Global mapping of Salmonella enterica-host protein-protein interactions during infection

Abstract

Intracellular bacterial pathogens inject effector proteins to hijack host cellular processes and promote their survival and proliferation. To systematically map effector-host protein-protein interactions (PPIs) during infection, we generated a library of 32 Salmonella enterica serovar Typhimurium (STm) strains expressing chromosomally encoded affinity-tagged effectors and quantified PPIs in macrophages and epithelial cells. We identified 446 effector-host PPIs, 25 of which were previously described, and validated 13 by reciprocal co-immunoprecipitation. While effectors converged on the same host cellular processes, most had multiple targets, which often differed between cell types. We demonstrate that SseJ, SseL, and SifA modulate cholesterol accumulation at the Salmonella-containing vacuole (SCV) partially via the cholesterol transporter Niemann-Pick C1 protein. PipB recruits the organelle contact site protein PDZD8 to the SCV, and SteC promotes actin bundling by phosphorylating formin-like proteins. This study provides a method for probing host-pathogen PPIs during infection and a resource for interrogating STm effector mechanisms.

Keywords: FMNL; NPC1; PDZD8; actin bundling; bacterial pathogen; cholesterol trafficking; effectors; protein-protein interactions.

Figure 1.. AP-QMS pipeline for mapping effector-host PPIs during Salmonella infection

(A) STm strains carrying C-terminally tagged effectors with STF (FLAG(2x)-TEV-STREP(3x)) were used to infect HeLa and RAW264.7 cells at a MOI ~100. Samples were either treated with DSP crosslinker or directly lysed for native anti-FLAG pulldowns. Eluates from the pulldowns were combined in a TMT-10plex labelling run and measured by LC-MS/MS. Elutions from nine different STF-tagged effectors and one untagged wildtype background control were combined. (B) Only proteins quantified with at least two unique peptides and identified in at least two biological replicates were used for analysis. A protein was a ‘hit’ when the false discovery rate (fdr) was < 1% and exhibited a fold increase of at least 20%. We further refined this list by loosening the fdr requirement to < 5% if a PPI passed the fold change (FC) requirement in both native and crosslinked conditions. Subsequently, only the strongest 20 PPIs per effector with respect to FC or fdr, as well as PPIs detected in both the native and crosslinked pulldowns, were kept for the final hit list. All analyzed data and hits are listed in Table S2. Volcano plots of all pulldowns can be found in Mendeley Data. PPI networks were built from hits and known host functional interactions.

Figure 2.. STm effector-host target physical interactions in RAW264.7 macrophages

(A) Network of PPIs identified between 12 STm effectors and their target proteins in RAW264.7 cells at 20 hpi. Only effectors identified as bait in AP-QMS and target proteins passing the criteria described in Fig. 1B are shown here. Host proteins from RAW264.7 cells are shown in gold or black (previously identified interactions; Table S3). STm are in grey. The edge color denotes the conditions interaction captured, and the thickness is proportional to the fold change (Log₂). Functionally related clusters are grouped and annotated accordingly. Network was generated using Cytoscape version 3.7.2 (Shannon et al., 2003). Murine-murine, as well as bacterial-bacterial functional interactions were extracted from the built-in STRING DB version 11 (Szklarczyk et al., 2019) protein query for Mus musculus and Salmonella with a confidence cutoff of 0.7. (B) Overview of identified PPIs in RAW264.7 cells at 20 hpi. Hits are grouped according to whether they are of murine or STm origin (upper histogram), or according to whether they were detected in native or cross-linked pulldown samples (lower Venn diagram). (C) GO-term analysis for enriched processes among all identified PPI partners. GO-term clusters are ordered according to the enrichment significance (negative logarithmic, corrected for multiple testing) (Benjamini and Hochberg, 1995; Bindea et al., 2009) and top 10 GO-clusters are displayed. n signifies the number of proteins present in cluster. Enrichments were normalized to the combined background proteome from AP-QMS experiments. Full enrichments lists can be found in Table S4.

Figure 3.. STm effector-host target physical interactions in HeLa epithelial cells

(A) Network of PPIs identified between 9 STm effectors and their target proteins in HeLa cells at 20 hpi. Network is generated and depicted as described in Fig. 2A except HeLa interactors are in blue and human data from STRING was used as input to generate edges. (B) Overview of identified PPIs in HeLa cells at 20 hpi - as in Fig. 2B. (C) GO-term analysis for biological processes across the full network as described in Fig. 2C.

