The bacterial virulence factor NleA undergoes host‐mediated O‐linked glycosylation

Enterohaemorrhagic and enteropathogenic Escherichia coli (EHEC and EPEC) are gastrointestinal pathogens responsible for severe diarrheal illness. EHEC and EPEC form “attaching and effacing” lesions during colonization and, upon adherence, inject proteins directly into host intestinal cells via the type III secretion system (T3SS). Injected bacterial proteins have a variety of functions but generally alter host cell biology to favor survival and/or replication of the pathogen. Non‐LEE‐encoded effector A (NleA) is a T3SS‐injected effector of EHEC, EPEC, and the related mouse pathogen Citrobacter rodentium. Studies in mouse models indicate that NleA has an important role in bacterial virulence. However, the mechanism by which NleA contributes to disease remains unknown. We have determined that the following translocation into host cells, a serine and threonine‐rich region of NleA is modified by host‐mediated mucin‐type O‐linked glycosylation. Surprisingly, this region was not present in several clinical EHEC isolates. When expressed in C. rodentium, a non‐modifiable variant of NleA was indistinguishable from wildtype NleA in an acute mortality model but conferred a modest increase in persistence over the course of infection in mixed infections in C57BL/6J mice. This is the first known example of a bacterial effector being modified by host‐mediated O‐linked glycosylation. Our data also suggests that this modification may confer a selective disadvantage to the bacteria during in vivo infection.


| INTRODUC TI ON
Escherichia coli is a ubiquitous member of the human intestinal microbiota that generally coexists with mutual benefits in digestion and immunity (Kaper et al., 2004). However, enterohaemorrhagic E. coli (EHEC) and enteropathogenic E. coli (EPEC) are pathogenic strains found in contaminated food and water that, when ingested, can infect the host gastrointestinal tract to cause potentially lethal diarrheal illness (Nataro & Kaper, 1998). EHEC, EPEC, and the related murine pathogen Citrobacter rodentium belong to a group of Gramnegative bacteria characterized by their ability to form "attaching and effacing" (A/E) lesions on the surface of host cells during intestinal colonization (Frankel et al., 1998;Knutton et al., 1987;Nataro & Kaper, 1998). A/E lesion development occurs through localized destruction of the normal epithelial microvilli architecture and subsequent cytoskeletal rearrangements for the formation of an actinrich pedestal structure, beneath the attached bacteria (Frankel et al., 1998;Kaper et al., 2004). The genes required for A/E lesion formation are clustered in the Locus for Enterocyte Effacement (LEE) which contains genes encoding a type III secretion system (T3SS) and multiple translocated effector proteins (Knutton et al., 1987).
The T3SS is found exclusively in Gram-negative bacteria, where it functions as a "molecular syringe" to deliver effector proteins across bacterial and host membranes directly into the cytosol of host cells (Deng et al., 2001Elliott et al., 1998;Perna et al., 1998).
However, while the genes encoding type III system components are conserved between bacterial species, the repertoire of translocated effectors tends to be unique to each pathogen (Hueck, 1998). EHEC, EPEC, and C. rodentium utilize an arsenal of virulence factors that have a variety of functions, but typically alter host cell biology to facilitate pathogen colonization of host tissue, suppression of the host immune system, or otherwise favor pathogen survival and/or replication inside the host. In addition, bacterial effector proteins may undergo several types of host-mediated post-translational modifications (PTM) following translocation that may trigger effector enzymatic activity, or aid in proper subcellular location targeting and functional regulation.
Several translocated bacterial effectors modified by host-mediated covalent modifications have already been described. After translocation, EPEC Tir, a LEE-encoded effector secreted by the T3SS essential for colonization, is tyrosine phosphorylated by host cell kinases upon its integration into the intestinal epithelial plasma membrane (Deng et al., 2003Kenny, 1999;Phillips et al., 2004). Phosphorylation of EPEC Tir is necessary for efficient actin pedestal formation and thereby intimate bacterial attachment by EPEC, indicating functional regulation by this host-mediated modification (Kenny, 1999;Kenny et al., 1997). The Salmonella type III effector, SopB, is a membraneassociated inositol polyphosphate phosphatase that is ubiquitinated by host machinery following translocation (Marcus et al., 2002). Although post-translational modification of proteins by ubiquitin has primarily been associated with protein degradation, ubiquitination of SopB regulates its enzymatic activity at the plasma membrane and intracellular localization rather than its intracellular stability (Mukhopadhyay & Riezman, 2007;Knodler et al., 2009).
One of the first-characterized T3SS-translocated non-LEE encoded effectors of EHEC, EPEC, and related pathogens are non-LEEencoded effector A (NleA; also known as EspI). The gene encoding NleA is absent from non-pathogenic E. coli and is preferentially found in those strains associated with outbreaks of human disease (Coombes et al., 2008;Mundy, Jenkins, et al., 2004a). In addition, it has been shown that NleA is present in C. rodentium, the murine model of EHEC and EPEC infections, and is required for virulence during infection of mice Mundy, Petrovska, et al., 2004b;Wickham et al., 2007). Genetically hyper-susceptible strains of mice infected with wild-type C. rodentium suffer high rates of mortality following infection. In comparison, mice infected with mutant C. rodentium in which the nleA gene is disrupted display mild symptoms but ultimately survive. In C57BL/6J mice, the nleA gene is also required for high-level bacterial colonization and associated histopathology . Although this indicates a clear effect of NleA on C. rodentium virulence, the functional role of NleA during infection is still unknown. NleA associates tightly with host cell membranes, although bioinformatic prediction of transmembrane domains yields inconclusive results .
Following translocation into host cells, NleA predominantly localizes to the host secretory pathway, colocalizing with markers of the Golgi apparatus (Creuzburg et al., 2005;Gruenheid et al., 2004;Lee et al., 2008) and has been shown to interact with COPII Kim et al., 2007). Among the dozens of effectors injected by EHEC, EPEC, and C. rodentium, NleA was noted as one of four effectors essential to bacterial pathogenicity (Ruano-Gallego et al., 2021).
In this study, we show that NleA undergoes an apparent size shift following translocation from the bacterium into the host cell, provide several lines of evidence consistent with the modification of NleA by O-linked glycosylation, and investigate the implications of the modification using analysis of human outbreak strains and in vivo infection models.

