Anaplasma phagocytophilum Hijacks Flotillin and NPC1 Complex To Acquire Intracellular Cholesterol for Proliferation, Which Can Be Inhibited with Ezetimibe

ABSTRACT The intracellular cholesterol transport protein Niemann-Pick type C1 (NPC1) and lipid-raft protein flotillin (FLOT) are required for cholesterol uptake by the obligatory intracellular bacterium Anaplasma phagocytophilum and for infection, and each protein localizes to membrane-bound inclusions containing replicating bacteria. Here, we found striking localization of FLOT2 in NPC1-lined vesicles and a physical interaction between FLOT2 and NPC1. This interaction was cholesterol dependent, as a CRAC (cholesterol recognition/interaction amino acid cholesterol-binding) domain mutant of FLOT2 did not interact with NPC1, and the cholesterol-sequestering agent methyl-β-cyclodextrin reduced the interaction. The stomatin-prohibitin-flotillin-HflC/K domain of FLOT2, FLOT21–183, was sufficient for the unique FLOT2 localization and interaction with NPC1. NPC1, FLOT2, and FLOT21–183 trafficked to the lumen of Anaplasma inclusions. A loss-of-function mutant, NPC1P691S (mutation in the sterol-sensing domain), did not colocalize or interact with FLOT2 or with Anaplasma inclusions and inhibited infection. Ezetimibe is a drug that blocks cholesterol absorption in the small intestine by inhibiting plasma membrane Niemann-Pick C1-like 1 interaction with FLOTs. Ezetimibe blocked the interaction between NPC1 and FLOT2 and inhibited Anaplasma infection. Ezetimibe did not directly inhibit Anaplasma proliferation but inhibited host membrane lipid and cholesterol traffic to the bacteria in the inclusion. These data suggest that Anaplasma hijacks NPC1 vesicles containing cholesterol bound to FLOT2 to deliver cholesterol into Anaplasma inclusions to assimilate cholesterol for its proliferation. These results provide insights into mechanisms of intracellular cholesterol transport and a potential approach to inhibit Anaplasma infection by blocking cholesterol delivery into the lumen of bacterial inclusions.

T he Gram-negative obligatory intracellular bacterium Anaplasma phagocytophilum primarily infects granulocytes and causes the emerging tick-borne zoonosis called human granulocytic anaplasmosis (HGA). The HGA cases reported to the CDC have increased greater than 10-fold during the past 10 years, reaching nearly 6,000 in 2017 (1). Early clinical signs of HGA are mild to moderate, including fever, chills, severe headache, muscle aches, nausea, vomiting, diarrhea, and loss of appetite, which are readily resolved in most cases with appropriate treatment. However, if treatment is delayed or if there are other medical conditions present, HGA can cause severe illness requiring hospitalization in 36% of cases, and life-threatening disease occurs in 3% with the case fatality rate at 0.6% (2). The only effective treatment is the broad-spectrum antibiotic doxycycline, and there is no vaccine.
Unlike most Gram-negative bacteria, A. phagocytophilum lacks lipopolysaccharide and peptidoglycan in its membrane, yet it contains a significant amount of membrane cholesterol (free cholesterol, not cholesterol esters or lipid droplets) to support its membrane structure and functions (3). Cholesterol is essential for this bacterium, and mice with high blood cholesterol develop more severe clinical signs with a 10-fold higher bacterial load in the blood than mice with a normal cholesterol level (4). A. phagocytophilum cannot synthesize or modify cholesterol; thus, it must acquire cholesterol from its host cell (3,5). Indeed, unlike most bacteria, A. phagocytophilum can readily take up exogenous cholesterol (3).
