Manipulation of IRE1-Dependent MAPK Signaling by a Vibrio Agonist-Antagonist Effector Pair

Vibrio parahaemolyticus is a seafood-borne pathogen that encodes two type 3 secretion systems (T3SS). The first system, T3SS1, is thought to be maintained in all strains of V. parahaemolyticus to maintain survival in the environment, whereas the second system, T3SS2, is linked to clinical isolates and disease in humans.


FIG 1 VopQ and VopS have antagonistic effect on T3SS1-specific pathway and network induction. (A)
Illustration of temporal effector function during T3SS1-mediated cell death. (B) Heat map of normalized differential expression of previously identified T3SS1-specific transcripts in uninfected (UN) primary human fibroblasts compared to primary human fibroblasts infected with either V. parahaemolyticus T3SS1 1 , T3SS1 1 DvopQ, or T3SS1 1 DvopS for 90 min. Yellow denotes transcripts with relative increased abundance infected cells compared to UN cells, and blue denotes decreased abundance. Clusters (color bars on the left) were assigned through hierarchical clustering of the differential expression data. (C) Heat map of predicted repression (blue) and activation (yellow) Z-scores calculated from differential expression data for UN versus V. parahaemolyticus T3SS1 1 , UN versus V. parahaemolyticus T3SS1 1 DvopQ, and UN versus V. parahaemolyticus T3SS1 1 DvopS using Qiagen's Ingenuity Pathway Analysis software. The color key correlates the displayed heat map color and calculated Z-scores, and gray denotes unaffected (P . 0.05) pathways. (D) Heat map of Ingenuity Pathway Analysis Z-score prediction of repression (blue) or activation (yellow) of biological networks after 90 min of POR3:T3SS1 1 , T3SS1 1 DvopQ, and T3SS1 1 DvopS infection.
interactions of VopQ with the c ring of V o is predicted to form an unconventional membrane pore through the juxtaposition of charged resides of VopQ against the hydrophobic lipid environment (19). This disruption is predicted to lead to the deacidification of the lysosomal membrane. We found that a fully functioning or assembled V-ATPase at the vacuole is not necessary to induce VopQ toxicity in Saccharomyces cerevisiae and that VopQ can interact with an assembly intermediate of the V-ATPase (V o c ring) in the endoplasmic reticulum (ER), resulting in cell death (19). As VopQ forms a pore in target membranes, the ER membrane is compromised, and this could lead to the induction of host cell signaling events, including the unfolded protein response (UPR).
Here, we found that VopQ activates the inositol-requiring enzyme 1 (IRE1) branch of the UPR in yeast and cultured cells. We demonstrate that the activation of IRE1 by VopQ results in an induction of extracellular signal-regulated protein kinase 1/2 (ERK1/ 2) signaling that is dependent on IRE1 kinase but not nuclease activity. We also found that another T3SS1 effector, VopS, dampens VopQ-mediated activation of ERK1/2 signaling by AMPylation-dependent inactivation of Rho GTPases, thereby limiting the activation of ERK1/2 signaling to early infection time points. Taken together, our results provide another example of the interplay between T3SS effectors and how they can temporally regulate host signaling pathways.

