RNAi Screen of Endoplasmic Reticulum–Associated Host Factors Reveals a Role for IRE1α in Supporting Brucella Replication

Brucella species are facultative intracellular bacterial pathogens that cause brucellosis, a global zoonosis of profound importance. Although recent studies have demonstrated that Brucella spp. replicate within an intracellular compartment that contains endoplasmic reticulum (ER) resident proteins, the molecular mechanisms by which the pathogen secures this replicative niche remain obscure. Here, we address this issue by exploiting Drosophila S2 cells and RNA interference (RNAi) technology to develop a genetically tractable system that recapitulates critical aspects of mammalian cell infection. After validating this system by demonstrating a shared requirement for phosphoinositide 3-kinase (PI3K) activities in supporting Brucella infection in both host cell systems, we performed an RNAi screen of 240 genes, including 110 ER-associated genes, for molecules that mediate bacterial interactions with the ER. We uncovered 52 evolutionarily conserved host factors that, when depleted, inhibited or increased Brucella infection. Strikingly, 29 of these factors had not been previously suggested to support bacterial infection of host cells. The most intriguing of these was inositol-requiring enzyme 1 (IRE1), a transmembrane kinase that regulates the eukaryotic unfolded protein response (UPR). We employed IRE1α−/− murine embryonic fibroblasts (MEFs) to demonstrate a role for this protein in supporting Brucella infection of mammalian cells, and thereby, validated the utility of the Drosophila S2 cell system for uncovering novel Brucella host factors. Finally, we propose a model in which IRE1α, and other ER-associated genes uncovered in our screen, mediate Brucella replication by promoting autophagosome biogenesis.


Introduction
Infectious diseases caused by intracellular bacterial pathogens are responsible for an enormous amount of worldwide pain, suffering, and mortality. Brucella spp., for example, cause brucellosis, a global zoonosis of profound importance [1,2]. Brucella melitensis, B. abortus, and B. suis are highly infectious and can be readily transmitted in aerosolized form [3,4]. In addition, they have eluded systematic attempts at eradication for more than a century, even in most developed countries, and a human vaccine against brucellosis is not available [3]. Therefore, Brucella spp. have been classified as potential bioterror threat agents [5], and have generated significant interest in the biosecurity and world health communities.
Understanding the molecular mechanisms of Brucella pathogenesis and host response is critical for brucellosis control, and intracellular trafficking and replication of Brucella spp. play important roles in these processes [6][7][8]. First, bacteria, internalized from the host cell plasma membrane, orchestrate the biogenesis of early Brucella-containing vacuoles (BCVs) [9,10]. Next, BCVs acidify but nevertheless fail to accumulate mannose 6-phosphate receptors (M6PRs) and cathepsin D, markers for late endosomes and lysosomes, respectively [8,11]. Instead, maturing BCVs fuse with membranes that contain endoplasmic reticulum (ER) resident proteins, including calreticulin and calnexin [7,8,11]. In addition, this trafficking involves BCV interactions with a compartment that contains the autophagosomal marker monodansylcadaverin [7,12]. Finally, Brucella spp. replicate in an ERlike compartment, and then presumably lyse the host cell to allow the infectious cycle to begin anew [8,13,14].
Bacterial lipopolysaccharides (LPS) play an important role in directing the bacterium along an intracellular trafficking pathway that enables a productive infection to be established. Brucella LPS also protects the bacterium from the harsh intracellular environment, suppresses pro-inflammatory and antibacterial host responses, and interferes with antigen presentation in macrophages [15]. Unlike their smooth wild-type (WT) counterparts, B. melitensis or B. abortus mutants harboring a deletion in the phosphomannomutase gene (DmanBA) lack LPS O-antigens, form rough colonies on solid medium, and are rapidly internalized by macrophages via a poorly understood pathway [16,17]. However, these mutants fail to establish an intracellular replicative niche and reportedly induce a necrotic cytopathic effect in these cells [18,19]. The bacterial type IV secretion system (T4SS) is also important for bacterial pathogenesis, and mutant strains lacking this system fail to traffic to, or replicate in, the ER [7,[20][21][22].
To date, relatively few host factors, including Rho1, Rac1, Cdc42 [23] and Sar1 [8], have been shown to be important for Brucella infection. Phosphoinositide 3-kinase (PI3K) activities have also been implicated in supporting Brucella infection [23]. Despite these advances, factors that mediate Brucella infection of host cells remain obscure. However, Brucella intracellular trafficking from the plasma membrane to an ER-associated replicative niche involves interactions with a membrane bounded compartment that contains autophagosome markers [7,12]. In addition, the organism replicates within a compartment that contains ER resident proteins [7,8,11]. These data thereby suggest that host cell autophagic pathway proteins, and ER-associated factors, may regulate the intracellular trafficking and replication of the pathogen.
Recent developments in the use of evolutionarily divergent Drosophila S2 cell model systems to study host-pathogen interactions, and RNA interference (RNAi) technology for knocking down host gene expression, have provided unprecedented opportunities for making significant progress in elucidating Brucella host factors. Drosophila S2 cells are macrophage-like cells that recapitulate conserved aspects of innate immunity [24] and that have been exploited for studying mammalian host-pathogen interactions. RNAi-based forward genetic screens in S2 cells have, for example, identified novel host factors involved in the recognition and replication of significant human bacterial pathogens, including E. coli [25], Listeria [26,27], Mycobacterium [28], Legionella [29], and Chlamydia [30,31]. Importantly, mammalian orthologs of hits identified in these screens have been shown to be important for bacterial infection of mammalian cells, thereby validating the utility of this Drosophila cell model for host-pathogen studies [28][29][30][31]. In this study, we show that the Drosophila S2 cell-Brucella interaction system recapitulates critical aspects of Brucella infection of mammalian cells.
In addition, we demonstrate the power of this system by identifying novel Brucella host factors, including IRE1a, a conserved transmembrane kinase that plays a key role in regulating the host cell unfolded protein response (UPR) [32][33][34]. Finally, we demonstrate that IRE1a is required for Brucella infection of mammalian cells, and discuss a possible mechanism by which this intriguing protein may regulate bacterial infection.

