [40], [41]. We noted a striking concentration of Arf1-GFP at the region of two closely apposed inclusion membranes (Figure 1B). To evaluate whether GBF1 or BIG1/2 activity was required for Arf1-GFP recruitment to the inclusion membrane, we treated cells with BFA, which inhibits both GBF1 and BIGs, and examined the localization Arf1-GFP and GBF1. Exposure of the cells to BFA resulted in dispersion of GBF1 and Arf1-GFP throughout the cell and loss of the circumferential rim staining on the inclusion; however, BFA had no effect on Arf1-GFP localization at abutting inclusion membranes (Figure 1B). This finding suggests that GBF1 and/or BIG1/2 activities are required for Arf1 recruitment to the outer region of the inclusion but that they are not necessary for Arf1 maintenance at closely apposed inclusions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. GBF1 function is required for SM acquisition but not for C. trachomatis replication.

(A) HeLa cells were infected with C. trachomatis L2 for 24 hrs and then fixed and stained with antibodies to GBF1 (red) and BIG1 (green). Bacteria and host DNA were detected using DAPI (blue). The cis and trans polarity of the Golgi was maintained in C. trachomatis L2-infected cells. N, host nucleus. *, inclusion. Scale bar = 5 µm. (B) HeLa cells were transfected with Arf1-GFP for 18 hrs, infected with C. trachomatis L2 for 24 hrs in the absence or presence of 10 µM BFA, and then fixed and stained with antibodies to GBF1 (red). Enlargements of boxed regions are shown to the right. Images represent a single z slice from confocal images. The exposure time for each filter set for all images was identical. Arf1-GFP localized to the region between two closely apposed inclusions (white arrow) and to a thin rim around the inclusion (red arrow) whereas GBF1 was excluded from these regions. *, inclusion. Scale bar = 5 µm. (C) Western blot analysis of siRNA-treated samples. GAPDH was used as a loading control. (D) HeLa cells were depleted of GBF1, BIG1, and/or BIG2 for 3 days, infected with C. trachomatis L2 for 24 hrs, and then labeled with BODIPY FL-Ceramide to visualize SM acquisition by the inclusion. The exposure time for all images was identical. Dashed red lines demarcate the inclusions. Scale bar = 5 µm. (E) HeLa cells were infected with C. trachomatis for 24 hrs, treated with 10 µM BFA or GCA during the last 3 hrs of infection, and then labeled with BODIPY FL-Ceramide to analyze SM acquisition by the inclusion. The exposure time for all images was identical. Dashed red lines demarcate the inclusions. Scale bar = 5 µm. (F) HeLa cells were depleted of GBF1, BIG1, and/or BIG2 for 3 days, infected with C. trachomatis L2 for 24 hrs, and then analyzed for progeny formation as described in Materials and Methods. Values (mean ± standard error) are shown as percentage of control siRNA-treated samples. No significant decrease in progeny formation was observed. IFU, inclusion forming units.

https://doi.org/10.1371/journal.ppat.1002198.g001

To determine the specific contribution of the cis- versus trans-Golgi compartments to SM acquisition, HeLa cells were depleted of GBF1 or BIG1/2 by siRNA, infected with C. trachomatis, and then examined by live-cell microscopy for SM acquisition as described in the Materials and Methods. Protein depletion was assessed by western blot analysis (Figure 1C). As shown in Figure 1D, labeled lipids were readily visible inside bacterial inclusions in control siRNA-treated cells and in cells depleted of BIG1 and/or BIG2. In contrast, depletion of GBF1 resulted in ∼60% reduction in SM acquisition by the inclusion (Figures 1D and S1A; p<0.001). SM acquisition was similarly reduced in cells treated with Golgicide A (GCA), a GBF1-specific inhibitor [37] and with BFA ([17] and Figures 1E and S1B). As GCA targets GBF1 selectively and does not inhibit BIGS, these results suggest that the relevant target of BFA during C. trachomatis L2 infection is GBF1.

We next tested whether GBF1 or BIG1/2 were required for intracellular growth. GBF1 depletion (Figure S1B) or inhibition with GCA or BFA (Figure S2) resulted in inclusions that were smaller than those observed in control siRNA- or untreated cells. Quantitation of inclusion size, as described in Materials and Methods, revealed that the average inclusion size in GBF1-depleted cells was decreased by 40% (p<0.001) compared to control siRNA-treated cells (Figure S1D). In contrast, depletion of BIG1 and/or BIG2 gave rise to inclusions that were slightly larger than those observed in control siRNA depleted cells (Figures S1C and S1D). Despite the changes in inclusion size, depletion of any of the Arf1 GEFS did not decrease production of progeny; in fact, a small increase in progeny was observed (Figure 1F). The reason for the increase in progeny is not currently known.

Together, these results demonstrate that GBF1, but not BIG1/2 function, is required for vesicle-mediated SM acquisition and inclusion growth but is dispensable for bacterial replication. These findings indicate that the impact of disrupting GBF1 function on SM acquisition is not an indirect result of altering Arf1-dependent trafficking in the trans-Golgi, since depletion of BIG1 and/or BIG2 did not reduce SM acquisition or inclusion growth. We note that in Drosophila S2 cells, depletion of either GBF1 or the single BIG1 was sufficient to decrease inclusion formation [53]; this finding may reflect differences in Golgi organization between Drosophila S2 and mammalian cells, and/or it may reflect the fact that these proteins may have additional roles beyond the secretory pathway in host cells.