RESULTS
VopQ and VopS have antagonistic effects on T3SS1-specific pathway and network induction. Previously, we discovered that T3SS1 activates host cell survival networks and represses cell death networks (18). Since autophagy is linked to prosurvival network signaling, our findings led us to ask if VopQ could be responsible for prosurvival signals observed during infection. To understand the contribution of individual effectors to the T3SS1-specific transcriptional response, we characterized infection of primary human dermal fibroblasts (PHDFs) with V. parahaemolyticus strains carrying deletions of either vopQ or vopS. We chose this cell line so that our data would be comparable with the data obtained in our previous transcriptomic analysis and because, being primary cells, PHDFs do not carry transforming mutations that could alter cell signaling pathways (18). We included VopS because it targets Rho GTPases that regulate MAPK signaling (9,17). We used V. parahaemolyticus strain POR3, a derivative of the clinical strain RIMD2210633 that does not produce functional hemolysins or a functional T3SS2 (DtdhAS DvcrD2) but maintains an active T3SS1 (5). This strain and its DvopQ and DvopS derivatives are referred to herein as T3SS1 1 , T3SS1 1 DvopQ, and T3SS1 1 DvopS, respectively (see Table S1 in the supplemental material). As observed with previously characterized cell types, cytotoxicity of PHDFs occurring within the first 4 h of infection was completely dependent on VopQ and independent of VopS (see Fig. S1A and B) (2). We then performed RNA sequencing on the PHDFs after 90 min of infection with V. parahaemolyticus T3SS1 1 , T3SS1 1 DvopQ, and T3SS1 1 DvopS. The sequencing data passed statistical quality control tests, and principal-component analysis indicated tight clustering of replicates (Fig. S2A). Complete differential expression data are reported in Data Set S1, sheet 1, but for this study, we focused on the 398 host genes previously found to be differentially expressed specifically in response to T3SS1 (18).
The hierarchically clustered expression heat map in Fig. 1B illustrates how T3SS1 causes changes in expression of these 398 genes in the absence of either VopQ or VopS. Of these T3SS1-specific genes, 146 were similarly differentially expressed in the uninfected (UN) versus T3SS1 1 -infected and UN versus T3SS1 1 DvopQ-infected cells, and 197 were similarly differentially expressed in the UN versus T3SS1 1 -infected and UN versus T3SS1 1 DvopS-infected cells ( Fig. S2B; Data Set S1, sheet 2). Two hundred fiftytwo and 201 T3SS1-specific genes either were not differentially expressed or changed direction during T3SS1 1 DvopQ and T3SS1 1 DvopS infection, respectively. Expression of many genes, especially those within clusters 1 and 4, was oppositely affected during infection with T3SS1 1 DvopQ compared to during infection with T3SS1 1 DvopS ( Fig. 1B; Data Set S1, sheet 2). Notably, expression of the EGR1 and FOS transcription factors, which are known to be regulated by MAPK signaling pathways, was reduced in T3SS1 1 DvopQinfected cells compared to T3SS1 1 -infected cells and highly elevated by T3SS1 1 DvopS infection (Data Set S1, sheet 2) (20). We validated these findings by quantitative reverse transcription-PCR (RT-PCR) and by using V. parahaemolyticus strains with deletions of multiple effectors (T3SS1 1 DvopQR Dvpa0450 and T3SS1 1 DvopRS Dvpa0450) and showed that VopQ is necessary and sufficient for the elevated expression of both FOS and EGR1 ( Fig. S2C and D).
We next used Ingenuity Pathway Analysis (IPA) to understand how the activities of VopQ and VopS contribute to the changes in host signaling events induced by the T3SS1 (Data Set S1, sheet 3). The T3SS1-specific induction or repression of many pathways was dependent on VopQ and enhanced in the absence of VopS (Fig. 1C). For example, induction of NF-κB signaling, actin cytoskeleton signaling, and Rho GTPase signaling by T3SS1 was greatly reduced in the absence of VopQ and enhanced in the absence of VopS. These observations are consistent with the opposing effects on differential expression patterns observed in Fig. 1B.
VopQ induces prosurvival signaling networks. To understand the relative contributions of VopQ and VopS to the host response to T3SS1 on the network level, we used IPA to perform biological function network analysis. Previously, we had shown that the T3SS1 activates cell survival networks and represses cell death networks (18). Strikingly, this effect was completely lost during T3SS1 1 DvopQ infection and amplified during T3SS1 1 DvopS infection ( Fig. 1D; Data Set S1, sheet 4). Specifically, we observed a loss in cell survival, viability signaling network activation, and death and mortality signaling network repression in PHDFs infected with V. parahaemolyticus T3SS1 1 DvopQ compared to V. parahaemolyticus T3SS1 1 , while infection with T3SS1 1 DvopS instead amplified these signaling changes (Fig. 1D). The apoptosis signaling network, which is normally repressed during T3SS1 1 infection, was activated during infection with T3SS1 1 DvopQ (Fig. 1D). These data support a model in which the activity of VopQ elicits transcriptional changes in the host cell that result in the activation of cell survival and repression of cell death networks, and VopS may function to dampen this response.