Brucella infection of Drosophila S2 and mammalian cells shares striking similarities
If Drosophila S2 cells are to provide a model system for studying Brucella infection, then they must support bacterial entry and replication. In addition, isogenic Brucella mutants with established entry, intracellular trafficking and replication properties should behave similarly in S2 cells and mammalian macrophages. Finally, Brucella should display similar infection phenotypes in S2 and mammalian cells that have been treated with compounds that disrupt host cell functions. With these ideas in mind, we employed gentamicin protection assays [18] to examine the entry and replication of different B. melitensis and B. abortus WT and mutant strains (listed in Table S1) in S2 cells. Because S2 cells require temperatures below 30uC for growth, all infection experiments were performed at 29uC, unless otherwise indicated. Importantly, J774A.1 cells supported Brucella entry and intracellular replication at this temperature (Fig. S2).
To easily visualize the intracellular trafficking and replication of Brucella spp., we exploited a GFP-expressing 16M strain (henceforth 16M-GFP) (Fig. S4). A comparison of the intracellular trafficking of Brucella spp. in S2 and mammalian cells indicated that the pathogen follows similar pathways in both host cell systems. BCVs trafficked to and replicated within an intracellular compartment that contained ER markers (e.g., mSpitz in S2 cells) [37], and was closely associated with COPII-coatomer (Sec 23) proteins ( Fig. S5A and data not shown) in both cell systems. Quantitative analysis also demonstrated that the bacterium failed to accumulate late endosome, Golgi marker (dGRASP) [38], or lysosomal markers in S2 or mammalian cells ( [7,8,12] and Fig. S5B). In addition, heat killed, formaldehyde fixed, and DvirB controls did not similarly colocalize with ER markers in either system ( Fig. S5A and data not shown). Therefore, the intracellular trafficking of B. abortus and B. melitensis in S2 and mammalian cells shared striking similarities.
Similar infection profiles were observed when B. abortus was used to infect mammalian or S2 cells that were treated with several Author Summary Brucella spp. are facultative intracellular pathogens that cause brucellosis in a broad range of hosts, including humans. Brucella melitensis, B. abortus, and B. suis are highly infectious and can be readily transmitted in aerosolized form, and a human vaccine against brucellosis is unavailable. Therefore, these pathogens are recognized as potential bioterror agents. Because genetic systems for studying host-Brucella interactions have been unavailable, little is known about the host factors that mediate infection. Here, we demonstrate that a Drosophila S2 cell system and RNA interference can be exploited to study the role that evolutionarily conserved Brucella host proteins play in these processes. We also show that this system provides for the identification and characterization of host factors that mediate Brucella interactions with the host cell endoplasmic reticulum. In fact, we identified 52 host factors that, when depleted, inhibited or increased Brucella infection. Among the identified Brucella host factors, 29 have not been previously shown to support bacterial infection. Finally, we demonstrate that the novel host factor inositol-requiring enzyme 1 (IRE1) and its mammalian ortholog (IRE1a) are required for Brucella infection of Drosophila S2 and mammalian cells, respectively. Therefore, this work contributes to our understanding of host factors mediating Brucella infection.
compounds; these compounds disrupted host cell functions and did not impair the bacterial growth in culture, or the viability of infected S2 cells (Fig. S6). These included: cytochalasin D [23], a compound that disrupts actin polymerization; bafilomycin A1, a specific inhibitor of vacuolar H + -ATPase activity and endolysosomal acidification [39]; brefeldin A (BFA), a fungal metabolite that prevents the assembly of COPI coated vesicles and disrupts vesicular transport [7,8] (Table S2 and Fig. S7A and S7B). Treatment of S2 and J774.A1 cells with the PI3K inhibitor wortmannin (WM) significantly reduced entry of B. abortus and B. melitensis ( Fig. S7A and data not shown). However, WM treatment of S2 and J774.A1 cells had no effect on the replication efficiency of the internalized bacteria (Fig. S7C, and data not shown). These findings were similar to those previously reported in mammalian cell systems [8,23,39,40]. In addition, we performed several experiments to assess the role of sphingolipids in supporting bacterial infection, and exploited myriocin (MR), a potent inhibitor of serine palmitoyltransferase (SPT), the first step in sphingosine biosynthesis [41], for these studies. B. abortus entry and survival were significantly inhibited when cells were treated with high MR concentrations ($1 mM). Low concentrations (#100 nM) of the compound had no effect on bacterial entry (Table S2 and Fig. S7A). However, the replication efficiency of the pathogen was decreased under these conditions (Fig. S7A).

