A CRISPR screen identified host factors contributing to C. trachomatis L2-induced cytotoxicity

To identify host genes crucial for the execution of the defensive cell death response suppressed by CpoS, we conducted a CRISPR/Cas9 knockout screen for host gene deficiencies that can protect cells from death triggered by a C. trachomatis cpoS null mutant. It is important to note that any deficiencies preventing bacterial entry into cells or significantly perturbing bacterial growth could also be assumed to protect cells from cytotoxicity following exposure to the bacteria. Hence, to aid in pinpointing those genes that contribute to the defensive cell death response specifically, but not to Chlamydia-induced cytotoxicity or Chlamydia infection generally, we concurrently screened for host gene deficiencies protecting cells from the wild-type strain of C. trachomatis.

To this end, we transduced HeLa cells, a human cervical epithelial cell line, with a genome-wide knockout library (Fig 1A). We then infected the cells with wild-type C. trachomatis L2/434/Bu (CTL2, WT) or a variant carrying an insertional disruption/knockout of cpoS (CTL2-cpoS::cat, KO). Subsequently, we incubated the cells until a large proportion had died to enrich for cells resistant to killing, that is, cells resistant to either infection or infection-mediated cell death. From cultures infected with CTL2-cpoS::cat, surviving cells were collected at 30 hpi when about 90% of the cells had died (KO30h) (S1A Fig and S1A Data). Yet, from cultures infected with CTL2, surviving cells had to be collected at 60 hpi (WT60h), when only about 75% of the cells had perished, because a later harvesting would have led to significant cell loss during collection due to the fragility of cells containing large inclusions. Uninfected cells were collected alongside the infected cells (UI30h and UI60h). Upon concluding the selection experiments, sgRNA sequencing confirmed a good library representation, with over 99.9% of all library sgRNAs detected in all samples (S1B Fig and S1B Data), and no drastic shifts in population composition (S1C Fig and S1C Data).

Download:

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

Fig 1. A CRISPR screen identified host factors contributing to C. trachomatis L2-induced cytotoxicity.

(A) CRISPR screening procedure. The screen was conducted in two independent replicates (R1 + R2). (B) Venn diagram illustrating the selection of screening hits. (C) STRING interaction network displaying interactions among and between “general infection hits” (lilac nodes) and “mutant-specific hits” (green nodes), refined by manual grouping and annotations. Mixed groups were deemed to represent biological processes of general importance for infection (lilac group labels), while groups containing only “mutant-specific hits” were considered to have specific importance during infection with the CpoS-deficient mutant (green group labels). The data underlying this figure can be found in S4 Data.

https://doi.org/10.1371/journal.pbio.3003297.g001

We then identified sgRNAs that were positively selected, i.e., sgRNAs that showed significant enrichment in infected compared to uninfected samples and can thus be assumed to target genes promoting infection or infection-mediated cell killing (S2A–S2E Data). By setting a false discovery rate (FDR) of ≤0.05 and an average fold change (FC) of ≥1.25 as cutoffs, we found 54 genes with sgRNAs enriched in KO30h versus UI30h in both replicates of the screen (S1D–S1E Fig and S3A–S3B Data). Using the same criteria, sgRNAs for only three genes showed enrichment in WT60h versus UI60h (S1D–S1E Fig and S3A–S3B Data), aligning with the weaker selection achieved for this strain (S1A Fig and S1A Data). Hence, for the latter comparison, we decided to consider sgRNAs enriched at lower confidence as well, in order to obtain a more comprehensive overview of genes that may contribute more generally to Chlamydia-induced cytotoxicity or infection, thereby also facilitating their later exclusion from our analyses focusing on CpoS (as described below).

Consequently, we classified genes targeted by sgRNAs enriched (FC ≥ 1.25) in both KO30h versus UI30h (FDR ≤ 0.05, both replicates) and WT60h versus UI60h (FDR ≤ 0.2, at least one replicate) as “general infection hits”, resulting in 21 genes (Fig 1B and S4A Data). In addition, we identified “mutant-specific hits” as genes whose sgRNAs were enriched in KO30h versus UI30h (FDR ≤ 0.05, both replicates) but not WT60h versus UI60h (FDR ≤ 0.2, any replicate), yielding 33 genes (Fig 1B and S4B Data). While genes with sgRNAs enriched in WT60h versus UI60h but not KO30h versus UI30h could theoretically be considered as “wildtype-specific hits” (Fig 1B and S4C Data), they were not further considered in this study given that they can be expected to be less reliable.

To comprehend the biological roles of the genes targeted by positively selected sgRNAs, we conducted a functional enrichment analysis in g:Profiler [13] using the Reactome pathway database [14]. This revealed that the “general infection hits” were enriched for genes involved in proteoglycan biosynthesis and intra-Golgi trafficking, while the “mutant-specific hits” were enriched for genes involved in various vesicular transport processes and sphingolipid metabolism (S2A–S2B Fig and S4D Data). To gain an even deeper understanding of the major biological processes represented by our hits, we further utilized the STRING database [15] to display a protein interaction network, which we then further refined by manual grouping and annotation based on literature. This in-depth analysis highlighted concrete molecular pathways and complexes in proteoglycan synthesis, vesicular transport, and lipid synthesis to have roles in infection or infection-mediated cell death (Fig 1C and S4E Data). Below, these will be discussed and in part validated and followed up on.

Deficiencies in heparan sulfate proteoglycan synthesis provide general protection against C. trachomatis L2

Since the primary goal of our CRISPR screen was to identify host factors that have specific roles during infection with the CpoS-deficient mutant (CTL2-cpoS::cat), we aimed to exclude from further in-depth analysis any genes that were either part of the “general infection hits” or involved in the same molecular machineries or pathways. This most notably included genes having key or auxiliary functions in the synthesis of heparan sulfate proteoglycans (HSPGs), cell surface molecules already previously shown to facilitate host cell attachment and invasion by C. trachomatis L2 [16–19]. Indeed, this group of genes was especially dominant among our overall pool of hits and included genes encoding all enzymes required for the synthesis of the tetrasaccharide linker in HSPGs (XYLT2, B4GALT7, B3GALT6, B3GAT3 [20]), the enzymes involved in sugar chain elongation (EXTL3, EXT1, EXT2 [20]), proteins involved in glycan chain modifications (NDST1, FAM20B, UST [20,21]), proteins mediating the synthesis, transport, or removal of precursors or side products (SLC35B2, UGP2, CANT1 [20,22,23]), proteins maintaining a functional glycosylation or glycan-modifying machinery (e.g., TM9SF2, TMEM165, SLC10A7, SLC39A9, PTAR1, SPPL3, GRP89A, VMA21 [24–32]), and proteins often found co-enriched in other screens with those acting in HSPG biosynthesis (UNC50 and TM9SF3, e.g., [25,27,33]) (Fig 1C and S3A Fig and S4E Data).

Several other molecular machineries and pathways also included both “mutant-specific hits” and “general infection hits” (Fig 1C and S4E Data), and we thus considered them also more likely to have general roles during C. trachomatis L2 infection rather than specific roles related to CpoS. A closer look even suggested that many of these proteins may also have roles in HSPG synthesis. For example, the guided entry of tail-anchored proteins (GET) pathway, which mediates membrane insertion of tail-anchored proteins [34], and the conserved oligomeric Golgi (COG) and Golgi-associated retrograde protein (GARP) complexes, which have roles in membrane trafficking [35,36], may regulate the localization of proteins involved in HSPG synthesis or modification. Indeed, deficiencies in COG and GARP have previously been shown to cause a mislocalization and destabilization of proteins of the glycosylation machinery at the Golgi apparatus [37,38]. To validate this finding, we generated HeLa cells deficient for a key component of the COG complex, COG3 (S3B Fig), and indeed observed them to have HSPG surface levels reduced to a degree similar as seen in cells deficient for EXT1 (S3C Fig and S5A Data). Moreover, infection with wild-type C. trachomatis L2 (CTL2) was significantly reduced in COG3-deficient cells, while a much more modest reduction was observed for C. trachomatis serovar E (CTE) (S3D Fig and S5B Data), known to invade cells in a less HSPG-dependent manner [39].

