Distinct Molecular Features of NleG Type 3 Secreted Effectors Allow for Different Roles during Citrobacter rodentium Infection in Mice

ABSTRACT

The NleGs are the largest family of type 3 secreted effectors in attaching and effacing (A/E) pathogens, such as enterohemorrhagic Escherichia coli (EHEC), enteropathogenic E. coli, and Citrobacter rodentium. NleG effectors contain a conserved C-terminal U-box domain acting as a ubiquitin protein ligase and target host proteins via a variable N-terminal portion. The specific roles of these effectors during infection remain uncertain. Here, we demonstrate that the three NleG effectors—NleG1_Cr, NleG7_Cr, and NleG8_Cr—encoded by C. rodentium DBS100 play distinct roles during infection in mice. Using individual nleG_Cr knockout strains, we show that NleG7_Cr contributes to bacterial survival during enteric infection while NleG1_Cr promotes the expression of diarrheal symptoms and NleG8_Cr contributes to accelerated lethality in susceptible mice. Furthermore, the NleG8_Cr effector contains a C-terminal PDZ domain binding motif that enables interaction with the host protein GOPC. Both the PDZ domain binding motif and the ability to engage with host ubiquitination machinery via the intact U-box domain proved to be necessary for NleG8_Cr function, contributing to the observed phenotype during infection. We also establish that the PTZ binding motif in the EHEC NleG8 (NleG8_Ec) effector, which shares 60% identity with NleG8_Cr, is engaged in interactions with human GOPC. The crystal structure of the NleG8_Ec C-terminal peptide in complex with the GOPC PDZ domain, determined to 1.85 Å, revealed a conserved interaction mode similar to that observed between GOPC and eukaryotic PDZ domain binding motifs. Despite these common features, nleG8_Ec does not complement the ΔnleG8_Cr phenotype during infection, revealing functional diversification between these NleG effectors.

INTRODUCTION

Gastrointestinal infections caused by attaching and effacing (A/E) pathogens feature distinctive histological lesions characterized by the effacement of epithelial microvilli and by formation of actin-mediated pedestals. A/E pathogens include the clinically important enterohemorrhagic (EHEC) and enteropathogenic (EPEC) Escherichia coli, which can cause outbreaks of bloody diarrhea and, in the case of EHEC, hemolytic-uremic syndrome in humans, and the murine pathogen Citrobacter rodentium (1). C. rodentium employs an infection strategy similar to that of human A/E pathogens, providing a natural and physiologically relevant model to study the pathogenesis of this bacterial group in mice (2). Most mouse strains clear C. rodentium infection in 14 to 21 days postinoculation (dpi), while C3H/HeJ mice, which have decreased ion transport at the crypt surface, develop severe diarrhea and succumb to infection, providing a useful model to study the development of severe symptoms during infection (3).

A/E pathogens contain a highly conserved pathogenicity island that encodes proteins that form a type 3 secretion system (T3SS), which enables the translocation of effector proteins into host cells. Most T3SS effector families are conserved across EPEC, EHEC, and C. rodentium; however, the total number and diversity of the effector repertoire are specific to each strain (4). Collectively, T3SS effectors manipulate host cell processes to support infection and are essential for virulence and development of disease symptoms (4, 5). For example, Tir and EspF effectors interact with host proteins N-WASP and Arp2/3 to hijack the host cytoskeleton system to construct the pedestal and enable attachment of the extracellular bacterium (6). EspB and EspJ effectors inactivate myosin and Src kinase, respectively, to prevent phagocytosis of the attached pathogen (7, 8). The EspG effector inhibits Rab1 to impair endoplasmic reticulum (ER)-Golgi trafficking and disrupt epithelial cell tight junctions (9, 10). The NleD effector cleaves Jun N-terminal protein kinase (JNK) and p38 to suppress the host inflammatory response to the invading pathogen (11). The NleF effector inhibits caspase-4 activation to suppress host cell apoptosis (12). Apart from the abovementioned repertoire of enzymatic activities to modify their host targets, bacterial effectors often take advantage of eukaryote-specific posttranslational modification mechanisms, such as protein ubiquitination (4, 13).

Ubiquitination is a conserved mechanism of cellular signaling in all eukaryotic organisms. Ubiquitination involves covalent attachment of one or multiple copies of a small ubiquitin protein to a lysine residue side chain of the target protein. Depending on ubiquitin chain length, branching architecture, and position on the target protein, ubiquitination can signal the proteasome degradation or affect the protein’s localization, stability, and enzymatic activity (14, 15). The ubiquitination cascade includes three sequential steps of activation, conjugation, and ligation, performed by a ubiquitin-activating enzyme (E1), a ubiquitin-conjugating enzyme (E2), and a ubiquitin protein ligase (E3), respectively (14). The E3 enzymes constitute two major classes and form one of the largest functional groups encoded in the human genome. The HECT-type E3s have a catalytic cysteine residue that forms an intermediate thioester bond with ubiquitin prior to its transfer onto lysine residues on a target protein, while the E3s with a RING/U-box domain mediate direct transfer of ubiquitin from E2 onto target proteins (16, 17). The E3s usually control the specificity of the ubiquitination process. A large number of bacterial effectors have evolved to functionally mimic E3 activity, targeting host or other effector proteins for ubiquitination (18). Specifically in A/E pathogens, the NleL effector functionally mimics HECT-type E3 ligase activity while the NleG effectors, representing the largest family found in EHEC strains (19), feature an active U-box domain (20).

NleG effectors are relatively small proteins (about 200 amino acids [aa]) that contain a C-terminally located U-box domain and a highly variable N-terminal domain that is required for interaction with specific host target proteins (20, 21). To date, several dozen NleG effector family members have been identified in EHEC, EPEC, Salmonella, Shigella, and Citrobacter species (22). However, the molecular structure and the specific host targets are known for only a few NleGs, including EHEC NleG2-3 and NleG5-1, which initiate ubiquitination and degradation of human hexose kinase and a subunit of the MEDIATOR transcriptional regulation complex, respectively (21).

A recent bioinformatic survey suggested the presence of a putative PDZ domain binding motif in several host-cell-modifying effectors, including NleG3′, NleG8 (NleG), and NleG9′ from EHEC strain O157:H7 Sakai (23). In eukaryotic cells, these short C-terminal sequences are recognized by the specific PDZ domains, constituting widely distributed protein-protein interaction modules (24). Based on their specific sequences, PDZ binding motifs are categorized into three major classes, -Xaa-(Ser/Thr)-Xaa-ψ (class I), -Xaa-ψ-Xaa-ψ (class II), and -Xaa-(Asp/Glu)-Xaa-ψ (class III), where Xaa is any amino acid and ψ is a hydrophobic residue. The role of the PDZ binding motif in NleG activity and the general role of NleG effectors during infection remain uncharacterized.

We recently completed the genome sequencing of C. rodentium strain DBS100 (25), which revealed a repertoire of T3SS effector genes, including three encoding NleG effectors. In this study, we used C. rodentium and a murine model of infection to demonstrate that each of the three NleG effectors plays an important and distinct role during infection. Through structural and in cellulo analyses, we show that NleG8 homologues in C. rodentium and EHEC contain a functional PDZ domain binding motif, which allows for physical interactions between these effectors and the specific host protein GOPC. We also demonstrate that both the PDZ binding motif and E3 ligase activity of NleG8 are necessary for its role during infection, shining new light on the activity of this important family of pathogenic factors.

