ARHGEF26 enhances Salmonella invasion and inflammation in cells and mice

Salmonella hijack host machinery in order to invade cells and establish infection. While considerable work has described the role of host proteins in invasion, much less is known regarding how natural variation in these invasion-associated host proteins affects Salmonella pathogenesis. Here we leveraged a candidate cellular GWAS screen to identify natural genetic variation in the ARHGEF26 (Rho Guanine Nucleotide Exchange Factor 26) gene that renders lymphoblastoid cells susceptible to Salmonella Typhi and Typhimurium invasion. Experimental follow-up redefined ARHGEF26’s role in Salmonella epithelial cell infection. Specifically, we identified complex serovar-by-host interactions whereby ARHGEF26 stimulation of S. Typhi and S. Typhimurium invasion into host cells varied in magnitude and effector-dependence based on host cell type. While ARHGEF26 regulated SopB- and SopE-mediated S. Typhi (but not S. Typhimurium) infection of HeLa cells, the largest effect of ARHGEF26 was observed with S. Typhimurium in polarized MDCK cells through a SopB- and SopE2-independent mechanism. In both cell types, knockdown of the ARHGEF26-associated protein DLG1 resulted in a similar phenotype and serovar specificity. Importantly, we show that ARHGEF26 plays a critical role in S. Typhimurium pathogenesis by contributing to bacterial burden in the enteric fever murine model, as well as inflammation in the colitis infection model. In the enteric fever model, SopB and SopE2 are required for the effects of Arhgef26 deletion on bacterial burden, and the impact of sopB and sopE2 deletion in turn required ARHGEF26. In contrast, SopB and SopE2 were not required for the impacts of Arhgef26 deletion on colitis. A role for ARHGEF26 on inflammation was also seen in cells, as knockdown reduced IL-8 production in HeLa cells. Together, these data reveal pleiotropic roles for ARHGEF26 during infection and highlight that many of the interactions that occur during infection that are thought to be well understood likely have underappreciated complexity.

The authors would like to thank the editors and reviewers for their thoughtful, insightful, and constructive feedback. The substantially revised manuscript that we are submitting has addressed the concerns below and is a significant improvement on our previous submission due to the helpful suggestions provided. Some of the experimental highlights include: • Identifying that polarized ARHGEF26 knockdown and DLG1 knockout MDCK cells are significantly protected from S. Typhimurium invasion in a SopB-SopE2independent manner • Finding that the differences in S. Typhi and S. Typhimurium's use of ARHGEF26 during HeLa cell invasion is partially, but not entirely, SopB, SopE, and SopE2 dependent • Determining that the impacts of ARHGEF26 on S. Typhimurium fitness in mice require the effectors SopB and SopE2 during the enteric fever model • Discovering that the impacts of SopB and SopE2 on S. Typhimurium fitness in mice require ARHGEF26 during the enteric fever model • Demonstrating that the Arhgef26 -/inflammation phenotype during the murine colitis model is SopB and SopE2 independent Below, each concern is addressed and the resulting changes to the manuscript are provided. Thank you again, Dennis Ko, Jeff Bourgeois, et al.
(1) Your method of testing "invasion" is unusual. It strictly does not measure invasion (total number of bacteria that were internalised in the monolayer) but rather the percentage of cells that are infected. And it is done at 3 h 15 min post-infection, which is well after the initial invasion event. By this time, Salmonella would have undergone 1-2 rounds of replication inside of cells. As pointed out by one reviewers, your differences in "invasion" are small (<15%). Please confirm whether you get similar results with an assay that quantifies the total number of bacteria that have been internalized e.g. gentamicin protection assay. Regarding our time point, we believe that this concern is addressed by altering our language to be host-cell centric, as the flow cytometric method only considers if a cell does or does not contain measurable amounts of GFP-tagged bacteria, i.e., it is agnostic to whether bacterial replication has begun. (2) The streptomycin-pretreatment mouse model is not a model of gastroenteritis. Rather it should be termed a mouse colitis model. See https://pubmed.ncbi.nlm.nih.gov/21343352/ for clarification. Please correct this throughout, including in figures.

Reviewer's Responses to Questions: Part I -Summary
Please use this section to discuss strengths/weaknesses of study, novelty/significance, general execution and scholarship.