RNAi-mediated inactivation of host factors required for Brucella infection
We employed RNAi technology to examine whether host proteins that are known to support bacterial infection of mammalian cells play similar roles in S2 cells. The evolutionarily conserved host proteins Rho, Rac, Cdc42 and Sar1 have been previously shown to be required for Brucella infection of mammalian cells [8,23], therefore, we examined whether these proteins were also required for Brucella entry and replication in S2 cells. Fluorescence microscopy image assays were employed for these studies because they offered a rapid and convenient method for assessing bacterial infection. Importantly, similar results were obtained when either fluorescence microscopy or gentamicin protection assays were performed (Table 1 and Fig. 2A). When S2 cells were depleted of Rac and Cdc42, the entry of B. abortus (S2308) or B. melitensis (16M) was impaired (Table 1 and Fig. 2A).
Rho1-depleted S2 cells appeared larger than untreated controls, contained numerous enlarged intracellular vacuoles, and also displayed significantly decreased levels of Brucella entry (Table 1, Fig. 2B). Sar1-depleted S2 cells also displayed dramatically reduced levels of Brucella replication ( Table 1, Table S3 and Fig. 2A) were observed in these cells. These findings were similar to results obtained when B. abortus was used to infect mammalian cells in which the activities of the corresponding human orthologous proteins had been depleted [8,23]. Therefore, the activities of these evolutionarily conserved GTP-binding proteins were required to support bacterial infection of both S2 and mammalian cells (Table 1, Table S3, Fig. 2 and Fig. S8).
To assess whether PI3Ks played similar roles in supporting bacterial infection of mammalian and S2 cells, we performed several experiments. First, we treated S2 and J774A.1 cells with WM and found that the levels of B. abortus and B. melitensis entry decreased in a similar fashion in both host cell systems (Table S2, Fig. S7A and data not shown). Second, we employed RNAi technology to deplete S2 cells of individual PI3K proteins and then measured bacterial entry and replication. These experiments revealed that multiple classes of PI3Ks are required to support B. abortus and B. melitensis WT strain infection (Table 1, Table S3, Fig. 2A, Fig. 3A and 3B). However, rough and smooth strains exploit separate host molecular pathways for entry [42]. when B. abortus rough strain CA180 was used to infect PI3K-depleted S2 cells, bacterial entry was dramatically enhanced (Fig. 3C). These data indicated that multiple PI3Ks play differential roles in mediating the entry of smooth and rough Brucella strains into S2 cells.

Experiments in mammalian cells confirm results obtained in Drosophila S2 cells
If Drosophila S2 cells are to serve as a useful model host cell system, then results obtained using this system should mirror corresponding mammalian cell findings. To test this possibility, we examined whether a murine ortholog (p85) of a model Drosophila gene (Pi3K21B) that supports Brucella infection of insect cells (Table 1, Table S3, Fig. 2A, Fig. 3A and 3B) mediates bacterial infection of murine cells. We used immortalized mouse embryonic fibroblasts (MEFs) derived from knockout mice harboring deletions in class I A PI3Ks (p85a and p85b) [43] for these studies. As expected, the levels of B. abortus and B. melitensis WT strains entry into MEF cells harboring PI3K gene deletions were dramatically reduced (Table 1 and data not shown). p85a 2/2 p85b 2/2 and p85b 2/2 MEFs supported lower levels of B. abortus and B. melitensis WT strains entry than p85 +/+ controls (Table 1 and data not shown). However, when these MEFs were infected with Brucella rough mutants (CA180 and S2308DmanBA), bacterial internalization significantly increased, especially in p85b 2/2 MEFs ( Fig. 3C and data not shown). These findings were similar to results obtained in experiments in which the entry of a Brucella rough mutant into class I A PI3K-depleted S2 cells was examined (Fig. 3C). Therefore, host cell PI3K isoforms differentially mediated the infection of smooth and rough organisms in both cell systems, and supported the use of the Drosophila cell system for elucidating novel Brucella host cell factors.

RNAi screen for ER-associated host factors
We were encouraged by our findings that previously described mammalian host proteins (i.e., Rho1, Rac, Cdc42 and Sar1) played similar roles in S2 cells. In addition, we noted that the Drosophila S2 cell system enabled the first molecular dissection of host cell PI3K isoform activity during Brucella infection. We therefore examined whether the Drosophila S2 cell system and RNAi technology could be combined to identify novel Brucella host factors. To focus our experiments, we constructed and screened 240 dsRNAs, including 110 dsRNAs that targeted the knockdown all of the genes annotated to be associated with the ER in the Drosophila RNAi Library Release 1.0 (Open Biosystems, Huntsville, AL, USA). The ER was ripe for examination because Brucella is known to replicate within a poorly characterized ER-like compartment, thereby suggesting that ER-associated host factors may be involved in regulating the intracellular replication of the pathogen.
Our ER-directed RNAi screen gave several interesting results. First, our screening approach successfully identified 52 hits. A hit was defined as a sample in which the relative infection differed by more than two standard deviations from the untreated control (Table S3). Importantly, control genes (i.e., Rho1, Rac, Cdc42, Sar1 and PI3Ks) were identified as hits in the screen (Table S3). Therefore, our screening strategy was sufficiently robust to uncover known or suspected host factors. We were curious whether the hit frequency obtained in our ER-targeted screen would be the same if a set of dsRNAs that were not associated with the ER were screened. We therefore screened 130 dsRNAs that were randomly picked from 2 of the 76 96-well plates in the Drosophila RNAi library. Because the manufacturer randomly arrayed dsRNAs into the source plates, this strategy for picking dsRNAs to be screen introduced no bias in the functions of the targeted genes in the screen. Notably, this experiment uncovered only 2 hits (,1.5% of the total) (Table S3), and therefore gave a hit frequency that was comparable to that observed in the Mycobacterium fortuitum and Listeria monocytogenes whole genome RNAi screens [26][27][28]. Interestingly, 14 out of 52 hits in our screen had been previously shown to mediate infection of S2 cells by Mycobacteria, Listeria, Legionella and Chlamydia infection [26][27][28][29][30][31] (Table S3 and Fig. 4). On the other hand, 29 genes were identified that had not been previously reported to be involved in supporting intracellular bacterial infection (Table S3). These novel genes were classified according to the gene ontology system of biological and molecular function, cellular component, or protein domains as reported in FlyBase (www.flybase.org). This classification revealed that the novel hits represented a variety of functional classes, including kinases, chaperones, and biosynthetic/metabolic enzymes. In addition, these 29 genes were localized to either the ER lumen (CG9429, CG30498) or ER membrane (CG6437, CG1063) (Table S3). We re-tested some of our most interesting hits in both fluorescence microscopy and gentamicin protection assays (Table  S3, repeat$3 times), and also employed quantitative reverse transcriptase polymerase chain reaction (Q-PCR) to verify that the expression of these genes in S2 cells was knocked down by dsRNA treatment. We typically obtained 60-90% knockdown of target gene expression in our screening plates ( Fig. 2C and data not shown).  Although each screen hit constituted a potential entry point for investigating the mechanism by which Brucella secures a replicative niche, we were particularly intrigued with IRE1 (CG4583), a key signal transducer that plays an important role in regulating the host cell UPR [32][33][34]. RNAi mediated knockdown of IRE1 gene expression resulted in significant reductions in Brucella replication (Fig. 5A, 5B and Table S3). In addition, IRE1 had not been previously implicated as a bacterial host factor. These data raised the intriguing possibility that IRE1 may play a novel role in regulating Brucella infection. We therefore examined whether IRE1a (the mammalian ortholog of Drosophila IRE1) was important for Brucella infection of mammalian cells.

