The type 3 secretion effector IpgD promotes S. flexneri dissemination

The bacterial pathogen Shigella flexneri causes 270 million cases of bacillary dysentery worldwide every year, resulting in more than 200,000 deaths. S. flexneri pathogenic properties rely on its ability to invade epithelial cells and spread from cell to cell within the colonic epithelium. This dissemination process relies on actin-based motility in the cytosol of infected cells and formation of membrane protrusions that project into adjacent cells and resolve into double-membrane vacuoles (DMVs) from which the pathogen escapes, thereby achieving cell-to-cell spread. S. flexneri dissemination is facilitated by the type 3 secretion system (T3SS) through poorly understood mechanisms. Here, we show that the T3SS effector IpgD facilitates the resolution of membrane protrusions into DMVs during S. flexneri dissemination. The phosphatidylinositol 4-phosphatase activity of IpgD decreases PtdIns(4,5)P2 levels in membrane protrusions, thereby counteracting de novo cortical actin formation in protrusions, a process that restricts the resolution of protrusions into DMVs. Finally, using an infant rabbit model of shigellosis, we show that IpgD is required for efficient cell-to-cell spread in vivo and contributes to the severity of dysentery.


Author summary
The intracellular pathogen Shigella flexneri is the causative agent of bacillary dysentery (blood in stool). Invasion of epithelial cells and cell-to-cell spread are critical determinants of S. flexneri pathogenesis. Cell-to-cell spread relies on the formation of membrane protrusions that project into adjacent cells and resolve into vacuoles. The molecular mechanisms supporting this dissemination process are poorly understood. In this study, we show that S. flexneri employs the phosphatidylinositol phosphatase activity of the T3SS effector protein IpgD to manipulate phosphoinositides in the protrusion membrane. Manipulation of phosphoinositide signaling restricts the formation of actin networks underneath the protrusion membrane, which would otherwise prevent the scission of protrusions into vacuoles. We also demonstrate that IpgD is required for efficient dissemination in the colon of infant rabbits and contributes to the severity of disease. This study exemplifies how manipulation of phosphoinositide signaling by intracellular pathogens supports bacterial pathogenesis.

Introduction
The intestinal pathogen Shigella flexneri is the causative agent of bacillary dysentery [1]. The disease is characterized by a dramatic ulceration of the colonic mucosa, massive infiltration of immune cells, and blood in stool [2]. There are 270 million cases of bacillary dysentery annually in the world, resulting in more than 200,000 deaths, especially in children under age five [3]. Until recently, infected patients were easily cured with antibiotic treatment. However, the isolation of multiple antimicrobial-resistant strains from patients is becoming the norm worldwide [4]. Seminal studies conducted in non-human primates have revealed that bacillary dysentery is associated with bacterial invasion of epithelial cells in the colon [5]. The development of in vitro tissue culture systems [6] and genetic approaches [7] led to the discovery that S. flexneri invasion properties rely on the type 3 secretion system (T3SS) [8], which delivers a panel of effector proteins that manipulate the actin cytoskeleton, leading to bacterial engulfment into primary vacuoles [9]. As they escape primary vacuoles, bacteria exploit the actin cytoskeleton to display actin-based motility in the cytosol of infected cells [10]. S. flexneri expresses a surface protein, IcsA [11,12], which recruits N-WASP [13], a host cell actin nucleation-promoting factor [14]. N-WASP in turn recruits the ARP2/3 complex, which leads to actin nucleation at the bacterial pole [15]. The forces generated by actin network assembly propel the bacteria throughout the cytosol of infected cells. As motile bacteria encounter cell-cell contact, they form membrane protrusions that project into adjacent cells [16]. The resolution of these protrusions leads to the formation of double-membrane vacuoles [17,18], from which the pathogen escapes by deploying its T3SS and specific effector proteins [19][20][21]. Although S. flexneri clearly relies on host factors such as N-WASP and the ARP2/3 complex to display actin-based motility, it has long been unclear whether the formation and resolution of membrane protrusions into DMVs is a mere consequence of actin-based motility or is indeed supported by specific host cell processes. Recently, however, several studies have highlighted the existence of specific mechanisms supporting S. flexneri cell-to-cell spread beyond the requirement of actin-based motility [22]. The proper formation of protrusions relies on Myosin-X and formins [23,24]. Moreover, the resolution process is dependent on host PI 3-kinase activity and subsequent noncanonical clathrin-mediated endocytosis [25]. Unlike Listeria monocytogenes, S. flexneri protrusions resolve into DMVs via formation of intermediate membrane compartments termed vacuole-like protrusions (VLPs), which requires tyrosine kinase and phosphoinositide signaling pathways [17,18,26]. In addition to host cell factors, the formation of protrusions relies on the T3SS translocase IpaC, whose insertion in the plasma membrane releases membrane tension at cell-cell contacts [27].The T3SS also supports the resolution of membrane protrusions into DMVs [18], through mechanisms that remain to be determined.
Here, we show that the resolution of membrane protrusions is facilitated by the T3SS effector IpgD. IpgD manipulates phosphoinositide signaling in protrusions, which prevents de novo formation of cortical actin in membrane protrusions, thereby facilitating the resolution of protrusions into DMVs. We also demonstrate that IpgD contributes to the efficiency of S. flexneri dissemination and severity of symptoms in vivo, using an infant rabbit model of shigellosis [28].