| NleA is modified by host-mediated Oglycosylation
We first noted a significant mobility change of NleA between bacterially-secreted proteins and those found within infected cells, resulting in a larger apparent size of host-translocated NleA on SDS-PAGE gels (Figure 1a). The bacterially secreted form of the protein is present at approximately 55 kDa, close to its predicted molecular weight of 48 kDa , while in the host cell lysate a band is observed at a higher apparent molecular weight (~60-65 kDa), suggesting a PTM of NleA inside the host cell. We assessed the phosphorylation of NleA by infecting HeLa cells with wild-type EPEC and incubating the lysates with λ phosphatase enzyme, which cleaves phosphates from serine, threonine, and tyrosine residues. We observed no change in apparent size of NleA after treatment with λ phosphatase (Figure 1b). In contrast, Tir, which has been previously reported to be phosphorylated upon host translocation (Kenny, 1999), was used as a positive control and showed a decrease in apparent size after enzyme treatment (Figure 1b). To assess the possibility of modification by ubiquitination, NleA was immunoprecipitated from Caco2/TC7 cell lysates infected with wild-type EPEC. There was an enrichment of NleA in the immunoprecipitate lysate, however, we detected no ubiquitin ( Figure 1c). Therefore, neither phosphorylation nor ubiquitination could account for the host-mediated modification of the NleA protein.
To further investigate the type of modification underlying the apparent size shift of the protein, a modified cell line with a reversible defect in glycosylation was used. Glycosylation, in which sugar moieties are attached to specific amino acids of a protein structure, plays a significant role in protein folding, targeted transport, cellular localization, and activity (Struwe & Robinson, 2019). Like phosphorylation and ubiquitination, glycosylation is a dynamic and reversible process. Under standard cell culture conditions, glucose is the only sugar directly provided in growth media. Wild-type Chinese Hamster Ovary (CHO) K1 cells maintain the ability to convert glucose to the required nucleotide sugars for normal glycosylation to occur. Mutant CHO LDLD cells, however, are defective in the synthesis of N-linked, O-linked, and lipid-linked glycoconjugates due to the absence of the UDP-Gal/UDP-GalNAc 4-epimerase, resulting in the inability to catalyze the reversible isomerizations of UDP-Glucose and UDP-N-acetylglucosamine to UDP-Galactose and UDP-N-acetylgalactosamine, respectively (Krieger et al., 1989). The mutant LDLD cells still possess sugar salvage pathways, however, and so these defects can be reversed by supplementing galactose and N-acetylgalactosamine (GalNAc) directly to the culture media (Krieger et al., 1989).
To determine whether glycoconjugate addition could explain the size shift of translocated NleA, CHO K1, and LDLD cells were infected with EPEC UMD207 with or without the addition of both galactose and GalNAc to the culture media. The UMD207 strain of EPEC was used as it is deficient in host cell adherence, but still functional for T3SS-mediated protein translocation (Donnenberg & Kaper, 1991), allowing us to wash away bacteria prior to preparing cell lysates. This strain, therefore, enables the analysis of only translocated bacterial protein, without or with minimal interference from NleA in remaining attached bacteria. Host-translocated NleA was visualized by Western blotting of the cell lysates. In the infected CHO K1 cells, two reactive bands were present: one at 55 kDa, corresponding to the expected size of unmodified NleA, and one at approximately 65 kDa, similar to the previously-observed size of the host-modified protein ( Figure 2a). Both these bands were present whether or not the media was supplemented with sugars. In the infected CHO LDLD cells, however, the appearance of NleA differed depending on whether the sugars were added to the growth media or not. Only the 55 kDa band was present without sugar supplementation. Upon supplementation with galactose and GalNAc, both the 55 and 65 kDa bands were observable. Since glycoconjugate addition is independent of sugar supplementation in the CHO K1 cells but dependent on the addition of galactose and GalNAc in the CHO LDLD cells, these results are consistent with the modification of NleA by glycoconjugate addition.
To further explore the potential modification of NleA by glycoconjugates, membrane preparations from cells infected with EPEC UMD207 were treated with a mix of deglycosylation enzymes that remove all N-linked, as well as simple O-linked glycans. This Given that we saw a restoration in the size shift of NleA when only GalNAc was added to the culture media, we posit that this implicates mucin-type O-linked glycosylation as the host-cell mediated modification of the bacterial protein upon translocation into the host cell.
Furthermore, the deglycosylation enzyme mix that did not remove the modification on the NleA protein ( Figure 2b) is known to be ineffective on mucin-like substrates. It should also be noted that when CHO LDLD cells are cultured with both galactose and GalNAc, the NleA apparent molecular weight is slightly larger than when only GalNAc is available, consistent with a restriction on the size of glycan structure with only GalNAc as a donor. to be modified by O-linked glycosylation, concentrated in a particularly serine-and threonine-rich portion of the EPEC NleA protein sequence. However, little is known about interactions between initial and subsequent sites of O-glycosylation, and the capability of predictive technology to precisely annotate substrate specificity of single GalNAc transferases remains limited. As a result, although NetOGlyc predicts specific sites of modification, the safe interpretation of a positive prediction is that the local region of the protein in which that residue is present is more likely to be O-glycosylated (Steentoft et al., 2013). Therefore, it is strongly suggested that NleA possesses at least one site of O-linked glycosylation in this serineand threonine-rich region.
NleA is one of many bacterial effectors injected by A/E patho-

