Elevated Cholesterol in the Coxiella burnetii Intracellular Niche Is Bacteriolytic

ABSTRACT Coxiella burnetii is an intracellular bacterial pathogen and a significant cause of culture-negative endocarditis in the United States. Upon infection, the nascent Coxiella phagosome fuses with the host endocytic pathway to form a large lysosome-like vacuole called the parasitophorous vacuole (PV). The PV membrane is rich in sterols, and drugs perturbing host cell cholesterol homeostasis inhibit PV formation and bacterial growth. Using cholesterol supplementation of a cholesterol-free cell model system, we found smaller PVs and reduced Coxiella growth as cellular cholesterol concentration increased. Further, we observed in cells with cholesterol a significant number of nonfusogenic PVs that contained degraded bacteria, a phenotype not observed in cholesterol-free cells. Cholesterol had no effect on axenic Coxiella cultures, indicating that only intracellular bacteria are sensitive to cholesterol. Live-cell microscopy revealed that both plasma membrane-derived cholesterol and the exogenous cholesterol carrier protein low-density lipoprotein (LDL) traffic to the PV. To test the possibility that increasing PV cholesterol levels affects bacterial survival, infected cells were treated with U18666A, a drug that traps cholesterol in lysosomes and PVs. U18666A treatment led to PVs containing degraded bacteria and a significant loss in bacterial viability. The PV pH was significantly more acidic in cells with cholesterol or cells treated with U18666A, and the vacuolar ATPase inhibitor bafilomycin blocked cholesterol-induced PV acidification and bacterial death. Additionally, treatment of infected HeLa cells with several FDA-approved cholesterol-altering drugs led to a loss of bacterial viability, a phenotype also rescued by bafilomycin. Collectively, these data suggest that increasing PV cholesterol further acidifies the PV, leading to Coxiella death.

C holesterol is known to play key roles in cardiovascular disorders, obesity, diabetes, and infectious diseases caused by numerous bacterial, viral, and protozoal pathogens. Intracellular pathogens in particular target cholesterol at various stages of infection. For example, Mycobacterium bovis and Helicobacter pylori directly target cholesterol as a "docking site" to stabilize interactions with the host cell membrane and initiate internalization (1)(2)(3). Mycobacterium spp., Brucella suis, Listeria monocytogenes, Leishmania donovani, and Plasmodium falciparum appear to target cholesterol-rich lipid rafts during entry into both phagocytic and nonphagocytic cells (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). Once inside the cell, cholesterol is often targeted during establishment of the intracellular niche and bacterial growth. For example, Mycobacterium tuberculosis and Mycobacterium leprae accumulate cholesterol in the early phagosome as a mechanism to inhibit phagosomelysosome fusion and promote pathogen survival (14)(15)(16). M. tuberculosis also utilizes a cholesterol import system to hijack host cell cholesterol as a carbon and energy source (17). Chlamydia trachomatis intercepts cholesterol trafficking from the Golgi apparatus and incorporates cholesterol into the Chlamydia-containing inclusion body as well as the bacterial cell wall (18). Thus, modulation of cellular cholesterol by diverse microbial pathogens appears to play an important role in promoting pathogen entry, survival, and subsequent disease.
Recent reports have implicated cholesterol as an important factor during infection by the intracellular bacterial pathogen Coxiella burnetii, a significant cause of culturenegative endocarditis in the United States (19)(20)(21). An obligate intracellular pathogen during natural infection, C. burnetii forms a unique niche in a modified acidic phagolysosome known as the parasitophorous vacuole (PV). After uptake by the host cell via phagocytosis, the bacterium resides in a tight-fitting nascent phagosome that matures through the default endocytic pathway (22,23). Approximately 24 to 48 h postinfection, the C. burnetii PV expands through fusion with early and late endosomes, lysosomes, and autophagosomes (24). As a result, the mature PV membrane is a hybrid of host vesicular membranes, and the vacuole displays various characteristics of a phagolysosome, including lysosomal hydrolases (acid phosphatase, cathepsin D, and 5=-nucleotidase) and an acidic pH of~4.5 to 5 (24). Establishment of a replicationcompetent PV requires the C. burnetii Dot/Icm type 4B secretion system (T4BSS), which manipulates host cell trafficking and signaling pathways via the activity of effector proteins secreted into the host cytoplasm (25).
Formation of the C. burnetii PV is a highly dynamic process involving vesicular trafficking and fusion events, with the PV membrane playing a central role. A distinguishing feature of the C. burnetii PV membrane, based on staining with the fluorescent sterol-binding compound filipin, is that it is rich in sterols (21). A role for cholesterol during C. burnetii infection was suggested by gene expression analysis of infected host cells, which found that genes involved in cholesterol efflux and storage are upregulated during C. burnetii infection (26,27). Further, a recent screen of a FDA-approved drug library identified 57 drugs that perturb host cell cholesterol homeostasis also block C. burnetii growth in THP-1 human macrophage-like cells (19). Intriguingly, these drugs had a more pronounced effect on C. burnetii than on Legionella pneumophila, Rickettsia conorii, or Brucella abortus, suggesting that C. burnetii may be uniquely sensitive to altered host cell cholesterol homeostasis. Additionally, when cholesterol transport from endosomes and presumably the C. burnetii PV was blocked through knockdown of the cholesterol transporter NPC-1, C. burnetii growth was significantly attenuated (19). Together, these studies suggest that cholesterol is an important player affecting the C. burnetii-host cell interaction.
