Epistatic Interplay between Type IV Secretion Effectors Engages the Small GTPase Rab2 in the Brucella Intracellular Cycle.

Bacterial pathogens with an intracellular lifestyle modulate many host cellular processes to promote their infectious cycle. They do so by delivering effector proteins into host cells via dedicated secretion systems that target specific host functions. While the roles of many individual effectors are known, how their modes of action are coordinated is rarely understood. Here, we show that the zoonotic bacterium Brucella abortus delivers the BspB effector that mitigates the negative effect on bacterial replication that the RicA effector exerts via modulation of the host small GTPase Rab2. These findings provide an example of functional integration between bacterial effectors that promotes proliferation of pathogens.

these pathogens to exploit specific host functions and yet do not encompass how the delivered effectors combine their modes of action to support the bacterium's intracellular cycle. Instances of cooperative or antagonistic interactions between effectors have highlighted the importance of their combined functions in modulating cellular pathways. These include effectors in Legionella pneumophila that directly or indirectly modulate the activity of other effectors to ensure spatiotemporal regulation of intracellular events (1)(2)(3)(4)(5)(6). Epistatic, antagonistic actions between effectors that indirectly compensate each other's activities on specific intracellular stages have also been characterized in Salmonella enterica and L. pneumophila (7,8).
The zoonotic bacterium Brucella abortus survives and proliferates within host phagocytes by generating a replication-permissive organelle derived from the host endoplasmic reticulum (ER), the replicative Brucella-containing vacuole (rBCV), via remodeling of its original endosomal BCV (eBCV) (9). Conversion of eBCV to rBCV is driven by the Brucella VirB type IV secretion system (T4SS) (10)(11)(12)(13), which delivers effector proteins (14)(15)(16)(17)(18)(19) that are thought to remodel BCV trafficking and promote bacterial replication. rBCV biogenesis and bacterial replication require functions of the host early secretory pathway, including vesicular transport steps controlled by the host small GTPases Sar1, Rab1, Rab2, and Arf1 (20)(21)(22). Although little is known about the roles and modes of actions of Brucella type IV effectors, recent studies have identified some that may target host secretory functions. These include BspA, BspB, and BspF, which impair host secretory trafficking and contribute to bacterial replication (17), and RicA, which binds Rab2 and downmodulates Brucella replication in HeLa cells (14). Among these, BspB alters secretory traffic via its interaction with the Golgi apparatus-associated conserved oligomeric Golgi (COG) complex, a regulator of vesicular traffic at, and within, the Golgi apparatus (23) and causes redirection of COG-dependent vesicular traffic to promote rBCV biogenesis and bacterial replication (22). These findings argue that Brucella delivers a series of effector proteins that modulate specific host secretory functions to mediate rBCV biogenesis and possibly bacterial replication within rBCVs. Whether, and how, Brucella T4SS effectors coordinate to integrate their respective modes of action toward promoting the bacterium's intracellular cycle is unknown.
RicA binds the GDP-bound inactive form of Rab2 in vitro and promotes Rab2 recruitment to the BCV (14), suggesting that it modulates Rab2-dependent transport as part of its mode of action. How RicA modulates Rab2 functions and whether it interferes with Rab2-dependent secretory transport remain unknown. Interestingly, the replication defect of a ΔbspB mutant is restored by depletion of Rab2a in macrophages (22), suggesting a functional link between BspB and Rab2-dependent transport and possibly RicA. We therefore examined the relationship between these two T4SS effectors during Brucella infection. Here, we reveal a functional epistatic interplay between these effectors, where BspB and RicA engage Rab2-dependent functions in the Brucella intracellular cycle and BspB compensates RicA's deleterious activity on rBCV biogenesis and Brucella replication. These findings uncover how Brucella T4SS effectors finely tune their actions to modulate host secretory transport for bacterial replication purposes.

