PIKfyve/Fab1 is required for efficient V-ATPase and hydrolase delivery to phagosomes, phagosomal killing, and restriction of Legionella infection

By engulfing potentially harmful microbes, professional phagocytes are continually at risk from intracellular pathogens. To avoid becoming infected, the host must kill pathogens in the phagosome before they can escape or establish a survival niche. Here, we analyse the role of the phosphoinositide (PI) 5-kinase PIKfyve in phagosome maturation and killing, using the amoeba and model phagocyte Dictyostelium discoideum. PIKfyve plays important but poorly understood roles in vesicular trafficking by catalysing formation of the lipids phosphatidylinositol (3,5)-bisphosphate (PI(3,5)2) and phosphatidylinositol-5-phosphate (PI(5)P). Here we show that its activity is essential during early phagosome maturation in Dictyostelium. Disruption of PIKfyve inhibited delivery of both the vacuolar V-ATPase and proteases, dramatically reducing the ability of cells to acidify newly formed phagosomes and digest their contents. Consequently, PIKfyve- cells were unable to generate an effective antimicrobial environment and efficiently kill captured bacteria. Moreover, we demonstrate that cells lacking PIKfyve are more susceptible to infection by the intracellular pathogen Legionella pneumophila. We conclude that PIKfyve-catalysed phosphoinositide production plays a crucial and general role in ensuring early phagosomal maturation, protecting host cells from diverse pathogenic microbes.


Introduction
Professional phagocytes must kill their prey efficiently if they are to prevent the establishment of infections [1]. Multiple mechanisms are employed to achieve this. Once phagosomes have been internalised they quickly become acidified and acquire reactive oxygen species, antimicrobial peptides and acid hydrolases. The timely and regulated delivery of these components is vital to protect the host from intracellular pathogens but is incompletely understood.
As in classical endocytosis, phagosome maturation is regulated by PIPs [26]. Phagosomes accumulate PI(3)P immediately after closure, and this is required for their subsequent maturation [27][28][29]. The recent identification of several PIKfyve inhibitors, including apilimod and YM201636 [30,31], has allowed researchers to demonstrate the importance of PI(3)P to PI (3,5)P 2 conversion for phagosomal maturation in macrophages and neutrophils [32][33][34][35]. However, there are conflicting reports on the roles of PIKfyve in key lysosomal functions such as acidification and degradation, with some studies reporting defective acidification upon PIKfyve inhibition [10,36,37] and others finding little effect [33,38,39]. Therefore the mechanistic roles of PIKfyve and its products, and their relevance to phagosome maturation, remain unclear and subject to debate.
To understand the function and physiological significance of PIKfyve, we have investigated its role in phagosome maturation and pathogen killing in Dictyostelium discoideum, a soildwelling amoeba and professional phagocyte that feeds on bacteria. Dictyostelium PIPs are unusual, with the lipid chain joined to the sn-1-position of the glycerol backbone by an ether, rather than ester, linkage: these PIPs should correctly be named as derivatives of plasmanylinositol rather than phosphatidylinositol [40]. This, however, makes no known difference to downstream functions, which are dictated by interactions between the inositol polyphosphate headgroups and effector proteins. Dictyostelium has thus been an effective model for analysis of phosphoinositide signalling [41][42][43][44]. For convenience, both the mammalian and Dictyostelium inositol phospholipids are referred to as PIPs hereafter.
We find that genetic or pharmacological disruption of PIKfyve activity in Dictyostelium leads to a swollen endosomal phenotype reminiscent of defects in macrophages. We provide a detailed analysis of phagosome maturation, and show that at least some of the defects in PIKfyve-deficient cells are due to reduced recruitment of the proton-pumping vacuolar (V-ATPase). Finally, we demonstrate that PIKfyve activity is required for the efficient killing of phagocytosed bacteria and for restricting the intracellular growth of the pathogen Legionella pneumophila.

