Yersinia pestis Targets the Host Endosome Recycling Pathway during the Biogenesis of the Yersinia-Containing Vacuole To Avoid Killing by Macrophages

ABSTRACT Yersinia pestis has evolved many strategies to evade the innate immune system. One of these strategies is the ability to survive within macrophages. Upon phagocytosis, Y. pestis prevents phagolysosome maturation and establishes a modified compartment termed the Yersinia-containing vacuole (YCV). Y. pestis actively inhibits the acidification of this compartment, and eventually, the YCV transitions from a tight-fitting vacuole into a spacious replicative vacuole. The mechanisms to generate the YCV have not been defined. However, we hypothesized that YCV biogenesis requires Y. pestis interactions with specific host factors to subvert normal vesicular trafficking. In order to identify these factors, we performed a genome-wide RNA interference (RNAi) screen to identify host factors required for Y. pestis survival in macrophages. This screen revealed that 71 host proteins are required for intracellular survival of Y. pestis. Of particular interest was the enrichment for genes involved in endosome recycling. Moreover, we demonstrated that Y. pestis actively recruits Rab4a and Rab11b to the YCV in a type three secretion system-independent manner, indicating remodeling of the YCV by Y. pestis to resemble a recycling endosome. While recruitment of Rab4a was necessary to inhibit YCV acidification and lysosomal fusion early during infection, Rab11b appeared to contribute to later stages of YCV biogenesis. We also discovered that Y. pestis disrupts global host endocytic recycling in macrophages, possibly through sequestration of Rab11b, and this process is required for bacterial replication. These data provide the first evidence that Y. pestis targets the host endocytic recycling pathway to avoid phagolysosomal maturation and generate the YCV.

in macrophages from canines and Mus spretus SEG mice, species that are relatively resistant to plague, compared to common laboratory murine macrophages, indicating that the ability of macrophages to kill Y. pestis may contribute to susceptibility to plague (32)(33)(34). These data, combined with studies showing the sensitivity of Y. pestis to PMN killing (20)(21)(22), suggest that Y. pestis infection of macrophages may provide an intracellular niche to avoid killing by PMNs during early stages of bubonic plague.
Upon phagocytosis by macrophages, Y. pestis actively inhibits phagosome-mediated killing (16,23,(25)(26)(27)(28). A hallmark of this process is the inhibition of phagosome acidification by Y. pestis (27). The bacterium remains within this phagosome throughout the course of the intracellular infection, eventually remodeling it into a compartment called the Yersinia-containing vacuole (YCV). In addition to maintaining a neutral pH, a subset of the YCVs eventually mature into autophagosome-like compartments, acquiring both LC3-II and double membranes (27). During late infection, the YCV expands from a tight-fitting vacuole to a spacious vacuole, coinciding with bacterial replication (23,27,33). Recently, we identified the first host factor required by Y. pestis for the YCV biogenesis process (35). The host GTPase Rab1b, which normally mediates endoplasmic reticulum (ER)-Golgi trafficking (36,37), is rapidly recruited to the YCV and is required for Y. pestis to inhibit vacuole acidification and phagosome maturation. These data demonstrate that Y. pestis manipulates host factors to subvert phagosomal maturation and to generate a protective replicative niche within the macrophage. Here we describe a genome-wide, RNA interference (RNAi)-based high-throughput screen to identify additional host factors required for intracellular survival of Y. pestis. Network analysis of these genes revealed enrichment for host factors involved in the endocytic recycling pathway. We further show that Y. pestis actively recruits Rab4 and Rab11 to remodel the YCV to resemble host recycling endosomes in order to avoid phagolysosomal maturation in a T3SS-independent manner. Moreover, we demonstrate that Y. pestis infection also disrupts global host cell recycling, likely through sequestration of Rab11, and that disruption in recycling is important for intracellular replication.

