Mycobacterium tuberculosis reprograms waves of phosphatidylinositol 3-phosphate on phagosomal organelles.

The potent human pathogen Mycobacterium tuberculosis persists in macrophages within a specialized, immature phagosome by interfering with the pathway of phagolysosome biogenesis. The molecular mechanisms underlying this process remain to be fully elucidated. Here, using four-dimensional microscopy, we detected on model phagosomes, which normally mature into phagolysosomes, the existence of cyclical waves of phosphatidylinositol 3-phosphate (PI3P), a membrane trafficking regulatory lipid essential for phagosomal acquisition of lysosomal characteristics. We show that mycobacteria interfere with the dynamics of PI3P on phagosomal organelles by altering the timing and characteristics of the PI3P waves on phagosomes. The default program of cyclical PI3P waves on model phagosomes is composed of an initial stage (phase I), represented by a strong PI3P burst occurring only upon the completion of phagosome formation, and a subsequent stage (phase II) of recurring PI3P waves on maturing phagosomes with the average periodicity of 20 min. Mycobacteria alter this program in two ways: (i) by inducing, in a cholesterol-dependent fashion, a neophase I* of premature PI3P production, coinciding with the process of mycobacterial entry into the macrophage, and (ii) by inhibiting the calmodulin-dependent phase II responsible for the acquisition of lysosomal characteristics. We conclude that the default pathway of phagosomal maturation into the phagolysosome includes temporally organized cyclical waves of PI3P on phagosomal membranes and that this process is targeted for reprogramming by mycobacteria as they prevent phagolysosome formation.

Phagocytosis is a fundamental biological process essential for proper tissue homeostasis, development, elimination of invading microorganisms, and processing of antigens for presentation to the immune system. Precisely how a phagosome forms and transforms into a phagolysosome has been the subject of evolving views (1)(2)(3)(4)(5)(6)(7)(8)(9). Maturation of the phagosome into the phagolysosome is a multistage membrane trafficking and pro-tein sorting process, and a significant portion of the current knowledge regarding phagosomal maturation comes from studies of phagosomes containing intracellular pathogens (10 -16).
Because intracellular pathogens interfere with various steps in phagolysosomal biogenesis, they have been employed as convenient tools to dissect different aspects of phagosomal maturation (10 -12). In general, upon phagocytosis or invasion, an intracellular pathogen that specializes in parasitizing host cells can follow one of the following routes: (i) lysis of the phagosome and escape into the cytosol (17); (ii) unscheduled fusion with intracellular compartments such as the endoplasmic reticulum (18); (iii) arrest of phagosomal maturation (19); or (iv) survival in lysosome-like or autophagic compartments (20). Mycobacterium tuberculosis parasitizes macrophages by arresting the default process of phagosomal maturation into the phagolysosome (19,21). The major defining features of the mycobacterial phagosome are the fewer H ϩ ATPase molecules with the attendant deficit in luminal acidification (22), absence of mature lysosomal hydrolases (23,24) and the mannose 6-phosphate receptor (23)(24)(25), lack of specific membrane tethering and docking molecules (23), and extended accessibility to transferrin receptors (24).
The maturation arrest of the M. tuberculosis phagosome has been linked to a block between the stages controlled by the early endosomal GTPase Rab5 and its late endosomal counterpart Rab7 (26). A recent follow-up study has examined the possible roles of Rab5 effectors in phagosome maturation (27), indicating that the majority of Rab5-interacting partners are present on mycobacterial phagosomes with the exception of the early endosomal autoantigen 1 (EEA1) 1 (27). EEA1 is a membrane-tethering molecule that is recruited to phagosomal surfaces via a bipartite mechanism, which comprises a binding process to Rab5 and association with phosphatidylinositol 3-phosphate (PI3P) on target membranes (28). EEA1 is required for the PI3P-dependent delivery of hydrolases and the H ϩ ATPase V o subunit from the trans-Golgi network (TGN) to the phagosome (23).
gosome and phagolysosome biogenesis. PI derivatives regulate diverse trafficking processes by affecting the localization and activity of effector proteins containing lipid binding modules such as C2, ENTH, FERM, FYVE, PH, and PX domains (29). A subset of PI derivatives and lipid-binding proteins also plays a role in phagocytosis and phagolysosome biogenesis. Phagocytosis triggers production of phosphatidylinositol 4,5-bisphosphate (PIP 2 ), which recruits actin to the phagocytic cup; this is followed by disappearance of PIP 2 as the phagosome seals (9,30), concomitant with activation of phospholipase C and conversion of PIP 2 to phosphatidylinositol 3,4,5-trisphosphate (PIP 3 ) (31,32), the latter being necessary for the Myosin X-dependent extension of pseudopods around the ingested particle (3). Another phosphoinositide, PI3P, has recently received more attention, because it is important for those stages in phagosomal maturation that follow the initial particle uptake (27,32,33).