Mammalian cells acquire cholesterol from two sources: serum lipoproteins and via biosynthesis at the endoplasmic reticulum. A. phagocytophilum captures host cholesterol derived exclusively from low-density lipoprotein (LDL) by upregulating the cellular level of LDL receptor and subverting the Niemann-Pick type C1 (NPC1) pathway of cholesterol transport to A. phagocytophilum-containing inclusions (5,6). In acidic endosomes, cholesterol esters in LDL are hydrolyzed by acid lipase to liberate free cholesterol, which enters intracellular vesicles containing the cholesterol-binding transmembrane protein, NPC1; NPC1-containing vesicles then transport cholesterol to the trans-Golgi network before cholesterol is distributed to various cellular destinations (7,8). Certain mutations in the NPC1 gene causes Niemann-Pick Type C disease, owing to a defect in the trafficking of endocytosed cholesterol with sequestration of free cholesterol in lysosomes and late endosomes (9). Garver et al. (10) pointed out that NPC1 is found in two morphologically distinct membrane compartments, namely, large vesicles (diameter, ;0.4 mm) that contain extensive internal membranes and caveolin-1 but lack lysosomal-associated membrane protein 1 (LAMP1), and a smaller diffusely distributed LAMP1-positive compartment. Only large NPC1-bearing vesicles devoid of lysosomal markers were found to be increased in the human promyelocytic leukemia cell line HL-60 infected with A. phagocytophilum, and this subset trafficked to the bacterial inclusions (6). This localization was abolished by the LDL-derived cholesterol-trafficking inhibitor U18666A, which, when administered to cells, mimics the molecular aspects of Niemann-Pick Type C disease (6). Studies using an NPC1-specific short interfering RNA (siRNA) and a cell line with dysfunctional NPC1 demonstrated that NPC1 function is required for cholesterol acquisition by A. phagocytophilum and infection (6).
Flotillin 1 (FLOT1) and FLOT2 are cholesterol-associated lipid-raft proteins that form a heterodimer and/or oligomer complex, and they are found in the plasma membrane, intracellular vesicles devoid of LAMP1, and exosomes (11 to 15). FLOTs are crucial for A. phagocytophilum replication in host cells, as siRNA-mediated knockdown of either FLOT1 or FLOT2 reduced A. phagocytophilum infection (16). FLOT-containing vesicles are enriched with free cholesterol and colocalize with endocytosed LDL and with acid lipase (16), and traffic to A. phagocytophilum inclusions (16). However, the relationship between NPC1 and FLOTs in uninfected or A. phagocytophilum-infected cells remains unknown, except for one study showing that the FLOT2-dependent release of cholesterol from exosomes ameliorates cellular cholesterol accumulation in Niemann-Pick type C disease (17). We investigated the interaction between NPC1 and FLOTs in the context of cholesterol transport within uninfected and A. phagocytophilum-infected host cells and explored the possibility of inhibiting this interaction to block A. phagocytophilum infection.

RESULTS
Lipid raft protein FLOT2 localizes to the lumen of NPC1-containing vesicles and interacts physically with NPC1. We first examined the topographical relationship between endogenous FLOT2 and endogenous NPC1-containing vesicles (here termed NPC1 vesicles) in thinly spread monkey endothelial RF/6A cells compatible for unambiguous localization analysis. By double immunofluorescence labeling, endogenous FLOT2 was distinctly localized in the lumen of endogenous NPC1-containing vesicles in RF/6A cells (Fig. 1A). Indeed, the fluorescence intensity profile analysis of red (FLOT2) and green (NPC1) signals along the length of the line revealed that the peak red signals were surrounded by peak green signals (Fig. 1D). Most of these NPC1 vesicles were found to be .0.4 mm in diameter. The majority of FLOT2 colocalized with NPC1 (Pearson's correlation coefficient [R ]= 0.78) (Fig. 1G). To avoid possible artifacts of membrane permeabilization used for immunofluorescence labeling or immune crossreactivity, we cotransfected RF/6A cells with FLOT2-mCherry and NPC1-GFP, which showed the same colocalization pattern as endogenous proteins (R = 0.83) (Fig. 1B, E, and H). This is not specific to RF/6A cells, as the colocalization pattern was observed with human kidney epithelial HEK293T cells (R = 0.77) (Fig. 1C, F, and I). Further, anti-FLOT2 IgG, but not negative-control mouse IgG, coimmunoprecipitated endogenous NPC1 in HEK293T cells (Fig. 1J). Immunoprecipitation with anti-GFP affinity gel resulted in pulldown of endogenous FLOT2 from the lysate of HEK293T cells transfected with NPC1-GFP but not cells transfected with the green fluorescent protein (GFP) negative control (Fig. 1K). These results demonstrate intraluminal localization of FLOT2 in NPC1 vesicles and physical interaction between FLOT2 and NPC1.
