Legionella-Containing Vacuoles Capture PtdIns(4)P-Rich Vesicles Derived from the Golgi Apparatus

The environmental bacterium Legionella pneumophila causes a life-threatening pneumonia termed Legionnaires’ disease. The bacteria grow intracellularly in free-living amoebae as well as in respiratory tract macrophages. To this end, L. pneumophila forms a distinct membrane-bound compartment called the Legionella-containing vacuole (LCV). Phosphoinositide (PI) lipids are crucial regulators of the identity and dynamics of host cell organelles. The PI lipid PtdIns(4)P is a hallmark of the host cell secretory pathway, and decoration of LCVs with this PI is required for pathogen vacuole maturation. The source, dynamics, and mode of accumulation of PtdIns(4)P on LCVs are largely unknown. Using Dictyostelium amoebae producing different fluorescent probes as host cells, we show here that LCVs rapidly acquire PtdIns(4)P through the continuous interaction with PtdIns(4)P-positive host vesicles derived from the Golgi apparatus. Thus, the PI lipid pattern of the secretory pathway contributes to the formation of the replication-permissive pathogen compartment.

demonstrate the vesicular nature of the PtdIns(4)P association (Fig. 1B). At 30 min p.i., net accumulation of PtdIns(4)P-rich vesicles was obvious, increasingly giving the appearance that the PtdIns(4)P around the LCV was a continuous membrane ( Fig. 1; see also Movie S2). However, this was not the case; individual vesicles could still be resolved, and the vesicle association did not show a continuous elliptical curvature, as typically observed with longer exposure times.
By 45 min p.i., the LCV took on the classic spherical appearance. The LCV membrane comprised a collection of slightly larger PtdIns(4)P-positive vesicles, compared to the previous time points (Fig. 1) (movie not shown). Importantly, the 45-min time series clearly illustrates that the PtdIns(4)P association is vesicular, as the vesicles could be observed to change position and deviate from the limiting LCV membrane, rather than forming a continuous PtdIns(4)P-positive membrane. The image inset of the final frame poignantly confirms these observations, as the individual PtdIns(4)P vesicle lumens became resolvable in their dynamic repositioning. In summary, the use of real-time 3D high-resolution resonant CLSM allowed the observation of the net accumulation of PtdIns(4)P-rich vesicles on LCVs. At around 45 min, vesicle lumens were still resolvable, and LCVs were not uniformly coated with a continuous PtdIns(4)P membrane. Rather, vesicles "stagnated" on most LCVs, thus apparently leading to a net accumulation of the PtdIns(4)P lipid.
Host-and T4SS-dependent association of PtdIns(4)P vesicles with LCVs. Applying real-time CLSM, we used dually labeled D. discoideum strains producing in tandem P4C SidC -mCherry and the PtdIns(3)P probe GFP-2ϫFYVE to analyze the PI patterns underlying the formation of vacuoles harboring L. pneumophila JR32 or the T4SSdeficient strain ΔicmT. The high-resolution approach revealed that vesicles positive for PtdIns(4)P or PtdIns(3)P both simultaneously and independently of one another interacted with the bacterial compartments, while the morphological appearances of the vesicles were similar ( Fig. 2; see also Movies S3 to S6).
At 15 min p.i., PtdIns(3)P vesicles associated with early LCVs harboring L. pneumophila JR32, which were not extensively overlapping with the PtdIns(4)P vesicles ( Fig. 2A; see also Movie S3). Overall, the PtdIns(3)P vesicles seemed to associate less tightly with the pathogen compartment than the PtdIns(4)P vesicles (Movie S3). Moreover, the net clearance of PtdIns(3)P appeared to take place through the shedding of PtdIns(3)P-rich vesicles. At 45 min p.i., the PtdIns(3)P-positive vesicles were compacted and remained clear of the LCV after their shedding (Movie S4). In contrast, a strong PtdIns(4)P signal was observed around the LCV at this time point, in agreement with the concept of dynamic stagnation and net accumulation of PtdIns(4)P-rich vesicles. Since the clearance of PtdIns(3)P vesicles coincided with the accumulation of PtdIns(4)P vesicles, our results indicate that the PI conversion from PtdIns(3)P to PtdIns(4)P on LCVs takes place through selective vesicle trafficking events rather than as a result of (or in addition to) a direct transformation of PtdIns(3)P into PtdIns(4)P.