RESULTS

C. rodentium secretes three NleG effectors in a T3SS-dependent manner.

Recent genomic analysis of the C. rodentium DBS100 strain (here referred to as C. rodentium) identified several genes encoding potential T3SS effectors (25). Among these genes, three encoded proteins—NleG1_Cr, NleG7_Cr, and NleG8_Cr—share significant similarity with NleG effectors in EHEC and other pathogenic E. coli strains (21). In line with the previously defined molecular topology of this effector family (21), NleG1_Cr, NleG7_Cr, and NleG8_Cr feature a variable N-terminal portion and a conserved C-terminal U-box domain typical of eukaryotic E3 ubiquitin ligases. Two additional alleles—nleG1ψ and nleG8ψ—were also identified but contained a premature stop codon in their coding region and thus are likely pseudogenes in this strain. To test if nleG1_Cr, nleG7_Cr, and nleG8_Cr were specifically activated under T3S-inducing conditions, we constructed luciferase transcriptional reporters in C. rodentium using the intragenic region upstream of the translational start site for each nleG gene (PnleG1_Cr, PnleG7_Cr, and PnleG8_Cr) or the T3SS operon LEE3 as a control. A promoter lacking the construct was used as a negative control. The luminescence of each reporter strain was measured under T3S-inducing conditions and normalized to basal luminescence in LB medium. The positive-control PLEE3 showed upregulation of the Lux transcription from 2 to 5 h postinduction. Similarly, all three PnleG reporters showed Lux transcriptional upregulation under T3SS-inducing conditions from 1.5 (PnleG7_Cr) or 2 (PnleG1_Cr and PnleG8_Cr) to 5 h postinduction, suggesting that nleG1_Cr, nleG7_Cr, and nleG8_Cr are coregulated with T3SS genes (see Fig. S1A in the supplemental material). Next, we determined if NleG1_Cr, NleG7_Cr, and NleG8_Cr were secreted in a T3SS-dependent manner by comparing the secretion and intracellular protein profiles between wild-type (WT) C. rodentium and a ΔescN mutant lacking the ATPase subunit of the T3SS that is deficient for effector translocation (26). Each strain, constitutively expressing hemagglutinin (HA)-tagged NleG1_Cr, NleG7_Cr, or NleG8_Cr in trans, was incubated in T3SS-inducing medium. The secreted and cellular lysate protein fractions were assayed by Western blot analysis with anti-HA antibodies (Fig. S1B). The HA-tagged NleG proteins were detected in the bacterial lysate fraction of both wild-type and ΔescN strains, showing that the T3SS mutant does not alter the expression level of the proteins. However, NleG1_Cr, NleG7_Cr, and NleG8_Cr were detected only in the secreted fraction for the wild-type strain, confirming their T3SS-dependent secretion.

NleG effectors are key virulence factors during infection.

Since our genetic and secretion analysis showed that NleG1_Cr, NleG7_Cr, and NleG8_Cr are T3SS effectors, we tested the effect of individual nleG_Cr gene knockouts on C. rodentium infection in mice. C3H/HeJ mice were challenged with C. rodentium wild-type, ΔnleG1_Cr, ΔnleG7_Cr, and ΔnleG8_Cr strains and monitored for (i) the development of diarrhea symptoms by analysis of fecal water content, (ii) pathogen burden by enumerating C. rodentium in feces, and (iii) infection survival rates. We observed a significant reduction in diarrheal symptoms in mice infected with the ΔnleG1_Cr strain and a minor attenuation trend for ΔnleG8_Cr (Fig. 1A). The ΔnleG7_Cr and ΔnleG1_Cr strains had significant and moderate reduction of colonization burden during infection, respectively (Fig. 1B). In the case of ΔnleG8_Cr infection, the bacterial burden was similar to that of infection by the wild type; however, the mouse cohort infected with this strain demonstrated a significantly higher survival rate (Fig. 1C). Taken together, these results demonstrated that each of the three NleG effectors played an important and distinct role in the infection process of C. rodentium. Given the dramatic effect of nleG8_Cr deletion on the ability of the pathogen to cause fatal infection in mice, we followed up with detailed molecular analysis of the NleG8 effector.

FIG 1.

Effect of individual nleG_Cr gene knockouts on C. rodentium infection in mice. C3H/HeJ mice were challenged with wild-type C. rodentium (WT) or strains with deletions of the indicated nleG_Cr genes. (A) Fecal water content in mice infected with WT (n = 9 in two biological repeats), ΔnleG1_Cr (n = 6 in one biological repeat), ΔnleG7_Cr (n = 5 in one biological repeat), and ΔnleG8_Cr (n = 14 in three biological repeats) strains. Asterisks indicate statistical significance (Dunn’s multiple comparison, P value ≤ 0.05). (B) Bacterial burden in feces of mice infected with WT (n = 13 in three biological repeats), ΔnleG1_Cr (n = 10 in two biological repeats), ΔnleG7_Cr (n = 10 in two biological repeats), and ΔnleG8_Cr (n = 18 in four biological repeats) strains. Only strain ΔnleG7_Cr displays a significant difference (Mann-Whitney, P value ≤ 0.05) from WT at 9 and 10 dpi. (C) Survival of mice infected with WT (n = 13 in three biological repeats), ΔnleG1_Cr (n = 10 in two biological repeats), ΔnleG7_Cr (n = 10 in two biological repeats), and ΔnleG8_Cr (n = 18 in four biological repeats) strains. Only strain ΔnleG8_Cr displayed a significant difference (Mantel-Cox, P value ≤ 0.05) from WT.

Identification of host proteins interacting with NleG8.

In order to determine the host-interacting partners of NleG8, we conducted mass spectrometry-assisted affinity pulldown assays (AP-MS) using immobilized NleG8_Cr against human U937 and murine CT26.WT cell lysates in accordance with the previously established assays that identified the host interactors for NleG2-3 and NleG5-1 effectors (21). Coprecipitates were enriched for eight proteins in U937 lysates and 40 proteins in CT26.WT cell lysates (Table S1). Furthermore, the orthologs of all eight identified human proteins—SLC9A3R1, GOPC, ezrin, moesin, MAPK1, MAPK3, SEPT7, and SEPT2—were present in the coprecipitates from the mouse cell line proteome, suggesting that this protein set may contain a conserved host interaction partner of NleG8. GOPC and SLC9A3R1, which were identified as potential NleG8 interactors, are PDZ domain-containing scaffold proteins known to interact with ezrin and moesin (26, 27). This suggests a possible explanation for coprecipitation of these proteins with NleG8_Cr as part of multiprotein complexes.

To determine which of the coprecipitated host proteins may be directly interacting with NleG8_Cr, we tested their pairwise interactions using yeast two-hybrid assays (28). As observed previously, the E3 ubiquitin ligase activity of NleG effectors may lead to ubiquitination and proteasome degradation of the interaction partner, or the effector itself, thus possibly confounding yeast-two hybrid assay results (21). To account for this possibility, we introduced the I138K mutation into NleG8_Cr, which we have previously shown blocks NleG interaction with E2 ubiquitin-conjugating enzyme, rendering it unable to mediate ubiquitination (21). Among the pairs tested, only yeast strains coexpressing NleG8_Cr or NleG8_Cr(I138K) with human GOPC, which shares 93% of primary sequence with the mouse homologue, consistently demonstrated growth above the negative controls on interaction-selective medium (Fig. 2A). The growth of the yeast strain coexpressing the wild-type NleG8_Cr with GOPC was observed only on His⁻ selective medium, whereas the NleG8_Cr(I138K) mutant grew on both His⁻ and Ade⁻ selective media, indicating that this ubiquitination-deficient mutation further promotes NleG8_Cr interaction with GOPC (Fig. 2A).

FIG 2.

The NleG8_Cr effector directly interacts with host GOPC protein. Yeast two-hybrid experiments were performed using GAL4 DNA binding domain alone (empty vector [EV]) or fused to C. rodentium NleG8_Cr (NleG8_Cr), NleG8_Cr(I138K) mutant (I138K_Cr), and NleG8_Cr(I138K)[1-209] truncation mutant (I138K_Cr[1-209]) as bait coexpressed with GAL4 transcription activation domain alone (EV) or fused to the indicated mammalian proteins as prey. (A) Determination of direct interactions between NleG8_Cr and GOPC. (B) Analysis of NleG8_Cr PDZ binding motif and GOPC PDZ domain (GOPC_[276-362]) interaction. (C) His-tagged GOPC protein (His-GOPC) was incubated with streptavidin-conjugated beads alone or coated with His and streptavidin-binding peptide (SBP)-tagged C. rodentium NleG8_Cr (His-SBP-NleG8_Cr) and NleG8_Cr[1-209] truncation mutant (His-SBP-NleG8_Cr[1-209]). Beads were boiled in sample buffer, and soluble protein fraction was subjected to Western blot analysis using anti-His tag antibodies.

The NleG8 effector interacts with GOPC through a conserved C-terminal PDZ domain binding motif.