FLOT2 colocalization with NPC1 requires the sterol-sensing domain of NPC1. A single-amino-acid mutation of the sterol-sensing domain of NPC1, namely, NPC1 P692S , results in decreased LDL-derived cholesterol delivery to endoplasmic reticulum and the plasma membrane, similar to what is observed in Niemann-Pick type C disease (20,21). Given the requirement of the FLOT2 CRAC motifs for FLOT2 colocalization with NPC1 ( Fig. 3A and B), we examined whether the NPC1 sterol-sensing domain is required for this colocalization. FLOT2-mCherry localized mainly to the plasma membrane in NPC1 P692S -GFP-cotransfected cells (Fig. 4A); consequently, colocalization of FLOT2-mCherry with NPC1 P692S -GFP vesicles was significantly reduced compared with the NPC1-GFP control ( Fig. 4A and B). Indeed, the amount of endogenous FLOT2 that was pulled down with anti-GFP from the lysate of HEK293T cells transfected with NPC1 P692S -GFP was significantly reduced compared with cells transfected with NPC1-GFP ( Fig. 4C and D). Thus, the sterol-sensing domain of NPC1, which is critical for normal intracellular cholesterol distribution, is required for FLOT2-NPC1 colocalization and interaction.
Cholesterol dependence of the colocalization/interaction of FLOT2 with NPC1. FLOTs have two major subcellular localization pools, namely, the plasma membrane and intracellular vesicles, and this differential localization is regulated by free cholesterol (15,17). An increase in cellular cholesterol above a certain threshold leads to transfer of FLOTs from the plasma membrane to intracellular vesicles, and, reciprocally, cholesterol depletion drives vesicular FLOT2 to the plasma membrane (17). Given that the CRAC domain of FLOT2 is required for FLOT2 colocalization with NPC1 ( Fig. 3A and B) and the sterol-sensing domain of NPC1 is critical for FLOT2 and NPC1 colocalization and interaction (Fig. 4), we examined the requirement of cellular cholesterol for the FLOT2-NPC1 interaction using methyl-b-cyclodextrin (MbCD), which, at 10 mM concentration, reduces membrane cholesterol abundance by inducing cellular cholesterol efflux (22). As 10 mM MbCD caused RF/6A cells to retract and partially detach from the substratum, we treated RF/6A cells with 2.5 mM MbCD for 40 min at 2 days posttransfection (dpt). This treatment greatly reduced FLOT2 localization to intracellular vesicles; consequently, the colocalization between NPC1-GFP and FLOT2-mCherry was significantly decreased compared with the control (Fig. 5A and B). After cholesterol replenishment in MbCD-treated cells for Ezetimibe blocks the colocalization/interaction of FLOT2 with NPC1. Ezetimibe, which blocks cholesterol absorption at the brush border of the small intestine, is an FDA-approved LDL cholesterol-lowering drug for the treatment of hypercholesterolemia (23,24). Altmann et al. (25) reported the discovery of the Niemann-Pick C1-like 1 protein (NPC1L1) as the human sterol transport protein that was expressed at the enterocyte lumenal (apical) surface as well as the hepatobiliary (canalicular) interface. NPC1L1 was then identified as the molecular target of ezetimibe, and the ezetimibebinding site of NPC1L1 was determined (26). FLOTs play a critical role in this NPC1L1mediated cholesterol uptake via formation of cholesterol-enriched membrane microdomains, which function as carriers for the bulk of cholesterol (27). Ezetimibe binding to NPC1L1 disrupts the formation of the NPC1L1-FLOT1/2 complex, resulting in reduced cholesterol absorption via NPC1L1-mediated endocytosis (27).
