Chlamydia exploits filopodial capture and a macropinocytosis-like pathway for host cell entry

Pathogens hijack host endocytic pathways to force their own entry into eukaryotic target cells. Many bacteria either exploit receptor-mediated zippering or inject virulence proteins directly to trigger membrane reorganisation and cytoskeletal rearrangements. By contrast, extracellular C. trachomatis elementary bodies (EBs) apparently employ facets of both the zipper and trigger mechanisms and are only ~400 nm in diameter. Our cryo-electron tomography of C. trachomatis entry revealed an unexpectedly diverse array of host structures in association with invading EBs, suggesting internalisation may progress by multiple, potentially redundant routes or several sequential events within a single pathway. Here we performed quantitative analysis of actin organisation at chlamydial entry foci, highlighting filopodial capture and phagocytic cups as dominant and conserved morphological structures early during internalisation. We applied inhibitor-based screening and employed reporters to systematically assay and visualise the spatio-temporal contribution of diverse endocytic signalling mediators to C. trachomatis entry. In addition to the recognised roles of the Rac1 GTPase and its associated nucleation-promoting factor (NPF) WAVE, our data revealed an additional unrecognised pathway sharing key hallmarks of macropinocytosis: i) amiloride sensitivity, ii) fluid-phase uptake, iii) recruitment and activity of the NPF N-WASP, and iv) the localised generation of phosphoinositide-3-phosphate (PI3P) species. Given their central role in macropinocytosis and affinity for PI3P, we assessed the role of SNX-PX-BAR family proteins. Strikingly, SNX9 was specifically and transiently enriched at C. trachomatis entry foci. SNX9-/- cells exhibited a 20% defect in EB entry, which was enhanced to 60% when the cells were infected without sedimentation-induced EB adhesion, consistent with a defect in initial EB-host interaction. Correspondingly, filopodial capture of C. trachomatis EBs was specifically attenuated in SNX9-/- cells, implicating SNX9 as a central host mediator of filopodial capture early during chlamydial entry. Our findings identify an unanticipated complexity of signalling underpinning cell entry by this major human pathogen, and suggest intriguing parallels with viral entry mechanisms.


Introduction
An essential early event in the lifecycle of many human and animal pathogens is entry into non-phagocytic host epithelial cells. Viruses, bacteria and parasites all engage with host cell surfaces prior to inducing the reorganisation of the plasma membrane and underlying cytoskeleton to promote their internalisation. Invasive bacteria like Salmonella and Listeria species are typically > 1 μm in diameter and promote their internalisation either by injecting virulence effector proteins that subvert host signalling to reversibly induce cytoskeletal reorganisation, or through surface ligand mimicry hijack receptor-mediated endocytosis, respectively [1]. By contrast, the infectious extracellular form of Chlamydia trachomatis, termed the elementary body (EB), is much smaller in diameter (~400 nm) than its archetypal Gram-negative cousins. Nevertheless, EBs must also induce actin-dependent internalisation into non-phagocytic cells, a pivotal step in the lifecycle of this obligate intracellular bacterium [2].
How C. trachomatis promotes cell entry is incompletely understood, but it is often considered as an example of the trigger mechanism epitomised by the enteroinvasive bacterium Salmonella typhimurium. Salmonella employs a type III secretion system (T3SS) to inject multiple, semi-redundant effectors into host cells that coordinate the reorganisation of the host actin cytoskeleton. Two effectors reversibly stimulate the cellular Rho-family GTPases Cdc42 and Rac1, two effectors bind to actin directly to modulate filament dynamics, and a further effector acts as a phosphoinositol phosphatase mimic, modulating membrane plasticity and co-stimulating the Rho GTPases [3][4][5][6][7]. C. trachomatis EBs also exploit a T3SS and deliver effectors into the host cell that reversibly stimulate Rac1. Although the mechanism remains incomplete, a major factor is the T3SS effector translocated actin recruiting phosphoprotein (TARP) that nucleates polymerisation directly by binding to actin, and indirectly upon tyrosine phosphorylation by acting as a scaffold for Rac1 guanine nucleotide exchange factors [8][9][10][11]. A second effector post-translationally modifies the GTPase itself, possibly to subsequently downregulate signaling [12]. A number of host receptors have also been linked to cell entry by different chlamydial species. For instance, C. pneumoniae Pmp21 binds epidermal growth factor receptor (EGFR) to induce EB entry by receptor-mediated endocytosis [13], an event more reminiscent of the zipper mechanism exemplified by Listeria [14]. The role of receptors in C. trachomatis entry is however less clear, as none are essential [15].
Although Rac1 stimulation is sufficient for the formation of lamellipodia, this signalling activity cannot exclusively account for the complex actin ruffles, pedestal-like structures and filopodia present at EB entry sites [2,16]. This view was further reinforced by our cryo-electron tomography of early interactions between C. trachomatis EBs and cultured cells [17], when EBs were captured in association with phagocytic cups, trapped by actin-rich filopodia and present within membrane ruffles reminiscent of macropinosomes. These combined data support a view that multiple or redundant entry pathways are likely to operate in parallel. As expected, chlamydial entry thus shares many similarities with other bacterial entry pathways. However, there are also striking parallels with viral entry mechanisms. The small size of EBs, their association with filopodia [2], entry-associated phosphorylation and signaling [16,18,19], and also the requirement for protein disulphide isomerase-associated reduction [20] and promiscuous receptor interactions [21][22][23] are all factors common to viral entry mechanisms [24][25][26][27][28]. In this study we have quantified the cytoskeletal rearrangements and membrane reorganisation at C. trachomatis entry foci, and systematically investigated the underlying signalling pathways, initially by employing inhibitor screens in a manner analogous to studies of viral entry [29,30].