IRE1a is required for efficient Brucella replication in mammalian cells
We performed several experiments to examine whether IRE1a played a critical role in supporting Brucella infection of mammalian cells. First, we infected IRE1a-null (IRE1a 2/2 ) and WT (IRE1a +/+ ) control MEF cells with 16M-GFP (Fig. 5C), and also performed gentamicin protection assays to assess bacterial entry and replication (Fig. 5D). The level of bacterial entry in IRE1a 2/2 was not statistically different from IRE1a +/+ controls (Fig. 5D). However, bacterial replication was significantly inhibited in IRE1a-depleted S2 cells and in IRE1a-null MEF cells (Fig. 5). Trypan blue dye exclusion analysis of 16M-infected MEF cells failed to reveal differences in host cell survival (data not shown). Therefore, the differences in bacterial replication efficiencies in these cell lines were not caused by the induction of host cell pro-apoptotic programs or by differences in the survival of Brucella-infected IRE1a 2/2 MEFs. Instead, they appeared to reflect a specific and important bacterial requirement for host cell IRE1a activity. Finally, the levels of entry and replication of Salmonella enterica serovar typhi, and the amounts of latex bead internalization, were similar in control and IRE1a 2/2 cells ( Fig. 6 and data not shown). These data supported the idea that IRE1a 2/2 cells do not possess general defects in phagocytosis, and that IRE1a activity is not required to support infection by all intracellular bacterial pathogens (Table S3 and Fig. 6).