Based on NCBI conserved domain structure analysis, human NPC1 and NPC1L1 proteins have similar Niemann-Pick C type protein family domains, which include an NPC1 N-terminus domain (pfam16414) with a cholesterol-binding pocket and a sterol- sensing domain (pfam12349) at the central region (see Fig. S1A in the supplemental material). In addition, alignment showed that these two proteins share 40% amino acid identity and 57% amino acid similarity ( Fig. S1B and C). Thus, we tested the effects of ezetimibe on the FLOT2-NPC1 interaction. When RF/6A cells were cotransfected with FLOT2-mCherry and NPC1-GFP and treated with 40 mM ezetimibe for 20 h at 2 dpt, FLOT2-mCherry localized mainly to the plasma membrane; consequently, colocalization between FLOT2-mCherry and NPC1-GFP was significantly reduced ( Fig. 6A and B). Treatment with ezetimibe significantly reduced the amount of endogenous FLOT2 that was pulled down with anti-GFP from the lysate of HEK293T cells transfected with NPC1-GFP ( Fig. 6C and D).
Endogenous FLOT2, endogenous NPC1, and HA-FLOT2 1-183 traffic to the lumen of A. phagocytophilum inclusions and encase individual bacteria. Given that each of FLOT2 and NPC1 localizes to A. phagocytophilum inclusions (6,16,28) and the interaction of FLOT2 and NPC1 in uninfected cells ( Fig. 1 to 6), we examined whether they colocalize on A. phagocytophilum inclusions. Immunofluorescence labeling showed FIG 4 NPC1 P692S interacts with FLOT2 at reduced levels. (A) RF/6A cells were cotransfected with C-terminal GFPtagged NPC1 wild type (NPC1 WT ) or P692S mutant (NPC1 P692S ) and FLOT2-mCherry. At 2 dpt, cells were fixed and stained with DAPI. The boxed area is enlarged 4Â on the right. Bar, 10 mm. (B) Colocalization of NPC1 P692S -GFP or NPC1-GFP with FLOT2-mCherry was quantified by counting vesicles in 30 cells per group from three independent experiments. *, P , 0.05, two-tailed t test. (C) HEK293T cells transfected with NPC1-GFP or NPC1 P692S -GFP at 3 dpt were lysed and immunoprecipitated (IP) with anti-GFP affinity gel. Whole-cell lysates and immunoprecipitates were analyzed by Western blotting (WB) with antibodies against GFP, FLOT2, and tubulin. (D) Relative ratio of WB band intensities, with the ratio of FLOT2/NPC1-GFP set as 100%. Results are presented as the mean 6 standard deviation from three independent experiments. *, P , 0.05, two-tailed t test.
We previously showed that the lumen of A. phagocytophilum inclusions is enriched with cholesterol based on staining with filipin, a free cholesterol-binding polyene antibiotic (5). Dipyrromethene difluoride-cholesterol (BODIPY-or TopFluor-cholesterol [TFcholesterol]) is a widely used cholesterol analog because it has greater intrinsic fluorescence (bright and photostable) than filipin and partitions in membranes similarly to natural cholesterol (29,30). When dissolved in solvent and applied to cells in growth medium containing lipoprotein-deficient serum, TF-cholesterol diffuses into eukaryotic cells and equilibrates slowly with intracellular membranes (30). We previously used this approach to more clearly visualize the distribution of host membrane cholesterol in inclusions of another cholesterol-dependent bacterium, Ehrlichia chaffeensis (31). In agreement with our previous study with filipin (5) and others (28), TF-cholesterol was highly enriched in A. phagocytophilum inclusion and colocalized with HA-FLOT2 1-183 within the inclusions (Fig. 7C).
NPC1 P692S cannot localize to A. phagocytophilum inclusions, and its overexpression reduces A. phagocytophilum infection. As NPC1 P692S had reduced colocalization or interaction with FLOT2 (Fig. 4A to D), we examined whether NPC1 P692S could localize to A. phagocytophilum inclusions. Indeed, compared to NPC-1-GFP, NPC1 P692S -GFP localization to inclusions (Fig. 8A and B) and the number of infected cells was significantly reduced (Fig. 8C), indicating that the sterol-sensing domain of NPC1 is required for both localization of NPC1 to inclusions and delivery of cholesterol to A. phagocytophilum to support its proliferation.