VopQ induces a pulse of ERK1/2 signaling that is dampened by VopS. To further dissect VopQ's effect on host signaling pathways in mammalian cells, we continued with a more genetically tractable model, mouse embryonic fibroblasts (MEFs). We characterized the cytotoxicity of the V. parahaemolyticus T3SS1 1 strain and its derivates in MEFs. The cell death induced by T3SS1 occurred over a similar time scale in MEFs as in PHDFs and was similarly dependent on VopQ (Fig. S3A). This result was expected, because T3SS1-mediated cell death is conserved across diverse cell types (2,8,17,18). We chose to examine the ERK1/2 signaling pathway because the RNA sequencing data suggested that VopQ activates Rho GTPase signaling (Fig. 1C), and in previous work we demonstrated that T3SS1-induced EGR1 and FOS expression requires active mitogenactivated protein kinase kinases 1 and 2 (MEK1/2), the kinases upstream of ERK1/2 (18). Furthermore, as we did not observe agonist and antagonist effects of VopQ and VopS, respectively, on the expression of c-Jun N-terminal kinase (JNK) signaling target genes, we focused our analysis on the ERK1/2 pathway (Table S2).
To test if VopQ induces EGR1 and FOS expression by activating the ERK1/2 MAPK pathway early during infection, we starved MEFs to remove basal ERK1/2 phosphorylation, infected them with V. parahaemolyticus T3SS1 1 or V. parahaemolyticus T3SS1 2 for 45, 60, 75, and 90 min, and probed for phospho-ERK1/2 as well as for the presence of downstream Egr1 by Western blotting. We found that the T3SS1 induced a pulse of ERK1/2 phosphorylation that peaked around 45 min and had completely disappeared by 90 min postinfection ( Fig. 2A). Total Egr1 protein levels began to rise 60 min postinfection and reached their maximum at 90 min postinfection, which is an expected pattern of expression when the time for transcription and translation of the induced Egr1 gene is taken into account (Fig. 2B). Using the MEK1/2 inhibitor U0126, we found that T3SS1-induced Egr1 expression was dependent on MEK1/2 activity in MEFs, similar to what was previously reported for PHDFs ( Fig. S3B) (18).
We repeated the infection time course with the V. parahaemolyticus T3SS1 1 DvopQ strain and found that the T3SS1-induced pulse of ERK1/2 phosphorylation was indeed dependent on VopQ, as was the increase in total Egr1 protein levels ( Fig. 2C and D). When MEFs were infected with T3SS1 1 DvopS, we observed not only an amplified induction of ERK1/2 phosphorylation and Egr1 production but also an extended duration of ERK1/2 phosphorylation ( Fig. 2C and D). Complementing T3SS1 1 DvopS with a wild-type copy of vopS reverted the induced ERK1/2 phosphorylation pattern to that observed during T3SS1 1 infection (Fig. S3C). Interestingly, complementation of T3SS1 1 DvopS with the catalytically dead vopS H348A allele did not result in a reversion of the included ERK1/2 phosphorylation pattern (Fig. S3C). This observation is consistent with previous work that found that VopS's AMPylation of Rho GTPases inhibits host ERK1/2 and JNK MAPK pathways (17). This early pulse of VopQ-dependent ERK1/2 MAPK signaling in MEFs is distinct from the previously reported VopQ-dependent ERK1/2 phosphorylation in Caco-2 cells at late infection time points, 3 to 4 h postinfection, as those assays were performed with V. parahaemolyticus strains that encoded functional T3SS1, T3SS2, and TdhAS hemolysins (21). Taken together, these data support the model that the combined actions of VopQ and VopS create a pulse of ERK1/2 MAPK signaling that is restricted to early infection time points, resulting in the controlled expression of downstream transcription factors EGR1 and FOS.