| Disruption of the host secretory pathway affects NleA modification
The canonical ER-Golgi pathway of glycan biosynthesis in eukaryotic cells involves co-translational protein translocation into the ER where proteins are folded and modified, followed by subsequent passage through the Golgi apparatus for further modification and traffic to their various destinations. However, an additional glycosylation pathway also occurs in the cytoplasm, primarily via O-GlcNAc transferase (OGT)-catalyzed addition of a single N-acetylglucosamine to Ser or Thr residues. Although the extent of the apparent molecular weight shift of host-modified NleA, as well as the requirement for GalNAc for modification to occur in LDLD cells argue against the implication of the OGT pathway in NleA glycosylation, we sought to formally rule out this possibility. We used CRISPR-Cas9 to generate an OGT knockout HEK293 cell line and infected these cells with EPEC UMD207. We observed that NleA translocated into either WT HEK293-or HEK293 OGT KO-infected cells displayed an increased apparent molecular weight compared to the bacterial lysate ( Figure 3b). Thus, OGT is not implicated in the mobility shift of NleA following its translocation into host cells.
Next, we investigated the implication of the host cell secretory pathway in the modification of NleA with brefeldin A and monensin treatment in infected CHO LDLD cells with and without sugar supplementation. Brefeldin A treatment induces rapid disassembly of the Golgi apparatus (Doms et al., 1989;Fujiwara et al., 1988;Lippincott-Schwartz et al., 1989) and blocks protein transport into cell compartments post-Golgi (Misumi et al., 1966;Oda et al., 1967). The retrograde transport of components back to the ER results in nonselective separation of Golgi enzymes from those located in the trans-Golgi network. Therefore, some glycans may remain truncated due to an uncoupling of core structures from later glycosylation reactions. Comparatively, treatment with monensin, a F I G U R E 3 NleA is modified by the host secretory pathway. (a) Western blot analysis of CHO LDLD cells transfected with NleA-EGFP. Cells were cultured in media without (−) or with galactose and GalNAc sugar supplementation (+). Cells were incubated for 24 h with mock or transfection conditions. Blots were probed with anti-NleA and anti-Actinin antibodies. (b) Western blot analysis of wild-type HEK293 (lane 1) or HEK293 OGT knockout (lane 2) cell lysates infected with EPEC UMD207, uninfected cells (lanes 4, 5), and bacterial lysate collected from culture supernatant (lane 3). Blots were probed with anti-NleA, anti-OGT, anti-Tir, and anti-Actinin antibodies. (c) Western blot analysis of CHO LDLD cells treated with brefeldin A and monensin, infected with EPEC UMD207. Cells were cultured in media without (−) or with galactose and GalNAc sugar supplementation (+). Migration of molecular weight markers (kDa) is indicated on the left of each blot. sodium-hydrogen ionophore, interrupts intra-Golgi protein trafficking and inhibits glycan processing by neutralizing the Golgi luminal pH, impairing the function of many glycosylation enzymes (Kubo & Pigeon, 1983;Mollenhauer et al., 1990;Tartakoff, 1983).
As expected, in the absence of sugars only a band at 55 kDa was present regardless of brefeldin A and monensin treatment ( Figure 3c). In lysate from cells cultured in the presence of sugar supplementation untreated with the protein transport inhibitors, two NleA bands were observed corresponding to unmodified and modified protein (Figure 3c). Following treatment with brefeldin A alone, there was a striking reduction in the modified form of the protein.
Upon monensin treatment, an intermediate NleA protein size was apparent in addition to the unmodified form. The molecular weight decreased further when brefeldin A and monensin were added in combination, although not to the same extent as when no sugars were supplemented. This intermediary modified form mimics what was observed when GalNAc alone was supplemented to the culture media, indicating that some degree of modification may be occurring even in the presence of protein transport inhibitors, albeit truncated.
Tir, which was used as a control due to modification independent of the secretory pathway, showed the expected pattern of modification ( Figure 3c). Together, these results implicate the host cell secretory pathway as the site of NleA modification.