In conclusion, this first mining of the data underscored the significance of HSPGs for CTL2 infection, which validated the performance of the screen and at the same time enabled us to focus our further attention on other processes promising specific insights into the function of CpoS.

Deficiencies in ceramide synthesis provide specific protection against the cpoS mutant

While we observed several host gene deficiencies to provide specific protection against CTL2-cpoS::cat in the CRISPR screen, they unexpectedly did not affect genes with known roles in immune signaling or regulated cell death. Instead, we found such impacting the Transport Protein Particle (TRAPP), a multi-subunit tethering complex involved in the vesicular trafficking between the endoplasmic reticulum (ER) and the Golgi apparatus [40], as well as the Rab GTPase RAB2A, also active in ER-Golgi trafficking [41], and GDI2 and TBC1D13, regulators of Rab protein function [41,42] (Fig 1C and S4E Data). Intriguingly, we also found specific protection to be conferred by deficiencies in the de novo synthesis pathway of ceramides, in particular such affecting the first enzymatic step, catalyzed by serine palmitoyltransferase (SPT; composed of SPTLC1, SPTLC2, and SPTSSA), or the second enzymatic step, catalyzed by 3-ketodihydrosphingosine reductase (KDSR) (Figs 1C and 2A and S4E Data). While all these screening hits deserve further investigations, in the present study, we decided to follow-up on the observations pertaining to ceramide synthesis, because they connected to our previous finding that CpoS, via its interactions with host Rab GTPases, mediates the vesicular transport of ceramide-derived sphingolipids to C. trachomatis L2 inclusions [11].

Download:

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

Fig 2. Deficiencies in ceramide synthesis provide specific protection against the cpoS mutant.

(A) Illustration of pathways in ceramide synthesis and metabolism. Enzymes whose deficiency protected cells from cpoS mutant-induced toxicity in the screen are highlighted in bold green, pharmacologic inhibitors used in this study in bold gray. Asterisks mark enzymes selected for genetic validation via individual CRISPR/Cas9-mediated gene knockout or knockdown. The names of the inhibitors are given in the main text. Enzymes: ACER/ASAH, ceramidase; CERK, ceramide kinase; CERS, ceramide synthase; CPP, ceramide-1-phosphate phosphatase; DEGS, dihydroceramide desaturase; GBA, glucosylceramidase; KDSR, 3-ketodihydrosphingosine reductase; LPP, lipid phosphate phosphatase; SGMS, sphingomyelin synthase; SGPP, sphingosine-1-phosphate phosphatase; SMPD, sphingomyelinase; SPHK, sphingosine kinase; SPL, sphingosine phosphate lyase; SPT, serine palmitoyl transferase; UGCG, UDP-glucose ceramide glycosyltransferase. (B–D) Selected inhibitors targeting sphingolipid metabolism protected cells against cpoS mutant-induced death. HeLa cells were treated with the indicated inhibitors (LCS, 125 µM; MYR, 1 µM; GW4869, 12.5 µM) or solvent only (DMSO), and were parallelly infected with the indicated strains (5 IFU/cell). Resorufin fluorescence (B) and nuclei count (C) at 25.5 hpi are displayed normalized to a CTL2-infected untreated control (mean ± SD, n = 4, two-way ANOVA with Dunnett’s post-hoc test; for each strain, indicated are significant differences compared to the DMSO control). (D) Representative images of DNA (Hoechst) staining (scale = 80 µm). (E) western blot analysis confirming the depletion (knockdown, KD) of SPTLC1 and absence (knockout, KO) of KDSR in the respective HeLa cell lines. (F–H) Depletion of SPTLC1 or deficiency in KDSR partially protected cells against cpoS mutant-induced death. The indicated HeLa cell lines were infected with the indicated strains (5 IFU/cell). Resorufin fluorescence (F) and nuclei count (G) at 24 hpi are displayed normalized to an uninfected control (mean ± SD, n = 3, one-way ANOVA with Dunnett’s post-hoc test; for each strain, indicated are significant differences compared to the parental (wild-type) cells). (H) Representative images of DNA (Hoechst) staining (scale = 80 µm). The data underlying this figure can be found in S6 Data.

https://doi.org/10.1371/journal.pbio.3003297.g002

To validate the screening hits in ceramide synthesis, we initially employed a pharmacologic approach leveraging the availability of well-established SPT inhibitors [L-cycloserine (LCS) and myriocin (MYR)] (Fig 2A). We additionally tested the effect of inhibitors targeting other enzymes in ceramide metabolism, including ceramide synthase [fumonisin B1 (FB1)], neutral sphingomyelinase [GW4869], acid sphingomyelinase [desipramine hydrochloride (DHY)], alkaline ceramidase [D-erythro-MAPP (DEM)], and ceramide kinase [NVP231] (Fig 2A). We treated HeLa cells with the respective inhibitors at non-toxic concentrations (S4A Fig and S6A Data) and simultaneously infected them with either CTL2 or CTL2-cpoS::cat. At 24 hpi, we quantified host cell viability based on the ability of viable cells to convert resazurin into fluorescent resorufin and based on the number of remaining cell nuclei. In agreement with the findings from the CRISPR screen, the SPT inhibitors LCS and MYR partially protected cells from death after infection with the cpoS mutant (S4B–S4C Fig and S6B–S6C Data). Additionally, our analyses indicated reduced cell death also in cultures treated with GW4869 (S4B–S4C Fig and S6B–S6C Data), although it should be noted that we observed some precipitation of GW4869, which may potentially have affected fluorescence recordings. The experiments with these three inhibitors (LCS, MYR, and GW4869) were subsequently repeated including as additional control also infections with a complemented mutant strain, i.e., a derivative of CTL2-cpoS::cat transformed with a plasmid driving expression of CpoS (CTL2-cpoS::cat/pCpoS). These additional analyses confirmed the inhibitor effects on cpoS mutant-induced cell death and demonstrated that complementation restored wild-type phenotypes (Fig 2B–2D and S6D–S6E Data).

To obtain genetic confirmation of our findings, we subsequently used CRISPR/Cas9 to generate a cell line depleted (knockdown, KD) for SPTLC1 and a cell line deficient (knockout, KO) for KDSR (Fig 2E). In agreement with the findings from the CRISPR screen and the experiments with the SPT inhibitors, these cell lines were partially protected against cpoS mutant-induced cell death (Fig 2F–2H and S6F–S6G Data).

In summary, these findings indicated that deficiencies in or the inhibition of ceramide-producing enzymatic reactions can protect cells from death following exposure to the cpoS mutant.

CpoS deficiency causes potentially cytotoxic alterations in sphingolipid metabolite levels

Ceramides and derived metabolites are not only precursors for complex sphingolipids, such as sphingomyelin, but function also as signaling molecules with key roles in cellular life/death decisions [43]. Therefore, it seemed plausible that a deficiency in cpoS may instigate host cell death by disturbing metabolite balance, for instance, by resulting in a cytotoxic accumulation of ceramides, an effect that impairments in the ceramide synthesis pathway may potentially mitigate. To investigate this possibility, we employed mass spectrometry to quantify the intracellular levels of 31 sphingolipid metabolites, encompassing a variety of common sphingoid bases, dihydroceramides, ceramides, dihydrosphingomyelins, sphingomyelins, hexosylceramides, and lactosylceramides. We focused on an early time point during infection, more specifically 14 hpi, which is just before premature host cell death becomes apparent in cultures infected with CTL2-cpoS::cat.