Reviewer #1: It has been known for many years that pathogenic Salmonella serovars manipulate host cell GTPases to drive bacterial internalization into non-phagocytic epithelial cells. Previous work has suggested that one of these GTPases is RhoG, activated by the host GEF ARHGEF26 (also known as SGEF) in response to phosphoinositide changes induced by the bacterial effector protein SopB. Here the authors identified ARHGEF26 in a focused genetic screen for natural genetic variants that enhance the infection of lymphoblastoid cells by S. Typhi or S. Typhimurium. Using a variety of in vitro and in vivo (mouse) assays, they demonstrate a small but statistically significant impact of ARHGEF26 on both bacterial internalization and the induction of inflammatory signaling (IL-8 production). In vitro, knockdown of ARHGEF26 appears to selectively affect S. Typhi as opposed to S. Typhimurium, presumably due to differences in the arrays of secreted effector proteins produced by each serovar. In contrast, the most physiologically significant difference is seen in mouse models of S. Typhimurium infection, where knockout of ARHGEF26 inhibits colonization of the ileum (in an enteric fever model) or inflammation, but not colonization in the colon (in a gastroenteritis model).
While the data certainly suggest a potential genetic basis for differential susceptibility to Salmonella infection, no clear unifying mechanism is identified. There are many "plausible hypotheses" but no firm conclusions. The in vitro data indicate that RhoG is not involved in invasion (contradicting an earlier study from the Galan lab), but that the bacterially encoded Rho-GEF SopE is, however no link between ARHGEF26 and SopE or its downstream targets Rac1 and Cdc42 is established. The data also implicate two scaffolding proteins, DLG1 and SCRIB, which are involved in the establishment and maintenance of epithelial junctions, however how they contribute to invasion is never defined, or even explored.
Through this work we have redefined the interactions of ARHGEF26 with S. Typhimurium and S. Typhi effectors during invasion. Previous work on ARHGEF26 has been limited to a single figure speculating that ARHGEF26 and SopE/SopE2 are interchangeable to activate RHOG and induce membrane ruffling (5). In this work, we have expanded and substantially revised this model to demonstrate that ARHGEF26 also contributes to SopE-mediated and RHOG-independent invasion of HeLa cells, as well as SopB-SopE2independent invasion of polarized MDCK cells. Given that this represents a significant difference from past literature, we are specifically very careful to avoid over interpretation of this data in order to encourage future work on this system and avoid speculative mechanisms being recorded as fact. That is to say, the transparency in what we have and have not experimentally demonstrated is intentional. The biology discovered in this work using 3 cellular and 2 mouse models, particularly the large effects observed on invasion into polarized MDCK cells and bacterial burden and inflammation in mice, reveal a central role for ARHGEF26 in Salmonella infection. While we agree that it will be useful for the field to further understand the mechanisms by which bacterial effectors, ARHGEF26, and RHO-GTPases connect, many of these questions are outside the scope of this paper and require highly technical cell biology approaches that can be utilized in future work.
Similarly, regarding SCRIB and/or DLG1, our contributions demonstrate SCRIB and DLG1 are required for HeLa cells and polarized MDCK cells, with knockdown or knockout nearly perfectly phenocopying ARHGEF26 knockdown in each situation tested. This is a critical leap forward in understanding ARHGEF26's contributions to invasion, as it ties the observations in Salmonella into the current paradigm shifts of ARHGEF26 regulation. Recent work from one of our labs has not only demonstrated that ARHGEF26 requires SCRIB and DLG1 interactions, but that the way in which these three proteins cooperatively regulate different cellular functions varies substantially depending on the specific context being studied (6). In that work, it is established that there are GEFdependent and a poorly understood GEF-independent roles for the protein complex. Given the significant heterogeneity we observe in our results with complex host by serovar by effector protein interactions, we hypothesize that it is very likely that ARHGEF26, SCRIB, and DLG1 may be performing dramatically different functions across our systems. Therefore, as above, we have defined a new, critical area where gaining greater understanding of these roles represents a very important future direction that is outside of the scope of the work presented here.
Most of the phenotypic effects of ARHGEF26 depletion in vitro are relatively small, suggesting that, whatever its function, it is not a primary driver of invasion or inflammation but may contribute in some secondary way.
The effects observed in vivo are somewhat larger, but confusing and somewhat contradictory. While depletion of ARHGEF26 had no effect at all on S. Typhimurium invasion in cultured cells, it had a significant effect in the ileum during oral infection of ARHGEF26-deficient mice with S. Typhimurium. It is possible that the interacting partners DLG1 and SCRIB have more important roles in the context of a polarized epithelium (i.e. the ileum) than in non-polarized cells in culture, but this issue is never explored.