If VopQ and VopS work together to fine-tune the host response, their co-occurrence in Vibrio genomes containing the T3SS1 gene cluster would be predicted to be high. To test this, we used the SyntTax server to identify all Vibrio strains that retained synteny in the T3SS1 gene neighborhood (Table S3). We identified 58 Vibrio strains representing 8 species containing the T3SS1 gene cluster and found that 91.4% of genomes containing vopQ also contained vopS (53/58) (Table S3). We found that these genes cooccur in diverse Vibrio species, including V. parahaemolyticus, Vibrio diabolicus, Vibrio antiquarius, Vibrio campbellii, and Vibrio alginolyticus (Fig. S4). Interestingly, the 5 genomes containing vopQ but lacking vopS belonged to two Vibrio species, Vibrio harveyi and Vibrio tubiashii.
VopQ-induced prosurvival signaling is independent of endosomal deacidification. Next, we wanted to understand how VopQ could activate ERK1/2 MAPK signaling in the host. The VopQ channel deacidifies vacuolar and lysosomal compartments but also inhibits homotypic fusion of yeast vacuoles, a model for Rab GTPase-and SNARE-dependent fusion between the lysosome and autophagosome (8,22). Mutation of serine 200 to a proline creates a mutant, VopQ S200P , that is still able to neutralize the vacuole or lysosome but can no longer block fusion (8). This observation is likely due to reduced binding of VopQ S200P to the V-ATPase (19). To test if VopQ's activation of ERK1/2 MAPK signaling was caused by one or both of these functions, we exchanged the chromosomal copy of the vopQ gene in the V. parahaemolyticus T3SS1 1 strain with a version encoding VopQ S200P , creating V. parahaemolyticus strain T3SS1 1 vopQ S200P . We tested the cytotoxicity of this strain during MEF infection and found that the VopQ S200P mutant was no less lethal than wild-type VopQ (Fig. S3A). However, unlike its parent strain, V. parahaemolyticus T3SS1 1 vopQ S200P was unable to induce ERK1/2 phosphorylation and downstream production of Egr1 in MEFs ( Fig. 2E and F). Notably, treatment of MEFs with chloroquine, a drug that prevents lysosomal acidification, was not able to induce phosphorylation of ERK1/2 (Fig. 2E). These data suggest that ERK1/2 MAPK signaling is not activated by lysosomal deacidification alone and is instead dependent on VopQ's strong physical interaction with the V-ATPase. We therefore considered two models by which VopQ could induce ERK1/2 MAPK signaling. In the first, VopQ's inhibition of lysosome-autophagosome fusion directly activates ERK1/2 MAPK signaling. In the second, VopQ manipulates another pathway upstream of both lysosome-autophagosome fusion and ERK1/2, thereby altering these two pathways in parallel.