PIKfyve -Dictyostelium have swollen endosomes
The Dictyostelium genome contains a single orthologue of PIKfyve (PIP5K3). Like the mammalian and yeast proteins, Dictyostelium PIKfyve contains an N-terminal FYVE domain, a CCT (chaperonin Cpn60/TCP1)-like chaperone domain, a PIKfyve-unique cysteine/histidinerich domain and a C-terminal PIP kinase domain [7]. In order to investigate the role of PI(3,5) P 2 in Dictyostelium we disrupted the PIKfyve gene in the axenic Ax3 background by inserting a blasticidin resistance cassette and deleting a portion of the central PIKfyve-unique region. Gene disruption was confirmed by PCR of the genomic locus and loss of mRNA demonstrated by reverse transcription PCR (RT-PCR) (S1 Fig). Two independent mutants were isolated PIKfyve-1 and -2 (strain ID's JSK06 and JSK07 respectively) While the unusual ether-linked chemistry of the Dictyostelium inositol phospholipids prevented direct measurement of PI(3,5)P 2 loss by either the standard method of methanolysis followed by HPLC of deacylation products or by mass spectrometry, we found that each mutant strain was highly vacuolated (Fig 1A and 1B), resembling the swollen vesicle phenotype observed upon PIKfyve knockdown or inhibition in mammalian cells, C. elegans, S. cerevisiae and D. melanogaster [10,15,20,45]. This effect was recapitulated by incubation with the PIKfyve-specific inhibitor apilimod [30], confirming that this phenotype was due to deficient PIKfyve activity, most likely via the production of PI(3,5)P 2 or PI(5)P ( Fig 1B).
The large vacuoles normally observable in axenically growing Dictyostelium are derived from either macropinocytic uptake of extracellular nutrients, or the contractile vacuole which aids osmoregulation by pumping water from the cytoplasm into specialised bladders for expulsion. To test if the swollen vesicles are macropinocytic in origin, cells were incubated with the fluid-phase marker TRITC-dextran for 2 hours. To further confirm macropinosome identity, we used cells expressing the PI(3)P reporter GFP-2xFYVE which specifically labels the early stages of this pathway ([46] and see later). Confocal microscopy showed that PIKfyvecells contained a dramatic increase in enlarged macropinosomes, at both early and late stages of maturation at a comparable size and number to the swollen vesicles described above (Fig 1A-1C). We therefore conclude that in Dictyostelium, the swollen vesicles observed upon loss of PIKfyve activity are macropinocytic in origin.
When PIKfyve-amoebae were hypotonically stressed in phosphate buffer, we observed a sustained increase in vacuolation for at least 5 hours. However, after 24 hours-when the cells became polarized indicating the onset of starvation-induced development-PIKfyve-mutants became indistinguishable from the random integrant and parental controls (Fig 1D). This is most likely due to the suppression of fluid-phase uptake that occurs when Dictyostelium cells enter starvation-induced development [47][48][49]. Consistent with this interpretation, PIKfyvecells had no observable delay or other defects in development, and formed morphologically normal fruiting bodies with viable spores (S2 Fig). Therefore, disruption of PIKfyve leads to endocytic defects but it is not required for Dictyostelium development.

PIKfyve is important for phagocytic growth but not uptake
Laboratory strains of Dictyostelium can obtain nutrients by macropinocytosis of liquid (axenic) medium or by phagocytosis of bacteria. Although intracellular macropinosomes were swollen, PIKfyvecells had normal rates of both endocytosis and exocytosis (Fig 2A and 2B). Furthermore, the fluid uptake of both wild-type and PIKfyve-cells reached a plateau at about 60 minutes. This steady-state occurs when the first vesicles complete their transit and are exocytosed and indicates that the overall transit time in the absence of PIKfyve is unperturbed. Despite this, axenic growth was slower than for wild-type cells, with mutants doubling every 16 hours compared to 10 hours for the controls (Fig 2C).
Growth on bacteria was more strongly affected. When we measured phagocytosis by following the ability of Dictyostelium cells to decrease the turbidity of an E. coli suspension over time we found no significant defects in PIKfyvecells ( Fig 2D). In contrast, they grew significantly more slowly on a lawn of Klebsiella pneumoniae ( Fig 2E). This indicates a specific role for PIKfyve activity in phagosome maturation rather than bacterial uptake.