The T3SS effector IpgD is required for S. flexneri dissemination in HT-29 cells
To address the role of the T3SS in S. flexneri dissemination, we investigated a collection of T3SS effector mutants for their ability to spread from cell to cell. We identified ipgD as a potential candidate whose deletion caused 35% reduction in the size of the infection foci formed in HT-29 cells (Fig 1A). In vitro and in vivo studies have shown that IpgD is a phosphatidylinositol 4-phosphatase [29]. To determine whether the phosphatidylinositol 4-phosphatase activity of IpgD is required for efficient cell-to-cell spread, we compared the size of the infection foci formed in HT-29 cells infected with the ΔipgD mutant expressing wild type (pIpgD) or catalytically dead IpgD (pIpgD C438S ). The spreading defect observed in cells infected with the ΔipgD mutant was complemented by expression of wild type IpgD, but not by expression of IpgD C438S (Fig 1A). These results suggest that the phosphatidylinositol 4-phosphatase activity of IpgD is required for S. flexneri spread from cell to cell.
IpgD was characterized as a T3SS effector protein that supports the formation of large membrane ruffles at sites of invasion in HeLa cells [30]. The role of IpgD in the invasion Circles represent data points and colors indicate independent biological replicate groups (A-E). Statistics: Panels A-C, one-way ANOVA followed by Tukey's multiple comparisons (WT, control); Panel D and E, unpaired t-test analysis; ���� P < 0.0001; NS, not significant. https://doi.org/10.1371/journal.ppat.1010324.g001

T3SS in Shigella dissemination
process per se is however controversial. Initial studies reported that the IpgD mutant was as invasive as the isogenic M90T parental strain [31]. However, recent studies using the 2457T strain reported a role for IpgD in invasion of HeLa cells through ARF6 activation [32]. In HT-29 cells, the ΔipgD strain was as invasive as the isogenic 2457T parental strain, as determined by infection foci counts ( Fig 1B) and gentamicin protection assay (Fig 1C).
In HeLa cells, an ipgD mutant was reported to exhibit a delayed escape from primary vacuoles within 30 min after invasion [33].To assess whether a delayed escape from primary vacuoles could account for the apparent cell-to-cell spreading defect displayed by the ΔipgD strain ( Fig  1A), we quantified the total number of bacteria per infection focus and the proportion of bacteria displaying actin tails 4 hours post-infection (hpi), as a readout of cytosolic bacteria. We observed similar bacterial counts and proportions of actin tails in cells infected with the wild type and ΔipgD strains ( Fig 1D). Furthermore, the wild type and ΔipgD strains formed actin tails of similar length (Fig 1E), indicating similar cytosolic motility, since the length of actin tails correlates with the speed of cytosolic bacteria [34]. Collectively, these data suggest that, in HT-29 cells, (i) the ΔipgD strain is as invasive as the wild type strain, (ii) delayed vacuole escape, if any, does not impact the number of cytosolic bacteria, and (iii) cytosolic bacteria display normal actin tails 4 hpi. These results therefore suggest that the impaired cell-to-cell spread of the ΔipgD strain probably originates from a defect in a subsequent step during the dissemination process, including membrane protrusion formation/resolution, and/or DMV escape.