A key finding of our paper is that the effect of ARHGEF26 on invasion varies substantially depending on the system used. In our previous submission, LCLs demonstrated a 10% proportional decrease in S. Typhimurium invasion susceptibility, while HeLa cells demonstrated no difference in S. Typhimurium invasion following knockdown. As the reviewer noted, in contrast to these results, the effect in vivo effect on burden was substantial (an approximately 2 log decrease in the ileum in the enteric fever model). This variation in effect size was somewhat puzzling but has now been resolved in following the Reviewer's suggestion to examine the role of ARHGEF26 in polarized epithelium. In polarized MDCK cells, we observed an 70% proportional decrease in the number of infected cells with ARHGEF26 knockdown, and a 92% reduction in relative invasion as measured by microscopy. While these results are not perfectly representative of the phenomena happening in vivo as the effects are SopBE2 independent in MDCKs but dependent in mice, it does set a precedent that ARHGEF26 can have substantial impacts on invasion susceptibility in vitro. We thank the reviewer for their suggestion to look at ARHGEF26 function in polarized epithelial cells and believe these experiments have substantially improved the paper.
Source: This work (Derived from Figure 6) In a different model of oral infection (streptomycin pre-treatment) where colonization occurs primarily in the colon, not the ileum, the data indicate no difference in colonization but a significant difference in inflammation (reduced in the absence of ARHGEF26). Although it is possible that regional differences in the intestinal epithelium itself (or regional expression of ARHGEF26) could account for this, the use of two different models of infection makes the result impossible to interpret from a mechanistic perspective.
We apologize for the confusion, as it appears our description of the two mouse models and, in particular, the implications of the equal bacterial burdens at two days post infection in the colitis model, may have caused confusion. The two models should be considered and interpreted separately, as one is a model for enteric fever and the other a model for colitis. We have expanded our description to clarify these points: "To test whether ARHGEF26 contributes to Salmonella-induced inflammation in vivo, we again examined Arhgef26-/-mice. Importantly, while the enteric fever model is able to measure SPI-1and invasion-dependent Salmonella fitness in vivo, it does not mimic the natural progression of most S. Typhimurium illness in humans as it causes very little inflammation. In order to mimic S. Typhimurium colitis disease in murine models, microbiota must be reduced with streptomycin pretreatment prior to infection (11) (Figure 9A). Interestingly, in this model, SPI-1 secretion does not impact bacterial burden at early timepoints but instead is required for inflammation onset (10, 11). These differences mean that data from the colitis model cannot be combined with data from the enteric fever model, but that interpreting the colitis model separately can provide important information about pathogenic processes that do not occur in the enteric fever model." The reduction of inflammation in the Arhgef26 -/cecum and colon following infection suggests that the knockout does have an effect on Salmonella pathogenesis in these tissues. We propose that Arhgef26 has tissue-and model-dependent roles during infection, and that not all of these roles are involved in bacterial abundance. In fact, these roles perfectly mimic the effects of SPI-1 ablation in each model. In the enteric fever model where spread and burden in the ileum depends on successful SPI-1-mediated invasion, we see differences in bacterial burden. In contrast, in the colitis model where SPI-1 secretion primarily drives inflammation without impacting burden at early timepoints, we observe differences in pathology. A comparable phenomenon is observed in our in vitro data where ARHGEF26 knockdown reduces IL-8 production following S. Typhimurium infection without impacting HeLa cell invasion. These pleiotropic effects are consistent with SPI-1 and ARHGEF26 having separate roles in invasion or inflammation depending on the in vivo model used.
Reviewer #2: Through the GWAS screen, Bourgeois et al. identified ARHGEF26 (also known as SGEF) as a host factor that up-regulates Salmonella invasion. In vitro invasion assays using HeLa cells, they found that ARHGEF26 contributes to SopB-and SopE-mediated S. Typhi (but not S.Typhimurium) invasion. They examined the molecular mechanism how ARHGEF26 activity is regulated, focusing on the ARHGEF interactors, DLG1 and SCRIB, as well as by the ARHGEF26 domain analysis. They further showed the possible role of ARHGEF26 in Salmonella-induced inflammation by monitoring IL-18 secretion. Finally, they demonstrated the contribution of ARHGEF26 to S. Typhimurium fitness using the mouse model.
Overall, the manuscript is well-written and gives new insights into a role of ARHGEF26 in SopE-mediated Salmonella invasion and inflammation, although there are some inconsistencies between results depending on the experimental systems (ex. HeLa cells vs LCLs) as the authors discussed. For the disagreement with a previous report, I think that some results need to be more clarified by conducting additional experiments.
Thank you for your positive feedback. We would like to state that while our report disagrees with the canonical roles of ARHGEF26 and RHOG during Salmonella uptake, we are not the first to dispute the involvement RHOG in invasion (7,8). Though, of course, we are happy to improve the rigor of this report through the additional experiments suggested by the reviewers.