NleG8_Cr shares sequence similarity (60% sequence identity) with a T3SS effector from EHEC (Sakai strain), which by analogy we will here call NleG8_Ec (Fig. S2). NleG8_Ec was previously identified as part of the large group of bacterial effectors carrying a potential C-terminal PDZ domain binding motif (Fig. S2) (23). Comparative sequence analysis between NleG8_Ec and NleG8_Cr shows that NleG8_Cr has a similar PDZ domain binding motif at its C terminus corresponding to residues Tyr211, Arg212, and Leu213. The PDZ binding motif also appears to be unique to the NleG8 effectors, as sequence alignments suggest the absence of such a motif in NleG5-1, sharing 34% sequence identity with NleG8_Cr. Modeling of NleG8_Ec and NleG8_Cr based on structurally characterized NleG5-1 (21) suggests that the potential PDZ domain binding motif in NleG8 effectors represents a C-terminal extension beyond the conserved U-box domain in line with its proposed function as a protein-protein interaction motif (Fig. S2 and S3). To establish the role of the NleG8_Cr PDZ binding motif (Fig. S2 and S3), we constructed a ubiquitination-deficient NleG8_Cr(I138K) mutant variant lacking the four C-terminal residues, called NleG8_Cr(I138K)[1-209]. GOPC and SLC9A3R1 have been shown to self-regulate the binding activity of their PDZ domains (GOPC[276-362] and SLC9A3R1[1-235]) (26, 27). Therefore, we compared the interactions between NleG8_Cr(I138K) or NleG8_Cr(I138K)[1-209] and human GOPC in yeast two-hybrid assays, as well as interactions of these NleG8 variants with GOPC[276-362] and SLC9A3R1[1-235] fragments. Coexpression of GOPC[276-362] with NleG8_Cr(I138K) supported yeast growth on selective medium, suggesting that the GOPC PDZ domain is involved in interaction with NleG8_Cr (Fig. 2B). Nevertheless, these interactions were insufficient to support yeast growth on Ade⁻ selective medium, suggesting that other domains of GOPC may also be contributing to interactions with NleG8_Cr. No yeast growth was observed in the case of SLC9A3R1[1-235] coexpressed with NleG8_Cr(I138K), corroborating results observed for full-length SLC9A3R1 (Fig. 2A and B). Coexpression of NleG8_Cr(I138K)[1-209] with either GOPC or GOPC[276-362] did not support yeast growth in selective medium, suggesting that the NleG8_Cr C-terminal residues are critical for interactions with this host protein (Fig. 2B). To directly show that the PDZ binding motif in NleG8_Cr is required for interaction with GOPC, we tested the ability of NleG8_Cr and NleG8_Cr[1-209] to coprecipitate with GOPC using an in vitro pulldown. In accordance with the yeast two-hybrid data, GOPC coprecipitated with NleG8_Cr but not with NleG8_Cr[1-209] (Fig. 2).

Next, we tested if the NleG8_Ec PDZ domain binding motif could specifically interact with GOPC. Similarly to NleG_Cr, NleG8_Ec, NleG8_Ec(I138K), or NleG8_Ec(I138K)[1-211] was coexpressed with GOPC, GOPC[276-362], SLC9A3R1, or SLC9A3R1[1-235] in yeast two-hybrid assays. Coexpression of all the NleG8_Ec derivate constructs with SLC9A3R1 and SLC9A3R1[1-235] did not mediate yeast growth, while coexpression of NleG8_Ec(I138K), but not of NleG8_Ec(I138K)[1-211], with GOPC supported yeast growth, suggesting a specific interaction between these proteins (Fig. S4A). Notably, coexpression of NleG8_Ec with GOPC was not able to support yeast growth on selective medium. This was likely due to autoubiquitination and degradation of NleG8_Ec.Taken together, these results demonstrate that both NleG8_Cr and NleG8_Ec directly bind GOPC through their C-terminal PDZ binding motifs.

Finally, we tested if the PDZ domain binding motif in NleG8_Cr and NleG8_Ec could interfere with their E3 ubiquitin ligase activity. According to the results of previously established in vitro ubiquitination assays (20), NleG8_Cr and NleG8_Ec and their PDZ domain binding motif truncations NleG8_Cr[1-209] and NleG8_Ec[1-211] demonstrated similar abilities to support the formation of polyubiquitinated species (Fig. 3). This suggests that C-terminal PDZ domain binding motifs in both NleG8_Cr and NleG8_Ec do not interfere with their interactions with host ubiquitination cascade components.

FIG 3.

In vitro E3 ubiquitin ligase activity of NleG8. Immunoblot analysis using antiubiquitin (anti-Ubn) antibodies (top panel) and anti-streptavidin-binding peptide (anti-SBP) antibodies (bottom panel) and visualization of reactions performed with ATP, with E1 ubiquitin-activating enzyme, and with (+) or without (−) E2D2 ubiquitin-conjugating enzyme, and SBP-tagged C. rodentium NleG8_Cr (NleG8_Cr) and EHEC NleG8_Ec (NleG8_Ec) as well as their I138K mutated and C-terminally truncated forms.

Structure of the GOPC PDZ domain in complex with the NleG8_Ec PDZ domain binding motif.

To determine the structural basis for NleG8 interaction with GOPC, we conducted cocrystallization trials of GOPC[276-362] with peptides corresponding to the C-terminal fragment of NleG8_Cr or NleG8_Ec. We were able to obtain diffraction-quality crystals only for a GOPC[276-362] fragment in complex with the peptide LATQNICTRI, corresponding to the residues at position 0 to −9 in the C terminus of NleG8_Ec (P⁰ refers to the C-terminal residue), and this complex structure was determined to 1.85-Å resolution (PDB identifier [ID] 6XNJ). The structure showed that the GOPC PDZ domain consists of six β strands and three α helices, where the β2 strand and α3 helix form a cleft with a hydrophobic pocket (Fig. 4A and B). An additional density corresponding to NleG8_Ec PDZ binding sequence TQNICTRI (P⁻⁷ to P⁰) was identified along the hydrophobic cleft formed by the β2 strand and the α2 helix (Fig. 4A and B). The GOPC hydrophobic pocket is formed by L291, L348, and I295 and is flanked by an electropositive wall formed from the amino backbone groups from I291 and G292 (Fig. 4B). This pocket specifically coordinates the C-terminal residue I215 of NleG8_Ec by forming three hydrogen bonds with the carboxylic acid and amino group of I215. In the center of the GOPC cleft, the carbonyl and amino groups of I295 from β2 interact with the carbonyl and amino groups of T213 of NleG8_Ec (Fig. 4C). Furthermore, this H341 side chain interacts with the side chain of T213 of NleG8_Ec (Fig. 4C). This type of binding mode between GOPC and NleG8_Ec is typical of type I PDZ binding motifs (29). Although only P⁰ and P⁻² are conserved in GOPC-interacting proteins, residues up to P⁻¹⁰ of the type I PDZ binding motif affect affinity for the GOPC PDZ domain (30). Consistent with P⁰ and P⁻² conservation and importance, the NleG8_Ec I215 and T213 main chain forms five hydrogen bonds with GOPC L291, G292, I293, and I295 residues in the β2 strand, while their side chain forms hydrophobic interactions with L291, I293, I295, I328, L348, and F357 in the GOPC hydrophobic pocket and a hydrogen bond with GOPC H341 in the α2 helix, respectively (Fig. 4B and C). NleG8_Ec R214 was stacked on GOPC H311, demonstrating a structural explanation for the experimentally observed positive effect of this residue at P⁻¹ (Fig. 4C) (30). NleG8_Ec C212, I211, N210, and Q209 demonstrate a van der Waals interaction with GOPC[276-362], while NleG8_Ec T208 localizes outside the LATQNICTRI-PDZ domain interface, suggesting that NleG8_Ec-GOPC interaction is mainly defined by P⁰ to P⁻⁶ residues of NleG8_Ec. Further supporting this, we modeled the NleG8_Cr LPRSDTRL C-terminal peptide into the GOPC PDZ binding pocket (Fig. 4D). NleG8_Cr T211, R212, and L213 are predicted to have the same interactions with GOPC as NleG8_Ec T213, R214, and I215. Moreover, the backbone carbonyl of NleG8_Cr P207 is predicted to interact with GOPC His301 as observed for NleG8_Ec Q209.

FIG 4.

Structural analysis of NleG8 PDZ binding motif interaction with GOPC PDZ domain. (A) Structure of GOPC[276-362] (magenta) cocrystallized with EHEC NleG8_Ec C-terminal peptide LATQNICTRI (cyan). Indicated are beta sheet (β1 to β6, purple) and α helix (α1 to α3, pink) secondary structure elements, N and C termini of GOPC fragment, and NleG8_Ec peptide LATQNICTRI. (B) GOPC[276-362] colored by surface electrostatic potential from red (−5 kT/e) to blue (+5 kT/e). NleG8_Ec C-terminal peptide LATQNICTRI (cyan sticks) is shown with the corresponding 2|F_o-F_c| electron density map contoured at 1.0 σ. (C and D) Interacting residues of GOPC PDZ domain (sticks) with the NleG8_Ec cocrystallized LATQNICTRI peptide (lines) (C) and modeled NleG8_Cr LPRSDTRL peptide (lines) (D). Indicated residues in GOPC (black) and LATQNICTRI (cyan) form a hydrogen bond (dashed black lines), stacking, and a hydrophobic interaction.

NleG8 does not affect the level of GOPC, suggesting another ubiquitination target.