IpgD supports membrane rotrusion resolution during S. flexneri cell-tocell spread
Our previous work revealed that S. flexneri spreads from cell to cell through the sequential formation of protrusions, vacuole-like protrusions (VLPs) and double-membrane vacuoles (DMVs), from which the pathogen escapes to gain access to the cytosol of adjacent cells [17] (Fig 2A). To determine which dissemination step(s) may be compromised in cells infected with the ΔipgD strain, we conducted time-lapse confocal microscopy of HT-29 cells expressing plasma membrane-targeted yellow fluorescent protein (mbYFP) infected with CFP-expressing S. flexneri. Tracking results showed that the majority of wild type bacteria (69%) successfully spread to adjacent cells through sequential formation of protrusions, VLPs and DMVs, and proceeded to DMV escape (Figs 2B-2D and S1 and S1 Movie). In stark contrast, the majority of ΔipgD bacteria (72%) failed to spread to adjacent cells ( Fig 2D). Although ΔipgD bacteria were able to form protrusions, the majority of the formed protrusions did not transition into the VLP stage, and ultimately collapsed back to primarily infected cells (Figs 2B-2D and S1 and S2 Movie). In the formed protrusions, the ΔipgD strain recruited comparable levels of ARP2/3 at the bacterial pole, suggesting that the ΔipgD mutant displayed normal actin-based motility in protrusions (S2 Fig). We noted that the reliance on IpgD for efficient spread was not absolute, and 26% of the bacteria successfully spread (Figs 2C and 2D and S1), suggesting the existence of redundancy and potential alternative mechanisms in this system. In those instances where the ΔipgD mutant formed DMVs, the time spent in DMV until escape was similar to wild type strain, showing that IpgD is not required for DMV escape (S3 Fig). In addition, wild type and mutant strains exhibited no difference in the time spent in protrusions until successful protrusion resolution or protrusion collapse (S3 Fig). Altogether, these results suggest that IpgD is specifically required for the step of protrusion resolution during S. flexneri spread from cell to cell.

PtdIns(4,5)P 2 clearance correlates with protrusion resolution success
Since our previous work showed a role for the T3SS in protrusion resolution [18] and previous studies showed that the T3SS effector protein IpgD hydrolyzes PtdIns(4,5)P 2 [29], we PLOS PATHOGENS T3SS in Shigella dissemination hypothesized that the regulation of PtdIns(4,5)P 2 levels in protrusions is critical for protrusion resolution. To test this hypothesis, we generated a HT-29 mbYFP cell line stably expressing mCherry fused to the PH domain of PLCδ (herein referred to as the mCherry-PH probe), which specifically binds PtdIns(4,5)P 2 [35,36]. We conducted time-lapse confocal imaging  experiments using CFP-expressing bacteria and mixed populations of HT-29 mbYFP/ mCherry-PH and HT-29 mbYFP cells. We used computer-assisted image analysis to quantify the signal levels corresponding to the mCherry-PH probe at the plasma membrane surrounding S. flexneri in protrusions projecting from HT-29 mbYFP/mCherry-PH-positive cells into HT-29 mbYFP-positive/mCherry-PH-negative cells (Fig 3A, white stars). Probe levels at protrusion membranes were corrected for local background (Fig 3A, white squares) and normalized with respect to the probe signal at cell-cell contact (Fig 3A, white diamonds), to account for cell-to-cell variation in probe expression levels. Tracking protrusions using the mbYFP probe allowed us to determine whether a given protrusion would fail to resolve into a DMV and collapse ( (Fig 2), these results indicate that IpgD-mediated clearance of PtdIns(4,5)P 2 is critical for protrusion resolution. We however note that PtdIns(4,5)P 2 clearance occasionally occurs in the protrusions formed in cells infected with ΔipgD

de novo cortical actin formation correlates with protrusion resolution failure
PtdIns(4,5)P 2 levels play a critical role in the assembly and dynamics of actin networks at the plasma membrane [37,38], herein referred to as cortical actin. Cortical actin is a critical determinant of cell shape. However, in the context of cell-to-cell spread, the presence of cortical actin in protrusions may impede protrusion scission into DMVs, leading to defective cell-tocell spread. To investigate cortical actin levels during S. flexneri cell-to-cell spread, we generated a HT-29 mbYFP cell line expressing β-actin fused to mCherry (mCherry-Actin). As expected, the probe accumulated at the bacterial pole and in the protrusion neck, reflecting actin-based motility in protrusions. Importantly, we also observed cortical actin underneath the plasma membrane surrounding the bacteria in protrusions (S5 Fig). Correlative analysis on fixed samples demonstrated that the probe mainly reports on polymerized actin (S5 Fig,  regression plot). To determine the dynamics of cortical actin in protrusions, we conducted time-lapse confocal imaging to quantify changes in mCherry-Actin levels at protrusion plasma membranes surrounding CFP-expressing bacteria, where cortical actin visualization is not obscured by the actin tails generated at the bacterial pole in the neck of protrusions. We applied the same quantification method described for mCherry-PH probe (Fig 3A and 3B) to determine changes in mCherry-Actin levels. All nascent protrusions displayed low cortical actin levels (Fig 4A and 4B). At later time points, we observed situations in which low levels of cortical actin were maintained (Fig 4A, WT,