GBF1 function is required for inclusion membrane stability

We observed that interference with GBF1 function resulted in a defect in SM acquisition by the inclusion, and recent studies have shown that the inclusion membrane integrity is compromised when host SM biosynthesis is inhibited [15]. Therefore, we tested whether GBF1 plays a role in maintaining the integrity of the inclusion membrane. Depletion of GBF1 (Figure 2A) or exposure to GCA or BFA (Figure S2A) resulted in a loss of inclusion membrane integrity, as evidenced by discontinuities in 14-3-3β staining (which binds to the inclusion membrane protein, IncG) and release of bacteria into the host cytoplasm at the broken membrane sites. Quantitative analysis (as described in Materials and Methods) revealed that ∼80% of infected cells in which GBF1 function was inhibited exhibited broken inclusions (Figures 2 and S2A). The loss of inclusion membrane integrity was observed at mid- to late times of infection (data not shown), after significant replication has occurred.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. GBF1 but not BIG1/2 function is required for inclusion membrane stability.

(A) HeLa cells were treated with the indicated siRNA for 3 days, infected with C. trachomatis L2 for 24 hrs, then fixed and stained with antibodies to 14-3-3β (green) to identify the inclusion membrane and GBF1 (red). Bacteria and host DNA were detected using DAPI (blue). The exposure time for each filter set for all images was identical. White arrows point to breaks in the inclusion membrane where the bacteria are released into cytoplasm in GBF1-depleted cells. Inclusions formed in BIG1 and/or BIG2 depleted cells remain intact. (B) HeLa cells were depleted of GBF1 for 3 days, infected with C. trachomatis L2 for 24 hrs, then fixed and stained with antibodies to MOMP (green) to identify bacteria and vimentin (red). Bacteria and host DNA were detected using DAPI (blue). The exposure time for each filter set for all images was identical. White arrows point to the region on the inclusion that is devoid of vimentin staining and where bacteria are released into the cytoplasm. N, host nucleus; *, inclusion. MOMP, C. trachomatis major outer membrane protein. Scale bar = 5 µm.

https://doi.org/10.1371/journal.ppat.1002198.g002

Recent studies have shown that the inclusion is stabilized by a cage-like structure composed of F-actin and vimentin [7]. Although it has been reported that inhibition of SM biosynthesis did not result in alteration of the vimentin or actin structure around the inclusion [15], experiments using BFA have revealed that perturbations in vesicular trafficking lead to reciprocal changes in the architecture of the vimentin network in the host cell, indicating a tight link between membrane trafficking and the cytoskeleton [39]. Furthermore, Arf1 has been shown to regulate actin dynamics at Golgi membranes [54]. We therefore considered whether the loss of inclusion integrity observed upon disruption of GBF1 function was due solely to the decrease in SM acquisition or whether disruption of GBF1 function also affected the vimentin and actin cage surrounding the inclusion. We examined the localization of these cytoskeletal proteins in infected cells treated with GBF1 siRNA or GCA. Disruption of GBF1 function caused both the vimentin and actin cages to collapse to a thin, hemispheric cup-like structure on the nuclear side of the inclusion (Figures 2B, S2B, and S2C). In cells treated with either BFA or GCA, we also observed less GBF1 staining surrounding the inclusion, and GBF1 displayed a more compact Golgi-like localization (Figure S2A). GBF1 inhibited cells were more sensitive to Triton extraction (data not shown), similar to what has been reported following disruption of the actin and vimentin cage during infection [7], suggesting that the cytoskeletal cage is compromised upon GBF1 inhibition.

Together, these results indicate that GBF1 function contributes to the stability of the inclusion membrane by a mechanism that involves SM acquisition and/or maintenance of the vimentin and F-actin cage around the inclusion; however, GBF1 function is not required for intracellular replication. Given that SM biosynthesis is essential for replication [6], [12], our findings suggest that other lipid transport mechanisms may be involved in the acquisition of host SM.

The ceramide transport protein, CERT and its ER binding protein, VAP-A, are recruited to the inclusion

Since ceramide transport from the ER to the trans-Golgi by the cytosolic lipid transporter, CERT, is the predominant pathway for SM biosynthesis [42], [43], [44], [45]), we tested whether CERT-dependent trafficking was involved in SM acquisition by C. trachomatis. We first examined the localization of CERT during C. trachomatis infection of HeLa cells that were transiently transfected with CERT-GFP. In uninfected cells, CERT-GFP preferentially localized to the Golgi as well as to punctate, vesicle-like structures outside the Golgi (Figure S4 and [45], [55]), which are thought to be ER-derived (K. Kumagai and K. Hanada, personal communication). In C. trachomatis-infected cells, CERT-GFP was recruited to the inclusion, exhibiting a non-homogeneous, patchy distribution on the inclusion membrane surface (Figure 3A, B, D, E, Video S1). CERT-GFP appeared on the same plane as but distinct from localization of the C. trachomatis inclusion membrane protein, IncA (Figure 3B). This pattern of CERT recruitment to the inclusion was confirmed using an antibody to endogenous CERT (Figure S3), demonstrating that localization was not an artifact of ectopic expression. CERT-GFP also localized to C. trachomatis serovar D inclusions (Figure 3B), suggesting that the recruitment to the inclusion was not serovar-specific. CERT-GFP was recruited to inclusions as early as 2 hpi, where nascent inclusions were enveloped by CERT-GFP-containing vesicles that partially overlapped with p58, a marker of the ER-Golgi intermediate compartment (ERGIC) (Figure S4A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. CERT and VAP-A are recruited to the inclusion.