PI3P is the product of the phosphatidylinositol 3-kinase (PI3K) hVPS34 (in yeast, Vps34p) (28,34), which is one of the Rab5 effectors (35), and is thought to localize predominantly to endomembranes (36). PI3P affects localization and function of proteins that contain FYVE, PH, or PX domains (37,38) and controls intracellular trafficking and signal transduction. Such functions include autophagy (39), early endosomal homotypic fusion (28), transit of internalized receptors to a late endocytic compartment (40), formation of endosomal multivesicular bodies involved in termination of signaling events (41), and a subset of transport pathways from the TGN (23,42,43). In this work, we show, using phagosomal organelles as a model system, that PI3P formation on endomembranes is not a uniform event and that it functions as a timer within a programmed pathway of phagolysosome biogenesis. We furthermore show that M. tuberculosis reprograms PI3P waves on phagosomal membranes as a strategy for blocking phagosomal conversion into the phagolysosome, underscoring the significance of the PI3P cycling on organellar membranes.

EXPERIMENTAL PROCEDURES
Cell and Bacterial Culture-Murine bone marrow-derived macrophages were isolated from the femurs of C57/BL6 mice and maintained in DME supplemented with 4 mM L-glutamine, 20% FBS, and macrophage-colony stimulating factor for 7 days prior to plating onto coverslips. The murine macrophage-like cell line RAW 264.7 was maintained in DME supplemented with 4 mM L-glutamine and 10% FBS. COS-7 cells were grown as described previously (44). M. tuberculosis var. bovis BCG harboring phsp60-DsRed was grown in 7H9 broth on a rotating platform and homogenized to generate a single cell suspension (Via et al. (26)).
Plasmid Constructs and Transfection-The plasmid constructs pPLCѨ PH-EGFP and pBtk PH-EGFP were from T. Balla (National Institutes of Health, Bethesda, MD), pFAPP PH-EGFP from M. A. De Matteis (Italy), p2xFYVE-EGFP from H. Stenmark (The Norwegian Radium Hospital, Oslo, Norway), and p40 phox PX-EGFP from M. Yaffe (Massachusetts Institute of Technology, Cambridge, MA). For transfection, 5 ϫ 10 6 RAW 264.7 cells were resuspended in a nucleoporator buffer supplied by the manufacturer (Amaxa Biosystems) with 2 g of plasmid DNA. Cells were nucleoporated according to the manufacturer's protocol. The cells were plated on 25-mm diameter coverslips and allowed to express the construct for 18 h prior to imaging experiments.
Preparation of Texas Red-labeled Latex Beads and M. tuberculosis var. bovis BCG-Streptavidin-conjugated polystyrene beads were washed with 50 mM NaHCO 3 , pH 9.6, and incubated in PBS with 5 mg/ml Texas Red-X (Molecular Probes) for 1-2 h. Beads were subsequently washed (5ϫ) with PBS and then opsonized with 100% FBS for an additional 30 min. M. tuberculosis var. bovis BCG were washed in PBS/0.5% Tween 20 and incubated in 5 mg/ml Texas Red-X in PBS for 1-2 h. Labeled mycobacteria were washed (5ϫ) with PBS and then opsonized with 100% FBS for an additional 30 min.
Treatment with Inhibitors-CaM was inhibited by treating RAW 264.7 or COS-7 cells with 25 or 100 M W7, respectively. W7 was added after the first PI3P burst (phase I) was ascertained following phagocytosis. CaMKII was inhibited using KN62 (2 M). KN62 was added prior to macrophage infection with latex beads and throughout the duration of the experiment. Cholesterol was extracted, and its removal was verified by using the methyl-␤-cyclodextrin (M␤CD) method (45). RAW 264.7 cells were incubated in 10 mM M␤CD in serum-free media for 1 h prior to infection with BCG. Cells were kept in M␤CD during infection and subsequent imaging.
Labeling with Fluid Phase Endocytic Tracers-Transfected RAW 264.7 cells were incubated in 2.5 mg/ml Texas Red-dextran (M r 10,000, Molecular Probes). For early endosomal labeling, cells were first infected with particles, washed (3ϫ) with PBS, and then subsequently incubated in Texas Red-dextran for 5 min. Cells were washed (6ϫ) with PBS prior to viewing under the microscope. For late endosomal labeling, cells were first incubated with Texas Red-dextran for 5 min, washed, and then infected with latex beads. For labeling of lysosomal compartments, cells were pulsed with dextran for 1 h, followed by a 2-h chase, and then infected with latex beads.