PLOS PATHOGENS
T3SS in Shigella dissemination spread and S3 Movie). These results show that the formation of nascent protrusions occurs as motile bacteria pierce through the existing cortical actin at sites of cell-cell contact, leading to deformation of the plasma membrane into a low cortical actin membrane-bound compartment that protrudes into adjacent cells. As protrusions elongate, IpgD confers to the bacteria the ability to restrict de novo cortical actin formation, which facilitates the resolution of protrusions into DMVs.

PtdIns(4,5)P 2 levels correlates with cortical actin levels
To further establish the correlation between the regulation of PtdIns(4,5)P 2 levels and de novo cortical actin formation in protrusions, we carried out infection experiments with CFPexpressing S. flexneri strains and monolayers of mixed HT-29 mbCFP/ mCherry-PH and HT-29 mbCFP cells. Using fixed samples, we assessed the co-localization of mCherry-PH probe with F-actin underneath the plasma membrane surrounding bacteria in protrusions (S6 Fig). The approach demonstrated a statistically significant relationship between mCherry-PH probe and F-actin signal levels for both strains (Fig 5A, regression plot, P<0.0001), showing that high levels of PtdIns(4,5)P 2 correlate with high levels of cortical actin in membrane protrusions. In support of the notion that cortical actin restricts the spread of the ΔipgD strain, we detected significantly higher mCherry-PH and F-actin levels in protrusions formed by the ΔipgD strain compared to wild type bacteria (Fig 5B and 5C). Collectively, these results support the notion that the 4-phosphatidylinositol phosphatase activity of IpgD decreases the levels of PtdIns(4,5)P 2 in protrusions, which prevents de novo cortical actin formation, and facilitates the resolution of protrusions into vacuoles.

IpgD is required for efficient cell-to-cell spread in vivo
Recent studies have established that many features of human shigellosis, including epithelial cell invasion and cell-to-cell spread, can be recapitulated in the colon of infant rabbits. [28,39].
To determine the role of IpgD in dissemination in vivo, we compared the size of infection foci formed by wild type and ΔipgD strains in the colon of infant rabbits 8 hpi. Similar to the results obtained in HT-29 cells (Fig 1), the ΔipgD strain formed significantly smaller infection foci compared to the wild type strain (Fig 6A). We also confirmed that IpgD is not essential for bacterial invasion, as the numbers of infection foci per colon and the number of bacteria per infection focus were comparable for wild type and ΔipgD strains 4 hpi (Fig 6B, detail quantification method in S7 Fig). We also investigated a potential role of IpgD in nonphagocytic cell apoptosis [40,41].To this end, we compared cell death as assessed by TUNEL assay in the colon of animals infected with wild type and ΔipgD strains. We observed no differences in the rate of cell death in either un-infected or infected regions, regardless of the bacterial strains used for infection (S8 Fig), suggesting that cell death is probably not the cause of the cell-tocell spreading defect observed in animals infected with the ΔipgD strain. Collectively, these results show that, similar to the results obtained in HT-29 cells, IpgD promotes cell-to-cell spread in vivo.

Role of IpgD in pathogenesis
We have recently determined that cell-to-cell spread is a critical determinant of S. flexneri pathogenesis [28]. Importantly, the severity of epithelial fenestration (absence of luminal epithelium) and bloody diarrhea was contingent on efficient cell-to-cell spread. We therefore evaluated the impact of the observed cell-to-cell spreading defect in animals infected with the ΔipgD strain on pathogenesis by measuring the level of epithelial fenestration and scoring dysentery (described in Materials and methods). In agreement with the role of cell-to-cell spread in ulceration, histopathological examination indicated that the ΔipgD mutant caused significantly less epithelial fenestration (Fig 7A and 7B). In addition, animals infected with ΔipgD bacteria exhibited significantly weaker bloody diarrheal symptoms than animals infected with wild type bacteria (Fig 7C and 7D). These results indicate that the ΔipgD mutant is attenuated in vivo, and further establish the importance of cell-to-cell spread in pathogenesis.