In contrast to LCVs harboring the virulent JR32 strain, vacuoles containing Icm/Dotdeficient ΔicmT mutant bacteria remained enriched for PtdIns(3)P and, at 15 min p.i. as well as at 45 min p.i., seemed to still acquire PtdIns(3)P-positive vesicles ( Fig. 2B; see also Movie S5 and S6). Interestingly, PtdIns(4)P-positive vesicles also temporarily associated with ΔicmT-containing vacuoles, in an obviously Icm/Dot-independent manner, but did not accumulate. At both 15 and 45 min p.i., vesicular PtdIns(4)P trafficking to the bacterial compartment was evident, and the early bacterial vacuole was literally dragged through a PtdIns(4)P-rich vesicle network. However, in contrast to vacuoles harboring strain JR32, ΔicmT-containing vacuoles remained essentially free of immobilized PtdIns(4)P ( Fig. 2B; see also Movie S5 and S6). Hence, real-time microscopy revealed the fast kinetics of in-coming and out-going vesicle trafficking on LCVs at unprecedented resolution. From these observations, we conclude that there is Icm/ Dot-independent "baseline" trafficking of PtdIns(4)P vesicles to vacuoles harboring newly internalized bacteria but the Icm/Dot T4SS is necessary for capturing and incorporating these vesicles, thus altering and defining the vacuole identity. Taking the results together, vesicular trafficking largely contributes to both the Icm/Dotdependent removal and segregation of PtdIns(3)P as well as the accumulation of PtdIns(4)P on LCVs.
PtdIns(3)P-positive vesicles interact with but do not fuse with PtdIns(4)Ppositive LCVs. LCVs harboring wild-type L. pneumophila shed their PtdIns(3)P identity early during the infection process through the net loss of PtdIns(3)P-positive vesicles (Fig. 2). To assess vesicle dynamics at later stages of LCV maturation, we infected dually labeled D. discoideum strains producing P4C SidC -mCherry and GFP-2ϫFYVE in tandem with the virulent JR32 strain and imaged the infection after 18 h. At that time point, all observed LCVs harboring several bacteria were exclusively PtdIns(4)P-positive. Under those conditions, PtdIns(3)P-rich vesicles still trafficked to PtdIns(4)P-positive LCVs but did not fuse or accumulate on the LCV membrane at all (Fig. 3A). Moreover, in heavily infected amoebae, PtdIns(3)P-positive vesicles also interacted with PtdIns(4)P-negative (likely newly formed) pathogen vacuoles but also did not fuse with these compartments (Fig. 3B).
In general, at a given late point during infection, PtdIns(3)P-positive vesicles were still vividly trafficking along microtubules and overall vesicle trafficking seemed intact ( Fig. 3; see also Movie S7). These observations indicated that the infection with L. pneumophila was relatively stealthy and did not severely compromise crucial cellular trafficking pathways. In contrast, the trafficking of PtdIns(4)P-rich vesicles was no longer observed at late stages of infection, likely because the probe was tied up on the massively PtdIns(4)P-positive LCVs at this time point. In summary, at late stages of infection, PtdIns(3)P-positive vesicles still interact with but do not fuse with PtdIns(4)Ppositive LCVs, and the trafficking of these vesicles as well as vesicle trafficking in general does not seem to be substantially compromised by the infection with L. pneumophila.