RESULTS
Y. pestis requires host cell signal transduction, transport, and localization pathways to survive in macrophages. RNAi has been used to identify host factors required for intracellular survival of several pathogens (38)(39)(40)(41)(42)(43)(44)(45)(46)(47)(48)(49). As macrophages are specifically infected during Y. pestis infection, our first goal was to select a macrophage cell line that was amendable to Lipofectamine-mediated transfection and RNAi necessary for high-throughput screening. Toward this end, we tested small interfering RNA (siRNA) transfection and knockdown in several human and mouse macrophage cell lines. While robust RNAi was observed in mouse macrophages, we were unable to reproducibly knock down gene expression in human cell lines (data not shown). On the basis of these results, we chose RAW264.7 mouse macrophages for further optimization. Using a combination of siRNAs targeting genes of variable expression levels, we optimized Lipofectamine/siRNA concentrations and transfection time to consistently achieve Ͼ70% knockdown of targets ( Fig. 1A and B). Next, we infected cells transfected with Rab2 siRNA and Cop␤1 siRNA (both siRNAs are known to inhibit intracellular survival of other pathogens [41,42,48]) with Y. pestis CO92 pCD1 (-) Lux PtolC , a bioluminescent bioreporter that can differentiate as little as a twofold difference in intracellular bacteria ( Fig. 1C and D; R 2 ϭ 0.89) (50), to demonstrate that this bioreporter can be used to kinetically monitor changes in intracellular survival (Fig. 1E). Finally, using Cop␤1 siRNA as a positive control, we calculated Z= factor values for this assay of 0.61 and 0.83 at 2 and 10 h postinfection, respectively (Fig. 1F). Together, these data indicate a highly reproducible assay amenable to high-throughput screening (Fig. 1G).
Using this RNAi assay, 17,370 host genes were screened using a pooled siRNA approach (three siRNAs for each target were pooled into one well), and we monitored changes in intracellular survival of Y. pestis. Each plate also contained control wells transfected with scrambled or COP␤1 siRNAs. Bioluminescence was measured at 2 and 10 h postinfection, and Z= factors were calculated from the control wells (the average Z= factors for the screen at 2 and 10 h were 0.57 and 0.66, respectively). Bioluminescence was normalized for each plate based on control wells, and changes in intracellular survival of Y. pestis were ranked by normalized scores ( Fig. 2A; see also Data Set S1 in the supplemental material). A total of 325 genes that inhibited bacterial growth and 39 genes that promoted bacterial growth were selected for secondary validation. For the secondary screen, a single siRNA (siRNA "A") was used to validate the primary screen results. Furthermore, the primary hits were validated against two different Y. pestis strains, one with the Ysc type three secretion system (T3SS) (KIMD19 pCD1 (ϩ) Lux PtolC ) and one without it (CO92 pCD1 (-) Lux PtolC ). A direct correlation was observed between the two strains ( Fig. 2B; r s ϭ 0.87), supporting previous studies showing that the T3SS is dispensable for intracellular survival (16,24,35,50). From the primary hits, 135 genes showed Ն40% inhibition of intracellular survival of Y. pestis and 7 showed a hypervirulent phenotype with Ն20% more growth than scrambled controls. These 142 genes were further screened using a second siRNA (siRNA "B"). Of the 142 genes, RNAi of 71 of these genes continued to show inhibition of intracellular Y. pestis survival, while one retained a hypervirulent phenotype (Ն10% more growth than scrambled controls; Data Set S1 and Table S1).
Gene Ontology (GO) clustering and network analyses of the 71 validated inhibition hits using all GO evidence codes, a minimum kappa score of 0.4, and a P value threshold of 0.05 revealed substantial clustering within the validated data set (Fig. 2D). Of the five enriched groups, the largest cluster was under vesicle-mediated transport ( Fig. 2D and E; P Յ 0.001). Under the parent GO term clusters, detailed GO terms significantly focused on host trafficking networks, with transport and localization as common themes. Additional enrichment included small-GTPase-mediated signal transduction and regulation of response to stress ( Fig. 2D and E). Within the small GTPase signal trafficking, six Rab GTPases appeared to be required for intracellular survival of Y. pestis, and two trafficking pathways were highlighted: (i) endocytic recycling (Rab4a and Rab20) and (ii) retrograde trafficking (Rab1b and Rab2b) (51)(52)(53)(54)(55)(56). Together, these data indicate targeting of specific host Rab-mediated signaling pathways by Y. pestis during infection of macrophages.