PIKfyve deficient cells have defective phagosome acidification and V-ATPase delivery
Next we investigated how the absence of PIKfyve affects phagosomal maturation. One of the first stages of maturation is the acquisition of the proton-pumping V-ATPase, leading to rapid acidification [50,51]. The influence of PIKfyve on endosomal pH regulation remains controversial: studies in C. elegans, plants, and mammalian epithelial cells have shown that PIKfyve is required for efficient acidification [37,45,[52][53][54], but RAW 264.7 macrophages are still able to acidify their phagosomes to at least pH 5.5 when treated with a PIKfyve inhibitor [33].
Phagosome maturation is well characterised in Dictyostelium, with most studies being performed using mutants in the Ax2 genetic background, rather than Ax3 as used above. For comparison with other studies we isolated additional mutants from Ax2 cells, which were used for all subsequent experiments unless stated otherwise. Ax2 background PIKfyve mutants also exhibited slow growth on bacterial lawns but normal phagocytosis of both beads and bacteria (S3 Fig), confirming that these phenotypes are robust across multiple genetic backgrounds. (B) Induction of swollen vesicles with 5 μM apilimod, a PIKfyve-specific inhibitor, images taken in HL5 medium after 5 hours of treatment. (C) Confocal images of cells expressing the PI(3)P reporter GFP-2xFYVE. Cells were incubated with 0.2 mg/ml TRITC-dextran for 2 hours to label macropinosomes indicating that the swollen compartments in PIKfyvecells derive from this pathway. (D) Changes in morphology upon incubation in low osmolarity starvation buffer (KK2) compared to full medium (HL5). Swollen vesicles in PIKfyvecells became initially more apparent but were lost as cells entered differentiation. https://doi.org/10.1371/journal.ppat.1007551.g001 We followed phagosome acidification in PIKfyve -Dictyostelium by measuring changes in the relative fluorescence of engulfed beads that had been labelled both with the pH-sensitive FITC and the pH-insensitive Alexa 594 succinimidyl ester [55]. Phagosomes from wild-type cells rapidly became acidified and remained acidic until~40 minutes before transitioning to neutral post-lysosomes, but phagosomes of PIKfyvecells acidified much more slowly and never achieved as low a pH ( Fig 3A). In contrast, when we incubated cells with a mixture of FITC-and pH-insensitive TRITC-conjugated dextrans to observed the acidication of macropinosomes, we found that they were still able to acidify sufficiently to quench FITC fluorescence (S4 Fig). However the concentration of dextrans within macropinosomes did appear delayed. This is consistent with the relatively mild defect in axenic growth and previous studies in mammalian cells [38], indicating a more stringent requirement for PIKfyve activity during phagosome acidification. Phagosomal acidification is driven by the rapid recruitment and activity of the V-ATPase. To differentiate between defective V-ATPase delivery and activity, we directly imaged recruitment of the VatM transmembrane subunit of the V-ATPase fused to GFP to nascent phagosomes by microscopy. By observing the phagocytosis of pH-sensitive pHrodo-labelled yeast we were able to simultaneously monitor acidification.
GFP-VatM began accumulating on phagosomes immediately following internalisation both in PIKfyveand control cells, but it accumulated more slowly in the mutants and to only about half of the levels observed in wild-type cells (Fig 3B and 3C). Defective acidification was further demonstrated by a reduced increase in pHrodo fluorescence ( Fig 3D). We conclude that there is some PIKfyve-independent lysosomal fusion with phagosomes, but that PIKfyve activity is required for the high levels of V-ATPase accumulation that are needed for efficient phagosomal acidification.
The V-ATPase consists of V 0 (transmembrane) and V 1 (peripheral) subcomplexes. It has previously been suggested that PI(3,5)P 2 can regulate V 0 -V 1 assembly at the yeast vacuole allowing dynamic regulation of activity [56]. VatM is a component of the V 0 subcomplex (subunit a in mammals and yeast). To test whether V 0 -V 1 association is also affected by loss of PIKfyve we observed the phagosomal recruitment of the V 1 subunit VatB fused to GFP [57]. Both GFP-VatM and VatB-GFP were expressed equally in wild-type and mutant cells (S5A Fig). As before VatB-GFP recruitment to nascent phagosomes was also significantly decreased in PIKfyve-cells (S5B and S5C Fig) It should be noted that expression of VatB-GFP (but not GFP-VatM) caused a partial inhibition of acidification in this assay, indicating caution should be taken in using this construct (S5D Fig). Nevertheless, the observation that both V-ATPase components were equally affected by PIKfyve deletion suggests that PIKfyve is required for delivery of the entire V-ATPase to the phagosome, rather than specifically affecting V 0 -V 1 association.

Proteolytic activity and hydrolase delivery are specifically affected in PIKfyvecells
Proper degradation of internalised material requires both acidification and the activity of proteases. To test if hydrolytic activity was also dependent on PIKfyve, we measured phagosomal proteolysis by following the increase in fluorescence due to the cleavage and unquenching of DQ-bovine serum albumin (DQ-BSA) coupled to beads [55] (Fig 4A). Strikingly, despite their partial acidification, phagosomes in PIKfyvecells exhibited an almost complete loss of proteolytic activity, an effect confirmed using the PIKfyve inhibitor apilimod (S6 Fig). To control for a potential general decrease in protease activity, we measured the unquenching of DQ-BSA beads in whole cell lysates ( Fig 4B) and the protein levels of lysosomal protease cathepsin D by Western blot (Fig 4C and 4D). Although activity in the phagosome is completely lost, we found that total proteolytic activity remained normal and cathepsin D levels were significantly increased upon loss of PIKfyve, consistent with a defect in delivery to the phagosome, rather than in lysosomal biogenesis.
To investigate whether PIKfyve activity was required to deliver proteases to the phagosome, we purified phagosomes at different stages of maturation and analysed their composition by Western blot (Fig 4E). In wild-type cells phagosomes acquired cathepsin D from the earliest time-point, but the protease was almost completely absent from phagosomes in PIKfyve-cells. Whilst the delivery of the vacuolar ATPase was also qualitatively reduced in this assay, consistent with decreased acidification, Actin-binding protein 1 (Abp1), an independent marker of phagosome maturation [58], was unaffected. Thus, although both hydrolase and to a lesser extent V-ATPase delivery requires PIKfyve activity, other aspects of maturation appear to progress normally.