Discussion
Several intracellular pathogens manipulate the actin cytoskeleton to display actin-based motility and form membrane protrusions at cell-cell contacts that project into adjacent cells [22,42].

T3SS in Shigella dissemination
The existence of structures such as cortical actin at cell-cell contacts has been suggested to act as a barrier to dissemination [43,44]. Listeria monocytogenes has evolved specific mechanisms to counter global restriction through secretion of Internalin C, a bacterial factor that overcomes membrane tension at cell-cell contacts by interfering with TUBA and N-WASP interactions [44]. Similarly, Rickettsia parkeri facilitates protrusion formation through secretion of Sca4, a T4SS effector protein that mediates global reduction in actomyosin cortical tension [43]. Moreover, recent studies have shown that Shigella flexneri utilizes the T3SS translocase IpaC to decrease membrane tension [27]. Our results indicate that S. flexneri forms protrusions by piercing through cortical actin, deforming the plasma membrane, and forming a membrane bound compartment with low cortical actin levels (Fig 4). Our results also indicate that de novo cortical actin formation in protrusions represents a barrier to protrusion resolution and therefore emerges as a potential mechanism of restriction on the dissemination

PLOS PATHOGENS
T3SS in Shigella dissemination process (Fig 4 and S3 Movie). We note that local restriction on protrusion resolution is conceptually different from global restriction on protrusion formation through membrane tension, as previously suggested for L. monocytogenes, R. parkeri and S. flexneri [43,44].
Our results indicate that S. flexneri manipulates PtdIns(4,5)P 2 levels to interfere with de novo cortical actin formation (Fig 3). It is likely that de novo cortical actin formation in protrusions may also restrict the dissemination of pathogens that spread through protrusion formation, including L. monocytogenes. Whether these pathogens resolve de novo cortical actin restriction through manipulation of PtdIns(4,5)P 2 signaling remains to be determined. Interestingly, the resolution of L. monocytogenes protrusions requires the expression of the bacterial metalloprotease Mpl [45]. It is thus possible that L. monocytogenes resolves cytoskeleton Our results indicate that S. flexneri utilizes the T3SS effector protein IpgD to manipulate PtdIns(4,5)P 2 levels and cortical actin in membrane protrusions (Figs 3 and 4). The role of IpgD was first characterized in the context of studies focused on elucidating the mechanisms supporting the invasion process [31]. The phosphatidylinositol 4-phosphatase activity of IpgD was demonstrated in vitro and in vivo [29]. The connection between the activity of IpgD and the dynamics of the actin cytoskeleton was established through the observation that the ruffles formed at sites of entry appear smaller in cells infected with the ipgD mutant [30]. Moreover, when expressed in cells, IpgD lessens plasma membrane-cortical actin interactions, leading to morphological alterations in cells owing to reduced membrane tether forces [29]. In spite of these obvious connections to the dynamics of the actin cytoskeleton, IpgD is not required for efficient invasion in non-polarized cells [31], a result that we have confirmed in polarized cells (Fig 1B-1D) and in the colons of infant rabbits (Fig 6B). Thus, our results in the context of cell-to-cell spread provide an important demonstration of the functional importance of IpgD in cytoskeleton manipulation.
Our results show that S. flexneri expresses IpgD in order to hydrolyze PtdIns(4,5)P 2 at the plasma membrane of the protrusions that project from primarily infected cells. Since PtdIns (4,5)P 2 is displayed on the inner leaflet of the plasma membrane, this suggest that IpgD is secreted into the cytosolic compartment of protrusions, as opposed to translocated across the membrane. We note that the local secretion of IpgD in protrusions is consistent with the notion that the activity of the S. flexneri T3SS is differentially regulated depending on the cellular localization of the pathogen, with critical activation observed in protrusions and DMVs [46]. We have previously shown that the T3SS is required for protrusion resolution and DMV escape [18]. Although specific T3SS effector proteins, such as IcsB, support DMV escape [19,20], T3SS effector proteins that support protrusion resolution remained elusive. Thus, IpgD is the first example of a T3SS effector protein displaying function in protrusions (Fig 2) and not in DMVs (S3 Fig) during S. flexneri cell-to-cell spread. Our work therefore supports the notion that S. flexneri cell-to-cell spread is a multi-step process that is orchestrated in space and time by the T3SS [18,21].
Our results show that the role of IpgD in facilitating protrusion resolution has a significant impact on infection foci formation in tissue culture cells (Fig 1) and in the colon of infant rabbits ( Fig 6A). Consistent with our previous work demonstrating the critical role of cell-to-cell spread in pathogenesis [28], the ΔipgD mutant conferred milder symptoms of disease, including epithelial fenestration and bloody diarrhea (Fig 7). Our results therefore support the notion that inhibiting the activity of bacterial T3SS effector proteins supporting dissemination, such as IpgD, or modulating the host cell processes hijacked by these effector proteins, may alter the course of disease and may constitute therapeutic alternatives to antibiotic treatment [47].