Host cell recycling is essential for Y. pestis survival. Rab4a, Rab11b, and Myo5b are well-characterized contributors to cell recycling (53). While Rab4a was a validated gene in our screen and interconnected to the largest enriched ontology (vesiclemediated trafficking), the other two genes did not pass the primary screen criteria. RNAi of Rab11b inhibited Y. pestis survival by only 50% in the primary screen, and Myo5B was cytotoxic (upon subsequent analysis, only one of the three Myo5B siRNAs used in the primary screen was cytotoxic; this siRNA was not included in further studies). However, because of the importance of these proteins in the recycling pathway, we chose to independently verify the contributions of Rab4a, Rab11b, and Myo5b to the intracellular survival of Y. pestis ( Fig. 3 and Fig. S1). RNAi resulted in Ͼ50% knockdown of each gene target (Fig. 3A) with no significant loss in cell viability (Fig. 3B). Subsequent infection confirmed that knockdown of all three genes impacted the ability of Y. pestis to survive within the macrophage. Knockdown of Rab4a had the largest impact, inhibiting Y. pestis survival by Ͼ40% at 2 h ( Fig. 3C; P Ͻ 0.001) and Ͼ80% at 10 h ( Fig. 3D; P Ͻ 0.001). Interestingly, knockdown of Rab11b and Myo5B had no significant impact on Y. pestis survival at 2 h ( Fig. 3C) but attenuated Y. pestis by Ͼ40% at 10 h ( Fig. 3D; P Ͻ 0.001). Bioluminescence data were confirmed at 10 h by conventional bacterial enumeration (Fig. 3E). Furthermore, while RNAi of Rab11b had a minor impact on intracellular survival between 2 h and 10 h postinfection, RNAi of Rab4a had a greater impact on the ability of bacteria to survive intracellularly between these two time points (Fig. 3F). Importantly, knockdown of Rab4a or Rab11b did not alter the expression of Rab GTPases involved with phagolysosome maturation or impact phagocytosis of Y. pestis (Fig. S2). Together, these data indicate that Y. pestis requires the host cell recycling pathway to avoid killing by macrophages. Y. pestis requires Rab4a to avoid YCV acidification and lysosomal fusion. We previously showed that RNAi of Rab1b significantly reduces intracellular survival of Y. pestis at 2 h postinfection, which directly correlated with an increase in the frequency of YCV acidification (35). Therefore, we next determined the impact of Rab4a, Rab11b, and Myo5b RNAi on YCV acidification using Lysotracker ( Fig. 4A to C). As previously shown (27,35), Y. pestis actively inhibited YCV acidification, with only~20% of YCVs containing live Y. pestis colocalizing with Lysotracker by 80 min postinfection. In contrast, YCVs containing paraformaldehyde (PFA)-killed Y. pestis rapidly acidified, with Ͼ75% colocalization by 20 min postinfection. As predicted by our relative light unit (RLU) data, Rab4a knockdown resulted in a significant increase in the frequency of YCV acidification (P Յ 0.001), approaching levels similar to PFA-killed bacteria. Unlike Rab4a, RNAi of Rab11b and Myo5b resulted in only a slight increase in colocalization, and by 80 min, colocalization remained significantly lower than both Rab4a siRNA-treated cells or cells infected with PFA-killed Y. pestis.
Phagosome acidification is directly linked to lysosomal fusion (57), so we next asked whether Rab4a is required for Y. pestis to avoid YCV fusion with lysosomal compart- Y. pestis Targeting of the Host Recycling Pathway ® ments by measuring YCV colocalization with the LAMP-1 lysosomal marker in siRNAtreated macrophages ( Fig. 4D to F). As observed for YCV acidification, Y. pestis actively inhibits lysosomal fusion, with only~20% of YCVs colocalizing with LAMP-1 during the first 80 min of infection. In contrast, YCVs containing PFA-killed Y. pestis reached Ͼ50% colocalization with LAMP-1 by this time point (P Յ 0.01). Knockdown of Rab4a also inhibited the ability of Y. pestis to avoid lysosomal fusion, resulting in a significant increase in colocalization with LAMP-1 (63.7%; P Յ 0.001). Together, these data suggest that Rab4a is required for Y. pestis to avoid phagosomal acidification and lysosomal fusion.