Interestingly, at this early time after infection, we did not detect any remarkable differences in the metabolite levels between uninfected cells and cells infected with either CTL2 or CTL2-cpoS::cat/pCpoS (Fig 3A and S5 Fig and S7A Data). Moreover, most metabolite levels remained similar also during infection with CTL2-cpoS::cat. Notably, ceramide levels were not significantly increased during infection with the CpoS-deficient mutant. However, we observed an increase in the levels of dihydroceramides and thus also in the dihydroceramide to ceramide ratio (Fig 3A and S5 Fig and S7A Data), an alteration previously reported to be cytotoxic [44].

Download:

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

Fig 3. CpoS deficiency causes potentially cytotoxic alterations in sphingolipid metabolite levels.

(A) Infection with the cpoS mutant enhanced the dihydroceramide (dhCer) to ceramide (Cer) ratio in infected cells. Cell extracts from HeLa cells infected with the indicated strains (10 IFU/cell) were prepared at 14 hpi, and the indicated lipids were quantified by LC–MS/MS. Shown are selected metabolite levels (expressed as “fmol/pmol total sphingolipids”) or ratios (mean ± SD, n = 3, one-way ANOVA with Tukey’s post-hoc test; if not specified otherwise, indicated are significant differences compared to uninfected cells). Full data are presented in S5 Fig. (B) Mass spectrometric analysis confirming SKI-II to increase the dhCer to Cer ratio. Cell extracts from uninfected and CTL2-infected (5 IFU/cell) HeLa cells, treated with SKI-II (6.25 µM) at 0 hpi, were prepared at 14 hpi, and the indicated lipids were quantified by LC–MS/MS and displayed as ratio (mean ± SD, n = 3, two-way ANOVA with Sidak’s post-hoc test; indicated are significant differences compared to the DMSO control). (C) SKI-II induced death in cells infected with CTL2. HeLa cells were treated with SKI-II (6.25 µM) or solvent only (DMSO) and were parallelly infected with the indicated strains (5 IFU/cell). Resorufin fluorescence at 25.5 hpi is displayed normalized to a CTL2-infected untreated control (mean ± SD, n = 4, two-way ANOVA with Sidak’s post-hoc test; for each strain, indicated are significant differences compared to the DMSO control). (D–E) SKI-II induced death specifically in infected but not uninfected cells. HeLa cells were treated with the indicated concentrations of SKI-II or with solvent only (DMSO) and were parallelly infected with CTL2 (5 IFU/cell) or left uninfected. (D) Resorufin fluorescence at 34 hpi is displayed normalized to the uninfected untreated control (mean ± SD, n = 3, two-way ANOVA with Sidak’s post-hoc test; if not specified otherwise, indicated are for each infection condition significant differences compared to the DMSO control). (E) Representative images of DNA (Hoechst) staining (scale = 80 µm). The data underlying this figure can be found in S7 Data.

https://doi.org/10.1371/journal.pbio.3003297.g003

To understand if the observed metabolic changes could contribute to the premature cell death seen during infection with CTL2-cpoS::cat, we employed the pharmacologic inhibitor SKI-II. This molecule inhibits sphingosine kinase (SPHK), but was recently shown to also inhibit dihydroceramide desaturase (DEGS), thereby increasing the dihydroceramide to ceramide ratio in treated cells (Fig 2A) [45]. Indeed, mass spectrometric analyses confirmed that SKI-II causes an accumulation of dihydroceramides in both uninfected cells and cells infected with CTL2 (Fig 3B and S7B Data). Intriguingly, the inhibitor also induced cell death in cultures infected with CTL2 or CTL2-cpoS::cat/pCpoS while not further augmenting death in cultures infected with CTL2-cpoS::cat (Fig 3C and S7C Data). This toxicity of SKI-II towards CTL2-infected cells was seen even at concentrations that were non-toxic to uninfected cells (Fig 3D–3E and S7D Data).

In summary, our findings indicated an increased dihydroceramide to ceramide ratio in cells infected with the cpoS mutant, a metabolic alteration that could potentially contribute to the observed premature host cell death.

The cpoS mutant displays an increased dependence on host de novo ceramide synthesis

Given the observed effects of the inhibitors on the viability of infected cells, we hypothesized that SKI-II would effectively eradicate CTL2 infection from infected cultures, while the inhibitors MYR, LCS, and GW4869, which can be expected to block formation of dihydroceramides and/or ceramides and partially blocked cpoS mutant-induced premature cell death (Fig 2B–2D and S4B–S4C Fig and S6B–S6E Data), may restore the intracellular growth of CTL2-cpoS::cat.

Hence, we treated HeLa cells with the inhibitors and concurrently infected them with GFP-expressing derivatives of CTL2 or CTL2-cpoS::cat, followed by microscopic detection of bacterial inclusions. In cultures infected with CTL2, SKI-II indeed caused a large drop in the numbers of both cells and inclusions, and the majority of cells surviving treatment did not harbor bacteria, in line with the inhibitor selectively killing infected but not uninfected cells (Fig 4A–4B and S8A Data). On the contrary, MYR, LCS, and GW4869 caused only minor reductions in the counts of both cells and CTL2 inclusions, as well as the size of the inclusions (Fig 4C–4D and S8B Data). In cultures infected with CTL2-cpoS::cat, the total number of inclusions detected in the absence of inhibitors was strongly reduced when compared to CTL2 infection, which was expected given the cell loss due to premature host cell death (Fig 4C–4D and S8B Data). Unexpectedly, LCS and MYR, while protecting cells infected with CTL2-cpoS::cat from death, not only failed to restore the growth of the bacteria but almost entirely eradicated the infection from treated cultures, with the majority of the remaining cells harboring no inclusions (Fig 4C–4D and S8B Data). In contrast, the presence of GW4869 did not modify the number of CTL2-cpoS::cat inclusions (Fig 4C–4D and S8B Data). A comparison of growth inhibition across a range of inhibitor concentrations, using GFP fluorescence derived from GFP-expressing strains as a measure for bacterial growth, confirmed that the cpoS mutant was considerably more dependent on SPT function (Fig 4E and S8C Data). Corresponding experiments with GW4869 suggested both strains respond similarly to this inhibitor (Fig 4E and S8C Data).

Download:

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

Fig 4. The cpoS mutant displays an increased dependence on host de novo ceramide synthesis.

(A–B) SKI-II treatment eradicated infected cells from cultures infected with CTL2. HeLa cells were treated with the indicated concentrations of SKI-II or with solvent only (DMSO), and parallelly infected with a GFP-expressing derivative of CTL2 (5 IFU/cell). Bacterial inclusions were detected microscopically at 34 hpi. (A) Inclusion numbers displayed relative to CTL2 inclusion count in the untreated control (mean ± SD, n = 2–3, one-way ANOVA with Dunnett’s post-hoc test; indicated are significant differences compared to the DMSO control). (B) Representative images showing inclusions (GFP) and DNA (Hoechst) staining (scale = 40 µm). (C–D) MYR and LCS, but not GW4869, reduced inclusion count after infection with the cpoS mutant. HeLa cells were treated with the indicated inhibitors (LCS, 125 µM; MYR, 1 µM; GW4869, 12.5 µM) or solvent only (DMSO), and parallelly infected with GFP-expressing derivatives of the indicated strains (1 IFU/cell). Bacterial inclusions were detected microscopically at 26.5 hpi. (C) Data displayed as percentage relative to inclusion count and average inclusion area detected for CTL2 in the absence of inhibitors (mean ± SD, n = 4, one-way ANOVA with Dunnett’s post-hoc test; for each strain, indicated are significant differences compared to the DMSO control). (D) Representative images showing inclusions (GFP) and DNA (Hoechst) staining (scale = 40 µm). (E) Differential susceptibility of cpoS mutant and wild-type bacteria to SPT inhibition. HeLa cells were treated with the indicated inhibitors and parallelly infected with GFP-expressing derivatives of the indicated strains (1 IFU/cell). GFP fluorescence measured at 25.5 hpi is displayed relative to fluorescence detected for the respective strain in the absence of inhibitors (mean ± SD, n = 3–4, 2-way ANOVA with Sidak’s post-hoc test; for each concentration, indicated are significant differences between the two strains). (F–G) Depletion of SPTLC1 or deficiency in KDSR partially restored inclusion count after infection with the cpoS mutant but increased the occurrence of inclusion-free cells. The indicated HeLa cells lines were infected with the indicated strains (5 IFU/cell). (F) Inclusion numbers detected at 24 hpi displayed normalized to the number of inclusions detected for CTL2 in the parental cells (mean ± SD, n = 3, two-way ANOVA with Dunnett’s post-hoc test; for each strain, indicated are significant differences compared to the parental cells). (G) Representative images showing inclusions (detected by immunofluorescence staining of the bacterial protein Slc1) and DNA (Hoechst) staining (scale = 40 µm). The data underlying this figure can be found in S8 Data.