Filopodial association is an early event during C. trachomatis entry
Although our cryo-electron tomography revealed an unexpected diversity of cellular structures at C. trachomatis entry sites [17], this technique did not enable the visualisation of sufficient events to statistically distinguish whether they represent sequential assemblies or distinct pathways. Consequently, we exploited confocal microscopy to categorise a significantly larger number of bacterial entry foci, initially by observing F-actin recruitment during C. trachomatis infection of human retinal pigment epithelial (RPE1) cells. Telomerase-immortalised RPE1 cells are widely applied to study endocytic pathways [31,32], and are permissive to C. trachomatis infection [33].
Distinct F-actin structures could be defined using light microscopy that correlated with those observed by cryo-electron tomography [17]. Differential fluorescence staining was used to discriminate extracellular and intracellular bacteria (S1 Fig, see Materials and Methods). From 10 minutes post-infection, C. trachomatis LGV2 EBs were captured in association with filopodia, F-actin cup, tail or ring-like assemblies (Fig 1A and additional examples in S2 Fig).
To examine the progression of these events over time, cells were additionally fixed 30 and 120 minutes post-infection. The most striking phenotype was the difference in the number of EBs in association with filopodia, which decreased from 33 ± 3% at 10 minutes to 13 ± 6% after 120 minutes (Fig 1B). This revealed that filopodial association is a significant early event during EB entry into cultured RPE1 cells.
To determine whether filopodial association is a conserved early event, F-actin structures were next equivalently quantified following the infection of HeLa cells with C. trachomatis LGV2 (S3A Fig). The frequency of filopodial association at 10 minutes post infection was even higher in these cells, at 51 ± 4% EBs, decreasing significantly by 120 minutes post infection to 23 ± 3% (S3B Fig). Comparable EB-associated F-actin structures formed when RPE1 cells were infected with C. trachomatis serovar D (S4A and S4B Fig). Under these experimental infection conditions, RPE1 and HeLa cells have similar overall infection kinetics, and comparable proportions of EBs in association with F-actin structures at 10, 30 and 120 minutes postinfection (S5 Fig). These quantitative data illustrate a spatio-temporal conservation in the Factin superstructures formed during C. trachomatis entry, which occur independently of bacterial serovar and host cell type. Furthermore, they highlight filopodial association as a conserved and quantitatively significant early event during initial EB-host cell interaction. These initial data thus support a model that Chlamydia might utilise similar mechanisms to viruses trachomatis EBs during early interaction with RPE1 cells. Cultured RPE1 cells were infected with C. trachomatis LGV2 for 30 minutes prior to fixation with 1% paraformaldehyde. Fixed cells were stained with an anti-Chlamydia primary antibody and an Alexa Fluor 488-conjugated secondary antibody. Cells were permeabilised with 0.05% Triton X-100 (v/v) and the bacteria stained using the same anti-Chlamydia primary antibody and an Alexa Fluor 633-conjugated secondary antibody. Intracellular bacteria were labelled with only Alexa Fluor 633 (dark blue; 'intracellular and extracellular' panel), extracellular bacteria were labelled with Alexa Fluor 488 and Alexa Fluor 633 (green + blue, cyan; 'extracellular' and 'intracellular and extracellular' panels). F-actin was stained with rhodamine-phalloidin. White arrowheads show typical examples of indicated classes of F-actin structure. Images are maximum projections of confocal xy sections. Scale bars, 5 μm. Right hand panels show diagrammatic representations of the defined classes of F-actin structures visualised by fluorescence microscopy of cultured RPE1 cells infected with C. trachomatis. (B) Quantification of F-actin structures associated with extracellular C. trachomatis EBs during infection of RPE1 cells. Cells were infected for 10, 30 and 120 min prior to fixation with 1% PFA. Fixed cells were stained as above and the association of EBs with the defined F-actin classes quantified. ! 200 bacteria were assessed at each time point and the percentage of EBs in association with each class of structure was calculated, expressed as the average ±SD (n = 3). Ã P<0.05, ÃÃ P<0.01 using one-way ANOVA followed by a Tukey's post hoc test.