(A) HeLa cells transfected with CERT-GFP for 18 hrs were left uninfected or infected with C. trachomatis L2 for 24 hrs. Images shown are maximum intensity projections of confocal z-stacks (0.4-µm slices). N, host nucleus; *, inclusions. Scale bar = 5 µm. (B) HeLa cells transfected with CERT-GFP were infected with C. trachomatis serovar D for 24 hrs and then fixed and stained with antibodies to IncA (red) to identify the inclusion membrane. Bacteria and host DNA were detected using DAPI (blue). Enlargements (inset) of boxed regions are shown to the right. *, inclusions. Scale bar = 5 µm. (C–E) HeLa cells were transfected with CERT-GFP and HcRedVAP-A for 18 hrs and infected with C. trachomatis L2 for (C) 2, (D) 8, or (E) 24 hrs. (C) Cells were stained with DAPI to visualize the nascent inclusions (red arrows). (D and E) Enlargements (inset) of boxed regions are shown to the right. At 8 and 24 hpi, CERT-GFP and HcRedVAP-A colocalize on the inclusion membrane and exhibit a patchy distribution. Images shown are maximum intensity projections of confocal z-stacks (0.4-µm slices). N, host nucleus; *, inclusions. Scale bar = 5 µm, except with insets from panels B, C, and D where scale bar = 2.5 µm.

https://doi.org/10.1371/journal.ppat.1002198.g003

CERT binding to VAP at the ER is required for CERT to extract ceramide from the ER and deliver it to the trans-Golgi [45], where ceramide serves as a substrate for SM synthesis by trans-Golgi localized SMS1 and SMS2 [46], [56]. Previous work suggests that there is an intimate association of the ER with the chlamydial inclusion [57], [58]. We examined the localization of CERT and VAP-A by co-expressing CERT-GFP and HcRed-VAP-A [45] in infected cells and found that HcRed-VAP-A colocalized with CERT-GFP at the inclusion at 2, 8, and 24 hpi (Figures 3C–E). These results suggest that CERT may function to bring the ER membrane in close apposition to the inclusion or that VAP-A is recruited directly to the inclusion.

CERT is required for bacterial replication and SM acquisition

Given the proximity of the Golgi and the striking recruitment of CERT to the inclusion, we used three complementary approaches to determine whether CERT plays a role in C. trachomatis infection: (1) pharmacological inhibition of CERT by HPA-12, a synthetic analog of ceramide that specifically inhibits CERT-mediated ceramide transfer [59], [60], (2) overexpression of Casein Kinase Iγ2 (CKIγ2), which plays a crucial role in down-regulating CERT activity [48], and (3) depletion of CERT by siRNA.

We first examined inclusion morphology and progeny production in cells treated with increasing concentrations of HPA-12. Cells were infected for 1 hr and then incubated in the absence or presence of HPA-12 for an additional 23 hrs. We observed a dose-dependent decrease in inclusion formation (Figures 4A) and progeny production (Figure 4B) in cells exposed to HPA-12, resulting in a 96% reduction in progeny at 10 µM HPA-12 compared to DMSO (p<0.001). At the doses employed, HPA-12 did not affect host or bacterial cell viability nor did it affect bacterial binding and entry (data not shown).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. CERT function is required for C. trachomatis replication.

HeLa cells were infected with C. trachomatis L2, treated with the indicated concentration of HPA-12 at 1–24 hpi, and then (A) fixed and stained with antibodies to MOMP (red) and with DAPI (blue) to visualize bacteria or (B) analyzed for progeny formation. Values (mean ± standard error) are shown as percentage of DMSO treated samples. Data are representative of 3 independent experiments. ***p<0.001 compared to DMSO treated cells (ANOVA). (C) HeLa cells were transfected with CERT-GFP and HA-CKIγ2 for 18 hrs, infected with C. trachomatis L2 for 24 hrs, and then fixed and stained with antibodies to HA (red). The exposure time for each filter set of all images was identical. Images shown are maximum intensity projections of confocal z-stacks (0.4-µm slices). Ectopic expression of HA-CKIγ2 decreased inclusion size but did not affect CERT-GFP recruitment to the inclusion membrane. Scale bar = 5 µm. (D) Western blot analysis of siRNA-treated samples. GAPDH was used as a loading control. (E) HeLa cells were depleted of CERT for 3 days and then labeled with BODIPY FL-Ceramide to analyze SM accumulation in the Golgi. The exposure time of all images was identical. CERT depletion reduced SM accumulation in the Golgi. Scale bar = 5 µm, (F) HeLa cells were depleted of CERT for 3 days, infected with C. trachomatis L2 for 24 hrs, and then fixed and stained with antibodies to MOMP (green) to identify the inclusion. Bacteria and host DNA were detected using DAPI (blue). The exposure time for all images was identical. CERT depletion reduced inclusion size. Red arrows point to inclusions. Scale bar = 5 µm. (G) HeLa cells were depleted of CERT for 3 days, infected with C. trachomatis L2 and analyzed for inclusion size and progeny formation. Values (mean ± standard error) are shown as percentage of control siRNA samples. CERT depletion significantly reduced inclusion size and progeny formation. Data are representative of 2 independent experiments. ***p<0.001 for CERT siRNA-treated cells compared to control siRNA-treated cells (ANOVA). N, host nucleus. IFU, inclusion forming units. Scale bar = 5 µm.

https://doi.org/10.1371/journal.ppat.1002198.g004

CKIγ2 regulates CERT by hyperphosphorylation of a SR motif located within the middle region, which induces an autoinhibitory interaction between the PH and START domains. This event interferes both with the ability of CERT to bind PI4P and with its ability to transport ceramide, though CERT is still able to interact with VAP-A at the ER [48], [61]. Overexpression of CKIγ2 is sufficient to inhibit CERT function and to cause dissociation of CERT from the Golgi complex [61]. In cells co-transfected with CERT-GFP and HA-tagged CKIγ2, we observed a significant defect in inclusion development, with the appearance of small inclusions (Figure 4C), similar to the results observed with HPA-12 (Figures 4A and 6C). Importantly, CERT-GFP was still recruited to the inclusion in cells expressing HA-tagged CKIγ2 (Figure 4C). These results are consistent with the notion that CERT activity is required for inclusion development and suggest that PI4P binding is not required for CERT recruitment to the inclusion.