To investigate localization of NleG8 effectors in host cells, we expressed FLAG-tagged EHEC NleG8_Ec, NleG8_Ec(I138K), and NleG8_Ec[1-211] variants and NleG8_Cr, NleG8_Cr(I138K), and NleG8_Cr[1-209] in human HEK 293 and murine CT26.WT cells, respectively. In HEK 293 cells, FLAG-tagged proteins were detected for all NleG8 expression constructs. The NleG8-specific signal was distributed throughout the cell cytosol (Fig. S5). Cytosolic localization of NleG8 effectors would allow for interactions with GOPC protein, which is associated with the cytosolic face of the Golgi apparatus (Fig. S5) (26). The Western blot analysis using anti-FLAG antibodies demonstrated that the FLAG-tagged NleG8_Ec(I138K) variant was expressed at a higher level than NleG8_Ec and NleG8_Ec[1-211] variants, probably due to autoubiquitination of the latter effector variants (Fig. S6). To analyze the NleG8_Ec interaction with GOPC in cellulo, we tested coprecipitation of the latter protein from HEK 293 cells expressing effectors using anti-FLAG antibodies. GOPC coprecipitated with FLAG-NleG8_Cr and FLAG-NleG8_Ec but not with effector variants lacking the C-terminal PDZ binding motif, in line with the observed role of these residues in interaction (Fig. S6).

Previously characterized NleG effectors triggered ubiquitination and degradation of their host interaction partners (21). Accordingly, we analyzed the level of GOPC proteins in HEK 293 cells expressing FLAG-tagged NleG8_Ec in comparison to the cells expressing NleG8_Ec[1-211] lacking the PDZ binding motif or NleG8_Ec(I138K), which were unable to engage with host E2 enzyme. Our Western blot analysis using GOPC-specific antibodies did not reveal any significant changes in GOPC protein level between these cell lines (Fig. S6). Only the unmodified version of GOPC was detected in cells expressing NleG8_Ec (Fig. S6), which together with our previous data is consistent with the formation of a GOPC-NleG8_Ec complex that does not lead to ubiquitination of this host protein. In CT26.WT cells, we were not able to detect FLAG-tagged NleG8_Cr after transfection, possibly due to the low transfection efficiency. To validate the NleG8_Cr-GOPC interaction in the murine proteome, we compared the proteins coprecipitated with immobilized NleG8_Cr or the NleG8_Cr[1-209] variant from murine CT26.WT cell lysate. GOPC was unambiguously identified in the coprecipitation mix of NleG8_Cr but not NleG8_Cr[1-209] (Table S1), consistent with the PDZ binding motif at the C terminus of this effector being critical for GOPC interaction.

GOPC is a scaffold protein shown to be involved in numerous cellular processes, including cell polarization, ion homeostasis, endocytosis, and formation of the mucus layer and tight junctions (26, 31–34). Since GOPC interactions with the NleG8 effector did not result in ubiquitination and degradation of the latter protein, we hypothesized this interaction grants NleG8 the ability to modulate the function of GOPC-interacting protein(s) and subvert corresponding cell processes. To test this, we did liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis of the proteome of human HEK 293 cells ectopically expressing NleG8_Ec in comparison with the same cell line expressing the NleG8_Ec(I138K) or NleG8_Ec[1-211] variant. Transfected HEK 293 cells were collected 30 h posttransfection, and their proteome was analyzed by mass spectrometry after trypsin digestion. The search against the human proteome of obtained spectra allowed for identification of over 3,100 proteins in each proteome sample. Each proteome analysis was performed in triplicate, and only proteins identified in all three biological repeats were considered for further analysis.

To identify host proteins whose level is negatively affected by NleG8_Ec, we considered proteins present in cells expressing NleG8_Ec variants defective either in ubiquitination or in interaction with GOPC but not in cells expressing the wild-type effector. This analysis revealed 22 human proteins (Table S2) that have been assigned to 11 distinct cellular processes. Specifically, P08758 is an inhibitor of blood coagulation (35), P05023 is a sodium and potassium pump that regulates the electrochemical gradient across the plasma membrane (36), P0DP24 is a positive regulator of calcium-activated potassium channel SK (37), and P51149 is a Ras-related protein, Rab7a, which plays a pivotal role in endocytosis by inducing maturation of an early endosome into a late endosome (38). The decrease in levels of P08758, P05023, and P0DP24 proteins triggered by expression of NleG_Ec provides a possible explanation for watery and bloody diarrhea symptom development during clinical EHEC infection (1, 39). In addition, specific depletion of Rab7a suggests that EHEC uses NleG8_Ec to subvert host cell endocytosis. This observation is in line with EHEC-induced accumulation of endosomal vesicles and the molecular function of the EHEC EspG effector that prevents early endosome maturation into a recycling endosome through inhibition of Rab35 (40).

Depending on ubiquitin chain type and the specific position of ubiquitination in the target protein, ubiquitination can trigger increased stability instead of proteasome degradation (41). Hence, we also queried proteomes expressing NleG8_Ec effector variants for proteins exhibiting increased stability in cells expressing the wild-type NleG8_Ec. We identified 24 host proteins present in the lysate of cells expressing the wild-type effector but not in cells expressing NleG_Ec variants (Table S2). These 24 host proteins can be assigned to nine cellular processes (Table S2). Specifically, P30040 is a positive regulator of protein biosynthesis in the endoplasmic reticulum including CFTR and ENaC, an epithelial chloride and a sodium channel, respectively (42), Q9BT67 is a negative regulator of proteins in the Golgi apparatus including the epithelial potassium channel hERG (43), and Q01518 plays a role in the establishment of cell polarity via inhibition of actin polymerization into filament (F-actin) and induction of disassociation of F-actin to its actin monomers (44). The NleG8_Ec-mediated increase in P30040 and Q9BT67 provides an additional mechanism for this effector to contribute to the watery diarrhea symptom development during EHEC infection. Additionally, observed upregulation of Q01518 in cells expressing NleG8_Ec can suppress F-actin, facilitating the EHEC attachment to the infected cell, as the inhibition of F-actin in the host cell has been shown to prevent the internalization of the invasive strain (45). In total, our analysis suggested that ectopic expression of NleG8_Ec altered the expression of 46 host proteins affecting at least 16 cellular processes, suggesting that this effector may have more than one ubiquitination target.

Both E3 ligase activity and the PDZ binding motif are essential for NleG8 function.

Since the ΔnleG8_Cr C. rodentium strain was attenuated in mouse infections, we further characterized the role of the NleG8_Cr PDZ domain in this phenotype. Introduction of an additional plasmid into C. rodentium severely hindered virulence, making transcomplementation studies unfeasible (data not shown). To overcome this, we constructed chromosomal cis complementation strains for nleG8_Cr (see Materials and Methods). To test individual roles of NleG8’s ability to engage host E2 ubiquitin-conjugating enzyme via its U-box domain and GOPC via the PDZ binding motif, we reintroduced the wild-type nleG8_Cr, nleG8_Cr(I138K), nleG8_Cr[1-209], and nleG8_Ec into the chromosome of the ΔnleG8_Cr strain. The expression of NleG8 variants in these strains, which we will refer to as ΔNleG8_Cr:NleG8_Cr, ΔNleG8_Cr:NleG8_Cr(I138K), ΔNleG8_Cr:NleG8_Cr[1-209], and ΔNleG8_Cr:NleG8_Ec, respectively, was confirmed by mass spectrometry analysis of corresponding bacterial strain proteomes (Table S3). To validate T3S-dependent secretion of the NleG8 variants, we expressed in trans HA-tagged NleG8_Cr, NleG8_Cr(I138K), NleG8_Cr[1-209], and NleG8_Ec in C. rodentium DBS100 wild-type and ΔescN strains as was described above. All four HA-tagged NleG8 variants were secreted by the C. rodentium DBS100 wild type but not by the ΔescN strain (Fig. S7).

To confirm T3S-mediated translocation of the NleG8 variants into the host cells, we constructed plasmids for expression of NleG8_Cr, NleG8_Cr(I138K), NleG8_Cr[1-209], and NleG8_Ec fused to the TEM-1 beta-lactamase reporter. TEM-1 translocation inside host cells can be tracked by monitoring the fluorescence signal due to the cleavage of the CCF2 substrate (46). Next, we infected HeLa cells preloaded with CCF2 with the EPEC wild type or the T3S-deficient ΔescN mutant strains expressing the NleG8 variants fused to TEM-1 and monitored the fluorescent signal up to 100 min postinfection. A significant increase in fluorescence was detected starting from 60 to 70 min postinfection in HeLa cells infected with EPEC expressing NleG8_Cr-TEM-1, NleG8_Cr(I138K)-TEM-1, NleG8_Cr[1-209]-TEM-1, and NleG8_Ec-TEM-1 compared to cells infected with the ΔescN strains (Fig. S8A). The Western blot analysis established that NleG8 variants fused to TEM-1 were expressed to similar levels in EPEC wild-type and ΔescN strains (Fig. S8B). These results confirmed the T3SS-dependent translocation of the four NleG8 variants.