Signalling diversity during C. trachomatis cell entry
Filopodia are induced by the activation of the GTPase Cdc42 and the associated actin nucleation-promoting factor (NPF) N-WASP [34,35], although additional efficiency factors are also required including IRTKS-and Ena/VASP-family proteins [36,37]. However, previous studies have shown Rac1-WAVE-Arp2/3 signalling is clearly required for C. trachomatis invasion, whereas Cdc42 is not [38,39]. To evaluate whether additional host factors are required for filopodial association and EB uptake, cells were infected in the presence of small molecule inhibitors targetting a broad panel of cellular factors known to influence actin dynamics and endocytic processes. Initially, to facilitate high-throughput screening, cultured RPE1 cells were infected with C. trachomatis LGV2 for two hours in the presence of three different concentrations of each inhibitor, where the mid concentration was the established IC 50 or effective concentration. Inhibitors were then removed by washing, and the infection allowed to proceed until 24 hours post-infection (hpi), when the infected cells were fixed and the number of inclusions present in comparison to control, mock-treated cells enumerated. Based on the initial results, a second smaller-scale inhibitor screen was conducted to directly assess the effects of a subset of the inhibitors, which influenced inclusion formation in the first screen, on bacterial entry by applying differential fluorescence staining to discriminate extracellular and intracellular bacteria directly. This allowed effects on entry and nascent inclusion formation to be distinguished. For the initial screening at 24 hpi, effects were arbitrarily considered as significant when the mid-concentration of an inhibitor reduced the number of inclusion-containing cells to 75% of the control (this threshold is represented by a dotted line on plots in S6 Fig). Consistent with the major roles for the Sos/Abi1/Eps8 and Vav2 guanine nucleotide exchange factors (GEFs) in the activation of the small GTPase Rac1 during C. trachomatis entry [11], the EHop inhibitor that specifically prevents Vav2-mediated Rac1 activation [40] induced a clear and significant dose-dependent decrease in the number of inclusion-containing cells, whereas NSC that alternatively targets Rac1-specific GEFs TrioN and Tiam1 had a lesser effect [41] (compare EHop and NSC in S6 Fig). Unexpectedly, inhibitors targeting the GTPase Cdc42 resulted in significant reduction in inclusion number (S6 Fig, ML141 and Casin). Conversely, the inhibition of RhoA and Arf GTPases did not reduce inclusion formation to 75% of the control, our significance criteria (S6 Fig, RhoA and Arf6), in agreement with previous studies [38], although Arf6 activity has been implicated in the internalisation of the related C. caviae [42]. Finally, all three dynamin inhibitors tested decreased inclusion formation significantly and dose-dependently, with dynasore and MiTMAB preventing inclusion formation at the highest concentration (S6 Fig, Dynasore, MiTMAB and OcTMAB), in agreement with the established role of dynamin in mediating lipid transport early during inclusion biogenesis [43].
Despite the well-recognised requirement for actin reorganisation, equivalent assays performed using small molecule inhibitors that target actin dynamics exhibited contrasting effects on inclusion formation in RPE1 cells. Cytochalasin D treatment induced a dose-dependent decrease in inclusion formation, with an 88 ± 4% reduction at mid concentration (S6 Fig,  actin polymerisation cytoD), whereas cells treated with latrunculin B only exhibited a 16 ± 7% decrease at the highest concentration tested (S6 Fig, actin polymerisation latB), despite the fact that both cytochalasin and latrunculin are classical inhibitors of actin polymerisation, albeit via distinct modes of action [44,45]. Jasplakinolide, which stabilises Factin, prevented inclusion formation at the mid concentration (S6 Fig, actin stabilisation) [46]. Consistent with a role for Rho GTPase signalling, CK636 and CK548 that target the Arp2/3 actin nucleation complex and inhibit actin polymerisation either by preventing the Arp2 and Arp3 subunits of the complex entering their active conformation or by binding to the hydrophobic core of Arp3 [47], induced a dose-dependent reduction in inclusion formation, reflected by a 37 ± 8% and 48 ± 9% decreases at the mid concentration, respectively (S6 Fig, CK636 and CK548). In addition to Arp2/3-directed nucleation of branched F-actin networks, unbranched filament nucleation by formins apparently also contributes, as the formin inhibitor SMIF induced a 76 ± 6% reduction in inclusion formation at the mid concentration, although this was excluded from further analysis as it did not reduce inclusion formation 75% of the control (S6 Fig, SMIF). These data demonstrate that our assays using C. trachomatis LGV2 and cultured RPE1 cells specifically recapitulate previous findings implicating the specific stimulation of Rac1 via a subfamily of cellular GEFs [11].
Given the early role of filopodia (Fig 1), we investigated the potential contribution of the NPF N-WASP, which stimulates filopodia formation by activating the Arp2/3 complex via Cdc42-dependent and -independent pathways [48,49]. We exploited wiskostatin that specifically inhibits N-WASP activity by stabilising the auto-inhibited conformation [50]. Intriguingly, N-WASP inhibition resulted in 49 ± 12% and 92 ± 5% decreases in inclusion formation at the mid-IC 50 and high-concentrations, respectively (S6 Fig, wiskostatin). These data implicate N-WASP as an apparently dominant mediator of early inclusion formation. Correspondingly, the macropinocytosis inhibitor EIPA induced a dose-dependent decrease in inclusion formation (S6 Fig, macropinocytosis), with~50% of the remaining inclusions containing RBs which were morphologically abnormal (S7 Fig, EIPA). By contrast, treatment with filipin and cholesterol oxidase, which target lipid raft and caveolae-mediated endocytosis [51,52], did not significantly affect the number of inclusions formed in comparison to the mock-treated controls (S6 Fig, filipin & cholesterol oxidase), or the morphology of the inclusions or RBs (S7 Fig, cholesterol oxidase). Tip, which targets myosin VI and consequently clathrin-mediated endocytosis (CME) [53], reduced inclusion formation by 15 ± 6% at the mid concentration, although the inhibitor itself was significantly cytotoxic when applied at higher concentrations (S6 Fig, Tip and S8 Fig, Tip). These initial inhibition experiments using inclusion formation 24 hpi as a phenotypic read-out in RPE1 cells, both confirmed recognised mediators of bacterial entry and early inclusion biogenesis (Rac1, dynamin, Arp2/3-dependent actin polymerisation) and implicated previously unrecognised factors (Cdc42, N-WASP, macropinocytosis).
However, these data alone are insufficient to distinguish a role for these factors in C. trachomatis entry rather than in the subsequent development and trafficking of early bacteria-containing vacuoles. Based on the initial results (S6 Fig), a more restricted inhibitor screen was performed using EHop (Rac1), Casin (Cdc42), MiTMAB (dynamin), wiskostatin (N-WASP), CK636 (Arp2/3), cytochalasin D and latrunculin B (actin polymerisation), EIPA (macropinocytosis) and Rhosin (RhoA), to assess the effect of these inhibitors directly on bacterial entry using differential 'inside-outside' immunofluorescence staining (S1 Fig, see Materials and Methods). In agreement with their effects on inclusion formation (S6 Fig and S7 Fig), inhibition of Rac1, N-WASP, macropinocytosis and the Arp2/3 complex each reduced bacterial entry in a dose-dependent manner (Fig 2 and S9 Fig). Indeed, the N-WASP inhibitor wiskostatin could block C. trachomatis entry, without inducing substantial cell cytotoxicity (Fig 2,  N-WASP and S8 Fig, Wiskostatin). Consistent with the dominant role of Rac1 signalling [2,16], entry was not as profoundly suppressed by the Cdc42 inhibitor (compare Rac1 and Cdc42 in Fig 2), in agreement with previous studies [38,39]. Conversely, RhoA and dynamin GTPase inhibitors had a limited effect on C. trachomatis LGV2 entry (Fig 2 RhoA and  dynamin), in agreement with a role for dynamin in mediating lipid transport post-entry, early during inclusion biogenesis [43]. Unexpectedly, neither cytochalasin D nor latrunculin B decreased bacterial entry as anticipated (Fig 2, cytochalasin D and S9 Fig, latrunculin B), although cytochalasin D treatment significantly inhibited the entry of Salmonella typhimurium into RPE1 cells, reducing internalisation to 18 ± 9% of mock-treated controls under equivalent conditions at IC 50 (S10A Fig), when F-actin organisation is clearly disrupted (S10B Fig). Intriguingly, treatment with cytochalasin D or latrunculin B reproducibly stalled a population of EBs in association with patches of F-actin at the cell periphery (immunofluorescence panels in Fig 2, cytochalasin D and S9 Fig, latrunculin B).
To extend the inhibitor screening, it was important to establish whether the implicated signal transducers were specifically recruited to C. trachomatis entry foci, and to gain insights into the spatio-temporal dynamics of this process. Initially, RPE1 cells transiently expressing Cdc42-GFP, Rac1-GFP, RhoA-GFP, Arf1-GFP or Arf6-GFP were infected with C. trachomatis LGV2 and observed over a two hour timecourse. In agreement with the inhibitor-based screen, Rac1-GFP and Cdc42-GFP were transiently recruited from 10 minutes post infection (Fig 3, Rac1-GFP and Cdc42-GFP, and fluorescence intensity plots through the central plane of the EB), whereas Arf1-GFP and RhoA-GFP were not enriched at entry foci (Fig 3, Arf1-GFP and RhoA-GFP). At this timepoint, only 7-10% of bacteria are intracellular (S5 Fig), thus the recruitment of Rac1 and Cdc42 (8.6 ±0.9% and 6.7 ±1.3% respectively) reflects the transient association of these with nearly every invasion-competent EB at 10 minutes post-infection. Arf6-GFP was observed with lower frequency in membrane ruffles adjacent to some EBs (Fig 3, Arf6-GFP), consistent with its role in membrane ruffling and Rac1 trafficking [54,55]. No enrichment of endogenous clathrin, caveolin-1 or flotillin-1 was evident at entry foci under equivalent conditions (S11 Fig, clathrin, caveolin-1 and flotillin-1), consistent with the lack of inhibition of inclusion formation by Tip, cholesterol oxidase and filipin (S6 Fig). These data support the view that Cdc42 and Rac1 GTPases are dominant early host mediators of C. trachomatis entry into RPE1 cells.
Since the NPFs N-WASP and WAVE bridge the activated Rho-family GTPases Cdc42 and Rac1 to the Arp2/3 complex, respectively, we next examined the recruitment of N-WASP and the WAVE complex following C. trachomatis LGV2 infection of RPE1 cells. The WAVE subunit PIR121-GFP was recruited to entry foci (Fig 4, PIR121), consistent with the reported requirement for Rac1 and the WAVE complex during C. trachomatis infection [38,39]. N-WASP-GFP was recruited more frequently to C. trachomatis entry foci than PIR121-GFP from as early as 10 minutes post infection (Fig 4, N-WASP), in agreement with our findings that N-WASP inhibition can also block bacterial entry (Fig 2, N-WASP). There was no apparent specific preference of the GTPase-GFP or NFP-GFP for specific EB-associated F-actin structures.
Given that N-WASP is transiently recruited to C. trachomatis entry foci and N-WASP inhibition has dose-dependent effects on bacterial entry, we validated the role of N-WASP in entry using knockout mouse embryonic fibroblasts (N-WASP -/-). In these cells, chlamydial adhesion was reduced by >70% compared to isogenic wild-type control cells (S12A Fig), while C. trachomatis entry was significantly reduced by 54 ± 17% (S12B and S12C Fig). These data confirm the importance of N-WASP during entry processes, yet imply the role of N-WASP is Cultured RPE1 cells were pre-treated for 5 min with the indicated concentration of inhibitor, followed by infection with Chlamydia trachomatis LGV2 in the presence of inhibitor. At 2 h postinfection, cells were washed then fixed with 1% PFA and stained with an anti-Chlamydia primary antibody and an AlexaFluor 488-conjugated secondary antibody (green). Cells were permeablised and stained again using anti-Chlamydia primary antibody and an AlexaFluor 633-conjugated secondary antibody (blue). Intracellular bacteria were labelled with only one fluorophore (AlexaFluor 633). Images are maximum projections of confocal xy sections. Scale bars, 10 μm. The entry efficiency in the presence of inhibitors relative to mock treated control cells was quantified. ! 300 bacteria were assessed at each time point and the percentage entry efficiency of EBs is expressed as the average ±SD (n = 3). P-values obtained from Student's unpaired two-tailed t-test, Ã P<0.05, ÃÃ P<0.01, ÃÃÃ P<0.001, 'ns' not significant.
https://doi.org/10.1371/journal.ppat.1007051.g002 GTPase localisation during C. trachomatis entry into cultured RPE1 cells. Cultured RPE1 cells were transfected with Rac1-GFP, Cdc42-GFP, RhoA-GFP, Arf1-GFP, Arf6-GFP or GFP alone and infected with C. trachomatis LGV2 18h later. Cells were infected for up to 120 min prior to fixation. Fixed cells were stained with an anti-Chlamydia primary antibody and an Alexa Fluor 633-conjugated secondary antibody and rhodamine phalloidin (red) to visualise F-actin. Scale bars, 5 μm. Representative images show maximum projections of confocal xy sections of the focal plane containing the EB and the nearest neighbour z-sections (±0.2 μm). Line intensity plots taken through the centre of the EB. ! 100 cell-associated bacteria were assessed for more complex than simply the requirement in entry alone. The fact that upstream adhesion of C. trachomatis is also affected may be indicative of a role for N-WASP and consequently the actin cytoskeleton in stablising chlamydial adhesion, which may subsequently impact bacterial entry efficiency. Alternatively, the absence of N-WASP in the knockout cells may disturb Factin organization and influence normal C. trachomatis adhesion processes indirectly. Nevertheless, these data clearly support the view that N-WASP plays a significant yet previously uncharacterised role in the early interactions between C. trachomatis and the host cell.

