Nucleolar c‐Myc recruitment by a Vibrio T3SS effector promotes host cell proliferation and bacterial virulence

Abstract

Pathogen type 3 secretion systems (T3SS) manipulate host cell pathways by directly delivering effector proteins into host cells. In Vibrio parahaemolyticus, the leading cause of bacterial seafood‐borne diarrheal disease, we showed that a T3SS effector, VgpA, localizes to the host cell nucleolus where it binds Epstein–Barr virus nuclear antigen 1‐binding protein 2 (EBP2). An amino acid substitution in VgpA (VgpA^L10A) did not alter its translocation to the nucleus but abolished the effector’s capacity to interact with EBP2. VgpA‐EBP2 interaction led to the re‐localization of c‐Myc to the nucleolus and increased cellular rRNA expression and proliferation of cultured cells. The VgpA‐EBP2 interaction elevated EBP2’s affinity for c‐Myc and prolonged the oncoprotein’s half‐life. Studies in infant rabbits demonstrated that VgpA is translocated into intestinal epithelial cells, where it interacts with EBP2 and leads to nucleolar re‐localization of c‐Myc. Moreover, the in vivo VgpA‐EBP2 interaction during infection led to proliferation of intestinal cells and heightened V. parahaemolyticus’ colonization and virulence. These observations suggest that direct effector stimulation of a c‐Myc controlled host cell growth program can contribute to pathogenesis.

The Vibrio parahaemolyticus type III secretion system effector VgpA enhances EBP2/c‐Myc interaction in the host cell nucleolus to promote gut barrier disruption and colonization during infection.

Introduction

Many Gram‐negative bacterial pathogens rely on type 3 secretion systems (T3SS), needle‐like organelles, to cause disease (Diepold & Wagner, 2014). T3SS machinery is typically composed of 20–30 proteins that are encoded within a gene cluster. These apparatuses, sometimes called injectisomes, span the bacterial inner membrane, periplasm, outer membrane, and extend as needle‐like structures beyond the bacterial surface (Deng et al, 2017). The injectisome directly delivers bacterial effector proteins into the cytosol of eukaryotic cells through pores formed by T3SS translocators in the host cell membrane (Pinaud et al, 2018). Within the host cell, effectors manipulate multiple host cell signaling pathways and processes (Coburn et al, 2007; Jennings et al, 2017). Typically, T3SS effectors target eukaryotic proteins associated with the cytoskeleton, innate immune signaling, Golgi network, endoplasmic reticulum, and lysosomes (Hicks & Galan, 2013; Pinaud et al, 2018).

Once delivered into host cells, effectors can localize to the cytosol (Jimenez et al, 2016), cell membrane (Weigele et al, 2017), or nucleus (Rivas & Genin, 2011; Bierne & Cossart, 2012). In the nucleus, effectors can function as transcription factors that directly bind DNA and regulate transcription (Buttner, 2016) or modify chromatin structure to regulate gene expression (Strahl & Allis, 2000). While a number of T3SS effectors that localize to the nucleus have been identified (Hicks & Galan, 2013), effector localization to the nucleolus, the cellular compartment for ribosomal RNA transcription and ribosome assembly, is comparatively rare. An example of a T3SS effector that localizes to the nucleolus is EspF, but the nucleolar target and function of this enteropathogenic Escherichia coli (EPEC) effector is unknown (Dean et al, 2010).

V. parahaemolyticus is the leading cause of seafood‐borne bacterial enteritis worldwide, and the incidence of infection has been increasing over the past 20 years (Yeung & Boor, 2004; Harth et al, 2009; Gavilan et al, 2013; Ma et al, 2014; Pazhani et al, 2014). The pandemic V. parahaemolyticus serotype, O3:K6, harbors two type 3 secretion systems (T3SS1 and T3SS2). T3SS2 is essential for intestinal colonization and causing pathological alterations in the small intestine of infant rabbits, a model host where V. parahaemolyticus elicits pathology resembling that observed in human infection (Ritchie et al, 2012; Zhou et al, 2013). In contrast, T3SS1 does not play a major role in V. parahaemolyticus’ pathogenicity in the infant rabbit model (Ritchie et al, 2012). Key pathologic changes caused by V. parahaemolyticus infection in infant rabbits include remodeling of the enterocyte brush border, disruption of epithelial barrier integrity, and increased intestinal cell proliferation (Ritchie et al, 2012). Eight T3SS2 effectors have been identified—VopZ, VopA, VopT, VopV, VopL, VopC, VopO, and Vpa1380 (Kodama et al, 2015). VopV directly interacts with filamin and actin and is required for intestinal colonization, bacterial adherence to enterocytes and remodeling of the enterocyte brush border (Hiyoshi et al, 2011; Zhou et al, 2014). VopZ promotes epithelial barrier disruption and cell sloughing during infection and inhibits the activity of TAK1, a regulator of innate immune pathways (Zhou et al, 2013). However, the bacterial effector(s) that stimulate intestinal cell proliferation have not been identified.

In many pathogens, T3SS expression is strictly controlled by regulatory systems that respond to host signals (De Nisco et al, 2018). In V. parahaemolyticus, T3SS2 is activated by bile salts through the bile salt receptor VtrC/VtrA complex. Binding of bile salts to the VtrC/VtrA complex leads to increased expression of VtrB, a key activator of expression of T3SS2 genes (Li et al, 2016). T3SS2 effector injection is also triggered by VgpA (Vpa1360), which senses host intracellular potassium (Tandhavanant et al, 2018). Here, we found that VgpA is also a T3SS2 effector translocated into host cells where it localizes to the nucleolus and interacts with EBP2 (Epstein–Barr virus nuclear antigen 1‐binding protein 2). This nucleolar protein is known to interact with c‐Myc (Liao et al, 2014), a transcription factor that plays a critical role in control of eukaryotic cell growth (van Riggelen et al, 2010). The VgpA‐EBP2 interaction leads to accumulation of c‐Myc in the nucleolus, apparently by enhancing the affinity of EBP2 for the oncoprotein. Our observations using tissue‐cultured cells suggest a model in which the nucleolar interaction between translocated VgpA and EBP2 promotes the retention of c‐Myc in the nucleolus, leading to elevated rRNA transcription and cell proliferation. VgpA translocation to nucleoli of intestinal epithelial cells was readily detectable during infection of infant rabbits. Moreover, the in vivo VgpA‐EBP2 interaction promoted nucleolar re‐localization of c‐Myc, intestinal cell proliferation, and enhanced V. parahaemolyticus colonization and virulence, suggesting that T3SS2 manipulation of a c‐Myc‐controlled growth pathway can promote pathogenesis.

Results

VgpA is a T3SS effector that localizes to the nucleolus

VgpA (VPA1360) has been described as a “gatekeeper” because its capacity to “sense” host cell potassium promotes V. parahaemolyticus effector translocation (Tandhavanant et al, 2018). Interestingly, VgpA is also a type 3 secreted protein (Zhou et al, 2012) and can be detected within host cells upon infection (Tandhavanant et al, 2018), suggesting it is not only a regulator of T3SS2 but also a substrate of this secretion system. We first confirmed that VgpA is a T3SS2 regulatory factor. Consistent with previous studies (Tandhavanant et al, 2018), inactivation of vgpA by introduction of an internal stop codon (referred to as vgpA’) diminished secretion of the T3SS2 translocator proteins (VopB2 and VopD2) (Appendix Fig S1A). Furthermore, using an adenylate cyclase (CyaA)‐based translocation assay that measures effector translocation by monitoring accumulation of intracellular cAMP concentration (Sory et al, 1995), the vgpA’ strain exhibited markedly reduced translocation of T3SS2 effectors (VopV, VopL, VopZ) into host cells (Appendix Fig S1B). Note that in these translocation assays, a ΔvscN1 (vscN1 is required for the assembly of a functional T3SS1) strain was used as the parental strain to minimize T3SS1‐mediated cytotoxicity. We next tested whether VgpA can be secreted by T3SS2 by immunoblotting supernatants from WT and T3SS2‐deficient (ΔvscN2) (vscN2 is required for the assembly of a functional T3SS2) V. parahaemolyticus with an anti‐VgpA antibody. VgpA was observed in supernatants from WT, but not in supernatants from the ΔvscN2 strain (Fig 1A). In contrast, the protein was detected in cell pellets from both strains (Fig 1A), indicating that while VgpA is produced by both strains, its secretion is dependent on T3SS2. No VgpA was observed in either the pellet or supernatant of a vgpA’ strain (Fig 1A), reflecting the antibody’s specificity.