Next, we challenged C3H/HeJ mice with the ΔNleG8_Cr C. rodentium strains complemented in cis with the nleG8 variants described above and monitored the development of infection for 14 days. Mice infected with strains expressing NleG8_Cr variants showed no difference in bacterial burden from the ones infected with the strain expressing the wild-type NleG8_Cr (Fig. 5A and Table S4). However, the cohort infected with the strains expressing NleG8_Cr(I138K) or NleG8_Cr[1-209] and the ΔNleG8_Cr strain showed on average a 26% (standard deviation [SD], ±10.4%) higher survival rate than the mice infected with the strain expressing wild-type NleG8_Cr. We also noticed that ΔNleG8_Cr:NleG8_Cr-infected mice succumbed to infection on average 26% (SD, ±4%) faster than mice challenged with ΔNleG8_Cr, ΔNleG8_Cr:NleG8_Cr(I138K), and ΔNleG8_Cr:NleG8_Cr[1-209] (Fig. 5B). These results suggested that both the C-terminal PDZ domain binding motif and the intact U-box domain were critical for NleG8_Cr function during infection.

FIG 5.

Complementation of nleG8_Cr gene in C. rodentium DBS100 with various functional mutants. C3H/HeJ mice were challenged with C. rodentium ΔnleG8_Cr (ΔNleG8_Cr) or strains complemented with nleG8_Cr (ΔNleG8_Cr:NleG8_Cr), nleG8_Ec (ΔNleG8_Cr:NleG8_Ec), nleG8_Cr(I138K) [ΔNleG8_Cr:NleG8_Cr(I138K)], or nleG8_Cr[1-209] (ΔNleG8_Cr:NleG8_Cr[1-209]). (A) C. rodentium burden in feces of mice infected with ΔNleG8_Cr (n = 18 in four biological repeats), ΔNleG8_Cr:NleG8_Cr (n = 9 in two biological repeats), ΔNleG8_Cr:NleG8_Ec (n = 8 in two biological repeats), ΔNleG8_Cr:NleG8_Cr(I138K) (n = 8 in two biological repeats), and ΔNleG8_Cr:NleG8_Cr[1-209] (n = 8 in two biological repeats) strains. (B) Survival of mice infected with ΔNleG8_Cr (n = 18 in four biological repeats), ΔNleG8_Cr:NleG8_Cr (n = 9 in two biological repeats), ΔNleG8_Cr:NleG8_Ec (n = 8 in two biological repeats), ΔNleG8_Cr:NleG8_Cr(I138K) (n = 8 in two biological repeats), and ΔNleG8_Cr:NleG8_Cr[1-209] (n = 8 in two biological repeats) strains. The cohort infected with the ΔNleG8_Cr:NleG8_Cr strain displayed significant susceptibility (Mantel-Cox, P value ≤ 0.05) compared to each of the other cohorts. (C) Fecal water content in mice infected with ΔNleG8_Cr:NleG8_Cr (n = 7 in two biological repeats), ΔNleG8_Cr (n = 14 in three biological repeats), ΔNleG8_Cr:NleG8_Ec (n = 7 in two biological repeats), ΔNleG8_Cr:NleG8_Cr(I138K) (n = 8 in two biological repeats), and ΔNleG8_Cr:NleG8_Cr[1-209] (n = 8 in two biological repeats) strains. Asterisks indicate statistical significance (Dunn’s multiple comparison, P value ≤ 0.05).

Mice infected with the ΔNleG8_Cr:NleG8_Ec strain showed bacterial burdens similar to the burdens of the mice infected with the ΔNleG8_Cr:NleG8_Cr strain (Fig. 5A and Table S4). However, in contrast to mice infected with the ΔNleG8_Cr:NleG8_Cr strain, those infected with the ΔNleG8_Cr:NleG8_Ec strain survived the infection on average 20% longer and up to 14 dpi (25%) (Fig. 5B). This suggested that despite significant similarity and the common ability to engage with GOPC protein through the PDZ domain binding motif, the expression of NleG8_Ec did not complement the ΔnleG8_Cr phenotype in the mouse infection model and these effectors may have distinct functions in EHEC and C. rodentium, respectively.

DISCUSSION

Delivering a specific set of effectors into the host during infection is a common strategy evolved by many human and animal pathogens. However, the specific role of many effectors remains elusive, hampering our progress in understanding this important and widespread pathogenic strategy. In pathogenic E. coli strains causing life-threatening gastroenteric infections, the NleG effectors represent the largest effector family secreted by the T3SS. Despite the progress in the characterization of common E3 ubiquitin ligase activity of these effectors, the role of specific members of this family during infection remained unclear. In this study, we tested the role of NleG effectors encoded by the murine pathogen C. rodentium, which serves as a model for studying A/E bacterial infections.

Our genomic, transcription, and secretion analysis identified three NleG effectors—NleG1_Cr, NleG7_Cr, and NleG8_Cr—as part of the T3SS arsenal of C. rodentium. We next demonstrated that deletion of each of the corresponding nleG genes resulted in significant reduction in C. rodentium virulence in mice. In line with functional diversification between members of the NleG family (21), deletion of individual nleG genes evokes distinct phenotypical aberrations during infection. The deletion of nleG1_Cr affected diarrheal symptoms, while deletion of nleG7_Cr negatively impacted the rate of C. rodentium colonization during infection. Finally, the nleG8_Cr deletion triggered an increase in survival rate in infected animals. Combined, these data provide the first direct evidence that NleG effectors are key virulence contributors in Citrobacter and open the opportunity to investigate their role in other A/E pathogens. Furthermore, they provide a functional insight into the specific role of certain NleG family members, which supports prior reports of correlation between NleG expression and increased virulence in non-O157 Shiga-toxin-containing E. coli strains isolated from human patients (22, 47).

The significant change in fecal water content during infection triggered by deletion of nleG1 highlights the importance of this infection-associated host condition as a virulence strategy of A/E pathogens. Mouse strains impaired in the gastric water absorption system develop severe diarrhea and die when infected with C. rodentium, while most mouse strains are able to clear this pathogen (3). In the case of human A/E pathogens, multiple EHEC virulence factors have been characterized to target the ion export in the colon (39). An increased ion concentration elevates the osmotic pressure which in turn impairs water absorption from feces, leading to diarrhea. Diarrhea symptoms can be advantageous for A/E pathogen infection by facilitating the pathogens’ dissemination, eliminating nonadherent microorganisms and thus reducing competition (48), and inhibiting or dampening host immune responses (49). Our data provide the first evidence that NleG effectors are important for C. rodentium virulence and pave the way to future studies delineating the exact function of each NleG effector during infection.

Through a combination of affinity purification and yeast two-hybrid assays, we demonstrate that the NleG8_Cr effector interacts with host coiled-coil motif-containing protein (GOPC) (26). This interaction is mediated by the C-terminal sequence extension in NleG8_Cr representing the class I PDZ domain binding motif. Via this additional molecular feature, the NleG8_Cr and EHEC NleG8_Ec effectors specifically interact with the PDZ domain of GOPC (26). Our structural analysis showed that the NleG8_Cr C-terminal sequence interacts with the GOPC PDZ domain following the general molecular architecture established for this domain’s recognition of eukaryotic PDZ binding motifs. In addition to NleG8, PDZ binding motifs have been predicted in NleG3 and NleG9, but their function has never been investigated (23). Based on our data, we propose that along with NleG8, NleG3 and NleG9 effectors are using their C-terminal sequence for interaction with yet-unidentified host proteins. The role these interactions play in the activity of NleG3 and NleG9 effectors will need further investigation.

NleG8 expressed in human cells did not affect the level of GOPC expression. We also were not able to detect any non-proteasome-related ubiquitination of GOPC in NleG-expressing cells. These data suggest that despite direct interactions NleG8 does not mediate ubiquitination of GOPC. In eukaryotic cells GOPC functions as the master regulator controlling the trafficking of cell surface receptors, ion channels, pumps, and adhesion molecules (31). Specifically, GOPC mediates lysosomal degradation of the CFTR Cl⁻ channel and Muc3 while facilitating the trafficking of Frizzled receptor to the cell surface (26, 32–34). Along the same lines, GOPC is involved in numerous cellular processes including cell polarization, formation of tight junctions, ion homeostasis, endocytosis, and mucus layer formation. Several of these processes are affected during infection by intestinal pathogens. Colonic epithelium mucus layers of mice and humans infected with Citrobacter and Salmonella, respectively, have been shown to be depleted of certain Muc proteins and enriched in Muc1 (50). A/E pathogens such as EPEC and EHEC bind to mucin proteins (51). EHEC actively subverts the host mucus layer by secretion of StcE zinc metalloprotease to digest mucin proteins (52). EPEC disrupts intestinal epithelial cell tight junctions via the function of several effector proteins including the T3SS effector EspG (10). Thus, interaction with GOPC protein can provide the NleG8 effector with a specific mechanism for interfering with these processes. Supporting this theory, we showed that ectopically expressed EHEC NleG8, but not the ubiquitination- or PDZ binding-impaired variants, affected the level of 46 host cell proteins involved in 16 cellular processes, including ion homeostasis, endocytosis, blood coagulation, and cell polarization. One possibility is that NleG8 uses the PDZ binding motif to initiate the ubiquitination of targets in the GOPC interaction network. A similar mechanism was observed for E3 ligase MARCH2, which uses interactions with the PDZ domain of GOPC to form a MARCH2-GOPC-CFTR complex and trigger ubiquitination of the CFTR Cl⁻ channel (53). Our observation that E3 ubiquitin ligase activity along with an intact PDZ binding motif is required for NleG8_Cr function during infection supports this hypothesis. Intriguingly, EHEC NleG8_Ec was not able to complement NleG8_Cr function during infection. This suggests that despite significant sequence similarity and the common PDZ binding motif, NleG8_Cr and NleG8_Ec effectors are targeting different host proteins for ubiquitination. In line with our findings, Citrobacter and EHEC use different Tir-based mechanisms for pedestal formation, and their Tir effectors are not functionally interchangeable (54). As mentioned above, the NleG8_Cr and NleG8_Ec interactions with GOPC may provide access to diverse signaling cascades in host cells, providing these otherwise homologous effectors with a variety of ubiquitination targets. Further studies are under way to identify possible ubiquitination targets of NleG8 effectors.

