Sequence Variability in NleG Effectors Suggests Divergent Targeting.

NleG effectors were originally divided into 12 distinct clades based on sequence diversity (10). The 14 NleG effectors identified in EHEC O157:H7 represent 8 of these clades, sharing between 23% and 99% of primary sequence identity in pairwise comparisons (excluding the truncated predicted pseudogenes nleG2-1 and nleG3) (4, 10). We hypothesized that sequence variability among NleG effectors reflects a diversification of host target-binding specificity and ubiquitination. We built a phylogenetic tree for the NleG family that includes several newly identified NleG effector sequences (SI Appendix, Fig. S1 A and B). The group represented by NleG2 from EHEC O157:H7 appears most prevalent, with at least one representative encoded in all EPEC or EHEC strains included in this analysis. In contrast, other NleG clades, such as the subfamily represented by NleG5-1 and NleG5-2 from EHEC O157:H7 are only present in a subset of EHEC strains.

This comparative analysis of NleG sequences revealed a conserved set of residues in the N-terminal region, including Arg28, Ile38, Gly42, Val45, Ile47, Phe56, Leu65, Ile70, Ile80, Leu84, Asn85, and Gly87 (numbering based on NleG5-1) (SI Appendix, Fig. S1A). We interpreted the prevalence of conserved hydrophobic residues within this N-terminal region as an indication of a structurally maintained hydrophobic core. Since the NleG N-terminal region lacked a known sequence motif or similarity with known functional domains, we proceeded with structural characterization of the NleG N-terminal region.

The NleG5-1 Structure Defines the Conserved Architecture of NleG Effectors.

To characterize the NleG N-terminal domain, we expressed and purified N-terminal fragments of NleG2-3, NleG5-1, and NleG8-1 and acquired (¹H-¹⁵N)-heteronuclear single quantum coherence (HSQC) spectra to evaluate whether they adopted a stable tertiary structure. These NleG effectors were chosen because they are secreted by EHEC (10) and none share greater than 30% pairwise amino acid identity to one another. The HSQC spectrum of the N-terminal region of NleG5-1 (residues 1–116) showed good peak dispersion and signal-to-noise, indicating this fragment adopts a stable structure suitable for determination by NMR spectroscopy (SI Appendix, Fig. S2). We therefore proceeded with the solution structure determination of the NleG5-1 N-terminal domain by NMR.

Standard 2D and 3D spectra were acquired with ¹⁵N/¹³C-labeled N-terminal NleG5-1 and the solution structure of NleG5-1 for residues 1–100 was determined (PDB ID code 6B3N). The structure of NleG5-1[1–100] revealed a two-layer α/β-sandwich fold, with an antiparallel β-sheet packed against the α1 and α2 helices (Fig. 1A). Consistent with our interpretation of conserved hydrophobic residues in the N-terminal domain, mapping these residues onto the structure of NleG5-1[1–100] confirmed that most are part of the hydrophobic core (Fig. 1B). Structural homology searches using the DALI (11) and PDBeFold (12) servers confirmed only general similarity between the N-terminal NleG5-1 domain and several layered α/β domains (SI Appendix, Tables S1 and S2). However, this analysis did not suggest possible functions that could be inferred from structural similarity.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

The NleG5-1 NMR and crystal structures reveal a conserved N-terminal fold. (A) Bundle of the 20 lowest-energy NMR models representing the solution structure of NleG5-1 for residues 1–100 (PDB ID code 6B3N). (B) The most conserved residues in the NleG5-1 N-terminal domain form the core of the fold. Depiction of the most conserved residues between NleGs made using the ConSurf server (40) and the multiple sequence alignment shown in SI Appendix, Fig. S1. (C) The crystal structure of full-length NleG5-1 (PDB ID code 5VGC) with the N-terminal domain colored in blue as in A and the U-box domain in purple. Due to slight secondary structure annotation differences the first β-strand in the NMR structure is annotated as β0.