https://doi.org/10.1371/journal.pbio.3003297.g004

Interestingly, in the cell lines depleted for SPTLC1 or deficient for KDSR, a slightly increased number of inclusions was detected after infection with CTL2-cpoS::cat (Fig 4F–4G and S8D Data). However, it was also apparent that a high proportion of the remaining cells in those cultures, i.e., the cells that were protected from the premature host cell death, were inclusion-free, in line with SPTLC1 depletion and KDSR deficiency having a growth-suppressing effect on CTL2-cpoS::cat (Fig 4G). It is likely that this growth suppression resulting from impairments in SPT or KDSR function contributes to the observed cytoprotective effects.

Previous research has shown that C. trachomatis L2 has a requirement for host-derived sphingomyelin [46] and can acquire the lipid in two ways: First, by redirecting sphingolipid-laden membrane vesicles to the inclusion [47]; second, by recruiting the host ceramide transfer protein CERT to inclusion-ER contact sites, facilitating non-vesicular acquisition of ceramides then converted to sphingomyelin at the inclusion [48]. We previously reported CERT recruitment, which others have shown to be mediated by the inclusion membrane protein IncD [49,50], to be unaffected by CpoS deficiency [11]. On the contrary, the vesicular transport of sphingomyelin to inclusions was compromised in the absence of CpoS [11]. Hence, we hypothesized that the cpoS mutant’s heightened dependence on de novo ceramide synthesis could be attributed to its reduced capacity to utilize the already available sphingolipid pool in the host cell, possibly resulting in a higher dependence on the CERT-mediated transport route.

Indeed, in the presence of golgicide A, an inhibitor known to block vesicular transport of Golgi-derived sphingomyelin to the inclusion [48], the dependence of CTL2 on host de novo ceramide synthesis intensified (S6A Fig and S8E Data). Moreover, the CRISPR screening data suggested a trend of CERT deficiency protecting cells from cpoS mutant-induced cell death, while interestingly enhancing the susceptibility of cells to CTL2-induced cytotoxicity (S6B Fig and S8F Data). Similar observations were made with a HeLa cell line depleted for CERT (S6C-S6D Fig and S8G Data). Moreover, despite being cytoprotective, CERT depletion did not significantly increase the number of inclusions detectable after infection with CTL2-cpoS::cat and many cells in cultures infected with this strain were inclusion-free, suggesting that CERT depletion, similar to SPTLC1 depletion and KDSR deficiency, may have a growth-suppressing rather than a growth-restoring effect on CTL2-cpoS::cat (S6E–S6F Fig and S8H Data).

In summary, our research indicated that the cytoprotective effects resulting from impairments in SPT or KDSR function in cultures infected with CTL2-cpoS::cat may not be explained by an effect on dihydroceramide levels alone, as CpoS deficiency heightens the bacteria’s reliance on host cellular ceramide synthesis for inclusion formation and/or growth.

Novel reporters for inclusion damage revealed instability of CpoS-deficient inclusions

We next sought to obtain a deeper mechanistic understanding of the ways by which sphingolipid deprivation can curb the growth of CTL2-cpoS::cat. It is important noting that CpoS deficiency has previously been proposed to cause a premature rupture of the inclusion [9]. Moreover, SPT inhibition has been suggested to destabilize CTL2 inclusions [51]. Hence, assuming that a potential instability of CpoS-deficient inclusions could stem from the mutant’s reduced ability to acquire sphingolipids via vesicular transport [11], we may further assume that SPT inhibition could aggravate and accelerate the destabilization of CpoS-deficient inclusions. This may then possibly modify the outcome for the host cell from death to infection clearance. However, live cell imaging suggested that CpoS-deficient inclusions rupture only at the onset of cell death [52], leaving unclear if CpoS deficiency indeed causes damage that later cumulates into rupture, possibly triggering cell death, or if described ruptures were merely a consequence of cell death.

Clarifying this question requires tools enabling the detection of early forms of inclusion membrane damage. Therefore, we developed fluorescence microscopic reporters that can visualize compromised inclusion membranes and bacteria released from damaged inclusions. These reporters employ the split-GFP principle, where the N-terminal part of GFP (GFP1–10) and the C-terminal part (GFP11) are expressed separately but can associate to form a functional fluorescent protein upon co-localization [53]. Previously, the system was successfully applied to detect secretion of C. trachomatis effectors or their insertion into the inclusion membrane [54]. Here, we used a modified approach. We generated derivatives of CTL2 and CTL2-cpoS::cat expressing distinct variants of the Inc protein IncB or the Chlamydia outer membrane protein OmpA tagged with both FLAG and GFP11 (Fig 5A). In IncB-GFP11_c, multiple copies of GFP11 were placed at IncB’s C-terminus, which is exposed to the host cell cytosol during infection. In IncB-GFP11_int, the GFP11 copies were placed between IncB’s two transmembrane domains, a region presumed to face the inclusion lumen, while in OmpA-GFP11_int, they were inserted into the predicted surface-exposed loop region between OmpA’s β-strands 11 and 12. Hence, during infection, the GFP11-tags of the constructs are expected to be displayed at the inclusion membrane, facing either towards the host cell cytosol (IncB-GFP11_c) or the inclusion lumen (IncB-GFP11_int), or at the surface of the bacteria (OmpA-GFP11_int). These constructs can thus serve as positive control for the split-GFP system (IncB-GFP11_c) or facilitate the detection of damaged inclusion membranes (IncB-GFP11_int) or bacteria exposed to the host cell cytosol (OmpA-GFP11_int) (Fig 5B).

Download:

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

Fig 5. Novel reporters for inclusion damage revealed instability of CpoS-deficient inclusions.