We next examined inclusion morphology in HeLa cells depleted of CERT for 72 hrs and subsequently infected with C. trachomatis for 24 hrs. The efficiency of CERT depletion was determined in uninfected cells by western blot analysis (Figure 4D). Loss of CERT activity was confirmed by demonstrating decreased accumulation of BODIPY FL-SM and BODIPY FL-Ceramide at the Golgi apparatus (Figure 4E). CERT depletion resulted in a ∼66% (p<0.001) reduction in inclusion size as well as a ∼60% (p<0.001) decrease in progeny production (Figures 4F–G). Depletion or inhibition of CERT sometimes resulted in the appearance of inclusions containing aberrant appearing RBs (Figure 4F and data not shown). Progeny formation was decreased more robustly in HPA-12 treated cells than in CERT-depleted cells, most likely because CERT depletion was not complete, although we cannot rule out the possibility that HPA-12 may have additional off-target effects.

We next explored whether CERT was required for the SM accumulation in the inclusion by monitoring incorporation of BODIPY FL-Ceramide. Since depletion of CERT led to a significant reduction in inclusion size (Figure 4F), we monitored lipid transport in the presence of the CERT inhibitor, HPA-12. HeLa cells were infected for 20 hrs to allow inclusions to develop to a large enough size for robust visualization by phase microscopy, then exposed to HPA-12 for 3 hrs, and subsequently pulse-labeled with BODIPY FL-Ceramide. The fluorescent lipid readily accumulated in inclusions within DMSO-treated cells, however, inclusions in HPA-12-treated cells showed ∼56% (p<0.001) decrease in fluorescence intensity (Figure 5). Similar results were observed in control cells treated with D609, an inhibitor of SMS1 and SMS2 [62], [63], [64] or with BFA (Figure 5; p<0.001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. CERT function is required for SM acquisition.

(A) HeLa cells were infected with C. trachomatis L2 for 24 hrs, treated with 5 µM HPA-12 for the last 3 hrs of infection, and then labeled with BODIPY FL-Ceramide to analyze SM accumulation by the inclusion. As a control for decreased SM acquisition by the inclusion, cells were also treated with 10 µM BFA or 25 µg/ml D609. The exposure time of all images was identical. Dashed red lines demarcate inclusions. The residual fluorescence likely represents Golgi staining. Scale bar = 5 µm. (B) Quantitation of SM acquisition following treatment with HPA-12, D609, or BFA. Values (mean ± standard error) are shown as percentage of mean fluorescence intensities relative to DMSO-treated samples. ***p<0.001 (ANOVA). HPA-12, D609, and BFA-treated cells displayed a significant decrease in fluorescence intensity of the inclusion and its contents compared to DMSO-treated samples.

https://doi.org/10.1371/journal.ppat.1002198.g005

Since CERT is localized to the Golgi during infection, it is possible that CERT is indirectly involved in C. trachomatis infection based on its role in host SM biosynthesis at the Golgi. In addition or alternatively, as CERT and VAP-A are also localized at the inclusion, CERT may be directly involved in C. trachomatis infection by mediating the transfer of a portion of host ceramide from the nearby ER directly to the inclusion. We examined the localization of ceramide in infected cells expressing CERT-GFP using a ceramide antibody. This antibody does not cross-react with SM, cholesterol, or other phospholipids under physiologic in vivo conditions [65]. Ceramide appeared to be highly enriched in the region adjacent to the inclusion in what is likely to be the surrounding ER (Figure S5). Importantly, we also found that a fraction of ceramide localized at the inclusion membrane as evidence by co-staining with IncA antibodies and Arf1-GFP (Figure S5A and S5B). Although we found ceramide in close association with CERT-GFP at the inclusion membrane, ceramide and CERT-GFP did not overlap on the inclusion membrane (Figure S5C), possibly reflecting conversion of ceramide to SM and loss of reactivity with the ceramide antibody. Taken together, these results indicate that CERT is required for C. trachomatis replication and that interference with CERT correlates with a block in the SM acquisition by the inclusion. In addition, the localization of ceramide during infection suggests that CERT may be functioning at the Golgi as well as at the inclusion.

CERT transfer activity and ceramide binding are required for its recruitment to the inclusion

To determine the mechanism of CERT recruitment to the inclusion, we examined the consequences of inactivating specific domains of CERT. The ability of CERT to transfer ceramide for SM biosynthesis requires: (1) interaction with the Golgi membrane (PI4P) via its PH domain, (2) an interaction with the ER membrane proteins VAP-A and VAP-B via its FFAT motif, and (3) the ability to bind ceramide via its START domain [47]. We were particularly interested in the PH domain, as recent studies have shown that multiple proteins that regulate PI4P metabolism are recruited to the inclusion, including OCLR (oculocerebrorenal syndrome of Lowe protein 1) and the PI4P kinases, PI4KIIα and to a lesser extent PI4KIIβ [40]. The presence of PI4P at the inclusion combined with our finding that VAP-A colocalizes with CERT at the inclusion (Figures 3C–3E) raises the possibility that CERT could be recruited to the inclusion by its interaction with PI4P and/or with VAP-A.

We examined the localization of two GFP-tagged CERT mutants: CERT (G67E), which contains a mutation in the PH domain that prevents CERT from binding PI4P at the Golgi, or CERT (D324A), which harbors a mutation in the FFAT motif that prevents CERT from interacting with VAPs at the ER [43], [45] during C. trachomatis infection. In uninfected cells, CERT (G67E)-GFP is distributed throughout the cell while CERT (D324A)-GFP is concentrated at the Golgi (data not shown and [45]). As shown in Figure 6A, both CERT (G67E)-GFP and CERT (D324A)-GFP localized to the inclusion at 24 hpi in a manner indistinguishable from CERT-GFP. While CERT-GFP and CERT (D324A)-GFP show some Golgi localization during infection, none is observed for CERT (G67E)-GFP (Figure 6A). Co-staining with antibodies to Calnexin, an ER marker, revealed that although the ER appears in close proximity to the inclusion, it does not overlap with CERT-WT, D324A, or G67E at the inclusion (Figure S6). CERT was still recruited to inclusions formed in a mutant cell line of Chinese hamster ovary cells, LY-A, which carries a mutation in the PH domain of endogenous CERT (G67E) (Figure S3) [43], [66]. Despite the fact that LY-A cells have reduced SM biosynthesis [43], [66], SM acquisition by C. trachomatis inclusions was normal in these cells (data not shown).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. CERT transfer and/or ceramide binding activity are required for its recruitment to inclusions.