| The suspected region of NleA modification is variable
Based on the concentration of predicted O-glycosylation sites, we investigated whether the serine-and threonine-rich region of the NleA sequence was conserved between isolates. Protein sequences of NleA from C. rodentium, EPEC, and EHEC acquired from NCBI, and from EHEC human outbreak strains collected by the Institut National de Santé Publique du Québec (INSPQ) were aligned using the Clustal Omega Multiple Sequence Alignment tool (Sievers et al., 2011).
Sequences showed considerable variability in the region encompassing the serine-and threonine-rich portion of the protein predicted to be modified by O-glycosylation (Figure 4a). Surprisingly, several of the strains from infected patients exhibited a partial or complete deletion of this region.
To gain the understanding of where this region exists within the structure of NleA, we used RoseTTAFold (Baek et al., 2021) for protein structure prediction. The resulting models were somewhat lowconfidence (0.45 confidence value), because there are few sequences of NleA homologs and no NleA structure has been determined experimentally. However, the models clearly indicate NleA possesses a central beta barrel, surrounded by one coiled-coil and several other helices and loops (Figure 4b). The serine-and threonine-rich region is an unstructured loop segment between two strands of the beta barrel. That the region is a protruding loop is consistent with it being exposed for modification and suggests that its deletion would not greatly affect the overall structure of NleA.
Using these natural variants as well as the structure and glycosylation site predictions, we designed and constructed a variant of C. rodentium NleA protein, in which this serine-and threonine-rich region is deleted. C. rodentium NleA was chosen for these studies in order to align with our ultimate goal of assessing any effects on virulence in vivo.
To test the prediction that the serine and threonine-rich region of NleA was the site of host-mediated modification, we first assessed whether the variant lacking this region appeared to be modified in host cells, using EPEC UMD207 as a delivery system as F I G U R E 4 The NleA region predicted to be modified is variable. (a) Multiple protein sequence alignment of NleA from homologs identified via BLAST search and EHEC clinical strains courtesy of INSPQ. Sequences were aligned using Clustal Omega. Absent residues are represented by a dash (−). (b) Structural model of C. rodentium NleA with residues 169-183 colored in red. (c) Western blot analysis of CHO K1 and LDLD cell lysates infected with EPEC UMD207 ∆nleA + wild-type C. rodentium nleA or EPEC UMD207 ∆nleA + C. rodentium nleA∆169-183. Blots were probed with anti-NleA, anti-Tir, and anti-Actinin antibodies. Migration of molecular weight markers (kDa) is indicated on the left. described above. CHO K1 and LDLD cells were infected with EPEC UMD207 ΔnleA + wild-type C. rodentium nleA or EPEC UMD207 ΔnleA + C. rodentium nleAΔ169-183 without sugar supplementation to the culture media. Consistent with the previous results, translocation of wild-type C. rodentium NleA into CHO K1 cells, but not CHO LDLD cells, led to an increase in its apparent molecular weight, with the appearance of an upper band present at approximately 60-65 kDa. In contrast, NleΔ169-183 did not appear to increase in apparent molecular weight upon translocation into CHO K1 or LDLD cells, consistent with this region being critical for host-mediated modification (Figure 4c). Thus, modification within host cells is conserved in the NleA protein of C. rodentium, and this modification is abrogated when the serine-and threonine-rich region is deleted.