A macropinocytosis-like pathway underlying C. trachomatis entry
The requirement for Rac1, Cdc42, N-WASP, the Arp2/3 complex and macropinocytosis-associated sodium-proton exchangers inhibited by EIPA, and the dynamic recruitment of Rac1, Cdc42, WAVE and N-WASP to entry foci, together with the association of EBs with phagocytic cups, suggested that C. trachomatis entry shared many similarities to growth factor and virus-induced macropinocytosis [56]. An additional hallmark of macropinocytosis is the associated activity of phosphoinositol-3-kinase (PI3K), and the sequential generation of the phosphoinositide signalling intermediates phosphatidylinositol-3,4,5-trisphosphate [PI(3,4,5)P3] and phosphatidylinositol-3-phosphate (PI3P) at the plasma membrane. Consequently, we next investigated PI3K activity and used reporters to determine the localisation of 3-phosphoinositide species during chlamydial internalisation. RPE1 cells were infected with C. trachomatis LGV2 in the presence of the pan-PI3K inhibitors wortmannin and LY294002 [57], and the effect on bacterial entry assessed using the fluorescence 'inside-out' assay. Neither wortmannin nor LY294002 significantly inhibited C. trachomatis entry (Fig 5A), in agreement with previous data in HeLa cells showing wortmannin-insensitive chlamydial internalization [2]. However, Akt-PH-GFP, which reports PI(3,4,5)P3 when expressed in cultured cells, was recruited to EBs 10 and 30 minutes post infection (Fig 5B, Akt-PH-GFP). Strikingly, the PI3P reporter PX-p40-GFP not only accumulated at entry sites, but also persisted around motile early vacuoles encapsulating EBs (Fig 5B, PX-p40-GFP and S1 Movie). These data suggest that C. trachomatis enters via a macropinocytosis-like route, with features subtly distinct from the canonical pathway. This would not be without precedent, as a wortmannin-and LY294002-insensitive pathway generates PI(3,4,5)P3 during Salmonella entry [58].
Despite this apparent difference in the requirement for PI3K, 40 ± 26% of C. trachomatis EBs co-localised with the fluid-phase marker 10,000 MW dextran after 30 minutes, whereas no equivalent association was observed with the CME marker transferrin (Fig 5C, and Fig 5D, compare dextran and transferrin), consistent with the lack of clathrin recruitment (S11 Fig,  clathrin). Fluid-phase and transferrin uptake were not significantly enhanced during infection (Fig 5C, compare NI and I panels), although dextran uptake was inhibited by EIPA treatment. Using this approach, it was not possible to determine whether dextran-labelled EB-containing macropinosomes were derived from a defined class of F-actin-rich surface structure. However, when cells were infected with C. trachomatis in the presence of 70,000 MW dextran, by comparison there did not appear to be equivalent coincidence between 70,000 MW dextran and C. trachomatis EBs (S13 Fig), suggesting a limit in the capacity of the uptake vesicle. This is in accordance with the tight encapsulation of C. trachomatis EBs observed by cryo-electron tomography at this time point [17]. Taken together, these data reveal that C. trachomatis LGV2 entry can proceed via atypical macropinocytosis-like events that share hallmarks of the archetypal cellular pathway, yet also exhibit key differences.

Sorting nexin 9 mediates filopodial capture during Chlamydia-induced macropinocytosis
Macropinosomes are derived from membrane ruffles and protrusions folding back and fusing with the plasma membrane to form large vesicles [59]. Although the signalling underlying cellular macropinocytosis is incompletely understood Rac1, WAVE, Cdc42 and N-WASP, along with other downstream effectors such as PAK1, are implicated in this process [56]. Phosphoinositide signalling is also central, with PI(3,4,5)P3 present in phagocytic cups being rapidly dephosphorylated to PI3P on the nascent macropinosome [60,61]. The sorting nexin (SNX) protein family is implicated in membrane trafficking, cargo sorting and endocytosis, and is characterised by a phosphoinositide-binding phox (PX) domain. The PX domain confers phosphoinositide binding specificity, with some PX domains binding PI(3)P [62]. The SNX-PX-BAR subfamily contain a BAR domain, and of these SNX1, SNX5, SNX9, SNX18 and SNX33 are implicated in macropinosome formation [63][64][65].
Of the twelve SNX-PX-BAR-Myc proteins analysed, SNX9-Myc was observed most frequently at entry sites (Fig 6A, SNX9 and Fig 6B, SNX9-Myc). However, since all twelve ectopically-expressed SNX-PX-BAR-Myc derivatives were also present at entry sites at low frequency using our applied scoring criteria (Fig 6A), we next verified the localisation of endogenous SNX9 with an anti-SNX9 polyclonal antibody by indirect immunofluorescence. This antibody recognised endogenous SNX9 in accordance with literature (S14A Fig anti-SNX9; [66,67]), and within F-actin pedestals generated following infection of RPE1 cells with enteropathogenic Escherichia coli (EPEC) (S14B Fig; [68]). Endogenous SNX9 was frequently present at sites of EB-host cell interaction in RPE1 cells infected with C. trachomatis (Fig 6B,  SNX9), in agreement with the recruitment of SNX9-Myc (Fig 6B, SNX9-Myc). In addition to these studies using fixed cells, RPE1 cells expressing GFP-SNX9 were also observed by live imaging. In trans expression of GFP-SNX9 expression generated two distinct phenotypes, dependent on the level of expression. High expression induced extensive membrane tubulation, whereas lower expression generated a punctate distribution reminiscent of the endogenous protein (compare left and right GFP-SNX9 panels in S14A Fig GFP-SNX9; [66,69]). Consequently, cells exhibiting membrane tubulation were excluded from further analysis. Functionality of GFP-SNX9 in low-expressing cells was additionally confirmed by verifying localisation within F-actin pedestals generated following infection with EPEC (S14C Fig). Cultured RPE1 cells expressing low levels of GFP-SNX9 were infected with C. trachomatis LGV2. GFP-SNX9 was clearly recruited to cell-associated EBs (Fig 6C; S2 Movie). Live imaging showed that EBs adhere in close proximity to peripheral membrane ruffles enriched in GFP-SNX9, following which intense GFP-SNX9 puncta form that directly overlap with the EB, dissipating~400 seconds later (S2 Movie). These combined approaches reveal the specific and transient recruitment of the SNX-PX-BAR family protein SNX9 to C. trachomatis entry sites.