Reviewer #3: In this manuscript the author have investigated the role of ARHGEF26 in response to Salmonella Typhi and Salmonella Typhimurium infection. They demonstrate that the effector proteins SopE and SopB of S. Typhi are required for efficient ARHGEF26-dependent invasion in HeLa cells, contrary to S. Typhimurium effectors SopE2 and SopB which appeared not to require ARHGEF26 to invade HeLa cells. They further demonstrate that activation of ARHGEF26 can induce inflammatory responses, and that ARHGEF26 plays a critical role in S. Typhimurium pathogenesis in both the enteric fever murine model and the murine gastroenteritis infection model.

Strengths:
Natural genetic variations in ARHGEF26 were identified in a cellular GWAS screen that contribute to susceptibility to Salmonella Typhi and Typhimurium invasion. Demonstrated new roles of ARHGEF26 during Salmonella invasion. Strong in vitro and in vivo data with a great deal of experimental repeats.
Weaknesses: Some controls that will strengthen the conclusions are missing. Additional time points in the in vivo experiments should be included.

Thank you for your informative feedback and for your complimentary assessment of the manuscript. Part II -Major Issues: Key Experiments Required for Acceptance
Please use this section to detail the key new experiments or modifications of existing experiments that should be absolutely required to validate study conclusions.
Generally, there should be no more than 3 such required experiments or major modifications for a "Major Revision" recommendation. If more than 3 experiments are necessary to validate the study conclusions, then you are encouraged to recommend "Reject".
Reviewer #1: 1. Fig. 2 shows a very small (<15%) reduction in invasion of LCLs by both S. Typhi and S. Typhimurium after knockdown of ARHGEF26. While Fig. 3 shows a slightly higher inhibition of invasion in Hela cells by S. Typhi (~30%), there is no effect on S. Typhimurium. How do the authors explain this discrepancy? Do both cell types express ARHGEF26 at similar levels? What about DLG1 or SCRIB? Are there differences in expression?
This was an interesting hypothesis that we had not previously considered. We have observed that ARHGEF26 amplifies at a lower CT in RNA from HeLa cells. In fact, measuring ARHGEF26 expression in LCLs is often challenging as it is often present in levels at or below the limit of detection of our qPCR assay. Thus, this could explain the differences between the magnitude of effect between the two cell lines. We have added the following sentence in the section where we introduce the HeLa cell data.
"ARHGEF26 RNAi knockdown (~75% reduction, Figure S1F) significantly reduced the proportion of cells susceptible to S. Typhi invasion ( Figure 3A). Notably this phenotype was larger than we observed with LCL knockdown. While a number of factors could contribute to this difference, we observed that LCLs expressed ARHGEF26 close to the technical limit of our TaqMan assay, while HeLa cells expressed easily detectable levels of ARHGEF26. We therefore hypothesized that increased ARHGEF26 expression in HeLa cells may promote a larger role for the protein in invasion." 2. Even in Hela cells, knockdown of ARHGEF26, DLG1 or SCRIB yielded only small differences in invasion (Fig. 4). Given the important roles of DLG1 and SCRIB in epithelial junction formation, it is surprising that the authors did not examine invasion in a polarized epithelial cell model (e.g. Caco-2 or HT-29 cells).
As mentioned above, we are very grateful for this suggestion. We find that ARHGEF26 and DLG1 are all critical for WT Salmonella Typhimurium and ∆SopBE2 S. Typhimurium uptake into polarized MDCK cells. This data is now included as Figure 6.
3. Dissection of ARHGEF26 suggests that interaction with DLG1 and/or SCRIB is important to its function. Do these interactions stimulate the GEF activity of ARHGEF26?
Previous work (9) has demonstrated that Dlg1 knockdown affects the ability for ARHGEF26 to activate RHOG. The figure from that paper is included below and we have added the following sentence to our description of the DLG1/SCRIB data: However, this has not been demonstrated in the context of Salmonella. As our data do not support a role for RHOG in contributing to Salmonella invasion, we have opted not to pursue examining how DLG1 and SCRIB contribute to Salmonella-induced RHO-GTPase activity until we have identified the causal GTPase connecting ARHGEF26 and Salmonella uptake, which is beyond the scope of this manuscript.
As we responded above, we thank you for raising these concerns about our mouse model descriptions, and we have improved the clarity of the models and their interpretation. The two models should be considered separately, as models of enteric fever and colitis, and each provides novel insight on the critical roles ARHGEF26 has in controlling burden and inflammation during infection.