Discussion
The study of host-Brucella interactions has suffered from the absence of a tractable genetic system to elucidate host factors. However, data obtained in this study indicate that Drosophila S2 cells provide a compelling model system for identifying and characterizing these important proteins. Brucella infection of Drosophila S2 cells recapitulates important aspects of mammalian cell infection. First, isogenic mutants of Brucella spp. behaved similarly in S2 and mammalian cells. In addition, these divergent host cell systems displayed similar trends in infection by smooth and rough strains with varied pathogenicity. Brucella rough mutants, such as CA180, were cytopathic to both mammalian and S2 host cells [18, 19, this study]. Therefore, these cells share conserved molecular mechanisms for recognizing and responding to Brucella LPS mutants. Second, Brucella entry and replication in S2 and mammalian cells were similarly sensitive to pharmacological perturbation by structurally diverse compounds. Of particular interest was the observation that MR, an inhibitor of STP, the rate-limiting enzyme in sphingolipid biosynthesis, dramatically reduced the amount of Brucella infection of S2 cells (this study). Previous studies have demonstrated an important role for sphingolipid enriched lipid rafts in pathogen infection [47][48][49][50], and our MR experiments support these observations. Third, Brucella infection of S2 cells required the activities of conserved GTP-binding proteins (Rho1, Rac, Cdc42, and Sar1), suggesting that Brucella infection of mammalian and Drosophila cells shared similar host molecular requirements. Finally, the activities of PI3Ks differentially regulate smooth and rough Brucella infection in both mammalian and Drosophila S2 cells (this study). Interestingly, the effects of PI3K knockdown in MEF cells were more dramatic than in S2 cells. In MEFs, PI3K genes are deleted, and thus the corresponding enzyme activities are absent. However, in Drosophila S2 cells, PI3K gene expression is knocked down (60-90%), and some residual activity may remain. These differences likely account for the differential infection of these cell types. Taken together, our data support the conclusion that S2 cells provide a useful model for investigating host-Brucella interactions.
Our demonstration that Drosophila S2 cells can be used to illuminate Brucella host factors is surprising because Brucella spp. do not occupy a described environmental niche outside of the mammalian host. In addition, the bacteria do not grow well in culture at temperatures below 35-37uC. However, previous reports have demonstrated B. suis multiplication within U937 cells at 30uC [51]. Therefore, Brucella growth below 37uC is not restricted to B. melitensis and B. abortus strains. Second, Brucella replication in J774A.1 and Drosophila S2 cells at 29uC share similar kinetics ( Fig. S2B and S2C). Although a difference in the replication efficiency of S2308DvirB2 in J774 and S2 cells at 24 and 48 h.p.i was observed, no difference was detected at 72 h.p.i. Therefore, the differential growth of B. abortus and B. melitensis in these host cell systems likely results from differences in the growth temperature, and not from differential subversion of conserved host cell functions. Third, the most important criterion for judging the utility of a model non-mammalian host-pathogen interaction system is whether it can be exploited to shed new insights into the interaction in mammalian cells. In this regard, it should be noted that bacterial pathogens, such as Listeria monocytogenes [52], grow more slowly in Drosophila S2 cells than in mammalian cells; however, many host factors required for entry and survival of these intracellular pathogens have been identified using Drosophila S2 cells as a platform [25][26][27][28][29][30][31]. We expect to garner similar insights through the use of our Drosophila S2 cell-Brucella interaction system, and our demonstration that PI3Ks and IRE1a mediate Brucella infection of Drosophila S2 cells and murine embryonic fibroblasts support this view.
Our RNAi screen in S2 cells for ER-associated Brucella host factors provides new insights into how Brucella secures an intracellular replicative niche. Our screen identified 52 genes that participate in this process, 29 of which had not been previously suggested to support bacterial pathogen infection. In addition, we dissected the role of 4 PI3K isoforms. The number of identified hits (50 out of 110 pre-selected ER-associated genes) was striking, and likely reflects that sustained and multi-faceted Brucella-ER interactions are required for Brucella replication in host cells. Interestingly, 14 of the genes identified in our screen were also required for infection of S2 cells by other intracellular bacterial pathogens, including Listeria, Mycobacteria, Legionella or Chlamydia [26][27][28][29][30][31]. The fact that Brucella and Legionella share several ERassociated host factors is perhaps not surprising, especially given that both organisms engage in sustained interactions with the host ER as part of their virulence and replication programs [53,54]. Finally, Brucella-specific ER-associated factors, such as IRE1 (CG4583), were uncovered in our screen. IRE1 may constitute a species-specific host factor that plays a role in mediating the unfolded protein response, thereby suggesting that the modulation of this stress-response system may be critical to bacterial intracellular survival and replication.
In eukaryotic cells, IRE1a mediated UPR induction is associated with enhanced expression of genes encoding ER chaperones and protein-folding catalysts, and proteins that participate in ERassociated degradation (ERAD) [55,56]. IRE1a activation also induces the biosynthesis of membrane phospholipids that increase the surface area and volume of rough ER [57,58]. In Brucella infected cells, IRE1a mediated activity may result in the biosynthesis of ER membrane that can be exploited by the pathogen to expand the size and enhance the quality of its replicative niche However, our data indicate that other UPR signal transducers, including PERK, are not required for Brucella infection in both Drosophila S2 and murine embryonic fibroblast cell systems. Therefore, not all UPR regulatory proteins are important for bacterial replication (Fig. 7 and Table S3), raising questions about the privileged status of IRE1a among these classes of molecules.
Recent reports have indicated an intriguing link between IRE1a activity and autophagic vacuole biogenesis [59,60]. For example, IRE1a is required for the autophagy observed after cells are treated with the ER stress-inducing agents DTT, tunicamycin or thapsigargin [59,60]. However, parallel experiments using PERKdeficient cells, and cells in which the expression of ATF6 had been knocked down, demonstrated that these UPR-associated signal transducers are not directly involved in the response to these drug treatments [60]. Therefore, IRE1a can regulate some autophagic events independently from input by these other ER associated signaling molecules. The differential participation of IRE1a, ATF6 and PERK in regulating the autophagy observed after cells are treated with stress-inducing agents is strikingly similar to their differential roles in mediating Brucella replication. IRE1a is required for Brucella to replicate efficiently; however, Brucella replication in PERK-, ATF6-, and BBF-2-depleted S2 cells was not significantly different from untreated controls. This differential participation therefore suggests a model in which IRE1a regulates Brucella infection by modulating the host cell autophagy pathway (Fig. 8).
Based on findings from our dsRNA screen, we propose a multistep model by which IRE1a regulates Brucella replication. First, BCVs traffic to a compartment that contains ER resident proteins. Concomitantly, BCVs trigger IRE1a activation, which in turn, stimulates the biogenesis of ER-associated autophagosomes (ERAs) [59,60]. ERAs then fuse with BCVs to form ERA-BCVs. This process is also regulated by the activities of PI3Ks. Finally, ERA-BCVs fail to fuse with lysosomes and hence avoid degradation; instead, they fuse with the ER to form ER-derived BCVs that are permissive for Brucella replication (Fig. 8).
Several pieces of evidence support this view. First, IRE1a, but neither PERK nor ATF6, is required for the induction of autophagy in response to treatment by ER stress-inducing agents [60]. Similar requirements for host proteins are observed during Brucella replication (Fig. 5, Fig. 7 and Table S3). Second, the assembly of ERAs is dependent upon early secretory pathway molecules [61][62][63]. In yeast, the COPII mutants sec16, sec23, and sec24, are defective in autophagy. However, mutations in two other COPII genes, sec13 and sec31, do not affect ERA biogenesis and autophagy [61,62]. In addition, PI3K activity is important for this process [64][65][66]. Our data demonstrate similar host factor requirements during Brucella infection of Drosophila cells. Specifically, depletion of Sec23, Sec24 and PI3Ks in host cells dramatically reduces Brucella replication (Table 1, Table S3, Fig. 2 and Fig. 3). However, depletion of Sec31 has no affect on this process (Table S3). Finally, Brucella trafficking to its intracellular replicative niche involves interactions with a compartment that contains the autophagosomal marker monodansylcadaverin [7,12]. These localization data thereby establish a physical interaction between internalized Brucella and the host cell autophagy pathway. It should be noted, however, that although we cannot rule out the possibility that Brucella trafficking in MEFs differs from professional phagocytes, Brucella trafficking in HeLa cells and phagocytes share striking similarities [12]. Therefore, our observations in MEFs likely shed light on Brucella infection of phagocytes. Taken together, the data are consistent with the idea that IRE1a activity plays an important role in supporting Brucella interactions with the host cell ERA biogenesis machinery in mammalian cells  (Fig. 8). Future studies will exploit the genetic power of the Drosophila S2 cell system to elucidate this intriguing possibility, and to define the precise molecular mechanisms by which Brucella secures an intracellular replicative niche.