VopQ-mediated cytotoxicity was also conserved in the Atf6 2/2 and PERK 2/2 MEFs ( Fig. S5B and C). Both the Atf6 2/2 and PERK 2/2 cell lines exhibited the same pattern of ERK1/2 phosphorylation and Egr1 accumulation upon T3SS1 1 infection as wild-type MEFs (Fig. 3B), indicating that VopQ's activation of ERK1/2 was specifically IRE1 . P values were calculated by an unpaired t test (****, P , 0.0001). (F) Immunoblot for p-Ire1a and total Ire1a in MEFs infected with T3SS1 1 , T3SS1 1 DvopQ, T3SS1 1 DvopQ1pvopQ, and T3SS1 1 vopQ S200P V. parahaemolyticus strains for 45 and 60 min. In panels A to C and F, the target band is marked with a red line, and background bands are indicated with a blue star. Blots are representative of 3 independent experiments. dependent. Finally, we stimulated wild-type, IRE1 2/2 , Atf6 2/2 , and PERK 2/2 MEFs with fetal bovine serum (FBS) to assess whether the well-described growth factor-stimulated ERK1/2 MAPK signaling was functional in these cell lines (29). FBS-stimulated ERK1/2 phosphorylation and downstream Egr1 expression were observed in all cell lines (Fig. 3C). These data strongly support our model that VopQ's IRE1-dependent activation of ERK1/2 occurs through a pathway that is separate from the established growth factor-stimulated pathway mediated by Ras and Raf (29).
VopQ expression results in IRE1 activation in yeast. Next, we asked if VopQ activates IRE1. VopQ toxicity is dependent on the assembly c subunit of the V o V-ATPase in yeast independent of the vacuolar localization of the complex. This led us to hypothesize that interaction of VopQ and the c subunit ring could also take place in the ER, where the V o complex initially forms (19). In addition, the ER is the source of membranes for autophagosomes; thus, a block in autophagic flux may also perturb the protein-to-lipid ratio of the ER (30). We predicted that if either of these scenarios occurred, the membrane perturbations caused by VopQ might activate the UPR, which is mediated by IRE1 in yeast, either through the disruption of the ER lumen environment or through the activation of lipid bilayer stress (31)(32)(33)(34). IRE1 is a type I ER-resident transmembrane protein that contains a protein kinase and an endoribonuclease domain in its cytoplasmic region (35). IRE1 also contains an amphipathic helix which can sense perturbations in the lipid bilayer, leading to activation and initiation of the UPR (32).
To test this hypothesis, we assessed the clustering of IRE1 in yeast by visualizing endogenously expressed IRE1p-GFP (9). Plasmids carrying galactose-inducible genes encoding wild-type VopQ, the V-ATPase binding mutant VopQ S200P , and VopA were transformed into the BY4741 IRE1-green fluorescent protein (GFP) yeast strain. VopA is a V. parahaemolyticus T3SS2 effector that kills yeast by a mechanism that is distinct from VopQ and was included as a control (36). Upon galactose induction, serial growth assays showed that VopQ and VopA both inhibited growth in the BY4741 IRE1-GFP strain while VopQ S200P and vector alone control did not (Fig. S5D). We then monitored IRE1p-GFP clustering, or focus formation, at 30 and 45 min after galactose induction by confocal microscopy. Treatment with dithiothreitol (DTT) for the same period was used as a positive control for UPR stress. Yeast expressing VopQ but not VopQ S200P or VopA induced IRE1p-GFP focus formation (Fig. 3D). IRE1p-GFP foci formation was observed, on average, in about 70% of DTT-treated cells compared to about 15% in cells expressing VopQ (Fig. 3E). This difference was expected, because DTT treatment is homogeneous, whereas expression of VopQ is stochastic (37). Our data indicate that expression of VopQ in yeast results in IRE1 activation.
VopQ causes induction of the IRE1 branch of the UPR of mammalian cells during infection. We next asked if VopQ also activates IRE1 in mammalian cells during infection. IRE1 is normally sequestered by BiP but is released upon UPR activation, when it oligomerizes, trans-autophosphorylates, and activates its endoribonuclease activity, resulting in the nonconventional splicing of X-box binding protein 1 (XBP1) mRNA in mammalian cells (38)(39)(40). To determine if the IRE1 branch of the UPR was activated by VopQ at early infection time points, we measured levels of phospho-IRE1a in MEFs by Western blotting at 45 and 60 min postinfection. Our results indicate that V. parahaemolyticus T3SS1 1 induces IRE1a phosphorylation in a VopQ-dependent manner (Fig. 3F) and that T3SS1 1 vopQ S200P is not able to induce IRE1a phosphorylation. Taken together, these data indicate that the V-ATPase binding activity of VopQ activates the IRE1 branch of the UPR in both MEFs and yeast.