| Modification of NleA is not required for virulence in susceptible mice
NleA has previously been shown to be absolutely required for pathogen virulence in the C. rodentium-infection model in both susceptible and resistant mice . Mice infected with wildtype C. rodentium develop diarrhea, colitis, and hyperplasia of the intestinal epithelium, all shared symptoms of EHEC and EPEC infections in humans. In contrast, mice infected with C. rodentium lacking NleA develop only mild symptoms and do not succumb to infection . To analyze the role of NleA's host-mediated modification in bacterial virulence, we infected susceptible C3H/ HeJ mice with wild-type C. rodentium (WT), C. rodentium ΔnleA, C.
However, the distinct role of glycans, especially O-linked structures, obtaining correct protein folding is ambiguous. While there is evidence for protein misfolding and aggregation in the absence of glycans, others have shown that the elimination of some or all glycans had no effect on folding (Mitra et al., 2006;Parodi, 2000). This implies that some glycosylation sites may be more important than others for folding and that glycan-induced conformation effects are more than likely local in nature. The finding that the non-modifiable protein confers full virulence in susceptible mice provides evidence that this protein is likely to be properly folded and translocated during in vivo infection.

| Mutant NleA-complemented strains outcompete wild-type-complemented
To test for a more subtle effect on virulence, we employed the competitive index assay in C57BL/6J mice, which undergo a selflimiting infection with wild-type C. rodentium. Mice were infected with a 1:1 ratio of C. rodentium ΔnleA + WT nleA: C. rodentium ΔnleA + nleA∆169-183 and the ratio of the two strains within the inoculum and within each animal was assessed at early, peak, and late timepoints post-infection by colony PCR. An index of 1 indicates that wild-type and mutant strains are present in equal proportions, while an index less than 1 means that the wild-type population outcompeted the mutant population and conversely with an index greater than 1, the mutant outcompeted the wild-type strain. Surprisingly, the mutant-complemented strain, lacking the serine-and threoninerich region, persisted longer than the WT-complemented strain, resulting in a median index significantly greater than 1 at day 26 post-infection ( Figure 6a). Likewise, the percent of total colonies skewed toward mutant-complemented at peak, day 12, and late, day 26, infection timepoints, despite total bacterial burden not differing markedly from typically observed CFU (Figure 6b,c).
These results indicate that the absence of the host-mediated modification of NleA may be somewhat favorable for the bacteria, or suggests that the contraction of this region may be associated with some kind of modest gain-of-function, which is consistent with the clinical isolate data. An intriguing hypothesis is that NleA is modified by the host to serve as a signal to the immune system of an active infection, and that the deletion of this region, as observed in human outbreak strains, is an adaptation of the bacteria to prolong clearance, consistent with superior fitness of the mutant observed in our competition assay. Among their repertoire of functions, glycans can serve as recognition markers as well as modulate immune responses (Dwek, 1996;Lis & Sharon, 1993). The consequences of the absence of the glycosylation of NleA on immune interplay have not yet been explored.
Our results indicate that the bacterial virulence factor NleA is modified by host-mediated O-linked glycosylation upon translocation into the host cell during infection. To our knowledge, this is the first example of a bacterial effector protein undergoing this type of modification inside the host cell. Our findings are, however, reminiscent of another bacterial effector modified inside host cells. Although Tir has been studied extensively in vitro, with epithelial cell culture demonstrating that tyrosine phosphorylation of EPEC Tir is necessary for efficient A/E lesion formation, it has been challenging to attribute the significance of this modification to Tir functionality in mouse models and in vitro organ culture. In intestinal explants, Tir phosphorylationdeficient mutants were still able to colonize, and tyrosine phosphorylation modification of Tir is not required for colonization, A/E lesion formation, or crypt hyperplasia in the mouse gut, despite being necessary in vitro (Deng et al., 2003;Schüller et al., 2007).
Nonetheless, we are intrigued by the potential role of this modification given the importance of the glycosylation in intestinal biology.
One important remaining question for future investigation is how the serine-and threonine-rich region of NleA gains access to the enzymes involved in glycosylation, which are located within the lumen of the secretory pathway. Notably, when NleA was used as bait for affinity purification of effector-host protein interactions from HeLa cell lysates, GalNAc transferase 7 (GALNT7) was identified with intermediate confidence (Sontag et al., 2016). GALNT7 is part of a large subfamily of enzymes residing in the Golgi apparatus that control the initiation step of mucin-type O-linked protein glycosylation and transfer of GalNAc to Ser and Thr amino acid residues. However, this putative interaction was not validated by the authors. We have previously shown that in addition to localizing to the secretory path- F I G U R E 6 Bacteria expressing mutant NleA outcompete those with wild-type NleA. C57BL/6J mice (n = 22) were infected by oral gavage with a 1:1 mixture of C. rodentium ∆nleA+ nleA∆169-183: C. rodentium ∆nleA + wild-type nleA. Data were pooled from four independent experiments. (a) Competitive index of B6 mice at various timepoints post-infection. The competitive index is defined as the mutant-to-wild type ratio within the output sample divided by the mutant-to-wild type ratio within the input (inoculum). The index was assessed at days 4, 12, and 26 post-infection by diluting and plating on selective MacConkey agar supplemented with kanamycin. 100 colonies were selected for polymerase chain reaction and run on a 1% agarose gel and the ratio of mutant:WT was determined. The median at each time point is indicated by a solid horizontal line. P values were determined via one sample t-test at a 95% confidence interval. (b) Change in percent of total colonies identified as of C. rodentium ∆nleA + nleA∆169-183 (red) or C. rodentium ∆nleA + WT nleA (blue) at various timepoints postinfection. Data are represented as mean ± S.E.M. (c) CFU/g over time for each group of mice infected with 1:1 v:v of C. rodentium ∆nleA + WT nleA and C. rodentium ∆nleA + nleA∆169-183. The median of each time point is indicated by a horizontal line. The limit of detection (LOD) is indicated by a dashed red line. interaction of NleA precedes or follows its modification. Together these studies lend credence to NleA being in the vicinity of and potentially interacting with glycosyltransferase enzymes. Intriguingly, Sec24 is part of a coat complex on the outside of transport vesicles, whereas the glycosylation enzymes reside within the lumen of the secretory pathway. This suggests that NleA must either span, cross, or disrupt the secretory pathway membranes. NleA was previously shown to be tightly associated with host membranes, resistant to extraction with high salt and high pH
To make EPEC UMD207 ∆nleA, sacB gene-based allelic exchange (Donnenberg & Kaper, 1991) was used to generate an inframe deletion mutant of nleA using the vector pRE112 with primers EPECDHKF1 and EPECDHKR2 (Table S1). This mutant was verified by PCR reaction spanning the deletion site, and by Western blot analysis with anti-NleA antiserum.
The complemented EPEC UMD207 ∆nleA + WT nleA and Citrobacter rodentium ∆nleA + WT nleA strains were created by PCR amplifying the entire nleA sequence and flanking region using primers EPECDHKF1 and EPECDHKR2 (Table S1) and, after sequence verification, was subcloned into the low-copy number vector pWSK129.
To make EPEC UMD207 ∆nleA + nleA∆169-183 and Citrobacter rodentium ∆nleA + nleA∆169-183, the nleA gene with a partial internal deletion was commercially synthesized and inserted into the pBluescript II SK (+) vector and transformed into chemically competent Subcloning Efficiency DH5α cells (ThermoFisher). pWSK129 plasmid containing the WT nleA gene and pBluescript II SK (+) containing the mutated nleA gene were digested with BglII and SphI. The resulting pWSK129 vector was ligated to the pBluescript II SK (+) digestion product encompassing the 45 base pair internal deletion.
The partial deletion was verified by Western blotting whole cell lysates with polyclonal anti-NleA antiserum, and by sequencing with primers NleA_1_F and NleA_4_R.
The supernatant was transferred into a new tube and centrifuged 16,000 × g, 2 min. Supernatant proteins were precipitated by the addition of trichloroacetic acid (10% [vol/vol]) for 1 h on ice. Precipitated proteins were pelleted by centrifugation at 4°C (16,000 × g, 30 min).
The supernatant was removed, and 1 ml of ice-cold acetone was added to each tube. The samples were stored at −80°C overnight.