(A) HeLa cells were transfected with CERT-GFP, CERT (D324A)-GFP, or CERT (G67E)-GFP, infected with C. trachomatis L2 for 24 hrs, and then fixed and stained with an antibody to p230 (red) to identify the trans-Golgi. The exposure time for each filter set of all images was identical. Images shown are maximum intensity projections of confocal z-stacks (0.4-µm slices). Mutation of the PI4P binding (G67E) or VAP-A binding (D324A) domains did not affect CERT-GFP recruitment to the inclusion. Scale bar = 5 µm. (B) HeLa cells expressing CERT-GFP were infected with C. trachomatis L2, treated with 50 µM Exo1 (Arf1 inhibitor) for 1–24 hpi, and then fixed and stained with antibodies to MOMP (red) to identify bacteria. The exposure time for each filter set of all images was identical. Images shown are maximum intensity projections of confocal z-stacks (0.4-µm slices). CERT-GFP localization to the inclusion was unaffected by Arf1 inhibition. Scale bar = 5 µm. (C) HeLa cells expressing CERT-GFP were infected with C. trachomatis L2, treated with 5 µM HPA-12 for 1–24 hpi, and then fixed and stained with antibodies to MOMP (red) to identify bacteria or to p230 (red) to identify the trans-Golgi. The exposure time for each filter set of all images was identical. Inhibition of CERT transfer and/or ceramide binding activity by HPA-12 treatment resulted in loss of CERT accumulation on the inclusion membrane. Images shown are maximum intensity projections of confocal z-stacks (0.4-µm slices). Scale bar = 5 µm. N, host nucleus; *, inclusion.

https://doi.org/10.1371/journal.ppat.1002198.g006

We next tested whether Arf1 activity was required for CERT recruitment, since Arf1 localizes to the inclusion and since the PH domains of some lipid transfer proteins, such as FAPPs (four-phosphate-adaptor protein 2) and OSBP (oxysterol binding protein), can simultaneously bind PI4P and Arf1 [67]. C. trachomatis-infected HeLa cells were treated with Exo1, a specific inhibitor of Arf GTPase activity [68], from 1 to 24 hpi, and CERT-GFP localization was examined at 24 hpi. Although quantitation of inclusion size in Exo1 treated cells revealed a ∼64% decrease (p<0.001), Exo1 had no effect on CERT recruitment to the inclusion (Figure 6B). Together, these results indicate that interaction with PI4P, VAP-A, or Arf1 alone is not essential for CERT recruitment to the inclusion, though PI4P, VAP-A, or Arf1 binding may function redundantly or there may be other determinants within CERT. In addition, CERT-GFP was still recruited to inclusions in cells treated with GCA, BFA, or Nocodazole, which disrupts microtubules (Figures S4B and 7C), indicating that GBF1 function, vesicular trafficking, and microtubules are not required for CERT recruitment.

To test whether CERT lipid transfer activity and/or ceramide binding is required for CERT recruitment to the inclusion, we treated cells expressing CERT-GFP with 5 µM HPA-12 from 1 to 24 hpi. In control experiments, HPA-12 treatment did not affect CERT-GFP localization to the Golgi region in uninfected cells (data not shown). In C. trachomatis-infected cells exposed to HPA-12, the patchy distribution of CERT-GFP was lost from the inclusion (Figure 6C, top row), although CERT-GFP was still present in the Golgi surrounding the inclusion, as evidenced by colocalization with p230, a trans-Golgi marker (Figure 6C, bottom row). These results indicate that ceramide binding and/or the transfer activity of CERT is required for its recruitment to the inclusion.

SMS1 and SMS2 are recruited to distinct compartments at the inclusion

Since CERT cannot directly transfer SM [44], we considered the possibility that SMS1 or SMS2 are directly recruited to the inclusion, where they could synthesize SM following ceramide transfer to the inclusion by CERT. In uninfected cells, both SMS1 and SMS2 are found at the trans-Golgi while SMS2 also localizes to the plasma membrane (Figure S7 and [46], [69]).

We examined the localization of CERT-GFP and C-terminally 3xFLAG-tagged SMS1/SMS2 or SMS1-V5 and C-terminally 3xFLAG-tagged SMS2 during C. trachomatis infection at 24 hpi. In uninfected cells, C-terminally 3xFLAG-tagged SMS1 and 3xFLAG-tagged SMS2 were located in close apposition to CERT-GFP at the Golgi (Figure 7A). In infected cells, C-terminally 3xFLAG-tagged SMS1 primarily localized to the fragmented Golgi surrounding the inclusion and displayed no overlap with CERT-GFP at the inclusion (Figure 7A). In contrast, C-terminally 3xFLAG-tagged SMS2 localized to both the fragmented Golgi as well as to the inclusion membrane, where it exhibited a distinct punctate pattern that partially overlapped with CERT-GFP (Figure 7A). Localization of SMS2 to the inclusion membrane was confirmed by co-staining infected cells expressing SMS2-V5 with antibodies to IncA; inclusion-membrane localized SMS2-V5 appeared to partially overlap with IncA whereas no overlap was seen between IncA and SMS1-V5 (Figure 7B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. SMS1 and SMS2 localization in C. trachomatis infected cells.