Bacterial Strains
Brucella melitensis strain 16M (WT) and B. abortus strain 2308 (WT), and their derived mutants are listed in Table S1. Bacteria were grown in tryptic soy broth (TSB) or on tryptic soy agar (TSA, Difco TM ) plates, supplemented with either kanamycin (Km, 50 mg/ml), or chloramphenicol (Cm, 25 mg/ml) when required. For infection, 4 ml of TSB was inoculated with a loop of bacteria taken from a single colony grown on a freshly streaked TSA plate. Cultures were then grown with shaking at 37uC overnight, or until OD 600 <3.0.

Cell Culture
Murine macrophage J774.A1 cells, MEFs and HeLa cells were routinely cultured at 37uC in a 5% CO 2 atmosphere in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum (FBS). S2 cells were maintained at 25uC in Drosophila-SFM medium or in Schneider's Drosophila medium (Invitrogen) supplemented with 10% FBS. Cells were seeded in 24well plates and cultured overnight before infection. For antibiotic protection assays, 2.5610 5 cells were seeded in each well; for fluorescence microscopy assays (see below), 5610 4 cells were seeded on 12-mm glass coverslips (Fisherbrand) placed on the bottom of 24-well microtiter plates before infection.

Brucella Infection
Host cells were infected with Brucella at an MOI of 100, unless otherwise indicated. Infected cells were then incubated at 29uC (S2 Figure 8. Model describing how Brucella may exploit IRE1a to secure a replicative niche in an ER-like compartment. After PI3Kdependent Brucella internalization (1), intracellular Brucella-containing vacuoles (BCVs) traffic to the ER in a T4SS-dependent fashion (2). Accumulation of BCVs in the ER may modulate IRE1a activities (3), which then may trigger the biogenesis of ER-containing autophagosomes (ERAs) [59][60] (4). ERA biogenesis is known to require the activities of early secretory pathway components, including members of the COPII complex (Sar1-Sec23-Sec24) [61][62][63] (indicated in red). This process is also regulated by the activities of PI3Ks [64][65][66] (5). ERAs may then fuse with BCVs (6) to form ERA-BCVs (7). In addition, ER expansion may occur in response to these events. Finally, ERA-BCVs may fuse with an expanded ER membrane (8) and intercept ER proteins such as calreticulin (indicated in black) to form ER-derived BCVs that are permissive for Brucella replication. Depletion of members of COPII complex, PI3Ks and ER proteins such as calreticulin disrupted these processes and dramatically reduced Brucella replication. doi:10.1371/journal.ppat.1000110.g008 cells) or 37uC (mammalian cells) after centrifugation for 5 min at 2006g. Thirty minutes post-infection, culture media was removed, and the cells were rinsed with 16phosphate buffered saline (PBS). Fresh media, supplemented with 40 mg/ml gentamicin, was then added for 1 hr to kill extracellular bacteria. Infected cells were continuously incubated in this antibiotic for various lengths of time at the indicated temperature. As indicated, viable bacteria in infected cells were analyzed using the antibiotic protection assay or the immunofluorence microscopy assay described below. In addition, Brucella replication efficiency ([# of CFUs at different time points post infection]/[# of CFUs of Brucella entry]) in the infected cells was also determined

Antibiotic Protection Assays
At various times post-infection, viable bacteria present in infected cells were analyzed using gentamicin protection assays [18]. Briefly, infected cells were washed twice with 16PBS buffer, lysed with 0.5% Tween 20 in sterile water, and the released bacteria were subjected to serial dilution in peptone saline [1% (wt/vol) Bacto peptone and 0.85% (wt/vol) NaCl]. Next, 10 ml of serial diluted cell lysate was plated on TSA plates. Finally, CFU were counted after three days of incubation at 37uC.

Viability Assay of Infected Host Cells
S2 cells were coincubated with or without various drugs before 1 hr of and during Brucella infection (See below). Next, the infected cells were centrifuged at 2006g for 5 min and then incubated (30 min) with assorted Brucella strains. Fresh Drosophila-SFM media, supplemented with drugs (as indicated) and 80 mg/ml gentamicin, was added to kill extracellular bacteria. The infected and gentamicin treated cells were then incubated at 29uC for various lengths of time. To quantify the viability of S2 cells, at various time points, a portion of the infected cells was removed and processed for 0.2% trypan blue vital stain analysis. At least 500 cells were counted per sample. For image analysis, infected cells were replated onto ConA (Sigma)-coated 12-mm coverslips in 24-well plates and allowed to adhere for 1 hr. Cells were stained with 0.2% trypan blue for 5 min and then fixed with 16PBS containing 3.7% formaldehyde for 1 hr. Viability of infected cells was assessed by analyzing images obtained with an Olympus IX70 fluorescence microscope. At least 500 infected cells per sample were used for the analysis.

S2 Cell Transfection
To visualize Brucella spp. trafficking, S2 cells were transfected with ER marker mSpitz-GFP [37], and Golgi marker dGRASP-GFP [38] before infection. Specifically, S2 cells were grown to ,80% confluence and then transfected using Effectene Transfection Reagent (Qiagen) as per the manufacturer's instructions. 0.25 mg of each pUAS-mSpitz GFP and pAcpA-Gal4 were employed in these transfection experiments. For the Golgi visualization experiments, 0.25 mg of dGRASP-GFP was used in the transfection. Typically, 1.5610 6 cells were transfected and then grown in 2.2 ml of Schneider's Drosophila medium supplemented with 10% FBS. Three days post-transfection, cells were replated onto ConA-treated 12-mm glass coverslips placed on the bottom of 24-well microtiter plates (for early time points of less than 8 hr) and immunofluorescence microscopy analysis was performed as previously described [18]. For later times points ($8 hr), the transfected cells were reseeded directly in 24-well plates and allowed to adhere for 2 additional hours before infection with Brucella. At different post-infection time points, the infected cells were replated onto ConA-coated 12-mm coverslips and allowed to adhere for 1 hr. The cells were then washed three times with 16PBS, fixed with 3.7% formaldehyde (pH 7.4) at room temperature for 1 hr and processed for immunofluorescence microscopy.

Immunofluorescence Microscopy Assay
To elucidate Brucella spp. intracellular trafficking, S2 cells were infected with the following strains: B. melitensis (strains 16M or 16M-GFP); B. abortus (strain S2308); S2308 virB2 deletion mutants; heat killed or 3.7% formaldehyde fixed WT strains. At various post-infection time points, S2 cells were replated onto ConAcoated 12-mm coverslips and allowed to adhere for 45 min to 1 hr. Cells were then washed, fixed as described above, and processed for immunofluorescence microscopy [18]. The primary antibodies used were as follows: goat polyclonal anti-Brucella; rabbit antihuman M6PR; rabbit anti-human cathepsin D; goat-anti rabbit Sec23 (COPII marker, Affinity BioReagents, Inc., CO, USA). Samples were stained with Alexa Fluor 488-conjugated and/or Alexa Fluor 594-conjugated donkey anti-goat/rabbit (Molecular Probes, 1:1000). Cover slips were then mounted in Vectashield mounting media (Vector Laboratories, Inc., CA, USA) and visualized with an Olympus BX51 confocal microscope. For quantitative analysis, single confocal section of random fields was acquired, and colocalization of markers was scored as positive when nonsaturated signals partially overlapped. Images for all immunofluorescence assays for Brucella spp. trafficking were acquired with a Hamamatsu ORCA-ER camera mounted on the Olympus BX51 microscope and driven by Simple PCI software (Compix Imaging Systems Inc., Cranberry Township, PA.). Images were processed with Adobe Photoshop CS Software (Adobe Systems Incorporated, San Jose, CA).

Drug Treatments
Drosophila S2 cells or J774.A1 murine macrophages were coincubated in 24 well plates with assorted drugs including bafilomycin A1 (BAF), brefeldin A (BFA), cytochalasin D (CD), myriocin (MR) and wortmannin (WM) at the indicated concentrations. Cells were treated with drugs 1 hr before, and during, infection with the indicated Brucella strains. After infection, the treated cells were incubated at 29uC (S2 cells) or at 37uC with 5% CO2 (J774.A1 macrophages). To evaluate Brucella internalization, after 30 min of infection, fresh media, supplemented with the same concentration of the drugs and 80 mg/ml gentamicin was added to kill extracellular bacteria. After 45 min of incubation, the cells were lysed and the CFU per well determined by plating dilutions on TSA plates as described above. To assess Brucella intracellular replication, CFU analysis was performed at 72 h.p.i. The effect that BAF-mediated inhibition of host cell endosomal acidification exerted on Brucella replication was also examined. Briefly, BAF was added to the culture media 2 h.p.i. and continuously coincubated with infected cells for 72 hr. Cells were lysed and analyzed using the gentamicin protection assay. To investigate whether the drugs inhibit Brucella growth, the drugs were individually added to Brucella TSB cultures at 29uC or 37uC and incubated for 1 and 72 hr. CFU plating was used to assess bacterial growth in the presence of drugs, and thereby to evaluate the potential inhibitory effects.

Generation of dsRNAs
Primers for generating RNAi that target the knockdown of Drosophila Rac1, Rac2, Rho1, Cdc42, Sar1 and PI3Ks were designed using sequence information present in flybase (http:// flybase.org/). The primers were used in RT-PCR reactions to generate cDNAs. dsRNAs targeting genes to be knocked down were generated using previously described methods [26]. Briefly, gene-specific RNAi primers were used to amplify target sequences from Drosophila cDNA mixtures. The PCR products were reamplified using the RNAi primers with T7 RNA polymerase promoter sequences in the 59 end. The reamplified PCR products were then used as templates for the generation of dsRNAs. For generation of dsRNAs targeting ER-associated and other genes, cDNAs from commercially available Drosophila RNAi Library Release 1.0-DNA templates (Open Biosystems, Huntsville, AL, USA) were directly used as templates. One or two microliters (total ,150 ng) of the PCR products were used to perform in vitro transcription reactions with the T7 MEGAscript kit (Ambion, Austin, TX) as per the manufacturer's instructions. Aliquots of in vitro transcription products were subjected to quality control by 1% agarose gel electrophoresis analysis and dsRNA concentrations were quantified using a NanoDropH ND-1000 UV-Vis spectrophotometer (NanoDrop Technologies, Inc. Wilmington, DE).