The Yersinia-containing vacuole acquires recycling endosome markers. To determine whether Rab4a or Rab11b is recruited to the YCV, primary peritoneal macrophages were infected with Y. pestis, PFA-killed Y. pestis, or Escherichia coli K-12, and Representative confocal microscopy images of RAW264.7 macrophages transfected with scrambled or Rab4a siRNAs that were infected with Y. pestis CO92 pCD1 (-) pGEN222 (MOI of 3) and stained with anti-LAMP-1 antibody. LAMP-1 is shown in red, and Y. pestis is shown in green. Yellow arrows indicate locations of bacteria that do not colocalize with LAMP-1 and white arrowheads indicate bacteria that colocalize with LAMP-1 based on Imaris COLOC function. Bars ϭ 5 m. (E and F) YCV colocalization with LAMP-1 was calculated using IMARIS at 20 min (E) and 80 min (F) postinfection. One-way ANOVA with Tukey's posthoc test was performed, and the results are indicated as follows: ns, not significant; *, P Յ 0.05; **, P Յ 0.01; ***, P Յ 0.001. The treatments are indicated as follows: Scr, scrambled siRNA; Killed, untransfected macrophages infected with PFA-killed Y. pestis CO92 pCD1 (-) pGEN222; Rab4a, Rab4a siRNA; Rab11b, Rab11b siRNA; Myo5b, Myo5b siRNA. recruitment of endogenous Rab4a and Rab11b to the YCV was determined by immunofluorescence using anti-Rab antibodies ( Fig. 5A to F). At 20 min postinfection, a significantly higher number of YCVs colocalized with Rab4a and Rab11b than vacuoles containing E. coli ( Fig. 5C and E; P Յ 0.05 and P Յ 0.01, respectively). By 80 min postinfection, YCVs retained significantly higher colocalization with Rab11b than E. colicontaining vacuoles (P Յ 0.01). However, while YCVs trended toward higher colocalization with Rab4A compared to vacuoles containing E. coli or killed Y. pestis, these differences were not statistically different. To confirm these results, RAW264.7 macrophages were transfected with plasmids expressing Rab4a or Rab11b fused to enhanced green fluorescent protein (EGFP) to monitor the localization of Rab proteins independently of antibodies (58) and 24 h later infected with bacteria. At 20 min postinfection, 50% of PFA-killed Y. pestis and E. coli K-12 containing phagosomes colocalized with Rab4a-EGFP (Fig. 5D). However, as observed in primary macrophages, a significantly higher number of vacuoles containing live Y. pestis colocalized with Rab4a-EGFP (76%; P Յ 0.01). By 80 min postinfection, both PFA-killed Y. pestis and E. coli K-12 decreased in colocalization with Rab4a-EGFP, indicating loss of the GTPase during phagosome maturation, but vacuoles containing live Y. pestis maintained Rab4a-EGFP at a statistically higher frequency (~61%; P Յ 0.01). As the infection continued, colocalization of the YCV with Rab4a-EGFP decreased, and by 2 and 20 h postinfection, it approached background levels (Fig. 5G). In contrast to Rab4a-EGFP, fewer vacuoles containing PFA-killed Y. pestis or E. coli K-12 colocalized with Rab11b-EGFP at 20 and 80 min postinfection (Fig. 5F). However, the majority of YCVs containing live Y. pestis colocalized with Rab11b-EGFP, approaching 85% by 80 min postinfection (P Յ 0.001). Furthermore, Y. pestis continued to colocalize with Rab11b-EGFP at a high frequency throughout the course of the infection (Fig. 5H).
Finally, to determine whether Rab proteins are specifically recruited to phagosomes containing Y. pestis and not to all phagosomes of Y. pestis-infected macrophages, RAW264.7 macrophages transfected with Rab4a-EGFP or Rab11b-EGFP were coinfected with Y. pestis and E. coli K-12. Colocalization between Rab4a-EGFP and bacteriumcontaining vacuoles at 20 min postinfection and between Rab11b-EGFP and bacteriumcontaining vacuoles at 2 h postinfection was determined by confocal microscopy (Fig. 5I to L). In cells infected with both bacteria, Y. pestis-containing vacuoles had a higher frequency of colocalization with both Rab proteins than E. coli K-12-containing vacuoles. Furthermore, the frequency of E. coli K-12 colocalization was not higher during coinfection than observed during single infection with just E. coli K-12 ( Fig. 5D and F). Together, these data indicate that Y. pestis actively recruits Rab4a and Rab11b to the YCV during early stages of macrophage infection and that recruitment is specifically to vacuoles containing live Y. pestis.