LCVs interact with PtdIns(4)P from the trans-Golgi network. To characterize the cellular compartment source of the PtdIns(4)P-positive vesicles accumulating on LCVs, we employed D. discoideum strains producing the well-characterized PtdIns(4)P/Golgi probe 2ϫPH FAPP -GFP (24,36). In keeping with the reported probe localization in mammalian cells, 2ϫPH FAPP -GFP principally localizes to the trans-Golgi network (TGN), with weak plasma membrane localization also in D. discoideum (Fig. 4A). Upon infection of D. discoideum producing 2ϫPH FAPP -GFP with L. pneumophila JR32, the 2ϫPH FAPP -GFP probe not only labeled the PtdIns(4)P-positive filaments of the Golgi apparatus but also accumulated on the limiting membrane of LCVs. Projections of the Golgi apparatus labeled by 2ϫPH FAPP -GFP made contact with and began to associate with the LCV around 15 min p.i. (Fig. 4B) and robustly enveloped the pathogen vacuole 30 min p.i. over the course of several minutes (Fig. 4C). In contrast, in D. discoideum infected with ΔicmT mutant bacteria, the probe still robustly labeled the PtdIns(4)P-positive filaments of the Golgi apparatus but did not localize to or accumulate on the membrane of vacuoles harboring the avirulent bacteria (Fig. 4A).
To validate the observed interactions of LCVs with Golgi membranes, we used an unrelated Golgi marker, golvesin (37). D. discoideum amoebae producing in parallel 2ϫPH FAPP -mCherry and the specific Golgi core probe Δ(1-75;119 -579)golvesin-GFP were infected with L. pneumophila JR32 or ΔicmT mutant bacteria for 1 h (Fig. 4D). Vacuoles harboring strain JR32 robustly stained positive for this set of Golgi markers, corroborating that the PtdIns(4)P decorating LCVs originated from a Golgi-derived source. In contrast, vacuoles containing ΔicmT mutant bacteria were totally devoid of either of the two Golgi markers. Taking the results together, the mammalian PtdIns(4)P probe 2ϫPH FAPP -GFP also labels Golgi PtdIns(4)P and LCVs in D. discoideum, and the D. discoideum Golgi marker golvesin accumulates on LCVs, indicating that PtdIns(4)P-rich Golgi membranes associate with LCVs.
The Icm/Dot T4SS determines sustained association of LCVs with the Golgi apparatus. Next, we sought to assess the contribution of the Icm/Dot T4SS to the accumulation of Golgi-derived PtdIns(4)P-positive vesicles on LCVs. To this end, we employed D. discoideum strains producing in tandem 2ϫPH FAPP -mCherry and Arf1-GFP. The Golgi-associated small GTPase Arf1 regulates Golgi-ER trafficking as well as intra-Golgi transport (38) and is recruited to LCVs by the Icm/Dot translocated effector protein RalF (14). LCV Interactions with Golgi-Derived PtdIns4P Vesicles ® Upon infection of the dually labeled D. discoideum strain with L. pneumophila JR32, both 2ϫPH FAPP -mCherry and Arf1-GFP associated with LCVs in a sustained manner, but the two probes did not strictly overlap and showed distinct accumulation kinetics ( Fig. 5A; see also Movie S8). While the amount of 2ϫPH FAPP -mCherry increased from 30 to 60 min p.i., Arf1-GFP association did not appear to intensify during this period. In contrast, upon infection of D. discoideum producing 2ϫPH FAPP -mCherry and Arf1-GFP with L. pneumophila ΔicmT, the Golgi membranes were inevitably brought into proximity of the compartment harboring the bacteria but did not engage in sustained interactions ( Fig. 5B; see also Movie S9). The video frames 30 min p.i. showed what appears to be co-localization of the ΔicmT-containing compartment and both Golgi probes, but approximately 700 ms later, the Golgi membranes were entirely clear of the compartment. Thus, the Golgi does not sustainably associate with the vacuole con- taining avirulent L. pneumophila. Finally, upon infection of D. discoideum producing 2ϫPH FAPP -mCherry and Arf1-GFP with L. pneumophila ΔralF, the PtdIns(4)P probe labeled LCVs harboring the mutant strain to the same extent as LCVs harboring the parental strain, while Arf1-GFP was not observable on pathogen vacuoles (Fig. 5C). These findings are in agreement with the notion that Golgi-derived PtdIns(4)P accumulates on LCVs independently of RalF-mediated Arf1 recruitment. In summary, the use of 2ϫPH FAPP -mCherry and Arf1-GFP revealed that the Icm/Dot T4SS determines sustained association of LCVs with the Golgi apparatus in an Arf1-independent manner.