Figure 4.. Comparison of STm interactomes in RAW264.7 and HeLa cells, and reciprocal PPI validation

(A) Venn diagram comparison of PPIs across the two cell lines and conditions. (B) Reciprocal pulldowns using antibodies against host targets were used to validate PPIs detected in the AP-QMS screen at 20 hpi. Effectors with similar expression levels were used in parallel pulldowns as negative controls. The PipB-PDZD8 reciprocal pulldown was performed by infecting PDZD8-myc transfected HeLa cells with STm^SL1344 ΔpipB complemented with PipB-2HA in trans; ΔpipB2 expressing PipB2–2HA in trans was used as a negative control. Two independent replicate experiments were performed, except for LASP1 and GroEL pulldowns which were performed once. One exemplary blot per interaction is shown, all blots and raw images are located in the accompanying Mendeley Data. Colored box around the Western Blot image corresponds to the cell background and condition tested (see panel A legend). Validated interactions are indicated by arrows and asterisk denotes antibody heavy chain (HC). (C) Violin plots of log₂ fold enrichments in AP-QMS for all effector-target proteins selected to be tested by reciprocal pulldowns (white), those that could (green) or could not (red) be validated, and those where the bait was not detected in the reciprocal pulldowns (grey). Dotted lines indicate median (bold) and interquartile range (light). Two-tailed Mann-Whitney test was used for significance testing. Enrichments of interactions identified in HeLa cells (blue) and RAW264.7 (orange) are shown for comparison. (D) Summary of the validation. In the upper table, interactions are considered irrespective of condition or cell line. In this case, 13 PPIs detected in the AP-QMS work, as well as 6 non-cognate controls for a total of 19 interactions was assessed. In the lower table, each cell line and condition is considered as a separate experiment wherein, 22 PPIs from the AP/MS work plus 14 non-cognate controls (i.e. 36 interactions in total) were assessed. All assessed individual reciprocal interactions are listed in Table S5.

Figure 5.. NPC1 is recruited to the SCV by SifA, where it physically interacts with SseJ and is functionally antagonized by SseL

(A) Enrichments after crosslinked pulldown of STF-tagged SseJ in HeLa cells at 20 hpi compared to untagged STm control (native pulldown in Figure S5A). Three replicates for SseJ-STF and wildtype were measured in a single TMT run. Dark blue: FC > 1.5, p-value < 0.001; light blue: FC > 1.2, p-value < 0.01. All hits can be found in the accompanying Mendeley Data. (B) Confocal microscopy images (60x) of HeLa cells infected with wildtype, ΔsseJ, ΔsseJΔsseL and ΔsifA STm expressing mCherry (pFCcGi) at 12 hpi (MOI = 100). Cells were stained with Hoechst 33342, anti-LAMP1 and anti-NPC1. Merge shows STm in red, NPC1 in green. For ΔsifA two examples are shown – a host cell in which STm grows in the SCV (upper) and one in which the SCV has ruptured and STm is in the cytosol (lower). Identically treated NPC1-K.O. cells were used to assess the anti-NPC1 specificity (blue border). Quantification of NPC1 localization to the SCV is shown in Figure S5C. Scale bar: 10 μm (C) Accumulation of cholesterol (stained with filipin) at the SCV. A total of 416 manually inspected fields of view (FOVs) across three independent experiments with four technical replicates in each run were analyzed (see Fig. S5D for representative images). The mean filipin intensity in regions of co-localization of intracellular STm with LAMP1 staining (to exclude cytosolic bacteria) was divided by the mean filipin intensity measured within the cell mask. Analysis was performed per FOV (n shown in boxplots). FOVs contained on average 20 infected cells. Boxplots (median and IQR, error bars are according to Tukey) with whiskers spanning Q10 to Q90. Unpaired, two-sided T-test with Welch correction was used to calculate p-values, without correction for multiple testing.