Although the structure of NleG5-1[1–100] along with the NleG2-3 U-box domain, reported previously by Wu et al. (4), provide structural details for understanding their function individually, how these domains are linked and their function in context of full-length NleG remained elusive. To understand the architecture of NleGs, we determined the crystal structure of full-length NleG5-1 to 2.6 Å by molecular replacement (PDB ID code 5VGC). Each NleG5-1 molecule featured distinct N- and C-terminal domains connected by two consecutive α-helices, designated α3 and α4, which are tethered by an unstructured linker (Fig. 1C). The NMR-derived NleG5-1[1–100] N-terminal domain and the C-terminal U-box domain of NleG2-3 (4) superimposed with the NleG5-1 full-length crystal structure with an rmsd of 2.14 (over 89 Cα atoms) and 1.35 Å (over 91 Cα atoms), respectively (SI Appendix, Fig. S3A). To interpret this full-length NleG structure in the context of host target ubiquitination, we modeled NleG5-1 against the structure of an activated eukaryotic E3–E2∼Ub complex, RNF4–UbcH5a∼Ub (13). Superimposition of the NleG5-1 U-box domain to the RING domain of RNF4 revealed no major clashes between NleG5-1 and the E2-conjugating enzyme UbcH5a, or the conjugated ubiquitin (SI Appendix, Fig. S3B). Thus, our structural analysis of NleG5-1 has established the two-domain architecture shared by NleG effectors and is consistent with eukaryotic ubiquitination processes.

Different NleG Effectors Target Unique Host Proteins.

To test the functional diversification among NleG effectors, we searched for human protein targets of the EHEC O157:H7 effectors NleG5-1 and NleG2-3, which share only ∼20% sequence identity in their N-terminal domain. For this, affinity-purified NleGs immobilized on magnetic beads were used to capture potential interaction partners from human U937 cell lysate, which were then identified using mass spectrometry (Fig. 2 A and B). For NleG5-1, we identified 23 subunits of human Mediator complex and several additional Mediator complex-affiliated proteins as possible interaction targets (Fig. 2A) (14). In contrast, NleG2-3 coprecipitated with only three human proteins, Hexokinase-2 (HK2), Synaptosomal-associated protein 29 (SNAP29), and Hexokinase-1 (HK1), with HK2 accounting for the majority of the identified peptides (Fig. 2B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Human protein targets of NleG5-1 and NleG2-3 are MED15 and HK2, respectively. AP-MS experiments with U937 cell lysate reveals the human protein target candidates for NleG5-1 (A) and NleG2-3 (B). AP-MS results are displayed as the average number of peptides identified for the indicated human proteins in at least four MS runs (two MS runs for each replicate AP experiment). To identify the direct interactors from the AP-MS candidate pools, Y2H experiments were performed with human protein candidates and NleG2-3 (C) or NleG5-1 (D), confirming interactions between MED15 and NleG5-1 and between HK2 and NleG2-3.

The human Mediator complex is composed of ∼26 subunits and plays a critical role in transcriptional regulation for eukaryotes (15). Hexokinase-2 (HK2) is an enzyme involved in glycolysis and regulation of intrinsic apoptosis (16, 17). SNAP29 is a SNARE protein involved in vesicle fusion and transport, as well as phagocytosis (18). To determine which of these proteins were involved in direct interactions with NleG5-1 and NleG2-3 we used a yeast two-hybrid (Y2H) approach (19). To control for the E3 ligase activity of the NleGs, which could lead to ubiquitination and degradation of their interacting partner, the NleG variants NleG5-1^V144K and NleG2-3^I121K were tested alongside the wild-type NleGs as they have previously shown to be deficient in ubiquitination due to disruption of the hydrophobic E2-binding surface (4). Coexpression of wild-type NleG2-3 with HK2 or SNAP29 did not result in yeast growth on selective media (Fig. 2C). However, the coexpression of NleG2-3^I121K with HK2, but not with SNAP29, consistently showed yeast growth, implying that NleG2-3 directly interacts with HK2, consistent with HK2 representing the most highly enriched NleG2-3 target identified in affinity purification (AP)-MS experiments. This observation suggested that the absence of growth for strains coexpressing wild-type NleG2-3 with HK2 is a false negative result, due to degradation of HK2 via NleG-triggered ubiquitination.