RESULTS
BspB is dispensable for rBCV biogenesis and bacterial replication in the absence of RicA. To determine whether BspB and RicA functionally interact, we first used a genetic approach to compare RicA's role in rBCV biogenesis and bacterial replication to that of BspB in primary murine macrophages. As previously described (22), ΔbspB bacteria underwent delayed rBCV biogenesis compared to wild-type bacteria, as visualized by a slower exclusion of LAMP1-positive endosomal membranes from BCVs between 8 and 12 h postinfection (pi) (Fig. 1A) and a replication defect at 24 h pi ( Fig. 1B and C). In contrast, ΔricA bacteria displayed rBCV biogenesis kinetics and replication similar to those of the wild-type bacteria (Fig. 1), indicating that RicA is not essential for rBCV biogenesis and intracellular replication of Brucella in primary macrophages. Similar replication results were obtained in HeLa cells (see Fig. S1A in the supplemental material), except that BspB was also dispensable for bacterial replication, ruling out this infection model as appropriate to study BspB-RicA interactions. Interestingly, the deletion of ricA in ΔbspB bacteria suppressed the delayed rBCV biogenesis and bacterial replication defect caused by deletion of bspB seen in bone marrowderived macrophages (BMMs), indicating that BspB is dispensable for these intracellular events in the absence of RicA. This suppressive effect was due to ricA deletion, as it was complemented by reintroduction of a chromosomal copy of ricA in trans (Fig. 1). The suppressive effect of ricA deletion was specific to BspB-dependent replication, as it failed to restore normal growth of a replication-deficient strain that lacks BspF (strain ΔbspF; Fig. S1B), another T4SS effector interfering with host secretion (17). Hence, these results demonstrate a genetic link between BspB and RicA functions and suggest that BMMs infected with DsRed m -expressing wild-type (strain 2308), ΔbspB, ΔricA, ΔbspB ΔricA, or ΔbspB ΔricA::ricA B. abortus strains, as measured by acquisition and then exclusion of LAMP1-positive membranes. Data are means Ϯ SD of results from 3 independent experiments. Asterisks indicate a statistically significant difference between wild-type (2308) and ΔbspB (red) or ΔbspB ΔricA::ricA (purple) bacteria, determined by two-way ANOVA with Dunnett's multiple-comparison test. *, P ϭ Ͻ0.05. (B) Intracellular replication of DsRed m -expressing wild-type (2308), ΔbspB, ΔricA, ΔbspB ΔricA, or ΔbspB ΔricA::ricA B. abortus strains in BMMs at 24 h pi. Values are means Ϯ SD of results from at least 3 independent experiments. Asterisks indicate a statistically significant difference compared to either the wild-type (2308; green) or ΔbspB (red) populations, assessed using a nonparametric Kruskal-Wallis test with Dunn's multicomparison statistical analysis. ns, not significant. (C) Representative confocal fluorescence micrographs of BMMs infected with DsRed m -expressing wild-type (2308), ΔbspB, ΔricA, ΔbspB ΔricA, or ΔbspB ΔricA::ricA B. abortus strains (pseudocolored in white) for 24 h pi. Nuclei were stained using Hoechst 33342 (blue). Scale bar, 20 m.

Epistasis of Brucella Type IV Effectors
® BspB may counteract RicA activities that interfere with rBCV biogenesis and Brucella replication.