In conclusion, our systematic analysis of the NleG arsenal in the murine pathogen C. rodentium DBS100 highlighted the importance of this effector family in the infection strategy of A/E pathogens. We also characterized the additional molecular features present in several members of this family, which mediate the specific interaction with host PDZ domain-containing proteins. This study advances our understanding of molecular mechanisms developed by pathogenic bacteria to control the host response during life-threatening infections.

MATERIALS AND METHODS

DNA manipulations.

Oligonucleotides and plasmids used in this study are listed in Tables S5 and S6 in the supplemental material, respectively. PCR was performed using Phusion HF DNA polymerase (New England Biolabs [NEB], Canada). Primers were used to introduce ligation-independent cloning (Lic) sequences, attB, and restriction sites at 5′ and 3′ ends of amplified fragments for Lic (55), Gateway, and restriction enzyme-mediated cloning, respectively. Point mutations were introduced by PCR-amplifying plasmids harboring cassettes for protein expression using primers encoding the mutation and self-ligation of the PCR product. Constructs were validated by sequencing, and plasmid DNA was extracted using the Presto Mini plasmid kit and Geneaid plasmid maxikit (Geneaid, Taiwan).

Luciferase assay.

Transcriptional reporter constructs were generated by amplifying the intergenic region upstream of the translational start site of nleG1_Cr (bp 975131 to 975630), nleG7_Cr (bp 4364899 to 4365379), nleG8_Cr (bp 3540515 to 3541014), and T3SS operon LEE3 (bp 2378037 to 2378522) in the C. rodentium strain DBS100 genome (CP038008). These fragments were then cloned into pGEN-em7-luxCDABE through restriction-based cloning, replacing the em7 constitutive promoter sequence (56). A promoterless negative control was generated similarly by digesting pGEN-em7-luxCDABE with PmeI and SnaBI, followed by religation. All plasmids were then transformed into C. rodentium DBS100 for analysis. Reporter strains were grown overnight at 37°C with shaking in LB supplemented with 50 μg/mL gentamicin and then subcultured 1:100 into either T3S-inducing medium (prewarmed Dulbecco’s modified Eagle’s medium [DMEM; Gibco]) (57) or LB in clear-bottom black 96-well plates (Corning) in technical triplicates. The 96-well plates were then incubated under standing conditions at 37°C with a 5% CO₂ atmosphere for DMEM or 37°C with shaking for LB, with readings of luminescence and optical density at 600 nm (OD₆₀₀) occurring every 30 min on a PerkinElmer Envision plate reader. The LB Lux/OD₆₀₀ data were used as the background level of promoter activation to blank the DMEM Lux/OD₆₀₀ data. This experiment was repeated twice with similar results.

In vitro secretion assay.

The genes encoding NleG1_Cr (bp 974546 to 975130), NleG7_Cr (bp 4364272 to 4364898), and NleG8_Cr (bp 3539873 to 3540514) were amplified from the C. rodentium DBS100 genome (CP038008) and cloned in frame with an HA tag. C-terminally HA-tagged effectors were expressed in the C. rodentium strain DBS100 wild type or escN deletion mutant, and the secretion assay was done essentially as described previously with minor modifications (57). Fifteen milliliters of DMEM (Gibco) supplemented with 15 μg/mL gentamicin was inoculated with 300 μL of overnight LB culture from the indicated strains and incubated for 5.5 h at 37°C in a 5% CO₂ atmosphere without shaking. For whole-cell lysate protein analysis, bacteria were collected by centrifugation (4°C, 30 s, 16,000 × g) from 1.5 mL of culture, resuspended in 50 μL of sample buffer, and boiled for 7 min. For secreted protein fraction analysis, the culture was centrifuged (4°C, 10 min, 4,300 × g), and the supernatant was filtered (0.2-μm filter), concentrated to 50 μL using an Amicon Ultra-4 10K concentrator (Millipore Sigma), resuspended in 75 μL sample buffer, and boiled for 7 min. For secreted and whole-cell lysate protein fractions, 20 μL/lane and 10 μL/lane, respectively, were separated on a 10% SDS-PAGE gel and visualized by Coomassie blue G250 staining and Western blotting using anti-HA antibody.

DBS100 mutagenesis.

The nleG1_Cr fragment from 64 to 364 bp was PCR amplified and cloned into the pRE112 suicide vector (58). To construct nleG7_Cr and nleG8_Cr deletion cassettes, the ≈1-kb fragments flanking the gene of interest were PCR amplified from the C. rodentium strain DBS100 genome and cloned into pRE112. For chromosomal complementation of ΔnleG8_Cr, the DNA fragment containing nleG8_Cr, nleG8_Cr(I138K), nleG8_Cr[1-209], and nleG8_Ec was PCR amplified and cloned into pRE112 harboring the nleG8_Cr deletion cassette. Chromosomal deletion, insertion, and complementation of genes in the DBS100 genome were done as previously described (58). Briefly, electrocompetent C. rodentium strain DBS100 was transformed with the pRE112 plasmid harboring the insertion, deletion, or complementation cassette enabling positive and negative selection using gentamicin and sucrose, respectively. Single recombination events were selected by plating transformed DBS100 on LB medium supplemented with gentamicin (15 μg/mL). For deletion and complementation only, a single colony was grown at 37°C in LB medium to an OD₆₀₀ of 0.6 and diluted in a 10-fold series and 100 μL of a 10⁻³ to 10⁻⁶ dilution was plated on LB supplemented with 2% sucrose. Double recombination events were selected by isolating single colonies that were gentamicin sensitive and sucrose resistant. Deletion and complementation mutants were verified by PCR and sequencing of the mutated locus.

Protein expression and purification.

Lic-compatible expression plasmids pMCSG68SBPTEV and pMCSG53 were used for heterologous protein expression fused to 6×His-streptavidin-binding peptide (SBP) and 6×His tag, respectively. Protein expression and purification were performed essentially as described previously (20). Briefly, E. coli strain BL21-Gold(DE3) (Stratagene) was transformed with the pMCSG68SBPTEV or pMCSG53 plasmid for expression of the genes of interest with N-terminal tandem 6×His and streptavidin-binding peptide (SBP) tag or 6×His tag, respectively. Bacteria were incubated at 37°C and 220 rpm in LB medium supplemented with 100 μg/mL ampicillin. At an OD₆₀₀ of 0.6 to 0.8, protein expression was induced with 1 mM isopropyl-beta-d-thiogalactopyranoside (IPTG; BioShop) overnight at 16°C and 220 rpm. Cells were harvested by centrifugation and lysed by sonication in binding buffer [50 mM HEPES (pH 7.5), 300 mM NaCl, 5% (vol/vol) glycerol, 0.5 mM tris-(2-carboxyethyl)phosphine] supplemented with 1 mM benzamidine and 0.5 mM phenylmethylsulfonyl fluoride (PMSF), and the soluble protein fraction was collected using centrifugation. The soluble protein fraction was incubated with nickel-nitrilotriacetic acid (Ni-NTA) agarose beads (Thermo Fisher Scientific) and washed using binding buffer with 30 mM imidazole, and the 6×His-tagged proteins were eluted using binding buffer with 0.5 M imidazole, 1 mM EDTA (pH 8.5), and 0.5 mM tris-(2-carboxyethyl)phosphine. Purified proteins were concentrated using an Amicon Ultra-4 10K concentrator (Millipore Sigma) and stored at −80°C.

Yeast two-hybrid assay.