The Y2H screens for interaction between NleG5-1 and individual Mediator complex subunits were complicated by autoactivation of yeast growth for DNA-binding (DB) domain fusions of NleG5-1^V144K (SI Appendix, Fig. S4A) and of the Mediator subunits MED1, MED4, MED15, MED26, and MED31 (Fig. 2D). Coexpression of the activating domain (AD) fusion of NleG5-1 with the rest of the tested Mediator subunits did not reveal any specific interactions. However, when plated on selective media containing 3-amino-1,2,4-triazole (3AT), which increases the stringency of the Y2H readout, coexpression of wild-type NleG5-1 with MED15 resulted in decreased yeast growth compared with NleG5-1^V144K. We interpreted this as a positive interaction between NleG5-1 and MED15 and the subsequent ubiquitination-mediated degradation of MED15, analogous to the results observed with coexpression of wild-type NleG2-3 and HK2 (Fig. 2C). This hypothesis was examined further by monitoring the growth of yeast coexpressing MED15 with wild-type NleG5-1 or NleG5-1^V144K in the presence of increasing concentrations of 3AT. Under these conditions, the MED15-triggered autoactivation of yeast growth was decreased only by the presence of wild-type NleG5-1, while growth with the NleG5-1^V144K mutant was amplified (SI Appendix, Fig. S4B). Considering that MED15 was the most represented human protein coprecipitated with NleG5-1 from human cell lysate (Fig. 2A), these results strongly suggest that the MED15 subunit of the Mediator complex is a direct target of NleG5-1.

To determine whether the N-terminal domain of NleGs mediate host target interactions, we tested the ability of the individual N- and C-terminal domains of NleG2-3 and NleG5-1 to bind their respective targets using Y2H. In this assay, yeast growth indicating direct binding was observed between the N-terminal domains of NleG2-3 with HK2 and NleG5-1 with MED15 (for details see SI Appendix, Supplementary Results and Fig. S5 A and B). Considering the NleG2-3 and NleG5-1 N-terminal domains mediate interactions with specific human protein targets, we then hypothesized that sequence variability in their N-terminal surface-exposed residues mediates the specificity of these interactions. A model of the N-terminal domain of NleG2-3 was constructed based on the crystal structure of NleG5-1 using MODELER (20) and used to select a panel of NleG2-3 N-terminal surface-exposed residues that are not conserved with NleG5-1, which were then probed for their role in HK2 interaction. These NleG2-3 residues were individually substituted with alanine and tested for their interaction with HK2 by Y2H in the NleG2-3^I121K ubiquitination-deficient variant. The substitutions D30A, T32A, G35A, T37A, V41A, Y42A, S44A, L63A, and L64A resulted in a detectable decrease of Y2H growth, particularly when tested on the medium containing 3AT (SI Appendix, Fig. S5 C and D). Of these, Y42A, L63A, and L64A mutations had the most severe effect on Y2H signal. Since these residues are conserved in a subset of NleG effectors, but are not found in NleG5-1, we postulated that these residues may be part of a common target interaction interface in several NleGs, while the rest of the identified NleG2-3 residues are responsible for specificity of NleG2-3 toward HK2. The large number of NleG2-3 N-terminal residues that appeared to be involved in interaction with HK2 is indicative of a broad NleG2-3–HK2 interface. Intriguingly, the mutation H10A in NleG2-3 resulted in increased Y2H growth on selective medium, suggesting that this substitution may strengthen the NleG2-3 interactions with HK2.

Our results identify distinct host targets for two EHEC NleG effectors, revealing the functional diversification in this family and identify a number of N-terminal residues specific to NleG2-3 that are critical for HK2 interaction, supporting the hypothesis that sequence plasticity of this domain is responsible for host target specificity of NleG effectors.

NleG Effectors Bind and Induce Degradation of Their Targets in Human Cells.

To determine whether NleG2-3 and NleG5-1 affected endogenous levels of HK2 and MED15, HEK293T cells were transiently transfected with constructs expressing wild-type and ubiquitination-deficient NleG5-1 or NleG2-3 and the levels of MED15 and HK2 were probed by immunoblotting. Cells expressing wild-type NleG5-1 or NleG2-3, but not their ubiquitination-inactive variants, showed significantly decreased levels of MED15 and HK2, respectively (Fig. 3A). In addition, the NleG5-1^V144K and NleG2-3^I121K mutants consistently coimmunoprecipitated with MED15 and HK2, respectively, further confirming specific interactions between NleG effectors and these human proteins (Fig. 3A). Transfection of HEK293T cells with a construct expressing the partially inactive NleG5-1^V144A variant also led to reduction of MED15 levels, although to a lesser degree than in the case of the wild-type NleG5-1. These studies confirmed that NleG5-1 and NleG2-3 specifically target the human proteins MED15 and HK2 and induce their degradation in human cells.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

NleG5-1 and NleG2-3 bind and degrade host targets in human cells. (A) NleG2-3 and NleG5-1 bind and degrade their host targets following ectopic expression in HEK293T cells using the vector pcDNA3.1/nFLAG-DEST and their subsequent immunoprecipitation. (B) The indicated FLAG NleG2-3 and NleG5-1 constructs were expressed in EPEC O126:H7 ΔnleG and used to infect HeLa cells, which were then detected by anti-FLAG immunofluorescence (green) and costained for nucleus (DAPI, blue) and F-actin (phalloidin, red). (C) NleG2-3, NleG5-1, and their indicated mutants were expressed in EPEC O126:H7 ΔnleG and used to infect HeLa cells in culture along with wild-type and secretion-deficient (ΔescN) EPEC and probed for target degradation. Immunoblotting of the targets reveals specific degradation by both NleG2-3 and NleG5-1 during infection. GAPDH is a loading control in A and C.

To further address individual roles of the NleG N- and C-terminal domains in interaction with host proteins, we constructed a chimeric NleG effector composed of the N-terminal binding domain of NleG5-1 (residues 1–112) and the U-box domain of NleG2-3 (residues 90–191), which we called NleG5-1_chi. This chimeric effector was transiently expressed in HEK293T cells and the level of MED15 was measured by immunoblotting (SI Appendix, Fig. S5E). While NleG5-1_chi did not degrade MED15 as efficiently as wild-type NleG5-1, the NleG5-1_chi/I121K coimmunoprecipitated with MED15 but not with HK2, confirming the ability of this chimeric effector to specifically interact with only the identified target of NleG5-1. Furthermore, the N-terminal fragment of NleG5-1 spanning residues 1–112 transiently expressed in HEK293T cells also coimmunoprecipitated with MED15 and not with HK2. In the reciprocal experiment, we constructed a chimeric effector composed of the N-terminal domain of NleG2-3 (residues 1–88) and the C-terminal U-box of NleG5-1 (residues 105–213), called NleG2-3_chi, although we were not able to obtain substantial expression of NleG2-3_chi or NleG2-3_chi/V144K. However, the N-terminal fragment of NleG2-3 individually expressed in HEK293T cells coimmunoprecipitated with HK2 and not with MED15 (SI Appendix, Fig. S5E). Combined, these results show that the N-terminal domain in NleG2-3 and NleG5-1 are sufficient for specific interactions with human HK2 and MED15, respectively.