RicA and BspB engage Rab2-dependent functions in Brucella replication. Given the role of Rab2 in Brucella replication (21,22) and RicA binding to Rab2 (14) and considering that BspB targets the COG complex, which orchestrates Rab2-dependent Golgi apparatus-ER vesicular transport, we next examined the effect of Rab2-dependent transport inhibition on the intracellular behavior of BspB-deficient and RicA-deficient bacteria in BMMs. In contrast with a previous study performed in HeLa cells showing a role of Rab2 in rBCV biogenesis (21), small interfering RNA (siRNA)-mediated depletion of Rab2a in BMMs (86.0% Ϯ 12.7% depletion; Fig. 2A) did not affect rBCV biogenesis by wild-type B. abortus (Fig. 2B), despite efficiently inhibiting Golgi apparatus-ER retrograde transport (22). However, Rab2 depletion (89.0% Ϯ 2.4%) significantly decreased bacterial replication (Fig. 2C), as previously reported (22). Hence, Rab2a appears to be dispensable for rBCV biogenesis in BMMs but required for Brucella replication once rBCVs are formed. The absence of RicA did not affect rBCV biogenesis in either small interfering nontargeting (siNT)-treated or siRab2a-treated BMMs (Fig. 2B) and yet rendered Brucella replication Rab2a independent (Fig. 2C), suggesting that RicA involves Rab2a functions in bacterial replication. Interestingly, Rab2a depletion suppressed the BCV maturation defect of ΔbspB bacteria and caused this otherwise replication-deficient mutant to grow to wild-type levels by 24 h pi ( Fig. 2B and C). Deletion of ricA in the ΔbspB mutant also suppressed BspB's requirement for rBCV biogenesis and replication and rendered replication of this strain Rab2a independent ( Fig. 2B and C). This phenotype was complemented by the chromosomal reintroduction of ricA in the ΔbspB ΔricA mutant ( Fig. 2B and C). Taken collectively, we interpret these findings as showing that RicA involves Rab2a-dependent functions toward rBCV biogenesis and bacterial replication and that BspB counteracts RicA activities on Rab2adependent functions that are otherwise deleterious to proper rBCV biogenesis and subsequent replication.
RicA alters host secretory traffic and Golgi apparatus morphology. RicA was identified as binding Rab2 in vitro (14), which regulates Golgi apparatus-ER retrograde traffic (24)(25)(26), and yet whether RicA activity impairs Rab2-dependent secretory transport has not been established. We first verified the occurrence of RicA-Rab2 interactions in mammalian cells, by coimmunoprecipitation of ectopically expressed myc-tagged RicA with wild-type, dominant-negative (Rab2 S20N ) or constitutively active (Rab2 Q65L ) alleles of green fluorescent protein (GFP)-tagged Rab2 in HeLa cells (Fig. S2), consistent with RicA binding to active or inactive Rab2 in vivo. Given that BspB interferes with secretory traffic (17,22), we then tested the effect of RicA on constitutive host secretion. Expression of myc-tagged RicA in HeLa cells significantly inhibited secretion of a secreted alkaline phosphatase (SEAP) secretory reporter, although not as drastically as expression of hemagglutinin-BspB (HA-BspB) ( Fig. 3A) (22), indicating that RicA functionally interferes with the secretory pathway. Coexpression of both myc-RicA and HA-BspB caused SEAP secretion inhibition levels similar to those induced by HA-BspB alone (Fig. 3A), indicating that BspB exerts an inhibitory effect stronger than that exerted by RicA. Since impairment of Rab2 functions causes fragmentation of the Golgi apparatus (27), we next tested the effect of RicA expression of Golgi apparatus morphology. While retroviral expression of HA-BspB in BMMs did not induce any Golgi morphological changes, that of GFP-RicA caused Golgi fragmentation similar to that observed upon expression of dominant-negative GFP-Rab2a S20N ( Fig. 3B and C), indicating that RicA expression phenocopies Golgi morphological changes caused by impairment of Rab2a functions. RicA-induced Golgi fragmentation was GFP or cell type independent, as myc-RicA also induced Golgi fragmentation in both BMMs and HeLa cells (Fig. S3). Coexpression of HA-BspB with GFP-RicA in BMMs did not alter the RicA-dependent Golgi fragmentation phenotype ( Fig. 3B and C). Taken together, these results show that ectopically expressed RicA interacts with Rab2 and functionally impairs the secretory pathway and yet does so independently of BspB.
BspB-mediated alterations of ER-to-Golgi transport are RicA and Rab2a independent. To gain a better understanding of the BspB-RicA functional interactions that we uncovered genetically ( Fig. 1 and 2), we examined the role of RicA in BspBdependent changes in secretory transport. BspB causes a redistribution of the ER-to-Golgi intermediate compartment (ERGIC) cargo receptor p58/ERGIC53 to the Golgi apparatus, as a result of altered ER-Golgi secretory transport during infection of BMMs (22). We therefore examined the effect of ricA deletion and Rab2a depletion on BspB-mediated p58 redistribution in BMMs. As previously shown (22), Brucella infection of BMMs induced p58 redistribution to the Golgi apparatus in a BspB-dependent Asterisks indicate a statistically significant difference compared to siNT-treated BMMs infected with either wild-type (2308; green) or ΔbspB (red) strains, assessed using a nonparametric Kruskal-Wallis test with Dunn's multicomparison statistical analysis; ns, not significant.