Bait and prey proteins were fused to the GAL4 transcriptional activating domain (AD) or the DNA binding domain (DB) using pDEST-AD-ccdB and pDEST-DB-ccdB constitutively active gateway destination plasmids (28). The pDEST-AD and pDEST-DB vectors were cotransformed into Saccharomyces cerevisiae strain Y8800 and plated onto SD plates omitting Trp and Leu with 2% glucose as carbon source at 30°C for 2 to 3 days. Each of the Trp and Leu autotrophic strains was grown overnight in liquid SD omitting Trp and Leu, centrifuged, washed with H₂O, and spotted at least in duplicate onto three plates: (i) SD omitting Trp and Leu to control for growth inhibition effect; (ii) SD omitting Trp, Leu, and His for selection; and (iii) SD omitting Trp, Leu, and adenine for strong selection. Plates were imaged 3 days after incubation at 30°C. The experiment was repeated four times using independent colonies. Expression in yeast of all eight NleG8_Cr-interacting protein constructs was verified by Western blot analysis (see Fig. S4B and C in the supplemental material), while NleG8_Cr and NleG8_Ec wild types and their mutated forms were expressed at an undetectable level (Fig. S4D).

Subcellular localization analysis.

Effector genes were PCR amplified and cloned in frame with FLAG tag into pcDNA3.1/nFLAG-DEST plasmid. HEK 293T/17 SF cells (ACS-4500) were grown in T75 flasks with 15 mL DMEM, 10% fetal bovine serum (FBS), and 1% (vol/vol/) penicillin-streptomycin at 37°C and 5% CO₂. A day prior to transfection, cells were trypsin digested and seeded at 6.3 × 10⁴ cells per well into a 12-well plate with a coverslip (Electron Microscopy Sciences; catalog no. 72222-01). A Lipofectamine 3000 kit (Invitrogen, USA) was used to transiently transfect the cells following the manufacturer’s protocol. Cells on the coverslip were fixed with 4% (vol/vol) paraformaldehyde (PFA) for 15 min at room temperature, permeabilized with 0.5% (vol/vol) Tween 20, blocked using phosphate-buffered saline (PBS) supplemented with 1% bovine serum albumin (BSA) for 30 min, and incubated 1 h with anti-FLAG (Sigma-Aldrich; F1804) and anti-GOPC (Abcam; ab37036) antibody at 2 μg/mL in 3% BSA-PBS at room temperature. The following steps were carried out in the dark. Coverslips were incubated for 1 h with anti-mouse IgG conjugated to Alexa Fluor 647 (Abcam; ab150107) and anti-rabbit IgG conjugated to Alexa Fluor 488 (Abcam; ab150073) diluted in 1% BSA-PBS, DAPI (4′,6-diamidino-2-phenylindole) stained (1 μg/mL, PBS) for 1 min, mounted with Prolong Gold antifade solution (Invitrogen, USA), sealed with nail polish, and stored at 4°C. A DMI4000 B microscope (Leica Microsystems, Canada) and Quorum Angstrom software (Quorum Technologies, Canada) were used to visualize the samples in phase mode and with a filter set of λ_ex of 390 nm, 490 nm, and 620 nm and λ_em of 455 nm, 525 nm, and 700 nm for DAPI, Alexa Fluor 488, and Alexa Fluor 647, respectively. The Icy software was used for image processing.

Structure modeling.

The three-dimensional (3D) structure model of NleG8_Cr and NleG8_Ec was predicted using the multiple-threading template modeling approach of the I-TASSER server (59). The models were refined using the GalaxyWEB server (60). Model refinement statistics are summarized in Table S7. Refined models were visualized and aligned with the NleG5-1_Ec protein structure (PDB ID 5VGC) using PyMOL software (https://pymol.org/2/).

Crystallization and structure determination.

Crystallization of the GOPC PDZ domain (276 to 362 aa) in complex with the NleG8_Ec peptide (206 to 215 aa) was obtained using sitting-drop vapor diffusion by mixing 0.6 μL of an equimolar mixture (1.5 mM [each] GOPC[276-362] and NleG8_Ec peptide [Bio Basic Inc., USA]) with 0.6 μL reservoir solution. Protein solution prior to crystallization was in 0.3 M potassium chloride, 10 mM HEPES, pH 7.5. Reservoir solution was 1.5 M ammonium sulfate, 0.1 M Tris (pH 8.5), and 12% (wt/vol) glycerol. Crystals were cryoprotected using Paratone oil prior to the collection of diffraction data at 100 K on a Rigaku HF-007 (laboratory X-ray diffraction system) with an R-Axis IV detector. Images were processed using HKL-3000 (PMID 16855301), and the structure (PDB 6XNJ) was solved by molecular replacement using the structure of the GOPC PDZ domain in complex with the iCAL36 peptide (PDB 4E34, PMID 23243314) using Phenix.phaser (PMID 20124702). The difference (F_o-F_c) in density was unambiguous and manually modeled as the NleG8_Ec peptide. Refinement was performed with Phenix.refine, and all B-factors were refined as isotropic with Translation-Libration-Screw-rotation model (TLS) parameterization. All geometry was verified using Phenix.molprobity and the wwPDB server. Structure determination statistics are summarized in Table S8. Structure figures were generated using the PyMOL Molecular Graphics System (DeLano Scientific), and quantitative electrostatics were calculated using APBS (61).

In vitro ubiquitination.

Ubiquitination reactions were performed as described previously with minor changes (20). Briefly, the 20-μL reaction mixture consisted of buffer (50 mM Tris, pH 7.5, 100 mM NaCl, 10 mM ATP, 10 mM MgCl₂, 0.5 mM dithiothreitol [DTT]), 4 μg of ubiquitin (Boston Biochem), 0.13 μg of E1, 2 μg of E2, and 2 μg of His₆-tagged NleG proteins. Reaction mixtures were incubated at 30°C for the indicated period of time and stopped by the addition of an equal volume of 2× Laemmli sample buffer (0.125 M Tris-HCl, pH 6.8, 20% glycerol, 4% SDS, 0.004% bromophenol blue, 100 mM DTT). Reaction mixtures were separated by SDS-PAGE and visualized by Western blot analysis and Coomassie blue staining.

AP-MS.

Affinity purification-mass spectrometry (AP-MS) experiments were done essentially as described previously (21). Human cell line U937 cells (1 × 10⁸) and murine CT26.WT cells (1 × 10⁸) were lysed using the freeze-thaw method in 500 μL AP buffer (50 mM HEPES, pH 8.0, 150 mM NaCl, 0.2% [vol/vol] NP-40, 1 mM DTT, and 0.5 mM EDTA) with a protease inhibitor cocktail (cOmplete, Mini [Roche]). The soluble fraction of U937 cell lysate was separated by centrifugation and depleted of biotinylated proteins using streptavidin magnetic beads (NEB). C. rodentium strain DBS100 His-SBP-tagged NleG8_Cr effector (200 μg) was bound to 100 μL streptavidin-coupled magnetic beads (ThermoFisher) in AP buffer. NleG8_Cr-bound beads were incubated with U937 cell lysate for 3 h and washed twice with 500 μL AP buffer following a wash with 500 μL 50 mM ammonium bicarbonate, pH 8.5 (AMBIC). NleG8_Cr-interacting proteins were eluted using 200 μL of 50 mM AMBIC supplemented with 2.5 mM biotin and digested by trypsin. Tryptic peptides were purified using C₁₈ OMIX tips (Agilent) and analyzed by LC-MS/MS (Southern Alberta Mass Spectrometry [SAMS] Centre, Canada). The mass spectra were searched against the combined human, C. rodentium, and E. coli proteomes using Mascot software (www.matrixscience.com) with a filter set for at least 95% confidence for the identified proteins. Scaffold software (www.proteomesoftware.com) was used to identify NleG8_Cr-specific interacting proteins by filtering for the number of unique peptides identified in comparison to U937 pulldown experiments using VipF and NleG1_Cr and NleG7_Cr from Legionella pneumophila and C. rodentium, respectively, and in comparison to CT26.WT pulldown experiments using NleG8_Cr[1-209] and empty beads.

For in vitro pulldown analysis, purified His-GOPC at 3 μg was depleted of biotinylated proteins using streptavidin magnetic beads in 300 μL AP buffer. His-SBP-NleG8_Cr and His-SBP-NleG8_Cr[1-209] were bound to 15 μL streptavidin-coupled magnetic beads in AP buffer. Empty streptavidin magnetic beads and those bound to His-SBP-NleG8_Cr or His-SBP-NleG8_Cr[1-209] were incubated with 100 μL of AP buffer or His-GOPC for 1 h, washed four times with 500 μL AP buffer, resuspended in 60 μL SDS-PAGE sample buffer, boiled for 5 min, separated on an SDS gel (20 μL/lane), and visualized by Western blot analysis.