The PtdIns(4)P probes, 2؋PH FAPP and P4C SidC show distinct LCV interaction dynamics. Based on the different spatiotemporal localization of 2ϫPH FAPP -mCherry and Arf1-GFP on LCVs, we decided to simultaneously assess the localization dynamics of the eukaryotic and bacterial PtdIns(4)P probes, 2ϫPH FAPP and P4C SidC , respectively. In D. discoideum producing in parallel 2ϫPH FAPP -GFP and P4C SidC -mCherry, the former predominantly labels the Golgi apparatus, while the latter in addition to the Golgi primarily localizes to the plasma membrane and (endosomal) vesicles surrounding the Golgi (Fig. 6A). Hence, aside from the plasma membrane where P4C SidC -mCherry localization is dominant, there is little obvious spatial overlap between the two probes recognizing the same PI lipid.
Upon infection of D. discoideum producing 2ϫPH FAPP -GFP and P4C SidC -mCherry with L. pneumophila JR32, the LCVs were marked by PtdIns(4)P-positive vesicles as indicated by P4C SidC -mCherry, but were also entangled by a dynamic meshwork of TGN labeled by 2ϫPH FAPP -GFP ( Fig. 6B; see also Movie S10). Noteworthy, while P4C SidC exclusively labeled the limiting LCV membrane, thus defining its identity, 2ϫPH FAPP not only labeled the LCV membrane (as seen in Fig. 4 and 5), but also extended into the TGN. The kinetics of LCV labeling of both probes, P4C SidC -mCherry and 2ϫPH FAPP -GFP, were very similar (80% to 90% positive LCVs 1 to 2 h p.i.), and the probes maintained their distinct labeling patterns throughout the infection with L. pneumophila from 2 h p.i. to 16 h p.i. (Fig. 6C).
Transient Arf1 recruitment to LCVs. Arf1-GFP robustly localizes to LCVs at early time points of pathogen vacuole formation (30 to 60 min p.i.) (Fig. 5A). To further assess the time window during which Arf1 is recruited to LCVs, we infected D. discoideum strains producing Arf1-GFP and 2ϫPH FAPP -mCherry (Fig. 7A) or Arf1-GFP and P4C SidC -mCherry ( Fig. 7B) with L. pneumophila JR32. These experiments confirmed Arf1 localization to LCVs at early time points; however, at 2 h p.i., the interaction of the Arf1-positive TGN with LCVs appeared to subside. This happened alongside the accumulation of the PtdIns(4)P/Golgi marker 2ϫPH FAPP , which remained on LCVs similarly to P4C SidC , likely reflecting the continuous accumulation of PtdIns(4)P on the LCVs. Hence, the interactions of LCVs with PtdIns(4)P-positive Golgi membranes occur early during infection (within 1 h p.i.) and later diminish. In summary, this high-resolution CLSM study using the Golgi markers Arf1, PH FAPP , and golvesin revealed that, during their maturation, LCVs interact with Golgi-derived PtdIns(4)P-positive vesicles at early time points of infection.

DISCUSSION
Using real-time 3D high-resolution resonant CLSM, we have shown that vesicular trafficking contributes to the Icm/Dot-dependent removal and segregation of PtdIns(3)P as well as to the accumulation of PtdIns(4)P on LCVs. The PtdIns(3)P-and/or PtdIns(4)P-positive vesicles investigated here might correspond to the "smooth vesi-cles" associating with LCVs originally observed by EM (39). At early time points (Ͻ1 h p.i.) LCVs were not uniformly coated with a continuous PtdIns(4)P membrane, and the lumen of PtdIns(4)P-positive vesicles was still resolvable. The association of small PtdIns(4)P-positive vesicles with LCVs correlates with the punctate PtdIns(4)P and SidC staining observed previously (26,28). The PtdIns(4)P-positive vesicles appeared to "stagnate" on the LCVs, thus leading to a net accumulation of the PI lipid. This process likely involves tethering and immobilization of PtdIns(4)P-positive vesicles on the LCVs, followed by fusion of the vesicle and the pathogen vacuole membrane. At present, the putative host and pathogen factors promoting the tethering of and interactions with PtdIns(4)P-positive vesicles are unknown.