Epistasis of Brucella Type IV Effectors ® manner ( Fig. 4A and C). Rab2a depletion (93.2% Ϯ 3.0% depletion; Fig. 4B) did not impact these changes (Fig. 4C), indicating that BspB-mediated redistribution of p58 is Rab2a independent. Deletion of ricA in either wild-type or ΔbspB bacteria did not affect p58 distribution either, compared to the respective parental strains (Fig. 4C). Consistently, retroviral expression in BMMs of HA-BspB, but not of myc-RicA, caused p58 redistribution similar to that seen during infection ( Fig. 4D and E). Altogether, these results demonstrate that BspB-mediated changes in ER-Golgi transport are RicA and Rab2 independent and are likely unrelated to BspB-RicA genetic interplay.
BspB and RicA modulate Golgi apparatus-associated vesicular traffic in a Rab2a-dependent manner. In addition to its effect on ER-Golgi apparatus traffic, BspB remodels COG-dependent, Golgi apparatus-derived vesicular transport during Brucella infection, as measured by the intracellular redistribution of the Golgi SNARE GS15 in BMMs (22). We therefore tested whether RicA and Rab2a play a role in this process, by examining the redistribution of the Golgi SNARE GS15 in control or Rab2a-depleted Epistasis of Brucella Type IV Effectors ® BMMs. Infection with wild-type bacteria caused a redistribution of GFP-GS15 in BMMs, as previously shown (22), which was partially dependent on Rab2a since its depletion (99.4% depletion; Fig. 5B) decreased GS15 redistribution induced by wild-type bacteria (Fig. 5A to C). Additionally, both ΔbspB bacteria and ΔricA bacteria failed to cause GS15 redistribution ( Fig. 5A and C) in control (siNT) BMMs, and these defects were genetically complemented in infections with ΔbspB::bspB and ΔricA::ricA bacteria (Fig. S4). This indicates that both BspB and RicA activities contribute to remodeling of Golgi apparatus-derived vesicular traffic. Consistently, the individual or combined forms of retroviral expression of HA-BspB and myc-RicA in BMMs were sufficient to cause GS15 redistribution, confirming the effect of these effectors on Golgi apparatus-associated vesicular traffic (Fig. 5D and E). Interestingly, Rab2a depletion did not impact GS15 redistribution by the ΔbspB mutant and yet fully suppressed the redistribution defect caused by the ΔricA mutation, indicating that the activity of RicA, but not that of BspB, in Golgi apparatus-derived vesicular traffic is Rab2a dependent. Furthermore, ΔbspB ΔricA bacteria caused GFP-GS15 redistribution similar to that seen with wild-type bacteria (Fig. 5C), indicating that these mutations exert suppressive effects on each other, as seen with rBCV biogenesis and bacterial replication ( Fig. 1 and 2). Taken together, these findings indicate that RicA and BspB comodulate Golgi apparatusderived vesicular transport in a compensatory manner and that RicA, but not BspB, engages Rab2a functions in this process.

DISCUSSION
Although the role of the Brucella VirB T4SS in rBCV biogenesis and bacterial replication has been long known (10)(11)(12)(13), only a few VirB effectors dedicated to these processes have been identified (14,17,18,22) and even fewer have been characterized (22), limiting our understanding of the bacterium's molecular strategies of intracellular proliferation. Here, we show that BspB and RicA, two effectors previously associated with rBCV biogenesis and bacterial replication (14,17,22), comodulate Golgi apparatusassociated vesicular transport and engage Rab2-dependent functions in Brucella intracellular replication. These findings highlight another level of complexity beyond the individual functions of effectors, whereby the integration of their modes of action finely tunes how the bacterium modulates specific cellular processes.