In order to further understand the role of cholesterol during C. burnetii infection, we developed a novel cholesterol-free host cell tissue culture system using cells lacking DHCR24, the final enzyme in cholesterol biosynthesis (20). When adapted to serum-free media, DHCR24 Ϫ/Ϫ mouse embryonic fibroblasts lack both endogenous and exogenous cholesterol sources, and instead, they accumulate desmosterol in cellular membranes. Cholesterol-free cells are an attractive model for deciphering the role of cholesterol in cellular processes, enabling cholesterol manipulation by the addition of exogenous cholesterol to the media. Our prior studies with this model system revealed that C. burnetii uptake into fibroblast cells was dependent on cholesterol-rich lipid rafts and the integrin ␣ v ␤ 3 (20). Strikingly, C. burnetii PV formation and intracellular replication did not require cholesterol. Further, the PV acquired the typical PV markers Rab7, flotillin-2, syntaxin 7, syntaxin 8, and Vamp7 and contained active cathepsin, indicating that the majority of PV maturation events occurred in the absence of cholesterol. However, the lack of the late endosomal marker CD63 in the PV lumen in cholesterolfree cells suggests that cholesterol regulates one or more intracellular trafficking pathways to the PV (20).
While studies thus far indicate that cholesterol plays a key role during C. burnetii infection, how cholesterol affects the formation and maintenance of the PV, as well as C. burnetii growth, is not yet known. Here, we utilized cholesterol-free cells to further decipher the role of cholesterol in C. burnetii-host cell interactions. Our studies surprisingly revealed that increasing cholesterol in the C. burnetii PV inhibits fusion between the PV and endosomes, acidifies the PV, and results in C. burnetii degradation. Our data demonstrating a cholesterol-mediated negative effect on an intracellular bacterial pathogen is novel and may have broader implications in the treatment of C. burnetii infection.

RESULTS
PV size and C. burnetii growth are sensitive to cholesterol. DHCR24 Ϫ/Ϫ mouse embryonic fibroblast cells (MEFs) lack the final enzyme in cholesterol biosynthesis. When adapted to serum-free media, these cells are cholesterol free and accumulate desmosterol in place of cholesterol (20). Cellular cholesterol levels can be controlled by adding cholesterol to the culture medium, which is then preferentially incorporated into cellular membranes over desmosterol (20). Using this model system, we can observe the effect of cellular cholesterol on C. burnetii-host interactions over a range of cholesterol concentrations and longer infection periods than those used in traditional methods of manipulating cellular cholesterol.
We previously found that while C. burnetii entry was reduced, bacterial replication did not appear to be significantly affected in the absence of cholesterol (20). To determine whether host cholesterol levels influence formation of the C. burnetii PV, we infected DHCR24 Ϫ/Ϫ MEFs grown under different cholesterol concentrations and assessed PV size over 6 days of infection using immunofluorescence microscopy. Infected cells were fixed and stained using antibodies against C. burnetii and LAMP-1 (lysosomeassociated membrane glycoprotein 1), a lysosomal protein found on the PV membrane. Surprisingly, at the beginning of PV expansion at 2 days postinfection, the average PV size in cholesterol-free MEFs was at least twice as large as PVs in MEFs with cholesterol ( Fig. 1A). While PVs in cholesterol-free MEFs continued to expand approximately 8-fold over the next 4 days, PVs in MEFs with cholesterol remained significantly smaller regardless of the cholesterol concentration ( Fig. 1B and C). Given the dramatic effect of cholesterol on PV size, we used quantitative PCR to measure C. burnetii growth. The fold change in C. burnetii growth over 6 days decreased in a cholesterol-dependent manner, with little growth seen at the highest cholesterol concentration (Fig. 1D). Together, these data indicate that C. burnetii PV size and bacterial growth are negatively affected by cholesterol.
Addition of cholesterol leads to C. burnetii lysis. Fixed-cell microscopy suggested that poor bacterial growth in the presence of cholesterol was associated with deficient PV formation, similar to the observed phenotype for C. burnetii T4BSS mutants (28,29). To further examine this phenotype, we used live-cell imaging of MEFs with or without cholesterol and infected with mCherry-expressing C. burnetii (mCherry-C. burnetii). Surprisingly, we observed in MEFs with cholesterol a significant number of PVs with free mCherry fluorescence in the PV lumen, a result of lysis or degradation of mCherryexpressing C. burnetii ( Fig. 2A, bottom panel). Because the bacteria have lysed, we have designated these PVs "lytic." By 2 days postinfection, approximately 20% of the PVs were lytic in MEFs with cholesterol, regardless of the cholesterol concentration. Over the next 48 h, the percentage of lytic PVs in MEFs with cholesterol increased in a dose-dependent fashion. Importantly, we never observed lytic PVs in cholesterol-free MEFs (Fig. 2B). To determine whether live-cell microscopy data correlated with bacterial viability, we measured viable bacteria using a fluorescent infectious focus-forming unit (FFU) assay (30). Bacteria were recovered from MEFs under different cholesterol conditions, replated onto a monolayer of Vero cells, and incubated for 5 days. After C. burnetii bacteria were stained, the numbers of fluorescent foci were counted, with one focus unit equivalent to one viable bacteria. Over a period of 6 days, the number of viable C. burnetii in MEFs with cholesterol decreased 80 to 98% compared to the number in cholesterol-free MEFs, depending on the cholesterol concentration (Fig. 2C). These data suggest that, as opposed to a lack of bacterial growth, the addition of cholesterol to cholesterol-free MEFs is bacteriolytic.