Figure 1. VgpA is a T3SS2 effector that localizes to the nucleolus.

1. Detection of VgpA in supernatants from indicated strains by immunoblotting (IB) with anti‐VgpA antisera.
2. T3SS2 translocation of VgpA into Caco‐2 cells using CyaA‐based assay; VopV is a known T3SS2 effector. Error bars represent mean ± standard deviation (n = 3 biologically independent experiments).
3. Localization of translocated VgpA (upper panel) and VgpA^L10A (lower panel) during infection.

Source data are available online for this figure.

We next determined if VgpA can be translocated into host cells using the adenylate cyclase assay. Following infection of Caco‐2 cells with a T3SS1‐deficient (ΔvscN1), T3SS2‐proficient strain harboring the VgpA‐CyaA fusion, similar amounts of intracellular cAMP were detected as cells infected with a strain harboring a VopV‐CyaA fusion (Zhou et al, 2013; Zhou et al, 2014) (Fig 1B). Like VopV‐CyaA, VgpA‐CyaA was not translocated by a strain with a defective T3SS2 (ΔvscN1ΔvscN2) (Fig 1B). As a negative control, CyaA alone was not translocated by either ΔvscN1 or ΔvscN1/vscN2 (Fig 1B). These results show that VgpA translocation into Caco‐2 cells requires a functional T3SS2. These observations also suggest that besides VgpA’s critical regulatory role in T3SS2 activity, it may also act as a T3SS2 effector, since it was translocated into host cells in a T3SS2‐dependent fashion.

We next investigated the subcellular localization of translocated VgpA within host cells. To minimize the cytotoxic effect of both T3SS1 and T3SS2, we used a V. parahaemolyticus strain (vopZ’ΔvscN1) deficient in vscN1 and vopZ, which encodes a T3SS2 effector important for T3SS2‐mediated cytotoxicity (Zhou et al, 2013), for these cell culture studies. Immunofluorescence microscopy using anti‐VgpA antisera revealed that VgpA appeared to localize primarily to the nucleoli of Caco‐2 cells following infection (Appendix Fig S1D, top panel). As expected, no VgpA was observed in Caco‐2 cells infected with a mutant strain (vopZ’ΔvscN1ΔvscN2) lacking a functional T3SS2 (Appendix Fig S1D, bottom panel). To further confirm VgpA’s nucleolar localization, we labeled infected cells with both anti‐nucleophosmin antibody [a nucleolar marker (Grisendi et al, 2006)] and anti‐VgpA antibody. VgpA co‐localized with nucleophosmin (Fig 1C, upper panel), demonstrating that VgpA is a T3SS2 translocated effector that localizes to the nucleolus of host cells.

VgpA interacts with EBP2

To identify candidate VgpA binding partners, we used a GST‐VgpA fusion protein to pull down VgpA‐interacting host proteins from Caco‐2 nuclear extracts. There was only one band in the co‐purification sample when GST‐VgpA was used as a bait (Fig 2A, lane 1, red arrow) that was not present in the sample that only contained VgpA‐GST (Fig 2A, lane 3); this band was not present in the co‐purification sample when GST was used as the bait (Fig 2A, lane 2), suggesting that it binds VgpA, and not GST. Mass spectrometry identified this protein as Epstein–Barr virus nuclear antigen 1 (EBNA1)‐binding protein 2 (EBP2). We validated this band co‐purified with GST‐VgpA (Fig 2A, red arrow) as EBP2 by immunoblotting with an anti‐EBP2 antibody (Fig 2B). Furthermore, purified recombinant GST‐VgpA pulled down purified recombinant EBP2‐6xHis (Appendix Fig S2A, lanes 1 and 3), suggesting that the two proteins directly interact. EBP2 is known to localize to the human nucleolus (Liao et al, 2014) and transfected VgpA co‐localized with EBP2 (Fig 2D, upper panel, white arrow), providing additional corroboration of the interaction of these two proteins and of VgpA’s nucleolar localization. Together, these observations suggest that VgpA interacts with EBP2 following its translocation into host cells. The function of EBP2 is incompletely understood, but studies with the yeast homolog (Ebp2p) have revealed that the protein is required for processing of ribosomal RNA and ribosome assembly (Huber et al, 2000). In mammalian cells, overexpression of EBP2 promotes ribosomal DNA transcription and cell proliferation (Liao et al, 2014).

Figure 2. Identification of EBP2 as a binding partner of VgpA.

1. Silver‐stained PAGE gel of proteins from a nuclear extract that interact with either GST‐VgpA or GST glutathione agarose. The protein bound by GST‐VgpA (red arrow), but not GST, was identified as EBP2 by mass spectrometry; GST‐VgpA is labeled with a blue arrow.
2. Western blot using anti‐EBP2 as antibody confirmed that EBP2 binds VgpA.
3. Interaction of VgpA or VgpA mutants (VgpA^Δ10 and VgpA^L10A) with EBP2.
4. Localization of VgpA (upper panel) or in VgpA^L10A (lower panel) in HeLa cells after transfection. EBP2 was detected with anti‐EBP2 antibody (red), VgpA and VgpA^L10A were detected with anti‐VgpA sera (green), and nuclei were stained with DAPI (blue). White arrows indicate co‐localization of EBP2 and VgpA (upper panel). VgpA^L10A localizes primarily to the nucleoplasm (white arrow, lower panel).

Source data are available online for this figure.

A set of VgpA truncation mutants was created to map the VgpA region required for its interaction with EBP2. These assays revealed that in the absence of VgpA’s N‐terminal 10 amino acids (aa 1–10), the protein’s capacity to interact with EBP2 was abolished (Appendix Fig S2B). Additional mapping revealed that the 10^th amino acid (leucine) of VgpA is critical for the VgpA‐EBP2 interaction because VgpA^aa10–206 interacts with EBP2, while VgpA^aa11–206 did not (Appendix Fig S2C). In addition, deletion of this leucine residue (VgpA^Δ10) or its substitution with alanine (VgpA^L10A) abolished VgpA’s capacity to bind to EBP2 (Fig 2C). Immunoprecipitation assay using anti‐VgpA showed that while purified GST‐VgpA pulled down purified EBP2x6His, purified GST‐VgpA^L10A did not (Appendix Fig S2A, compare lanes 3 and 4). Furthermore, transfected VgpA^L10A did not co‐localize with EBP2 and form foci; instead, VgpA^L10A was observed throughout the nucleus (Fig 2D, lower panel, white arrow). Similarly, after translocation during infection, VgpA^L10A was present throughout the nucleus of Caco‐2 cells (Fig 1C, lower panel), and not in the nucleolus, in contrast to VgpA (Fig 1C, upper panel). Fractionation analyses confirmed that following infection, VgpA was primarily found in nucleoli, whereas VgpA^L10A was mainly found in the nucleoplasm (Appendix Fig S3). Together, these observations reveal that VgpA’s L10 residue is critical for its interaction with EBP2 and nucleolar localization. Although the substitution of Ala for Leu at VgpA position 10 severely compromised the effector’s capacity to interact with EBP2, the VgpA^L10A mutant (ΔvscN1vgpA ^L10A) sustained wild‐type levels of substrate (VopB and VopD) secretion (Appendix Fig S1C) and translocation of other T3SS2 effectors (VopV, VopZ, and VopL) (Appendix Fig S1B), suggesting that different residues in VgpA mediate its roles in T3SS2 and in interacting with the nucleolar protein EBP2.