The Golgi protein FAPP1 binds both PtdIns(4)P and Arf1 (24,36,40). Producing 2ϫPH FAPP -mCherry and Arf1-GFP or 2ϫPH FAPP -GFP and P4C SidC -mCherry, respectively, in D. discoideum indicated that the LCVs associate with the Golgi apparatus and accumulate Golgi-derived rather than plasma membrane-derived PtdIns(4)P. Most of the cellular PtdIns(4)P is found in the Golgi apparatus, the secretory vesicles, and the plasma membrane (22,24), but there are additional pools of this lipid found in (late) endosomes (41,42), which might contribute to the acquisition of vesicle-bound PtdIns(4)P by nascent LCVs. However, the fact that LCVs deviate from the endosomal route early during formation, together with the accumulation on LCVs of the Golgispecific probes 2ϫPH FAPP -mCherry and Arf1-GFP, strongly suggests that the PtdIns(4)Ppositive vesicles interacting with LCVs are indeed derived from the Golgi apparatus. Overall, these results also emphasize the importance of performing live-cell rather than fixed-sample experiments and strengthen the notion of the LCV as a dynamic compartment (co)defined by the frequency and/or duration of vesicular interactions. From a technological standpoint, the speed of the resonant scans and of multi-Z-plane imaging allowed us to decipher these processes.
At later stages of infection, the PtdIns(4)P-positive LCVs still interacted but did not fuse with PtdIns(3)P-positive vesicles. The Icm/Dot-translocated effector VipD shows Rab5-activated phospholipase A 1 activity, removes PtdIns(3)P from endosomal membranes, and reduces Rab5 levels on early LCVs (43). Thus, VipD might contribute to limit the interactions of LCVs with endosomes throughout pathogen vacuole maturation. The putative Icm/Dot substrates promoting the observed early interactions of LCVs with PtdIns(4)P-positive vesicles and their sustained accumulation on the pathogen vacuole are unknown. In any case, the Icm/Dot substrate RalF is dispensable for the accumulation of PtdIns(4)P on LCVs. While Arf1-GFP was not observable on pathogen vacuoles harboring L. pneumophila ΔralF (Fig. 5C), as published previously for mam-malian cells (14), the accumulation of 2ϫPH FAPP -mCherry and, hence, PtdIns(4)P was not compromised. These results also indicate that Arf1, which recruits PI 4-kinase (see below), is dispensable for the accumulation of PtdIns(4)P on LCVs.
Further adding to the complexity of the process, a number of host PI-metabolizing enzymes have been implicated in the production of PtdIns(4)P on the LCV membrane. The PI 4-kinase class III␤ (PI4K III␤) is recruited by the small GTPase Arf1 and promotes traffic along the secretory pathway (54). Both Arf1 and PI4K III␤ promote accumulation of SidC on the LCV, suggesting that these host factors contribute to PtdIns(4)P accumulation (30,55). Arf1 localizes to LCVs (14), but the association of PI4K III␤ with the pathogen vacuole remains to be assessed. Another host factor potentially involved in shaping the LCV PI pattern is the PI 5-phosphatase Oculocerebrorenal syndrome of Lowe (OCRL), which localizes to the TGN and endosomes and regulates retrograde trafficking between the two compartments (56). OCRL promotes intracellular replication of L. pneumophila (57) and determines LCV composition, including Rab1 and retrograde trafficking components (58). The PI 5-phosphatase preferentially dephosphorylates PtdIns(4,5)P 2 and also PtdIns(3,4,5)P 3 , yielding PtdIns(4)P and PtdIns(3,4)P 2 . Based on the SidC localization assay, OCRL produces PtdIns(4)P on LCVs (57). Moreover, the PI 3-phosphatase effector SidF possibly cooperates with OCRL to produce PtdIns(4)P from PtdIns(3,4)P 2 .