In the enteric fever model, Salmonella invasion plays a critical role in promoting Salmonella burden in the ileum (10). In this model, ARHGEF26 deletion causes a ~2-log reduction in Salmonella burden in the ileum ( Figure 8B). The magnitude of this effect is equal to the effect of sopB/sopE2 double mutant and is entirely dependent on the presence of these two effectors ( Figure 8C). Further, in the absence of ARHGEF26, sopB/sopE2 double mutant has no effect on bacterial burden in the ileum. This result reveals a central role for ARHGEF26 in the function of sopB and sopE2 in the enteric fever mouse model. "Given our finding that Arhgef26-/mice have substantially reduced inflammation, we were surprised to find that bacterial burden was not affected in the large intestine at late time points during the colitis model ( Figure 9F). We propose four hypotheses for why we observe this result… The fourth hypothesis is that differences in ARHGEF26 expression across the different gastrointestinal compartments could lead to the observation that S. Typhimurium burden is reduced in the ileum, but not the cecum or colon, at four days post infection in the colitis model."

In contrast, Salmonella
Reviewer #2: (1) The in vitro invasion assay using HeLa cells demonstrated that ARHGEF26 contributes to both SopB-and SopE-mediated S. Typhi invasion (Figure 3), while the previous work (Ref 7) showed that ARHGEF26 is involved in SopB-, but not SopE-mediated (S. Typhimurium) invasion. As the genetic background of Salmonella strains (as well as cell-lines) are different between the two works, it is important to ensure that ARHGEF26 works solely on the SopB/SopE axis in the strains used in this study. The double knockout (ΔsopB ΔsopE) S. Typhi strain (which should give no difference between NT and siARHGEF26) needs to be included in Figure 3D as well as in Figure 4C). This strain may give the passive invasion as observed with ΔprgH ( Figure 7C), but should be a required control.

In addition to this experiment, we performed an experiment suggested by reviewer three to test whether moving the sopB and sopE2 genes from S. Typhimurium into S. Typhi could cause S. Typhi to lose its sensitivity to host ARHGEF26 knockdown. Curiously, we found that expressing S. Typhimurium SopB rendered S. Typhi sensitive to ARHGEF26, while S. Typhi expressing SopE2 actually invaded siARHGEF26 transfected cells better than NT transfected cells. Given that knocking out sopE2 does not render S. Typhimurium resistant to ARHGEF26 mediated invasion, we conclude that the effects of ARHGEF26 on HeLa cell invasion is only partially dependent on the SopB/SopE/SopE2 effectors and that other effectors also determine whether Salmonella strains will be sensitive to HeLa cell ARHGEF26 knockdown.
In contrast, in polarized epithelia, we find that the ARHGEF26 and DLG1 S. Typhimurium phenotypes are completely independent of SopB and SopE2, as the invasion deficits are present regardless of the two effectors.
Source: This work (Derived from Figure 6) Further, in our murine enteric fever model, the effects of ARHGEF26 on S. Typhimurium fitness depended on SopB and SopE2. Shockingly, we also found that the effects of SopB and SopE2, in turn, require ARHGEF26.

Together, these data demonstrate a fascinating and complex relationship between ARHGEF26/DLG1/SCRIB and SopB/SopE/SopE2. In some contexts, particularly in HeLa cells and the enteric fever murine model, SopB/E/E2 appear to be critical for ARHGEF26 manipulation and infection. In other systems, particularly in polarized MDCKs (which have recently been described to partially invade by SopB/SopE/SopE2 independent mechanisms (13)) and regulating inflammation, ARHGEF26's contributions appear to be
SopB/SopE2 independent. We have added these data and discussion of these points throughout the text.
(2) (Ref 48), detail of the molecular interaction between ARHGEF26 and DLG1/SCRIB has been analyzed. DLG1-and SCRIB-binding regions seem to locate just upstream of the DH domain of ARHGEF26. I wonder if the 415-871 construct in Figure 5A possesses the ability to interact with DLG1and SCRIB. As the authors propose that SCRIB-DLG1-ARHGEF26 complex guides the GEF to the plasma membrane (lines 699-700), the interaction should be disrupted experimentally by introducing mutations in the binding regions to see if the interaction is essential for enhancing invasion.

We have wanted to perform this experiment for some time, however, the 415-871 construct still retains high ruffling and invasion activity, which we feel is a nonphysiological artifact of overexpression. We show that the proposed mutant (415-871), despite not being able to interact with DLG1 and SCRIB (6), still causes spontaneous membrane ruffling (which we hypothesize is due to its intact catalytic domain and accidental interactions with RHOG and other GEFs that occur in the overexpression system regardless of specific localization) and uptake of non-invasive bacteria.