SNX9 is central to filopodial association early during C. trachomatis entry
Given that SNX9 is transiently recruited to C. trachomatis entry foci, we next investigated the possible roles for SNX9 during infection. The effect of SNX9 knockdown on C. trachomatis entry was therefore assessed. Cultured RPE1 cells were treated with pooled SNX9 siRNA for 72 hours or equivalently with control non-targeting scrambled siRNA, prior to infection with C. trachomatis LGV2. Infection was allowed to proceed for two hours prior to fixation and quantification of intracellular bacteria by inside-outside differential fluorescence staining (Fig 7A). Knockdown of SNX9 resulted in a 20 ±8% reduction in the number of internalised bacteria  To establish whether the effect of SNX9 knockdown is limited to invasion, cells were treated with siRNA for 48 hours, prior to infection with C. trachomatis LGV2. Infection was allowed to proceed for 24 hours prior to fixation and quantification of the number of inclusion containing cells (Fig 7B, Fig 7C). No significant difference was observed in the overall number of infected cells (Fig 7C). However, RPE1 cells treated with SNX9 siRNA had a significant number of smaller-sized inclusions (Fig 7D). While the spread of inclusion diameters in control non-targeting siRNA (10.4 ± 3.4 μm) and SNX9 siRNA (8.1 ± 3.8 μm) treated cells were similar, the minimum and maximum inclusion diameters in the SNX9 siRNA treated cells were smaller than those in control non-targeting siRNA treated cells. This suggests that SNX9 influences inclusion growth, or the phenotype could be a secondary effect arising from the reduction in C. trachomatis entry (Fig 7A).
However, as these knockdown experiments inevitably represent a mixed population, we exploited human adherent HAP1 cells that recapitulate the chlamydial infection cycle [70] to generate a SNX9 knockout cell line (SNX9 -/-). While bacterial adhesion remained unchanged in comparison to isogenic controls when SNX9 -/cells were infected C. trachomatis LGV2 (Fig  8A), EB entry was significantly reduced by 25 ± 9% (Fig 8B). As the membrane-scission protein dynamin interacts with SNX9 [71,72], we examined the structure of early C. trachomatiscontaining vacuoles at 3 hours post infection of HAP1 wild type and SNX9 -/by cryo-EM tomography [17]. The vacuoles formed in SNX9 -/cells were morphologically indistinguishable from those formed in the wild type background (Fig 8C), suggesting SNX9 does not participate in membrane scission or early vacuole formation subsequent to the entry defect. Furthermore, there was no apparent change in appearance or formation of the inclusion (Fig  8D), the generation of infectious progeny (Fig 8E), the number of infected cells (Fig 8F) or the inclusion diameter in the SNX9 -/cells (Fig 8G). These phenotypic data using knockout cells thus demonstrate that the functional role of SNX9 is limited to the early phase of EB entry.
SNX9 has an established role in the reorganisation of the actin cytoskeleton, as it interacts directly with multiple proteins that regulate filament dynamics, including the Arp2/3 complex, N-WASP, Cdc42 and RhoA [66,73,74], and is recruited to F-actin-rich structures during fluidphase endocytosis [73]. To examine whether SNX9 might contribute to cytoskeletal reorganisation during C. trachomatis entry, F-actin morphology during bacterial entry into wild type and SNX9 -/-HAP1 cells was compared using the assays developed previously to assess RPE1 and HeLa cells (Fig 1 and S3 Fig). Adherent wild type and SNX9 -/cells appear phenotypically similar prior to infection, and both populations exhibited protrusions, filopodia and lamellipodia; however, SNX9 -/cells exhibited a small yet significant reduction in the numbers of scrambled control siRNA. 48 h later cells were infected with C. trachomatis LGV2. At 24 h post-infection the cells were fixed. Cells were stained with anti-SNX9 and anti-Chlamydia primary antibodies, followed by Alexa Fluor 488 (green) and 546 (red)-conjugated secondary antibodies respectively and DRAQ5 (blue). Scale bars, 10 μm. (C) Fewer cells are infected in SNX9 knockdown cells. Cells were infected with C. trachomatis LGV2 for 24 h prior to fixation. Cells were stained with anti-SNX9 and anti-Chlamydia primary antibodies, followed by Alexa Fluor 488 (green) and 546 (red)conjugated secondary antibodies respectively. 10 fields of view SNX9 siRNA or scrambled control siRNA treated cells were analysed for the number of inclusion containing cells. The average percentage of inclusion containing cells in the SNX9 siRNA treated cells relative to scrambled control was calculated ±SD (n = 3). P-values obtained from Student's unpaired two-tailed t-test, Ã P<0.05. (D) Box and whiskers plot comparing inclusion diameter (μm) in SNX9 siRNA treated cells. Cultured RPE1 cells treated with SNX9 siRNA or scrambled control siRNA were infected with C. trachomatis LGV2 for 24 h prior to fixation and quantification of inclusion diameter. The whiskers of each plot extend to the minimum and maximum inclusion diameters of the population, whereas the box indicates the interquartile range 'IQ' which consists of the middle 50% of the data. Box and whiskers contains pooled data from three independent experiments each analysing ! 30 infected cells. P-value obtained from Student's unpaired two-tailed ttest, ÃÃ P<0.01.
https://doi.org/10.1371/journal.ppat.1007051.g007 filopodia per cell in non-infected cells (Fig 9A, compare WT and SNX9 -/in 'NI' cells). During infection, there was a modest increase in overall filopodia numbers per cell in the WT background, while this was not observed in the SNX9 -/cells, and the reduction in filopodia numbers per cell compared to the WT was greatly reduced (Fig 9A, compare WT and SNX9 -/cells during infection 'I'). This is likely to be an oversimplification however, due to the limitations of confocal microscopy to analyse filopodia on the cell surface, and additional information about the cell surface structures during infection would likely be visible using higher resolution microscopy techniques. When both populations were infected with C. trachomatis LGV2 and the resulting actin structures compared 30 minutes post infection, the number of bacteria in contact with filopodia decreased by > 50% in the SNX9 -/background, whereas the number of EBs in association with cup, tail, or ring-like structures were not significantly different (Fig  9B, sedimentation). As filopodial capture is an early event, we investigated whether this effect occurred only when EBs were artificially sedimented onto the HAP1 cells, by comparing the phenotypes when cells were infected by co-incubation with a population of cells from which bacteria are actively egressing, previously developed for electron tomography [17]. This natural infection captures all the progressive processes that occur during cell entry, rather than experimentally imposing synchronous contact between EBs and the host cell plasma membrane by sedimentation. When wild type and SNX9 -/-HAP1 cells were infected in this more natural way, fewer EBs were again observed in contact with filopodia in the SNX9 -/relative to the WT cells (Fig 9B, egress). Moreover, entry was strikingly reduced by 59 ± 8% in the SNX9 -/cells (Fig 9C), compared to the 25 ± 9% reduction following sedimentation (Fig 8B). Taken together, these data demonstrate that SNX9 is required for early filopodia formation during C. trachomatis entry, an effect that can be partially compensated by artificial sedimentation of the bacteria into contact with the host cells.