(A) Composition of the three GFP11-tagged constructs. (B) The principle of detecting inclusion damage using the split-GFP approach. (C) Fluorescence microscopic detection of inclusion damage during infection with CTL2-cpoS::cat. HeLa cells were transfected with a plasmid driving GFP1–10 expression, infected with the indicated strains (5 IFU/cell), and then fixed, stained (DNA (Hoechst) staining), and imaged at 26 hpi (scale = 20 µm). (D) Confocal fluorescence microscopic images displaying different grades of inclusion damage. GFP1–10-expressing HeLa cells were infected with the indicated strain (10 IFU/cell) and then fixed, stained (DNA (Hoechst) staining), and imaged at 24 hpi (scale = 10 µm). (E) FIB-SEM analysis validating inclusion damage at the ultrastructural level. GFP1–10-expressing HeLa cells were infected with the indicated strain and fixed at 24 hpi. A cell containing green-fluorescent bacteria was identified by fluorescence microscopy (left, scale = 10 µm, asterisks highlight inclusions) and subjected to FIB-SEM analysis (right, scale = 1 µm, one selected slice of the volume shown in S1 Movie, green arrows highlight bacteria in the host cell cytosol). (F) Quantitative analysis of inclusion damage in cells infected with the cpoS mutant. GFP10-expressing HeLa cells were infected with the indicated strains (10 IFU/cell) and then fixed, stained, and imaged at the indicated time points. The percentage of infected cells, surviving cells, and infected cells containing cytosolic bacteria was determined by manual image analysis (mean ± SD, n = 3, at least 100 cells per condition and replicate counted, 2-way ANOVA with Sidak’s post-hoc test; for each time point, indicated are significant differences compared to CTL2/pOmpA-GFP11_int). (G) Complementation of the cpoS mutant restored inclusion integrity. GFP10-expressing HeLa cells were co-infected with the indicated strains (5 IFU/cell of each strain) and then fixed, stained, and imaged at 24 hpi. The percentage of infected cells, surviving cells, and infected cells containing cytosolic bacteria was determined by manual image analysis (mean ± SD, n = 4, at least 200 cells per condition and replicate counted, one-way ANOVA with Tukey’s post-hoc test; if not indicated otherwise, indicated are significant differences compared to CTL2/pOmpA-GFP11_int + CTL2). Note that due to the fusogenic nature of inclusions, co-infection results in mixed inclusions containing both of the co-infecting strains. The data underlying this figure can be found in S9 Data.

https://doi.org/10.1371/journal.pbio.3003297.g005

In infected HeLa cells, all three constructs were expressed and localized correctly to either the inclusion membrane or the bacteria (S7A Fig). In cells expressing cytosolic GFP1–10, IncB-GFP11_c at the membrane of CTL2 inclusions became green fluorescent (S7B Fig). In contrast, no green fluorescence was observed in cells infected with CTL2 expressing IncB-GFP11_int or OmpA-GFP11_int (S7B Fig), consistent with the barrier function of the inclusion membrane. Critically, upon treating cells with membrane-permeabilizing agents, such as digitonin or high concentrations of DMSO, green-fluorescent spots emerged at IncB-GFP11_int-positive inclusion membranes (S7C Fig). Moreover, inclusions of OmpA-GFP11_int-expressing bacteria displayed green fluorescence post-treatment (S7C Fig). This fluorescence was faint and diffuse, implying a rupture of both the inclusion and bacterial membranes. With shorter treatment durations, individual green-fluorescent bacteria could still be discerned within some of the inclusions (S7D Fig).

Intriguingly, green-fluorescent spots, indicative of membrane damage, were also observed at CpoS-deficient IncB-GFP11_int-positive membranes in the absence of permeabilizing agents (Fig 5C). Moreover, a subset of cells infected with OmpA-GFP11_int-expressing CTL2-cpoS::cat harbored green-fluorescent bacteria (Fig 5C), often adjacent to seemingly intact non-fluorescent inclusions, likely representing early signs of inclusion damage (Fig 5D). Inclusion rupture, resulting in the majority of the bacteria exhibiting green fluorescence, was observed occasionally, but typically only in cells that appeared to be in the process of dying (Fig 5D). The release of CTL2-cpoS::cat into the host cell cytosol could further be confirmed at the ultrastructural level, when we selected cells with green-fluorescent bacteria for focused ion beam scanning electron microscopy (FIB-SEM), a technique that can record volumes of infected cells (Fig 5E). Moreover, a quantitative analysis by fluorescence microscopy revealed that at 14 hpi, coinciding with the onset of cell death in cultures infected with CTL2-cpoS::cat, and at 24 hpi, when extensive cell death was observed, about 12.7% and 31.4% of the infected cells, respectively, contained cytosolic bacteria (Fig 5F and S9A Data). Critically, this analysis also confirmed the near absence of cytosolic bacteria at these times during infection with CTL2. Furthermore, complementation of the cpoS mutant restored inclusion stability (Fig 5G and S9B Data).

In summary, our novel microscopic reporters, which enable the detection of early signs of inclusion damage and cytosolic Chlamydia, provided evidence that inclusions formed by CTL2 are usually stable and that CpoS indeed plays a role in preserving inclusion integrity, as its absence leads to subtle forms of damage preceding inclusion rupture and cell death.

Early destabilization of inclusions clears infection while keeping host cells alive

After having confirmed that CpoS-deficient inclusions are unstable, we sought to investigate whether SPT inhibition indeed exacerbates inclusion damage. We infected GFP1–10-expressing HeLa cells with OmpA-GFP11_int-expressing CTL2-cpoS::cat and parallelly treated them with MYR or LCS. In line with our previous findings, we observed robust protection from cell death and a substantial decrease in the proportion of cells containing inclusions at 24 hpi (Fig 6A–6B and S10A Data). Unexpectedly, when observed at this time, the percentage of infected cells containing cytosolic bacteria appeared unaffected by the presence of SPT inhibitors (Fig 6A and S10A Data). However, when we quantified inclusion damage at an earlier time point, at 18 hpi, we observed a strong inclusion-destabilizing effect due to SPT inhibition, with up to 70% of the treated infected cells harboring cytosolic bacteria (Fig 6A and S10A Data). Collectively, these findings confirmed that SPT inhibition aggravates the instability of CpoS-deficient inclusions, with a majority of the inclusions becoming destabilized early on, apparently leading to infection clearance rather than host cell death.

Download:

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

Fig 6. Early destabilization of inclusions clears infection while keeping host cells alive.

(A–B) Destabilizing effect of SPT inhibition on inclusions. GFP10-expressing HeLa cells were infected with the indicated strains (5 IFU/cell) and parallelly treated with MYR (10 µM) or LCS (125 µM) or left untreated (control). Cells were fixed, stained, and imaged at the indicated time points. (A) The percentage of infected cells, surviving cells, and infected cells containing cytosolic bacteria was determined by manual image analysis (mean ± SD, n = 3, at least 100 cells per condition and replicate counted, 2-way ANOVA with Sidak’s post-hoc test; for each time point, indicated are significant differences compared to the untreated control). (B) Representative images of whole cell (HCS) staining (scale = 100 µm). (C) Inclusion-destabilizing effects of SPTLC1 depletion or KDSR deficiency. The indicated cell lines were transfected with a GFP1–10-expressing plasmid and then infected with the indicated strains. Cells were fixed, stained, and imaged at the indicated time points. The percentage of infected cells, surviving cells, and infected cells containing cytosolic bacteria was determined by manual image analysis (mean ± SD, n = 3, at least 200 cells per condition and replicate counted, one-way ANOVA with Dunnett’s post-hoc test; for each time point (and strain), indicated are significant differences compared to the parental cells). (D–E) Treatment with SKI-II destabilizes CTL2 inclusions. GFP10-expressing HeLa cells were infected with the indicated strains (5 IFU/cell) and parallelly treated with SKI-II (10 µM) or solvent-only (DMSO). Cells were fixed at 24 hpi or 30 hpi, stained (Hoechst, DNA; Slc1, bacteria), and imaged. (D) The percentage of infected cells displaying inclusion damage (ruptured inclusions or early damage indicated by individual cytosolic bacteria) was determined by manual image analysis (mean ± SD, n = 3, at least 200 cells per condition and replicate counted, unpaired t test). (E) Representative images displaying cytosolic bacteria and inclusion rupture (scale = 10 µm). The data underlying this figure can be found in S10 Data.

https://doi.org/10.1371/journal.pbio.3003297.g006

When GFP1–10-expressing HeLa cells were infected with CpoS-proficient OmpA-GFP11_int-expressing CTL2 and parallelly treated with MYR or LCS, we observed inclusion destabilization only at later stages of infection. Indeed, at 24 hpi, cytosolic bacteria were undetectable (Fig 6A and S10A Data), although we observed an also previously reported [51] inclusion fusion defect in inhibitor-treated cells, as indicated by a high proportion of cells containing multiple inclusions (Fig 6B). Yet, at 30 hpi, cytosolic bacteria were detectable in a small fraction of cells treated with SPT inhibitors, which in cultures treated with MYR also coincided with a slight increase in host cell death (Fig 6A and S10A Data). Overall, this implies that CpoS-proficient inclusions exhibit greater resistance to the destabilizing effects of SPT inhibition. Moreover, destabilization at later stages seems to be linked to cell death, partially mirroring the phenotype of CpoS deficiency.