| Cell culture
HeLa (ATCC no. CCL2) and Caco2/TC7 (Chantret et al., 1994) cells were cultured in Dulbecco's minimal Eagle's medium (DMEM; Wisent) supplemented with 2 mM glutamine and 5% and 10% (v/v) heat-inactivated fetal calf serum (Wisent). Chinese Hamster Ovary (CHO) cells were cultured in Ham's F-12 medium (Wisent) with 5% inactivated fetal calf serum and 1% L-glutamine. Cells were grown at 37°C in a humidified atmosphere of 5% CO 2 . 48 h before infection of CHO cells, dishes were washed with phosphate-buffered saline (PBS) once and changed to Ham's F-12 media supplemented with 1% inactivated fetal calf serum with or without galactose (20 μM) and N-acetylgalactosamine (200 μM) sugar supplementation. For each sample, at least one confluent 100 mm dish of cells were infected with EPEC UMD207 ∆nleA + WT nleA or EPEC UMD207 ∆nleA + nleA∆169-183 using an initial multiplicity of infection of 1:100 and incubated for 4 h at 37°C in a 5% CO 2 atmosphere. EPEC strain UMD207 lacks intimin and the bundle-forming pilus and therefore delivers bacterial effector proteins without stable bacterial adherence (Donnenberg & Kaper, 1991). in the presence of brefeldin A and monensin, and whole cell lysates were collected for resolution by SDS-PAGE as outlined above.