Immunofluorescence Microscopy-Bone marrow-derived macrophages grown on 25-mm glass coverslips were fixed with 1% paraformaldehyde followed by membrane permeabilization using 0.2% saponin. Permeabilized cells were incubated with the probe 2xFYVE-GST (10 g/ml, Echelon, Inc.) followed by anti-GST antibody conjugated to Alexa 488 (Molecular Probes). Epifluorescence microscopy was performed using an IX70 Olympus microscope. Images were collected and processed using LSR Esprit software. Colocalization of the PI3P probe with phagosomes was determined using methods of unbiased counting (27).
Four-dimensional Confocal Microscopy-Live cell confocal microscopy was performed on a rotating disc confocal UltraView LCI microscope system (PerkinElmer Life Sciences) using argon or argon/krypton lasers and appropriate excitation dichroic and emission filters for red and green fluorescence. Transfected RAW 264.7 cells grown on 25-mm glass coverslips were allowed to phagocytose latex beads or M. tuberculosis var. bovis BCG. Infection was achieved by centrifugation of particles onto macrophages adherent to coverslips at 800 rpm for 5 min. Coverslips were transferred immediately to DME in a perfusion chamber (Harvard Apparatus) set at a constant temperature of 37°C. Transfected cells chosen for observation were selected based on the presence of moving (Brownian motion) particles to be phagocytosed by adjacent macrophages. Rapid z-stack (controlled by using a piezoelectric device; millisecond speed range) of confocal sections spanning a total thickness of 10 -15 m were taken to ensure that all sides of the phagocytosed particles were examined for each time point regardless of object movement in or out of any given single plane of focus. Images were collected and processed using an UltraView system. Data were collected as a time lapse sequence of two separate channel/color z-stacks and processed for three-dimensional rendering (particle internalization), x-y plane projection of a range of z stacks (maximum intensity for rendering and mean-intensity for quantitative analysis), and a time-lapse sequence for supplemental movies. All procedures followed the previously published approaches for four-dimensional microscopy (46 -50).
Particle Entry Determination-The entry of latex beads into the macrophages was identified by a characteristic brief immobilization, followed by a centripetal movement during internalization (confirmed by three-dimensional reconstruction). Mycobacterial entry was identified by immobilization of the bacilli upon their firm attachment to the macrophage surface, followed by the centrifugal movement of macrophage membrane along the bacterial cells (see Supplemental Movie 1 showing PIP 2 -labeled membranes in the process of engulfing mycobacteria). Latex bead and mycobacterial internalization and intracellular localization were additionally confirmed by differentially staining extracellular and intracellular objects (by permeabilizing or not permeabilizing the plasma membrane) in independent correlative studies.
Quantitative Analysis of Four-dimensional Confocal Microscopy Data-The previously published approaches (46 -50) for analyzing three-dimensional data by projections in the x-y plane built into the UltraView software were applied. Maximum intensity projections and mean-intensity projections were used for rendering and quantitative analysis, respectively (46).

Diminished Levels of PI3P on Mycobacterial Phagosomes-
The previously reported changes in EEA1 recruitment to M. tuberculosis phagosomes (MBPs) (27) led us to hypothesize that PI3P profiles may be affected on mycobacterial phagosomes, because EEA1 binds to target membranes through association of its PI3P-binding FYVE domain with the PI3P on organellar membranes. To compare the PI3P levels on model, latex bead phagosomes (LBPs) with those on mycobacterial phagosomes, we examined by immunofluorescence microscopy the PI3P staining in macrophages using a PI3P affinity probe (2xFYVE-GST) (36). Murine bone marrow-derived macrophages were allowed to phagocytose complement-opsonized 1-m latex beads or M. tuberculosis var. bovis BCG for 10 min followed by 20-or 60-min chase periods, after which the cells were fixed and processed for immunostaining. Under these conditions, LBPs were positive for the PI3P probe ( Fig. 1). In contrast, MBPs displayed reduced association with the PI3P probe ( Fig. 1). Identical results were obtained with murine macrophage cell lines J774 and RAW264.7. Thus, the previously reported phenomenon of altered EEA1 recruitment to MBPs (27) can be, at least in part, explained by altered levels of PI3P on mycobacterial phagosomes. This is in keeping with our earlier demonstration that treatments with phosphatidylinositol 3-kinase inhibitors prevent phagosomal maturation (23).
Four-dimensional Live Microscopy Analysis of PI3P Dynamics: Multiple Waves of PI3P on Phagosomes-Immunostaining of fixed cells has technical limitations in precluding a reliable investigation of the very early time points of phagocytosis or transient events at any other stage. To examine in a timeresolved manner the different stages of phagocytosis and phagolysosome biogenesis, we resorted to live microscopy observations. An approach was employed that afforded low photocytotoxicity and low photobleaching imaging by using a rotating disk UltraView confocal microscope. Time-lapse studies using the UltraView system allowed monitoring of both the very early phagocytic events and the phagosome maturation progression over long periods of time. In addition, z-sections encompassing complete thickness of the cell were obtained at millisecond speeds for each time point (see "Experimental Procedures" for details). This allowed three-dimensional tracking of objects as they moved in and out of a single focal plane. For analysis and presentation, z-sections were collapsed following published procedures (46) into a single image (projection), af-fording four-dimensional investigations of a volume over time.
RAW 264.7 macrophages were nucleoporated with DNA encoding GFP fused to either the p40 phox PX domain (P40PX-EGFP) or to two tandem FYVE domains from Hrs (2xFYVE-EGFP), both of which bind exclusively to PI3P (36,51). Complement-opsonized, fluorescent latex beads or M. tuberculosis var. bovis BCG were added to coverslips, and imaging was carried out by monitoring events from the onset of phagocytosis, indicated by the initiation of the uptake of a particle by the macrophage. The characteristic particle movement (see "Experimental Procedures"), further confirmed by three-dimensional image analysis, was scored as the time of entry (0 time point). LBPs began accumulating 2xFYVE-EGFP (and in separate experiments P40PX-EGFP) 5-10 min after the uptake.
The initial burst of PI3P on phagosomes was followed by its rapid decline. Unexpectedly, LBPs displayed repeated cycles of increase and decline in the amounts of the PI3P probe associated with the phagosome (Fig. 2, A-H, and Supplemental Movie 2). Upon examination of multiple profiles in over 10 independent experiments, two distinct phases became apparent with the standard phagosomes, LBPs. The plots of fluorescence data on the acquisition of PI3P (Fig. 3) illustrate the two distinct phases invariably associated with latex bead phagosomes. The first peak, strong in intensity, of probe association with LBPs, i.e. the initial burst of PI3P after phagocytosis, was designated as phase I (Fig. 3, A-C, red bar). Following the initial phase I burst, subsequent multiple waves of PI3P on LBPs were collectively designated as phase II (Fig. 3, A-C, green bar). In comparison to phase I PI3P bursts, those during phase II had a lower peak value of fluorescence intensity, with the average maximum amplitude corresponding to 41% (Ϯ 4%) of the phase I peak (corrected for a background photobleaching decline) and the average periodicity (time elapsed between consecutive peaks) of 1,183 Ϯ 193 s.
Mycobacteria Displace phase I to the Stage of Bacterial Entry-Unexpectedly, nascent mycobacterial phagosomes showed a transient PI3P positivity during live microscopy observa- tions. The initial phase of PI3P presence on MBPs occurred earlier than on LBPs, i.e. it was shifted to the time points associated with particle entry into the macrophage (Fig. 4 and Supplemental Movie 3). This early time shift by mycobacteria was defined as phase I* (Fig. 3, D-F). With latex beads, there was a 5-to 10-min lag time after the particles were fully inside the macrophages before the initial burst of PI3P could be detected on LBPs (Fig. 3 and 2B). In contrast, PI3P appeared on MBPs before the phagosomal cup closed around the mycobacteria, as parts of individual mycobacteria could be observed outside the macrophages during the initial PI3P appearance. In most cases, there was a zipper-like effect of PI3P appearance on MBPs (Fig. 4 and Supplemental Movie 3). This zipper-like progression of the section positive for PI3P was typically initiated at one end of the mycobacterial phagosome followed by a continued escalation of the PI3P-positive region up the tubular structure. Thus, PI3P progressively spread from one to the other end of the nascent mycobacterial phagosome. In conclusion, an ascending zipper-like rapid appearance and subsequent permanent loss of PI3P were observed on MBPs as they formed. In sharp con-trast, LBP acquisition of PI3P occurred only following the completion of the uptake process and phagosome closure.
Because phase I* PI3P formation was associated with the mycobacterial entry into the macrophages, and previous studies had shown that cholesterol depletion inhibited the entry of M. tuberculosis var. bovis BCG into the macrophages (52), we next tested whether cholesterol depletion affects phase I* PI3P formation (Fig. 5). Macrophages were incubated with 10 mM M␤CD for 1 h in serum-free medium, and complement-opsonized mycobacteria were added. In two independent experiments (n ϭ 5 total recorded BCG), none of the observed mycobacteria that became firmly associated with macrophages (immobilized on the surface of macrophages) were PI3P-posi- FIG. 4. Mycobacterial phagosomes acquire PI3P synchronously with bacterial entry. A-L, a sequence of UltraView images of a macrophage in the process of engulfing a mycobacterium. E-K, PI3P phase I* occurs on the plasma membrane as the mycobacterium is phagocytosed by the macrophage. Note the formation of a ring-like structure as mycobacteria enter the nascent phagosome, zippering action along the phagosome as it closes, and removal of PI3P shortly upon phagosome closure. L-P, PI3P disappears from mycobacterial phagosomes following phase I*; phase II PI3P is absent from these organelles. Arrows indicate the positions of the nascent mycobacterial phagosome. Images represent collapsed confocal optical sections corresponding to a total thickness of 4.6 m.
tive over the observation period of 1 h. In the untreated control, 100% (n Ͼ 20) of the bacilli that firmly associated with macrophages (signifying entry) became PI3P-positive. Both the entry process, as previously shown (52), and the phase I* PI3P formation, as shown here, depended on cholesterol. Thus, phase I* PI3P formation was associated with mycobacterial entry in the macrophage.
Absence of phase II on Mycobacterial Phagosomes-In addition to a shift in phase I on mycobacterial phagosomes to the very early stages coinciding with the bacterial entry (phase I*), the corresponding vacuoles, once formed, lacked the dynamic phase II consisting of multiple PI3P waves (Fig. 3, D-F). With some variation in the length of phase I*, the oscillatory increases and decreases of P40PX-EGFP and 2xFYVE-EGFP binding to phagosomal membrane, which were associated with phase II and routinely seen with latex bead phagosomes, were completely absent from mycobacterial phagosomes (Fig. 3). A quantification of PI3P positivity of LBPs and mycobacterial phagosomes is given in Fig. 6. The results show that: (i) 100% of LBPs (n ϭ 59) were positive for phase II PI3P in comparison to only 13% of M. tuberculosis var. bovis BCG-containing phagosomes (n ϭ 45) and (ii) 63% of the LBPs positive for phase II PI3P were tethered to a PI3P-positive profile (vesicle) at the time of imaging and 37% of LBPs did not have vesicles tethered and displayed either punctate (32%) or uniform, homogeneous (5%) staining.
Profiles of Other Phosphoinositides on LBPs and MBPs-Using additional probes, Btk PH-EGFP (PIP 3 -specific) (53), PLC␦PH-EGFP (PIP 2 -specific) (54), and FAPP PH-EGFP (PI4P-specific) (55), we also examined other phosphoinositides on LBPs and MBPs. The rationale behind these experiments was to determine whether potentially altered interconversions of other phosphoinositides might be responsible for the observed differences in PI3P dynamics between LBPs and MBPs.
The PIP 3 probe appeared only sparingly on LBPs (Fig. 7, A-C) but was well visible on MBPs (Fig. 7, J-L), decorating the mycobacterial phagosomal cup at the start of phagocytosis. PIP 3 accumulation was ablated once the phagosomal cup had sealed, regardless of whether the phagosome contained latex beads or mycobacteria. This is consistent with earlier observations made by Grinstein and colleagues (56) in that PIP 3 was lost from the nascent phagosome as soon as the phagosomal cup had closed. The PIP 2 probe PLC␦ PH-EGFP also appeared at the phagosomal cup in a localized manner prior to the phagosomal closure (Fig. 7, D-F) (9). We also noted that the PIP 2 probe lingered on a significant fraction of LBPs for prolonged periods of time (up to 2 h) even after the phagosomal cup had sealed, albeit such LBPs seemed to remain juxtaposed and possibly tethered to the plasma membrane. Similarly, PIP 2 was found on forming mycobacterial phagosomes by first accumulating and then disappearing from nascent phagosomal cups (Fig. 7, M-O). During occasional frustrated phagocytosis of mycobacteria (ascribed to a convoluted particle geometry), PIP 2 disappeared from the phagosomal cup prior to phagosomal closure (Supplemental Movie 1). As a control, FAPP PH-EGFP, a probe for PI4P that is abundant on the Golgi membranes, did not appear on LBPs or MBPs (Fig. 7, G-I and P-R). Based on these analyses we concluded that PI3P anomalies on mycobacterial phagosomes are not an indirect consequence of differences in conversions of other phosphoinositides, PIP 2 , PIP 3 , and PI4P, tested here.
Mechanism of Phase II Increase in PI3P Levels on LBPs-In 37% of events (Fig. 6), the phase II-multiple PI3P waves were attributed to de novo synthesis of PI3P on latex bead phagosomes, because LBPs displayed increased fluorescence of the probes without prior or concomitant association with any surrounding vesicles. In the remaining 63% of events (Fig. 6), asymmetric bursts of PI3P could be observed upon a transient contact or fusion with PI3P-positive vesicles tethered to LBPs. To examine the nature of the PI3P-positive vesicles interacting with latex bead phagosomes, we labeled endosomes with a fluid phase tracer using dextran-Texas Red. Endosomal compartments containing dextran were positive for PI3P immediately after pinocytosis (Fig. 8, A-C). 25-min-chased dextran-labeled endosomes (5-min pulse and 20-min chase), which were positive for PI3P, were found to tether to LBPs (Fig. 8, D-F). Dextran-containing compartments that were chased for 50 min did not tether to LBPs (Fig. 8, G-I). When dextran was chased into the lysosomal organelles (1-h pulse and 3-h chase), no vesicles that were both PI3P-and dextran-positive could be seen to tether to LBPs (Fig. 8, J-L). Thus, some of the PI3Ppositive vesicles that interacted with LBPs were 25-min dextran-positive endosomes. In conclusion, LBPs acquired PI3P by two mechanisms: (i) by de novo synthesis and (ii) by contact or exchange of membranes with PI3P-positive endosomal profiles. It should be noted that, by collecting z-stacks for each time point and making x-y projections of the collected z-stacks for quantitative analysis, we bypassed completely the problem of particles moving into and out of view in a single focal plane.
Generation of Phase II PI3P Waves on Phagosomes Requires Calmodulin and Calmodulin-dependent Kinase II-It has been recently shown that calmodulin (CaM) and calmodulin-dependent kinase II (CaMKII) are localized to the phagosome (57) and can recruit hVPS34 (33). Thus, we tested whether CaM and CaMKII control changes in PI3P levels on phagosomes observed in our studies. Two inhibitors, W7 and KN62 (specific for CaM and CaMKII, respectively), were used to study the effects on oscillatory PI3P bursts. W7, a naphthalene sulfonamide specific for inhibiting CaM (58), had to be added only after the initial stages of phagosome formation were completed. This compound inhibits uptake by phagocytosis and could not be used to study phase I. However, when W7 was added following phagocytosis (after the completion of a phase I event was ascertained by microscopy), W7 inhibited phase II PI3P generation, as shown in Fig. 9, G-L (W7-treated cells; see also Supplemental Movie 4), versus Fig. 9, A-F (untreated control).
KN62 is a quinolone sulfonamide specific for CaMKII in that it prevents its binding to CaM (59). Because KN62 did not inhibit phagocytosis, it was added during the entire course of the experiment, including particle uptake. Fig. 9 (panels M-R) shows that KN62 did not prevent phase I burst but did prevent phase II PI3P waves (Fig. 9, panels M-R showing KN62treated cells, versus Fig. 9, panels A-F with untreated control). Whereas W7 eliminated PI3P from all PI3P-positive endomembranes (Supplemental Movie 4), KN62 prevented only the formation of PI3P on LBPs. KN62 also prevented tethering/docking of PI3P-positive vesicles with LBPs. In six independent experiments (n ϭ 12 for the total number of bead phagosomes observed) with W7, 100% of bead phagosomes did not enter the phase II. In 13 independent experiments (n ϭ 59 for the total number of bead phagosomes observed) with KN62, 80% of latex bead phagosomes showed phase I but no phase II, 5% displayed both phase I and phase II, and 15% showed neither phase I nor phase II. In the untreated control, of phagosomes that displayed the phase I, all subsequently entered phase II. Thus, generation of PI3P on phagosomes during phase II depends on CaM and CaMKII. As previously demonstrated (21,57), CaM and CaMKII are critical for phagosomal maturation. Thus, the interference of M. tuberculosis var. bovis BCG with the CaMKIIdependent oscillations on phagosomes is a crucial event in the M. tuberculosis phagosome maturation arrest.

DISCUSSION
In this study, we report the previously unappreciated waves of PI3P on maturing phagosomes in macrophages. This process represents an intracellular trafficking/signaling program organized into two major phases, with a strong phase I burst of PI3P followed by multiple phase II waves of waxing and waning PI3P levels. M. tuberculosis inhibits the phase II of the programmed PI3P schedule, underlying the previously described phenomenon of mycobacterial interference (23,27) with PI3K-dependent processes controlling phagosomal maturation into the phagolysosome (27,32). The findings presented here, combined with the previously reported observations (23,27,33), explain the mechanisms underlying the marquee virulence determinant of M. tuberculosis, namely inhibition of phagosomal maturation, that enables this pathogen to infect over a billion people worldwide.
The existence of PI3P waves is consistent with the notion that conversion of phagosomes into phagolysosomes is a preprogrammed, evolutionarily tuned, and automated pathway in phagocytic cells. The generation of PI3P, which in this work was shown to be a temporally ordered series of discrete events on maturing phagosomes, is likely to dictate the initiation and termination of the recruitment of PI3P-binding effectors. One such effector, already shown to play a significant role in phagosomal maturation, is the membrane-tethering and SNARE-organizing factor EEA1 (27). EEA1 is required for delivery of lysosomal hydrolases and the V o domain of the vacuolar H ϩ -ATPase, which are essential components of the phagolysosome (23).
The generation of multiple PI3P waves is regulated by CaM and CaMKII in macrophages. CaM and CaMKII have been recently shown to recruit the PI3K hVPS34 (33), and this The data presented here, combined with the previous reports demonstrating the role of CaM in phagosome maturation (21,33,57) and showing that CaM is affected by M. tuberculosis (57), indicate that a significant portion of the M. tuberculosis phagosome maturation block is due to CaM-dependent events that control phase II PI3P waves.
The effects of CaM on PI3P are most likely cell-type specific, because a recent study has indicated that addition of W7 to COS-7 cells did not remove or redistribute the 2xFYVE-EGFP probe from the endosomes to the cytosol (44). To address the apparent discrepancy between our observations and the report by Lawe et al. (44), we treated COS-7 cells and macrophages with 100 M W7 in parallel (concentrations used by Lawe et al. (44)). Our results confirmed that, in COS-7 cells, levels of PI3P were not abrogated upon W7 treatment, because the 2xFYVE-EGFP probe remained associated with endomembranes (data not shown). Nevertheless, W7 treatment removed the 2xFYVE-EGFP probe from the endomembranes in macrophages. These results are best explained by a conclusion that the role of CaM in the regulation of PI3P levels is cell type-specific, and, al-though it has no effect on PI3P in COS-7 cells, it is critical for PI3P generation on phagosomes in macrophages.
The increases of PI3P levels on phagosomes following its decline within a given cycle can be explained either by the action of hVPS34 (21,27,32,33) or by fusion with PI3P-positive organelles. The process of PI3P loss from phagosomal membrane is not known at present but may involve the recruitment of a PI3P phosphatase. For example we observed by fourdimensional microscopy that the PI3P phosphatases MTM1-GFP (60) and MTMR3-GFP (61) are recruited to phagosomes (data not shown). Alternatively, enzymes converting PI3P into phosphatidylinositol 3,5 bisphosphate, such as PIKfyve (62), may, in combination with a mammalian Fig4 phosphatase-like protein, which dephosphorylates phosphatidylinositol 3,5 bisphosphate at the 5 position (63), cause oscillations of PI3P levels. The continuous production and breakdown of PI3P (or its masking by conversion into phosphatidylinositol 3,5 bisphosphate) on phagosomes most likely dictate orderly interactions with other compartments and trafficking pathways governing phagosomal maturation.
Previous studies have shown that mycobacterial phagosomes display a reduced association with EEA1 (Fratti et al. (27)) but at the same time show increased presence of Rab5 (Via et al. (26)). Based on these opposing properties, and the fact that both PI3P and Rab5 associate with EEA1 (Simonsen et al. (28) and Patki et al. (67)), it was not possible to predict with certainty the status of PI3P on MBPs. Although the experiments with fixed cells have initially indicated that PI3P is not generated on MBPs, the live confocal microscopy experiments revealed that PI3P is present on MBPs at an unexpected, very early time point associated with mycobacterial uptake. However, PI3P is excluded from phagosomes harboring mycobacteria during phase II, when, as previously shown (23), lysosomal constituents are delivered from the TGN.