Ezetimibe blocks A. phagocytophilum infection in HL-60 and RF/6A cells. Because NPC1 and FLOT2 are required for A. phagocytophilum infection (6, 16) and we found that ezetimibe blocks NPC1-FLOT2 colocalization and interaction (Fig. 6), we examined whether ezetimibe could block infection of cells with A. phagocytophilum. First, we demonstrated that ezetimibe at 2 to 40 mM was not toxic to HL-60 cells (Fig. S2). Ezetimibe then was added to HL-60 cells in culture at 2 h postinfection (hpi; just after internalization and when bacterial inclusions were not discernible in the cells under a light microscope). Cells were then harvested at 2 days postinfection (dpi). Ezetimibe indeed blocked A. phagocytophilum proliferation in a dose-dependent manner (1 to 20 mM; Fig. 9A to C). In addition, the ezetimibe-mediated inhibition of proliferation could be achieved even when ezetimibe was added to the cell culture at 2 dpi with subsequent incubation for only 20 h (Fig. 9D). Strikingly, with ezetimibe treatment, many vacuoles containing lightly stained materials and a few bacteria, were seen in the cytoplasm of the infected HL-60 cells, in contrast to the presence of numerous A. phagocytophilum inclusions tightly packed with numerous bacteria in nontreated control cells (Fig. 9D). Notably, ezetimibe did not directly inhibit A. phagocytophilum, as demonstrated by the fact that infection was not blocked when host cell-free A. phagocytophilum was treated with ezetimibe (10 mM) for 30 min and then added to HL-60 cells after removal of ezetimibe (Fig. S3). The inhibition of A. phagocytophilum infection and vacuolation caused by ezetimibe were not specific to the host cell type, as similar results were obtained with RF/6A cells (Fig. S4). The vacuolation was observed specifically when A. phagocytophilum-infected cells were treated 1 to 2 dpi with ezetimibe, as 0% of cells developed vacuolation when uninfected cells were treated (Fig. S4F), or infected cells were treated starting 2 hpi with ezetimibe ( Fig. 9). Ezetimibe-induced inhibition of A. phagocytophilum proliferation and vacuolation were reversible, as demonstrated by the fact that when ezetimibe was added at 24 hpi and removed at 36 hpi, regrowth and reduction of vacuoles were observed at 72 hpi (Fig. S5). Vacuolization was confirmed as being a consequence of ezetimibe-mediated modifications of A. phagocytophilum inclusions, as intravacuolar A. phagocytophilum could be detected by immunofluorescence staining with monoclonal antibody 5C11, which is specific for the A. phagocytophilum outer membrane protein P44 (32) (Fig. 10A). Lastly, we used filipin staining to examine whether free cholesterol levels associated with A. phagocytophilum inclusions were altered with/without ezetimibe treatment. Indeed, the level of inclusion-associated free cholesterol was significantly lower in ezetimibe-treated cells (Fig. 10B). Taken together, these data indicated that ezetimibe could block A. phagocytophilum infection in host cells by preventing (or even reversing) cholesterol trafficking to A. phagocytophilum inclusions.
Ezetimibe blocks trafficking of host cell membrane lipids to A. phagocytophilum inclusions. Using DiI [3,DiIC 18 (3)] to label host cell membranes (31,33,34), and using TF-cholesterol to visualize the distribution of membrane cholesterol (31), we examined whether ezetimibe affects the trafficking of hostderived membrane lipids to bacterium-replicating inclusions and A. phagocytophilum.
To prevent diffusion of DiI between membranes, cells were fixed with paraformaldehyde (PFA) without permeabilization, and a coverslip sealant containing no organic solvents was used. Fluorescence microscopy and line profile analysis showed that DiI strongly labeled membranes of all A. phagocytophilum inclusions ( Fig. 11A and C, open arrows) and most individual bacterium ( Fig. 11A and C, solid arrows) in infected RF/6A cells, suggesting that host cell membranes were trafficked and incorporated into membranes of A. phagocytophilum inclusions and individual bacteria. Similarly, membranes Inhibition of Cholesterol Traffic to Anaplasma of A. phagocytophilum inclusions and individual bacteria were extensively labeled by TF-cholesterol ( Fig. 11B and C, open and solid arrows, respectively).