| λ phosphatase treatment
The λ phosphatase treatment was performed as per the manufacturer's instructions (New England Biolabs). Briefly, HeLa cells were infected with EPEC UMD207. Cells were harvested and lysed in B150 buffer without phosphatase inhibitors (20 mM 150 mM KCl,5 mM MgCl2,10% glycerol,protease inhibitors). Lysates were incubated with or without enzyme for 2 h at 30°C. The samples were run on an SDS-PAGE gel, transferred onto a PVDF membrane, and immunoblotted.
The supernatant was transferred to a clean Eppendorf tube, containing 40 μl of protein G Plus (Santa Cruz; prewashed in lysis buffer) per sample, and incubated for 1 h on the labquake at 4°C. Following incubation, the samples were centrifuged at 4°C, 1500 × g for 5 min.
15 μl of 5X Laemmli was added to 75 μl of the supernatant, boiled for 5 min, and stored as the pre-IP lysate. NHS-activated sepharose beads (GE Healthcare Life Sciences) coupled to rat anti-NleA affinity purified antibody were prepared as per manufacturer's instructions and added to the remaining supernatant, followed by incubation for 2 h on a labquake at 4°C. Following incubation, the samples were centrifuged at 4°C, 1500 × g, and 75 μl of supernatant was stored as the post-IP lysate as described above. The beads were washed with lysis buffer, each wash consisting of 10 min on the labquake at 4°C followed by centrifugation at 4°C, 1500 × g for 5 min. Beads were resuspended in 100 μl of 5X Laemmli sample buffer to elute the protein and boiled for 5 min. All samples were then run on an SDS-PAGE gel, transferred to a PVDF membrane, and immunoblotted for total ubiquitin.

| Deglycosylation assay
HeLa cells were infected with 100 μl of overnight culture of wild-type EPEC for approximately 4-5 h. The cells were harvested in PBS+/+.
Cells were centrifuged at 1000 × g for 5 min at 4°C. The supernatant was aspirated, and the pellets were resuspended in 1.5 ml homogenization buffer (250 mM sucrose, 3 mM pH 7.4 imidazole, 0.5 mM EDTA, 1 mM VO 4 , 1 mM NaF). Samples were centrifuged at 3000 × g for 10 min at 4°C. The supernatant was aspirated and the samples were resuspended in 300 μl of homogenization buffer. The samples were mechanically lysed by passing through a 22-gauge needle and centrifuged at 3000 × g for 15 min at 4°C. The supernatant was transferred to an ultracentrifuge tube, and centrifuged at 41,000 × g in a TL100 Beckman centrifuge in TLS55 rotor for 20 min at 4°C.
The supernatant was aspirated, and the pellet (membrane fraction), was resuspended in 125 μl of dH 2 O. The deglycosylation assay was performed as per the manufacturer's instructions (New England Biolabs, P6039). Briefly, 2 μl of 10X Glycoprotein Denaturing Buffer was added to 18 μl of each sample and denatured at 100°C for 10 min. The samples were cooled on ice and centrifuged at maximum speed for 10 s. Following centrifugation, 5 μl of 10X GlycoBuffer 2 was added to each sample, followed by 5 μl of Deglycosylation Enzyme Cocktail or dH 2 O. The contents were mixed gently and incubated at 37°C for 4 h. Following incubation, samples were run on an SDS-PAGE gel and immunoblotted with an anti-NleA antibody.
Fetuin was processed as a positive control, run on an SDS-PAGE gel, and stained by Coomassie.

| Transfection
LDLD cells were plated 2.5 × 10 6 cells per 10 cm dish 1 day prior to transfection so that the cells were approximately 80% confluent on the day of transfection. Cells were transiently transfected with 2 μg of DNA using 25 μl FuGENE HD Transfection Reagent (Promega) according to the manufacturer's instructions. Transfected cells were briefly washed with sterile ice-cold PBS to remove residual media.
Cells were harvested 24 h post-transfection by scraping into 1 ml cold PBS, centrifuged at 1500 rpm for 5 min at 4°C and resuspended in 0.5 ml RIPA buffer. Cell suspensions were sonicated and centrifuged at maximum speed for 15 min at 4°C. The supernatant containing the whole cell lysate was made up to 1X Laemmli buffer using a 5X stock for resolution by SDS-PAGE.