VgpA promotes host cell proliferation

Recently, the c‐Myc oncoprotein was shown to directly interact with EBP2, consequently promoting rDNA transcription and cell proliferation (Liao et al, 2014). To address whether the VgpA‐EBP2 interaction also promotes cell proliferation, we first measured BrdU incorporation in Caco‐2 cells infected with various V. parahaemolyticus strains. Approximately 15% of uninfected control Caco‐2 cells were BrdU‐positive, compared to ~ 55% of Caco‐2 cells infected with the vopZ’∆vscN1 strain (Fig 3A and C). In the absence of vgpA (vopZ’∆vscN1vgpA’) cell proliferation remained at control levels (Fig 3A and C). Complementation with wild‐type vgpA (vopZ’∆vscN1vgpA’:pvgpA) restored cell proliferation, whereas complementation with vgpA ^L10A (vopZ’∆vscN1vgpA’:pvgpA ^L10A) did not (Fig 3A and C). Together, these observations suggest that the VgpA‐EBP2 interaction is important for V. parahaemolyticus to promote cell proliferation. Since the L10A mutation in VgpA does not affect the secretion (Appendix Fig S1C) or the translocation of T3SS2 effectors into host cells (Appendix Fig S1B), it is unlikely that the reduced cell proliferation observed with infection by vopZ’∆vscN1vgpA’:pvgpA ^L10A was due to the reduced T3SS2 functionality.

Figure 3. The VgpA‐EBP2 interaction promotes Caco‐2 cell proliferation.

1. Fluorescence micrographs of uninfected Caco‐2 cells or Caco‐2 cells infected with the indicated V. parahaemolyticus strains. Cells were stained with DAPI (blue) and anti‐BrdU antibody (green).
2. Fluorescence micrographs of Caco‐2 cells transfected with siRNA targeting EBP2 (siEBP2) or control siRNA (siNC) and subsequently infected with vopZ’ΔvscN1. Cells were stained with anti‐BrdU antibody (green) and DAPI (blue).
3. The percentage of BrdU‐positive cells (green) among the total cells (green + blue) was analyzed for 100 cells in three experiments. Error bars represent mean ± standard deviation (n = 3 biologically independent experiments). One‐way analysis of variance (ANOVA) was used for statistical analysis. *P < 0.05 (when compared to the uninfected Caco‐2 cells).
4. The percentage of BrdU+ cells was analyzed as above for cells treated with siNC or siEBP2. Error bars represent mean ± standard deviation (n = 3 biologically independent experiments). Two‐way ANOVA was used for statistical analysis. *P < 0.05 (when compared to siNC‐transfected cells that are infected with vopZ’ΔvscN1).
5. Immunoblot of EBP2 and Actin in Caco‐2 cells transfected with siEBP2 or siNC.

Source data are available online for this figure.

Additional analyses of the association between cell proliferation and VgpA translocation bolster the idea that this effector promotes cell proliferation. Approximately 75% of cells that contained VgpA translocated from vopZ’∆vscN1 were BrdU+; in contrast, among cells that did not contain detectable VgpA after infection with vopZ’∆vscN1, only 15% (i.e., the frequency of BrdU+ among uninfected cells) were BrdU+ (Appendix Fig S4A, upper panel, Appendix Fig S4B). The frequency of BrdU+ cells was approximately the same (15%) in cells that contained translocated VgpA^L10A or not (Appendix Fig S4A, lower panel, Appendix Fig S4B), enforcing the idea that the VgpA‐EBP2 interaction is critical for V. parahaemolyticus to induce epithelial cell proliferation.

To further assess the requirement for the VgpA‐EBP2 interaction for stimulating cell proliferation, the expression of EBP2 in Caco‐2 cells was knocked down with siRNA (siEBP2). Caco‐2 cells transfected with siEBP2 reduced the expression of EBP2 compared to those transfected with a control siRNA (siNC) (Fig 3E). Infection of siNC‐transfected (control) cells with the VgpA+ V. parahaemolyticus strain (vopZ’∆vscN1) resulted in ~ 55% BrdU‐positive cells; in contrast, only 17% of siEBP2‐transfected infected cells were BrdU‐positive (Fig 3B and D). Together, these observations support the idea that VgpA‐induced cell proliferation is dependent on EBP2.

VgpA‐EBP2 interaction promotes the nucleolar localization of c‐Myc

The overexpression of EBP2 has been reported to recruit c‐Myc into the nucleolus and ultimately promote cell proliferation (Liao et al, 2014). We hypothesized that VgpA‐induced cell proliferation might also depend on a similar mechanism. In uninfected Caco‐2 cells, c‐Myc localized in a diffuse fashion in the nucleoplasm, but rarely (< 1%) in the nucleolus (Fig 4A, first row), whereas c‐Myc was observed in the nucleolus of > 70% of the Caco‐2 cells following infection with vopZ’∆vscN1 (Fig 4A, second row). In contrast, in cells infected with a VgpA mutant (vopZ’ΔvscN1vgpA’), c‐Myc was detected in the nucleolus in < 1% of cells (Fig 4A, third row). Complementation of the vgpA’ mutant with wild‐type VgpA (Fig 4A, fourth row), but not with VgpA^L10A (Fig 4A, fifth row), restored the nucleolar localization of c‐Myc in > 75% cells. Cellular fractionation analyses showed that c‐Myc was primarily present in the nucleoli of cells infected with vopZ’∆vscN1, whereas c‐Myc was mainly present in the nucleoplasm of cells infected with vopZ’∆vscN1vgpA ^L10A (Appendix Fig S3). These analyses also showed that the nucleolar localization of EBP2 did not change in cells infected with either vopZ’∆vscN1 or vopZ’∆vscN1vgpA ^L10A (Appendix Fig S3). Together, these observations suggest that the VgpA‐EBP2 interaction is critical for c‐Myc recruitment into and/or retention in the nucleoli. Consistent with this idea, we found that knockdown of EBP2 with siRNA nearly abolished infection‐associated nucleolar re‐localization of c‐Myc (Appendix Fig S5, upper panel), whereas c‐Myc was observed in the nucleolus of > 70% cells treated with control siRNA (Appendix Fig S5, lower panel).

Figure 4. The VgpA‐EBP2 interaction promotes EBP2‐c‐Myc interaction and c‐Myc nucleolar localization.

• A
  Localization of c‐Myc in uninfected Caco‐2 cells (first row), Caco‐2 cells infected with V. parahaemolyticus vopZ’ΔvscN1 (wild‐type VgpA, second row), Caco‐2 cells infected with vopZ’ΔvscN1vgpA’ (third row), Caco‐2 cells infected with vopZ’ΔvscN1vgpA’:pvgpA (fourth row), and Caco‐2 cells infected with vopZ’ΔvscN1vgpA’:pvgpA ^L10A (fifth row). Fixed cells were stained with DAPI (blue), and anti‐EBP2 (red), and anti‐c‐Myc (green) antibodies. White arrows are examples indicating the co‐localization of EBP2 and c‐Myc.
• B, C
  Detection of c‐Myc or EBP2 in nuclear extracts from uninfected Caco‐2 cells or after infection with indicated strains. Following infection, nuclear extracts were obtained and loaded on protein A/G beads cross‐linked with anti‐EBP2 ((B), IP: anti‐EBP2), or anti‐C‐Myc ((C), IP: anti‐c‐Myc) antibodies, respectively. Input and eluates were blotted with anti‐EBP2, anti‐c‐Myc, or anti‐VgpA antibodies, respectively. Quantified value (mean ± standard deviation) as measured by band intensity (normalized to uninfected) from three independent experiments is presented under the signals of the blot of IP samples. n.d., not determined.
• D
  Purified proteins EBP2, c‐Myc, and GST‐VgpA or GST‐VgpA^L10A were mixed and then pulled down with protein A/G beads cross‐linked with anti‐EBP2. Input and eluates were blotted with anti‐EBP2, anti‐c‐Myc, or anti‐VgpA antibodies, respectively. Quantified value (mean ± standard deviation) as measured by band intensity (normalized to EBP2 + c‐Myc sample without GST‐VgpA and GST‐VgpA^L10A) from three independent experiments is presented under the signals of the blot of IP samples. n.d., not determined.
• E
  Caco‐2 cells were infected with the indicated strains and treated with cycloheximide (CHX) for the indicated time. Whole cells lysates were analyzed by Western blot using anti‐c‐Myc and anti‐Actin antibody.