Discussion
In this study, we aimed to further understand the pathways exploited by C. trachomatis to enter eukaryotic host cells. We demonstrated that during internalisation, distinct F-actin structures interact with cell-associated EBs, including an initial contact between EBs and filopodia, and bacterial association with more complex F-actin cup, tail and ruffle structures. In addition to the recognised roles of the Rac1 GTPase and its associated NPF WAVE, we have revealed an additional unrecognised pathway sharing key hallmarks of macropinocytosis: i) amiloride sensitivity, ii) fluid-phase uptake, iii) recruitment and activity of the NPF N-WASP, iv) the localised generation of phosphoinositide-3-phosphate (PI3P) species and v) involvement of the SNX-PX-BAR protein SNX9 in early filopodial capture.
Macropinocytosis-like pathways underlie cell entry by diverse viruses, including HIV, Vaccinia, Adenovirus, Ebola, Influenza A and the Herpes Simplex Virus [29,[75][76][77][78][79]. Indeed, Chlamydiae were originally described as a virus due to the absolute requirement of the host cell for survival, yet in many ways Chlamydiae remain atypical bacteria, particularly as EBs do not exclusively utilise either zipper or trigger mechanisms of entry [1], and for bacteria they are relatively small. In this study we extend viral parallels to C. trachomatis entry. We have shown that filopodia associations with EBs occur at a high frequency during invasion in a process reminiscent of the filopodia-mediated capture of viruses, including Vesicular Stomatitis Virus, Murine Leukaemia Virus and HIV [24,80]. In this process, viral particles associate with and either 'surf' along or induce filopodial retraction towards the cell body where they are internalised at endocytic 'hot spots' [24]. In the context of our data which show a decrease in EB association with filopodia over time, filopodial capture is likely the initial point of contact for EBs and may explain the distinct orientation of the T3SS towards the cell during infection [17], likely facilitating interaction of EBs with the surface of microvilli-dense cells of mucosal membranes they preferentially infect. Indeed, many viruses follow this filopodia-mediated capture with membrane ruffling and uptake via macropinocytosis, furthering similarities between EB and viral uptake. The involvement of a macropinocytosis-like pathway in C. trachomatis entry is supported by several observations: requirement for Rho GTPases, sensitivity to EIPA,