Ethics statement
All experiments described in this study were reviewed and approved by the University of Virginia Institutional Biosafety Committee and the Institutional Animal Care and Use Committee (protocol #4161).

Bacterial strains, cell lines and growth conditions
The wild type (WT) S. flexneri strain used in this study is S. flexneri 2457T [6]. The ΔipgD strain was generated by replacement of the coding region with a kanamycin cassette by PLOS PATHOGENS T3SS in Shigella dissemination homologous recombination, as previously described [48]. For complementation purposes, ΔipgD was transformed with pACYC184-IpgD and pACYC184-IpgD C438S kindly provided by Dr. Marcia Goldberg [32]. The pMMB207 plasmid harboring the gene encoding cyan fluorescent protein (pCFP) under the control of an isopropyl-β-D-thiogalactopyranoside (IPTG)inducible promoter was introduced into S. flexneri strains by electroporation, resulting in WT pCFP and ΔipgD pCFP. LB medium with appropriate antibiotics was used to culture Escherichia coli strains (DH5α and Δnic35) for cloning steps. Primers and generated constructs are listed in Tables A and B in S1 Text, respectively. Exponential phase S. flexneri strains cultured at 37˚C were used for infecting cell monolayers. S. flexneri and S. flexneri pCFP glycerol stocks were re-streaked on LB agar containing 10 μg/ml Congo red dye (Fisher Chemical) supplemented with 10 μg/ml chloramphenicol when necessary and incubated at 37˚C. The next day, a single Congo red positive colony of each strain was transferred to 2 ml LB (-/+ chloramphenicol) for overnight growth at 30˚C. Overnight cultures were transferred to fresh LB (-/+ chloramphenicol) and cultured to exponential phase (37˚C) to infect cell monolayers. Tryptic Soy Broth (TSB) supplemented with 10 μg/ml chloramphenicol was used to culture S. flexneri pCFP strains for 12-13 h at 37˚C on a rotating wheel for animal infections. The growth of HT-29 cell lines was carried out in McCoy's 5A medium (Gibco) containing 10% heat inactivated fetal bovine serum (hiFBS) (Invitrogen) at 37˚C with 5% CO 2 . HEK 293 cells were cultured in DMEM (Gibco) supplemented with 10% hiFBS. Lentiviruses were produced from HEK 293 cells to generate HT-29 cell lines stably expressing either cyan or yellow fluorescence proteins targeted to the plasma membrane (mbCFP or mbYFP), mCherry fused to the PH domain of PLCδ (referred to as the mCherry-PH probe), or mCherry-fused to β-actin (mCherry-Actin), using packaging vectors pCMVΔ8.2Δvpr and pMD2.G [49]. The pMXsIP-mCherry vector was used for cloning; corresponding primers and plasmids are listed (Tables A and B in S1 Text), respectively.

Infection of HT-29 cell monolayers and image analysis
Cell-to-cell spread phenotypes were tested by infecting confluent HT-29 cell monolayers in 96-well plates (Corning) with exponential phase S. flexneri. Following the delivery of S. flexneri inoculated media to each well, 96-well plates were centrifuged (1000 rpm, 5 min, room temperature) to initiate infection. Plates were incubated at 37˚C with 5% CO 2 for 1 h, then fresh McCoy's medium supplemented with gentamicin was added to eliminate extracellular S. flexneri (final gentamicin concentration, 50 μg/ ml). Plates were incubated for 15 h before paraformaldehyde fixation (4%, 20 min, room temperature). Infection foci were detected by immunostaining using a primary unconjugated rabbit S. flexneri antibody (ViroStat, 1:1000, 90 min, room temperature) and secondary goat anti-rabbit IgG conjugated with Pacific blue (Invitrogen, 1:1000, 90 min, room temperature). Images were acquired with ImageXpress Micro imaging system (Molecular Devices). The size of each S. flexneri infection focus was measured with the Region function of the MetaMorph software. Three independent biological replicates were performed and 50 foci per strain were analyzed for each biological replicate. To assess invasion phenotype, foci count per image frame was determined using the same images from infection foci size experiments. As another measure for invasion, gentamicin protection assays were conducted in 96-well plates to determine the number of intracellular bacteria. Following the invasion period (37˚C, 5% CO 2 , 1 h), the medium was replaced with gentamicinsupplemented medium (3 wells per strain). After 1 h in the presence of gentamicin, wells were washed with DPBS and cells were lysed in 0.1% triton-X on ice for 12 min. After lysis, serial dilutions were plated on LB Congo red plates and plates were incubated at 37˚C overnight for colony forming unit determination.

Live imaging
Time lapse confocal microscopy was applied to characterize the dynamics of S. flexneri pCFP spread, PtdIns(4,5)P 2 and cortical actin at protrusion membranes in monolayers of HT-29 mbYFP, mixed populations of HT-29 mbYFP and HT-29 mbYFP/mCherry-PH (1:1), and mixed populations of HT-29 mbYFP and HT-29 mbYFP/mCherry-actin (1:1), respectively. Confluent monolayers of cells in 8-chamber coverglasses (Lab-TEK II, Thermo Fisher) were infected with S. flexneri pCFP. Coverglasses were centrifuged (800 rpm, 4 min, room temperature) to initiate invasion. After 1 h invasion period and inducing CFP expression with IPTG, live imaging was conducted every 2 min for 6 h on a Leica DMI 8 spinning-disc 474 confocal microscope controlled by the iQ software (Andor). Imaris software (Bitplane) was used for analysis. Three independent biological replicates were conducted and movies generated for S. flexneri spreading dynamics were analyzed according to Fig 2A. Movies generated for dynamics of PtdIns(4,5)P 2 (mCherry-PH) levels and cortical actin (mCherry-Actin) levels at protrusion membranes were analyzed as described in Fig 3. Briefly, mCherry-PH and mCherry-Actin levels at protrusion membrane were reported as percentage of corresponding signal levels at membrane of mCherry-PH and mCherry-Actin expressing cells, respectively. Supporting movies (S1-S3 Movies) were generated using the Imaris software (Bitplane).

Immunostaining
Co-localization of PtdIns(4,5)P 2 /F-actin and mCherry-Actin/F-actin were tested at 4 h post infection (hpi) in mixed HT-29 mbCFP and HT-29 mbCFP/mCherry-PH (1:1), and HT-29 mbCFP and HT-29 mbCFP/mCherry-Actin cells, respectively. Confluent cell monolayers were established on glass coverslips in 24-well plates and infected with overnight S. flexneri pCFP cultures (in McCoys's-10% hiFBS). After 1 h of invasion and washing steps, McCoys's-10% hiFBS supplemented with gentamicin and IPTG was delivered to each well to eliminate extracellular S. flexneri and to induce CFP expression, respectively. Following paraformaldehyde fixation, incubation with Phallodin-514 was performed for detection of F-actin. Formation of actin tails by S. flexneri pCFP strains in cells were visualized at 4 hpi in HT-29 cell monolayers after incubation with Alexa Fluor Phalloidin (either 514 or 594) (Invitrogen). Coverslips were mounted and images were acquired with a Leica DMI 8 spinning-disc 474 confocal microscope controlled by the iQ software (Andor). Protrusions projecting from probeexpressing cells (PH+ or Actin+) into probe-negative (PH-or Actin-) cells were evaluated as described in Fig 3.

Infant rabbit infections
Pregnant New Zealand White Rabbits were obtained from the Charles River breeding company. Infant rabbits were handled as previously described [28]. Two days prior to infection, S. flexneri pCFP strains were re-streaked on LB-Congo red-chloramphenicol plates and incubated at 37˚C for overnight. Congo red-positive colonies were transferred to 5 ml TSB-chloramphenicol and incubated as described in the first section of Materials and methods. On infection day, overnight cultures were pooled in 50 ml falcon tubes, centrifuged (4000 rpm /10 min / room temperature) and resuspended in TSB-chloramphenicol. Prior to infection, S. flexneri pCFP suspensions were centrifuged, bacterial pellets were resuspended in PBS and diluted 10-fold in PBS for rectal inoculation of 1.5x10 8 cfus in anesthetized, 10-15 day old infant rabbits. Bacillary dysentery pathological scores were determined blindly. Dysentery scores were determined by presence (1) or absence (0) of bloody stain on the fur, and severity of diarrhea scores (wet body parts) were determined as (1)

Histological examinations of infant rabbit colons
Animals infected with S. flexneri pCFP strains were euthanized by CO 2 asphyxiation followed by euthasol injection. Bacterial invasion and cell death phenotypes were evaluated in colon samples collected 4 hpi. S. flexneri dissemination and infection-associated fenestration of colonic epithelium were evaluated in colon samples collected 8 hpi. Harvested animal colons were prepared for paraffin sections as previously described [28]. Colon paraffin sections were further processed for immunofluorescence or hematoxylin and eosin stains. To detect S. flexneri pCFP and epithelial cells, colon paraffin sections were sequentially processed via deparaffinization, re-hydration, antigen retrieval, permeabilization, and blocking as previously described [28]. The processed sections were incubated with rabbit anti S. flexneri antibody (ViroStat, 1:100) and mouse anti E-cadherin antibody (BD Biosciences, 1:100) at 4˚C overnight in a humidified chamber. Sections were then incubated with goat anti-rabbit IgG conjugated with Pacific blue and (1:500) and Alexa Fluor goat 514 anti-mouse (1:500) secondary antibodies for 2 h at room temperature. Coverslips were mounted with ProLong Gold Antifade Mountant (Thermo Fisher). Cell death in colon sections were determined by using in situ cell death detection kit (Roche) along with DAPI staining. Entire colons (4 hpi) were imaged using a Nikon TE2000 microscope equipped for multi-color imaging including motorized stage and filter wheels and a Hamamatsu Orca ER Digital CCD Camera. Entire colons (8 hpi) were scanned for bacterial infection foci and imaged using a Nikon TE2000 microscope. Each image was processed with thresholding and masking functions of the MetaMorph software to detect intracellular Shigella that invaded E-cadherin positive epithelial cells. As described in S7  Fig, total infection foci count and average number of bacteria per focus were reported for a whole colon section (4 hpi; WT, n = 6; ΔipgD, n = 7). The size of the infection foci was determined with images of 8 hpi colon sections (6 colons for each strain) using the Region function of the MetaMorph software (Molecular Devices, Inc.). Hematoxylin and eosin stains of colon sections were conducted at the Research Histology Core Facility at University of Virginia School of Medicine. H&E staining were scanned with an Aperio ScanScope Slide Scanner (Leica Biosystems), and analyzed to determine the amount of fenestration using the Aperio ImageScope software (Leica Biosystems). The percentage of fenestration for each colon was calculated as: (length of colon with fenestrated epithelium/ total length of colon) x 100.

Statistical analyses
Analysis of a single variable across two groups was performed with unpaired t-test to determine statistical significance. Paired t-test was used to for the comparison of probe levels at different time points (nascent vs late) within the same protrusion. Experiments with multiple groups were analyzed with one-way ANOVA followed by Tukey's multiple comparisons. Two-factor analysis was performed with two-way ANOVA followed by Sidak's multiple comparisons. Two linear regression models were built to analyze data pertinent to co-localization at protrusion membranes: F-actin at protrusion membranes was explained by mCherry-PH probe recruitment, strain and their interaction or by mCherry-Actin recruitment, strain, and their interaction.  Table A in S1 Text. Primers used in this study. Table B in S1 Text. Plasmid constructs used in this study. (DOCX) S1 Movie. Successful Shigella dissemination in colonic epithelial cells. HT-29 mbYFP (yellow) cells were infected S. flexneri WT pCFP (blue) for 6 h. This movie shows the formation of canonical membrane compartments and escape from a DMV during Shigella dissemination. Images were acquired every 2 min. This movie corresponds to the successful spread by the WT pCFP strain shown in Fig 2B. (MP4) S2 Movie. A failed Shigella dissemination in colonic epithelial cells. HT-29 mbYFP (yellow) cells were infected S. flexneri ΔipgD pCFP (blue) for 6 h. This movie shows that ΔipgD pCFP strain forms a protrusion yet cannot form VLP and eventually collapses back to the primarily infected cell. Images were acquired every 2 min. (MP4) S3 Movie. A failed ΔipgD dissemination in which actin accumulation occurs at protrusion membrane. Initially, a blue ΔipgD pCFP forms an actin-negative protrusion projecting from HT-29 mbYFP Cherry-Actin (+) cell to HT-29 mbYFP cell. However, cortical actin starts to accumulate at the protrusion membrane which coincides with the pause in elongation. Eventually, the protrusion enriched with polymerized cortical actin collapses back to the primarily infected cell. Images were acquired every 2 min. (MP4)