(A) HeLa cells co-transfected for 18 hrs with CERT-GFP and C-terminally 3xFLAG-tagged SMS1/SMS2 (upper 2 rows) or SMS1-V5 and C-terminally 3xFLAG-tagged SMS2 (bottom row) were infected with C. trachomatis L2 for 24 hrs, and then fixed and stained with antibodies to FLAG (red) and/or to V5 (green). Single channel images of uninfected cells are shown to the right. Enlargements of the boxed regions (inset) in infected samples are shown to the right of infected set. The exposure time for each filter set of all images was identical. Images shown are maximum intensity projections of confocal z-stacks (0.4-µm slices). CERT and SMS1 localization at the inclusion are distinct while CERT and SMS2 partially overlap. SMS2 localization partially overlaps with SMS1 around the inclusion, however SMS2 also localizes to the inclusion. At longer exposure times, SMS2 plasma membrane localization is evident. Scale bar = 5 µm, except the insets where scale bar = 2.5 µm. (B) HeLa cells transfected with SMS1-V5 (green) or SMS2-V5 (green) infected with C. trachomatis serovar D for 24 hrs and then fixed and stained with antibodies to IncA (red) to identify the inclusion membrane. Enlargements of the boxed regions (inset) are shown to the right. Images represent a single z slice from confocal images. The exposure time for each filter set for all images was identical. SMS2 but not SMS1 partially overlaps with IncA on the inclusion. Scale bar = 5 µm, except the insets where scale bar = 2.5 µm. (C) HeLa cells transfected for 18 hrs with CERT-GFP, SMS1-V5, or SMS2-V5 were infected with C. trachomatis L2, treated with 10 µM BFA or Nocodazole for 1–24 hpi, and then fixed and stained with antibodies to V5 (green) and to p230 (red) to identify the trans-Golgi. BFA and Nocodazole disrupted SMS1 localization around the inclusion but had no effect on SMS2 or CERT localization at the inclusion. Images represent a single z slice from confocal images. Scale bar = 5 µm. N, host nucleus; *, inclusion.

https://doi.org/10.1371/journal.ppat.1002198.g007

To further distinguish between inclusion membrane versus Golgi localization, we examined the distribution of SMS1-V5 or SMS2-V5 in infected HeLa cells following disruption of the Golgi by BFA or by Nocodazole. As shown in Figure 7C, exposure of C. trachomatis-infected cells to either BFA or to Nocodazole caused SMS1-V5 to disperse throughout the cell while SMS2-V5 remained primarily associated with the inclusion, where a rim-like staining was observed. These results indicate that SMS2 is tethered to the inclusion during infection whereas SMS1 is primarily associated with the Golgi around the inclusion. In addition, SMS2 recruitment to the inclusion is independent of microtubules and of Arf1/GBF1-dependent vesicular trafficking.

The mechanism by which SMS1 or SMS2 localize to the trans-Golgi has not yet been elucidated. However, it was recently shown that SMS2 is palmitoylated at the carboxyl terminus and that this modification plays a role in targeting SMS2 to the plasma membrane [69]. We therefore tested whether SMS2 recruitment to the inclusion required palmitoylation by examining the localization of a C-terminally 3xFLAG-tagged SMS2 construct in which all four of the potential palmitoylation sites were mutated (C331, 332, 343, 348A) [69]. Although this mutant displays a decrease in PM distribution (data not shown and [69]), it was recruited to the inclusion in a similar fashion as C-terminally 3xFLAG-tagged SMS2 wild-type (Figure S7). This result suggests that palmitoylation is not required for SMS2 recruitment to the inclusion.

SMS1 and SMS2 are required for C. trachomatis replication

Both SMS1 and SMS2 are required for host SM production, however their relative contributions to subcellular SM levels are distinct, with SMS1 generating the bulk of SM at the Golgi while SMS2 is responsible for SM biosynthesis at the plasma membrane [46], [56]. Our observation that SMS2 but not SMS1 is recruited to the inclusion membrane prompted us to determine their relative contributions to infection. Cells were depleted of SMS1 or SMS2 by siRNA [56] for 72 hrs and subsequently infected with C. trachomatis for 24 hrs. The efficiency of depletion was verified by western blot analysis in cells expressing C-terminally 3xFLAG-tagged SMS1 and 3xFLAG-tagged SMS2 (Figure 8A). Loss of SMS1 activity was confirmed by demonstrating decreased accumulation of BODIPY FL-Ceramide and BODIPY FL-SM at the Golgi apparatus (Figure 8B) [56]. Depletion of either SMS1 or SMS2 resulted in a significant reduction in relative inclusion size and progeny formation (Figures 8C–D). These results demonstrate that both SMS1 and SMS2 contribute to C. trachomatis replication and inclusion growth. Our results support the hypothesis that at least a portion of SM acquired by C. trachomatis may be directly synthesized on the inclusion membrane. In addition, SM synthesized at the nearby Golgi by SMS1 and SMS2 transported by a BFA-insensitive pathway may also contribute to replication.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 8. SMS1 and SMS2 contribute to C. trachomatis replication.