In the cell lines depleted for SPTLC1 or deficient for KDSR we made similar observations, including aggravated destabilization of CTL2-cpoS::cat inclusions at 18 hpi, as well as some inclusion destabilization and host cell death induction at 30 hpi with CTL2 (Fig 6C and S10B Data). Moreover, as DEGS inhibition by SKI-II can also be expected to curb sphingolipid supply, we were not surprised to see that this inhibitor also destabilized CTL2 inclusions (Fig 6D–6E and S10C Data), likely explaining its selective toxicity towards infected cells.

Taken together, these findings validate that impairments in the ceramide synthesis pathway cause inclusion destabilization and demonstrate that they in combination with CpoS deficiency, presumably disrupting both sphingolipid supply routes simultaneously, destabilize inclusions at an early infection stage, thereby clearing infection while maintaining the host cells alive.

Supplementation with sphingosine stabilizes CpoS-deficient inclusions

To confirm that the observed destabilization of CpoS-deficient inclusions resulted from diminished sphingolipid acquisition, we provided cells with an excess of sphingolipid metabolites. The addition of C6-D-erythro-ceramide (C6-Cer) to the growth medium proved highly toxic to both infected and uninfected cells (Fig 7A and S11A Data), aligning with the pro-death signaling function of ceramides [43]. Interestingly, C6-D-erytho-dihydroceramide (C6-dhCer) showed no significant toxicity in this assay, which, however, may be explained by the observation that such synthetic cell-permeable short-chain dihydroceramides are generally less toxic than their natural long-chain counterparts [55]. Notably, the addition of D-erythro-sphingosine (Sph) or D-erythro-dihydrosphingosine (dhSph) protected cells from cpoS mutant-induced cell death and restored the growth of the mutant to a significant extent (Fig 7A–7B and S11A–S11B Data). The addition of sphingosine also stabilized CpoS-deficient inclusions (Fig 7C–7D and S11C Data).

Download:

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

Fig 7. Supplementation with sphingosine stabilizes CpoS-deficient inclusions.

(A–B) Supplementation of growth media with sphingoid bases protected cells from cpoS mutant-induced death and restored bacterial growth. HeLa cells were infected with the indicated strains (4 IFU/cell), and parallelly treated with 5 µM of the indicated metabolites (dhSph, dihydrosphingosine; Sph, sphingosine; C6-dhCer, C6-dihydroceramide; C6-Cer, C6-ceramide; DMSO, solvent only) or left untreated (control). Nuclei count (A) at 25.5 hpi is displayed normalized to the uninfected untreated control, while inclusion count (B) is displayed normalized to the CTL2-infected untreated control (mean ± SD, n = 3, one-way ANOVA with Dunnett’s post-hoc test; for each infection condition, indicated are significant differences compared to the DMSO control). (C–D) Stabilizing effect of sphingosine on CpoS-deficient inclusions. GFP10-expressing HeLa cells were infected with the indicated strains (5 IFU/cell) and parallelly treated with 5 µM sphingosine (Sph) or left untreated (control). Cells were fixed, stained (HCS, cells; Slc1, bacteria), and imaged at 24 hpi. (C) The percentage of infected cells, surviving cells, and infected cells containing cytosolic bacteria was determined by manual image analysis (mean ± SD, n = 3, at least 100 cells per condition and replicate counted, 2-way ANOVA with Sidak’s post-hoc test; for each strain, indicated are significant differences compared to the untreated control). (D) Representative images (scale = 100 µm). The data underlying this figure can be found in S11 Data.

https://doi.org/10.1371/journal.pbio.3003297.g007

Collectively, the findings obtained in this study thus underscore CpoS’ role in preserving the intracellular hideout of C. trachomatis L2 by ensuring an adequate sphingolipid supply.

Cell culture

HeLa (ATCC CCL-2), Vero (ATCC CCL-81), and HEK293T (ATCC CRL-3216) cells were cultivated in Dulbecco’s Modified Eagle’s Medium (Gibco), supplemented with 10% heat-inactivated fetal bovine serum (Gibco). In selected experiments, including those involving an analysis of cell death via the resazurin assay, a DMEM medium devoid of phenol red was used to avoid interference with measurements of fluorescence. Cultures were maintained in a humidified incubator (37°C, 5% CO₂) and routinely tested for contamination with Mycoplasma spp. using a commercial PCR-based Mycoplasma detection assay (VenorGeM, Minerva).

Treatment with pharmacological inhibitors or metabolites

The following inhibitors targeting enzymes in ceramide synthesis, metabolism, or transport were used: fumonisin B1 (Cayman), NVP-231 (Cayman), L-cycloserine (Cayman), myriocin (Cayman), D-erythro-MAPP (Cayman), GW4869 (Cayman), SKI-II (BioVision), desipramine hydrochloride (BioVision), and golgicide A (Cayman). Additionally, we used the following sphingolipid metabolites: D-erythro-sphingosine (Merck), D-erythro-dihydrosphingosine (Merck), C6-D-erythro-dihydroceramide (Merck), C6-D-erythro-ceramide (Merck). Stock solutions of inhibitors and metabolites were prepared in DMSO, except for fumonisin B1 and D-erythro-MAPP, which were dissolved in ethanol. The inhibitors and metabolites were added to cell cultures at concentrations and for durations as specified in the respective figure legends.

Chlamydia strains and general infection procedure

Infection experiments were carried out with C. trachomatis L2/434/Bu (ATCC VR-902B, CTL2), a variant carrying an insertional disruption of cpoS (CTL2-cpoS::cat), and further genetic derivatives as listed in S1 Table. Two different procedures were used to prepare infection inocula, including crude preparations (used in infection experiments involving the split-GFP system or GFP-expressing bacteria) and highly pure, density-gradient purified EB preparations (used in all other experiments, including the screen). These procedures have been described in detail in a previous study [11]. Bacteria obtained through either of the two methods were resuspended in SPG (sucrose-phosphate-glutamate) buffer (75 g/l sucrose, 0.5 g/l KH₂PO₄, 1.2 g/l Na₂HPO₄, 0.72 g/l glutamic acid, pH 7.5) and stored at −80°C. Furthermore, the bacterial preparations were titered (as previously described [11]) and confirmed to be free from Mycoplasma contamination (as described above for cell lines). To conduct infections, cells were typically seeded in multi-well plates, followed by the addition of bacteria (number of inclusion forming units (IFUs)/cell as specified), centrifugation (1,500 × g, 30 min, 23°C), and incubation (37°C, 5% CO₂) for the indicated periods of time. The centrifugation step was omitted for infections conducted in MatTek dishes. Moreover, infections conducted in the context of the CRISPR screen were done in suspension, as described below.

CRISPR screen

Generation of cell lines with individual gene knockouts or knockdowns

To generate cell lines with individual gene knockouts or knockdowns, the sgRNAs listed in S2 Table were cloned into vector pLentiGuide-Puro (Addgene 52963). Lentiviral particles were harvested from supernatants of HEK293T cells that had been co-transfected with psPAX2 (Addgene 12260), pMD2.G (Addgene 12259), and the respective sgRNA-encoding derivative of pLentiGuide-Puro. After filtration (0.45 µm), virus-containing supernatants were used to transduce Cas9-expressing HeLa cells in the presence of 8–10 µg/ml polybrene (Merck). Cells were co-transduced with the lentiviral vectors encoding the distinct sgRNAs that target the same gene (S2 Table). Transduced cells were selected in presence of puromycin (8–16 µg/ml; Gibco), first added at 72 hrs post transduction. Cells deficient for EXT1 were cloned by limiting dilution. Cells deficient or depleted for other proteins were used as polyclonal cultures.