Figure 6.. PipB binds to and recruits PDZD8 to the SCV during infection

(A) Y2H with truncated versions of PipB. A direct interaction to PDZD8, as indicated by growth in -His conditions, is abolished by deletion of the 20 amino acid C-terminus of PipB. (B) Y2H with truncated versions of PDZD8. Deletion of the C-terminal 225 amino acids of PDZD8 impairs its interaction with PipB. (C) Western Blot analysis of immunoprecipitates from HeLa cells transfected with EGFP, EGFP-PipB, EGFP-PipB (Δ270–291) or EGFP-PipB2 fusions. Anti-GFP immunoprecipitates were analyzed by immunoblotting for endogenous PDZD8 using anti-PDZD8 peptide antibodies and anti-GFP antibodies. The PipB-PDZD8 interaction requires the last 22 amino acids of PipB. PipB2 was used as negative control. (D) HeLa cells were transfected with EGFP-PipB or EGFP-PipB(Δ270–291) and immunostained for endogenous PDZD8 (red). DNA was detected with Hoechst 33342 (blue). Scale bars: 10 μm (E) Fluorescence microscopy of HeLa cells transfected with a plasmid expressing myc-tagged PDZD8 and infected with STm^SL1344 ΔpipB complemented in trans with PipB-2HA (and constitutively expressing mCherry from the chromosome) at 12 hpi. Upper row: Translocated PipB was visualized by immunostaining with anti-HA antibodies and PDZD8 using anti-myc antibodies. Infected cells display a clear redistribution of PDZD8-myc to SCV. Lower row: HeLa cells as per the upper row, immunostained for LAMP-1. Representative image demonstrates a redistribution of PDZD8-myc to the SCV (decorated with LAMP-1), whereas PDZD8-myc remains localized to the ER in uninfected cells. Scale bars: 10 μm. (F) Fluorescence microscopy images showing that translocated 2xHA-tagged PipB co-localizes with PDZD8. HeLa cells were transfected with a plasmid expressing PDZD8-myc and infected with STm^SL1344 ΔpipB complemented in trans with PipB-2HA, and immunostained at 12 hpi. Overlay shows PipB-2HA in green, PDZD8-myc in red and LAMP-2 (decorates SCV and SIFs) in blue. PDZD8 localizes predominantly to the SCV rather than SIFs. (G) Quantification of PDZD8 recruitment in infected cells. HeLa cells were transfected with a plasmid expressing PDZD8-myc and infected with STm^SL1344 wildtype, ΔpipB, ΔpipB complemented with PipB-2HA or PipB-2HA(Δ270–291) for 12 hpi. All strains are constitutively expressing mCherry from the chromosome. PDZD8 localization was scored as the fraction of infected cells displaying recruitment of PDZD8-myc to the SCV. Data from three independent experiments with roughly 100 cells per experiment, n indicates the total cell number. Error bars denote standard deviation. Localization of PDZD8-myc to the SCV depends on full-length PipB expression in trans.

Figure 7.. SteC directly targets FMNL proteins to promote actin cytoskeleton rearrangement

(A) Size exclusion chromatograms obtained from purified recombinant FMNL1_1–385, SteC (upper panel) or catalytically inactive SteC_K256H (lower panel). Pre-incubation of FMNL1_1–385 with SteC or SteC_K256H leads to a leftward shift in elution volume compared to the individual purified proteins, as indicated by the dotted lines. (B) Autoradiography after in vitro kinase assay. FMNL1_1–385 was purified and incubated with purified SteC kinase, as well as SteC_K256H in the presence of radioactively labelled [³²P]-γ-ATP. Protein inputs were visualized by Coomassie blue staining. Only catalytic active SteC is capable of autophosphorylation and of phosphorylating FMNL1. (C) Protein maps of FMNL1 and FMNL2, including functional regions, secondary structure elements, and phosphosites identified in the in vitro kinase assay, followed by phosphoproteomics. Coloring: Blue: identified in the first replicate (using 50μM ATP), red: identified in the second replicate (5mM ATP) black: identified in both. Phosphorylation by SteC occurs mostly in flexible loops of FMNL1 and FMNL2. Results are found in Table S6. (D) Representative fluorescence microscopy images after infection of 3T3 fibroblasts (8 hpi) with mCherry-expressing STm strains (MOI 100), and staining with DAPI (blue) and phalloidin (purple). Data from three independent experiments for FMNL2/3 knockout cells (Kage et al., 2017b), and two independent experiments for wildtype 3T3 fibroblasts, spanning 162 FOVs (20 infected cells on average per view). Arrows indicate intracellular STm microcolonies. Scale bar: 30 μm. Quantification is shown in E. (E) Average actin signal intensity at STm microcolonies divided by overall average actin signal intensity as a measure of actin-STm co-localization. 162 FOVs across 60 wells were analyzed and displayed as boxplots. Boxplots and statistical tests as in Fig. 5C. (F) Model of SteC-FMNL interaction: (i) SteC directly binds FMNL subfamily formins, and is necessary and sufficient for its phosphorylation. (ii) The interaction between phosphorylated FMNL formins and Cdc42 induces actin polymerization (Kühn et al., 2015).