RNAi-mediated Gene Knock Down and Assays
1.0610 6 S2 cells were seeded in 12-well plates. dsRNAs (i.e., Rho1, Rac, Cdc42, Sar1 and PI3Ks) were added to each well at a final concentration of 15 mg/ml. After 4 days of incubation with dsRNA, an aliquot of the S2 cells was removed to check the efficiency of dsRNA mediated gene knock down by quantitative RT-PCR (Q-PCR). dsRNA-treated S2 cells in the same well were also re-plated in 24-well plates and allowed to adhere for at least 2 hr before infection. At the selected time points, the dsRNA-treated and Brucella infected cells were lysed and antibiotic protection assays or fluorescence microscopy image assays were performed as described.
To evaluate the utility of the combination of S2 cells and dsRNA technology, and the consistency of the results from antibiotic protection assays, we analyzed Brucella infection using fluorescence microscopy image assays. dsRNAs that target ERassociated genes or other known or unknown genes were added to 96-well microplates at a final concentration of 15 mg/ml (dsRNAs were added in duplicate in two different plates). S2 cells were then seeded in the plates with 5.0610 4 cells/well in 200 ml Drosophila-SFM medium. dsRNA-treated cells were incubated at 25uC for 4 days to allow for knockdown of target gene expression. The dsRNA-treated cells (100 ml) were replated into 96 well plates, infected with B. melitensis 16M-GFP at an MOI of 50. After 30 min of infection, the same amount of fresh media supplemented with 80 mg/ml gentamicin was added to each well and the infected cells were incubated at 29uC. At 72 h.p.i., infected cells were replated onto 96 well glass bottom plates (Greiner), that had been coated with ConA, and allowed to adhere for 1 hr. The infected S2 cells were washed 3 times with 16PBS, fixed with 3.7% formaldehyde in 16PBS at 4uC overnight, and stained with phalloidin-Texas red (1:1000) for 1 hr to visualize the host cell actin cytoskeleton. Brucella infected S2 cells were viewed with an Olympus IX70 inverted microscope and two 4006 images from each well were acquired for image analysis. Images were analyzed using NIH Image J software (http://rsb.info.nih.gov/ij/), and the relative infection (RIF) [1006(% of infected dsRNA-treated cells)/(% of infected cells in the untreated control)] was determined. More than 1,000 S2 cells were counted to obtain the percentage of infection or infection index [(number of infected cells (at least 10 brucellae within the cell))/(number of total cells)] in a sample. The detailed process by which image analysis was performed is shown in Fig.  S1. dsRNA screen was repeated once, and some of hits identified in both two round of screens were picked out to re-test in triplicate in fluorescence microcopy and gentamicin protection assay as described above.

Mammalian Cell Infection
MEFs deficient of the two regulatory isoforms of class I A PI3Ks (p85a 2/2 p85b 2/2 and p85b 2/2 ) [43], IRE1a (IRE1a 2/2 ) [67] and PERK (PERK 2/2 ) [68] and their corresponding WT control p85 +/+ , IRE1a +/+ and PERK +/+ MEFs, were seeded in 24-well plates. After overnight culture, cells were infected with 16M-GFP and/or S2308, and their derived mutant strains. Infected cells were centrifuged for 5 min at 2006g and then incubated at 37uC for 60 min. Cells were washed with 16PBS buffer, and fresh media supplemented with 40 mg/ml gentamicin was added. Cells were incubated for an additional 1 hr (entry) and 48 hr (replication) at 37uC. The amount of viable bacteria present in infected cells was assessed using gentamicin protection assays. For fluorescence microscopy and viability assays, 5610 4 cells were seeded onto 12mm coverslips in 24-well plates. At 48 h.p.i., infected cells were subjected to the appropriate assays as described above.

Statistical Analysis
All quantitative data were derived from results obtained in triplicate wells for at least three independent experiments. The significance of the data was assessed using Student's t-test, and all the analyzed data were normalized with internal controls before analysis.    Figure S4 GFP expression has no effect on bacterial entry and replication. The entry (1.5 h.p.i) and replication (72 h.p.i.) of Brucella melitensis strains 16M and 16M-GFP in Drosophila S2 and J774.A1 murine macrophages at 29uC and 37uC, respectively, were compared using gentamicin protection assays. The number of 16M CFUs of entry and replication in Drosophila A. The indicated drugs were added into fresh TSB and Brucella (S2308) was incubated in this medium for the indicated periods of time. The effects of the drugs on Brucella growth were determined using gentamicin protection assays. 1 hr (white bars) and 72 hr (gray bars) represents the relative amount of Brucella in the drug-treated media (CFU/ml) compared with the untreated control at 1 and 72 hrs post coincubation, respectively. Replication efficiency (black bars) indicates relative Brucella replication efficiency in drug-treated medium and no drug treated control. B. Viability of Brucella (S2308) infected and drug-treated S2 cells. S2 cells were pretreated with drugs for 1 hr, and then infected with bacteria. At 72 h.p.i., cells were stained with trypan blue, fixed, and the percentage of viable cells was determined (Panel C). Two images, containing a total of at least 500 cells in each sample, were analyzed in each experiment. C. Images of infected S2 cells coincubated with the indicated drugs at 72 h.p.i.,