PIKfyve does not regulate PI(3)P dynamics
We next wanted to confirm whether loss of PIKfyve disrupts specific aspects of phagosome maturation or causes a general trafficking defect. PI(3)P is one of the best characterised early markers of maturing endosomes and phagosomes in both mammalian macrophages and Dictyostelium. Immediately following particle internalization, PI(3)P is generated on the phagosome by the class III PI 3-kinase VPS34 [27, 33,46] and interacts with a number of important regulators of maturation such as Rab5 [59]. PIKfyve is both recruited by PI(3)P and phosphorylates it, forming PI(3,5)P 2 . Loss of PIKfyve activity could perturb phagosome maturation by reducing PI(3)P consumption, by eliminating the actions of PI(3,5)P 2 , or both. Studies on macrophages have indicated that inhibition of PIKfyve can cause prolonged PI(3)P signalling [33].
PI(3)P can be visualised in cells using the well-characterised reporter GFP-2xFYVE [41,60]. Expression of GFP-2xFYVE in control cells demonstrated that PI(3)P is present on Dictyostelium phagosomes immediately following engulfment, consistent with previous reports [46] ( Fig 4F). However, we found no abnormalities in either the recruitment to or the dissociation from phagosomes of this reporter in PIKfyve-mutants ( Fig 4F and 4G). PIKfyve activity seems not to influence the steady-state levels of PI(3)P in Dictyostelium, and the functional defects in PIKfyve-cells therefore are likely to be caused by a lack of the product(s) of PIKfyve activity.
Overall, the normal transition of PIKfyve-phagosomes into a PI(3)P-negative compartment indicates that much of their phagosome maturation process continues normally-but with the product(s) of PIKfyve action playing specific role(s) in the delivery of certain important components, including the V-ATPase and hydrolytic enzymes.

PIKfyve is essential for efficient killing of bacteria
Acidification and proteolysis are important mechanisms used by phagocytes to kill engulfed microbes. We therefore asked whether PIKfyve was physiologically important for killing. Bacterial death leads to membrane permeabilisation and intracellular acidification, so survival time within phagosomes can be inferred by observing the phagocytosis and subsequent quenching of GFP fluorescence when expressed by a non-pathogenic Klebsiella pneumoniae strain [61]. In this assay, the phagocytosed bacteria survived more than three times longer in PIKfyvecells (median survival 12 min) than in wild-type cells (3.5 min) (Fig 5A and 5B). This effect was again recapitulated by treatment with 5 μM apilimod (S6 Fig). These benign bacteria did eventually die in PIKfyvecells, indicating either that the residual acidification is eventually sufficient or that enough other elements of the complex bacterial killing machinery remain functional in PIKfyvephagosomes.
These defects in bacterial killing and digestion (Fig 4) explain why PIKfyve is important for Dictyostelium to grow on a lawn of K. pneumoniae (Fig 2E). To test whether this is general to a broad range of bacteria, we employed an assay whereby serial dilutions of amoebae are plated on lawns of a panel including both Gram-positive and Gram-negative bacteria [62]. In this assay, PIKfyve-deficient cells were severely inhibited in growth on all bacteria tested, demonstrating a general role for PIKfyve in bacterial killing and digestion (Fig 5C and 5D).

PIKfyve activity restricts the persistence of Legionella infection
Many pathogenic bacteria infect host immune cells by manipulating phagosome maturation to establish a replication-permissive niche or to escape into the cytosol. To avoid infection, host PIKfyve in microbial killing cells must efficiently kill such pathogens; hence PIKfyve might be critical to protect host cells from infection.
Legionella pneumophila is a Gram-negative opportunistic human pathogen that normally lives in the environment by establishing replicative niches inside amoebae such as Acanthamoeba. This process can be replicated in the laboratory using Dictyostelium as an experimental host [63]. Following its phagocytosis, Legionella can disrupt normal phagosomal maturation and form a unique Legionella-containing vacuole (LCV). This requires the Icm/Dot (Intracellular multiplication/Defective for organelle trafficking) type IV secretion system that delivers a large number of bacterial effector proteins into the host (reviewed in [64]). These effectors modify many host signalling and trafficking pathways, one of which prevents the nascent Legionella-containing phagosome from fusing with lysosomes [65]. Phosphoinositide signalling is heavily implicated in Legionella pathogenesis, with Legionella-containing phagosomes rapidly accumulating PI(3)P. Its concentration then declines within 2 hours and PI(4)P accumulates [66]. Multiple effectors introduced through the Icm/ Dot system bind PI(3)P or PI(4)P [67][68][69][70][71][72][73]. The role of PI(3,5)P 2 (and/or PI(5)P) in Legionella infection is yet to be investigated, so we tested whether PIKfyve was beneficial or detrimental for the host to restrain Legionella infection.
When we measured the rate of uptake of GFP-expressing Legionella by flow cytometry, we found PIKfyve -Dictyostelium were indistinguishable from wild-type. Both strains phagocytosed many more of the virulent wild-type Legionella strain (JR32) than an avirulent strain that is defective in type IV secretion (ΔicmT) [74] (Fig 6A). These results are in agreement with the previous finding that expression of the Icm/Dot T4SS promotes uptake of Legionella [75]. When we measured the ability of Dictyostelium to kill Legionella ΔicmT, the bacteria survived for significantly longer in PIKfyve-cells (Fig 6B).
We next tested the role of PIKfyve on the outcome of infection. When Ax2 and PIKfyve mutants were infected with either wild-type or ΔicmT Legionella, using a MOI of 0.1 to compensate for the greater uptake of wild-type Legionella, both amoeba strains suppressed the avirulent bacteria, although the reduction in bacteria was slower in the PIKfyve mutants. In contrast, wild-type Legionella grew substantially faster in cells lacking PIKfyve (Fig 6C, note that the CFUs scales in Fig 6B and 6C are logarithmic). We independently confirmed these results using flow cytometry of cells infected with GFP-producing bacteria in our Ax3-background mutants. The only Dictyostelium cells that accumulated substantial GFP fluorescence over several days were those infected by wild-type Legionella, and this happened sooner and to a greater degree in the PIKfyve-cells (Fig 6D).
Unlike PI(3)P and PI(4)P, the lipid products of PIKfyve are thus not required for Legionella to subvert phagosome maturation and generate its replicative vacuole. Rather, the role of PIKfyve in ensuring rapid phagosomal acidification and digestion is crucial for the host to prevent Legionella, and presumably other pathogens, from surviving and establishing a permissive niche.

Discussion
In this study, we have characterised the role of PIKfyve during phagosome maturation using the model phagocyte Dictyostelium. The roles of PI 3-kinases and PI(3)P signalling during phagosome formation and early maturation have been studied extensively but the subsequent actions of PIKfyve and roles of PI(3,5)P 2 and PI(5)P are much less well understood [3,26]. In non-phagocytic cells such as fibroblasts and yeast, PI(3,5)P 2 production is important for endosomal fission and fragmentation of endolysosomal compartments [10,18,37,45], and PIKfyve inhibition has been shown to cause macrophage lysosomes to coalesce by an unknown mechanism [34]. PIKfyve also regulates macropinosome maturation [38], intracellular replication of the vaccinia virus and Salmonella [38,53,76] and production of reactive oxygen species (ROS) in neutrophils [35]. In this paper we show that PIKfyve is critical to ensure efficient phagosomal acidification and proteolysis via delivery of specific components, and we demonstrate its physiological importance in the killing of bacteria and suppression of intracellular pathogens.
Complex effects of PIKfyve inhibition on PIP-mediated signalling have hampered clear interpretation of PIKfyve function in some mammalian studies. For example, some studies report that disruption of PIKfyve both prolonged PI(3)P-mediated signalling and eliminated PI(3,5)P 2 production [30,32], making it difficult to determine which phosphoinositide change is responsible for the observed phenotypes. In contrast, and in agreement with other reports in mammalian cells [18,37], we found that deletion of PIKfyve had no impact on phagosomal PI (3)P dynamics in Dictyostelium. The defects in phagosome maturation that we observed in this system are thus due to lack of PI(3,5)P 2 /PI(5)P formation and not to prolonged PI(3)P signalling. There is limited evidence that PIKfyve might exhibit protein kinase activity [77], but whether this is relevant in vivo remains to be shown.
Whilst PIKfyve appears to be the major route of PI(3,5)P 2 synthesis, there remains the caveat that other pathways exist in Dictostelium or mammalian cells. Indeed 20% of PI(3,5)P 2 levels remain after PIKfyve inhibition in fibroblasts [37] and whilst inhibition may not be complete it is also theoretically possible that a pool of PI(3,5)P 2 can be generated by 5-dephosphorylation of PIP 3 , or 3-phosphorylation of PI(5)P. Whilst PIKfyve is the only route for PI(3,5)P 2 synthesis in yeast, until we are able to specifically measure PI(3,5)P 2 in Dictyostelium it is possible that alternative pathways exist.
The role of PIKfyve in lysosomal acidification and degradation is currently disputed. Several studies which have measured vesicular pH at a single time point have shown that PIKfyve is required for acidification [10,37,45,54], whereas others found that disruption of PIKfyve had little effect on phagosomal pH [33,38,39]. In contrast, we followed the temporal dynamics of V-ATPase delivery and of phagosomal acidification and proteolysis, and showed that V-ATPase delivery to PIKfyve-deficient phagosomes was substantially decreased and delayed, with consequent defects on initial acidification and proteolysis. PI(3,5)P 2 has also been proposed to regulate V-ATPase V 0 -V 1 subcomplex association dynamically at the yeast vacuole [56], but we found no evidence for this during Dictyostelium phagosome maturation.
It is still not clear how PIKfyve-generated PI(3,5)P 2 regulates V-ATPase trafficking, and few PI(3,5)P 2 effectors are known. One of these is the lysosomal cation channel TRPML1/mucolipin, which is specifically activated by PI(3,5)P 2 [78]. This interaction was recently shown to partly underlie the role of PIKfyve in macropinosome fragmentation, although not acidification [38]. TRPML1 is also required for phagosome-lysosome fusion [79], and PI(3,5)P 2 and TRPML1 have been proposed to mediate interactions between lysosomes and microtubules [80]. PIKfyve may therefore drive V-ATPase delivery to phagosomes both by microtubuledirected trafficking and by regulating fission. However, the sole mucolipin orthologue in Dictyostelium is only recruited to phagosomes much later, during the post-lysosomal phase, and its disruption influences exocytosis rather than acidification [81].
Effective phagosomal acidification and proteolysis is essential if phagocytes are to kill internalised bacteria. Many clinically relevant opportunistic pathogens, including Legionella [64,82], Burkholderia cenocepacia [83] and Cryptococcus neoformans [84], have developed the ability to subvert normal phagosome maturation so as to maintain a permissive niche inside host phagocytes. This ability is likely to have evolved from their ancestors' interactions with environmental predators such as amoebae [85][86][87].
Legionella are phagocytosed in the lung by alveolar macrophages. After internalisation, they employ effectors secreted via their Type IV secretion system, some of which interfere with PI (3)P-signalling, to inhibit phagosome maturation [68,88,89]. We have shown that the products of PIKfyve are not required for Legionella to establish an intracellular replication niche. Rather, Legionella survive much better in PIKfyve-deficient cells, suggesting that PI(3,5)P 2 helps Dictyostelium to eliminate rather than harbour Legionella. Consistent with this, recent studies have shown that the Legionella virulence effector SidK directly binds to and suppresses the activity of the phagosomal V-ATPase, further underlining the importance of acidification in restricting pathogen survival [90,91]. This is in contrast to the non-phagocytic invasion of epithelia that occurs during Salmonella infection. In that case, PIKfyve activity is necessary to promote the generation of a specialised survival niche within which the bacteria replicate [53]. Salmonella has thus evolved a specific requirement for PIKfyve in generating a survival niche-likely through using phagosome acidification as a cue for virulence factor expression-whereas Legionella and other bacteria are suppressed by rapid and PIKfyve-driven phagosomal maturation.
The molecular arms race between host and pathogens is complex and of great importance. The very early events of phagosome maturation are critical in this competition; host cells aim to kill their prey swiftly whilst pathogens try to survive long enough to escape. Although its molecular effectors remain unclear, PIKfyve and its products are crucial to tip the balance in favour of the host, providing a general mechanism to ensure efficient antimicrobial activity.

Cell strains and culture
Dictyostelium discoideum cells were grown in adherent culture in plastic Petri dishes in HL5 medium (Formedium) at 22˚C. PIKfyvemutants were generated in both Ax2 and Ax3 backgrounds, with appropriate wild-type controls used in each case. Cells were transformed by electroporation and transformants selected in 20 μg/ml hygromycin (Invitrogen), 10 μg/ml G418 (Sigma) or 10 μg/ml blasticidin S. Apilimod was from United States Biological.
Growth in liquid culture was measured by seeding log phase cells in a 6 well plate and counting cells every 12 hours using a haemocytometer. Growth on bacteria was determined by plating~10 Dictyostelium cells on SM agar plates (Formedium) spread with a lawn of nonpathogenic KpGe K. pneumoniae [92].
Plaque assays were performed as previously described [93]. Briefly, serial dilution of Dictyostelium cells ( The Dictyostelium development was performed by spreading 10 7 amoebae on nitrocellulose filters (47 mm Millipore) on top of absorbent discs pre-soaked in KK2 (0.1 M potassium phosphate pH 6.1) and images were taken at 20 hours [94].

Gene disruption and molecular biology
PIKfyvecells in an Ax2 background were generated by gene disruption using homologous recombination. A blasticidin knockout cassette was made by amplifying a 5' flanking sequence of the PIKfyve gene (DDB_G0279149) (primers: fw-GGTAGATGTTTAGGTGGTGAAGT, rv-gatagctctgcctactgaagCGAGTGGTGGAATTCATAAAGG) and 3' flanking sequence (primers: fw-ctactggagtatccaagctgCCATTCAAGATAGACCAACCAATAG, rv-AGAATCAGAA-TAAACATCACCACC). These primers contained cross over sequences (in lower case) allowing a LoxP-flanked blasticidin resistance cassette (from pDM1079, a kind gift from Douwe Veltman) to be inserted between the two arms.
For PIKfyve gene disruption in an Ax3 background a knockout cassette was constructed in pBluescript by sequentially cloning fragment I (amplified by TAGTAGGAGCTCGGATCCG GTAGATGTTTAGGTGGTGAAGTTTTACCAAC and TAGTAGTCTAGACGAGTGGTGG AATTCATAAAGGTACGTTCAT) and fragment II (amplified by TAGTAGAAGCTTCCAT TCAAGATAGACCAACCAATAGTAGTCCTGC and TAGTAGGGTACCGGATCCCAGT GTGTAAATGAGAATCAGAATAAACATCACC). The blasticidin resistance gene was inserted between fragment I and II as an XbaI-HindIII fragment derived from pBSRδBam [95]. Both constructs were linearised, electroporated into cells and colonies were screened by PCR. Genomic DNA was extracted from lysing 1 x 10 6 cells in 100 μl lysis buffer (10mM Tris-HCl pH8.0, 50 mM KCl, 2.5mM MgCl1, 0.45% NP40, 0.45% Tween 20) and addition of 2ul of 20mg/ml Proteinase K (NEB) for 5 minutes incubation at room temperature. Proteinase K was inactivated at 95˚C for 10 minutes prior to PCR analysis. RT-PCR analysis was performed as previously described [96]. Briefly, total RNA was extracted from 1 x 10 7 cells using the Roche High Pure total RNA isolation kit. The Thermo RevertAid first strand synthesis kit was then used to synthesise cDNA from 2 μg RNA using random hexamer primers, and used as template for PCR analysis of transcription. Primers used for gDNA and cDNA PCR screens are listed in S1 Table. After validation, the Ax2-derived PIKfyve mutant strain was designated JSK01, and the two independent Ax3-derived mutants JSK05 and 06.