Source: This Work ( Figure 5) Therefore, because bacterial uptake and spontaneous membrane ruffling cannot be decoupled in this system ( Figure 5), we cannot determine the impact of the ARHGEF26 Nterminus on invasion without direct manipulation of the host genome, which would allow assessing the terminus' function without driving spontaneous membrane ruffling through overexpression. This will be an interesting future direction but is outside the scope of this study.
(3) For clarifying the role of ARHGEH26 in Salmonella invasion, it is crucial to demonstrate the recruitment of the protein to the site of invasion. Infection experiments can be conducted using HeLa cells expressing Myc-tagged ARHGEF26 and its derivatives, not only showing transfection-induced ARHGEF26-positive membrane ruffles ( Figure 5DE).
Once again, we have tried multiple approaches to address this question both before the original submission and during the revision process in light of your review. Regardless of the approach used, we have found that the overexpression system makes it impossible to distinguish between enrichment of ARHGEF26 at the membrane ruffle and the high amounts of ARHGEF26 present in the cytosol (which fills the Salmonella induced ruffle). We have reached this conclusion by overexpressing mCherry and comparing localization to a catalytically dead ARHGEF26-mCherry construct. The catalytically dead construct was chosen because the wild-type construct causes such widespread spontaneous ruffling that Salmonella induced ruffles cannot be distinguished.
As you can see below, with confocal microscopy we see near equal presence of the mCherry florescent protein at the membrane ruffle. Thus, it would be misleading to draw any conclusions about localization from our overexpression system. Source: Unpublished Data. Cells expressing mCherry-C1 or an mCherry tagged CD-ARHGEF26 construct were infected with S. Typhi for 30 minutes, immediately fixed with 4% PFA and 1% glutaraldehyde before mounting. Actin was stained with Alexa Fluor™ 647 Phalloidin (Thermo). Micrographs were taken with the 63x objective on a Zeiss LSM 510 confocal microscope. Z-stacks were compressed into Z-projections by sum slices using Fiji (14).
Given this finding, we conclude that advanced cell biology and genetic techniques to localize endogenous levels of a low abundance protein will be required to understand exactly how ARHGEF26 contributes to invasion, and which domains are involved in that process. While an important question, we believe the complexity of this question leads it to be outside the scope of this paper.
Despite the technical limitations, your suggestion is well taken, and we have removed all mention of DLG1 and SCRIB directing ARHGEF26 to the ruffle. Thus, our discussion of ARHGEF26, DLG1, and SCRIB throughout the paper more accurately reflect our findings. This was a fascinating suggestion that we took great joy in executing. We opted to perform the second described experiment due to the availability of plasmids and strains. Curiously, we found that complementing S. Typhi with S. Typhimurium SopB did not impact the ability for ARHGEF26 knockdown to impact S. Typhi invasion. In contrast, complementing SopE2 did overcome ARHGEF26 knockdown and actually resulted in increased invasion compared to non-targeting. This was slightly unexpected, as we showed in the original submission of this manuscript that knocking out SopE2 in S. Typhimurium did not restore ARHGEF26 sensitivity. Together, these data demonstrate that that while the SopE vs SopE2 effector regiments in this study help to explain differences in strain dependence on ARHGEF26 in HeLa cells, they are not the only contributors to the phenotype.
Source: This work (Derived from Figure 3) 2. The in vivo role of ARHGEF26 was determined by infecting Arhgef26-/-mice using two different mouse models, the enteric fever murine model and the murine gastroenteritis infection model. Arhgef26-/-mice have reduced bacterial numbers in ileum and spleen (Fig 7), and this is dependent on a functional T3SS-1, since a DprgH mutant no longer showed reduced bacterial numbers in Arhgef26-/mice compared to wild type mice. A DsopB mutant, however, still had significantly lower numbers in the ileum, suggesting S. Tm SopB does not contribute to ARHGEF26-dependent bacterial colonization. It is unclear why the authors state that this is dependent on SopB (lines 601). Is this conclusion solely based on the reduced ratio (0.04) compared to mice infected with wild type Salmonella (0.004)? What is the bacterial load of the DsopB mutant in the spleen of wild type and Arhgef26-/-mice? And does SopE play a significant role in the murine enteric fever model?
Thank you for pointing out that this is unclear in the text. Your interpretation is correct that we were referring to the change in ratio. However, in light of our new data showing that the phenotype is dependent on SopB and SopE2, we agree that referring to the ARHGEF26 phenotype as being partially SopB-dependent based on this minor shift is distracting. We have now revised the text to reflect this.