VopQ-induced prosurvival ERK signaling is dependent on IRE1 kinase activity. Next, we wanted to further determine if the catalytic activities of IRE1 are required for VopQ-induced ERK1/2 signaling. IRE1 contains both protein kinase and an endoribonuclease domain in its cytoplasmic region (35). To test if the kinase or endonuclease activity of IRE1 is required for ERK signaling, MEFs were treated before V. parahaemolyticus infection with KIRA6 or 4m8c, an IRE1-specific kinase inhibitor and a potent inhibitor of IRE1 RNase activity, respectively (Fig. 4A). Strikingly, ERK1/2 phosphorylation and Egr1 protein expression were not observed in infected MEFs treated with the kinase inhibitor KIRA6.
However, 4m8c treatment appeared to have no effect on VopQ-induced ERK1/2 phosphorylation and Egr1 protein expression. Taken together, these data indicate that the prosurvival ERK signaling induced by VopQ during V. parahaemolyticus infection is dependent on IRE1 kinase activity.

DISCUSSION
By studying the function of two effectors of the T3SS1 of the seafood-borne pathogen V. parahaemolyticus, we have shown that T3SS effectors act together to systematically manipulate the host response. We observe that the effector VopQ is responsible for the T3SS1-mediated activation of cell survival and repression of cell death networks and another effector, VopS, is responsible for dampening this response. Of note, VopS has been determined to AMPylate and thereby inactivate Rac, a known activator of MEK1/2 mediated ERK signaling (9). Furthermore, in Vibrio alginolyticus, a Vibrio species closely related to V. parahaemolyticus, VopS was found to be required for the rapid induction of apoptosis in infected cells (41). The association of VopQ with the V-ATPase elicits early activation of ERK1/2 phosphorylation that is turned off by the delayed temporal action of VopS (Fig. 4B).
Manipulation of the V-ATPase to affect signaling in the cell was previously observed with oncogenic Ras, which induces a Rac-dependent plasma membrane ruffling and micropinocytosis. However, in the case of oncogenic Ras, activation of Rac is dependent on the activity of the V-ATPase along with its relocation to the plasma membrane (42). Importantly, the activation ERK1/2 MAPK signaling by VopQ does not occur through the growth factor-inducible Ras-mediated pathway but instead is dependent on the kinase activity of the ER stress sensor and cell fate executor IRE1 (Fig. 4B). Activation of ERK1/2 MAPK signaling specifically through the IRE1 branch of the UPR by a bacterial effector has not previously been reported.
The evolutionary conservation of vopQ and vopS in Vibrio species retaining synteny in the T3SS1 neighborhood suggests that their concerted function is important for proper modulation of the host response. Effector pairs with opposing effects yet synergistic effects during infection are not unknown. YopJ and YopM of Yersinia pestis were previously described as having opposing effects on interleukin signaling and caspase-1 processing which synergistically suppressed proinflammatory cytokines during infection (43). In Salmonella, the effectors SptP and SopE act as respective GTP-activating proteins (GAPs) and guanine nucleotide exchange factors (GEFs) for Cdc42 and Rac. Rapid degradation of SopE allows the temporal regulation of Cdc42 and Rac activity during Salmonella infection (44). Furthermore, Legionella pneumophila utilizes several such pairs of effectors, such as SidM/DrrA and LepB, SidH and LubX, and AnkX and Lem3 (45)(46)(47)(48)(49)(50)(51). Together, these effector pairs coordinate the establishment, maintenance, and properly timed escape from the legionella-containing vacuole (LCV), the environmental niche Legionella requires to replicate (52,53).