The existence of waves of PI3P suggests that controlling PI3P generation in space and time can be used as a timer in organellar maturation and may constitute a programmed sequence of events. As shown in this work, CaM-dependent downstream events are necessary for the PI3P waves on phagosomes, and thus Ca 2ϩ may provide the pacemaker for these processes. It is worth mentioning that oscillations of PI3P have previously been observed in the context of the Salmonellacontaining vacuole (64). In that study (64), however, the authors attributed the PI3P oscillations to a putative Salmonella phosphatase that they proposed specifically breaks down PI3P. As shown here with latex bead phagosomes, this is entirely a function of the host cell. Our data may also help explain a recently reported apparent paradox (65) suggesting that phagosomes, even within the same macrophages, behave differently with regards to PI3P levels, i.e. that some phagosomes have PI3P while others do not. Our data show that a phagosome undergoes cyclical changes in PI3P levels and that PI3P positivity can be missed when applying conventional confocal microscopy methods.
What controls phase I? Here, we have demonstrated that the PI3P phase II oscillations specifically depend on the CaM pathway but that the initial burst of PI3P (phase I) is CaM-independent. Although we could not directly test whether CaM affected phase I, because W7 inhibited phagocytosis, interference with a downstream CaM effector CaMKII, which is also necessary for hVPS34 action (33), did not prevent phase I from occurring. This leads us to conclude that only the phase II PI3P formation depends on the CaM pathway for hVPS34 activation (33). Instead, Rab5 alone may control phase I, because Rab5 is proximal to the membrane trafficking events occurring at the plasma membrane and within the nascent early endosomes. FIG. 8. Analysis of the identity of PI3P-positive vesicles tethered to latex bead phagosomes. RAW 264.7 cells were transfected with either 2xFYVE-EGFP or P40PX-EGFP and allowed to endocytose Texas Red-dextran (A, D, G, and J) or phagocytose latex beads (D, G, and J) for the times indicated. A-C, Texas Red-dextran was added to cells for 5 min, excess dextran removed by washing and endocytosed dextran viewed immediately under the microscope. Arrows in A-C point to a PI3P-positive pinosome containing freshly endocytosed Texas Reddextran. D-F, latex beads were spun down onto adherent cells for 5 min at 1000 rpm, and unphagocytosed latex beads were removed by washing. Texas Red-dextran was then added for 5 min, excess was removed by washing, and the cells were subjected to imaging. The dashed circles outline latex beads. Arrows point to PI3P-positive, dextran-positive vesicles tethered to latex bead phagosomes. Time of imaging corresponds to the time of chase indicated. G-I, Texas Red-dextran was added to adherent cells for 10 min, and excess was removed by washing. Latex beads were spun for 5 min onto the adherent cells on coverslips viewed under the microscope. Time of imaging corresponds to the chase time indicated. Arrows indicate PI3P-positive dextran-negative profiles tethered to latex bead phagosomes. J-L, macrophages were pulsed for 1 h with Texas Red-dextran and chased for 3 h prior to phagocytosis of latex beads. Arrows indicate PI3P-positive dextran-negative profiles. All images are of 1.8-m thick optical sections.
Because Rab5 has the ability to recruit either p85/p110 or p150/hVPS34 (35,66), the phase I PI3P burst could be due to one of the PI3Ks recruited by Rab5.
What processes contribute to anomalous phase I* on nascent mycobacterial phagosomes? Our finding, that cholesterol depletion, which has been shown to block M. tuberculosis var. bovis BCG entry into the macrophages (52), prevents phase I*, suggests that plasma membrane re-organization, most likely re-quiring cholesterol-rich lipid rafts, is involved in the premature generation of PI3P on the plasma membrane during the entry of mycobacteria into the macrophage. This too involves a wortmannin-sensitive PI3K, because the phase I* formation, just like the normal phase I and II, was inhibited by 100 nM wortmannin (data not shown).
Mycobacteria interfere with both phases of PI3P production on phagosomes by: (i) shifting phase I to an unscheduled early time point and (ii) silencing phase II. The physiological significance of CaM-and CaMKII-dependent processes (phase II) as a key target for mycobacterial interference have been unequivocally established (33,57). The mechanism underlying the lack of phase II on mycobacterial phagosomes may be related to changes in Ca 2ϩ flux regulation of PI3P (33) or recruitment of host PI3P-modifying enzymes such as PI3P phosphatases or kinases. Alternatively, phase II PI3P absence on mycobacterial phagosomes may be due to a mycobacterial product responsible for removal of PI3P. The shift of phase I to a much earlier time point (phase I*) by mycobacteria is most likely important, because it is intimately associated with the cholesterol-dependent process of mycobacterial entry (52). The mechanism for the premature generation of PI3P may involve an early activation of Rab5 and its downstream effectors or may represent a completely new process that engages cholesterol-dependent lipid domains (rafts). In terms of providing an advantage to the pathogen, the premature generation of PI3P renders the M. tuberculosis phagosome distinct from the very beginning of phagosomal formation but may play additional functions that remain to be defined.