Source data are available online for this figure.

Since the VgpA interaction with EBP2 promotes the re‐localization of c‐Myc to the nucleolus (Fig 4A, Appendix Figs S3 and S5) and overexpression of EBP2 also promotes the re‐localization of c‐Myc to nucleolus (Liao et al, 2014), we initially hypothesized that VgpA recruits c‐Myc to the nucleolus by increasing the expression of EBP2. However, there was no detectable difference in EBP2 expression in Caco‐2 cells infected with vopZ’∆vscN1 or vopZ’∆vscN1vgpA ^L10A (Appendix Fig S6), arguing against the idea that the interaction of VgpA with EBP2 leads to nucleolar re‐localization of c‐Myc by increasing EBP2 expression. We then tested whether the VgpA interaction with EBP2 enhanced the apparent binding between EBP2 and c‐Myc using immunoprecipitation assays. In these assays, nuclear extracts from uninfected Caco‐2 cells or Caco‐2 cells infected with vopZ’ΔvscN1 or vopZ’ΔvscN1vgpA ^L10A were added to agarose beads bound to either anti‐EBP2 antibody or anti‐c‐Myc antibody. Eluted material was subsequently immunoblotted for detection of EBP2, c‐Myc, and VgpA (Fig 4B and C). In the uninfected cells or cells infected with vopZ’∆vscN1vgpA ^L10A, there were weak signals for c‐Myc in eluates from anti‐EBP2 cross‐linked agarose (Fig 4B), and weak signals (low band intensity) for EBP2 in eluates from anti‐c‐Myc cross‐linked agarose (Fig 4C). In contrast, in the cells infected with vopZ’∆vscN1, there was a prominent signal for c‐Myc in the eluates from anti‐EBP2 cross‐linked agarose (Fig 4B), and a prominent signal (high band intensity) for EBP2 in the eluates from anti‐c‐Myc cross‐linked agarose (Fig 4C). Together, these observations suggest that binding between c‐Myc and EBP2 is enhanced by VgpA, but not by VgpA^L10A. The possibility that the increased c‐Myc in the eluates in Fig 4B is due to the direct binding between VgpA and c‐Myc was excluded in a pull‐down assay using purified c‐Myc and VgpA; neither VgpA nor VgpA^L10A pulled down c‐Myc (Appendix Fig S7). To further determine if the VgpA interaction with EBP2 directly enhances the capacity of EBP2 to bind c‐Myc, we mixed all three purified proteins (EBP2‐6xHis, c‐Myc‐6xHis, and GST‐VgpA or GST‐VgpA^L10A) and performed a pull‐down assay using anti‐EBP2 cross‐linked agarose. EBP2 pulled down more (roughly 31‐ to 34‐fold based on the band intensity) c‐Myc in the presence of VgpA than in the absence of VgpA or in the presence of VgpAL^10A (Fig 4D), indicating that VgpA, but not VgpA^L10A, directly enhances the binding between EBP2 and c‐Myc. Thus, the re‐localization of c‐Myc to the nucleolus appears to be a consequence of the nucleolar interaction between VgpA and EBP2 enhancing EBP2’s affinity for c‐Myc. Moreover, cycloheximide (CHX) chase assays showed that the half‐life of c‐Myc in vopZ’ΔvscN1‐infected Caco‐2 cells was prolonged compared to that in vopZ’ΔvscN1vgpA ^L10A‐infected Caco‐2 cells (Fig 4E), suggesting that VgpA interaction with EBP2 also increases c‐Myc’s stability.

The VgpA‐EBP2 interaction increases ribosomal RNA transcription

c‐Myc can enhance transcription of ribosomal RNA (rRNA) by activating RNA polymerase I (RNA Pol I) transcription (Grandori et al, 2005) and transcription of rRNA is important for cell proliferation (Zhang et al, 2014). Therefore, we analyzed the effect of VgpA on the transcription of rRNA and RNA Pol I (POLR1A). We found that Caco‐2 cells infected with vopZ’ΔvscN1 or vopZ’ΔvscN1vgpA’:pvgpA had significantly higher rRNA levels than uninfected cells (Fig 5A); in addition, cells infected with vopZ’ΔvscN1vgpA’ or vopZ’ΔvscN1vgpA’:pvgpA ^L10A had rRNA levels that were not different from those measured in uninfected cells (Fig 5A). Moreover, knockdown of EBP2 abolished the ability of vopZ’ΔvscN1 to elevate rRNA levels (Fig 5B). Similar results were observed for the relative transcription of RNA Pol I gene (Fig 5C). These observations strongly suggest that the VgpA‐EBP2 interaction elevates RNA polymerase I gene transcription and subsequent rRNA transcription, and support a model in which the nucleolar interaction between translocated VgpA and Ebp2 recruits c‐Myc to nucleolus, leading to elevated transcription of rRNA and cell proliferation. This model is supported by the observation that knockdown of c‐Myc abolished the ability of vopZ’ΔvscN1 to promote cell proliferation (Appendix Fig S8).

Figure 5. VgpA promotes rRNA transcription in vitro and in vivo .

1. Amounts of rRNA were from uninfected Caco‐2 cells or cells infected with indicated strains. *P < 0.05 (when compared to the uninfected Caco‐2 cells).
2. Amounts of rRNA in uninfected or infected Caco‐2 cells transfected with control (siNC) or EBP2‐targeting (siEBP2) siRNAs. *P < 0.05 (when compared to the siNC‐transfected cells that are infected with vopZ’ΔvscN1).
3. Amounts of RNA polymerase I (POLR1A) were from uninfected Caco‐2 cells or cells infected with indicated strains. *P < 0.05 (when compared to the uninfected Caco‐2 cells).
4. Amounts of rRNA in intestinal homogenates from infant rabbits infected with indicated strains 18 and 38 h after infection. *P < 0.05 (when compared to the uninfected rabbits at 18 h). ^Δ P < 0.05 (when compared to the uninfected rabbits at 38 h post‐infection).

Data information: All error bars represent mean ± standard deviation (n = 3 biologically independent experiments). One‐way ANOVA was used for statistical analysis for (A) and (C). Two‐way ANOVA was used for statistical analysis for (B) and (D).

VgpA is translocated into the nucleolus of intestinal epithelial cells during infection of infant rabbits and promotes intestinal cell proliferation

Infant rabbits were oro‐gastrically inoculated with WT V. parahaemolyticus or V. parahaemolyticus expressing VgpA^L10A (from the native locus) to investigate the in vivo consequences of the VgpA‐Ebp2 interaction. Initially, we tested if VgpA translocation into the nucleolus of intestinal cells could be detected. Tissue sections were obtained from rabbits 18 h after inoculation, a time point when WT V. parahaemolyticus exhibits maximal colonization of the small intestine, but when tissue damage is less severe than at later points (Ritchie et al, 2012). Both VgpA and VgpA^L10A were translocated in vivo and detectable within intestinal epithelial cells at similar frequencies; however, VgpA was observed as punctae within nuclei of epithelial cells (Fig 6A, upper panel), whereas VgpA^L10A was seen diffusely in the nucleoplasm (Fig 6A, lower panel). Thus, although in vivo translocation of both VgpA and VgpA^L10A was readily detectable, only VgpA appears to localize in nucleoli of infected intestinal epithelial cells.

Figure 6. VgpA promotes cell proliferation within the intestine of infected rabbits.

1. VgpA (upper panel: animals infected with WT) and VgpA^L10A (lower panel: animals infected with vgpA ^L10A) localization in intestinal epithelial cells of infected animals. Arrows in upper panel indicate VgpA localized as punctae in the nucleus; the arrow in the lower panel indicates VgpA^L10A localized diffusively in the nucleoplasm.
2. VgpA and VgpA^L10A interactions with EBP2 in intestinal epithelial cells isolated from infant rabbits infected with WT or vgpA ^L10A V. parahaemolyticus. Inputs represent proteins detected in intestinal epithelial cells of uninfected rabbits or rabbits infected with the indicated strains. Outputs represent eluates from protein A/G beads cross‐linked with anti‐EBP2 (IP: anti‐EBP2) or anti‐VgpA (IP: anti‐VgpA) antibodies and analyzed with anti‐VgpA and anti‐EBP2 antibodies.
3. Frozen sections from uninfected rabbits or rabbits infected with the indicated strains for 18 h were stained with anti‐Ki67 antibodies to detect actively dividing cells (green), phalloidin to visualize F‐actin (red), and DAPI to detect nuclei (blue).
4. Eighteen hours after infant rabbits were infected with WT (middle panel) or vgpA ^L10A (lower panel) V. parahaemolyticus, fixed sections from the small intestine were stained with antibodies to c‐Myc (green), EBP2 (red), and DAPI (blue). Uninfected rabbits were processed similarly and were included as a control (upper panel). White arrows indicate co‐localization of c‐Myc and EBP2 (middle panel).

Source data are available online for this figure.

To determine if VgpA interacts with EBP2 in the epithelium of infected rabbits, we carried out pull‐down assays using anti‐EBP2 or anti‐VgpA cross‐linked agarose on cell lysates from intestinal epithelial cells isolated from infant rabbits infected with WT or VgpA^L10A V. parahaemolyticus. Both VgpA and VgpA^L10A were detectable in the intestinal epithelial cells (Fig 6B, input), indicating that both proteins were translocated into epithelial cells during infection. EBP2 pulled down VgpA, but not VgpA^L10A (Fig 6B, IP: anti‐EBP2). Similarly, VgpA, but not VgpA^L10A, pulled down EBP2 (Fig 6B, IP: anti‐VgpA). These findings provide additional evidence that VgpA is translocated into intestinal epithelial cells; moreover, they demonstrate that VgpA interacts with EBP2 during infection in vivo.

In infant rabbits, V. parahaemolyticus infection stimulates proliferation of intestinal cells (Ritchie et al, 2012), but the bacterial factor(s) that trigger proliferation are not known. Since VgpA promoted proliferation of Caco‐2 cells in culture, we investigated if this effector contributes to intestinal cell proliferation in vivo. Infant rabbits were inoculated with one of four isogenic V. parahaemolyticus strains: WT, vgpA’, vgpA’:pvgpA or vgpA’:pvgpA ^L10A. Intestinal cell proliferation was assessed by staining tissue sections from infected rabbits with Ki‐67, which labels actively proliferating cells (Alferez & Goodlad, 2007). Uninfected animals had minimal Ki‐67 labeling (Fig 6C, first panel). Animals infected with the WT and the complemented (vgpA’:pvgpA) strain had similar prominent labeling with Ki‐67 (Fig 6C, second and fourth panels); in marked contrast, there was minimal Ki‐67 labeling in the rabbits infected with the vgpA’ strain (Fig 6C, third panel). The absence of proliferation in animals infected with vgpA’ mutant was expected, because vgpA is required for T3SS2 function, a key requirement for V. parahaemolyticus colonization in the small intestine. However, although animals infected with the vgpA’:pvgpA ^L10A strain had similar colonization as WT and vgpA’:pvgpA at this time point (Fig 7A), there was only minimal Ki‐67 labeling observed in rabbits infected with this strain (Fig 6C, fifth panel). The size of the zone of proliferating cells in animals infected with the WT strain was significantly larger than in animals infected with vgpA’:pvgpA ^L10A (0.55 ± 0.04 and 0.15 ± 0.01, P < 0.05). Furthermore, infant rabbits infected with either WT or vgpA’:pvgpA for 18 or 38 h had significantly higher rRNA levels in intestinal homogenates than uninfected rabbits (Fig 5D). In contrast, rabbits infected with vgpA’ or vgpA’:pvgpA ^L10A for 18 or 38 h had rRNA levels that were not different from those measured in uninfected rabbits (Fig 5D). Notably, consistent with the findings observed in tissue‐cultured cells, c‐Myc co‐localized with Ebp2 in nucleoli of epithelial cells of animals infected with V. parahaemolyticus expressing VgpA but not VgpA^L10A (Fig 6D, compare the middle and lower panel). Collectively, these observations are concordant with the model generated from experiments with tissue‐cultured cells and suggest that in vivo VgpA‐EBP2 interactions within nucleoli of epithelial cells stimulate intestinal cell proliferation during infection, and that heightened c‐Myc‐EBP2 interactions contribute to this process.

Figure 7. The Vgp‐EBP2 interaction promotes V. parahaemolyticus intestinal colonization and virulence.

1. CFU in homogenates from the distal small intestines of infant rabbits infected with the indicated strains 18, 28, and 38 h following inoculation. Black horizontal line indicates mean CFU (n = 7–10 animals/group). One‐way ANOVA was used for statistical analysis. *P < 0.05 for the indicated comparison.
2. H&E‐stained sections from the distal small intestine from the rabbits infected with the indicated strains 18 h after infection. Black arrows point to bacterial colonies closely attached to the epithelium.
3. Fluid accumulation in rabbits infected with the indicated strains 38 h post‐inoculation. Error bars represent mean ± standard deviation (n = 7 animals/group). One‐way ANOVA was used for statistical analysis. *P < 0.05 (when compared to the rabbits infected with WT).

The VgpA‐EBP2 interaction contributes to V. parahaemolyticus intestinal colonization

Vibrio parahaemolyticus’ T3SS2 is essential for the pathogen to colonize the small intestine and to cause disease in infant rabbits (Ritchie et al, 2012). Since VgpA is required for T3SS2 function (Appendix Fig S1A and B), we anticipated that the mutant lacking VgpA (vgpA’) would exhibit a severe colonization and virulence defect in this model. Indeed, the vgpA’ exhibited as profound an intestinal colonization defect (> 5‐log reduction in recoverable cfu from intestinal homogenates compared with WT) as that of a mutant lacking the T3SS2 ATPase component (VscN2) at 18, 28, and 38 h of infection (Fig 7A). Expression of VgpA from a plasmid complemented the colonization defect (Fig 7A). There was no significant difference in colonization between vgpA’:pvgpA and vgpA’:pvgpA ^L10A 18 h post‐inoculation, but by 28 and 38 h post‐inoculation, the vgpA’:pvgpA ^L10A strain exhibited an ~ 100‐fold colonization defect (Fig 7A). Thus, the interaction between VgpA and EBP2 does not appear to be required for initial V. parahaemolyticus colonization of the intestinal epithelium, but sustained robust colonization depends on this interaction.

In H&E‐stained sections of the intestines from animals infected with the WT, vgpA’:pvgpA, and vgpA’:pvgpA ^L10A strains 18 h earlier, the pathogen was found within characteristic cavities in the intestinal epithelium (Fig 7B) (Ritchie et al, 2012). As expected, the ΔvscN2 or vgpA’ mutants did not generate close associations with the epithelium and epithelial damage was not detected (Fig 7B). Despite the similar disruption of the epithelium caused by the WT and vgpA’:pvgpA ^L10A strains at 18 h, by 38 h post‐inoculation, there was more fluid accumulation observed in rabbits infected with the WT vs. the vgpA’:pvgpA ^L10A mutant (Fig 7C). Thus, the VgpA‐EBP2 interaction appears to promote virulence as well as sustained intestinal colonization.

Discussion

V. parahaemolyticus’ T3SS2 is critical for this enteric pathogen to colonize and cause disease in animal models of infection (Ritchie et al, 2012). In infant rabbits, V. parahaemolyticus intestinal colonization is associated with damage to the epithelium as well as elevated intestinal cell proliferation (Ritchie et al, 2012). Here, we found that VgpA, a gene product that is critical for T3SS2 activity, is also a T3SS2 effector that is translocated to the host cell nucleolus where it interacts with the nucleolar protein EBP2. Identification of a VgpA residue (leucine 10) that is required for the VgpA‐EBP2 interaction but not for the effector’s translocation or role in T3SS2 activity, enabled us to compare phenotypes of strains expressing either VgpA or VgpA^L10A to assess the significance of the VgpA‐EBP2 interaction for V. parahaemolyticus pathogenicity. Remarkably, the interaction between VgpA and EBP2, its host nucleolar target, markedly increased (i) the amount of c‐Myc detectable in the nucleolus, (ii) levels of host cell rRNA, and (iii) proliferation of both tissue‐cultured cells and epithelial cells in the infant rabbit intestine. Moreover, V. parahaemolyticus expressing VgpA had more robust intestinal colonization at late time points during infection and elicited more fluid accumulation than a strain expressing VgpA^L10A. Collectively, these observations suggest that direct pathogen effector manipulation of a c‐Myc‐associated cell growth control program can promote pathogenesis.

The intestinal epithelium undergoes rapid turnover. Epithelial cells shed from the villi are replaced with cells that are derived from proliferating crypt stem cells (Parker et al, 2017). The sloughing of villous epithelial cells, a key component of intestinal homeostasis, is thought to protect the gut from adherent pathogens, which are released into the lumen along with the detached host cell (Ashida et al, 2011). Several mechanisms by which gut pathogens inhibit epithelial cell death and sloughing to sustain colonization have been described. For example, Helicobacter pylori injects the apoptosis inhibitor CagA into the epithelial cells, enhancing its colonization of the stomach (Mimuro et al, 2007), while Shigella flexneri enhances its colonization by stabilizing focal adhesions and blocking epithelial cell sloughing (Kim et al, 2009). Our findings suggest that V. parahaemolyticus has adapted a distinct strategy to preserve its close association with villous epithelial cells. Rather than block epithelial shedding, V. parahaemolyticus translocates VgpA into host cells, triggering the production of new epithelial cells to which it can adhere.

Unlike effectors that promote host cell survival by inhibiting cell death pathways, VgpA appears to stimulate host cell proliferation. The nucleolar interaction of VgpA with EBP2 promotes the accumulation of c‐Myc in the nucleolus. In quiescent cells, c‐Myc is primarily localized to the nucleoplasm but when cells are stimulated with serum or during the G0/G1 transition, c‐Myc can be recruited to the nucleolus where it activates RNA polymerase I transcription of ribosomal RNA expression, promoting cell growth and proliferation (Grandori et al, 2005). Factors that promote c‐Myc nucleolar localization, such as nucleophosmin, have been described (Li & Hann, 2013). Since the binding of VgpA to EBP2 appears to increase the affinity of the EBP2‐c‐Myc interaction (Fig 4B–D), we hypothesize that the nucleolar accumulation of c‐Myc results from its heightened retention there due to its enhanced affinity for the VgpA‐EBP complex. Within the nucleolus, c‐Myc is usually targeted to the ubiquitin–proteasome system for degradation (Welcker et al, 2004; Li & Hann, 2013). Our observation that infection with VgpA⁺ (but not VgpA^L10A+) V. parahaemolyticus strain increased the half‐life of c‐Myc (Fig 4E) suggests that c‐Myc binding to VgpA‐EBP2 complex protects the oncoprotein from degradation. Enhanced stability of c‐Myc has also been observed when EBP2 is overexpressed (Liao et al, 2014) but we found that EBP2 levels were not altered by VgpA (Appendix Fig S6). Additional studies elucidating the biochemical and structural bases for the VgpA‐mediated increase in EBP2 binding to c‐Myc and enhanced stability of c‐Myc will be valuable, given the fundamental role of c‐Myc in controlling cell growth. Also, although pathogen manipulation of c‐Myc expression levels has been described (Borth et al, 2011; Colineau et al, 2018), effector‐mediated alteration in c‐Myc localization has not.

Type III effector protein localization to the nucleolus is unusual. EspF, an EPEC effector, has been shown to target the nucleolus after transfection, leading to its disruption and the redistribution of the nucleolar protein nucleolin to the cytoplasm (Dean et al, 2010); however, EspF is a multifunctional effector that primarily targets mitochondria and the significance of its nucleolar localization in EPEC pathogenesis is unclear (Holmes et al, 2010). LegAS, a Legionella pneumophila type IV secretion system effector, also targets the host cell nucleolus where it increases transcription of ribosomal DNA by methylating rDNA chromatin (Li et al, 2013). Li et al’s (2013) bioinformatic analyses identified additional LegAS‐like proteins, including a type III effector from Burkholderia thailandensis that was also shown to localize to the nucleolus and activate rDNA expression. They proposed that nucleolar‐localizing “SET”‐domain containing effectors promote rDNA expression (Li et al, 2013). The contribution of this family of effectors to in vivo pathogenesis remains to be defined. Viral proteins often localize to the nucleolus where they can lead to the re‐localization of host nucleolar proteins (Weeks et al, 2019).

V. parahaemolyticus utilizes T3SS2 to attach to and markedly remodel the brush border of the epithelial surface of the small intestine, where it colonizes (Zhou et al, 2014). Presumably, this mode of attachment facilitates pathogen growth, but it damages the epithelium, ultimately leading to cell extrusion and death (Ritchie et al, 2012; Zhou et al, 2014; Blondel et al, 2016). We propose that V. parahaemolyticus deploys the T3SS2 effector VgpA to stimulate the host to produce new epithelial cells that are capable of being exploited by the pathogen. Consistent with this idea, we found that the colonization defect of V. parahaemolyticus expressing VgpA^L10A was not observed until later time points during infection, when pathology becomes apparent (Fig 7); at an early time point, when there is minimal epithelial damage (Ritchie et al, 2012), V. parahaemolyticus expressing either VgpA or VgpA^L10A had similar colonization.

Finally, VgpA’s capacity to stimulate intestinal cell proliferation could potentially be harnessed for the treatment of diseases, such as inflammatory bowel disease, associated with intestinal epithelial injury. It may be possible to deliver VgpA to damaged tissue by a V. parahaemolyticus mutant that can robustly colonize the intestine without causing disease (Zhou et al, 2013) or via a probiotic strain like E. coli Nissle engineered to express a heterologous T3SS (Gonzalez‐Prieto & Lesser, 2018).

Materials and Methods

Bacterial strains and growth conditions

A clinical isolate of V. parahaemolyticus (RMID2210633) (Makino et al, 2003) was used as the wild‐type strain in this study. Inactivation of T3SS2 (ΔvscN2) and VopZ (vopZ’) was described previously (Ritchie et al, 2012; Zhou et al, 2013). vgpA was inactivated (vgpA’) by introducing a stop codon at aa39 of vpa1360 using the suicide vector pDM4 (pDM4‐vgpA^stop). The fusion of cyaA to vgpA, which was used to monitor the translocation of VgpA into host cells, was made as described previously (Zhou et al, 2013). The fusion of GST and VgpA was constructed using a pGEX4T1 vector as described previously (Zhou et al, 2014). V. parahaemolyticus was used as host to express GST‐VgpA or GST‐VgpA mutants. Recombinant EBP2 with 6xHis was synthesized from pET28a (pET28a‐EBP2) using E. coli BL21(DE3) as host strain. Full‐length (without its native promoter) vgpA or vgpA ^L10A was inserted into pMMB207 (resulting in pMMB207‐vgpA and pMMB207‐vgpA ^L10A) for complementation studies. The L10A mutation was introduced to the genome of WT, ΔvscN1 and vopZ’ΔvscN1 V. parahaemolyticus using the suicide vectors (pDM4‐vgpA ^L10A) as described previously (Zhou et al, 2013). All strains were grown in LB medium containing appropriate antibiotics: 50 µg/ml carbenicillin and 30 µg/ml chloramphenicol. IPTG (0.1 mM) were used to induce the expression of proteins encoded in pMMB207 derivatives. Sodium cholate (0.04%) was used to activate T3SS2. Bacterial strains and plasmids are listed in Appendix Table S1.

Experimental animals

Protocols for animal studies were approved by the Harvard Medical Area Standing Committee on Animals (Animal Welfare Assurance of Compliance #A3431‐01) and by the Institutional Animal Care and Use Committee (IACUC) of University of Connecticut (Protocol #A13‐060). Infant rabbits were infected as described previously (Ritchie et al, 2012). Colonization and fluid accumulation in the small intestine were measured as described previously (Ritchie et al, 2012). H&E staining was performed as described previously (Ritchie et al, 2012).

Protein secretion and translocation

Cell pellets and supernatants from V. parahaemolyticus cultures were prepared as described previously (Zhou et al, 2013). Western blots of the pellets and supernatants were performed using anti‐VgpA, anti‐VopD2, and anti‐VopB2 antibodies. Cyclic AMP‐based protein translocation assays were performed as described previously (Zhou et al, 2013). To visualize the localization of VgpA, Caco‐2 cell monolayers on coverslips were infected with V. parahaemolyticus for 2 h at MOI of 100 and washed with PBS three times. Subsequently, the infected cells were maintained in DMEM containing 20 µg/ml gentamicin for 12 h. Following thorough wash with PBS, the infected cells were fixed with 4% paraformaldehyde. Cells were then incubated with anti‐VgpA (1:100) or anti‐nucleophosmin (NPM1) (1:1,000) antibody (Thermo Fisher Scientific) overnight. FITC‐conjugated anti‐rabbit IgG (1:100) or Alexa Fluor 568‐conjugated anti‐mouse IgG antibody (1:200) (Thermo Fisher Scientific) was then added and incubated for additional 1 h before imaging with a confocal microscope. Representative confocal microscopic images were taken from three biological experiments.

Protein pull‐down assays

To identify the proteins that bind VgpA, a pull‐down assay was performed using recombinant GST fused to VgpA (GST‐VgpA) or GST alone as bait and Caco‐2 cell nuclear extracts as prey, as described previously (Zhou et al, 2014). Nuclear extracts were isolated using a Nuclear Extract Kit (Active Motif, Carlsbad, CA). GST‐VgpA was expressed in V. parahaemolyticus using IPTG (0.5 mM). GST‐VgpA containing whole cell lysates were loaded on to glutathione agarose as prey and incubated for 1 h with gentle shaking to allow binding. Following extensive washing, nuclear extracts from Caco‐2 cells were added as prey and incubated for additional 1 h. Following extensive washing, proteins were eluted and separated by SDS–PAGE. The protein band that is only present in the elution from the sample using GST‐VgpA as bait and nuclear extract as prey was submitted for mass spectrometry identification (Taplin Mass Spectrometry Facility, Harvard Medical School). Western blot of the eluted samples was performed using anti‐EBP2 at the dilution of 1:100 (Sigma) to verify the identity of proteins. GST was fused to series of VgpA truncations, and similar pull‐down assay was performed with these truncated VgpA as bait and nuclear extract as prey to identify the VgpA region that binds Ebp2.

To determine the influence of VgpA on the interaction between EBP2 and c‐Myc, Caco‐2 cells were infected for 2 h. Infected or uninfected cells were scraped off the 6‐well plates and lysed by sonication. Then, lysed samples were loaded on to protein A/G agarose that had been cross‐linked with anti‐EBP2 or anti‐c‐Myc (Thermo Fisher Scientific) antibodies using disuccinimidyl suberate (DSS) (Thermo Fisher Scientific). Following incubation at 4°C for 12 h, agarose was washed with PBS three times and proteins were eluted using SDS loading buffer. Both input and eluates were Western blotted using anti‐EBP2 (1:100), anti‐c‐Myc (1:100), and anti‐VgpA (1:100) antibodies.

To determine if VgpA directly influences the binding between EBP2 and c‐Myc, we expressed and purified GST‐VgpA, GST‐VgpA^L10A, and EBP2‐6xHis. Briefly, GST‐VgpA and GST‐VgpA^L10A were expressed in V. parahaemolyticus and purified using glutathione agarose (Thermo Fisher Scientific) according to the manufacturer’s instruction. EBP2‐6xHis was expressed by E. coli BL21 (DE3) and purified using nickel‐charged affinity resin (Thermo Fisher Scientific) according to the manufacturer’s instruction. Purified c‐Myc‐6xHis was purchased from RayBiotech (Peachtree Corners, GA). Purified EBP2‐6xHis (5 µg), 6xHis‐c‐Myc (5 µg), and GST‐VgpA (10 µg) or GST‐VgpA^L10A (10 µg) were mixed together for 2 h to allow protein binding. Subsequently, the mixture was loaded onto protein A/G agarose that had been cross‐linked with anti‐EBP2 and incubated for 12 h at 4°C. Agarose was washed with PBS three times, and proteins were eluted using SDS loading buffer. Both input and eluates were Western blotted using anti‐EBP2, anti‐c‐Myc, and anti‐VgpA antibodies.

To determine the direct interaction between EBP2 and VgpA, equal amounts of purified GST‐VgpA (10 µg) or GST‐VgpA^L10A (10 µg) were mixed with purified recombinant EBP2‐6xHis (5 µg) and incubated for 2 h to allow binding. Subsequently, the mixtures were loaded onto protein A/G agarose that had been cross‐linked with anti‐VgpA and incubated for 12 h. Following extensive washing, proteins were eluted using SDS loading buffer. Both inputs and eluates were separated by SDS–PAGE and visualized by Coomassie blue staining.

To determine the interaction between VgpA and EBP2 in vivo, we performed pull‐down assay using protein samples isolated from intestinal epithelial cells of infant rabbits. Following 18 h of infection, infant rabbits were sacrificed and 1–2 cm pieces of small intestinal tissue was retrieved. The tissue was rinsed with wash media (RPMI1640 containing 2% FCS, 10 mM HEPES, 100 µg/ml penicillin/streptomycin and 50 µg/ml gentamicin) and cut longitudinally. The tissue was thoroughly rinsed again with Hanks' balanced salt solution (HBSS) to remove bacteria and then incubated with epithelial dissociation solution (HBSS containing 10 mM EDTA, 100 µg/ml penicillin/streptomycin, 10 mM HEPES and 2% FCS) for 5 min at 37°C. Following gentle shaking, the supernatant, containing dead/dying cells as well as colonized bacteria, was removed. The tissue was then incubated with epithelial dissociation solution containing EDTA (50 mM) for 30 min. Following vortexing, the supernatant, containing sheets of epithelial cells, was collected. Protein samples were extracted by incubating the epithelial cells with lysis buffer (1% Triton X‐100, 150 mM NaCl, 50 mM Tris–HCl) and then used in a pull‐down assay using A/G agarose that had been cross‐linked with anti‐EBP2 or anti‐VgpA. Western blots on the input epithelial lysates and output eluates were performed using anti‐VgpA and anti‐EBP2 antibodies. Representative blots of three biological experiments were shown.

Cycloheximide chase assay

Caco‐2 cells were infected with vopZ’ΔvscN1 or vopZ’ΔvscN1vgpA ^L10A at an MOI of 100 for 2 h and washed with PBS three times. The infected cells were maintained in DMEM containing 20 µg/ml gentamicin for 12 h. Subsequently, cells were treated with 50 µg/ml cycloheximide (CHX) for 5, 15, 30, 45, and 60 min. Treated cells were lysed, and Western blots were performed using anti‐c‐Myc antibody (1:100). Anti‐actin antibody (1:1,000) (Abcam) was used as a control. Representative blots of three biological experiments were shown.

Confocal microscopy

To visualize the localization of VgpA and EBP2, HeLa cells were transfected with a pCMV plasmid harboring VgpA (pCMV‐vgpA) or VgpA^L10A (pCMV‐vgpA^L10A). Subsequently, HeLa cells were labeled with anti‐VgpA antibody (1:100) and anti‐EBP2 antibody (1:100). Secondary antibody was FITC‐conjugated anti‐mouse IgG (1:200) and Alexa Fluor 568‐conjugated anti‐rabbit IgG (1:200). Confocal microscopy was used to visualize the co‐localization of VgpA and EBP2. To determine the influence of VgpA‐EBP2 interaction on c‐Myc localization, Caco‐2 cells were infected for 2 h and maintained in DMEM containing 20 μg/ml gentamicin for 12 h. Subsequently, infected cells or uninfected control cells were labeled with primary anti‐c‐Myc (1:100) and anti‐EBP2 (1:100) antibodies. FITC‐conjugated anti‐mouse IgG and Alexa Fluor 568‐conjugated anti‐rabbit IgG were used as secondary antibodies. Confocal microscopy was performed to visualize the localization of c‐Myc and EBP2. To visualize VgpA in vivo, infant rabbits were infected with WT or vgpA ^L10A (L10A was introduced into VgpA in the chromosome of WT). After 18 h of infection, distal small intestine samples were obtained and frozen sections were prepared as described previously (Ritchie et al, 2012). Sections were labeled with anti‐VgpA (1:100) and rhodamine phalloidin. FITC‐conjugated anti‐rabbit IgG was used as the secondary antibody to detect VgpA or VgpA^L10A. To analyze the localization of c‐Myc in vivo, infant rabbits were infected with WT or vgpA ^L10A. After 18 h of infection, distal small intestine samples were obtained and frozen sections were prepared. Sections were labeled with anti‐c‐Myc and anti‐EBP2. FITC‐conjugated anti‐mouse IgG and Alexa Fluor 568‐conjugated anti‐rabbit IgG were used as secondary antibody to visualize c‐Myc and EBP2, respectively. Representative images of three biological experiments were shown.

Quantitative RT–PCR

Total RNA was extracted from the Caco‐2 cells infected with V. parahaemolyticus strains for 2 h and maintained in DMEM containing 20 μg/ml gentamicin for 12 h. Following extraction, equal amounts of RNA (500 ng) were used to generate cDNA using cDNA synthesis kit (Bio‐Rad, Hercules, CA). The cDNA was then amplified with SYBR Green qPCR Master Mix (Bimake, USA) with primers that were specific to human rDNA and RNA Pol I gene on a Real‐Time PCR Detection System (Applied Biosystems, Foster City, CA). The transcript levels were normalized to those of GAPDH in each of the sample, and three independent experiments were performed. The 2^{−ΔΔ C T} method was used to compare the relative transcript level between infected and uninfected samples. To quantify the level of rRNA in rabbits, we extracted RNA from the distal intestine of the infant rabbits as described previously (Ritchie et al, 2012). Quantitative RT–PCR was performed for intestinal samples from three rabbits of each group, and the transcript levels were normalized to those of GAPDH. Comparison of the transcript levels in infected rabbits with those in the uninfected rabbits was performed using the 2^{−ΔΔ C T} method. Primers used in this study are listed in Appendix Table S2.

Fraction of infected Caco‐2 cells

Caco‐2 cells were infected with VgpA⁺ (vopZ’∆vscN1) or VgpA^L10A+ (vopZ’∆vscN1vgpA ^L10A) V. parahaemolyticus at an MOI of 100 for 2 h. After washing with PBS three times, the infected cells were maintained in the DMEM for 12 h in the presence of 20 µg/ml gentamicin to kill any bacteria remained in the medium. Approximately 10⁸ cells were washed twice with PBS and resuspended in 2 ml buffer (10 mM HEPES–KOH, pH 7.9, 1.5 M MgCl₂, 10 mM KCl, and 0.5 mM DTT) for 30 min at 4°C before phenylmethylsulfonyl fluoride at a final concentration of 0.2 mM was added. Cells were homogenized and centrifuged at 228 g for 5 min at 4°C. The supernatant was used as cytoplasmic fraction. The pellet was resuspended in 2 ml S1 buffer (0.25 M sucrose, 10 mM MgCl₂) and layered onto 2 ml of S2 buffer (0.35 M sucrose, 0.5 mM MgCl₂). After centrifugation at 1,430 g for 5 min, pelleted nuclei were resuspended in 2 ml of S2 buffer and sonicated to disrupt the nuclear membrane. The sonicated samples were layered over 2 ml of S3 buffer (0.88 M sucrose, 0.5 mM MgCl₂) and centrifuged at 2,800 g for 10 min. The pellet contains the nucleoli, and the supernatant was used as the nucleoplasm fraction. Whole cell lysates were obtained by resuspending the infected cells directly onto SDS Laemmli loading buffer. Each fraction was blotted with anti‐histone (Thermo Fisher Scientific), anti‐upstream binding factor (UBF) (Abcam), and anti‐tubulin (Thermo Fisher Scientific) antibodies to indicate successful isolation of nucleoplasm, nucleoli, and cytoplasm, respectively. Representative blots of three biological experiments were shown.

Proliferation assay

Caco‐2 cells were cultured on coverslips for 4 days to reach monolayers, with media change once every day. Cells were then infected with indicated V. parahaemolyticus strains for 2 h, followed by washing with PBS three times. Subsequently, the infected cells were maintained in DMEM for 12 h in the presence of 20 µg/ml gentamicin to kill any bacteria remaining in the medium. The BrdU assay was performed according to the manufacturer’s instruction (Sigma). The percentage of proliferating cells was calculated by dividing the number of BrdU‐positive cells by the total number of cells analyzed. Each experiment was performed three times, and the average percentage with standard deviation is presented. Knockdown of EBP2 or c‐Myc was carried out by transfecting Caco‐2 cells with siEBP2 (Thermo Fisher Scientific) or sic‐Myc (Santa Cruz Biotechnology, Dallas, Texas). Caco‐2 cells transfected with siEBP2 or sic‐Myc or control siRNA (siNC) were infected with vopZ’∆vscN1 or uninfected and proliferation was determined similarly as described above.

To determine the association of cell proliferation with VgpA translocation, Caco‐2 cells were infected with vopZ’∆vscN1 or vopZ’∆vscN1vgpA ^L10A as described above. Subsequently, VgpA was labeled with anti‐VgpA antibody and proliferating cells were labeled by BrdU. The percentage of BrdU‐positive cells among cells that are VgpA‐positive (cells containing detectable VgpA following infection with vopZ’∆vscN1) or VgpA‐negative (cells where VgpA was not observed) was calculated. Similarly, the percentage of BrdU‐positive cells among cells that were VgpA^L10A‐positive (cells containing detectable VgpA^L10A+ following infection with vopZ’∆vscN1vgpA ^L10A) or VgpA^L10A‐negative (cells where VgpA^L10A+ was not observed) was also calculated.

To analyze proliferation in vivo, infant rabbits were infected with the indicated strains for 18 h. Distal small intestine samples were obtained and processed for frozen sections. Tissue frozen sections were incubated with anti‐Ki67 monoclonal antibody at the dilution of 1:50 (Abcam) overnight at 4°C. After washing with PBS, the sections were incubated with Alexa Fluor 488‐conjugated anti‐mouse IgG. Rhodamine phalloidin and DAPI were used to label actin and nuclei, respectively. Relative proliferation (the length of the proliferating zone divided by the villus length) was calculated as described previously (Ritchie et al, 2012).

Statistical analysis

The percentage of BrdU‐positive cells (Fig 3C), relative rRNA level (Fig 5A) and RNA Pol I transcript level (Fig 5C), colonization (Fig 7A), and fluid accumulation ratios (Fig 7C) was statistically analyzed using one‐way analysis of variance (ANOVA) and Bonferroni’s test for multiple comparisons. Statistical analysis involving two independent factors (Figs 3D, and 5B and D, Appendix Figs S4 and S8) was carried by two‐way analysis of variance (ANOVA) and Bonferroni’s test for multiple comparisons.

Author contributions

XZ and MKW conceived the project. MH, YZ, DG, XC, and XZ performed the experiments. MH, XZ, and MKW analyzed the data. XZ and MKW wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Source Data for Appendix

Review Process File

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 6

The EMBO Journal (2021) 40: e105699.

Data availability

This study includes no data deposited in external repositories.