To test whether ubiquitination of the identified targets by NleG effectors triggers their degradation by the proteasome, we transiently expressed NleG2-3 or NleG5-1 in HEK293T cells and tested the effect of the proteasome inhibitor MG132 on the levels of HK2 and MED15. Treatment of cells with MG132 for 2 h resulted in restoration of HK2 levels in cells carrying NleG2-3^WT, although the levels of HK2 slowly lowered after this point, potentially showing the rate of HK2 turnover (SI Appendix, Fig. S6A). Similarly, a 2-h treatment with MG132 restored the levels of MED15 in cells expressing NleG5-1^WT (SI Appendix, Fig. S6B), together supporting our hypothesis that NleG-mediated ubiquitination directs their specific host protein targets for degradation by the proteasome. To confirm that NleGs are directly ubiquitinating their targets, we transiently expressed NleG5-1^WT or NleG5-1^V144K in combination with HA-tagged ubiquitin in HEK293T cells and treated these cells with MG132 for 6 h. Next, MED15 was immunoprecipitated under denaturing conditions (21) and probed for ubiquitinated MED15 species by immunoblotting using anti-HA antibodies. The amount of ubiquitinated MED15 was noticeably higher in cells expressing the NleG5-1^WT relative to the cells expressing NleG5-1^V144K variant and the vector control (SI Appendix, Fig. S6C), thus confirming that NleG5-1 directly ubiquitinates MED15 to trigger its degradation by the proteasome.

NleG5-1 and NleG2-3 Induce Specific Host Target Degradation During EPEC Infection.

The repertoire of NleG effectors varies between pathogenic E. coli, ranging from 14 members in EHEC O157:H7 to a single NleG (also known as NleI) in the EPEC strain E2348/69 (22, 23). We decided to take advantage of the low representation of NleGs in EPEC to test the individual function of NleG5-1 and NleG2-3 during infection. An EPEC ΔnleG strain was constructed and transformed with plasmids expressing FLAG-tagged NleG2-3, NleG5-1, or their ubiquitination-inactive variants. The secretion of NleG5-1 and NleG2-3 by the EPEC ΔnleG strain was confirmed by stimulating T3SS-dependent secretion in defined media (SI Appendix, Fig. S7 A and B) (24). Next, to test whether the identified interactions of NleG2-3 and NleG5-1 with host proteins were in line with their cellular localization during infection, we infected HeLa cells with EPEC expressing FLAG-tagged NleG2-3, NleG5-1, or empty vector. Consistent with its interaction with the nuclear Mediator complex, NleG5-1 predominantly localized to the nucleus, while NleG2-3 displayed nonspecific localization and was generally present in the cytosol, in line with the fact that HK2 is cytosolic and commonly localized to the surface of mitochondria (Fig. 3B) (14, 16, 25).

To determine whether NleG2-3 and NleG5-1 triggered degradation of their identified targets in the context of infection, we infected HeLa cells with EPEC strains expressing NleG2-3, NleG5-1, or empty vector, and monitored the levels of HK2 and MED15. The infection of HeLa cells by EPEC carrying wild-type NleG2-3, but not the inactive NleG2-3^I121K variant, resulted in a decrease of HK2 levels (Fig. 3C). Unexpectedly, the level of SNAP29 was also reduced during infection with EPEC expressing NleG2-3. Similarly, infection by EPEC expressing NleG5-1^WT led to complete degradation of MED15 (Fig. 3C). Additional probing for the levels of another Mediator subunit, MED6, showed no change between infections with EPEC expressing NleG5-1^WT or NleG5-1^V144K, indicating that NleG5-1–mediated degradation of MED15 does not lead to degradation of the entire Mediator complex. We did not observe any detectable changes in the level of MED15 for EPEC carrying NleG2-3, nor for HK2 or SNAP29 following infection by EPEC expressing NleG5-1. Next, we followed the infection of HeLa cells with EPEC expressing FLAG-tagged NleG2-3, NleG5-1, or a vector control with an anti-FLAG immunoprecipitation and immunoblotting for HK2, MED15, and SNAP29. We detected HK2 and MED15 coimmunoprecipitating with NleG2-3 and NleG5-1, respectively, consistent with our other results, suggesting that these are the direct targets of NleG2-3 and NleG5-1 during infection. This is in line with our hypothesis that NleG2-3 and NleG5-1 bind and ubiquitinate HK2 and MED15 during infection, respectively (SI Appendix, Fig. S7C). Taken together, our results demonstrate the EHEC effectors NleG2-3 and NleG5-1 are functionally diverse E3 ubiquitin ligases that are able to trigger the ubiquitination-mediated degradation of specific human proteins during infection.