The concept of interplay between bacterial effectors has been proposed in the context of spatiotemporal regulation of their functions. Several "meta-effectors" have been characterized in L. pneumophila and Salmonella enterica which either regulate the activity of another effector (1,2,6) or exert their effect on the activity of another effector via compensatory modification of the same host target (3)(4)(5) or via compensatory activity in the same cellular processes (7,8,28). Our results argue that the functional interaction between BspB and RicA fits the scenario of compensatory effects of effectors on the same cellular processes. We previously reported that deletion of bspB impairs rBCV biogenesis and optimal bacterial replication in macrophages (22), suggesting a direct involvement of BspB in these stages of the bacterium's infectious cycle. And yet the additional deletion of ricA suppresses these phenotypic defects, indicating that BspB's role in these processes strictly depends upon RicA function, which consequently appears deleterious to rBCV biogenesis and bacterial replication in the absence of BspB. Deletion of ricA alone did not confer rBCV biogenesis or replication defects under our experimental conditions, which we interpret as BspB masking negative effects of RicA on rBCV biogenesis and bacterial replication and therefore exerting an epistatic effect on RicA activity. This model of epistatic interplay is supported by the restoration of normal rBCV biogenesis and bacterial replication upon deletion of both ricA and bspB, which is reminiscent of the genetic and functional interactions between the type III secretion effectors SifA and SseJ in Salmonella enterica and the T4SS effectors SdhA and PlaA in Legionella pneumophila. In these instances, SifA and SdhA promote bacterium-containing vacuole integrity by compensating for the vacuole-disrupting activities of SseJ and PlaA, respectively, and inactivation mediated by the combined effects of both partners in each pair of effectors restores vacuolar integrity (7,8). In this context, we propose that the combined activities of RicA and BspB contribute to optimal rBCV biogenesis and replication of Brucella.
A role for Rab2 in both rBCV biogenesis and bacterial replication was originally uncovered in HeLa cells (21), and we reported its requirement in macrophages previously (22) and in this study but only for optimal bacterial replication of wild-type bacteria. This discrepancy was unlikely to have resulted from an incomplete inactivation of Rab2a functions under our experimental conditions, as we previously showed that siRNA-mediated depletion of Rab2a inhibited Rab2-dependent secretory transport (22). The discovery that RicA is a Rab2-binding effector that negatively modulates rBCV biogenesis and Brucella intracellular growth in HeLa cells (14) supports a model whereby RicA negatively affects Rab2 functions that are otherwise required for proper rBCV biogenesis and bacterial replication. Under this assumption, inactivation of Rab2 should phenocopy RicA function. We found instead that Rab2a depletion mimicked RicA deficiency in suppressing rBCV biogenesis and replication defects caused by BspB inactivation, arguing that RicA positively engages Rab2-dependent functions and that BspB activity counteracts RicA's effect on Rab2-mediated transport to promote rBCV biogenesis and bacterial replication. These findings are consistent with a model in which BspB's function in rBCV biogenesis and bacterial replication depends upon RicA's effects on Rab2a-dependent transport. We therefore propose that BspB-RicA functional interplay consists of BspB-mediated compensation for RicA-induced activation of Rab2 functions (Fig. 6). This model predicts that Rab2-dependent functions are dispensable for rBCV biogenesis and bacterial replication. In agreement, we found that Rab2 depletion does not affect rBCV biogenesis by wild-type or RicA-deficient bacteria, only suppressing defects caused by BspB inactivation. However, inactivation of Rab2a functions reduced replication of wild-type bacteria (this study and reference 21) but not that of RicA-or RicA/BspB-deficient bacteria, indicating a requirement for Rab2a functions in replication only in the presence of RicA. This is consistent with effects of RicA activity on Rab2a functions being deleterious to replication in the presence of BspB but not in its absence, further arguing for a combined role of these two effectors during bacterial replication. Alternatively, Rab2a depletion or inactivation may cause indirect effects on secretory transport that are deleterious to Brucella intracellular growth, so further studies are required to test this hypothesis.
Our previous studies revealed that BspB alters ER-to-Golgi transport, affecting steady-state localization of ERGIC cargo receptor p58, and also redirects COGdependent Golgi apparatus-derived vesicular traffic (22). While RicA and Rab2a were not involved in BspB-mediated changes in ER-to-Golgi transport, both contributed to changes in Golgi apparatus-associated traffic, with RicA's requirement abolished by Rab2a depletion. Similarly to their epistatic interaction in rBCV biogenesis and bacterial replication, ricA deletion suppressed the phenotypic defects in GS15 redistribution caused by bspB deletion, consistent with counteracting effects of these effectors on this trafficking pathway. These findings highlight this specific stage in secretory transport as a target of BspB and RicA, but whether this common targeting accounts for the functional interplay between these two effectors in rBCV biogenesis and bacterial replication is unknown. The molecular basis of the BspB-RicA functional interaction remains to be understood. BspB interacts with, and alters functions of, the COG complex (22), which coordinates Rab2-dependent vesicular transport from the Golgi apparatus, functionally linking its host target to a Rab2-dependent process. The mode of action explaining the effect of RicA on Rab2 remains elusive. While it is required for Rab2 recruitment on BCVs in HeLa cells (14), whether and how it modulates Rab2 activity are unknown. While a previous study observed a binding preference for GDP-bound, inactive Rab2 in vitro (14), our functional data suggest a positive effect of RicA on Rab2 function. Additional work is required to determine whether RicA inhibits or activates Rab2 function during infection. Defining how BspB modulation of COG functions antagonizes RicA's effect on Rab2 will reveal the molecular basis of their

MATERIALS AND METHODS
Brucella strains. Brucella abortus strains 2308, 2308 ΔbspB, 2308 ΔbspB::bspB, and 2308 ΔbspF have been described previously (17,20). To make an in-frame deletion of ricA, 1,000-bp regions of the Brucella genome upstream and downstream of ricA (BAB1_1279) were amplified using primers pairs WSU0345/ WSU0346 and WSU0347/WSU0348, respectively. The two resulting fragments were joined by overlapextension PCR and inserted into pJC80 (20) using XbaI and SacI restriction sites and sequenced to confirm ricA in-frame deletion. Strains 2308 ΔricA, 2308 ΔbspB ΔricA, and 2308 ΔbspF ΔricA were constructed by electroporation of pJC80 ΔricA into B. abortus 2308, B. abortus 2308 ΔbspB, and B. abortus 2308 ΔbspF strains using SacB-assisted allelic-replacement-based selection as described previously (20). Deletion was confirmed by PCR using primers WSU0345 and WSU0348. Brucella strains were modified to express DsRed m through chromosomal insertion of miniTn7K-dsRed at the attTn7 locus through electroporation of Brucella strains with pUC18T-miniTn7K-dsRed (29) and the corresponding helper plasmid pUC18T-Tn7-tnp, as described previously (17). For genetic complementation of B. abortus 2308 ΔbspB ΔricA, ricA preceded by a 489-bp promoter region was amplified from B. abortus genomic DNA using primers WSU0391 and WSU0410 and cloned into pUC18T-miniTn7K-dsRed using HindIII and KpnI restriction sites. For genetic complementation of B. abortus 2308 ΔbspF, bspF preceded by a 340-bp promoter region was amplified from pUC18T-miniTn7K-bspF (17) using primers WSU0218 and WSU0219 (Table 1) and cloned into pUC18T-miniTn7K-dsRed using EcoRI and KpnI restriction sites. The resulting pUC18T-miniTn7K-dsRed-ricA and pUC18T-miniTn7K-dsRed-bspF were electroporated into Brucella as described above. All Brucella strains were grown on tryptic soy agar (TSA) (BD Difco) for 3 days at 37°C and 5% CO 2 and in tryptic soy broth (TSB) (BD Difco) at 37°C with shaking until cultures reached an optical density at 600 nm (OD 600 ) of ϳ1.0 for BMM infections. All experiments with B. abortus strains were performed in a biosafety level 3 facility following CDC Division of Select Agents and Toxins regulations and in compliance with standard operating procedures approved by the Washington State University Institutional Biosafety Committee. For cloning, Escherichia coli strain DH5␣ (Invitrogen) was grown in Luria-Bertani broth or on Luria-Bertani agar (BD Difco) at 37°C and supplemented with either 50 g/ml kanamycin or 100 g/ml ampicillin (Fisher BioReagents) when required.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.