Y. pestis infection disrupts host recycling. Recruitment of Rab4a and Rab11b to the YCV indicated that Y. pestis remodels its phagosome to resemble a recycling endosome. Because of these links to host cell recycling, we next tested whether infection with Y. pestis impacts host cell recycling by monitoring recycling of the host transferrin receptor (TfR). RAW264.7 macrophages were infected with Y. pestis CO92, PFA-killed Y. pestis, or E. coli K-12. At 2 and 24 h postinfection, intracellular TfRs were differentially labeled from extracellular TfRs, and intracellular TfR intensity was determined by microscopy ( Fig. 6A and B). Unlike PFA-killed bacteria, infection with live Y. pestis significantly disrupted recycling of TfR as early as 2 h postinfection, resulting in accumulation of intracellular TfR, and continued to impact recycling for 24 h (Fig. 6C to F). Furthermore, intracellular TfR intensity also increased as the Y. pestis multiplicity of infection (MOI) increased. Inhibition of TfR recycling was also specific for macrophages containing intracellular Y. pestis, as uninfected cells from the same cultures did not show elevated retention of TfR (Fig. S5). Importantly, infection with 10-fold-higher numbers of E. coli K-12 did not result in a significant change in TfR retention compared to uninfected macrophages ( Fig. 6G and H), indicating that recycling disruption is not a general response to macrophage activation. Next, we confirmed that disruption of recycling occurs during infection of primary human monocyte-derived macrophages (hMDMs). As observed for RAW264.7 macrophages, infection of hMDMs with Y. pestis resulted in a significant increase in intracellular TfR intensity, while infection with PFA-killed Y. pestis or E. coli K-12 had no significant impact on recycling ( Fig. 6I and J). In contrast to Y. pestis, infection with Salmonella enterica serotype Typhimurium, another intracellular pathogen, had no   Fig. S3]). Importantly, intracellular growth of Y. pestis was not impacted by incubation with TfR antibody (Fig. S4). Together, these data demonstrate that Y. pestis actively disrupts host cell recycling, disruption is not a default response by macrophages to bacteria, and disruption is pathogen specific.
Disruption of host cell recycling is required for Y. pestis replication. Since Rab11b is recruited to and retained on the YCV, we next explored whether Y. pestis infection disrupts host cell recycling through sequestration of Rab11b and depletion of available cellular Rab11b for vesicular trafficking. If this was occurring, then overexpression of Rab11b may be able to restore depleted Rab11b levels and host cell recycling. To test this hypothesis, RAW264.7 macrophages were transfected with a plasmid overexpressing wild-type Rab11b-EGFP (58), infected with Y. pestis, and TfR recycling was monitored. Importantly, overexpression of Rab11b-EGFP did not alter TfR recycling in uninfected cells (Fig. S6). At both 2 and 24 h postinfection, intracellular TfR intensity and TfR-positive endosomes per cell were significantly lower in cells overexpressing Rab11b-EGFP ( Fig. 7A and B), demonstrating that overexpression of Rab11b restored host cell recycling during infection. To determine whether restoration of recycling impacted intracellular survival of Y. pestis, we quantified bacterial numbers as a function of bacterial fluorescence ( Fig. 7C and D). At 2 h postinfection, there were no differences in the bacterial numbers of Rab11b-overexpressing and untransfected cells. However, by 24 h postinfection, bacterial numbers significantly increased in the untransfected cells, while the signal area did not increase in transfected cells (P Յ 0.001). Furthermore, overexpression of Rab4a-EGFP, which is not sequestered to the YCV (Fig. 5G), did not alter bacterial replication ( Fig. 7E and F). Together, these data suggest that Y. pestis infection limits Rab11b availability through sequestration to the YCV, resulting in disruption of host cell recycling. Furthermore, disruption of recycling is required in order for Y. pestis to replicate in macrophages.

DISCUSSION
While it has been known for decades that Y. pestis survives within a vacuolar compartment within macrophages (27,28), the mechanisms leading to subversion of phagolysosome killing by macrophages have not been defined. To better understand the processes involved in the biogenesis of the protective YCV by Y. pestis, we conducted an RNAi genome-wide screen that identified 71 host proteins necessary for survival of Y. pestis inside macrophages. Bioinformatic analysis showed enrichment for three key cellular processes: vesicular trafficking, vesicular transport, and vesicular localization. Refining the interactome generated from this screen suggested that the host endocytic recycling pathway is key for Y. pestis to survive in macrophages. Expanding on these findings, we demonstrated for the first time that Y. pestis remodels the YCV by recruiting endocytic recycling compartment (ERC) markers Rab4a and Rab11b. Importantly, trafficking of endosomes to the ERC, which is mediated by Rab4 and Rab11, is thought to prevent cargo within these compartments from entry into degradative compartments such as phagolysosomes (53). Therefore, by remodeling the YCV, Y. pestis may take advantage of this normal, nondegradative pathway to avoid lysosomal fusion and escape killing by the macrophage. A model summarizing additions to the YCV biogenesis process based on these data is outlined in Fig. 8.