Confirmation of gene knockouts or knockdowns by western blot analysis

Protein extracts were prepared by cell lysis in boiling 1% SDS buffer, as previously described [11], separated by SDS PAGE (Mini-PROTEAN TGX 4–20% gels, Bio-Rad), and transferred onto nitrocellulose membranes (pore size of 0.2 µm, Bio-Rad). Membranes were blocked for 20 min with 3% bovine serum albumin (BSA) in Tris-buffered saline with Tween (TBST; 20 mM Tris (pH 7.5), 150 mM NaCl, 0.1% Tween 20) and then incubated overnight at 4°C with primary antibodies diluted in blocking buffer. The following primary antibodies were used: mouse-anti-β-actin (1:2000; Cell Signaling, 3700), rabbit-anti-COG3 (1:2000; Proteintech,11130-1-AP), rabbit-anti-EXT1 (1:1000; Novus, H00002131-D01P), rabbit-anti-KDSR (1:2000; Proteintech, 16228-1-AP), rabbit-anti-SPTLC1 (1:1000; Proteintech, 15376-1-AP), and rabbit-anti-CERT (1:2000; Proteintech, 15191-1-AP). After incubation, membranes were washed thrice with TBST, incubated for one hour with horse radish peroxidase (HRP)-conjugated secondary antibodies (anti-mouse: Thermo-Fisher-Scientific, A10551; anti-rabbit: Thermo-Fisher-Scientific, G-21234; 1:10,000 diluted in blocking buffer), and washed again thrice with TBST. Membranes were then incubated for 1 min with HRP substrate (SuperSignal West Pico PLUS, Thermo-Fisher-Scientific; or ECL Prime Western Blotting Detection Reagent, Cytiva) and chemiluminescent signals were recorded with an Amersham Imager 680RGB (GE Healthcare). Membranes were stripped for 15–30 min with Restore Plus western blot stripping buffer (Thermo-Fisher-Scientific) and blocked with 3% BSA in TBST before detection of additional targets. Based on the results, we considered the cell lines to be either target protein deficient (knockout, KO) when the protein was undetectable, or target protein depleted (knockdown, KD) when reduced protein levels were observed.

Quantification of surface-exposed HSPG by flow cytometry

For flow cytometric analyses, 2 × 10⁵ cells of each cell line to be analyzed were transferred to a 1.5 ml tube. The cells were washed once with blocking solution (2% BSA in DPBS), then incubated for 30 min with blocking solution and subsequently for 1 hr with primary antibodies (mouse-anti-heparan sulfate, AmsBio, 370255-S, 1:100 diluted in blocking solution). Afterwards, the cells were washed twice with blocking solution, incubated for 30 min with secondary antibodies (goat-anti-mouse-DyLight 650, Thermo-Fisher-Scientific, SA5-10153, 1:400 diluted in blocking solution), and washed again twice. All incubation steps described above were carried out on ice and cells were collected by centrifugation (9,000 × g, 5 min, 4°C) between the individual incubation and wash steps. Finally, the cells were resuspended in blocking solution containing propidium iodide (1 µg/ml, Thermo-Fisher-Scientific) prior to measurements at a BD Accuri C6 Plus flow cytometer (BD Biosciences). Data analysis was carried out using FlowJo software (version 10.10.0). Cell debris and doublets were excluded based on forward and side scatter characteristics. Moreover, dead cells were excluded based on propidium iodide fluorescence. Mean DyLight 650 fluorescence intensity was then quantified to assess HSPG surface expression levels.

Transient and stable expression of GFP1–10 in human cells

Plasmid pcDNA3.1-Cyto-GFP1–10 [54], enabling transient expression of cytosolic GFP1–10, was kindly provided by Kevin Hybiske (University of Washington). HeLa cells were transfected with this plasmid using jetPRIME transfection reagent (Polyplus) with a medium exchange after 2–3 hrs. To enable stable expression of GFP1–10 in HeLa cells, a lentiviral expression system was used. The sequence encoding GFP1–10 was PCR-amplified from pcDNA3.1-Cyto-GFP1–10 using the primers GFP1–10_F (5′-TATGTCGACATGGTTTCGAAAGGCGAGGA-3′) and GFP1–10_R (5′-GGCCTCGAGTTATTTCTCGTTTGGGTCTT-3′). The PCR product was then digested with EcoRI and BamHI and ligated into EcoRI/BamHI-digested lentiviral vector pLVX-Puro (Takara). Subsequently, HEK293T cells were co-transfected with the resulting vector pLVX-Puro-GFP-1–10 and the packaging and envelope plasmids using Lenti-X Packaging Single Shots (Takara). Supernatants containing lentiviral particles were collected at 48 and 72 hrs post-transfection, filtered (0.45 µm), and used to transduce HeLa cells in the presence of 4 µg/ml polybrene (Merck). At 24 hrs post-transduction, the medium was replaced with a medium containing 1 µg/ml puromycin (Gibco) to select for cells that had been stably transduced. Cell clones were generated by limiting dilution and screened for robust GFP1–10 expression. This was based on the fluorescence microscopic detection of GFP signals following infection with strain CTL2/pIncB-GFP11_c (described below).

Generation of GFP11-expressing Chlamydia strains

Vector pTL2-tetO-IncB-GFP11x7-FLAG [54], enabling inducible expression of IncB (CT232) tagged at its C-terminus with seven copies of GFP11 (RDHMVLHEYVNAAGIT) and the FLAG epitope (DYKDDDDK), i.e., construct IncB-GFP11_c, was a kind gift from Kevin Hybiske (University of Washington). Vector pTL2-tetO-CTL0050-GFP11x4-FLAG-CTL0050, enabling inducible expression of OmpA (CTL0050) with four copies of GFP11 (RDHMVLHEYVNAAGIT) and a FLAG epitope (DYKDDDDK) inserted into OmpA’s predicted surface-exposed loop region between the β-strands 11–12, i.e., construct OmpA-GFP11_int, and vector pTL2-tetO-IncB-GFP11x3-IncB-FLAG, enabling inducible expression of IncB (CTL0484) with a C-terminal FLAG epitope and three copies of GFP11 inserted into IncB’s predicted inclusion lumen-exposed sequence, i.e., construct IncB-GFP11_int, were generated as follows: Gene blocks encoding the modified OmpA and IncB genes (S1 Methods) were purchased (Thermo-Fisher Scientific). The genes were PCR-amplified (OmpA: 5′-TATAGGTATGCGGCCGCATGAA-3′, 5′-GCGGTAAGTGTCGACTTATTAG-3′; IncB: 5′-ATTATCGGCCGATGGT-3′, 5′-CGCGTGTCGACTTATTAC-3′), digested with NotI and SalI (OmpA) or EagI and SalI (IncB), and inserted into EagI/SalI-digested vector pTL2-tetO-IncB-GFP11x7-FLAG. The vectors encoding the three constructs were then transformed into CTL2 and CTL2-cpoS::cat using the CaCl₂ approach [86]. Transformed bacteria were selected in the presence of penicillin G (Merck) and plaque-purified [8]. The presence of the correct plasmid in the clones was verified by PCR (pTL2-F: 5′-TCGATTTTTGTGATGCTCGTCAG-3′, pTL2-R1: 5′-ATATGATCTCTACCTCCAGAGCC-3′) and analysis of PCR products by agarose gel electrophoresis. In experiments using these strains, the expression of GFP11-tagged IncB or OmpA was induced by the addition of 4 ng/ml anhydrotetracycline (Clontech) at 0–2 hpi.