Taken together, the available data are in agreement with a model stipulating that LCV PI conversion involves host factors as well as pathogen factors and is the sum of processes occurring in trans (at a distance from the LCV) and others occurring in cis (on the LCV directly). As documented in this study, vesicle identity and trafficking in trans seem to set the stage and determine early events of LCV formation. The L. pneumophila PI-modulating effectors appear to preferentially act in cis. Yet the issue of whether some of these effectors also act in trans, like several other L. pneumophila effectors, modifying, e.g., ribosomes, mitochondria, or histones (9, 10), has not been addressed. The work presented here provides an outline to address these issues and to search among the more than 250 uncharacterized L. pneumophila Icm/Dot substrates for effectors modulating early steps of LCV formation by interfering with host cell vesicle trafficking.
was performed with 2 pulses of 1 ms and 1 mV separated by a 5-s gap. Directly after electroporation, cells were transferred into a T75 flask containing 10 ml HL-5. At between 12 and 24 h after electroporation, the medium was replaced with fresh HL-5 and the required selection antibiotics were added. The medium was changed 72 h later. Upon the obvious appearance of several microcolonies (usually 6 to 7 days after transformation), cells were dislodged into fresh medium and transferred to a new flask.
Sample preparation for microscopy. D. discoideum amoebae producing the desired fluorescent probes were harvested from approximately 70%-confluent cultures. HL-5 medium was removed, and cultures were washed with 5 ml LoFlo medium (ForMedium) and resuspended in fresh LoFlo medium. The cells were seeded (300 l) at a density of 2.5 ϫ 10 5 /ml to 4 ϫ 10 5 /ml in eight-well -slides (Ibidi). Cells were allowed to adhere for 1 h, after which the LoFlo medium was replaced. Infections (at a multiplicity of infection [MOI] of 5) with early stationary-phase cultures of L. pneumophila JR32 harboring pNP099 (mCerulean) or pNP102 (mCherry) were initiated in -slides already in position for imaging.
Confocal laser scanning fluorescence microscopy setup. All imaging was performed with living cells, carried out with a Leica TCS SP8 X CLSM with the following setup: white-light laser (WLL), 442-nm diode, HyD hybrid detectors for each channel used, HC PL APO CS2 63ϫ/1.4 oil objective with Leica type F immersion oil, Leica LAS X software. mCerulean was excited at 442 nm and detected at around 469 nm. Enhanced GFP (EGFP) was excited at 487 nm and detected at around 516 nm. mCherry was excited at 587 nm and detected at around 622 nm. The microscope stage thermostat was set to hold the temperature at between 22°C and 25°C. Images were captured with a pinhole at between 0.6 and 0.9 Airy units (AU) and with a pixel/voxel size at or close to the instrument's Nyquist criterion of approximately 39.5 ϫ 39.5 ϫ 118 nm (xyz).
Resonant scanning at 8,000 Hz (bidirectional scan) was used to capture videos corresponding to Fig. 1, 2, 3, 5, and 6B. Capture rates for 2 scans with 2 to 8 line averages were between approximately 2.5 and 5 frames per second. For Fig. 1, four Z-slices with 110-nm spacing were captured per time interval. Standard scanning at frequencies between 200 and 600 Hz (bidirectional scan with 2 to 3 line averages) was used to capture images and videos corresponding to Fig. 4, 6A, C, and D, and 7.
Video and image processing. All images were deconvolved with Huygens Professional version 17.10 (Scientific Volume Imaging, The Netherlands) using the CMLE algorithm with 40 iterations and a 0.05 quality threshold. Signal-to-noise ratios were estimated from the photons counted for a given image. Video captures and their snapshots were finalized with Imaris 9.1.0 software (Bitplane, Switzerland). Still images were finalized and exported with ImageJ software (https://imagej.nih.gov/ij/).

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.