(A) Western blot analysis of siRNA-treated samples. GAPDH was used as a loading control. (B) HeLa cells were depleted of SMS1 or SMS2 for 3 days and then labeled with BODIPY FL-Ceramide to analyze SM accumulation in the Golgi. The exposure time of all images was identical. SMS1 but not SMS2 depletion reduced BODIPY FL lipid accumulation in the Golgi. Scale bar = 5 µm. (C and D) HeLa cells were depleted of SMS1 or SMS2 for 3 days, infected with C. trachomatis L2 for 24 hrs, and then fixed and stained with an antibody to MOMP (red). Bacteria and host DNA were detected using DAPI (blue). (D) SMS1 and SMS2-depleted cells were analyzed for inclusion size and progeny formation. Values (mean ± standard error) are shown as percentage of control siRNA samples. SMS1 and SMS2 depletion reduced inclusion size and production of infectious progeny. Data are representative of 2 independent experiments. ***p<0.001, all samples compared to control siRNA treatment (ANOVA). N, host nucleus; red arrows point to inclusions. IFU, inclusion forming units. Scale bar = 5 µm.

https://doi.org/10.1371/journal.ppat.1002198.g008

Reagents

D609 was purchased from Calbiochem. Brefeldin A, Nocodazole, and Exo1 were purchased from Sigma. Golgicide A (GCA) was kindly provided by D. Haslam (Washington University School of Medicine, St. Louis, MO). BODIPY FL C5 Ceramide was obtained from Molecular Probes. Construction of Arf1-GFP, Arf1-mCherry, and GFP-GBF1 were previously described [82], [83]. CERT-GFP, CERT (D324A)-GFP, CERT (G67E)-GFP, HcRed-VAP-A, HA-CKIγ2, and HPA-12 were kind gifts from K. Hanada (National Institute of Infectious Disease, Tokyo) and have been previously described [45], [59]. C-terminally V5-tagged SMS1 and SMS2 constructs were kind gifts from J. Holthuis (Utrecht University) and have been described previously [46]. C-terminally 3xFLAG-tagged SMS1 and SMS2 constructs were kind gifts from M. Tani (Kyushu University) and have been previously described [69]. GFP-p58 was kindly provided by C. Roy (Yale University). Unless otherwise indicated, the following concentrations of inhibitors were used: 5 µM HPA-12, 25 µg/ml D609, 10 µM GCA, 10 µM BFA, 10 µM Nocodazole, and 50 µg/ml Exo1. Antibodies were obtained from the following sources: mouse anti-Chlamydia FITC conjugate (Meridian Diagnostics), goat anti-C. trachomatis MOMP (Cortex Biochem), mouse anti-GAPDH (Chemicon), mouse anti-GBF1 (BD Transduction Laboratories), rabbit anti-BIG1 (Santa Cruz), rabbit anti-BIG2 (Bethyl Laboratories, Inc.), chicken anti-CERT (Sigma), rabbit anti-14-3-3β (Santa Cruz), mouse anti-vimentin (Sigma), mouse anti-FLAG (Sigma), rabbit anti-V5 (Sigma), mouse anti-HA (Covance), rabbit anti-p58 (Sigma), mouse anti-ceramide IgM (Clone15B4), rabbit anti-calnexin (Cell Signaling), rabbit anti-goat IgG horseradish peroxidase (HRP) (Calbiochem), goat anti-rabbit IgG HRP (Amersham Biosciences), goat anti-mouse HRP (Amersham Biosciences), donkey anti-goat Alexa 594 (Molecular Probes), donkey anti-goat Alexa 488 (Molecular Probes), chicken anti-mouse 594 (Molecular Probes), goat anti-mouse IgG1 Alexa 594 (Molecular Probes), donkey anti-goat Alexa 488 (Molecular Probes), Texas red-conjugated donkey anti-chicken IgY (Jackson Laboratories), and donkey anti-rabbit Alexa 488 (Molecular Probes). Texas red-conjugated phalloidin was obtained from Molecular Probes. Mouse anti-IncA antibodies were kind gifts from T. Hackstadt (Rocky Mountain Laboratories), D. Rockey (Oregon State University), and G. Zhong (University of Texas Health Science Center at San Antonio). All siRNAs, including non-targeting negative controls, were obtained from Dharmacon: Human GBF1 (J-019783-06), human BIG1 (D-012207-02, D-012207-03), and human BIG2 (MQ-012208-01) and have been previously described [33], [84]. Human SMS1 and SMS2 siRNA sequences have been previously described [56]. Human CERT siRNA was performed using siGENOME SMARTpool.

Cell culture and C. trachomatis propagation

HeLa 229 cells and L929 cells were obtained from ATCC and passaged as previously described [85]. LY-A and LY-A/hCERT cells were obtained from RIKEN BRC through the National Bio-Resource Project of MEXT, Japan and have been previously described [43], [66]. CHO cells were maintained at 37°C with 5% CO₂ in Ham's F12 containing 10% fetal bovine serum (FBS). C. trachomatis serovar L2 (LGV 434) and D (UW3-Cx) were propagated in L929 cells. C. trachomatis EBs were harvested from infected cells and purified using a Renografin step-gradient as described [86]. Unless otherwise indicated, experiments were performed with serovar L2.

Immunofluorescence studies

Cells were grown on glass coverslips in 24-well plates and infected at a multiplicity of infection (MOI) of approximately 1–5 with C. trachomatis as described in the text. For expression of tagged constructs, cells were transfected with the indicated plasmid constructs using Effectene (Invitrogen) for 18 hrs prior to infection, following manufacturer's instructions. Cells were fixed in 4% paraformaldehyde (PFA) or methanol and stained with the appropriate antibodies. For anti-GBF1, anti-14-3-3β, anti-MOMP, anti-HA, anti-ceramide, anti-calnexin, anti-p58, and anti-V5 staining, cells were permeabilized with 0.05% saponin/0.2% bovine serum albumun (BSA) in phosphate buffered saline (PBS) for 30 minutes and blocked with 0.2% BSA/PBS for 30 minutes. For anti-CERT staining, cells were permeabilized with 0.5% Triton X-100/PBS for 3 minutes and blocked with 1% BSA/PBS for 20 minutes. For phalloidin staining, cells were permeabilized with 0.2% Triton/PBS and blocked with 1% BSA/PBS for 1 hr. For anti-vimentin and anti-IncA staining, cells were post-fixed in methanol for 5 minutes and blocked in 2% BSA/PBS for 1 hr. Cells were incubated with the appropriate primary and fluorophore-tagged secondary antibodies for 1 hr each, followed by washing 3 times in PBS. Anti-FLAG staining was performed as previously described [69]. Coverslips were mounted in Vectashield mounting media containing DAPI (Vector Laboratories) to identify bacteria and host cell nuclei. Images were acquired under 20, 40 or 100× oil objective using a Nikon Eclipse TE2000-E fluorescence microscope with Simple PCI imaging software (Compix, Inc.) or acquired under 100× oil objective using Nikon TE2000-PFS spinning disk confocal inverted microscope with Nikon Elements software. For each set of experiments, the exposure time for each filter set for all images was identical. Images were processed with Adobe Photoshop CS, Nikon Elements, or Imaris Software.