While Rab4a and Rab11b are recruited to the YCV, RNAi of these two genes resulted in very distinct phenotypes. Knockdown of Rab4a resulted in rapid killing of Y. pestis within 2 h of infection, while changes in bacterial survival in Rab11b siRNA-treated macrophages were not apparent until later. These findings suggest that Rab4a is required to avoid early steps in phagosome maturation, while Rab11b interactions contribute to later stages in YCV biogenesis. Supporting this hypothesis, we observed that Rab4a is required to avoid phagosome acidification and lysosomal fusion, which begins within 20 min of phagocytosis. We have also shown that Rab1b is also required to avoid acidification (35), highlighting that avoidance of YCV acidification is a key step in the intracellular survival of Y. pestis. Future studies to better understand the kinetics of Rab1b, Rab4a, and Rab11b recruitment and retention will help understand the dynamics of these early steps in YCV biogenesis and whether these proteins are dependent on one another for stepwise recruitment. Importantly, direct recruitment of Rab1b and Rab4a to the YCV in order to inhibit acidification could also explain why live Y. pestis did not inhibit the acidification of phagosomes containing PFA-killed Y. pestis in coinfected macrophages previously reported by Pujol et al. (27). This is further supported by our data here that live bacteria recruit Rab4a and Rab11b only to the YCV they are contained within (Fig. 5K and L). Furthermore, these data also suggest the bacterial factors that mediate Rab recruitment are likely located at the YCV and not distributed throughout the cell. In contrast to Rab1b and Rab4a, knockdown of Rab11b had only a minor impact on YCV acidification, with the majority of YCVs still avoiding acidification (58% versus 22% for Rab4A RNAi at 80 min). This change in YCV acidification frequency correlated with increased bacterial survival at 2 h postinfection in Rab11b siRNA-treated cells (Fig. 3C). However, we observed significantly lower bacterial numbers in Rab11b siRNA-treated macrophages at 10 h postinfection. Notably, while the number of Y. pestis in Rab11b siRNA-treated macrophages decreased slightly between 2 and 10 h postinfection, bacteria were beginning to replicate in the scrambled-siRNA-treated cells (Fig. 3F). These data indicate that Rab11b knockdown may restrict intracellular replication of Y. pestis as opposed to survival. Anaplasma phagocytophilum (59) and Chlamydia species (60) have also been shown to recruit Rab4 and Rab11 to their vacuoles, and knockdown of Rab11 in cells infected with Chlamydia trachomatis (61) and Coxiella burnetii (39) inhibits bacterial replication of these species. During infection, these three pathogens develop large vacuoles, which require acquisition of vesicular membranes. It has been suggested that interception of Rab11-positive recycling endosomes could provide both membranes for vacuole expansion and nutrients for bacterial replication (59,61). The YCV also expands into a spacious vacuole after~8 h of infection, though vacuoles do not expand in size to the degree observed for the aforementioned pathogens (16,27,28). This time frame coincides with the beginning of intracellular replication of Y. pestis, suggesting that Y. pestis may also intercept recycling endosomes for similar purposes. During these studies, we did not observe evidence for the formation of spacious YCVs in Rab11b siRNA-treated macrophages, supporting that hijacking of recycling endosomes may be occurring during Y. pestis infection and suggesting convergent evolution by multiple intracellular pathogens. However, if direct interception of recycling endosomes is the only function of Rab11b targeting during Y. pestis infection, we would not expect that overexpression of Rab11b would inhibit bacterial growth, as the YCV still acquires Rab11b in overexpressed cells and should still be able to intercept recycling endosomes. Importantly, while Huang et al. and Rzomp et al. used overexpression of Rab-GFPs to localize the Rab4 and Rab11 to the Anaplasma-and Chlamydia-containing vacuoles, they did not report inhibition in bacterial growth during overexpression of Rab11 (59,60). These data indicate that inhibition of intracellular replication is not a general artifact of Rab11 overexpression, but perhaps a pathogen-specific phenotype. Therefore, while our current data do not exclude interception of recycling endosomes through Rab11b, our overexpression data suggest that a novel mechanism used by Y. pestis to manipulate the biology of the . Y. pestis engages the host endosome recycling pathway by recruiting Rab GTPases to the YCV in order to generate a protective replicative niche in a two-step process. First, Rab1b and Rab4a are recruited to the YCV, which is required for the bacterium to inhibit phagosome acidification and fusion with the lysosome. While these two Rab proteins are eventually lost from the YCV, Rab11b is retained on the YCV over the entire course of infection. Retention of Rab11b leads to a global inhibition of host recycling, which is required for Y. pestis to replicate in macrophages. Rab4, Rab11, and Rab1b proteins are shown in the figure as gray circles labeled 4, 11, and 1B, respectively. macrophage, which is yet to be defined, contributes to intracellular replication of Y. pestis.
We also demonstrated for the first time that macrophage recycling is disrupted during Y. pestis infection. This disruption does not appear to be a general response of macrophages to encountering bacteria, as infection with PFA-killed Y. pestis, E. coli K-12, or S. enterica Typhimurium did not disrupt recycling in macrophages. While other bacterial pathogens have been shown to recruit recycling markers to their vacuole, to our knowledge, this is the first description of a bacterial infection disrupting global endosome recycling during infection. Disruption of host recycling by Y. pestis could alter several different aspects of macrophage biology that could impact infection. For example, Longatti et al. have shown that disruption of recycling inhibits starvationinduced autophagy and autophagy induction is dependent on active Rab11 (62,63). More recently, Szatmári et al. demonstrated that Rab11b interacts with Hook, a negative regulator of endosome maturation, to facilitate cross talk between recycling endosomes and induction of autophagy (64). These links between Rab11, recycling, and autophagy could be important in the context of Y. pestis infection. Several studies have shown that both Y. pestis and Yersinia pseudotuberculosis induce autophagy during infection, and eventually YCVs take on characteristics of autophagosomes (65). Interestingly, autophagy does not seem to be detrimental to Y. pestis (27), but it may be required for bacterial replication (66). Furthermore, Y. pestis does not appear to escape the YCV, raising questions of why is autophagy triggered and how is the YCV targeted for autophagosome formation. Our discovery that Y. pestis interacts with Rab11b and the host recycling pathway provide potential answers to these questions. For example, disruption of endocytic recycling may induce conditions/signals similar to starvation that trigger autophagy pathways in the macrophage. Additionally, sequestration of Rab11b by Y. pestis may trigger autophagy nucleation at the YCV, resulting in autophagosome formation. Therefore, the consequence of inhibiting host cell recycling by Y. pestis may be to specifically induce autophagy to promote replication. Future studies to address these hypotheses are ongoing and important to define the contribution of Rab11b sequestration and recycling disruption on the pathogenesis of Y. pestis.
In summary, we have demonstrated for the first time that Y. pestis remodels the YCV to resemble a recycling endocytic vacuole in a T3SS-independent manner, and the bacterium disrupts host recycling during infection. These findings suggest that interactions with the recycling pathway are important for multiple steps in the YCV biogenesis process. We also showed that overexpression of Rab11b overcomes the ability of Y. pestis to disrupt recycling and this prevents bacterial replication, suggesting that recruitment of Rab11b to the YCV has trans-acting effects on the cell that are beneficial to the bacterium. Future studies to further define the impact disruption of host cell recycling has on the biology of the macrophage will be important to understand how Y. pestis avoids killing by these phagocytes.

MATERIALS AND METHODS
Eukaryotic cells, bacterial strains, and plasmids. RAW264.7 macrophages were obtained from ATCC and cultured in Dulbecco modified Eagle medium (DMEM) containing 100 mM glucose plus 10% fetal bovine serum (FBS) (HyClone). Peritoneal macrophages were isolated from C57BL/6 mice as previously described (67). Human monocyte-derived macrophages (hMDMs) were isolated with minor modifications as previously described (68)(69)(70). Briefly, hMDMs were isolated from peripheral blood samples from healthy, antibiotic-free adult donors (institutional review board [IRB] protocol 04.0358) using a Ficoll-Hypaque gradient. Monocytes were suspended in RPMI 1640 plus 20% FBS (Biowest), then aliquoted into six-well ultralow attachment plates (Greiner Bio One), and incubated for 3 days at 37°C and 5% CO 2 . On day 4, nonadherent cells were aspirated, monolayers were scraped and resuspended in RPMI 1640 plus 10% FBS, and 2.5 ϫ 10 5 cells/well were transferred into the wells of a 24-well tissue culture plate (Greiner Bio One). On days 6 and 7, the medium was removed, and replaced with RPMI 1640 plus 5% FBS and RPMI 1640 plus 1% FBS, respectively. hMDMs were used for studies on day 8.
RNAi primary and validation screens. RAW264.7 macrophages were forward transfected with small interfering RNAs (siRNAs) from the Silencer siRNA mouse genome library v3 (Ambion). The library contains three siRNAs targeting each gene, which were pooled for the primary screen. siRNAs were suspended in 20 l Opti-MEM at a final concentration of 1 M and then mixed with 10 l of 0.03% (vol/vol) Lipofectamine RNAiMax/Opti-MEM and added to each well of a white flat-bottom 96-well plate (Greiner Bio One). For each plate, column 12 included wells containing scrambled siRNA (negative control; n ϭ 3) and Cop␤1 siRNA (positive control; n ϭ 3) as controls for transfection efficiency and plate-to-plate variation. The plates were incubated at room temperature for 10 min, and then 1 ϫ 10 4 RAW264.7 macrophages suspended in 80 l of DMEM plus 10% FBS (HyClone) were added. The cells were incubated for 48 h and then infected with Y. pestis CO92 Lux PtolC pCD1 (-) (multiplicity of infection [MOI] of 10) and synchronized by centrifugation (200 ϫ g) for 5 min. Twenty minutes after infection, extracellular bacteria were killed with gentamicin as described above. Intracellular bacteria were quantified as a function of bioluminescence at 20 min and 2 h and 10 h postinfection using a Synergy 4 plate reader (BioTek; 1-s read with sensitivity set at 150). After the 10-h bioluminescent read, cell viability was determined using alamarBlue (Life Technologies). Briefly, 10 l alamarBlue was added directly to each well and incubated at 37°C and 5% CO 2 for 2 h, and fluorescence was determined using a Synergy 4 plate reader (excitation wavelength, 560 nm; emission wavelength, 600 nm) and compared to the average of the scrambled-siRNA control wells. For each plate, a Z factor (Z=) was calculated from the control wells using the formula: 1 Ϫ (3ϫ (SD Cop␤1 RLU -SD scrambled RLU)/(AVG scrambled RLU -AVG Cop␤1 RLU)) where SD is the standard deviation, AVG is the average. Plates with Z= of Ͻ0.3 were repeated. Intracellular survival was normalized for each plate using the following formula: (siRNA RLU/AVG Cop␤1 RLU)/(AVG scrambled RLU/AVG Cop␤1 RLU). Primary screen selection criteria was set at Ն60% inhibition of Y. pestis survival and Յ50% cytotoxicity. Primary hits were validated using two single siRNAs in independent validation screens. Selection criteria for the validation screen were Ն40% inhibition of Y. pestis survival and Յ50% cytotoxicity. Bioinformatic analysis. Validated and primary screen hits were stored with both Entrez Gene and MGI identifiers. Interacting partners were identified using all experimental evidence codes from BioGRID (75) and STRING (76) databases using MGI and Entrez Gene identifiers. Interactors for the input data sets were validated and primary hits were characterized by (i) direct interactions within each individual data set and (ii) direct interactions from validated to primary hits. These interactions were stored and imported into Cytoscape (v3.30) to generate interaction maps (77). For Gene Ontology (GO) clustering, the validated hits were imported by Entrez Gene identifier to Cytoscape plug-ins ClueGO (78) and CluePedia (79). Genes were clustered using all GO evidence codes, a minimum kappa score of 0.4, and a P value threshold of 0.05.
Differential staining of intracellular TfR. To monitor TfR recycling, macrophages were infected with bacteria as described above in the presence of 1 g/ml of anti-TfR antibody, which was maintained throughout the course of infection. At 2 and 24 h postinfection, the cells were gently washed three times with ice-cold stripping buffer (Hanks buffered salt solution [HBSS] with 50 mM glycine, 150 mM NaCl, and 0.2% bovine serum albumin [BSA] [pH 4]) to remove anti-TfR antibody bound to the cell surface (extracellular receptors) and nonspecific binding antibody. The cells were then fixed with 2.5% PFA for 15 min at room temperature and incubated with permeabilization buffer (0.5% Tween 20, 3% BSA) overnight at 4°C. For indirect immunofluorescence labeling of internal TfR, cells were incubated for 1 h with anti-rabbit antibody labeled with Alexa Fluor 488 diluted in permeabilization buffer. TfR intensity per cell was calculated as follows: (TfR signal ϫ number of endosomes)/number of nuclei.
Statistics. All experiments were repeated three times to ensure reproducibility. Unless noted, data are shown as the means Ϯ standard deviations (SDs) from three independent experiments. For microscopy, each experiment analyzed at least 50 YCVs, 50 individual cells, or 25 fields, and power analyses were performed posthoc to ensure that appropriate sample sizes were analyzed. P values were calculated using Student's t test or one-way analysis of variance (ANOVA), with appropriate posthoc testing, using GraphPad Prism software.