Fig 9. Filopodial capture is significantly attenuated in SNX9 -/cells. (A) Non-infected HAP1 WT and SNX9 -/cells have similar Factin organisation and filopodia number. Cultured HAP1
WT and SNX9 -/cells were fixed and stained with rhodamine-phalloidin to visualise F-actin. Scale bars, 10 μm. The number of filopodia per cell was quantified in maximum projections of confocal xy sections with the criteria that the cells quantified has at least three 'edges' not adherent to a neighbouring cell, and that filopodial projections were !1 μm in length. 15 cells were assessed for WT and SNX9 -/cells and the results are expressed as the average ±SD (n = 3). Pvalues obtained from Student's unpaired two-tailed t-test, Ã P<0.05. (B) The numbers of EBs in association with filopodia during entry is decreased in the HAP1 SNX9 -/cells after infection by sedimentation or egress. Cultured HAP1 WT and SNX9 -/cells were infected with C. trachomatis LGV2 for 2h. At 2 h post-infection cells fixed with 1% PFA and stained with an anti-Chlamydia primary antibody and an Alexa Fluor 488-conjugated secondary antibody. Cells were permeablised and stained again using anti-Chlamydia primary antibody and an Alexa Fluor 633-conjugated secondary antibody. Intracellular bacteria were labelled with only one fluorophore (Alexa Fluor 633). Extracellular EBs in association with filopodia, ring, cup/tail, or other F-actin structures were scored, and expressed as the relative percentage of EBs in association with each structure ±SD for !150 cell-associated bacteria (n = 3). Pvalues obtained from Student's unpaired two-tailed t-test, Ã P<0.05, 'ns' not significant. Data from one representative experiment is shown for egress as infection is asynchronous therefore it is inappropriate to cross-compare data from individual experiment for statistical analysis. Scale bars, 10 μm. A macropinocytosis-like pathway of Chlamydia entry coincidence of EBs and a fluid phase marker during entry, dynamin independence and the association of EBs with PI (3,4,5)P3 and PI3P during early entry [56]. We have also shown a requirement for N-WASP activity during chlamydial entry, suggestive of N-WASP involvement in F-actin rearrangements during internalisation, including filopodia formation [35]. Furthermore, while Cdc42 does not have a major role in C. trachomatis entry ( [38]; this study), there is a limited, although significant reduction in entry upon Cdc42 inhibition. Co-activation of Rac1 and Cdc42 is central to membrane ruffling [81,82] and macropinocytosis [59,83], so the coordinated inter-dependent action of Rac1 and Cdc42 would not be unexpected in membrane ruffling and macropinocytosis-like uptake of C. trachomatis, especially as a similar pathway underlies C. caviae entry [84]. Indeed, our data suggest that there was no specific preference of Rac1-, Cdc42, N-WASP or PIR121 (WAVE) GFP reporter fusions for particular EBassociated F-actin superstructures. Further careful investigation of endogenous regulators is now required, as this either reflects a limitation induced by the expression of the reporters or implies functional redundancy.
Further similarities remain to be assessed, such as whether PAK1/2, myosin II, and CtBP1 are involved in entry [85][86][87][88], yet Chlamydia-specific adaptations would not be unexpected. Typical macropinocytic ruffles are much larger and differences in signalling requirements during macropinocytic uptake of smaller-sized cargoes have been described [87,89,90]. Indeed, despite the clear association of EBs with 3-phosphoinositide species, paradoxically C. trachomatis entry is largely PI3K-independent. However, this is not without precedent, as for example the association of Salmonella with PI(3,4,5)P3-rich membrane ruffles is PI3K-independent, and mediated by as yet unknown mechanisms involving the inositol phosphatase mimic SopB [58].
As expected, cytochalasin D and latrunculin B significantly inhibited inclusion formation and Salmonella entry in our hands. However, phalloidin staining revealed F-actin patches accumulated beneath cell-associated EBs, which were reported as internalised by the 'insideoutside' assay despite the presence of the inhibitors. Both the F-actin accumulation and this apparent internalisation were unexpected. The latter may merely reflect a difference in antibody accessibility under these conditions, but the accumulation of F-actin may indicate that effector-mediated actin polymerisation might locally limit the effect of actin depolymerising agents. This is not without precedent, as a similar phenomenon was observed after cells infected by Salmonella typhimurium were treated with cytochalasin D, where stabilised F-actin patches were similarly evident beneath adherent bacteria [91], and even during bead uptake by phagocytes treated with cytochalasin B [92]. Cytochalasin D may therefore not prevent F-actin accumulation, despite inhibiting actin-dependent processes. As with PI3K-and myosin Xdependent processes, the efficiency of drug-mediated inhibition of actin polymerisation may be dependent on particle size [89,90,93], and consequently the short dense F-actin present beneath~400 nm diameter EBs represent relatively poor targets. Despite the 'dogma' of cytochalasin D-mediated inhibition of pathogen entry, conflicting effects have been reported for Chlamydia. Ward and Murray [94] reported only a 50% reduction in chlamydial entry into cytochalasin D-treated cells, whereas Carabeo and colleagues later reported a 41-fold reduction in entry [2]. However, entry of related C. psitacci was only decreased by 10% [95]. As TARPmediated actin polymerisation can occur both directly and via Rac1-dependent pathways [8,11], these redundant mechanisms may be differentially susceptible to inhibitors in different strains and target cells. It is clear that the mechanisms by which a small bacterium like C. trachomatis triggers host actin polymerisation now demands further investigation.
We reveal a functional relationship between EB internalisation and PX-BAR-domain containing protein SNX9, which mediates F-actin rearrangements during the early entry process. Indeed, global increases in filopodia formation during infection (Fig 9A), similar to those observed previously [2], as well as specific association between C. trachomatis EBs and filopodia (Fig 9B), are decreased in the SNX9 -/cells. How SNX9 is facilitating this process remains an open question, yet one attractive hypothesis is that SNX9 acts as a scaffold for the recruitment and activation of N-WASP to bring about filopodia formation, membrane ruffling and macropinocytic uptake of the EB [66,96,97]. SNX9 interaction with factors important for filopodia formation have been previously identified, and include N-WASP and Arp2/3 [66,73,96]. A more recent study also described a direct interaction between Cdc42 and SNX9 and intriguingly linked SNX9 expression to increased filopodia formation [74], suggesting our observation that initial filopodial capture of EBs is impaired in SNX9 -/cells is a direct consequence of impaired F-actin rearrangements (Fig 10).
Similar to the effect of SNX9 depletion on Salmonella invasion [58], in our study SNX9 knockdown or knockout did not abrogate entry completely. As some SNX-PX-BAR proteins are reported to have redundant roles, for instance SNX18 can compensate for SNX9 deficiency in CME, which may in part arise from the ability of SNX-PX-BAR family proteins to form heterodimers [98,99], there may be compensatory effects during entry into these SNX9 depleted cells that might also allow chlamydial entry to proceed.
SNX9 is recognised not only for coordination of membrane remodelling, but also as a scaffold to integrate F-actin reorganisation, endocytic traffic and Rho GTPase activity to fulfill roles in both cellular homeostasis and disease [74,100]. Consequently, SNX9 is an attractive candidate for hijack by opportunistic bacterial pathogens. SNX9 has been implicated in both the entry and infection of Salmonella and pedestal formation by EPEC and related EHEC [68,69,101,102]. These pathogens utilise T3SS effectors to subvert SNX9 activity, for example during Salmonella invasion localised SopB-mediated increases in PI (3,4)P2 recruit SNX9 to membranes to facilitate ruffling and N-WASP signalling. In this process, comparable decreases in Salmonella entry are observed when SNX9 is depleted [101]. It is tempting to speculate that a chlamydial effector interacts with SNX9 during cell entry, triggering oligomerisation of SNX9 to amplify SH3-domain mediated interactions akin to the signal amplification induced by EPEC/EHEC [68,69,102]. Whether or not SNX9 is recruited to membranes through an association with a chlamydial T3SS effector remains to be established, but an additional possibility is that manipulation of phosphoinositides during entry temporally and spatially control SNX9 recruitment, similar to the indirect recruitment of SNX9 mediated by Salmonella SopB [101]. To date no chlamydial entry effectors that directly interfere with phosphoinositide signalling have been have been identified, unlike in most other bacterial pathogens studied, for example phosphoinositide phosphatase mimics such as Salmonella SopB or Shigella IpgD [7,103]. However, the T3SS effector TARP binds to the p85 subunit of PI3K, whereas the T3SS effector TepP can interact with both p85 and p110 PI3K subunits and contributes to PI3K activation on early inclusions [11,104]. These effects are intriguing given the apparent PI3K-independent PI3P and PI (3,4,5)P3 interaction with C. trachomatis we observe during early entry and now warrant further investigation.
Chlamydial entry and early T3SS effectors are not well defined, a fact reinforced by the observation that only C. trachomatis TARP harbors the N-terminal repeat regions required for Rac1 activation, suggesting additional factors are required by other species [10,16]. In this study we identified roles for N-WASP and macropinocytosis during the internalisation of C. trachomatis using a panel of small molecule inhibitors, in addition to the established roles of Rac1 and Arp2/3 [38,39]. This is in contrast to other studies which have implicated clathrin mediated endocytosis in EB uptake [105][106][107]. However, consistent with the data presented here, including a lack of clathrin recruitment to chlamydial entry sites (S11 Fig), our detailed dissection of entry structures by cryo-EM never revealed an electron-dense clathrin coat present at chlamydial entry foci [17].
Our data reveal new insights into the diversity of signalling underlying the entry of C. trachomatis into host cells. We revealed a key initial interaction between EBs and host cell filopodia mediated by the SNX-PX-BAR protein SNX9. This shares similarities with virus-like entry routes, and precedes a macropinocytosis-like pathway. Further studies of the underlying Fig 10. SNX9 knockout decreases C. trachomatis entry and influences filopodial association early during host F-actin reorganisation. Schematic depicting the entry defects associated with entry into SNX9 -/cells. When bacteria are sedimented onto cells there is a 25 ±9% decrease in C. trachomatis internalisation, whereas when cells are infected via egress 59 ±8% fewer EBs are internalised. Both methods of infection involve fewer EBs in contact with filopodia. We propose it is this filopodia defect which results in a discrepancy in entry efficiency by the two different methods of infection. This seems likely as filopodial capture would result in more efficient infection when cells are exposed to EBs from infected cells by egress, yet would have less impact on efficiency when cells are infected by sedimentation, consequently partially compensating for SNX9 loss.
https://doi.org/10.1371/journal.ppat.1007051.g010 molecular mechanisms will reveal insights into the hijack of host cell function by this important obligate intracellular pathogen.

Infection of cultured cells with Chlamydia
C. trachomatis LGV2 or C. trachomatis D serovars were propagated in HeLa cells as previously described and stored in SPG buffer at −80˚C [108]. Infections were carried out by diluting the stored LGV2 serovar in infection medium (DMEM/F12 or DMEM, 10% FCS, 25μg ml −1 gentamicin) resulting in a multiplicity of infection (MOI) of~5-30 EBs per cell. In sedimentation infection cells seeded 24 h previously were overlaid with infection medium and centrifuged at 900 x g for 10 min. At an appropriate time-point, cells were fixed with paraformaldehyde (PFA) for immunostaining. Bacterial inclusion-forming units (IFU) were determined as described [109]. For the egress method, cells were infected according to [17] with minor modifications. Briefly, RPE1 or HAP1 cells were seeded into 100 mm dishes. 24 h later cells were infected with C. trachomatis LGV2 (MOI 5-30) by sedimentation. The following day, noninfected cells (RPE1 or HAP1) were seeded onto 12 mm coverslips in a separate 24-well plate. At 48 hpi when infected cells begin to release new EB progeny, the coverslips were introduced to the 100 mm dish and incubated for an appropriate time-point at 37˚C to allow infection directly with the released EBs.