Fluorometric quantification of host cell viability and bacterial growth

The ability of cells to convert non-fluorescent resazurin into fluorescent resorufin was used as an indicator of cell viability. In brief, resazurin (Merck; 0.15 mg/ml in DPBS) was added to the cells at a volumetric ratio of 1:5, followed by further incubation for 1.5–2.5 hrs (37°C, 5% CO₂). Resorufin fluorescence was then measured (excitation: 560 nm, emission: 590 nm, bandwidth: 10 nm) and normalized to the fluorescence detected in control cells to calculate cell viability as a percentage. To quantify intracellular bacterial growth fluorometrically, infections were conducted using rsGFP-expressing derivatives of CTL2 and CTL2-cpoS::cat (S1 Table). At indicated times, the cells were fixed with 4% formaldehyde for 20 min at room temperature. Immediately afterwards, GFP fluorescence was measured (excitation: 490 nm, emission: 510 nm, bandwidth: 10 nm) and normalized to the fluorescence detected in infected control cells to calculate bacterial growth as a percentage. All assays described above were conducted in 96-well plates, and fluorescence measurements were made using a Spark plate reader (Tecan).

Quantification of sphingolipids

For the quantification of sphingolipids, HeLa cells seeded in 6-well plates were left uninfected or were infected with CTL2, CTL2-cpoS::cat, or CTL2-cpoS::cat/pCpoS (10 IFU/cell). In experiments including SKI-II treatment, cells were left uninfected or were infected with CTL2 (5 IFU/cell) in the presence or absence of SKI-II (6.25 µM). At 14 hpi, cell extracts were prepared (from about 1 × 10⁶ cells per sample). In brief, cells were washed with ice-cold DPBS and lysed in ice-cold methanol (HPLC-grade, ≥99.9%, Merck). Lysates were then collected with a cell scraper and stored at −80°C until further processing. Sphingolipids were extracted using 1.5 ml methanol/chloroform (1:1, v:v), containing the internal standards d₇-dihydrosphingosine (d₇-dhSph), d₇-sphingosine (d₇-Sph), d₇-sphingosine 1-phosphate (d₇-S1P), 17:0 ceramide (d18:1/17:0), d₃₁-16:0 sphingomyelin (d18:1/16:0-d₃₁), 17:0 glucosyl(β) ceramide (d18:1/17:0) and 17:0 lactosyl(β) ceramide (d18:1/17:0) (all from Avanti Polar Lipids) [87]. Final extracts were subjected to sphingolipid quantification by liquid chromatography tandem‐mass spectrometry (LC–MS/MS) applying the multiple reaction monitoring (MRM) approach. Chromatographic separation was achieved on a 1290 Infinity II HPLC (Agilent Technologies) equipped with a Poroshell 120 EC-C8 column (3.0 × 150 mm, 2.7 µm; Agilent Technologies) guarded by a pre-column (3.0 × 5 mm, 2.7 µm) of identical material. MS/MS analyses were carried out using a 6495C triple-quadrupole mass spectrometer (Agilent Technologies) operating in the positive electrospray ionization mode (ESI+). Chromatographic conditions and settings of the ESI source and MS/MS detector have been published elsewhere [88]. The mass transitions used for analysis of sphingolipid subspecies are given in S3 Table. Peak areas of Cer, dhCer, SM, dhSM, HexCer and LacCer subspecies, as determined with MassHunter software (Agilent Technologies), were normalized to those of their internal standards followed by external calibration. DhSph, Sph, and S1P were directly quantified via their deuterated internal standards. Quantification was performed with MassHunter Software (Agilent Technologies).

Fluorescence microscopy

For standard fluorescence microscopy evaluations (not coupled to electron microscopy), cells were cultivated in 96-well plates or on glass coverslips placed into 24-well plates and fixed for 20 min with 4% formaldehyde in DPBS. Subsequently, the cells were permeabilized for 15 min with 0.2% Triton X-100 in DPBS, incubated for 20 min in blocking solution (2–3% BSA in DPBS), and then for 1 hr in blocking solution containing primary antibodies (rabbit-anti-Slc1 (gift from Raphael Valdivia [89]; 1:1000) or rabbit-anti-FLAG (Merck, F7425; 1:1000)). After washing thrice with DPBS, the cells were incubated for 1 hr in blocking solution containing the DNA dye Hoechst 33342 (Invitrogen; 2−10 µg/ml), Alexa Fluor 555-conjugated secondary antibodies (Invitrogen; 1:250–1:1000), and the whole cell dye HCS CellMask Deep Red Stain (Invitrogen; 5 ng/ml). Subsequently, the cells were washed thrice with DPBS again. Cells grown and stained on glass coverslips were transferred to microscope slides and embedded in ProLong Glass Antifade Mountant (Invitrogen). All incubation steps were conducted at room temperature. In some experiments, staining with the whole cell dye or with antibodies was omitted. Images were captured using various microscopy systems, including an epifluorescence microscope (Zeiss Axio Imager.Z2), a confocal microscope (Leica SP8), and a high-content imaging platform (Molecular Devices ImageXpress Micro Confocal system). Automatic processing of images recorded at the ImageXpress system and detection of chlamydial inclusions and host cell nuclei were performed with CellProfiler (version 4.0.7) [90]. To quantify the extent of inclusion damage, images were analyzed manually. A minimum of 100 cells were analyzed per condition and biological replicate. Cells were classified as “uninfected” or “infected” based on the presence of inclusions. Within the group of “infected” cells, we further differentiated between cells that “contained cytosolic bacteria” and those that “did not contain cytosolic bacteria”.

FIB-SEM imaging

GFP1–10-expressing HeLa cells were seeded in gridded 35 mm glass bottom MatTek dishes (MatTek, P35G-1.5-14-C-GRD) and subsequently transfected with pcDNA3.1-Cyto-GFP1–10, infected with the indicated strains, and treated with anhydrotetracycline, using above-described standard procedures. At 23.5 hpi, Hoechst 33342 (10 µg/ml) was added to the dishes, and at 24 hpi, the cells were fixed with 4% paraformaldehyde in 0.1 M PHEM buffer. Fluorescence microscopic analysis was conducted using a Leica DMi8 inverted Thunder widefield microscope, equipped with a LEICA DFC 9000 GTC camera and operated by Leica Application Suite X 3.8.2. Subsequently, the cells were postfixed with 2.5% glutaraldehyde in 0.1 M PHEM buffer. Samples were processed for FIB-SEM using a PELCO Biowave Pro+ microwave tissue processor (Ted Pella) following a previously described procedure [91], with minor modifications: calcium was not used during fixation and the contrasting step with lead aspartate was omitted to reduce the risk of overstaining. The samples were embedded in a thin disc of Durcupan resin, mounted on an SEM stub with epoxy and silver glue, and coated with 5 nm platinum. Before imaging, the area of interest was coated with a 1 µm thick protective layer of platinum. The volume was imaged using a FEI Scios DualBeam system, with the acquisition automated using the Auto Slice & View software (ver. 4.1.1.1582). Images were captured with the electron beam operating at 2 kV and 0.2 nA, detected with the T1 in-lens BSE detector. The voxel size (X × Y × Z) of the volume was 9.6 × 9.6 × 20 nm. The volume was further registered and processed in ImageJ using the plugins “Linear Stack Alignment with SIFT” and “MultiStackReg”. After registration, the volume was converted into an MRC file and the header was modified to include pixel size information.

Statistical analysis

Statistical analyses, other than those described for the CRISPR screen, were conducted in GraphPad Prism 8. The following statistical tests were used when indicated in the figure legends: unpaired t test or one-way or two-way ANOVA with the indicated multiple comparison post-hoc tests. The following significance levels were considered: * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001; ns, not significant. Functional enrichment analysis was conducted  in g:Profiler [13] using the Reactome pathway database [14]. Enrichment maps were drawn using Cytoscape [92,93].