Cholesterol traffics to the C. burnetii PV. We next examined cholesterol trafficking in cholesterol-free MEFs to determine how the addition of exogenous cholesterol in our Coxiella burnetii Cholesterol Sensitivity ® model system might lead to C. burnetii death. Exogenous cholesterol could be internalized into the cell through two mechanisms: insertion into the plasma membrane or uptake of cholesterol-bound low-density lipoprotein (LDL). In the first case, cholesterol intercalates into the plasma membrane and is then distributed throughout the cell via vesicular and nonvesicular pathways. This pathway also mimics trafficking of endogenous host cell cholesterol, which is synthesized in the endoplasmic reticulum (ER) and then transported to the plasma membrane for cellular distribution. To examine whether plasma membrane-derived cholesterol travels to the PV, we incubated C. burnetiiinfected DHCR24 Ϫ/Ϫ MEFs with fluorescent BODIPY-cholesterol complexed to methyl-␤-cyclodextrin, which leads to incorporation of the BODIPY-cholesterol into the plasma membrane. After 24 h incubation at 37°C to allow for cellular trafficking, live-cell microscopy revealed fluorescent cholesterol in vesicular structures and the C. burnetii PV (see Fig. S1A in the supplemental material). We next examined trafficking of exogenous LDL, a major cholesterol-binding protein internalized by cells through receptor-mediated endocytosis. Following incubation of infected DHCR24 Ϫ/Ϫ MEFs with fluorescent LDL for 4 h, fluorescent LDL was found in the PV membrane and PV lumen (Fig. S1B). Thus, at least two sources of cholesterol travel to the PV, suggesting that cholesterol supplementation of cholesterol-free MEFs increases PV cholesterol.

U18666A-induced cholesterol accumulation in the PV is bacteriolytic.
On the basis of our cholesterol trafficking data, we hypothesized that accumulation of cholesterol in the PV leads to C. burnetii death. The cholesterol-altering drug U18666A blocks cholesterol transport from endosomes and lysosomes and results in cholesterol accumulation in the endolysosomal system (31). To determine whether U18666A also leads to cholesterol accumulation in the PV, we treated infected HeLa cells with U18666A and then stained with filipin to label cholesterol and antibodies against LAMP-1 and C. burnetii. As previously shown, filipin staining of untreated infected cells shows the presence of cholesterol or other sterols in the PV membrane (Fig. 3A, top). However, after U18666A treatment, there is a significant increase in filipin labeling in and around the PV (Fig. 3A, bottom), suggesting an increase in PV cholesterol. By live-cell microscopy, we observed a significant number of lytic PVs shortly after adding U18666A to mCherry-C. burnetii-infected HeLa cells. When quantitated after a 6 h treatment of either 1 M and 5 M U18666A, approximately 30% and 85% of PVs were lytic, respectively (Fig. 3B). Furthermore, the presence of lytic PVs corresponded with a 30 to 60% loss of bacterial viability, depending on the U18666A concentration (Fig. 3C). Cumulatively, these data suggest that U18666A-induced cholesterol accumulation in the PV leads to C. burnetii death.
To address the possibility that cholesterol could be directly killing C. burnetii, we measured growth of axenic C. burnetii cultures in the presence or absence of cholesterol. Compared to the carrier protein bovine serum albumin (BSA), cholesterol had no effect on C. burnetii growth (Fig. S2A). Further, the bacteriolytic effect of U18666A is not due to a direct effect on C. burnetii, as the addition of U18666A to C. burnetii axenic cultures did not affect viability (Fig. S2B). These data indicate that the toxic effect of cholesterol and U18666A is specific to the C. burnetii intracellular niche.
C. burnetii growth is most sensitive to cholesterol during the early stages of PV biogenesis. During the first 24 to 48 h of infection, C. burnetii resides in a tight-fitting PV, as the nonreplicating small-cell variant (SCV) transitions into the replicationcompetent large-cell variant (LCV). During this time, the C. burnetii T4SS also secretes effector proteins into the host cell cytoplasm, with secretion detected as early as 8 h postinfection in HeLa cells and 1 h postinfection in mouse bone marrow-derived macrophages (32). PV expansion around 48 h postinfection coincides with the beginning of log-phase growth; LCVs transition back to SCVs around 5 or 6 days postinfection (33). To determine whether C. burnetii is sensitive to cholesterol at specific stages of infection, we added cholesterol to MEFs at 24 h intervals postinfection and assessed the effect on PV size and bacterial viability. When cholesterol was added at the time of infection or 1 day postinfection, the final PVs at 6 days postinfection were significantly smaller than those in cholesterol-free MEFs (Fig. 4A). However, there was no significant effect on final PV size when cholesterol was added at any time after 1 day postinfection, regardless of the cholesterol concentration. Similarly, the fold change in recoverable bacteria over 6 days was sensitive to cholesterol only during the first 2 days of infection, with an approximately 80% decrease in bacterial viability in MEFs with cholesterol compared to cholesterol-free MEFs (Fig. 4B). While not statistically significant, cholesterol addition after day 4 trended toward slightly larger PVs, although there was no effect on bacterial viability. Based on these data, C. burnetii is sensitive to cholesterol only during the early stages of PV expansion and log growth. Coxiella burnetii Cholesterol Sensitivity ® Lytic PVs are nonfusogenic. The C. burnetii PV is a highly dynamic vacuole, promiscuously fusing with vesicles from the endocytic pathway (34). We previously observed cholesterol-dependent fusion between late endosomes and the PV, based on a lack of the late endosome marker CD63 in the PV lumen in cholesterol-free MEFs (20). To further characterize the effect of cholesterol on endosomal trafficking to the PV, we developed a quantitative fusogenicity assay utilizing fluorescent dextran. Dextran is internalized by cells through non-receptor-mediated endocytosis and accumulates in the PV lumen following fusion between endosomes and the PV (35). To measure PV-endosome fusion, we quantitated dextran accumulation in the PV lumen in MEFs with or without cholesterol. MEFs infected with mCherry-C. burnetii were pulsed for 10 min with fluorescent dextran and then imaged for 40 min using live-cell confocal microscopy ( Fig. 5A and B). The accumulation of dextran in the PV lumen in individual PVs was determined by measuring the fold change in fluorescence intensity over  40 min. An average of 4.2-fold increase in fluorescent dextran was observed in PVs from cholesterol-free MEFs (Fig. 5C). In cells with cholesterol, PVs that were not yet lytic were less fusogenic, with an average of 2.5-fold increase in fluorescent dextran. In contrast, lytic PVs (defined as having a twofold increase in PV lumen mCherry fluorescence compared to background) accumulated little to no fluorescent dextran, indicating that PVs containing degraded bacteria are no longer fusogenic with the endocytic pathway.
Cholesterol further acidifies the C. burnetii PV. Experiments in cholesterol-rich macrophages have shown that cholesterol affects the ability of lysosomes to maintain an acidic pH (36), raising the possibility that cholesterol influences PV pH. Prior studies using ratiometric pH measurements with fluorescein found the C. burnetii PV pH to be approximately 4.8 (37,38). Further, C. burnetii metabolism is activated by acidic pH (39,40), and blocking PV acidification during phagosome maturation using the vacuolar ATPase (vATPase) inhibitor bafilomycin A1 inhibits bacterial growth (35). Collectively, these data demonstrate that PV pH is a critical component of C. burnetii pathogenesis.
To determine whether PV pH is affected by cholesterol, we measured PV pH under different cholesterol conditions using a microscopy-based ratiometric fluorescence assay (41). The pH-sensitive fluorophore Oregon Green 488 is a fluorescein derivative with a pK a of 4.7, making Oregon Green 488 more accurate in acidic environments compared to fluorescein (pK a of 6.4) (41). MEFs with and without cholesterol were infected with mCherry-C. burnetii for 3 days and then incubated with pH-sensitive Oregon Green 488 dextran and pH-insensitive Alexa Fluor 647 dextran for 4 h to allow for dextran trafficking to the PV. It is important to note that because this assay relies on dextran trafficking to the PV, only nonlytic PVs in MEFs with cholesterol can be analyzed, as lytic PVs are nonfusogenic (Fig. 5). The fluorescence intensities of Oregon Green 488 and Alexa Fluor 647 were measured for each PV, and a ratio of Oregon Green 488 to Alexa Fluor 647 was compared to a standard curve to generate individual PV pH measurements. In cholesterol-free MEFs, the average PV pH was 5.1 with a range of 3.95 to 7.26 ( Fig. 6A and Fig. S3). Interestingly, PVs in MEFs with cholesterol were significantly more acidic, ranging between 2.33 and 6.02 and an average pH of 4.4.
Given that U18666A also leads to lytic PVs, we measured PV pH before and after the addition of U18666A to infected HeLa cells. After incubation with Oregon Green 488 dextran and Alexa Fluor 647 dextran for 4 h, PVs were imaged prior to adding dimethyl sulfoxide (DMSO) or U18666A and then imaged every 15 min for 2 h. Similar to cholesterol-free MEFs, the average PV pH prior to drug addition was approximately 5.1 (Fig. 6B and Fig. S4). While DMSO-treated cells maintained this pH over 2 h, in cells treated with U18666A, the PV further acidified to an average pH of 4.4 within 30 min. While these data suggest that cholesterol accumulation leads to PV acidification, we next tested whether increased acidification was responsible for C. burnetii degradation. The proton pump vacuolar ATPase is responsible for PV acidification and can be blocked using bafilomycin A1 (35). Bafilomycin A1 recovered bacterial viability in the presence of U18666A (Fig. 6C), indicating that increased acidification of the PV leads to C. burnetii death.
Cholesterol-altering FDA-approved drugs lead to C. burnetii lysis. A previous study discovered that numerous drugs from a FDA-approved drug library altered cholesterol homeostasis in HeLa cells and inhibited intracellular C. burnetii growth (19). To determine whether these drugs worked through a mechanism similar to that of U18666A, we tested a small subset of drugs for their ability to induce lytic PVs in HeLa cells. These drugs were chosen based on their ability to (i) cause cholesterol accumulation in endosomes, (ii) block intracellular C. burnetii growth, (iii) not affect C. burnetii growth in axenic media, and (iv) have low toxicity to host cells (19) (Table S1). Six of the eight drugs we tested led to lytic PVs after a 3 h treatment (Fig. S5A). In particular, loperamide, clemastine, and amiodarone resulted in nearly all of the PVs containing degraded bacteria, with few intact bacteria ( Fig. S5B and data not shown). These observations correlated with almost no bacteria recovered after 3 h, as measured by a FFU assay (Fig. 7A). Furthermore, concurrent incubation with bafilomycin A1 at least partially rescued killing by amiodarone, clemastine, haloperidol, and spiperone, as measured by a decrease in the number of lytic PVs (Fig. S5A) and an increase in bacterial viability (Fig. 7A). This suggests that similar to U18666A, these drugs increase the acidity of the PV.
To determine whether these drugs led to altered cholesterol levels in the PV, we observed cholesterol by staining with filipin. While all of the drugs tested appeared to alter cholesterol distribution, several phenotypes were observed (summarized in Table S1). Compared to DMSO-treated cells, amiodarone did not significantly alter overall fillipin labeling intensity of endosomes. However, unlike filipin labeling being restricted to the PV membrane of control cells, amiodarone treatment led to significant filipin labeling of the PV lumen (Fig. 7B, insets). In contrast, filipin primarily labeled the PV membrane in clemastine-treated cells, but there was a significant increase in endosomal filipin labeling intensity. Haloperidol-treated cells had a phenotype similar to that of clemastine-treated cells, while the phenotype of loperamide-treated cells was more similar to that of U18666A-treated cells (Fig. S5). Intriguingly, spiperone-treated cells appeared to have filipin levels similar to those in control cells, with some of the PVs showing labeling within the PV lumen (Fig. S5). Together with the bafilomycin data, this suggests that there might be multiple mechanisms by which altering host cholesterol leads to increased acidification of the C. burnetii PV, and ultimately, death of the bacteria. Interestingly, bafilomycin did not rescue the effects of loperamide, suggesting that this drug may act through a different mechanism.

DISCUSSION
Cholesterol is a critical lipid constituent of cellular membranes, regulating membrane dynamics, trafficking, and signaling. Due to its involvement in important host cell processes, an increasing number of pathogens, including Leishmania spp., Salmonella enterica, Staphylococcus aureus, Mycobacterium spp., and Listeria monocytogenes, have been reported to exploit host cell cholesterol (3)(4)(5)(11)(12)(13). To understand the role of cholesterol during C. burnetii-host interaction, we utilized a cholesterol-free tissue culture model system that lacks both endogenous cholesterol (from biosynthesis) and exogenous cholesterol (from serum). In a previous study, we used this system to establish that C. burnetii entry into fibroblasts occurred through lipid raft-mediated ␣ v ␤ 3 signaling (20). In addition, with the exception of CD63, endolysosomal markers were associated with the C. burnetii PV regardless of the presence or absence of cholesterol, indicating that PV maturation was not cholesterol dependent. Here, we made the surprising discovery that increasing cellular cholesterol is detrimental to C. burnetii survival. C. burnetii is most sensitive to cholesterol during the early stages of infection, with increasing cholesterol levels leading to altered PV fusion, increased acidity, and bacterial degradation. Cholesterol traffics to the PV and drugs that trap cholesterol in the endolysosomal system are bactericidal, suggesting that PV cholesterol influences PV biology and C. burnetii pathogenesis. These data strongly support the conclusion that manipulating cholesterol in the bacterium-containing PV kills C. burnetii.
C. burnetii growth is sensitive to drugs that target cholesterol biosynthesis and uptake (19,21). Further, Czyz et al. reported that treatment of THP-1 cells with FDA-approved drugs that alter cellular cholesterol distribution similar to U18666A also inhibit C. burnetii intracellular growth (19). In support of these data, they also showed decreased C. burnetii growth in THP-1 macrophage-like cells deficient in NPC-1, a cholesterol transporter that facilitates cholesterol export from late endosomes and lysosomes (19). We found that both plasma membrane cholesterol and the cholesterolbinding exogenous protein LDL travel to the PV. We hypothesize that cholesterol supplementation of cholesterol-free MEFs leads to increased cholesterol in the PV membrane compared to cholesterol-free MEFs. Our model system further revealed that C. burnetii PV size and bacterial growth are sensitive to cellular cholesterol, further supporting the hypothesis that manipulating host cell cholesterol homeostasis adversely affects C. burnetii infection. Importantly, our data show that rather than simply blocking bacterial growth or PV formation, increasing PV cholesterol leads to C. burnetii lysis. Remarkably, treatment for only 3 h with U18666A, which traps cholesterol in the PV, killed 80% of the bacteria. These data, along with data from Czyz et al. (19), suggest that C. burnetii is sensitive to altered cholesterol distribution within the cell, particularly accumulation of cholesterol in the endosomal trafficking pathway and the PV. With other bacteria, including Chlamydia trachomatis, Staphylococcus aureus, and Mycobacterium spp., the presence of cholesterol is reported to be beneficial to the bacterium, with cholesterol depletion leading to reduced bacterial growth (11,18,(42)(43)(44)(45)(46). The unique sensitivity of C. burnetii to host cell cholesterol may reflect the distinctive intracellular niche this bacterium occupies.
Cholesterol is a key regulator of endosomal trafficking and fusion (47)(48)(49). We previously found that the endosome marker CD63 was absent in the PV lumen in cholesterol-free MEFs, suggesting that cholesterol is required for late endosomal trafficking to the PV (20). In this study, we discovered that lytic PVs containing degraded C. burnetii, which are found only in cells with cholesterol, were no longer fusogenic with the endosomal pathway. Most likely, bacterial degradation leads to a loss of the T4BSS effector proteins required to maintain PV fusogenicity. The C. burnetii T4BSS is not required for short-term intracellular survival, with a T4BSS mutant persisting for several days in a viable form (25). Thus, it is unlikely that a cholesterol-dependent loss in PV fusogenicity would lead to bacterial degradation. However, cholesterol is specifically toxic to C. burnetii during the initial stages of PV biogenesis and expansion, and it is possible that cholesterol plays a role in activating T4BSS secretion early during PV development. Most likely, both the timing and amount of PV cholesterol are tightly regulated by the bacteria to regulate PV-endosome fusion.
In addition to altered fusogenicity, the PVs from MEFs with cholesterol are significantly more acidic than PVs from cholesterol-free cells. Further, U18666A treatment also increased PV acidity in a vATPase-dependent manner. Importantly, blocking acidification by vATPase also rescued bacterial viability, demonstrating that increasing the PV acidity kills C. burnetii. This is a surprising finding, given that C. burnetii metabolism is activated by acid (35,39,40) and the PV pH has previously been reported to be approximately 4.8 (37,38). Our studies found the PVs in cholesterol-free MEFs and HeLa cells to be slightly more alkaline at pH 5.2. The differences in measured pH between studies might be a result of different pH-sensitive reagents. We utilized the fluorescein derivative Oregon Green 488, given the improved pH sensitivity of Oregon Green 488 in acidic environments (41). A recent study on the role of lysosome-associated membrane glycoprotein (LAMP) proteins in PV maturation found the pH of the PV to be between 4.0 and 4.5 using LysoSensor Yellow/Blue DND-160 (50). In our hands, this reagent stained the bacteria and was not free in the PV lumen, prompting us to utilize the dual dextran labeling approach (41). Regardless, it is clear that the PV is more acidic in MEFs with cholesterol and HeLa cells treated with U18666A, and this increased acidity kills C. burnetii. However, the mechanism behind C. burnetii degradation is not known. Host cathepsin D and lysosomal acid phosphatases accumulate in the PV (35), and the PV is proteolytically active, presumably due to the presence of host proteases (51). It is possible that increased acidity further activates lysosomal degradative enzymes beyond the threshold the bacteria can survive. Detailed characterization of the PV proteolytic activity is needed to fully understand how C. burnetii survives in this environment.
Previously, a drug screen revealed that C. burnetii growth is sensitive to 57 FDAapproved drugs that perturb host cell cholesterol homeostasis (19,21). We used several drugs from this screen to further validate our hypothesis that C. burnetii lysis was due to cholesterol-induced changes in the PV pH. Treatment with six of the eight selected drugs resulted in significant C. burnetii lysis which could be at least partially rescued by blocking acidification through vATPase. Importantly, these drugs were shown to have little to no effect on C. burnetii growth in axenic media or on host cell viability (19). While the mode of action differs between these drugs, we confirmed the results of previous studies that they altered cholesterol distribution within the cell (19). Treatment with several of these drugs appeared to increase cholesterol within the PV. Together, these findings reveal a potential vulnerability in the C. burnetii lifestyle which could be targeted with currently available drugs.
The cholesterol-mediated negative effect on intracellular C. burnetii raises intriguing questions as to how C. burnetii successfully colonizes cholesterol-containing cells during natural infection, given that cholesterol is an essential lipid for host cells outside the laboratory setting. Accumulating evidence suggests that C. burnetii possesses multiple mechanisms to manipulate host cholesterol metabolism. For example, Howe and Heinzen reported differential expression of cholesterol biosynthesis-related genes in C. burnetii-infected Vero cells (21). Expression profiling of C. burnetii-infected THP-1 cells suggests that C. burnetii actively upregulates expression of apoE and plin2, which are involved in cholesterol efflux and storage, respectively (26,27). Beyond gene expression, cholesterol storage organelles called lipid droplets have been observed in and around the PVs of infected primary human alveolar macrophages (52). It is possible that C. burnetii targets the carefully regulated host cholesterol homeostasis, upregulating storage and efflux while also decreasing biosynthesis. In addition, we recently showed that C. burnetii recruits the host cell sterol-binding protein ORP1L to the PV, where it participates in membrane contact sites between the PV and endoplasmic reticulum (53). Finally, C. burnetii expresses two eukaryote-like sterol reductase enzymes that could modify cholesterol (54). This intriguing possibility might explain the intense filipin labeling of the PV, with a bacterium-derived ␤-hydroxysterol other than cholesterol dominating the PV membrane.
In summary, our data suggest that the presence of cholesterol in the PV during the initial phases of PV formation negatively affects PV formation and C. burnetii survival. While not absolutely required for C. burnetii growth, some cholesterol is needed for optimal PV development through fusion with late endosomes (20). However, too much PV membrane cholesterol leads to increased PV acidification, decreased fusion with endosomes, and eventual bacterial degradation. We propose that the amount of cholesterol in the PV membrane regulates key aspects of PV function, and C. burnetii must maintain a delicate balance of PV membrane cholesterol. This would explain the unique sensitivity of C. burnetii to drugs that target different aspects of host cholesterol metabolism: any slight shift in host cholesterol homeostasis would impact PV membrane cholesterol levels. Identifying both the bacterial and host pathways involved in this delicate balance may yield novel targets to treat or prevent C. burnetii pathogenesis.
PV measurements. A total of 5 ϫ 10 4 MEFs were plated onto ibidi-treated channel slide VI 0.4 (3 ϫ 10 3 cells per channel; ibidi USA Inc., Verona, WI) and allowed to adhere overnight. After the MEFs were infected with C. burnetii for 1 h, they were washed with phosphate-buffered saline (PBS) to remove extracellular bacteria and incubated in media containing the indicated cholesterol concentrations. At different time points postinfection, cells were fixed with 2.5% paraformaldehyde (PFA) on ice for 15 min and then permeabilized/blocked for 15 min with 0.1% saponin and 1% bovine serum albumin (BSA) in PBS. The cells were incubated with rat anti-LAMP1 (catalog no. 553792; BD Biosciences, San Jose, CA) and rabbit anti-C. burnetii primary antibodies in saponin-BSA-PBS for 1 h, followed by Alexa Fluor secondary antibodies (Invitrogen) for 1 h. Following washing with PBS, ProLong Gold with 4=,6=-diamidino-2phenylindole (DAPI) (Invitrogen) was added, and the cells on the slides were visualized on a Leica inverted DMI6000B microscope (63ϫ oil immersion objective). Images were captured and processed identically, and a cross-sectional area through the middle of the PV was measured using ImageJ software. Approximately 20 PVs were measured per condition for each of three independent experiments.
C. burnetii growth in MEFs. MEFs were plated at 1 ϫ 10 5 cells/well in a six-well plate under different cholesterol conditions and allowed to adhere overnight. After the MEFs were infected with C. burnetii for 1 h in 500 l medium, the wells were washed with PBS to remove extracellular bacteria and then gently scraped into 3 ml of medium. For the day 0 sample, 1 ml of the cell suspension was centrifuged at 20,000 ϫ g for 10 min, and the pellet was frozen at Ϫ20°C. The remaining cells were left in the six-well plate in medium supplemented with cholesterol. The medium was changed daily to ensure constant cholesterol concentrations. At 6 days postinfection, the cells were harvested by scraping the cells into the growth medium and centrifuging at 20,000 ϫ g for 10 min. Bacterial DNA was extracted from the pellets using the UltraClean microbial DNA isolation kit (Mo Bio Laboratories, Carlsbad, CA) according to the manufacturer's instructions. Quantitative PCR for genome equivalents was performed using a primer set specific for dotA (30) and Luminaris Color HiGreen quantitative PCR (qPCR) master mix (Thermo Scientific) with an Applied Biosystems 7500 real-time PCR cycler. Each experiment was done in duplicate.
Quantitation of lytic PVs containing lysed C. burnetii. DHCR24 Ϫ/Ϫ MEFs were plated under different cholesterol conditions at 5 ϫ 10 4 cells per well of a six-well plate and infected with mCherryexpressing C. burnetii (mCherry-C. burnetii) for 1 h as described above. Approximately 24 h later, the cells were scraped into fresh medium, resuspended to 1 ϫ 10 5 cells/ml, and plated onto ibidi-treated channel slide VI 0.4 (3 ϫ 10 3 cells per channel). The medium was changed daily, and cells were examined live every 24 h on a Leica inverted DMI6000B microscope with a 63ϫ oil immersion objective. PVs with visible mCherry fluorescence in the PV lumen were scored as "lytic PVs" with 50 PVs scored for each condition for three individual experiments.
HeLa cells (5 ϫ 10 4 ) were infected with mCherry-C. burnetii in 6-well plates for 1 h. At 2 days postinfection, the cells were trypsinized and resuspended to 1 ϫ 10 5 cells/ml, and plated onto ibidi-treated channel slide VI 0.4 (3 ϫ 10 3 cells per channel; Ibidi). At 3 days postinfection, dimethyl sulfoxide (DMSO) control, U18666A (1 or 5 M), or the indicated FDA-approved drugs (see Table S1 in the supplemental material; obtained from Sigma and used at a final concentration of 20 M) with or without vATPase inhibitor bafilomycin A1 (100 nM) were added to the cells and incubated for the time indicated prior to counting lytic PVs as described above. At least 50 PVs were scored for each condition for three individual experiments. C. burnetii viability by fluorescent infectious focus-forming unit (FFU) assay. To test viability of C. burnetii in MEFs under different cholesterol conditions, 1 ϫ 10 4 cells/well were infected with C. burnetii for 1 h in a 48-well plate, washed extensively with PBS, and incubated with media containing different cholesterol concentrations. At the indicated time points, cells were incubated for 5 min with sterile water, pipetted up and down to lyse cells, and diluted 1:5 in RPMI 1640 with 2% FBS (2% FBS-RPMI). Serial dilutions were added to confluent monolayers of Vero cells in a 24-well plate and incubated for 5 days. Plates were fixed with methanol and stained with rabbit anti-C. burnetii antibody and DAPI to confirm monolayer integrity. Four fields per well were captured on an Evos automated microscope (Thermo Fisher) with a 4X objective, and fluorescent focus units were quantitated using ImageJ. Each experiment was done in duplicate.
To determine bacterial viability in drug-treated cells, HeLa cells were plated at 5 ϫ 10 4 cells/well in a six-well plate and infected with mCherry-C. burnetii. At 2 days postinfection, the cells were trypsinized and replated in 24-well plates at 5 ϫ 10 4 cells/well. Approximately 16 h later, the cells were treated with DMSO or drug with or without the vATPase inhibitor bafilomycin A1 (100 nM) for the time indicated, at which point the medium was aspirated from the 24-well plate and the cells were lysed by incubation in sterile water for 5 min. After the cells were pipetted up and down, the released bacteria were diluted 1:5 in 2% FBS-RPMI and plated in 10-fold serial dilutions onto confluent Vero cell monolayers in a 96-well ibidi-treated plate (ibidi). The plate was fixed with 2.5% PFA 5 days later and stained with DAPI, and the number of fluorescent foci was determined as described above. Each experiment was done in duplicate.
Microscopy for cholesterol trafficking. To monitor trafficking of plasma membrane cholesterol, fluorescent cholesterol (TopFluor cholesterol; Avanti Polar Lipids) was resuspended at 20 mg/ml in ethanol. Twenty microliters of this solution was added to 1 ml of 10% methyl-beta-cyclodextrin (Sigma) in serum-free RPMI 1640 medium. The solution was sonicated in a water bath sonicator (Avanti) for 30 s, and insoluble material was pelleted by spinning for 2 min at 20,000 ϫ g. MEFs with cholesterol were infected with mCherry-C. burnetii and plated onto an ibidi slide as described above. At 3 days postinfection, fluorescent cholesterol (final concentration of 30 g/ml) was added to the cells for 24 h. Live-cell images were taken with a modified PerkinElmer UltraView spinning disk confocal connected to a Nikon Eclipse Ti-E inverted microscope with a 63ϫ oil immersion objective.
For trafficking of BODIPY-LDL, MEFs were infected and plated onto an ibidi slide as described above. The cells were incubated for 5 min on ice with 25 g/ml BODIPY-LDL (Invitrogen), washed twice with medium, and visualized after 4 h of incubation at 37°C.
To visualize endogenous free sterols after drug treatment, infected cells were plated onto ibidi slides and treated as described above. The cells were fixed with 2.5% PFA on ice for 15 min and incubated with 1:100 filipin (Cayman Chemicals, Ann Arbor, MI) in PBS with 1% BSA for 1 h. After the cells were washed with PBS three times, they were incubated with rat anti-LAMP1 for MEFs (catalog no. 553792; BD Biosciences, San Jose, CA) or mouse anti-CD63 for HeLa cells for 1 h, followed by three washes in PBS and a 1 h incubation with Alexa Fluor 488 anti-mouse secondary antibody. After the cells were washed three times with PBS, ProLong Gold was added to the wells, and samples were visualized on a Leica inverted DMI6000B microscope (63ϫ oil immersion objective). Images were captured under identical capture settings and processed identically using ImageJ.
C. burnetii growth in cell-free media. To complex cholesterol to BSA, 500 g of cholesterol (10 mg/ml chloroform stock; Avanti) was dried down in a glass tube under a nitrogen stream. The lipid film was resuspended into 2.5 ml of 5% fatty acid-free BSA using a water bath sonicator (Avanti). The resulting 200 g/ml stock was sterile filtered and added to a final concentration of 5 g/ml in ACCM-2 (40). C. burnetii bacteria were diluted to approximately 1 ϫ 10 5 genomes/ml in ACCM-2 with BSA or BSA-cholesterol, and 7 ml was transferred to a T25 flask and incubated as previously described (40). Every 24 h, 50 l was removed and added to a tube with 150 l PBS and a half volume of 0.1-mm zirconia-silica beads (BioSpec Products, Bartlesville, OK). Bacteria were lysed by bead beating in a FastPrep FP120 (Thermo Scientific) and analyzed by qPCR as previously described (20). Each experiment was done in duplicate.
To test bacterial sensitivity to U18666A, ACCM-2 was inoculated at approximately 1 ϫ 10 5 bacteria/ml with mCherry-C. burnetii and grown for 5 days as previously described (40). Bacteria (500 l) were treated for 6 h with DMSO or U18666A in 24-well plates under normal C. burnetii culture conditions. The bacteria were diluted 1:10 in 2% FBS-RPMI prior to the FFU assay in 96-well ibidi-treated plates as described above.
Dextran trafficking. Cells were infected with mCherry-C. burnetii in six-well plates and replated onto ibidi slides at 2 days postinfection as described above. PVs were selected and marked in Elements software on the spinning disk confocal microscope in a live-cell environmental chamber. Individual ibidi channels were pulsed with 1 mg/ml Alexa Fluor 488-dextran (molecular weight [MW] of 10,000) for 10 min in medium, followed by four washes with medium to remove uninternalized dextran and finally replaced with either basal medium or basal medium with cholesterol (5 g/ml). The PVs were then focused, and confocal images through the entire PV were obtained every 6.33 min for 38 min. The fluorescence intensity of dextran inside the PV was calculated using the average intensity multiplied by the PV volume using ImageJ.
PV pH measurements. The pH measurement was performed as previously described with slight modifications (57). Briefly, MEFs were infected with mCherry-C. burnetii in six-well plates, incubated with and without cholesterol, and replated onto ibidi slides at 2 days postinfection as described above. For measurement of PV pH in U18666A-treated cells, HeLa cells were infected with mCherry-C. burnetii in six-well plates and replated onto ibidi plates at 2 days postinfection. Under both conditions, at 3 days postinfection, cells were incubated with pH-sensitive Oregon Green 488 dextran (MW, 10,000; Invitrogen) and pH-stable Alexa Fluor 647 dextran (MW, 10,000; Invitrogen) for 4 h at a concentration of 0.5 mg/ml. MEFs were imaged directly with a 63ϫ oil immersion objective under identical capture settings.
To measure the time-dependent change in PV pH, individual PVs were selected and imaged, and then treated with 5 M U18666A or DMSO as a vehicle control. Starting from 15 min after the treatment, cells were then imaged every 15 min for the next 2 h. The PV fluorescence intensity was measured using ImageJ, and the Oregon Green 488/Alexa Fluor 647 ratio was calculated. To generate a standard curve for MEFs and HeLa cells, the respective infected cells were incubated with the ionophores nigericin (10 M) and monensin (10 M) for 5 min at room temperature, followed by buffers with different pHs (pH 4.0 to 7.0) before imaging. At least 20 PVs were imaged at each pH for every experiment, and the ratio of fluorescence intensity at 488/647 nm were plotted against the pH of the respective buffer to obtain a sigmoidal standard curve.
Data analyses. Image processing and analysis were done with ImageJ software (W. S. Rasband, National Institutes of Health, Bethesda, MD) (58). Statistical analyses were performed using unpaired two-tailed t test, ordinary one-way or two-way analysis of variance (ANOVA) with Tukey's or Dunnett's multiple-comparison test in Prism (GraphPad Software, Inc., La Jolla, CA).