Infection of cultured cells with enteropathogenic E. coli
Enteropathogenic E. coli (EPEC) was inoculated into Luria-Bertani broth and incubated overnight at 37˚C with shaking. EPEC overnight cultures were diluted 1:25 into DMEM supplemented with 10% FCS and 100 mM HEPES pH 7.4 and grown to mid-log phase, conditions which maximise T3SS activity [110]. Cell monolayers were infected with an appropriate dilution of mid-log phase cultures in DMEM supplemented with 10% FCS and 25 mM HEPES pH 7.4 without antibiotics at an MOI of 50. Infected cells were incubated for 4 h at 37˚C prior to fixation.

Infection of cultured cells with Salmonella enterica
Salmonella enterica overnight cultures were diluted 1:25 in 2 ml of Luria-Bertani broth and incubated for 4 h at 37˚C with shaking. Cell monolayers were pre-treated with media without antibiotics containing 1% (v/v) serum and 2μm cytochalasin D. After 5 minutes, this media was replaced with medium containing appropriately diluted bacterial stock and cytochalasin D at an MOI of 50. Cells were centrifuged at 160 × g, 10 minutes, room temperature to synchronize the infection. After incubation at 37˚C 5% CO2 for 1 h, cells were washed three times and remaining extracellular bacteria killed by incubation of the cells in infection media supplemented with 25 μg/ml gentamicin 37˚C 5% CO2, for 1 h. Cells were washed again and lysed in 0.05% (v/v) Triton-X-100. Serial dilutions of cell lysates were plated on LB agar and the percentage of intracellular bacteria compared with the original inoculum was determined for both cytochalasin D treated and mock treated cells.

Immunofluorescence and live-cell microscopy
For immunolabelling, cells were cultured on 12 mm coverslips in 24-well dishes. When appropriate, cells were fixed by exchanging media for 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS) at room temperature. Cells were incubated in PFA for 20 min before neutralisation using an equal volume of 50 mM NH 4 Cl in PBS for at least 20 min prior to antibody labelling. Following fixation, cells were permeabilized in 0.05% Triton X-100 (10 min), rinsed in PBS and then washed in PBS containing 0.1% (w/v) BSA. Primary antibodies were diluted in PBS / 1% (w/v) BSA, added to coverslips and incubated for 2 h. Secondary antibodies (Alexa Fluor 488, 546 and 633) or Texas Red phalloidin (for visualisation of F-actin) were diluted in PBS / 1% (w/v) BSA, added to coverslips and incubated for 1 h. Coverslips were mounted with Mowiol (Sigma) and observed using a confocal microscope (TCS Sp5 AOBS; Leica). For live-cell imaging of GFP-SNX9 expressing cells, cells were infected as described above with CellTracker CMTR (Thermofischer) labelled C. trachomatis LGV2 EBs, according to the method devised by Boleti and colleagues [111].

Invasion efficiency assay
At an appropriate time-point, infected cells were fixed in 1% PFA for 20 min and quenched with 50 mM NH 4 Cl in PBS. Following fixation, cells were rinsed in PBS and then in PBS containing 0.1% (w/v) BSA. Cells were stained as above, incubating with anti-Chlamydia primary antibody, followed by secondary antibody (Alexa Fluor 488). Cells were then permeablised in 0.05% Triton X-100 for 10 min. Cells were then stained again with the same anti-Chlamydia primary antibody followed by secondary antibody (Alexa Fluor 633, pseudocoloured blue) and Texas-Red phalloidin. Extracellular bacteria are therefore labelled twice with Alexa Fluor 488 (green) and 633 (blue) appearing cyan and allowing them to be distinguished from intracellular bacteria labelled only with Alexa Fluor 633 which appear blue. The percentage of intracellular bacteria was then calculated [total bacteria-extracellular bacteria]/total bacteria x 100) and expressed relative to control.

Small molecule inhibitor assays
RPE1 cells were plated into 24-well plates containing glass coverslips. 24 h later the cell media was removed and replaced with media containing 1% (v/v) serum and inhibitors at the indicated concentrations. After 5 min, this media was replaced with media containing C. trachomatis EBs and the inhibitor. Cells were infected as described above and incubated at 37˚C for 2 h. For entry experiments at 2 h post-infection, quantification of invasion efficiency was carried out as described previously in inhibitor and mock-treated control cells. For quantification of the effects on inclusion formation, inhibitor was washed out at 2 h post-infection, and the infection allowed to proceed until 24 h post-infection when cells were fixed and stained.

Dextran and transferrin uptake assays
Cells were seeded at an appropriate density in 24-well plates containing 12 mm coverslips. 24 h later cells were infected as described above in infection media containing 1% serum and either TRITC-dextran at a concentration of 1 mg ml -1 for 30 min or transferrin Alexa Fluor 647 conjugate at a concentration of 200 μg ml -1 for 30 min. Following ligand uptake, the plates were incubated on ice and the medium was removed. Cells were then washed twice with ice-cold PBS and for 2 min in ice-cold stripping buffer containing 150 mM NaCl, 100 mM glycine, 5 mM KCl and 1 mM CaCl 2 at pH 4.5. Cells were washed twice with ice-cold PBS prior to fixation in PFA. The total fluorescence of transferrin Alexa Fluor 647 conjugate within cell boundaries was corrected by background fluorescence in ImageJ. Mean transferrin fluorescence in arbitrary units was quantified in infected and non-infected cells. Dextran uptake was measured according to a modified version of the protocol reported by Wang and colleagues [112]. Briefly, the 3D objects counter tool in ImageJ was used to quantify dextran uptake as 3D objects or puncta (macropinosomes) within z-stacks, using a minimum size filter of 0.2 μm2-20 μm 2 for specificity of macropinosome identification and to minimise background fluorescence. Localisation of endogenous SNX9 and GFP-SNX9 in non-infected RPE1 cells. Cultured RPE1 cells were fixed. Fixed cells were stained with an anti-SNX9 primary antibody followed by an Alexa Fluor 488-conjugated secondary antibody. Cultured RPE1 cells were transfected with GFP-SNX9 and 18 h later were fixed. Representative images show greyscale maximum projections of confocal xy sections and inset zoom panels. Left: punctate; Right: tubular. Scale bars, 10 μm. (B) Cultured RPE1 were infected with EPEC for 4 h prior to fixation. Fixed cells were stained with an anti-SNX9 primary antibody followed by an Alexa Fluor 488-conjugated secondary antibody (green) and DRAQ5 (DNA stain, blue) to visualise the bacteria. Images show a maximum projection of confocal XY sections. Scale bar, 10 μm. (C) Cultured RPE1 cells were transfected with GFP-SNX9 and 18 h later were infected with EPEC for 4h prior to fixation. Fixed cells were stained with rhodamine-phalloidin (red) to visualise F-actin and DRAQ5 (DNA stain, blue) to visualise the bacteria. Images show a maximum projection of confocal xy sections. Scale bar, 10 μm. (TIF) S1 Movie. Dynamic association of EBs and PI3P during C. trachomatis infection. Cultured RPE1 cells were transfected with p40-PX-GFP (PI(3)P reporter) and 18h later were infected with C. trachomatis LGV2 CMTR labelled EBs (red). Cells were incubated at 37˚C for 15 min prior to imaging. Confocal z-stacks were captured every 10 s for 15 min. (MP4) S2 Movie. GFP-SNX9 is transiently recruited during C. trachomatis infection. Cultured RPE1 cells were transfected with GFP-SNX9 and 18h later were infected with C. trachomatis LGV2 CMTR labelled EBs (red). Cells were incubated at 37˚C for 15 min prior to imaging. Confocal z-stacks were captured every 49 s for 16 min. (MP4)