"Supporting that a functional SPI-1 secretion system is required for ARHGEF26 to affect S. Typhimurium fitness, the difference in burden between wild-type and Arhgef26-/-mice was significantly reduced when mice were infected with ∆prgH bacteria ( Figure 7C). In contrast, differences between wild-type and Arhgef26-/-mice were slightly reduced but broadly maintained when infected with a ∆sopB S. Typhimurium strain, once again demonstrating that the previous model that ARHGEF26 contributes specifically to SopB-mediated fitness is incomplete." Regarding the spleen, the phenotype in the spleen is very small given the noise we observe and is therefore very difficult to make mechanistic conclusions about, which is why we did not include the data displayed below in the manuscript. With the moderate number of mice used in these experiments we did not see statistically significant differences or ratios resembling what we see in ileum where the phenotype is substantially larger. Further, the very well-defined role for SPI-1 in the ileum and not systemic sites makes the ileum the more relevant site to examine for sopB dependence.
Source: Unpublished data, paired with Figure 7C Furthermore, in light of your suggestion and suggestions for Reviewer 2, we have performed experiments using the ∆sopB∆sopE2 S. Typhimurium mutant. Excitingly, we found that the entire ARHGEF26 phenotype depends on ∆sopB∆sopE2. Perhaps even more unexpected, we found that, in turn, the entire ∆sopB∆sopE2 phenotype depends on ARHGEF26. This suggests that despite the GEF activity of SopE2, it requires ARHGEF26 involvement in order to induce S. Typhimurium fitness in the ileum. 3. In the gastroenteritis model is there a reduction in cytokine (eg. KC) production that would correlate with the findings in HeLa cells? It has been demonstrated that Salmonella-induced inflammation is beneficial for Salmonella to grow to high numbers. Does the reduction in total pathology scores result in reduced bacterial colonization in the Arhgef26-/-mice at later days post infection (days 3 and 4)? And do the effector proteins SopB and/or SopE contribute this ARHGEF26-dependent inflammation?
Due to the COVID-19 pandemic, we have been limited in the amount of mouse breeding that was able to occur over the last 15 months. To try to minimize the numbers of mice we needed for this revision process, we attempted to collect fecal biomarkers of inflammation that would allow us to measure inflammation while also measuring bacterial burden and/or pathology. Therefore, in order to address this suggestion, we examined secreted lipocalin-2 (15). Unfortunately, we did not observe statistically differences between wild-type and ARHGEF26 mice at two days post infection due to very high variation in the assay. We predict that if we had measured luminal cytokines in sufficiently large numbers, we would have been able to generate corroborating data for the pathology differences which were collected in a blinded fashion and are unequivocal. In the text we provide a number of explanations for why we observed this result, and have reproduced that section here: "Given our finding that Arhgef26-/-mice have substantially reduced inflammation, we were surprised to find that bacterial burden was not affected in the large intestine at late time points during the colitis model ( Figure 9F). We propose four hypotheses for why we observe this result. The first is that compared to previous work on inflammation-promoted bacterial burden (65) Finally, regarding your question regarding ∆sopB∆sopE2 influencing the colitis model phenotype, we find that ARHGEF26 actually reduces pathology in all three tissues (ileum, cecum, colon) more than we observed ∆sopB∆sopE2 reducing pathology, and that these reductions appear to be at least largely independent (full independence cannot be assessed due to reaching the limit of detection of the assay).
Source: This work (Derived from Figure 9)

Part III -Minor Issues: Editorial and Data Presentation Modifications
Please use this section for editorial suggestions as well as relatively minor modifications of existing data that would enhance clarity.

Reviewer #1: (No Response)
Reviewer #2: (1) I wonder why the ΔsopB S. Typhi strain shows the higher invasion late than the wildtype strain in HeLa infection experiments ( Figure 3D, Figure 4CD). Are there any reasonable explanations?
We have observed that the ∆sopB mutant typically invades equal to slightly higher than the wild-type strain, though occasionally we do see less invasion. Figures 3D (Now Figure 3B) and the experiments for Dlg1 knockdown in 4C were from the same experiments where ∆sopB did, by chance, have some of the most consistent increased invasion we've seen. As a result, it did have a significant p-value by a t-test.
By contrast, the experiments matched with siSCRIB knockdown did not meet significance, though there was slightly smaller N for these experiments (6 replicates per group vs 9 in the experiment above) Source: This work (Derived from Figure 4) Because the phenotype is, typically, not consistent, we do not comment on the phenotype in the study. From a biological perspective, knocking out sopB may allow for a greater amount of secretion of other effectors, such as sopE2, that may be better at stimulating invasion.
(2) The authors did not get experimental data supporting that RhoG has any roles in Salmonella invasion and in "infection-induced" IL-8 production. I wonder if the authors think that the genetic diversity of SopE in the Salmonella strains can cause the discrepancy between studies (lines 720-722).