| Generation of HEK293 OGT knockout cell line
Synthesis of gRNAs: The Alt-R CRISPR-Cas9 sgRNA for generation of OGT knockout was synthesized by IDT (gRNA target sequence: CATCGATGGTTATATTAACC).
Electroporation: 1 day prior to transfection, HEK293 cells were split into a new flask with fresh growth medium such that the cells reach 70%-90% confluent the following day. On the day of electroporation, single cells were prepared, counted, and appropriate amounts of cells (1 × 10 5 cells per transfection) were transferred to a 1.5 ml microcentrifuge tube. The cells were washed once with PBS by centrifugation at 500 × g for 5 min. At the same time as the preparation of cells for electroporation, 2 μg Cas9 protein and 400 ng sgRNA were mixed in 10 μl of resuspension buffer R and incubated at room temperature for 10 min, after which 12.3 pmol of Alt-R Cas9 electroporation enhancer (IDT) was added. Prepared cells were re-suspended in the buffer R containing Cas9-gRNA complex and electroporation enhancer, and cell mixture was then transferred into a 10 μl Neon tip with Neon pipette and electroporation were performed using the parameters as following: pulse voltage 1100 V, pulse width 20 ms and pulse number 2. After electroporation, cells were added into 1 ml prewarmed growth medium in a 24-well plate and cultured for 3 days.
T7EI assay: Genomic DNA was extracted from HEK293 cells transfected with Cas9-gRNA and was then PCR amplified with primers flanking the gRNA target region (Fwd: ACACTTGTCGCCTTTTCCAGA; Rev: GACCCATTATCCACCATTCCTTG). The amplification was carried out with AmpliTaq Gold 360 master mix (ThermoFisher), using the following cycling condition: 95°C for 10 min for initial denaturation; 40 cycles of 95°C for 30 s, 60°C for 30 s and 72°C for 35 s; and a final extension at 72°C for 7 min. The PCR product was used for the assay of the success of genome editing by T7EI assay using the ALT-R genome editing detection kit (IDT).
Single-cell clone analysis: single cells were prepared, counted, and serially diluted to 2 × 10 4 , 5 × 10 2 , and 5 cells/ml. Then, 200 μl of 5 cells/ml was dispensed to each well of a 96-well plate using a multichannel pipette. Plates had been incubated at 37°C in a 5% CO 2 incubator until single cells grew into colonies.
Screening KO: Cell lysates were prepared from single-cell clones and Western blot analyses were used to screen knockout clones.

| Immunoblot analysis
Samples for Western blot analysis were resolved by mini-Protean pre-cast gels (7.5% polyacrylamide; BioRad) at 140 V for 75 min and transferred electrophoretically at 10 V for 30 min onto an activated PVDF membrane. Immunoblots were blocked in 5% non-fat dried milk (NFDM) in TBS, pH 7.2, containing 0.1% Tween 20 (TBST) overnight at 4°C. Membranes were then incubated with primary antibody in 5% NFDM TBST overnight at 4°C before being washed five times in TBST and then incubated with horseradish peroxidase-conjugated secondary antibody for 45 min at room temperature. Membranes were washed as described above and visualized by developing in Immobilon Western ECL Substrate (Millipore), followed by exposure to autoradiography film (Denville). Primary and secondary antibodies used are listed in Table S1.

| Animal studies
This study was conducted in accordance with the recommendations of the Canadian Council on Animal Care. The protocol was approved by the McGill University Animal Care Committee. Five-week-old female C3H/HeJ and C57BL/6J mice (Jackson Laboratory) were housed in a specific-pathogen-free facility at McGill University. Wild-type or mutant C. rodentium DBS100 strains were grown overnight in 3 ml LB broth, 220 rpm, at 37°C, and 100 μl of the cultures was used to infect mice by oral gavage, containing 2-3 × 10 8 CFU. The infectious dose was confirmed by plating of serial dilutions. For survival experiments, the highly susceptible C3H/HeJ mice were monitored daily over the course of infection and were euthanized if any of the following clinical endpoints were met: 20% body weight loss, hunching, and shaking, inactivity, or body condition score <2 (Ullman-Culleré & Foltz, 1999). To determine bacterial colonization, fecal pellets were homogenized in PBS using a MagNA Lyser (Roche) and serially plated on MacConkey agar (Difco Laboratories). Plates with colonies between 30 and 300 were enumerated, and C. rodentium was identified by its distinctive morphology on the selective medium. For competitive index experiments, equal volumes of the inoculum of C. rodentium ∆nleA + WT nleA and C. rodentium ∆nleA + nleA∆169-183 were mixed, and 100 μl of the mixture was used to infect five-week-old female C57BL/6J mice by oral gavage. At the indicated times throughout the course of infection, fecal pellets were collected, homogenized in PBS as described, and serially diluted before being plated on MacConkey agar. After overnight incubation, individual colonies were patched onto LB plates supplemented with Kan and, the following day each clone was genotyped by PCR using primers NleA_1_F and NleA_3_R flanking the deleted region of the nleA∆169-183 mutant strain. 100 colonies were analyzed per mouse per timepoint, and three to five mice were analyzed per experiment. PCR products were visualized on 1% agarose gels containing ethidium bromide. The competitive index was determined as the output ratio (mutant:wild type) divided by the input ratio (mutant:wild type).

| Structural modeling
C. rodentium NleA was modeled by providing its sequence (NCBI accession code WP_012904700.1) to the web-based program RoseTTAfold (Baek et al., 2021). The five models each had a confidence score of 0.45, and had average per atom error estimates of 6.8-8.0 Å. The models were overall similar to one another and model 1 is shown in Figure 4b.

| Statistical analysis
Data analyses were performed using GraphPad Prism v9.0 software.
Statistical comparisons were carried out using tests described in the figure legends with a p < 0.05 considered significant.

CO N FLI C T O F I NTE R E S T
We have no known conflict of interest to disclose.

E TH I C S S TATEM ENT
All animal experimentation was carried out in accordance with the principles outlined in the most recent policies and Guide to the

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials.