- Record: found
- Abstract: found
- Article: found
N-Desmethylclozapine, Fluoxetine, and Salmeterol Inhibit Postentry Stages of the Dengue Virus Life Cycle
research-article
Author(s):
Guruprasad R. Medigeshi
a
,
,
Rinki Kumar
a
,
b ,
Ekta Dhamija
a ,
Tanvi Agrawal
a ,
Meenakshi Kar
a
Publication date (Electronic):
21 October 2016
Publisher:
American Society for Microbiology
Read this article at
There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.
Abstract
Around 10,000 people die each year due to severe dengue disease, and two-thirds of the world population lives in a region where dengue disease is endemic. There has been remarkable progress in dengue virus vaccine development; however, there are no licensed antivirals for dengue disease, and none appear to be in clinical trials. We took the approach of repositioning approved drugs for anti-dengue virus activity by screening a library of pharmacologically active compounds. We identified N-desmethylclozapine, fluoxetine hydrochloride, and salmeterol xinafoate as dengue virus inhibitors based on reductions in the numbers of infected cells and viral titers. Dengue virus RNA levels were diminished in inhibitor-treated cells, and this effect was specific to dengue virus, as other flaviviruses, such as Japanese encephalitis virus and West Nile virus, or other RNA viruses, such as respiratory syncytial virus and rotavirus, were not affected by these inhibitors. All three inhibitors specifically inhibited dengue virus replication with 50% inhibitory concentrations (IC 50s) in the high-nanomolar range. Estimation of negative-strand RNA intermediates and time-of-addition experiments indicated that inhibition was occurring at a postentry stage, most probably at the initiation of viral RNA replication. Finally, we show that inhibition is most likely due to the modulation of the endolysosomal pathway and induction of autophagy.
Related collections
Most cited references38
- Record: found
- Abstract: found
- Article: not found
Efficacy and Long-Term Safety of a Dengue Vaccine in Regions of Endemic Disease.
A candidate tetravalent dengue vaccine is being assessed in three clinical trials involving more than 35,000 children between the ages of 2 and 16 years in Asian-Pacific and Latin American countries. We report the results of long-term follow-up interim analyses and integrated efficacy analyses.
- Record: found
- Abstract: found
- Article: not found
Dissecting the Cell Entry Pathway of Dengue Virus by Single-Particle Tracking in Living Cells
Hilde van der Schaar, Michael J. Rust, Chen Chen … (2008)
Introduction Dengue virus (DENV) is a mosquito-transmitted, enveloped RNA virus that belongs to the family Flaviviridae. This family also includes West-Nile virus (WNV) and tick-borne encephalitis virus (TBEV). DENV causes the most common arthropod-borne infection worldwide with 50–100 million cases annually [1]–[3]. Despite its threat to human health, there are presently neither vaccines nor antiviral drugs to prevent or treat dengue infection. The development of novel therapies requires insight into the viral life cycle. A potential target for intervention strategies is the infectious cell entry pathway. The infectious entry of DENV is mediated by the viral envelope glycoprotein E, which is organized in 90 homodimers on the surface of the virion [4],[5]. The E glycoprotein is involved in interaction with cellular receptors as well as the subsequent membrane fusion process [6]–[8]. In vitro studies with TBEV indicate that membrane fusion is triggered upon exposure of the virus to low pH [8]. At low pH, the E proteins undergo a dramatic re-organization which leads to the formation of E trimers [9]. The crystal structure of the E protein has been solved in its dimeric pre-fusion, and trimeric post-fusion configurations [10],[11]. Although much is known about the molecular mechanisms involved in the membrane fusion process, many critical questions regarding the cell entry pathway of flaviviruses remain unanswered. The cell entry mechanism of DENV remains controversial. Early electron microscopy studies provided evidence for direct fusion with the plasma membrane [12],[13], whereas a recent study indicates that DENV enters cells via clathrin-mediated endocytosis [14]. Clathrin-mediated endocytosis involves internalization of ligands and receptors through a clathrin-coated pit, which buds into the cell cytosol and delivers its cargo to early endosomes and subsequently to late endosomes and lysosomes [15]–[17]. Other flaviviruses have also been described to infect their host cell via clathrin-mediated endocytosis [18]–[21]. Evidence for flavivirus entry via this pathway is based on the use of inhibitors of clathrin-mediated uptake, such as chlorpromazine and dominant-negative mutants of Eps15 [18],[20],[22]. Furthermore, addition of acidotropic reagents to cells has been observed to dramatically reduce viral infectivity and membrane fusion activity, suggesting that flaviviruses mediate membrane fusion from within acidic endosomes [23]–[26]. A recent study on the entry of WNV particles demonstrates that WNV colocalizes with the early endosome marker EEA-1 (Early Endosome Antigen-1), and at later time points with the late endosome/lysosome marker LAMP-1 (Lysosome-Associated Membrane Protein-1) [27]. Taken together, these studies suggest clathrin-mediated endocytosis as a viable pathway for flavivirus entry, but the exact manner in which DENV virus particles enter cells and traffic through the endocytic network remains unclear, as does the identity of the organelle in which viral fusion occurs. In this study, we dissected the cell entry pathway of DENV by tracking fluorescently labeled DENV particles in living cells expressing various fluorescent cellular markers using real-time multi-color fluorescence microscopy. These experiments demonstrate that DENV infects its host cell via clathrin-mediated endocytosis. DENV particles move on the cell surface in a diffusive manner until they join a pre-existing clathrin-coated pit. Following clathrin-mediated uptake, the majority of DENV particles enter early endosomes that progress to late endosomes, where membrane fusion occurs. Results Dengue Virus Enters Cells via Clathrin-Mediated Endocytosis In order to visualize single DENV particles in living cells, we labeled the virus with the lipophilic fluorescent probe DiD. The concentration of the DiD dye in the viral membrane is sufficiently high so that its fluorescence is largely quenched, but still allows single DiD-labeled virions to be detected. Membrane fusion can be observed as fluorescence dequenching. We have shown previously that this labeling procedure does not affect the infectious properties of DENV [26]. The tracking experiments were performed in African green monkey kidney cells (BS-C-1), which are highly permissive to DENV infection [26],[28]. To test whether DENV is internalized through clathrin-mediated endocytosis, BS-C-1 cells stably expressing enhanced yellow fluorescent protein (eYFP) fused to the light chain of clathrin (LCa-eYFP) were used. We and others have previously shown that LCa-eYFP highlights more than 95% of the coated pits and vesicles in living cells and that this fusion protein does not disturb the functional integrity of clathrin molecules [29],[30]. DiD-labeled DENV was added in situ to these cells at 37°C and fluorescent images were recorded at 2 frames per second for 25 min. Figure 1A shows the distribution of LCa-eYFP (green) and DiD-labeled DENV particles (red) in a cell. The LCa-eYFP signal appeared as discrete structures in cells. A typical example of a DENV entry event is shown in Figure 1B. In this example, the virus particle first binds to and moves along the cell surface. Forty-eight seconds post-binding, the virus particle associates with a discrete spot containing the LCa-eYFP signal. Thereafter, the clathrin signal around the virus particle increases, indicating maturation towards a clathrin-coated vesicle. At 94 seconds, the clathrin signal rapidly disappears, presumably due to uncoating of the clathrin-coated vesicle. Membrane fusion eventually occurs at 512 seconds post-infection. Quantitative analysis of 47 virus trajectories revealed that 98% of the DENV particles that fused with endosomes entered through LCa-eYFP positive clathrin-coated pits. On average, the clathrin signal colocalized with the virus particle for 83 seconds (Figure 1C), which is consistent with previously observed dynamics of clathrin-mediated endocytosis [30],[31]. 10.1371/journal.ppat.1000244.g001 Figure 1 DENV enters cells via clathrin-mediated endocytosis. (A) Fluorescent image of a LCa-eYFP-expressing cell (green) infected with DiD-labeled DENV particles (red). Scale bar is 10 µm. (B) Selected frames from a single DENV particle (red), indicated with the arrow, entering a cell via clathrin-mediated endocytosis. The numbers designate seconds post-binding. (C) Histogram of the time that DENV particles colocalized with clathrin-eYFP. (D) DENV infectivity in BS-C-1 cells in the absence or presence of 15 µM chlorpromazine. Viral infectivity was measured at 24 hours post-infection by counting the number of E protein–expressing cells using immunofluorescence microscopy. The experiment was performed in triplicate, and the bars represent the average±SD. * P<0.01. (E) DENV infection in HeLa cells expressing a dominant-negative Eps15 mutant (E95/295) or its vector control (D3Δ2). At 30 hours post-transfection, HeLa cells were infected with DENV for 21 hours and subsequently stained for expression of E-proteins. Cells were analyzed by flow cytometry, and the results are expressed as the percentage infectivity in transfected cells. The experiment was performed in triplicate, and the bars represent the average±SD. * P<0.01. To confirm that DENV specifically enters cells via clathrin-mediated endocytosis, we investigated the effects of chlorpromazine, a cationic amphiphilic drug that inhibits the formation of clathrin-coated pits [32], and of a dominant-negative mutant form of Eps15 (E95/295), a protein required for clathrin-dependent uptake [33], on DENV infectivity. Viral infectivity was severely impaired in cells treated with chlorpromazine (Figure 1D) and significantly reduced in cells expressing dominant-negative Eps15 (Figure 1E). Furthermore, no membrane fusion events were seen in real-time virus tracking experiments in chlorpromazine-treated cells (results not shown). Taken together, these results indicate that DENV requires clathrin-mediated endocytosis for its infectious entry. Dengue Virions Are Delivered to Pre-Existing Clathrin-Coated Pits by Diffusion Tracking individual particles also allowed us to determine how DENV particles recruit clathrin-coated pits. A detailed characterization of the individual trajectories showed that virus particles associate with clathrin on average at 111 s post-attachment to the cell surface. Nearly all particles (92%) were observed to move along the cell surface and join a pre-existing clathrin-coated pit. The remaining minor fraction either appeared to land directly on a pre-existing clathrin-coated pit or a clathrin-coated pit was formed directly at the site of the virus particle. An example of the surface motion of DENV towards a clathrin-coated pit is depicted in Figure 2A and Video S1, which is published as supporting information on the web site. 10.1371/journal.ppat.1000244.g002 Figure 2 DENV diffuses along the cell surface and joins a pre-existing clathrin-coated pit. (A) Overlay of a series of time-lapsed images of a virus particle prior to entering a pre-existing clathrin-coated pit. Each depicted virus image (red) is taken 4 seconds apart, and the positions are connected in time with white arrows. For clarity, the LCa-eYFP signal (green) was averaged over 10 seconds, starting at the moment when the virus first overlaps the clathrin-coated pit. Scale bar is 10 µm. (B) MSD-plot for 5 example virus trajectories prior to association with clathrin-coated pits. The diffusion constants are 0.003 µm2/s (squares), 0.009 µm2/s (asterisks), 0.047 µm2/s (triangles), 0.107 µm2/s (diamonds), and 0.159 µm2/s (circles). (C) Diffusion constants of virus particles prior to entrance into a clathrin-coated pit (light gray bars), during colocalization with clathrin (black bars), and on cells treated with chlorpromazine (dashed bars). Subsequently, we investigated whether the surface motion of DENV is characterized by random diffusion or directed movement. To this end, we plotted the mean-square displacement (MSD) of each particle prior to association with a clathrin-coated pit as a function of time. A linear relationship between MSD and time would indicate simple diffusion, an upward curvature designates directed motion, and a downward curvature implies diffusion within a confined region. Figure 2B gives the MSD plot for 5 typical virus trajectories. The apparent linear relationship between the MSD and time indicates that DENV moved on the cell surface in a diffusive manner. During the tracking experiments, we noticed that the mobility of the virus drops when the particle overlaps with a clathrin-coated pit. To obtain a quantitative insight into this behavior, we calculated the diffusion constants from the MSD plots for each particle prior to or during colocalization with clathrin and used that as a measure for surface mobility of the virus. The results show that many DENV particles that were associated with the cell surface were quite mobile, but once they were captured by a clathrin-coated pit their mobility was highly reduced (Figure 2C). Furthermore, treatment of cells with chlorpromazine revealed that DENV particles remained migrating along the cell surface throughout the duration of the experiment in a manner similar to that seen for particles prior to clathrin-mediated entry in untreated cells (Figure 2C). Endocytic Trafficking of Dengue Virus Particles Following clathrin-mediated internalization, virus particles are typically trafficked along an endocytic pathway, which comprises a network of highly dynamic vesicles and endosomes. Endocytic trafficking is regulated by a large family of small Rab GTPases [34]–[38]. Specific Rab GTPases are often enriched in distinct intracellular vesicles and may be used to identify endocytic vesicles and endosomes. For example, Rab5 and Rab7 primarily decorate early and late endosomes, respectively [39]–[41]. Recent live-cell imaging studies have also revealed a small fraction of the endosomes containing both Rab5 and Rab7, which likely indicates intermediate endosomes that are maturing towards late endosomes [34],[35],[37]. To study the itinerary of endosomal compartments visited by DENV, we tracked single DiD-labeled virus particles in BS-C-1 cells co-expressing Rab5 fused to enhanced cyan fluorescent protein (Rab5-eCFP) and Rab7 fused to enhanced yellow fluorescent protein (Rab7-eYFP). We have used this approach before and observed that low level expression of Rab5-eCFP and Rab7-eYFP in cells does not adversely affect endocytic trafficking inside the cell [34]. A typical example of a cell co-transfected with Rab5-eCFP and Rab7-eYFP early after infection with DiD-labeled DENV particles is depicted in Figure 3A. Rab5- and Rab7-positive endosomes can be observed as clear distinct spots that are localized in the cell periphery as well as in the perinuclear region of the cell. We analyzed 51 virus trajectories in total and observed that 86% of the particles first enter Rab5-positive early endosomes; the other 14% of the virions are directly delivered to Rab5/Rab7-positive intermediate endosomes. An example of DENV endocytic trafficking is shown in Figure 3B and Video S2, which is published as supporting information on the web site. At 42 seconds post-binding, this particular virus particle moves with a velocity of 0.42 µm/s towards an intermediate endosome, as shown by the colocalization with Rab5-eCFP and Rab7-eYFP. Subsequently, the intermediate endosome enclosing the virus particle matures into a late endosome as detected by the disappearance of the Rab5 signal. At 342 seconds, the virus particle resides in a Rab7-positive late endosome and induces membrane fusion from within this organelle at 508 seconds post-infection as detected by a five-fold increase in the DiD-intensity. An example of DENV intracellular trafficking via Rab5-positive endosomes is shown in Video S3, which is published as supporting information on the website. 10.1371/journal.ppat.1000244.g003 Figure 3 DENV particles are transported to early endosomes that mature into late endosomes. (A) Fluorescent images of a cell expressing Rab5-eCFP (light blue) and Rab7-eYFP (green) upon infection with DiD-labeled DENV (red). Scale bar is 10 µm. (B) The selected frames show the endocytic trafficking behavior of a single DiD-labeled DENV particle (surrounded by a white circle) in a cell. The numbers designate seconds post-infection. (C) Snapshots of a virus trajectory showing the modes of Rab7 accumulation. At all time frames, the virus particle colocalizes with Rab5, but for clarity the signal is not depicted. Two modes of Rab7 accumulation were observed: a gradual recruitment of Rab7 (upper panel) or merging of the endosome with an existing Rab7-positive endosome (lower panel) was observed. The numbers indicate the time frame (in seconds) of endosome maturation. (D) Snapshots of a virus showing the modes of Rab5 exit. At all time frames, the virus particle colocalizes with Rab7, but for clarity the signal is not depicted. The exit of Rab5 signal also appears to take place in two different modes: a gradual release of Rab5 (upper panel) or splitting off a Rab5-containing endosome (lower panel). We observed different modes of Rab7 accumulation and Rab5 dissociation during endosome maturation. Rab5-positive early endosomes carrying DENV particles were found to mature either through a gradual appearance of Rab7 (55%) or by merging with an existing Rab7-positive endosome (45%). Typical examples of these modes of Rab7 accumulation are depicted in Figure 3C, and Videos S4 and S5. Likewise, the exit of the Rab5 signal also appears to take place in different modes. About 70% of the intermediate endosomes complete the maturation process by a gradual release of Rab5 (Figure 3D, Video S6). In the remaining cases, the virus particle appeared to be sequestered into a Rab7-enriched domain of the intermediate endosome, which subsequently pinched off and moved away as a late endosomal compartment (Figure 3D, Video S7). We have previously identified two distinct populations of Rab5-positive early endosomes [34]. A group of dynamic early endosomes are transported on microtubules and rapidly mature towards late endosomes, while the remaining are relatively static and mature much more slowly. Influenza virus, low density lipoproteins, and epidermal growth factors were previously found to be preferentially targeted to the dynamic, rapidly maturing population, whereas transferrin is non-selectively delivered to both populations. DENV was non-selectively delivered to both endosome populations (data not shown). Dengue Virus Fuses Primarily from within Rab7-Positive Late Endosomes Enveloped viruses escape from the endocytic pathway by a membrane fusion reaction. In our experimental set-up, membrane fusion can be detected as a sudden increase of DiD fluorescence due to the dilution of the DiD-probes from the viral membrane into the endosomal membrane. This assay allowed us to directly examine the nature of the endosomes from which DENV mediates membrane fusion. Individual virus trajectories showed that the majority of the virus particles first joined a Rab5-positive endosome, which then matured through the Rab5/Rab7-copositive intermediate stage into a Rab7-positive late endosome, were membrane fusion was observed. Figure 4A gives a quantitative kinetic analysis of the endocytic trafficking behavior and membrane fusion events of all analyzed DENV particles. Internalization of DENV particles appeared to be relatively quick, since 50% of the particles localized to early endosomes at 3.5 minutes post-attachment to the cell surface. Thereafter, the particles started to associate with Rab7-positive endosomes. The first membrane fusion events were detected at 5 minutes post-infection, and nearly all fusion events occurred within 17 minutes post-infection. The average time point of membrane fusion was 12.5 min, which is in agreement with our previous results [26]. 10.1371/journal.ppat.1000244.g004 Figure 4 DENV fuses primarily from within late endosomes. (A) Individual DENV trajectories were aligned to the time of binding to the cell surface. Fraction of virus particles that colocalize with Rab5 and Rab7, and that undergo membrane fusion are shown as a function of time. (B) Quantification of the percentage of membrane fusion events in the different endocytic compartments. (C) Histogram of the time that virus particles spend in Rab7-positive late endosomes prior to the onset of membrane fusion. (D) The number of DENV membrane fusion events in BS-C-1 cells transfected with Rab7T22N or with a control plasmid. The experiment was carried out more than 10 times and the bars represent the mean±SD. ** P<0.001. The vast majority (80%) of particles induced membrane fusion from within Rab7-positive late endosomes devoid of any detectable Rab5 signal, while 15% of the particles fused from within Rab5/Rab7-copositive intermediate endosomes. Only 5% of the virus particles fused from within Rab5-positive early endosomes that lack Rab7 (Figure 4B). Membrane fusion was initiated at on average 5.5 minutes post-entry of DENV into the Rab7-positive endosomes (Figure 4C). Our observation that DENV (serotype 2, strain S1) primarily fuses from within Rab7-positive late endosomes is somewhat surprising, since a recent report showed that expression of dominant-negative Rab7T22N did not affect DENV (serotype 2, strain New Guinea C) infectivity [22]. To investigate whether this discrepancy is related to the different virus strains used, we analyzed the infectious properties of both viruses on HeLa cells expressing dominant-negative Rab5S34N and Rab7T22N. In agreement with the above results, viral infectivity of S1 was severely impaired in cells expressing dominant-negative Rab7, whereas the infectious properties of NGC were unaffected under the conditions of the experiments (Figure S1). To investigate the requirement for Rab7 of S1 infectivity in more detail, we performed single-particle tracking experiments in cells transiently expressing the dominant-negative Rab7T22N mutant [42],[43]. Figure 4D shows that the number of membrane fusion events was significantly reduced by a factor of 4 in these cells (T-test: P<0.001), which indicates that S1 needs to travel to Rab7-positive endosomes to undergo membrane fusion. Discussion Despite the medical importance of DENV, little information is available about the infectious cell entry pathway of the virus. In this study, we investigated the cell entry process of single DENV particles in real-time by simultaneous tracking of fluorescently labeled DENV particles and endocytic structures in cells. This approach allowed us to obtain mechanistic and kinetic insights into the route of internalization and endocytic trafficking behavior of individual DENV particles in living cells. Previous electron-microscopy studies suggested that DENV penetrates both mammalian and insect cells by direct fusion with the plasma membrane [12],[13]. In contrast, this report shows that DENV enters cells via clathrin-mediated endocytosis and fuses from within late endosomes. We observed that more than 98% of the particles that underwent membrane fusion, first associated with a clathrin-coated structure for a substantial time period. Furthermore, treatment of cells with chlorpromazine as well as expression of a dominant-negative Eps15 mutant significantly suppressed the number of DENV-infected cells. It is not clear what the explanation is for the discrepancy, but it might be related to the methodology that was used to investigate the cell entry process of the virus. The conclusion that DENV utilizes clathrin-mediated endocytosis for internalization is in agreement with recent observations of Acosta and co-workers [14]. During the course of this study, these investigators published that DENV infectivity in C6/36 mosquito cells is severely inhibited after treatment of the cells with a variety of chemical and molecular inhibitors of clathrin-mediated endocytosis. Real-time imaging studies showed that macromolecules either induce de novo formation of clathrin-coated pits or are recruited to pre-existing clathrin-coated pits [31],[44],[45]. For example, influenza virus particles land on the cell surface and induce de novo formation of clathrin-coated pits at the site of binding [30]. This study indicates that DENV particles first diffuse along the cell surface before they encounter pre-existing clathrin-coated pits. After the virus associates with the pit, the clathrin signal around the virus particle increases, which implies maturation of the clathrin-coated pit and formation of a clathrin-coated vesicle. Thereafter, the clathrin signal rapidly disappears again, typically within a time scale of a few seconds. This behavior is similar to that of reoviruses, which have been shown to stabilize and induce maturation of pre-existing clathrin-coated pits [31]. Recently, several modes of endosome maturation have been described. Rink et al. showed that Rab5-positive vesicles, which have split off from a dynamic early endosomal network, accumulate Rab7 and subsequently gradually lose Rab5 [35]. Vonderheit et al. found that Rab5-positive endosomes, containing Semliki Forest Virus (SFV) particles, gradually acquire Rab7 in a separate domain. The SFV particles are sequestered into this Rab7 domain, which pinches off as a Rab7-positive late endosome, leaving a Rab5-positive endosome behind [37]. We observed both modes of endosome maturation. Most DENV particles progressed from early to late endosomes by gradual appearance of Rab7 and a gradual loss of Rab5. In addition, 45% of Rab5-positive endosomes carrying DENV merged with existing Rab7-positive endosomes. Occasionally, we observed that DENV particles sequestered into a distinct Rab7 domain, similar to the behavior observed from endosomes containing SFV [37]. DENV particles predominantly fused from within Rab7-positive endosomes. Furthermore, the membrane fusion activity was significantly impaired in cells expressing dominant-negative forms of Rab7, which indicates that progression of DENV to Rab7-positive endosomes is important for its infectious entry. In contrast, Krishnan et al. have recently demonstrated that the infectivity of DENV-2 strain NGC was not affected by dominant-negative Rab7, while ablation of Rab5 severely reduced the number of infected cells [22]. A direct comparison between both virus strains revealed that viral infectivity of S1 was severely impaired in cells expressing dominant-negative Rab7, whereas the infectivity of NGC was unaffected. These results suggest that both virus strains have distinct entry characteristics. In this respect it is interesting to note that DENV-2 strain NGC induces syncytium formation in a fusion from without assay at pH 6.4, whereas the pH threshold for the DENV-2 S1 strain is around pH 5.8 (personal communication, Dr. P. Young, University of Queensland, Australia). The different pH-dependent properties of these virus strains may therefore reflect the distinct requirements for functional endocytic trafficking in cells. Future experiments should reveal whether the pH threshold determines in which organelle membrane fusion occurs. DENV particles reside on average for 5.5 min in Rab7-positive endosomes prior to the onset of membrane fusion. This result is surprising considering that TBEV efficiently fuses with liposomes in a model system in a time scale of seconds after low-pH exposure [46]. Pre-exposure of TBEV to low pH for 10–20 seconds in the absence of liposomes completely abolishes the membrane fusion activity of the virus [46]. Similar results were obtained for WNV (unpublished results, J. Wilschut and J. M. Smit). Our finding that DENV fuses several minutes after entering a late endosome might therefore suggest that, in addition to exposure to the acidic lumen of the late endosome, other cellular factors are involved in the activation of the membrane fusion machinery of DENV. Another possibility is that the accumulation of Rab7 significantly precedes acidification to the fusion pH. Taken together, we propose the following model for cell entry of DENV S1 strain. First, the virus particle binds to a cellular receptor. Subsequently, DENV diffuses as a virus-receptor complex or rolls over multiple receptors along the cell surface towards a clathrin-coated pit. Upon capture by a pre-existing clathrin-coated pit, the virus particles loses its mobility. Then, the clathrin-coated pit matures and pinches off into the cell cytoplasm to deliver the particles to Rab5-positive early endosomes. In general, the early endosome carrying the virus matures into a late endosome by gradual accumulation of Rab7, followed by a gradual loss of Rab5. Finally, the DENV particles localize to Rab7-positive late endosomes and move through the cytoplasm of the cell until the onset of membrane fusion allows the genetic material of the virus to be delivered into the cytoplasm. Single-particle tracking has substantially enriched our knowledge on viral cell entry mechanisms and has revealed previously unknown aspects of virus-host interactions [30],[47],[48]. The mechanistic and kinetic insights offered by this technique provide a better understanding of disease pathogenesis and may lead to a rational design of antiviral drugs and vaccines. This is the first study that describes the cell entry pathway of DENV at a single-particle level. The parameters obtained in this study will serve as a framework for our current study on the fate of individual antibody-opsonized DENV particles into Fc receptor-bearing to elucidate the molecular basis of antibody-dependent enhancement of DENV infection. Materials and Methods Cells Aedes albopictus C6/36 cells were maintained in Minimal Essential Medium (MEM; Life Technologies, Breda, The Netherlands) supplemented with 10% fetal bovine serum, 25 mM HEPES, 7.5% sodium bicarbonate, 200 mM glutamine, 100 µM non-essential amino acids, penicillin (100 U/ml), and streptomycin (100 µg/ml) at 30°C, 5% CO2. HeLa cells were cultured in a 1∶1 mix of DMEM (Life Technologies) and HAM (Life Technologies) supplemented with 10% fetal bovine serum, 25 mM HEPES, penicillin (100 U/ml), and streptomycin (100 µg/ml) at 37°C, 5% CO2. BS-C-1 cells were maintained in MEM (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum at 37°C, 5% CO2. BS-C-1 cells stably expressing LCa-YFP were created by use of the RetroMax retroviral expression system (Imgenex, San Diego, CA, USA) and cultured in BS-C-1 medium [30]. BS-C-1 cells were grown on glass coverslips (MatTek, Ashland, MA, USA), and prior to the tracking experiments washed with serum-free, phenol red-free medium. DNA Transfection of Cells The plasmid encoding Rab5-eCFP was a gift from Dr. M. Zerial (Max Planck Institute, Dresden, Germany). The Rab7-eYFP plasmid was previously constructed by Dr. M. Lakadamyali [34]. The GFP-tagged dominant-negative Eps15 mutant E95/295 and its empty vector D3Δ2 were kindly provided by Dr. A. Benmerah and Dr. A. Dautry-Varsat (Institute Pasteur, Paris, France). The plasmids encoding GFP-tagged dominant-negative Rab5 mutant Rab5S34N, wild-type Rab5-GFP, myc-tagged dominant-negative Rab7 mutant Rab7T22N, and wild-type Rab7-myc were gifts from Dr. P. van der Sluijs (University Medical Center, Utrecht, The Netherlands). Cells were transfected with the plasmids using the transfection reagent FuGENE, according to the manufacturer's protocol (Roche, Nutley, NJ, USA). Viral Infectivity To analyze the route of DENV cell entry, viral infectivity was measured in HeLa cells expressing dominant-negative Eps15 mutants. At 30 hours post-transfection, cells were infected at MOI 5. At 21 hours post-infection, cells were washed with PBS, trypsinized, fixed with 4% paraformaldehyde, and permeabilized with 0.5% saponin in PBS containing 2% FBS. Expression of the myc-tagged plasmids was detected with monoclonal antibody A-14 (Santa Cruz Biotechnology, Santa Cruz, CA, USA). DENV infection was measured using the monoclonal antibody MAB8702 directed against the viral E protein (Chemicon, Hampshire, United Kingdom). Cells were analyzed on a FACS Calibur cytometer. The effect of chlorpromazine on DENV infectivity was determined by an infectious center assay in BS-C-1 cells as described before [25]. Chlorpromazine (15 µM) was added to the cells 30 min prior to addition of the virus. At 1 hour post-infection, cells were washed and fresh medium containing 20 mM ammonium chloride was added. At 24 hours post-infection, cells were fixed and stained intracellularly with MAB8702 to measure infection [26]. Preparation of DiD-Labeled Virus DENV serotype 2 strain PR159 S1, generously provided by Dr. Richard Kuhn (Purdue University, Lafayette, IN, USA), was produced and labeled with DiD as described previously [26]. Briefly, monolayers of C6/36 cells were inoculated with DENV at MOI 0.1. At 72 hours post-infection, the progeny virions were harvested, purified by ultracentrifugation, and cleared from tartrate using 100 kD filter devices (Millipore, Amsterdam, The Netherlands). Subsequently, 2 nmol DiD (Molecular Probes, Eugene, OR, USA) dissolved in dimethyl sulfoxide (DMSO) was mixed with approximately 5×109 genome-containing DENV particles while vortexing in a total DMSO concentration of less than 2.5%. After 10 min, the unincorporated dye was removed by gel filtration. DiD-labeled virus was stored at 4°C and used within 2 days. Virus preparations were analyzed with respect to the infectious titer and the number of physical particles, as described previously [26]. Live-Cell Imaging Tracking experiments were carried out 24 to 48 hours post-transfection as described previously [34]. Briefly, fluorescent images were recorded by exciting CFP with a 454 nm Argon laser (Melles-Griot, Carlsbad, CA, USA), YFP with a 532 nm Nd∶YAG laser (Crystalaser, Reno, NV, USA), and DiD with a 633 nm helium-neon laser (Melles-Griot). For the clathrin experiments, simultaneous images were recorded of DiD-labeled virions and LCa-eYFP at 2 frames per second. In case of the Rab5/Rab7 experiments, the excitation of DiD was continuous, whereas the excitation of CFP and YFP were alternated at 0.5 Hz. The fluorescent emission was spectrally separated by 650 nm long-pass dichroic mirrors (Chroma, Rockingham, VT, USA) and imaged onto two separate areas of charge-coupled device camera (CoolSNAP HQ, Roper Scientific). A 665 nm long-pass filter was used for the emission of DiD. For the emission of CFP and YFP, bandpass filters of 480/40 nm and 585/35 nm were used, respectively. The CFP and YFP filters were toggled by a motorized wheel at 0.5 Hz synchronically with the 454 nm and 532 nm lasers. Image analysis and single-particle tracking was performed using custom-written IDL software as described previously [26],[49]. Briefly, background and noise were reduced by convolution with a Gaussian spatial filter. Viral trajectories were generated by pairing virus spots in each frame according to proximity and similarity in intensity. Colocalization of viruses with fluorescent cellular markers was identified with an automated program and confirmed by eye, the criteria for colocalization being that the objects move together and have at least partial overlap. Only those particles that moved roughly within the focal plane and showed more than a fivefold increase in fluorescence intensity after membrane fusion were used for image analysis. Characterization of the movement of DENV particles on the cell surface was done by generating MSD-plots. The MSD at time interval τ is the average of all squared displacements throughout the virus trajectory prior to or during association with clathrin. The diffusion constants were calculated from the slope of the MSD-plot. Supporting Information Figure S1 Effect of functional repression of Rab5 or Rab7 on the infectivity of DENV S1 strain and NGC strain. DENV S1 (light gray bars) and NGC infectivity (black bars) in HeLa cells expressing either wild-type Rab5 and Rab7 or dominant-negative Rab5 (Rab5S34N) and Rab7 (Rab7T22N). At 30 hours post-transfection, HeLa cells were infected with DENV for 21 hours and subsequently stained for E protein-expression. Cells were analyzed by flow cytometry, and the results are expressed as the percentage infectivity in transfected cells. The experiment was performed in triplicate, and the bars represent the average±SD. (0.39 MB TIF) Click here for additional data file. Video S1 Surface motion of a DENV particle before capture by a clathrin-coated pit. DENV particle (red, surrounded by a white circle) that moves along the cell surface until it associates with clathrin (green). Upon binding, the clathrin intensity increases around the virus particle, indicating the maturation of the clathrin-coated pit. Subsequently, the clathrin coat disappears and the particle exhibits a rapid movement towards the perinuclear region of the cell. Snapshots of this video are shown in Figure 2A. The playback speed is 15× real-time. (8.35 MB AVI) Click here for additional data file. Video S2 Endocytic trafficking of DENV. Upon cell entry, the DENV particle (red, surrounded by a white circle) moves rapidly towards an intermediate endosome enriched in Rab5 (blue) and Rab7 (green). This moment is clarified by the appearance of “(A)”, the arrow points at the intermediate endosome. Subsequently, the intermediate endosome carrying the virus particle moves around and merges with other endosomes (indicated by the arrows): a Rab5-positive endosome (B) and an intermediate endosome (C). Next, the intermediate endosome carrying the virus matures into a late endosome through a gradual disappearance of Rab5 (D). Membrane fusion occurs from within the Rab7-positive endosome (E). Snapshots of this video are shown in Figure 3B. During real-time imaging, DiD-labeled DENV particles were recorded at 2 Hz, whereas the excitation of Rab5-eCFP and Rab7-eYFP were alternated at 0.5 Hz. Due to this recording scheme, the CFP and YFP frames are updated every 4 frames in the video, whereas the DiD signal is updated every frame, which might give the impression that the endosomal markers are lagging behind the virus particle. The hallmarks of colocalization are that the DiD spot and the Rab signal at least show partial overlap when they are simultaneously excited, and that both signals are linked and move together in time. The playback speed is 10× real-time. (9.66 MB AVI) Click here for additional data file. Video S3 Endocytic trafficking of DENV. DENV particle (red, surrounded by a white circle) that rapidly moves to a Rab5-positive early endosome (blue) on time point (A). The early endosome carrying the DENV particle acquires Rab7 (green) by merging with an intermediate Rab5- and Rab7-positive endosome (B). Subsequently, the endosome gradually accumulates more Rab7 (C) and gradually loses Rab5 (D). After merging of another large late endosome (E), the virus finally fuses from within the Rab7-positive late endosome (F). Real-time tracking was performed as described in the legend of Video S2. The playback speed is 10× real-time. (10.20 MB AVI) Click here for additional data file. Video S4 Some early endosomes gradually accumulate Rab7 molecules. Endosome enclosing a DENV particle (red, surrounded by a white circle) that shows a gradual increase in the Rab7 signal (green). Although during real-time imaging DiD-labeled particles were excited at 2 Hz and Rab7-eYFP at 0.5 Hz, this video is composed of only those frames in which DiD and YFP were excited simultaneously. During the whole video, the virus particle colocalizes with Rab5, but for clarity the signal is not depicted. Snapshots of the maturation event are shown in Figure 3C. The playback speed is 10× real-time. (8.00 MB AVI) Click here for additional data file. Video S5 Some early endosomes accumulate Rab7 by joining a pre-existing Rab7-positive endosome. Endosome enclosing a DENV particle (red, surrounded by a white circle) that merges with a Rab7-positive endosome (green). The arrow indicates the moment of the merging. Real-time imaging and the preparation of the video are described in the legends of Videos S2 and S4, respectively. During the whole video, the virus particle colocalizes with Rab5, but for clarity the signal is not shown. Snapshots of the maturation event are shown in Figure 3C. The playback speed is 10× real-time. (4.43 MB AVI) Click here for additional data file. Video S6 Some endosomes complete the maturation process by a gradual release of Rab5 molecules. Endosome enclosing a DENV particle (red, surrounded by a white circle) that shows a gradual release of the Rab5 signal (blue). Real-time imaging was performed as described in the legend to Video S2, but only those frames in which DiD and CFP were excited simultaneously were used in the video. Furthermore, the virus particle colocalizes with Rab7 throughout the time trace, but for clarity the signal is not depicted. Snapshots of the maturation event are shown in Figure 3D. The playback speed is 10× real-time. (5.97 MB AVI) Click here for additional data file. Video S7 Some endosomes complete the maturation process by splitting off a Rab5-positive endosome. Endosome enclosing a DENV particle (red, surrounded by a white circle) that splits off and leaves the Rab5-positive endosome (blue). The moment of splitting is indicated by the arrow, which points at the Rab5 endosome that is left behind. Real-time imaging and the preparation of the video are described in the legends to Videos S2 and S6, respectively. During the time trace, the virus particle colocalizes with Rab7, but for clarity the signal is not depicted. Snapshots of the maturation event are shown in Figure 3D. The playback speed is 10× real-time. (2.93 MB AVI) Click here for additional data file.
- Record: found
- Abstract: found
- Article: found
Identification of Host-Targeted Small Molecules That Restrict Intracellular Mycobacterium tuberculosis Growth
Sarah Stanley, Amy Barczak, Melanie Silvis … (2014)
Introduction Tuberculosis continues to be a cause of significant morbidity and mortality world-wide due to numerous factors, including the rise of drug resistance and the absence of an effective vaccine. The challenge of adherence to long treatment regimens and the limited number of effective therapeutics drive the need for innovative therapeutic strategies that are predicated on a better understanding of the biology of infection. For example, the fact that the majority of people infected with the bacillus never develop active disease suggests that human immunity is actually quite effective at controlling M. tuberculosis; a novel, potentially more effective therapeutic strategy could emerge if we were able to understand and leverage the basis for this control. The cellular interaction between M. tuberculosis and macrophages is crucial for determining the outcome of infection. Early in infection, macrophage microbicidal mechanisms actively work to try to clear the bacteria; however, macrophage responses that are adequate to kill other bacterial pathogens often fail to clear M. tuberculosis. In a majority of individuals, the activation of macrophages by IFN-γ can result in control but not sterilization of infection, instead driving it into a latent state. During latency, the macrophage can be a niche where the bacteria are protected from assaults by the immune system and antibiotic therapy, thus facilitating their persistence and ultimate dissemination. Recent studies have uncovered a number of processes that are important to tubercular infection. The ability of M. tuberculosis to arrest the normal progress of phagosome maturation is critical for its survival in macrophages [1]; however, the molecular mechanisms on both the pathogen and host sides that account for this arrest are unclear. For example, while calcium signaling in macrophages appears to be important in this process, the nature of the calcium signal and the mechanisms by which M. tuberculosis actively affects calcium signaling are debated [2], [3]. In addition to phagosome maturation arrest, M. tuberculosis may actively suppress many other macrophage innate immune responses. For example, virulent strains of M. tuberculosis actively prevent apoptosis of infected macrophages, thus preventing bacterial killing by macrophage efferocytosis and avoiding activation of T-cells through cross-presentation of antigens by dendritic cells [4], [5], [6]. M. tuberculosis may also actively prevent activation of the inflammasome and induction of autophagy [7], [8]. In addition to subversion of immune responses, M. tuberculosis manipulates the host microenvironment in order to acquire nutrients to promote its own survival. For example, virulent mycobacteria are able to induce the development of intracellular lipid bodies which fuse with M. tuberculosis containing phagosomes and provide a critical source of carbon [9]. Although we have some insight into the pathways that are important for M. tuberculosis infection of macrophages, our current understanding of the mechanisms that determine whether the macrophage controls bacterial infection or succumbs to its virulence is incomplete. In order to obtain greater insight into host factors involved in M. tuberculosis infection, unbiased screening using RNAi or small molecules targeting host proteins have recently been performed. Two published RNAi screens, one genome-wide and one focused on kinases and phosphatases, identified mammalian proteins that are candidate regulators of M. tuberculosis infection [10], [11]. To provide a functional context for the identified regulators, the authors constructed a signaling network by integrating the RNAi screening data with data from transcriptional profiling. Over half of identified genes were found to be negative regulators of autophagy, affirming the importance of this pathway for host defense against M. tuberculosis [10]. In addition to regulators of autophagy, the networks implicated were enriched for modules that govern metabolism and signal transduction, with many of these modules centered around the serine/threonine kinase AKT. Kinases are central to mammalian signaling pathways. AKT/PKB is a key modulator of cellular processes such as growth and proliferation, glucose metabolism, apoptosis, and autophagy. AKT is specifically activated during Salmonella infection of host cells by the bacterial effector SopB and promotes bacterial survival by prevention of phago-lysosome fusion [12]. Treatment of M. tuberculosis macrophages with the AKT and PKA inhibitor H-89 also results in inhibition of bacterial growth. However, in contrast to Salmonella infection, the role of AKT is unknown in M. tuberculosis infection [12]. Importantly, although AKT was identified in the network that emerged from the genome-wide RNAi screen of M. tuberculosis infected THP-1 macrophages, the kinase itself was neither identified in the primary genome-wide screen nor in a more directed kinase/phosphatase screen conducted by the same group [11]. Of note, in the RNAi screens that have been reported, the siRNAs used to decrease host factor expression were added only after M. tuberculosis had already entered and adapted to the macrophage microenvironment; thus, these screens were not designed to identify factors that are crucial for the earliest events in the host-pathogen interaction. Effective silencing of gene expression using transfection of siRNA is in part dependent on the half-life of the targeted protein and occurs on the timescale of hours to days after transfection. In contrast, a chemical biological approach has some advantages over RNAi with regard to studying early events. The rapid binding of small molecules to proteins facilitates probing the early period immediately after phagocytosis. Because this period is accompanied by the most significant transcriptional responses on the part of both the bacterium and the macrophage [13], events during this time frame are likely important determinants of the outcome of infection. Further, in general, unlike genetic approaches which can target only one isoform or homologue at a time, small molecules can inhibit multiple, closely related isoforms of the same target, thus facilitating the identification of host activities performed by two or three closely related targets with redundant function. In addition, unlike RNAi, small molecules can inhibit enzymatic function without disrupting larger complexes with subsequent pleiotropic effects, or can inhibit a specific function of a protein while leaving other functions intact [14]. Thus small molecule based screens can provide a valuable complement to existing datasets obtained using siRNA based knockdown. Three recently conducted screens were designed to identify small molecules that disrupt M. tuberculosis replication in macrophages. The first screen used a high-content imaging approach to identify compounds that directly target bacterial processes during macrophage growth [15]. A subsequent study by the same group and using the same screening approach identified an inhibitor of M. tuberculosis cytochrome bc1 [16]. More recently, a group reported a screen in which microscopy was used to identify host-targeted inhibitors that prevent replication of Mycobacterium bovis BCG in macrophages. This study characterized existing neurotropic drugs that diminish replication of M. tuberculosis in infected macrophages by inducing autophagy or altering endosomal trafficking, however the targets of these drugs and their modes of action have yet to be elucidated [17]. We performed a complementary screen of a library of known bioactive small molecules to identify inhibitors of M. tuberculosis replication in macrophages that are biased towards disrupting host functions. Our goal was to identify small molecules that specifically target host proteins to gain insight into host functions required for controlling M. tuberculosis infection. To focus on early events crucial for the adaptation of the macrophage for bacterial survival but not on uptake itself, we added molecules to infected macrophages immediately after phagocytosis. From this screen, we identified several classes of host-targeted compounds that limit the ability of M. tuberculosis to proliferate in macrophages, including kinase inhibitors, G-protein coupled receptor modulators, and ion channel inhibitors. Our results are complementary to previously reported screens, expanding our knowledge of host pathways that are crucial determinants of infection. Based on the bioactive molecules identified in our screen, we determined that the SSRI fluoxetine inhibits M. tuberculosis growth in macrophages, induces enhanced expression of TNF-α, and enhances autophagy. We additionally demonstrated a new role for the tyrosine kinase EGFR during M. tuberculosis infection. Importantly, we showed that the clinically used EGFR inhibitor gefitinib has efficacy for preventing bacterial replication in both infected macrophages and mice, suggesting that EGFR is relevant in vivo. The inhibitors we have identified will be important tools for studying the host-pathogen interaction and for testing the importance of host pathways during in vivo infection in animal models. Results Development of a high-throughput, high-content imaging assay to identify compounds that inhibit M. tuberculosis growth in macrophages For identification of small molecules that inhibit M. tuberculosis replication in macrophages, after significant assay testing (see Methods S1) we designed a microscopy-based assay with simultaneous imaging of macrophages and mycobacteria that provides the sensitivity to detect modest growth inhibition, easily identifies compounds with significant macrophage toxicity, and can be largely automated to allow for the rapid testing of thousands of compounds (for assay design and development details, see Methods S1). We used this high-content imaging assay to monitor the growth of GFP-expressing M. tuberculosis in infected macrophages. In designing the assay, we sought to establish growth conditions that would allow for robust and reproducible replication of M. tuberculosis in macrophages and a readout that would accurately reflect bacterial number. Macrophages were infected with M. tuberculosis constitutively expressing GFP (H37Rv-GFP [18]), and at various time-points after infection, the cells were fixed and stained with DAPI to allow for enumeration of macrophages by microscopy. Because intracellular mycobacteria grow in clumps rather than as discrete, easily quantifiable spots, we utilized the open-source image analysis software CellProfiler, which can simultaneously determine and integrate multiple parameters from our images [19] to determine an optimal visual parameter that correlates with bacterial number. We found that GFP pixel intensity integrated across the field and normalized to macrophage number best represented growth while also accounting for differences in surviving macrophage number, and most accurately reflected bacterial census enumerated by plating for colony forming units (CFU). Using this parameter, we tested growth of H37Rv-GFP in a number of cell lines including THP-1 cells, RAW 264.7, and J774A.1 cells, and found that J774 murine macrophages allowed for the most consistent and robust growth, with a bacterial doubling time equivalent to that measured in axenic culture. The optimal time-point for measuring growth was 3 days post infection, when intracellular mycobacterial growth was most homogeneous and little apparent macrophage death was observed. We next assessed the heterogeneity of mycobacterial growth and ability of the assay to reproducibly detect growth inhibition across large numbers of wells in 96-well format. Using a gradient of concentrations of rifampicin to model varying degrees of growth inhibition, we demonstrated that the assay was able to distinguish growth inhibition of 50% inhibition or less (Figure 1A, Figure S1). With the highest dose of rifampicin we consistently obtained Z′-factors of 0.4–0.5, which is borderline in robustness for a high-throughput screen. However, because bacterial growth based on imaging across the wells did not follow a Gaussian distribution, standard statistical methods to assess high-throughput assay quality, such as Z′-factor, that assume a normal distribution for high-throughput assays may not accurately assess the robustness of the assay. We therefore used a bootstrap Monte Carlo analysis as another means to assess the quality of the assay [20], [21] and found that we were able to consistently distinguish 50% inhibition of growth as established with our positive controls with a predicted false negative rate of 0.2% and a false positive rate of 0.13%. We found these rates to be acceptable, allowing us to progress to high-throughput screening. Ultimately we compared the list of hits obtained using either composite z-scores or p-values obtained using the Monte Carlo bootstrap analysis (both metrics included in Table S1); we found that the list of hits was essentially the same using both metrics. 10.1371/journal.ppat.1003946.g001 Figure 1 High-content imaging assay development and categorization of screen hits. (A) J774 cells in 96-well dishes were infected with GFP-expressing H37Rv at an MOI of 1∶1. Cells were then treated with rifampin at the indicated concentrations. Day 3 after infection, cells were fixed, stained with DAPI, and imaged for both DAPI and GFP. Images were analyzed using CellProfiler to determine the mean integrated GFP intensity across each well normalized to macrophage number. The graph represents one of four independent experiments with data represented as an average +/− SEM for each condition for 84 wells. (B) Of 133 unique hits in the screen, 41 were agonists or antagonists at G-protein coupled receptors, 16 were inhibitors or activators of ion channels, 15 were kinase activators or inhibitors, 5 were membrane transport modulators, 7 were anti-inflammatory agents, and 46 did not fall into one of the above major categories. Screening a library of bioactive small molecules and screen hits We used the assay to screen a library of 1920 small molecules from the Broad Institute bioactives collection and an additional 159 kinase inhibitors obtained from Nathanael Gray at the Dana Farber Cancer Institute (Table S1, Figure S2). One advantage of screening a library of known bioactives is that compound annotation can provide initial suggestions about mechanisms of action, thus facilitating the considerable challenge of target identification from potential host and bacterial targets. Additionally, the bioactives library is enriched for molecules that target mammalian proteins, thus allowing for elucidation of novel host/pathogen biology by focusing on macrophage functions. In a 96 well format, ∼3000 macrophages were plated in each well and infected with M. tuberculosis strain H37Rv-GFP at a multiplicity of infection (MOI) of 1∶1 for 4 hours, after which time extracellular bacteria were removed by washing. To facilitate identification of compounds that restrict intracellular M. tuberculosis growth rather than inhibit uptake, the compound library was added after phagocytosis. The average final concentration of small molecules was 5 µM. After three days, cells were washed, fixed, stained with DAPI, and imaged with a 4× objective lens on an ImageXpress Micro high-throughput microscope (Molecular Devices). We analyzed our data using both a bootstrap Monte Carlo analysis and composite z-score and were reassured that both methods identified the same small molecule hits. (The z-score cutoff was 2 fold decrease in bacterial load at the highest tested concentration of each inhibitor in both J774 macrophages (Figure 4A) and mouse bone marrow macrophages (Figure 4B). 10.1371/journal.ppat.1003946.g004 Figure 4 Inhibitors of protein kinases impair mycobacterial growth in macrophages. (A) J774 cells or (B) BMDM were infected with M. tuberculosis strain H37Rv at an MOI of 1∶1, and treated with each kinase inhibitor at the indicated concentrations (µM) after a 4 h phagocytosis period. At day 3 (J774) or day 5 (BMDM) after infection, cells were washed, lysed, and plated for CFU. Each column represents the mean and standard deviation of four biological replicates and each graph represents one of three independent experiments. For (A) all p<0.001 with the exception of GNF-2 at 2.5 µM, imatinib at 5 µM and gefitinib at 5 µM. For (B) inhibitors were used at the following concentrations: AKTi1/2 5 µM (p<0.03), H-89 5 µM (p<0.03), GNF-2 10 uM (not significant), imatinib 5 µM (p<0.03), gefitinib 5 µM (p<0.03), lapatinib 5 µM (p = <0.06). All p values were calculated using the Mann Whitney U test. AKT/PKB and M. tuberculosis infection of macrophages The serine/threonine kinase AKT/PKB has been previously implicated in the ability of M. tuberculosis and other bacterial pathogens to replicate in host cells [10], [12]. Treatment of infected cells with the AKT1 inhibitor H-89 was shown to limit proliferation of M. tuberculosis in infected mouse macrophages [12] and knocking down AKT1 and AKT2 in human THP-1 cells also led to decreased replication of intracellular M. tuberculosis [10]. In our screen, in addition to previously identified ATP competitive inhibitor H-89, which preferentially targets PKA in addition to AKT, we found that a more specific allosteric inhibitor of AKT, AKTi1/2 [30], [31] also restricts intracellular M. tuberculosis replication. To confirm that AKT plays a role in M. tuberculosis infection of macrophages, we first tested to see whether AKT was activated by infection with virulent M. tuberculosis. AKT activation requires phosphorylation at two residues, Thr-308 and Ser-473 [32]; Western blot analysis for this phosphorylation is a standard assay for activation. Using Western blot analysis for phosphorylation of Ser-473, we observed rapid activation of AKT upon infection of J774 macrophages with M. tuberculosis (Figure 5A). Importantly, the activation of AKT was completely abrogated by treatment with the allosteric inhibitor AKTi1/2 that is known to block the phosphorylation and activation of AKT1 and AKT2 (Figure 5B). To genetically confirm the role of various AKT isoforms in macrophage infection, we knocked down the expression of AKT1, AKT2, and AKT3 in J774 macrophages (Figure S10). Similar to previously reported results [10], we found that the maximum effect was observed with simultaneous knockdown of AKT1 and AKT2. We also found that knocking down AKT3 provided little additional benefit (Figure 5C). That silencing both AKT1 and AKT2 is required for maximal effect may account for the failure to identify AKT in the original RNAi screens that target only one isoform at a time, while a small molecule inhibitor inhibits both isoforms simultaneously. 10.1371/journal.ppat.1003946.g005 Figure 5 AKT and ABL are important for mycobacterial infection of macrophages. (A) J774 cells were infected with H37Rv at an MOI of 1 or an MOI of 10. Cells were harvested two hours after infection, and lysates were probed for AKT activation using an antibody to phospho-serine at position 473. (B) J774 cells were infected with H37Rv at an MOI of 1 for 4 hours, then washed and treated with compound. Cells were harvested at 3 hours after treatment and lysates were probed for AKT activation using an antibody to phospho-serine at position 473. (C) akt1 and akt2 or akt1, akt2, and akt3 were silenced with siRNA in J774 cells. Cells were then infected with H37Rv at an MOI and 1, and infection was allowed to progress for 3 days. Cells were then lysed and plated for CFU. (D) abl1 was silenced in J774 macrophages using siRNA. Cells were then infected with H37Rv. After 4 hours of phagocytosis, wells were lysed and plated for CFU to determine uptake (day 0). Infection was allowed to progress in the remaining wells; day 3 after infection, cells were lysed and plated for CFU. Each experiment was repeated a minimum of three times, and a representative experiment is shown. Error bars are standard deviation, *p = 0.0195 by Mann Whitney U for (C) and (D). The role of ABL tyrosine kinase during infection with M. tuberculosis Imatinib mesylate (Gleevec, STI-571) is a tyrosine kinase inhibitor that is currently used therapeutically for treating chronic myelogenous leukemia (CML). Imatinib inhibits the ABL family tyrosine kinases ABL1 and ABL2 and other related tyrosine kinases [33]. In fact, imatinib's activity against CML may require activity at more than one target [34]. Recently, imatinib was shown to diminish entry of Mycobacterium marinum in mouse fibroblasts and M. tuberculosis in mouse macrophages [28]. In addition, imatinib treatment decreased bacterial replication in M. marinum and M. tuberculosis infections of mice [28]. Identification of the ABL inhibitors GNF-2 and imatinib in our screen, which was designed to avoid inhibition of initial phagocytosis, supports the idea that ABL is important not simply for cell entry but also for later replication and survival of mycobacteria in macrophages. A subsequent study has shown that imatinib increases acidification of the lysosomal compartment of macrophages. This effect was required for the anti-mycobacterial activity of the compound [29]. As imatinib inhibits both ABL1 and ABL2, in addition to other tyrosine kinases, the relevant target during M. tuberculosis infection of macrophages is not clear. Further, in apparent contradiction, silencing of abl1 using siRNA has been reported to increase M. tuberculosis replication in macrophages [11] suggesting that this family of kinases may play complicated roles during infection. We therefore sought to characterize the role of ABL1 in entry and survival of M. tuberculosis in macrophages. To specifically abrogate ABL1 function, we transfected macrophages with siRNA targeting abl1. Although we observed significant silencing of ABL1 expression (Figure S10), we observed no difference in M. tuberculosis entry into cells treated with abl1-specific siRNA, demonstrating that ABL1 may not in fact play a role in M. tuberculosis entry (Figure 5D, Figure S10B). However, after three days of infection we observed a ∼1.9 fold decrease in bacterial replication in cells treated with siRNA targeting abl1 compared to cells treated with a non-specific siRNA control, indicating that ABL1 likely plays a role in M. tuberculosis intracellular survival post entry into macrophages (Figure 5D). Importantly, knocking down expression of only ABL1 gave approximately the same magnitude effect observed in imatinib treated cells (∼2.1 fold), suggesting that the effects of imatinib are likely mediated primarily through ABL1 during M. tuberculosis infection of macrophages. The role of epidermal growth factor receptor during infection with M. tuberculosis Although primarily studied in the context of cancer, EGFR has been linked to influenza uptake [35], to regulation of inflammation following rhinovirus infection [36], and to prevention of apoptosis in host cells in bacterially infected gastric epithelial cells [37]. EGFR has not previously been linked to mycobacterial infection. We identified gefitinib (Iressa, ZD-1839), an EGFR inhibitor used for treating non-small cell lung cancer [38], and two other EGFR inhibitors in our primary screen (Table S1). Subsequent testing in J774 macrophages and BMDM confirmed that gefitinib treatment controls M. tuberculosis intracellular growth (Figures 4A–4B); we confirmed that EGFR is in fact expressed in macrophages using reverse transcription and PCR-based detection of transcript (Figure S11). To confirm that the small molecules function by specifically inhibiting EGFR signaling, we sought to perturb EGFR signaling in an alternative manner. As EGFR is a well-validated target for anticancer therapy, several monoclonal antibodies for blocking human EGFR have been developed. In particular, treatment of cells with combinations of non-competitive antibodies against EGFR results in synergistic receptor downregulation via recycling inhibition in a manner that does not result in activation of EGFR signaling [39]. We first confirmed that gefitinib effectively blocks M. tuberculosis replication in primary human monocyte derived macrophages using a total of four donors (data not shown). Next, we treated infected human macrophages with two EGFR-neutralizing antibodies [39]. Treatment with neutralizing antibodies inhibited intracellular M. tuberculosis growth to the same degree as treatment with gefitinib. (Figure 6A), confirming the role of EGFR in M. tuberculosis intracellular survival and replication in primary human macrophages. 10.1371/journal.ppat.1003946.g006 Figure 6 EGFR is important for mycobacterial infection of macrophages and in a mouse model of infection. (A) human primary PBMC derived macrophages were infected with a luciferase-expressing Erdman strain, then treated with EGFR-neutralizing antibodies (αEGFR) and/or gefitinib at 10 µM (gef) 4 h after the phagocytosis period. Antibodies and inhibitor was refreshed daily, and bacterial growth was determined by luminescence on day 5 after infection. The data is representative of two independent experiments. (B) J774 cells were infected with M. tuberculosis strain H37Rv. 4 hours after infection, cells were washed and gefitinib or DMSO was added. Cells were lysed at 15 minutes, 30 minutes, or 60 minutes after drug treatment and lysates were probed for phospho-p38 and total p38 by Western blot. The data is representative of three independent experiments. (C) J774 cells were infected with H37Rv-GFP. 4 hours after infection, cells were washed and DMSO or AMG548 was added at the indicated concentrations. Day 3 after infection, cells were washed and fixed with 4% paraformaldehyde. Plates were imaged and analyzed using our CellProfiler pipeline to determine integrated GFP intensity normalized to macrophage number. The data represent one of two independent experiments. (D) Mice were infected with M. tuberculosis strain Erdman via the aerosol route with ∼200 CFU and infection was allowed to progress for 7 days. Beginning day 8 after infection, mice were treated with DMSO or gefitinib 100 mg/kg by intraperitoneal injection daily for six days. Day 14 after infection, mice were sacrificed and lungs were plated for CFU. Data combined from three independent experiments is shown. ***p<0.001, *p = 0.034 by non-parametric Mann-Whitney U test. Error bars are standard deviation. Gefitinib inhibits phosphorylation of the downstream signaling molecule p38 after infection There are several signaling pathways that can be triggered downstream of EGFR activation, including activation of AKT, ERK, JNK, and MAPK p38. To determine whether the effect of EGFR inhibition was mediated by any of these signaling pathways, we first sought to determine whether they were activated by M. tuberculosis infection of macrophages. Both AKT and p38 were phosphorylated upon infection (Figures 5A and 6B), consistent with activation. ERK and JNK did not appear to have increased phosphorylation upon infection (data not shown). Surprisingly, AKT phosphorylation was not inhibited by gefitinib treatment (Figure 5B), suggesting that the effect of inhibiting EGFR during M. tuberculosis infection is independent of AKT signaling. To determine whether p38 signaling is impacted by gefitinib, we measured phosphorylation of p38 and found that it was consistently inhibited in the presence of gefitinib treatment (Figure 6B). While p38 phosphorylation decreased over time after infection even in the absence of compound, gefitinib-treated cells had significantly less p38 phosphorylation than DMSO-treated cells at multiple timepoints. To determine whether inhibition of p38 phosphorylation could be the mechanism by which EGFR inhibition of gefitinib treatment restricts M. tuberculosis growth in macrophages, we tested whether selective p38α inhibitor AMG548 [40] would also restrict intracellular mycobacterial growth. We found that inhibition of p38α by AMG548 does in fact restrict mycobacterial growth in a dose-dependent fashion (Figure 6C). MAPK p38 has been demonstrated to regulate autophagy via p38IP and mATG9, with signaling through p38 associated with inhibition of autophagy and depletion of p38 with activation of autophagy [41]. Consistent with our finding that gefitinib induces autophagy (Figure 3A), this report that p38 inhibition can induce autophagy [41], and our implication of p38 downstream of EGFR activation after M. tuberculosis macrophage infection, we suggest that inhibition of p38 activity downstream of gefitinib-mediated EGFR inhibition may trigger an increase in autophagy within the cell, resulting in enhanced clearance of M. tuberculosis. Gefitinib is effective for treating M. tuberculosis in vivo As there are few studies implicating EGFR in bacterial pathogenesis, we sought to determine whether EGFR signaling is relevant to M. tuberculosis replication in vivo. Given the problem of drug-resistant tuberculosis and the long courses of therapy currently required for treatment, there is growing interest in host-directed therapies as adjunctive therapy, with the potential to shorten therapy and increase the efficacy of current antitubercular antibiotics [11], [17], [28], [42]. To determine whether gefitinib has efficacy against M. tuberculosis in vivo, we examined the efficacy of this inhibitor during early acute infection with M. tuberculosis in the mouse model. We chose to test during early infection prior to the recruitment of IFN-γ producing T cells to the lungs, as this is the time period that is best modeled by in vitro infection of unactivated macrophages. BALB/c mice were infected with M. tuberculosis via aerosol inoculation. Infection was allowed to progress for 7 days prior to initiation of treatment. Mice were subsequently treated with gefitinib at a dose of 100 mg/kg every day for six days. On day 14 after infection, 7 days after initiation of gefitinib treatment, mice were sacrificed, and lungs were homogenized and plated for CFU. The experiment was performed three times and the combined results are shown. Relative to DMSO-treated control mice, mice treated with gefitinib had statistically significantly less bacterial burden (Figure 6D). These results support the in vivo relevance of EGFR in infection and the idea that host-targeted therapies, and kinase inhibitors specifically, may be useful as adjunctive treatment for tuberculosis. Further work will clearly be required to demonstrate efficacy in additional, more chronic models of M. tuberculosis as well as to determine any potential benefit when used in combination with traditional anti-tuberculosis chemotherapeutic agents. Discussion We have developed a high-content, high-throughput imaging assay to identify small molecules that inhibit the growth of M. tuberculosis within macrophages. We applied our assay to screen a library of known bioactive molecules, and from that screen identified several compound classes with diverse annotated targets that result in reproducible, dose-dependent intracellular M. tuberculosis growth restriction. Out of the compounds of interest in our five primary categories, several hits validated previously described targets [12], [28], while others pointed to the involvement of potentially novel factors important for infection, including EGFR, the serotonin transporter targeted by SSRIs, serotonin and dopamine receptors, and sodium channels. An important caveat is that an annotated mechanism of action may not be relevant to the actual activity of a compound in a given assay; observed activity may be due to some other “off-target” effect. Thus, we have used complementary genetic approaches to validate a few exemplary small molecules and implicate their annotated targets in our identified phenotype of M. tuberculosis growth restriction within macrophages. Three previous chemical screens had identified compounds that restrict M. tuberculosis growth within the context of macrophages. The first, by Christophe et al., focused on screening compounds that are predominantly without known activities [15]. Moreover, the compounds selected for follow-up target essential bacterial processes rather than host proteins relevant to the host-pathogen interaction. A similar screen by the same group also focused on compounds without known activities that targeted essential bacterial processes [16]. A third chemical screen was, like our screen, skewed toward bioactive compounds that target host proteins [17]. Of our 133 identified active compounds, 21 were screened by Sundamarthy et al., and of those compounds 13 were identified as active by their criteria. While distinct libraries screened explain some of the differences between the lists of hits, more subtle differences contribute as well. Of the 8 compounds identified in our screen that were screened but not identified as hits in the screen by Sundamarthy et al., several score close to their threshold for consideration as a positive result. The discrepancies for those compounds likely reflect differences in the precise cutoff for determining hits and the sensitivity of the respective assays. However, some clear hits from our screen scored poorly in their assay, possibly reflecting biological differences between the assays. While our assay used M. tuberculosis and was designed to identify a visual output that approximated differences in mycobacterial growth as optimized using typical antimycobacterial agents, the assay used by Sundamarthy et al. used Bacille Calmette Guerin and had a complex visual output based on phenotypic changes including host cell morphologic changes, phenotypes felt to represent toxicity to the host cell, bacterial size, bacterial number, and bacterial intensity. In fact, they report that traditional antimycobacterial agents did not perform well in their screen, suggesting that it was not designed primarily to identify molecules that affect bacterial load when compared to the gold standard of CFU. Their phenotype likely represented growth restriction mediated by the particular host processes captured by their visual output, such as enhanced autophagy, rather than overall growth restriction. Overlapping hits between the screens may all have a similar mechanism of growth restriction, while compounds that were hits in our screen but not in the screen by Sundamarthy et al. may have distinct mechanisms of action. We identified several ion channel inhibitors in our screen, including niguldipine and verapamil, which block calcium channels, and flecainide and quinidine, which block sodium channels. A role for voltage-gated calcium channels in mycobacterial infection has previously been suggested [43]. In addition to altering calcium signaling, at least one calcium channel blocker, verapamil, has been shown to inhibit a bacterial efflux pump that is important in the context of intracellular infection [44]. Multiple calcium channel blockers, including verapamil, are used extensively in the clinic for cardiac applications. Our work adds to the growing body of work that suggests a role for calcium channel blockers as adjunctive therapy. In fact, in a recent paper, Gupta et al. tested the addition of verapamil to the standard background regimen of isoniazid, rifampin, and pyrazinamide to treat tuberculosis in a murine model, and found that addition of verapamil both reduced CFU during the active phase of infection and reduced rates of relapse [45]. Whether these reductions are the result of the effect of verapamil on the host or the bacterium is unclear. Nevertheless, this work validates the possibility that compounds identified in this screen may ultimately be useful as adjunctive agents for the treatment of tuberculosis. A potential role for sodium channels has not previously been described and merits further investigation. Our second class of targets, GPCRs, has only recently been suggested to have a role in mycobacterial infection. While recent work implicated a single GPCR, the D-3-hydroxybutyrate receptor GPR109A [46], our work identified primarily inhibitors of central nervous system-associated GPCRs, such as serotonin receptors and dopamine receptors. Although their function in macrophages is not well-understood, several central-nervous system associated GPCRs have been described to be expressed in macrophages [47], [48], [49]. While some of these receptors have previously been shown to have a role in modulating inflammatory cytokines in response to specific stimulation [50], [51], none have explicitly been shown to be involved in infection or the host-pathogen interface. Our results raise the intriguing possibility that these receptors play roles in the host immune response that are quite distinct from their roles in the central nervous system. The membrane transport proteins identified in this screen, including the serotonin transporter targeted by SSRIs, a dopamine transporter, and an acetylcholine transporter, have similarly not been shown to be involved in infection previously. Fluoxetine has been noted to increase systemic TNF-α [26], consistent with our results. The role of TNF-α during infection with M. tuberculosis is complex. Low TNF-α levels are clearly detrimental for the host and lead to impaired control of bacterial replication. However, overproduction of TNF-α can have host detrimental effects resulting from excessive tissue damage, induction of macrophage necrosis, and potentially from signaling bacteria to enter a nonreplicating antibiotic tolerant state [52], [53], [54]. Whether the levels of TNF-α produced during infection of humans with M. tuberculosis are optimal for bacterial replication or for host protection is not clear, and this likely varies with the genotype of both the infecting strain and the infected individual. Tobin et al. demonstrated in a zebrafish model of M. marinum infection that particular host genotypes of LTA4H resulted in either high or low levels of TNF-α production in response to infection, either of which was detrimental for outcomes. They went on to demonstrate that promoter polymorphisms for LTA4H in humans similarly resulted in high or low levels of TNF-α production. In an elegant translational component of their study, they demonstrated that among patients with tuberculosis meningitis, only those with the high- TNF-α producing genotype benefited from the standard addition of corticosteroids to their therapeutic regimen [52]. Conversely, it is likely that individuals that produce low levels of TNF-α upon infection with M. tuberculosis may benefit from adjunctive therapy that enhances production. Given the safety profiles of SSRIs, their widespread use, and their ability to modulate TNF-α levels and induce autophagy during infection, they are attractive candidates for further exploring the possibility of tailoring TNF-α levels to optimize host response in the infected individual. The bulk of the compounds in our anti-inflammatory category were non-steroidal anti-inflammatories (NSAIDs). Their identification is somewhat surprising as their inhibition of cyclooxygenase should block the production of a metabolite, prostaglandin E2, previously shown to be protective for infected macrophages and to reduce the bacterial burden in infected cells [5]. In fact, one would expect an increase in metabolites generated by competing pathways, including lipoxygenase-mediated production of lipoxins such as lipoxin A4, which has been shown to impair the host response to infection [5]. Thus, an NSAID-induced shift in metabolic balance away from prostaglandin E2 toward lipoxins might be expected to worsen the outcome of infection. In contrast, the identification in our screen of multiple distinct NSAIDs that restrict M. tuberculosis growth in macrophages suggests that the role of eicosanoid pathways in mycobacterial infection is potentially more complex than we understand thus far. Given how inexpensive, well-studied, and generally well-tolerated NSAIDs are, they certainly merit further study in animal models of infection. Corticosteroids were also represented in our anti-inflammatory category. Anti-inflammatory corticosteroids have been used as adjunctive therapy for tuberculosis for decades [55]. The use of corticosteroids for tuberculosis meningitis has been accepted into clinical practice; however, the potential benefit of adjunctive steroids for any other form of tuberculosis has been a matter of debate within the literature. A recent meta-analysis found that steroids significantly reduced mortality associated with tuberculosis infection of all organ groups [56]. Our finding that corticosteroids reduce bacterial burden in a cellular model of infection raises the question of whether steroids might impact infection at the level of infected macrophages that counteract the immunosuppressive effects on the whole organism level. We identified multiple kinase inhibitors in our screen. The identification of AKT and ABL family inhibitors validated our screen, as inhibitors of both classes have been previously shown to disrupt mycobacterial proliferation in macrophages. While a previous siRNA screen had identified AKT indirectly through association with genes identified in their screen [10], AKT was not directly identified using siRNA. Our study demonstrates the inherent advantage of small molecules over RNAi in simultaneously targeting multiple redundant isoforms of a given protein. Imatinib, an inhibitor of ABL family kinases, had similarly been demonstrated to be important for mycobacterial infection of macrophages, predominantly by both inhibiting bacterial uptake and subsequent replication [28]. Using primarily a model with M. marinum, which has significant differences from M. tuberculosis in its intracellular lifestyle, Napier et al. found that the effects of imatinib were to restrict M. marinum uptake and limit its intracellular growth, thus implicating imatinib-sensitive tyrosine kinases as important for virulence. In contrast, a previous study targeting ABL1 with siRNA suggested that depletion of ABL1 results in increased proliferation of M. tuberculosis in macrophages [11]. Here we demonstrate that ABL1 is specifically required for proliferation of M. tuberculosis in macrophages. Our study does not show reduced uptake of M. tuberculosis in cells treated with imatinib or with reduced ABL1 expression. Instead, reduction of ABL1 expression or inhibition of ABL function seems to specifically restrict intracellular growth at a later stage. Finally, we identified EGFR/p38 MAPK signaling pathway as a novel regulatory pathway during mycobacterial infection that functions to suppress effective antimicrobial responses. Inhibition of EGFR appears to restrict growth of intracellular mycobacteria through induction of autophagy in an AKT-independent mechanism, potentially through downstream inhibition of p38 MAPK. Previous studies have demonstrated a role for p38 MAPK as a negative regulator of both basal and starvation-induced autophagy via p38IP and mATG9 [41]. Our data suggests that EGFR signaling during M. tuberculosis infection activates a similar p38 dependent pathway that prevents clearance of the bacteria by autophagy. Using an EGFR inhibitor in a murine model of infection, we show the relevance of EGFR signaling in vivo and demonstrate that targeting the host with compounds already in clinical use for other applications holds potential for novel therapeutics for tuberculosis. This work thus demonstrates a possible therapeutic strategy of targeting host factors that modulate intracellular M. tuberculosis infection and replication. Repurposing of agents already in clinical use for other indications could expedite the testing of this strategy for tuberculosis. The general idea of modulating the host response to improve the outcome of tuberculosis treatment has circulated for some time. This concept however, is enormously complex as any benefit to the patient must integrate the impact of host-targeted intervention across all cell and tissue types and all systemic responses, and throughout the whole course of infection. Thus, benefits on a cellular, macrophage level may be counterbalanced by detrimental systemic effects involving numerous cell types and cytokine responses, or vice versa. Further, interventions with benefits at one point in infection, for example early in an inflammatory process, may not necessary be beneficial late in infection. Nevertheless, treatments that limit inflammation have been used both clinically and in experimental settings as adjunctive therapy for conventional antibiotics. Treatment of human patients with corticosteroids results in a modest decrease in mortality and is helpful in some forms of extrapulmonary tuberculosis including meningitis and pleural disease [56], possibly by limiting host inflammatory related tissue damage and/or by allowing M. tuberculosis to transition into a state of active replication in which it is sensitive to antibiotics [57]. Similarly, while anti-TNF-α agents lead to reactivation of mycobacterial disease, at the same time, blocking TNF-α has been suggested to favor a host microenvironment that favors bacterial clearance, particularly in the face of tubercular chemotherapy. Specifically, agents that either directly block TNF-α or inhibit signaling mechanisms that indirectly result in TNF-α production have been used to enhance the responsiveness of bacteria to conventional antibiotics. The TNF-α blocking agent etanercept increased the efficacy of conventional antibiotics during the chronic phase of infection, when the bulk of the bacterial are thought to be replicating slowly [58]. FDA-approved phosphodiesterase (PDE) inhibitors, which alter intracellular levels of cAMP resulting in reduced TNF-α secretion, likewise have been shown to reduce bacterial burden in rabbit and mouse models of infection when combined with current antimycobacterial antibiotics [42], [59]. Of note, an inhibitor of PDE-4 enhanced the effect of isoniazid on clearance of bacteria, but did not have an effect alone [42]. Inhibitors of PDE-3 and PDE-5 were not tested in the absence of standard tuberculosis therapy, so whether they would have an effect alone is unknown [59]. In contrast to work studying therapies that potentially manipulate immunity at a systemic level, a growing body of literature including this current work supports the idea that targeting specific host-pathways to enhance molecular mechanisms for bacterial clearance on the cellular level may be an effective adjunctive strategy for treatment [60]. Based on known host-pathogen biology or the RNAi screens described above, a variety of host targets have been identified and tested in animal models. Targeting lipid metabolism, a liver X receptor inhibitor reduced bacterial burden in the lungs of mice infected with M. tuberculosis [61]. As noted above, imatinib, an ABL inhibitor, has been shown to reduce bacterial burden in a murine model of infection [28]. Additionally, a TGF-β receptor inhibitor [11] and a GPR109A inhibitor [46], selected for study based on findings from the RNAi screen of M. tuberculosis infected THP-1 cells, were both shown to be similarly effective in reducing the burden of disease in mice. As the models for infection and treatment differ from study to study, it is difficult to directly compare efficacy between studies. Unlike the PDE inhibitors however, the liver X receptor inhibitor, ABL inhibitor, TGF-beta inhibitor, GPR109A inhibitor, and now an EGFR inhibitor all have efficacy even in absence of traditional tuberculosis therapy. These results suggest that enhancing host mechanisms of intracellular killing is a viable option for novel TB therapies, while reiterating the need to move quickly from in vitro cellular models to infected animals to determine the effect of an intervention on the complex, intact immune response of a whole organism. Additionally, efforts to determine the timing during an infectious process or the host genetic background where host intervention might be most beneficial will be important to such approaches. As the problems of MDR- and XDR-TB grow globally, identifying new therapeutic approaches will be critical for decreasing morbidity and mortality, and potentially disrupting the transmission of these highly resistant strains. Traditional drug development pipelines for anti-tuberculosis antibiotics have proven slow to move from lead compounds to clinically deployed medications. As several host-acting compounds that are already in clinical use with well-understood pharmacology and side effects have been shown to be effective in animal models of disease, such compounds could be rapidly tested in expanded animal models of infection and perhaps even moved to humans directly, given the limitations of existing animal models, to clarify the role of host-targeted therapies on treatment efficacy and duration. Alternatively, for some host-modulating compounds that are commonly used within patient populations (i.e, calcium channel blockers, NSAIDs, SSRIs), retrospective data may exist to provide initial support for any potential benefit of these host-targeted inhibitors. While some host-targeted therapies, including some kinase inhibitors, would currently be relatively expensive to administer, they are within the cost range of other medications being studied for use against for highly drug-resistant tuberculosis, including linezolid, and with time, their costs will drop as they become available off patent [62]. Others, such as some SSRIs or calcium-channel blockers, are currently inexpensive and would be amenable to inclusion in treatment regimens. Given the current state of tuberculosis, with rising incidence of MDR, XDR, and even TDR-TB, novel strategies are required that move beyond the conventional paradigm of an antibiotic that kills the bacterium in axenic culture. Targeting the host is one such strategy that can be tested as a feasible path forward exists, facilitated by the repurposing of current drugs. Materials and Methods Ethics statement Animal work was approved by Massachusetts General Hospital IACUC (protocol number 2009N000203) or the Harvard Medical School HMA Standing Committee on Animals (protocol number 03000). All protocols conform to the USDA Animal Welfare Act, institutional policies on the care and humane treatment of animals, the “ILAR Guide for the Care and Use of Laboratory Animals,” and other applicable laws and regulations. Imaging assay protocol J774 macrophages were seeded into 96-well black clear-bottom plates. M. tuberculosis strain H37Rv constitutively expressing GFP [18] was grown to mid-log phase in axenic culture, washed in PBS, briefly opsonized in heat-inactivated horse serum, and used to infect cells at an MOI of 1∶1. Infection was allowed to progress for four hours, then media was aspirated, cells were washed once with PBS, and media containing the screening compounds at an average concentration of 5 uM was added back to cells. Three days after infection, media was aspirated, cells were washed once with PBS, and fixed with 4% paraformaldehyde with Triton X-100 and DAPI. Image acquisition and analysis Plates were imaged using an Image Xpress Micro high-throughput microscope (Molecular Devices). Images were taken with a 4× objective at four sites per well. Images were then analyzed using CellProfiler open-source software [19]. The imaging-analysis pipeline is openly available (http://cellprofiler.org/published_pipelines.shtml) and included correction to homogenize illumination over each field, a filter to remove any large debris from the analysis, identification and quantitation of DAPI-stained nuclei, identification, quantitation, and pixel intensity calculation for GFP-expressing bacteria. The final output was calculated as (average GFP pixel intensity per bacterium across the field)×(number of bacteria identified in the field)/number of nuclei per field. The four images sites per well were averaged. To identify hits we used p-values obtained by bootstrap Monte Carlo and composite z-scores (see Methods S1). J774 infections for CFU J774 cells were seeded into 24 well plates. M. tuberculosis strain H37Rv was prepared as described above, then used to infect cells at an MOI of 1∶1. Phagocytosis was allowed to progress for 4 hours; cells were then washed once with PBS and fresh media containing compound was added back. Day 3 after infection, cells were washed once with PBS, lysed in 0.5% Triton X-100, and plated on 7H10 plates in serial dilutions. BMDM isolation and infections Bone marrow was obtained from C57BL/6 mice. In brief, adult male mice were euthanized and femurs and tibias were harvested. Bone marrow was flushed from the cells, resuspended in DMEM, and plated non-tissue culture treated dishes in DMEM media containing 2 mM L-glutamine, 20% fetal bovine serum, and 25 ng/ml recombinant mouse M-CSF. Cells were harvested day 6 after bone marrow isolation and either plated for subsequent infection or frozen. For infections, cells were plated in 24 well plates and infected as above with H37Rv. Media was changed every two days. Day 5 after infection, cells were washed once with PBS, lysed with 0.5% Triton X-100, and plated for CFU. siRNA and blocking antibodies For siRNA silencing experiments, J774 cells were plated in a 6-well dish. 20 pmol siRNA duplex was added in Optimem (Gibco) with 9 µl Lipofectamine RNAiMax (Invitrogen) at 24 h and again at 48 h after plating. The following day, cells were harvested, counted and re-plated for M. tuberculosis infections. 24 h after replating, the cells were infected with H37Rv as described above. To assess the efficiency of silencing, lysates were prepared at the same timepoint that the cells are infected with M. tuberculosis. For experiments blocking EGFR signaling, anti-EGFR antibody 225 (Millipore) and EGFR antibody Ab-5 (Thermo Scientific) were used at a final concentration of 20 nM. Media containing antibodies was replenished every 24 h. Macrophage assays For monitoring LC3 conversion by Western blot analysis, 2×105 J774 macrophages were plated per well in 12 well dishes and were infected with H37Rv at an MOI = 1. After 4 h the infected monolayer was washed once with PBS and media containing inhibitors was added. Three hours later the cells were washed, and protein lysates were prepared and run on a 15% SDS-PAGE gel. LC3-I and LC3-II were detected using an antibody from Cell Signaling Technologies. For TNF-α ELISAs, J774 or mouse BM macrophages were plated at 5×104 cells per well in a 96 well plate and infected at an MOI = 1. Monolayers were washed with PBS after 4 h phagocytosis and media with inhibitors was added. Cell supernatants were collected 24 h later, and assayed for TNF-α using an ELISA kit (Invitrogen). For phospho-p38 and p38 westerns, J774 cells were infected with H37Rv at an MOI of 1. Phagocytosis was allowed to proceed for 4 hours. Cells were then washed with warmed media, and then treated with gefitinib or DMSO carrier. Cells were harvested at 0 minutes, 15 minutes, 30 minutes, and 60 minutes after drug treatment and probed for phospho-p38 or p38 using antibodies from Cell Signaling Technologies. Mouse infections BALB/c mice were infected in a Madison aerosol chamber as previously described [63]. 5 mice were sacrificed day 1 after infection, and their lungs were homogenized and plated for CFU to determine the number of implanted bacteria. Infection in the remaining mice was allowed to progress for 7 days. Beginning day 8 after infection, mice were then given intraperitoneal injections of gefitinib at 100 mg/kg or DMSO carrier for 6 days. Day 14 after infection, 5 mice in each experimental group were sacrificed, and lungs were homogenized and plated for CFU. Supporting Information Figure S1 Positive controls for development of high-content imaging assay. J774 cells in 96-well dishes were infected with GFP-expressing H37Rv at an MOI of 1∶1. Cells were then treated with rifampin at the indicated concentrations. Day 3 after infection, cells were fixed, stained with DAPI, and imaged. (A) Nuclei of macrophages are in blue; GFP-expressing M. tuberculosis are in green. (B) GFP channel only. For each image set image contrast was adjusted equally for every image to promote print quality. Images were not adjusted prior to analysis. (PDF) Click here for additional data file. Figure S2 Percent representation of compound categories in input library. (PDF) Click here for additional data file. Figure S3 Fluoxetine re-testing in J774 macrophages using the image analysis assay. J774 cells were seeded into 96-well plates at ∼3000 cells/well and allowed to adhere overnight. They were then infected with H37Rv-GFP at an MOI of 1∶1. After 4 hours of phagocytosis, cells were washed, and fluoxetine at the indicated concentrations or DMSO control was added to wells. Day 3 after infection, cells were fixed and stained with DAPI. The imaging pipeline described in detail in Methods S1 was used to measure normalized integrated GFP intensity (A) and quantify DAPI-stained nuclei compared to DMSO control for each condition (B). (PDF) Click here for additional data file. Figure S4 Gefitinib re-testing in J774 macrophages using the image analysis assay. J774 cells were seeded into 96-well plates at ∼3000 cells/well and allowed to adhere overnight. They were then infected with H37Rv-GFP at an MOI of 1∶1. After 4 hours of phagocytosis, cells were washed, and gefitinib at the indicated concentrations or DMSO control was added to wells. Day 3 after infection, cells were fixed and stained with DAPI. The imaging pipeline described in detail in Methods S1 was used to measure normalized integrated GFP intensity (A) and quantify DAPI-stained nuclei compared to DMSO control for each condition (B). (PDF) Click here for additional data file. Figure S5 GNF2 re-testing in J774 macrophages using the image analysis assay. J774 cells were seeded into 96-well plates at ∼3000 cells/well and allowed to adhere overnight. They were then infected with H37Rv-GFP at an MOI of 1∶1. After 4 hours of phagocytosis, cells were washed, and GNF2 at the indicated concentrations or DMSO control was added to wells. Day 3 after infection, cells were fixed and stained with DAPI. The imaging pipeline described in detail in Methods S1 was used to measure normalized integrated GFP intensity (A) and quantify DAPI-stained nuclei compared to DMSO control for each condition (B). (PDF) Click here for additional data file. Figure S6 AKTi1/2 re-testing in J774 macrophages using the image analysis assay. J774 cells were seeded into 96-well plates at ∼3000 cells/well and allowed to adhere overnight. They were then infected with H37Rv-GFP at an MOI of 1∶1. After 4 hours of phagocytosis, cells were washed, and AKTi1/2 at the indicated concentrations or DMSO control was added to wells. Day 3 after infection, cells were fixed and stained with DAPI. The imaging pipeline described in detail in Methods S1 was used to measure normalized integrated GFP intensity (A) and quantify DAPI-stained nuclei compared to DMSO control for each condition (B). (PDF) Click here for additional data file. Figure S7 FTT re-testing in J774 macrophages using the image analysis assay. J774 cells were seeded into 96-well plates at ∼3000 cells/well and allowed to adhere overnight. They were then infected with H37Rv-GFP at an MOI of 1∶1. After 4 hours of phagocytosis, cells were washed, and FTT at the indicated concentrations or DMSO control was added to wells. Day 3 after infection, cells were fixed and stained with DAPI. The imaging pipeline described in detail in Methods S1 was used to measure normalized integrated GFP intensity (A) and quantify DAPI-stained nuclei compared to DMSO control for each condition (B). (PDF) Click here for additional data file. Figure S8 Ritanserin re-testing in J774 macrophages using the image analysis assay. J774 cells were seeded into 96-well plates at ∼3000 cells/well and allowed to adhere overnight. They were then infected with H37Rv-GFP at an MOI of 1∶1. After 4 hours of phagocytosis, cells were washed, and ritanserin at the indicated concentrations or DMSO control was added to wells. Day 3 after infection, cells were fixed and stained with DAPI. The imaging pipeline described in detail in Methods S1 was used to measure normalized integrated GFP intensity (A) and quantify DAPI-stained nuclei compared to DMSO control for each condition (B). (PDF) Click here for additional data file. Figure S9 Testing selected hits for activity against M. tuberculosis growing in axenic culture. M. tuberculosis was grown to mid-log phase, then diluted back to an OD600 of 0.05. Compounds were added at the concentrations indicated, and the cultures were incubated at 37°C. On days 3, 7, and 14 after inoculation, cells were mixed, and OD600 was recorded. At the tested concentrations, which are the maximum concentrations used in macrophages, no compounds had significant activity against M. tuberculosis in axenic culture. (PDF) Click here for additional data file. Figure S10 Targeting protein expression in J774 macrophages using siRNA. J774 macrophages were transfected with a pool of 5 siRNAs targeting AKT1, AKT2, AKT3, or ABL1 on two consecutive days. Control samples were transfected with a pool of 5 nonspecific siRNAs. 24 h after the second transfection the cells were harvested and split into fresh plates for 24 h at which time lysates were prepared for (A) Western blot analysis. To assess protein levels present in the cells at the time of infection, lysates were prepared at the same time-point after transfection that the cells are infected with M. tuberculosis. Samples were blotted with antibodies specific for AKT1, AKT2, AKT3 or ABL1. Blots were stripped and reprobed with α–actin for loading control. (B) To assess efficiency of phagocytosis the cells were infected with wild-type H37Rv cells for a period of 4 h at which time the infected monolayers were washed, lysed, and CFU were enumerated by plating on agar plates. (PDF) Click here for additional data file. Figure S11 Detection of EGFR transcript by PCR in J774 cells. RNA was isolated from J774 cells. 1 µg of RNA was used as template for cDNA production with (+RT) or without (−RT) addition of reverse transcriptase. The cDNA was used as a template for standard PCR using primers to amplify a small fragment of cDNA crossing a splice junction. A band corresponding to transcript was detected in the +RT but not −RT samples after 25 cycles. Shown are the samples after 35 cycles. For comparison, 20 µl of 50 bp ladder (NEB) were run in lane 1. See Methods S1 for full protocol details including primer sequences. (PDF) Click here for additional data file. Figure S12 CellProfiler identification of bacteria and nuclei. A. Illumination-corrected DAPI image (nuclei). B. CellProfiler identification of nuclei (green outlines: nuclei, red outlines: excluded shapes based on size criteria). C. Illumination-corrected GFP image (bacteria). D. CellProfiler identification of bacteria (green outlines: bacteria). (PDF) Click here for additional data file. Methods S1 Additional methods including statistics, details of assay development, CellProfiler analysis and EGFR PCR. (DOCX) Click here for additional data file. Table S1 Compounds included in screened library and complete hit list with composite Z-scores and Monte Carlo bootstrap-determined p-values. (XLSX) Click here for additional data file.
Author and article information
Journal
Journal ID (nlm-ta): Antimicrob Agents Chemother
Journal ID (iso-abbrev): Antimicrob. Agents Chemother
Journal ID (hwp): aac
Journal ID (pmc): aac
Journal ID (publisher-id): AAC
Title:
Antimicrobial Agents and Chemotherapy
Publisher:
American Society for Microbiology
(1752 N St., N.W., Washington, DC
)
ISSN
(Print):
0066-4804
ISSN
(Electronic):
1098-6596
Publication date
(Electronic preprint):
29
August
2016
Publication date
(Electronic):
21
October
2016
Publication date Collection:
November
2016
Publication date PMC-release: 21
October
2016
Volume: 60
Issue: 11
Pages: 6709-6718
Affiliations
Author notes
Address correspondence to Guruprasad R. Medigeshi,
gmedigeshi@
123456thsti.res.in
.
Citation Medigeshi GR, Kumar R, Dhamija E, Agrawal T, Kar M. 2016. N-Desmethylclozapine, fluoxetine, and salmeterol inhibit postentry stages of the dengue virus life cycle. Antimicrob Agents Chemother 60:6709–6718. doi: 10.1128/AAC.01367-16.
Author information
Guruprasad R. Medigeshi
http://orcid.org/0000-0001-5333-9743
Article
Publisher ID: 01367-16DOI: 10.1128/AAC.01367-16
PMC ID: 5075077
PubMed ID: 27572397
SO-VID: 416c65d5-610d-4192-9b30-6806c75503da
Copyright © Copyright © 2016 Medigeshi et al.
License:
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license.
History
Date
received
: 26
June
2016
Date
revision requested
: 8
July
2016
Date
accepted
: 21
August
2016
Page count
Figures: 7,
Tables: 0,
Equations: 0,
References: 47,
Pages: 10,
Words: 7049
Funding
Funded by: Wellcome Trust-DBT India Alliance
Award ID: IA/S/14/1/501291
Award Recipient
: Guruprasad R. Medigeshi
Additional funding support was from intramural funds.
Categories
Subject:
Antiviral Agents
ScienceOpen disciplines: Infectious disease & Microbiology
Data availability:
ScienceOpen disciplines: Infectious disease & Microbiology