Immunoblot analysis

Cells were lysed in lysis buffer (50 mM Tris HCl, pH 7.5, 150 mM NaCl, 0.1% SDS, 1% Nonidet P-40, 1% sodium deoxycholate, 1 mM sodium orthovanadate, 1 mM sodium fluoride, 1 mM okadaic acid, and Complete protease inhibitors (Roche Diagnostics)). After centrifugation at 20,800× g for 5 minutes to remove cell debris, the supernatants were transferred to fresh tubes. Proteins were separated on 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels and transferred to 0.45-µm pore Trans-blot nitrocellulose membranes (BioRad Laboratories). Membranes were blocked with 5% milk and probed with the indicated antibodies. Proteins were detected by enhanced chemiluminescence (ECL) (Amersham Biosciences) according to the manufacturer's protocol.

RNAi studies

HeLa cells grown in 6-well plates were transfected with specific or control siRNAs according to manufacturer's protocol. At 72 hrs post transfection, cells were infected with C. trachomatis for 1 hr and then incubated for an additional 24 hrs. Knockdown efficiency was measured by western blot analysis for endogenous protein, or in the case of SMS1 and SMS2, knockdown efficiency was determined using HeLa cells expressing C-terminally 3xFLAG-tagged SMS constructs as previously described [56]. Knockdown efficiency for SMS1 was also determined using BODIPY FL-Ceramide (see below).

Quantitation of primary inclusion formation, inclusion size, inclusion membrane disruption, and production of infectious progeny

HeLa cells were incubated with the indicated siRNA for 3 days and subsequently infected with C. trachomatis (MOI of 1–5) for 24 hrs. For drug treated samples, HeLa cells were infected with C. trachomatis for 1 hr and then incubated for an additional 23 hrs in the presence of the indicated drug. To quantify primary inclusion formation, cells were fixed and processed by immunofluorescence and processed as described below. To quantify production of infectious progeny, infected cells were scraped into media, lysed with sonication, and 5-fold serial dilutions were used to infect fresh HeLa monolayers that had been plated on coverslips for 24 hrs. For assessment of drug effects on bacterial viability and internalization, progeny was harvested from DMSO treated samples and left untreated or treated with the various inhibitors during the 1 hr infection of fresh HeLa monolayers. Cells were then washed free of the inhibitor and incubated for an additional 24 hrs. Inclusions were visualized by immunofluorescence with goat anti-MOMP antibodies and donkey anti-goat Alexa 594 or mouse anti-Chlamydia FITC conjugate. Images were analyzed using MetaMorph software (Molecular Devices, Sunnyvale, CA) from a minimum of 10 fields per sample. The relative inclusion forming units (IFUs) were expressed as a percentage of control siRNA or DMSO controls. The data is presented as mean ± standard error. To quantify the relative size of inclusions, approximately 200–350 cells were counted from 7–10 fields, the inclusion area analyzed by Metamorph using the mean area function, and the data is normalized to the percent area of control siRNA or DMSO-treated cells. To quantify inclusion disruption, approximately 350 cells were scored for apparent breaks in the inclusion membrane by staining with 14-3-3β and for concomitant release of bacteria into the cytosol. The data is normalized to percentage of similarly scored control or DMSO-treated cells.

Sphingomyelin accumulation assay

HeLa cells were infected with C. trachomatis for 21 hrs and then treated with the indicated drug for an additional 3 hrs. At 24 hpi, live cells were then incubated for 30 minutes in serum-free MEM containing 1 µM BODIPY FL-Ceramide (Molecular Probes) complexed with 0.034% defatted bovine serum albumin (dfBSA) in MEM at 4°C in the dark. Cells were washed 3 times with serum free MEM and back exchanged with MEM/0.34% dfBSA for 1 hr. Cells were immediately visualized by fluorescence microscopy under a 40× objective. To assess SM accumulation in the Golgi upon depletion of CERT or SMS1, BODIPY FL-Ceramide labeling was performed at 72 hrs post siRNA treatment and fluorescence intensity in the Golgi region of siRNA-treated samples was compared to control siRNA. To assess SM accumulation by the inclusion upon depletion and inhibition of GBF1 and BIGs, cells were treated with the appropriate siRNA for 72 hrs, infected with L2, and BODIPY FL-Ceramide labeling was performed at 24 hpi. For quantitation of SM accumulation in the inclusion, mean fluorescence intensity of the inclusion was determined by defining regions of interest within the inclusion and dividing by the mean fluorescence intensity of the nucleus, which remained constant with all drug treatments. At least 5 fields were examined, and values were obtained from a total of 10–15 inclusions.

Statistical analysis

Data represented the mean ± standard error of n experiments. Statistical analysis was performed using the software program Instat. The significance between groups was determined by ANOVA. A p value less than 0.05 was considered to be statistically significant.

Gene IDs

Arf1 (GeneID 375), GBF1 (GeneID 8729), BIG1 (GeneID 10565), BIG2 (GeneID 10564), CERT (GeneID 10087), VAP-A (GeneID 9218), SMS1 (GeneID 259230), and SMS2 (GeneID 166929).