Host cell real-time translocation assay.

The nleG8 fragments were PCR amplified with flanking NdeI and KpnI restriction sites and cloned into pCX340 upstream of the beta-lactamase gene TEM-1 (62). The various nleG8-TEM-1 fragments were then PCR amplified with flanking SnaBI and SacI restriction sites and subcloned into pGEN-Pem7-CmR expression vector. The expression constructs were transformed into EPEC E2348/69 wild-type (WT) and ΔescN mutant strains.

The effector-TEM-1 translocation levels were analyzed as previously reported with some modifications (46). Briefly, NleG8-TEM-1 fusion strains were grown in LB supplemented with 34 μg/mL chloramphenicol overnight at 37°C with shaking and then subcultured 1:100 into preequilibrated DMEM (Sigma) and incubated for 3.5 h standing at 37°C with 5% CO₂. Bacterial strains were washed and normalized to an OD₆₀₀ of 0.25 in FluoroBrite DMEM (Gibco). For each bacterial strain tested, HeLa cells at 5 × 10⁴ were seeded in four wells in black, clear-bottom 96-well plates and grown for 24 h in DMEM supplemented with 10% fetal bovine serum (FBS) (Sigma). Then, cells in two wells were loaded with CCF2/AM dye (Invitrogen loading kit), which can undergo TEM-1-mediated cleavage that produces an increased ratio of blue to green fluorescence due to loss of fluorescence resonance energy transfer (FRET) within the CCF2 dye. An additional two wells remained untreated to control for background fluorescence. The four wells were washed with PBS and infected with the bacterial strain at a multiplicity of infection (MOI) of 1,000:1. Immediately after infection, the plate was sealed and placed into a Cytation 5 plate reader set to 37°C. HeLa cell fluorescence was measured every 5 min for 100 min postinfection using an excitation filter λ of 409 nm and an emission filter λ of 450 nm (blue) and λ of 520 nm (green). For blue and green fluorescence separately, the fluorescence of cells without CCF2/AM dye was subtracted from the fluorescence of cells loaded with CCF2/AM to control for cell autofluorescence background. The ratio between blue and green fluorescence derived from CCF2 was used to determine NleG8 variant-TEM-1 fusion translocation into HeLa cells.

To validate the translocation assay setup, we used an EPEC strain with a chromosomal Tir-TEM-1 fusion as well as WT and ΔescN EPEC strains expressing the empty vector pGEN-Pem7-CmR. In line with previously reported translocation dynamics of Tir (46), EPEC expressing Tir-TEM-1 fusion induced an increase of the blue/green fluorescence ratio in infected HeLa cells starting at about 30 min after infection, and this effect increased steadily until completion of the assay (Fig. S8A), whereas EPEC and ΔescN strains expressing the empty vector control showed no significant increase in the blue/green fluorescence ratio throughout the assay (Fig. S8A).

Western blotting.

Proteins were separated by electrophoresis in SDS-10% polyacrylamide gels and blotted onto a polyvinylidene difluoride (PVDF) membrane. HA fusion proteins were detected using anti-HA and anti-rabbit horseradish peroxidase (HRP)-conjugated antibodies (Millipore Sigma, Canada). GAL4 fragment fusion proteins were detected using anti-GAL4[1-147], anti-GAL4[768-881], and anti-mouse HRP-conjugated antibodies (Abcam, Canada). GOPC and FLAG-tagged proteins were detected using anti-PIST and anti-FLAG (Abcam, Canada) and anti-rabbit HRP-conjugated antibodies, respectively. His-tagged proteins were detected using anti-His HRP-conjugated antibodies (ThermoFisher, Canada). TEM-1 fusion proteins were detected using anti-TEM-1 (Abcam) and anti-mouse HRP-conjugated antibodies (Bio-Rad). HRP activity was visualized with Immobilon Western chemiluminescent HRP substrate (Millipore Sigma, Canada) and PicoPLUS enhanced chemiluminescence (ECL) detection reagent (ThermoFisher).

Animal experiments.

All animal experiments were performed according to the Canadian Council on Animal Care guidelines using protocols approved by the Animal Review Ethics Board at McMaster University under animal use protocol number 17-03-10.

For in vivo survival experiments, 6- to 8-week-old female C3H/HeJ (000659; Jackson Laboratory) mice were infected via oral gavage with 2 × 10⁸ CFU/mouse of C. rodentium DBS100 WT, ΔnleG1_Cr, ΔnleG7_Cr, ΔnleG8_Cr, ΔnleG8:nleG8_Cr, ΔnleG8:nleG8_Ec, ΔnleG8:nleG8_Cr(I138K), or ΔnleG8:nleG8_Cr[1-209] suspended in PBS. Survival rate was calculated as percentage of mice alive at 14 dpi. Longevity was calculated as average dpi on which mice were sacrificed. Fecal burden was monitored three times weekly and then daily as mouse health approached the endpoint. Fecal pellets were homogenized in PBS with a mixer mill (5 min, 30 Hz) (Retsch), serially diluted in PBS, and plated on MacConkey agar without antibiotic selection. Pink colonies were enumerated after 48 h.

Fecal water content from C. rodentium-infected mice was determined as described previously (63). Briefly, fecal pellets were collected into a preweighed 2-mL microtube on the day of sacrifice. The tube was reweighed (wet weight) and dried in a 37°C incubator for 48 h and reweighed (dry weight). The differences in wet versus dry weights were a measure of the amount of fecal water content.

Statistical analysis.

Mouse treatment groups were compared using either a nonparametric one-way analysis of variance (ANOVA), Kruskal-Wallis test, Dunn’s multiple comparison, Mann-Whitney test, or Mantel-Cox test. All analyses were performed using Graph Prism 8 (GraphPad Software Inc., San Diego, CA). The number of animals used in each experiment is reported in the figure legends. Luciferase assay statistical analysis was done in Excel as a two-tailed t test for each promoter at each time point compared to negative-control measurement. A P value of 0.05 or less was considered significant. Translocation assay statistical analysis was done in Graph Prism 5 using two-way ANOVA and Bonferroni posttest.

Proteomics.

Total proteome analysis was done essentially as previously described (64). Briefly, HEK 293T/17 SF cells (1.2 × 10⁶) were transiently transfected with constructs for expression of FLAG-tagged NleG8_Ec, NleG8_Ec(I138K), and NleG8_Ec[1-211]. Transfected cells were collected 30 h posttransfection with PBS, 0.25% trypsin, and 1 mM EDTA and lysed in 500 μL radioimmunoprecipitation assay (RIPA) buffer (25 mM Tris-HCl, pH 8, 150 mM NaCl, 1%, NP-40, 1% SDS, 1 mM EGTA, cOmplete protease inhibitor cocktail) by passing the cells five times through a 27-G needle using a 1-mL syringe. The cell lysate was centrifuged at 4°C for 30 min at 20,000 × g, and 300 μL of clear supernatant was transferred into a new 1.5-mL tube. Proteins were precipitated by addition of 12.5% (vol/vol) trichloroacetic acid (TCA) on ice for 30 min, collected by centrifugation at 4°C for 30 min at 20,000 × g, and washed twice with 1 mL acetone. Residual acetone was air dried at room temperature, pellets were trypsin digested, and tryptic peptides were purified using C₁₈ OMIX tips (Agilent) and analyzed by LC-MS/MS (BioZone Mass Spectrometry Facility, University of Toronto). The mass spectra were searched against the human proteome using Mascot software (www.matrixscience.com) to identify on average 3,133 proteins (SD = 549). Each protein construct was analyzed in triplicate, and only proteins identified in all three biological repeats were considered for further analysis. To identify the NleG8 upregulated and downregulated proteins, we compared the proteome of cells expressing wild-type NleG8 with that of cells expressing the I138K and PDZ binding domain deletion mutant.

Immunoprecipitation.

Coimmunoprecipitation (co-IP) was performed by immunoprecipitation as described previously (21). Briefly, HEK 293T/17 cells (2 × 10⁶) were seeded in 100-mm tissue culture (TC)-treated culture dishes in antibiotic-free complete medium. Transfections were performed using Lipofectamine 3000 reagent according to the manufacturer’s protocol. Cells were harvested 36 h after transfection and passively lysed for 30 min on ice in TNN buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.5% [vol/vol] Nonidet P-40, 1 mM EDTA, 1 mM DTT with Roche cOmplete Mini protease inhibitor). The cell lysate was clarified by centrifugation at 18,000 × g for 20 min at 4°C. Clear supernatant was incubated with mouse anti-FLAG M2 magnetic agarose beads (Sigma-Aldrich; M8823) for 2 h. The beads were magnetically separated, resuspended in 50 μL 2× Laemmli sample buffer, and boiled for 5 min. Immunoprecipitated proteins were separated on SDS-PAGE gels and visualized by Western blot analysis.