This idea of SopE/E2 diversity contributing to phenotypes observed in this study was explored through a major experiment suggested by reviewer three above. We "switched" the S. Typhimurium sopE2 gene into S. Typhi and found that it did render S. Typhi insensitive to host ARHGEF26 knockdown-mediated invasion resistance. However, knocking out SopE2 in S. Typhimurium did not restore ARHGEF26 sensitivity, suggesting that while SopE/E2 diversity might help to explain differences in strain dependence on ARHGEF26 in HeLa cells, it is not the only contributor.
Source: This work (Derived from Figure 3) In terms of RHOG, SopE/E2 diversity is unlikely to explain differences between our study and the Patel and Galán 2006 study, as in their work they used both wild-type bacteria and a ∆sopE∆sopE2 mutant and showed an effect on RHOG membrane ruffling (5), though their work on invasion was only performed with a wild-type (SopB+SopE+SopE2+) strain (16). In terms of why they observed this phenotype while we do not, we propose three non-mutually exclusive hypotheses: 1 (3) I assume that Figure 4 experiments were conducted using HeLa cells, but I cannot find the description.
We have added "HeLa cells" to the figure title and legend.
(4) Description of Figure S1 (D) is missing in the legend.
We have corrected the figure legend.
Reviewer #3: 1. In Figure 3D, it seems that the DsopB strain is more invasive compared to the S. Typhi wild type strain Ty2 in both NT and siARHGEF26 HeLa cells. Is this difference significant, and if so, how can this be explained? It shows that SopE-dependent invasion is more efficient (when SopB is deleted), also in the absence of ARHGEF26.
This was also asked above by Reviewer 2. We have copied the response, exactly, here for convenience: We have observed that the ∆sopB mutant typically invades equal to slightly higher than the wild-type strain, though occasionally we do see less invasion. Figures 3D (Now Figure 3B) and the experiments for Dlg1 knockdown in 4C were from the same experiments where ∆sopB did, by chance, have some of the most consistent increased invasion we've seen. As a result, it did have a significant p-value by a t-test.
By contrast, the experiments matched with siSCRIB knockdown did not meet significance, though there was slightly smaller N for these experiments (6 replicates per group vs 9 in the experiment above).
Source: This work (Derived from Figure 4) Because the phenotype is, typically, not consistent, we do not comment on the phenotype in the study. From a biological perspective, knocking out sopB may allow for a greater amount of secretion of other effectors, such as sopE2, that may be better at stimulating invasion.
2. Would a DsopB/sopE S. Typhi mutant invade NT and siARHGEF26 cells similarly? Or are there other effectors also required for ARHGEF26-dependent invasion?
This experiment was requested and explained above. We have copied the answer here for convenience.

While it is difficult to measure Salmonella uptake in ∆sopB∆sopE S. Typhi by HeLa cells due to their inefficient invasion, we infected HeLa cells at a high MOI and measured ~0.5% infection, which did not vary between NT and siARHGEF26 conditions, suggesting that in HeLa cells the ARHGEF26 infection decrease is occurring by SopB/SopEmediated mechanisms.
Source: This work (Derived from Figure 3) 3. Figure 6A shows that silencing of ARHGEF26 reduces IL-8 production in unstimulated cells by approximately 2-fold. SiARHGEF26 cells infected with S. Typhi of S. Tm also had reduced IL-8 production. Is this moderately reduced IL-8 production a result of the lower basal IL-8 production, or is it an effect of Salmonella-induced inflammation. Did the authors correct for the IL-8 reduction in unstimulated cells?
The magnitude of the siARHGEF26-mediated cytokine reduction for both S. Typhi and S. Typhimurium induced IL-8 was ~1,400pg/mL and ~700pg/mL, accordingly. In contrast, the absolute change in uninfected IL-8 abundance was only ~200pg/mL. Thus, while our data are not background subtracted, the differences in basal levels cannot explain our differences following stimulation. Further demonstrating this point, while we see that siARHGEF26 leads to changes in unstimulated and stimulated amounts of IL-8, we see that siRHOG only impacts unstimulated levels, demonstrating specificity in our system.
We have added discussion of this in the paper: "Importantly, the magnitude of the siARHGEF26-mediated cytokine reduction for both S. Typhi and S. Typhimurium induced IL-8 was much larger than the reduction seen in uninfected cells (~1,400pg/mL and ~700pg/mL, accordingly vs ~200pg/mL. This substantially larger phenotype with infection leads us to conclude that our results are not simply consequences of lower basal levels of cytokine production following ARHGEF26 knockdown. Instead, we propose that there are multiple levels of cytokine regulation by ARHGEF26, and that this could have important implications during Salmonella infection."