Microscopy and image analysis
Fluorescence microscopy was performed on a Perkin-Elmer Ultraview VoX spinning disk confocal microscope running on an Olympus 1x81 body with an UplanSApo 60x oil immersion objective (NA 1.4). Images were captured on a Hamamatsu C9100-50 EM-CCD camera using Volocity software by illuminating cells with 488 nm and 594 nm laser lines. Quantification was performed using Image J (https://imagej.nih.gov).
To image PI(3)P dynamics, cells were incubated in HL5 medium at 4˚C for 5 mins before addition of 10 μl of washed 3 μm latex beads (Sigma LB30) and centrifugation at 280 x g for 10 seconds in glass-bottomed dishes (Mat-Tek). Dishes were removed from ice and incubated at room temperature for 5 mins before imaging. Images were taken every 30 s across 3 fields of view for up to 30 mins.
V-ATPase recruitment and acidification was performed using Saccharomyces cerevisiae labelled with pHrodo red (Life Technologies) as per the manufacturers instructions. Dictyostelium cells in HL5 medium were incubated with 1x10 7 yeast per 3 cm dish, and allowed to settle for 10 mins before the medium was removed and cells were gently compressed under a 1% agarose/HL5 disk. Images were taken every 10 s across 3 fields of view for up to 20 mins. Yeast particles were identified using the "analyse particles" plugin and mean fluorescence measured over time. V-ATPase recruitment was measured as the mean fluorescence within a 0.5 μm wide ring selection around the yeast. The signal was then normalised to the initial fluorescence after yeast internalisation for each cell.

Endocytosis and exocytosis
To measure endocytosis, Dictyostelium at 5 × 10 6 cells/ml were shaken at 180 rpm for 15 mins in HL5 before addition of FITC dextran to a final concentration of 2 mg/ml (molecular mass, 70 kDa; Sigma). At each time point 500 μl of cell suspension were added to 1 ml ice-cold KK2 on ice. Cells were pelleted at 800 x g for 2 mins and washed once in KK2. The pellet was lysed in 50 mM Na 2 HPO 4 pH 9.3 0.2% Triton X-100 and measured in a fluorimeter. To measure exocytosis, cells were prepared as above and incubated in 2 mg/ml FITC-dextran overnight. Cells were pelleted, washed twice in HL5 and resuspended in HL5 at 5 × 10 6 cells/ml. 500 μl of cell suspension were taken for each time point and treated as described above.

Phagocytosis and phagosomal activity assays
Phagocytosis of E. coli was monitored by the decrease in turbidity of the bacterial suspension over time as described [100]. An equal volume of 2 × 10 7 Dictyostelium cells was added to a bacterial suspension with an OD 600 nm of 0.8, shaking at 180 rpm, and the decrease in OD 600 nm was recorded over time. Phagocytosis of GFP-expressing M. smegmatis and 1 μm YG-carboxylated polystyrene beads (Polysciences Inc.) was previously described [55,101]. 2 × 10 6 Dictyostelium/ml were shaken for 2 hours at 150 rpm. Either 1 μm beads (at a ratio of 200:1) or M. smegmatis (multiplicity of infection (MOI) 100) were added, 500 μl aliquots of cells were taken at each time point and fluorescence was measured by flow cytometry [55].
Phagosomal pH and proteolytic activity were measured by feeding cells either FITC/TRITC or DQgreen-BSA/Alexa 594-labelled 3 μm silica beads (Kisker Biotech) [55]. Briefly, cells were seeded in a 96 well plate before addition of beads and fluorescence measured on a plate reader over time. pH values were determined by the ratio of FITC to TRITC fluorescence using a calibration curve and relative proteolysis was normalised to Alexa594 fluorescence. To measure proteolytic activity in cell lysates 4 x 10 7 cells/ml were resuspended in 150 mM potassium acetate pH 4.0 and lysed by 3 cycles of freeze/thaw in liquid nitrogen. After pelleting cell debris at 18,000 x g for 5 minutes at 4˚C, 100 μl of lysate was added per well. Proteolytic activity was measured on a plate reader in triplicate, as described above, using 1 x 10 8 DQ-BSA/ Alexa594 beads per cell. A 5 x final concentration of HALT protease inhibitor cocktail (Life Technologies) was added to samples as a negative control.