The targeting of autophagy, UPR, and MAPK signaling together is not unprecedented for pathogens. For example, the mycotoxin patulin was found to manipulate these pathways though the inhibition of cathepsin B and cathepsin D, which leads to an accumulation of p62. Increased p62-mediated autophagy activates the PERK and IRE1 branches of the UPR through increased reactive oxygen species production. UPR activation then results in activation of ERK1/2 and phosphorylation of BAD, resulting in increased survival of host cells (54). Several viruses rely on the interplay of UPR and MAPK signaling for virulence as well. Dengue virus (DENV) relies on UPR activation of JNK signaling to induce autophagy and increase viral load of infected cells. Treatment with a JNK inhibitor decreased viral titers and reduced symptoms of DENV2 in mice (55). Recently the coronavirus infectious bronchitis virus (IBV) was found to rely on the activation of IRE1 and ERK1/2, but not XBP1 or JNK, for the induction of autophagic flux and prosurvival signaling during IBV infection (56).
We and others have observed that the ERK1/2 MAPK signaling can be regulated either directly or indirectly through the IRE1 branch of the UPR (54). However, it remains unclear if VopQ's induction of IRE1 signaling is caused by the manipulation of autophagy or via localized perturbations in the ER membrane caused by VopQ's interaction with assembly intermediates of the V-ATPase V o subcomplex in the ER. Deciphering the epistatic relationship between VopQ's association with the V-ATPase and activation of IRE1 will be the subject of future studies that will be important for understanding the role of the V-ATPase-UPR-MAPK feedback network in both cellular homeostasis and bacterial infection. Similarly, the interaction of VopQ with the V-ATPase at the ER should be further studied to determine if this interaction is sufficient to activate IRE1's lipid bilayer stress response and ERK1/2 MAPK signaling independently of autophagy (33,57). Other groups have hypothesized that Ire1's interaction with the adapter protein Nck plays an important role in activation of ERK1/2 signaling upon ER stress; however, whether IRE1's kinase or endonuclease activity is needed was unknown (27). Our studies indicate that the kinase activity of IRE1 is required, but the mechanism by which IRE1 kinase activity leads to ERK1/2 MAPK signaling is still poorly understood. Future experiments to dissect the role of Nck and Rho GTPase activation in this process would be a valuable addition to understanding this molecular mechanism of IRE1induced, growth factor-independent ERK1/2 MAPK signaling.

MATERIALS AND METHODS
Bacterial strains and culture conditions. The Vibrio parahaemolyticus POR3 (POR1DvcrD2) and POR4 (POR1DvcrD1/vcrD2) strains were generously provided by Tetsuya Iida and Takeshi Honda of Agonist-Antagonist Pair Manipulates Host Signaling Osaka University. Vibrio strains were cultured at 30°C in MLB (Luria-Bertani broth with 3% NaCl). All V. parahaemolyticus strains except V. parahaemolyticus T3SS1 1 vopQ S200P were from previous studies (Table S1). The T3SS1 1 vopQ S200P strain was created by cloning the vopQ S200P allele (8) flanked by the nucleotide sequences 1 kb upstream and 1 kb downstream of vopQ (vp1680) into pDM4, a Cm r OriR6K suicide plasmid. Escherichia coli S17 (lpir) was used to conjugate the resulting plasmid into the POR3 strain, and transconjugants were selected on medium containing 25 mg/ml chloramphenicol. Bacteria were then counterselected on 15% sucrose, and insertion of the vopQ S200P allele was confirmed by PCR.
Infection of PHDFs for RNA sequencing and quantitative RT-PCR. For RNA sequencing, PHDFs were seeded onto 6-well plates at a density of 1 Â 10 5 cells/ml and grown for 18 to 20 h to ;80% confluence. Overnight V. parahaemolyticus cultures were normalized to an optical density at 600 nm (OD 600 ) of 0.2 and subcultured to an OD 600 of 0.6. Bacteria were pelleted, resuspended in unsupplemented DMEM, and grown at 37°C for 45 min to preinduce T3SS1 expression (16). PHDFs were washed with unsupplemented DMEM and then infected with preinduced V. parahaemolyticus strains at a multiplicity of infection (MOI) of 10. Plates were centrifuged at 1,000 Â g for 5 min to synchronize infection and incubated at 37°C and 5% CO 2 . At 90 min postinfection, RNAprotect cell reagent (Qiagen) was added to stop the infection and preserve the RNA. Cells were harvested by scraping and pooled, and pellets were resuspended in RLT-plus buffer (Qiagen) and stored at 280°C. The same infection protocol was followed for quantitative PCR (qPCR) experiments with PHDFs and MEFs. For MEK1/2 inhibition qPCR experiments, MEFs were incubated with 10 mM U0126 (Cell Signaling) or 10 mM dimethyl sulfoxide (DMSO) (vehicle) during and for 1 h prior to infection.
RNA isolation and RNA sequencing. RNA isolation was performed as for PHDFs and MEFs. Cells were lysed with 27-gauge, 1/2-in. needles and then homogenized with QIAshredder columns (Qiagen). Total RNA from triplicate experiments was purified with the RNeasy Plus kit (Qiagen). The quality of purified total RNA samples was determined with an Agilent 2100 Bioanalyzer, and only samples with an RNA integrity number (RIN) of 9 or higher were used. RNA concentration was measured with a Qubit fluorimeter prior to library prep. Four micrograms of total DNase-treated RNA was run through the TruSeq Stranded Total RNA LT sample prep kit from Illumina as previously described (18). Samples were quantified by Qubit before being normalized, pooled, and then sequenced on the Illumina HiSeq 2500 sequencer with SBS v3 reagents. Each sample was sequenced at a depth of at least 25 million 50-nucleotide single-end reads.
Infection of MEFs for immunoblotting. For ERK1/2 and Egr1 immunoblotting experiments, MEFs were starved in unsupplemented DMEM for 1 h prior to infection/treatment to remove background growth factor-stimulated MAPK signaling. MEFs were not starved for in phospho-IRE1a experiments in order to prevent background activation of the UPR by nutritional stress. Cells were then infected at an MOI of 10 with V. parahaemolyticus as described above or with treated DMEM supplemented with 10% (vol/vol) fetal bovine serum (FBS; Sigma-Aldrich). To inhibit IRE1 nuclease and kinase activities, cells were treated with 100 mM 4m8c for 24 h and 100 mM KIRA6 for 1 h before infection, respectively.
IRE1p-GFP clustering assay. Yeast strains were grown to mid-log phase (OD 600 ; 0.5) in CSM medium lacking uracil with 2% raffinose. Cultures were then treated with 2% galactose, 2% raffinose, or 2% raffinose with 5 mM DTT for 30 or 45 min. Cultures were collected, resuspended in 1Â PBS, and fixed with 4% paraformaldehyde. Confocal images were acquired using a Zeiss LSM800 with Zen software. Images were processed with ImageJ (National Institutes of Health) and Adobe Photoshop CS6. For quantification, the presence of IRE1-GFP foci was scored in 100 cells per experiment over three independent experiments.
Statistical methods. For RNA sequencing DE analysis, statistical cutoffs were as follows: false discovery rate (FDR), #0.01; log 2 counts per million (log 2 CPM), $0; and absolute value of fold change (FC), $1.5. For IPA and biological network analysis, the P values are presented as 2log(P value) and the cutoff for significance was a P value of ,0.05. FDRs and P values are reported in Data Set S1, sheets 1 to 4. For quantitative RT-PCR, P values were calculated by one-way analysis of variance (ANOVA) and Dunnett's multiple-comparison test. For the quantification of IRE1-GFP clustering, P values were calculated by Student's unpaired t test (two tailed). Additional materials and methods are available in the supplemental material.
Data availability. Complete RNA-sequencing data have been deposited on the Gene Expression Omnibus server (GSE120273).