When A. phagocytophilum-infected RF/6A cells at 1 dpi were treated with ezetimibe for 20 h, bacterial growth and inclusion numbers were severely reduced, similar to results shown in Fig. 9 and Fig. S4, and trafficking of host membranes labeled by DiIor TF-cholesterol to A. phagocytophilum inclusions were significantly reduced more than 2-fold ( Fig. 11A and B). Interestingly, when A. phagocytophilum-infected cells were treated with ezetimibe for 14 h and then labeled with DiI for 6 h in the presence of ezetimibe, DiI labeling of A. phagocytophilum bacterial membrane within the inclusions was completely prevented (Fig. 11A and C). On the other hand, when TF-cholesterol was incubated with infected cells at 20 hpi for 8 h and then treated with ezetimibe for 20 h, a small number of surviving bacteria retained TF-cholesterol labeling ( Fig. 11B and C). Taken together, these data suggest that A. phagocytophilum in the inclusions incorporates host membrane lipids and cholesterol from the host membrane vesicles, which can be blocked by ezetimibe.

DISCUSSION
A. phagocytophilum captures cholesterol exclusively from LDL (5), and NPC1 and FLOT2 both target A. phagocytophilum inclusions and are required for infection (6,16). However, the relationship between NPC1 and FLOTs was unknown in uninfected cells, not to mention infected cells. One of the important findings of the present study is the striking presence of FLOT2 in the NPC1-containing vesicles and their physical interaction. We previously showed that FLOT2 localizes on LDL-and acid lipase-containing vesicles (16). Collectively, our results revealed the critical role of FLOT in the normal physiological process of intracellular membrane cholesterol trafficking, including LDLderived cholesterol from acidic endosomes to the NPC1 compartment, which is the main sorting compartment for cellular cholesterol. FLOT2-GFP bearing two mutations in the CRAC motif cannot target to A. phagocytophilum inclusions (16). The cholesterol-sequestering agent MbCD abrogated FLOT2 localization to A. phagocytophilum inclusions and cleared infection (16). The present study revealed the FLOT2 CRAC mutant could not interact with NPC1 vesicles, and MbCD reduced FLOT2 interaction with NPC1. The NPC1 sterol-sensing domain mutant NPC1 P692S sequesters cholesterol within phagolysosomes and lysosomes rather than redistributing it to the plasma membrane and multivesicular bodies (20,21); consequently, FLOT2 did not interact with NPC1 P692S , and NPC1 P692S could not target A. phagocytophilum inclusions. Similar to the effects of U18666A treatment, which sequesters cholesterol within phagolysosomes (6), overexpression of NPC1 P692S also reduced A. phagocytophilum infection. Taken together, the present study illuminated that the cholesterol-dependent NPC1 and FLOT2 interaction is critical for A. phagocytophilum to hijack the intracellular LDL-cholesterol transport process to capture cholesterol for proliferation.
The presence of multitransmembrane cholesterol-binding protein NPC1, lipid raftbinding protein FLOT2, or lipid raft-binding FLOT2 SPFH domain within A. phagocytophilum inclusions, encasing individual bacteria, suggests there are cholesterol-rich intraluminal membranes in addition to membranes of A. phagocytophilum. We recently reported that intrainclusion membranes are abundant in vacuoles containing another cholesterol-dependent bacterium, Ehrlichia chaffeensis (31). Labeling of host cell membrane lipids, glycerophospholipids, and cholesterol, revealed that host cell membrane lipids are delivered into the lumen of E. chaffeensis inclusions, and this process is driven by bacterial factors (31). The present study revealed host membrane lipids and cholesterol are similarly transported into A. phagocytophilum inclusions, encasing individual bacteria. As NPC1 trafficking to A. phagocytophilum inclusions requires bacterial protein synthesis (6), A. phagocytophilum actively utilizes the NPC1-FLOT2-coordinated intracellular cholesterol transport mechanism to divert membrane cholesterol as well as membrane lipids into inclusions for bacterial proliferation.
Recently, NPC1 was shown to consist of two forms that localize to different organelles: the mannose-rich form (NPC1h), which is sensitive to endo H treatment (and, thus, is cleaved to a smaller polypeptide of ;130 kDa), is located in the endoplasmic reticulum, and another form is the endo H-resistant complex glycosylated form (NPC1c) of ;190 kDa, which is processed from NPC1h in the Golgi and ultimately traffics to lysosomes (35). The investigators showed that NPC1c is the exclusive form of NPC1 that fractionates in lipid rafts containing FLOT2. Although that study did not address the potential colocalization or physical interaction of NPC1 and FLOT2, their results suggest that the NPC1 physically interacting with FLOT2-containing vesicles in our study is the Golgi-processed NPC1c. The relative abundance of highly glycosylated NPC1 observed in A. phagocytophilum-infected cells (6) corroborates that the NPC1c-and FLOT2-containing lipid raft membrane is trafficked to A. phagocytophilum inclusions.
Another important finding of the present study is that the topological and physical association of NPC1 with FLOT2, and A. phagocytophilum proliferation can be blocked by ezetimibe. Unlike uninfected cells, treatment of infected cells with ezetimibe resulted in multiple intracellular vacuoles that were derived from former A. phagocytophilum inclusions. Similar phenomena were observed when A. phagocytophiluminfected cells were treated with oxytetracycline (6), which inhibits bacterial translation, or when E. chaffeensis-infected cells were treated with 3-methyl adenine, an autophagy inhibitor that prevents bacterial proliferation (36). In all these experimental scenarios, lysosomal fusion with bacterium-containing vacuoles was not observed, suggesting the bacteria could not maintain the inclusion integrity and starved to death owing to the inability to take up sufficient amounts of critical host cell factors, e.g., cholesterol, membrane lipids, and/or amino acids (6,36). Ezetimibe inhibition of A. phagocytophilum proliferation and induction of vacuolation in infected cells were reversible. It seems once these cholesterol-dependent bacteria stop replicating, the intraluminal membranes, including the bacteria membrane, are integrated back to inclusions, which creates an enlarged vacuole.
Hydrophobic amines such as U18886A and imipramine are known to accumulate in acidic cellular compartments, particularly lysosomes, and block the postlysosomal transport of cholesterol in the LDL uptake pathway. We previously showed U18886A and imipramine significantly inhibit A. phagocytophilum infection and replication in HL-60 cells in a dose-dependent manner (5). Recently, it was reported that desipramine, a metabolite of imipramine and an acid sphingomyelinase inhibitor, blocks LDL-derived cholesterol efflux, similar to U18886A or imipramine, and significantly inhibits A. phagocytophilum infection in cell culture and in mice (28,37). Taken together, these studies extend previous findings on the critical role of LDL-derived cholesterol for A. phagocytophilum proliferation (5,6,16); thus, LDL-derived cholesterol traffic can be a potential target of host-directed anti-A. phagocytophilum chemotherapy.

MATERIALS AND METHODS
Antibodies and plasmids. The following antibodies were used: mouse monoclonal anti-FLOT2 (BD Pharmingen, San Jose, CA), rabbit anti-NPC1 (Novus Biologicals, Centennial, CO), mouse monoclonal Inhibition of Cholesterol Traffic to Anaplasma ® anti-GFP (Santa Cruz Biotechnology, Santa Cruz, CA), mouse monoclonal anti-a-tubulin (Cell Signaling, Danvers, MA), and mouse monoclonal anti-hemagglutinin (HA) (Santa Cruz Biotechnology). Normal mouse IgG was purchased from Santa Cruz Biotechnology. Secondary antibodies conjugated to fluorescent probes (Alexa Fluor 488-conjugated goat anti-mouse IgG and Alexa Fluor 555-conjugated goat anti-mouse IgG) were obtained from Life Technologies (Eugene, OR), and peroxidase-conjugated secondary antibodies were obtained from KPL (Gaithersburg, MD).
A. phagocytophilum and cell culture. Cultivation of A. phagocytophilum HZ strain in HL-60 cells (ATCC, Manassas, VA), preparation of host cell-free A. phagocytophilum, and infection were performed as described previously (6). The degree of bacterial infection in host cells was assessed by Diff-Quik staining (Baxter Scientific Products, Obetz, OH), and the number of A. phagocytophilum cells was scored in 200 host cells in triplicate culture wells as described previously (40) or by quantitative PCR to detect the A. phagocytophilum 16S rRNA gene normalized by the human GAPDH gene (6). RF/6A monkey endothelial cells (ATCC) were used in immunofluorescence labeling for unambiguous localization analysis owing to their tight adherence and thinly spread morphology (5). Human embryonic kidney HEK293T cells (ATCC) were used for transfection and pulldown assays, as they can be transfected with high efficiency (41).
For endogenous FLOT2 and NPC1 labeling, uninfected cells or A. phagocytophilum-infected cells treated with DFP for 2 h prior to harvesting were fixed in methanol-acetone (80:20) at -20°C for 10 min and incubated with mouse anti-FLOT2 and rabbit anti-NPC1 in PBS containing 0.1% gelatin for 1 h at 37°C, followed by incubation with Alexa Fluor 488-conjugated goat anti-rabbit IgG and Alexa Fluor 555conjugated goat anti-mouse IgG for 30 min at room temperature.
TF-cholesterol, DiI, and filipin labeling. RF/6A cells were transfected with HA-FLOT2 1-183 , seeded onto glass coverslips in a 6-well plate, and cultured in AMEM supplemented with 5% fetal bovine serum and 2 mM L-glutamine at 37°C for 4 h prior to addition of host cell-free A. phagocytophilum. For TF-cholesterol (Avanti Polar Lipids, Alabaster, AL) labeling, infected RF/6A cells at 1 dpi were washed three times with serum-free AMEM, and the medium was replaced with AMEM supplemented with 5% lipoprotein-depleted serum (LPDS; Kalen Biomedical, Germantown, MD). After culturing for 8 h, TF-cholesterol (1 mM final concentration) was added to cells with subsequent incubation for 1 day (31). For cholesterol labeling with filipin (Sigma), RF/6A cells were fixed with 4% PFA for 15 min and stained with 50 mg/ ml filipin in PBS at room temperature for 1 h and observed under a DeltaVision microscope.
For DiI and TF-cholesterol labeling with ezetimibe treatment, RF/6A cells were seeded onto a coverglass for 3 h and infected with A. phagocytophilum. At 1 dpi, cells were treated with dimethyl sulfoxide (DMSO) control or 40 mM ezetimibe for 14 h and then incubated with 5 mM Vybrant DiI cell-labeling solution (Thermo Fisher Scientific, Waltham, MA) for 6 h in the presence of ezetimibe. Alternatively, at 20 hpi, cells were incubated with 2 mM TF-cholesterol in AMEM containing 5% LPDS for 8 h and then treated with DMSO control or 40 mM ezetimibe for 20 h in the presence of TF-cholesterol. Cells were washed 3 times with PBS, fixed in 4% PFA for 15 min, and incubated with a cell-permeable DNA dye (1 mg/ml Hoechst 33342; Invitrogen) for 15 min to stain A. phagocytophilum and host DNA. The coverslip was mounted onto a slide using SlowFade Diamond antifade mountant (Invitrogen) and sealed with a coverslip sealant containing no organic solvents (Biotum, Fremont, CA).
DeltaVision microscopy and image analysis. Fluorescence and differential interference contrast images were captured with a DeltaVision personal DV deconvolution microscope system (GE Healthcare, Marlborough, MA). Data were processed using softWoRx (GE Healthcare) and Adobe Photoshop (Adobe Systems, Mountain View, CA). Colocalization was analyzed with softWoRx for the calculation of Pearson's correlation coefficient. Colocalization of NPC1-GFP and FLOT2-mCherry in MbCD-treated or ezetimibetreated cells was analyzed in 30 cells per group in each of three independent experiments. Colocalization of NPC1 P692S -GFP and FLOT2-mCherry was analyzed in 30 cells per group in each of three independent experiments. Line-profile analyses were performed on a single z-section using ImageJ (NIH,