Phagosome isolation and blotting
Dictyostelium phagosomes were purified at different stages in maturation after engulfment of latex beads as previously described [102]. Briefly 10 9 cells per timepoint were incubated with a 200-fold excess of 0.8 μm diameter beads (Sigma) first in 5 ml ice-cold Soerensen buffer containing 120 mM sorbitol (SSB) pH 8 for 5 minutes, then added to 100 ml room-temperature HL5 medium in shaking culture (120 rpm) for 5 (first time point) or 15 minutes to allow phagocytosis (pulse). Engulfment was stopped by adding cells to 300ml ice-cold SSB and centrifugation. After washing away non-engulfed beads, cells were again shaken in room-temperature HL5 for the times indicated (chase) to allow maturation. At each time point, maturation was stopped using ice-cold SSB as above, and cells pelleted. Phagosomes were purified as in [58], after homogenization using 10-passages through a ball homogeniser (void clearance 10 μm). The homogenate was incubated with 10 mM Mg-ATP (Sigma) for 15 minutes before loading onto a discontinuous sucrose gradient. Phagosomes were collected from the 10-25% interface, normalised by light scattering at 600 nm and analysed by Western blot. Antibodies used were anti-VatA mAB 221-35-2 (gift from G. Gerisch), anti-VatM mAb N2 [103]; rabbit anti-cathepsin D [104] and anti-Abp1 [105]. All blots were processed in parallel from the same lysates with identical exposure and processing between cell lines.

Bacteria killing assay
Killing of GFP-expressing K. pneumoniae was measured as previously described [61]. Briefly, 10 μl of an overnight culture of bacteria in 280 μl HL5 was placed in a glass-bottomed dish and allowed to settle before careful addition of 1.5 ml of a Dictyostelium cell suspension at 1 × 10 6 cells/ml. Images were taken every 20 s for 40 min at 20x magnification. Survival time was determined by how long the GFP-fluorescence persisted after phagocytosis.

Western blotting
Ax2 or PIKfyvecells expressing GFP-VatM or VatB-GFP were analysed by SDS-PAGE and Western blot using a rabbit anti-GFP primary antibody (gift from A. Peden) and a fluorescently conjugated anti-rabbit 800 secondary antibody, using standard techniques. The endogenous biotinylated mitochondrial protein Methylcrotonoyl-CoA Carboxylase 1 (MCCC1) was used as a loading control using Alexa680-conjugated Streptavadin (Life Technologies) [106].

Legionella infection assays
The following L. pneumophila Philadelphia-1 strains were used: virulent JR32 [107], the isogenic ΔicmT deletion mutant GS3011 lacking a functional Icm/Dot type IV secretion system [74], and corresponding strains constitutively producing GFP [75]. L. pneumophila was grown for 3 days on charcoal yeast extract (CYE) agar plates, buffered with N-(2-acetamido)-2-aminoethane-sulfonic acid (ACES) [108]. For infections, liquid cultures were inoculated in AYE medium at an OD 600 of 0.1 and grown for 21 h at 37˚C (post-exponential growth phase). To maintain plasmids, chloramphenicol was added at 5 μg/ml. Uptake by D. discoideum, intracellular replication or killing of GFP-producing L. pneumophila was analyzed by flow cytometry as described [72]. Exponentially growing amoebae were seeded onto a 24-well plate (1 × 10 6 cells/ml HL5 medium per well) and allowed to adhere for at least 1-2 h. L. pneumophila grown for 21 h in AYE medium was diluted in HL5 medium and used to infect the amoebae at an MOI of 50. The infection was synchronised by centrifugation (10 min, 500 x g), and infected cells were incubated at 25˚C for 30 min before extracellular bacteria were removed by washing twice with SorC (2 mM Na 2 HPO 4 , 15 mM KH 2 PO 4 , 50 μM CaCl 2 , pH 6.0). Infected amoebae were detached by vigorously pipetting and fixed (PFA 2%, sucrose 125 mM, picric acid 15%, in PIPES buffer, pH 6.0), and 1 × 10 4 amoebae per sample were analyzed using a using a LSR II Fortessa analyser. The GFP fluorescence intensity falling into a Dictyostelium scatter gate was quantified using FlowJo software (Treestar, http://www.treestar.com).
Alternatively, intracellular replication of L. pneumophila in D. discoideum was quantified by determining colony forming units (CFU) in the supernatant as described [72,109]. Exponentially growing amoebae were washed and resuspended in MB medium (7 g of yeast extract, 14 g of thiotone E peptone, 20 mM MES in 1 l of H 2 O, pH 6.9). Amoebae (1 × 10 5 cells per well) were seeded onto a 96-well plate, allowed to adhere for at least 2 h, and infected at an MOI of 0.1 with L. pneumophila grown in AYE medium for 21 h and diluted in MB medium. The infection was synchronised by centrifugation, and the infected amoebae were incubated at 25˚C. At the time points indicated, the number of bacteria released into the supernatant was quantified by plating aliquots (10-20 μl) of appropriate dilutions on CYE plates. Intracellular bacteria were also quantified by counting CFU after lysis of the infected amoebae with saponin. At the time points indicated host cells were detached by vigorous pipetting and lysed by incubation with saponin (final concentration-0.8%) for 10 min. The number of bacteria released into the supernatant was quantified by plating 20 μl aliquots of appropriate dilutions on CYE plates.

Statistics
Statistical analysis was performed using Graphpad Prism 7 software. Biological replicate numbers and statistical tests used for each experiment are detailed in each figure legend. A p-value of <0.05 were deemed significant with � indicating p<0.05, �� p<0.01 and ��� p<0.005 throughout.