Research ArticleHost-Microbe Interactions

β-Barrel outer membrane proteins suppress mTORC2 activation and induce autophagic responses

See allHide authors and affiliations

Science Signaling  27 Nov 2018:
Vol. 11, Issue 558, eaat7493
DOI: 10.1126/scisignal.aat7493

Barreling over mTORC2

Autophagy can be beneficial for clearing pathogen infection. Gram-negative bacteria commonly have outer membrane proteins with a β-barrel tertiary structure, which is also found in proteins localized to the outer membrane of mitochondria (which are evolutionarily descended from symbiotic bacteria). Chaudhary et al. showed that β-barrel outer membrane proteins were recognized by specific receptors on macrophages and epithelial cells, resulting in suppression of the multiprotein complex mTORC2 and induction of autophagy. Adding outer membrane proteins to mouse macrophages enabled clearance of Salmonella Typhimurium infection. Thus, the signaling pathways activated by outer membrane proteins may contribute to immune responses that promote pathogen clearance or autoimmune diseases.


The outer membranes of Gram-negative bacteria and mitochondria contain proteins with a distinct β-barrel tertiary structure that could function as a molecular pattern recognized by the innate immune system. Here, we report that purified outer membrane proteins (OMPs) from different bacterial and mitochondrial sources triggered the induction of autophagy-related endosomal acidification, LC3B lipidation, and p62 degradation. Furthermore, OMPs reduced the phosphorylation and therefore activation of the multiprotein complex mTORC2 and its substrate Akt in macrophages and epithelial cells. The cell surface receptor SlamF8 and the DNA-protein kinase subunit XRCC6 were required for these OMP-specific responses in macrophages and epithelial cells, respectively. The addition of OMPs to mouse bone marrow–derived macrophages infected with Salmonella Typhimurium facilitated bacterial clearance. These data identify a specific cellular response mediated by bacterial and mitochondrial OMPs that can alter inflammatory responses and influence the killing of pathogens.


The β-barrel outer membrane proteins (OMPs) of Gram-negative bacteria are an important and abundant component of the bacterial outer membrane and released vesicles (OMVs). Such proteins from commensals may interact with host tissues at mucosal surfaces and are presented to deeper tissues by invasive pathogens, upon mucosal damage, in mitochondrial disorders, and as vaccine components. Mitochondrial, chloroplast, and bacterial OMPs share a structural fold, the β-barrel, because these eukaryotic organelles are evolutionarily derived from endosymbiotic bacteria. Bacterial OMPs from different species induce specific mammalian responses (13), which may explain their utility as components of vaccines, independently of their specific antigenicity or ability to be targets for cytotoxic antibodies that fix complement (2, 4). OMPs can induce inflammatory cytokine secretion by macrophages (5), trigger costimulatory molecule expression on dendritic cells (6), and induce phagocyte migration to secondary lymphoid organs (4). Therefore, understanding specific responses to OMPs is of physiologic and practical relevance.

Bacterial and mitochondrial OMPs may induce host autophagy-like pathways. The mitochondrial OMP VDAC1 (voltage-dependent anion channel 1) is a requisite for PINK1- and Parkin-mediated selective autophagy of damaged mitochondria (7). The cell surface costimulatory molecule SlamF1 can regulate activity of the NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) oxidase Nox2 complex after direct interaction with Escherichia coli porins (8). In addition, SlamF4 can inhibit biochemical autophagic responses upon engagement of its ligand SlamF2 through the Beclin-1/Vps34 complex (9). Therefore, we investigated the ability of endotoxin-free preparations of purified OMPs from bacterial and mitochondrial membranes to trigger specific autophagy-associated responses in mammalian cells.


OMPs trigger Atg16L1-dependent endosomal acidification in human monocyte–derived THP1 cells

We have previously used endosomal acidification as a quantitative trait to screen Hapmap cells for human variation in cellular autophagy (9). We therefore tested the ability of increasing doses of OMPs to influence endosomal acidification in mammalian THP1 cells (Fig. 1A) using OMPs purified from various Gram-negative bacteria by affinity purification followed by gel filtration (fig. S1, A to C). The porin OmpF from Salmonella Typhimurium, the smaller OMP OprF from Pseudomonas aeruginosa, OmpA from E. coli, and OmpA from Acinetobacter baumannii were either solubilized from outer membrane preparations derived from spheroblasts or refolded from inclusion bodies. The treatment of THP1 cells with up to 1 μM purified OMPs did not trigger cell death or tumor necrosis factor–α (TNF-α) secretion (fig. S2, C and D), indicating that host Toll-like receptor (TLR) or the inflammasome pathways were not activated. Treatment of THP1 cells with nanomolar concentrations of OMPs triggered an increase in LysoTracker staining, consistent with increased endosomal acidification (Fig. 1, A and B, and fig. S2, A and B). Furthermore, the OMP-mediated induction of endosomal acidification was decreased by silencing Atg16L1, an important component of the autophagy protein complex required for LC3 recruitment to membranes (Fig. 1, C and D). Denaturation of OMPs by boiling prevented the induction of endosomal acidification in cells (Fig. 1, A and B), indicating that this effect was mediated by protein activity or folding, but not by associated copurified lipid moieties. Consistently, the treatment of THP1 cells with increasing doses of purified Salmonella lipopolysaccharide (LPS) and peptidoglycan (PGN) did not induce similar staining (fig. S2E). Similar preparations of Salmonella flagellin FliC and E. coli K12 fimbriae 1 subunit K12FimH did not induce cellular endosomal acidification (Fig. 1A), indicating that this cellular response was specific to bacterial OmpA, OprF, and OmpF proteins, and not due to associated bacterial macromolecules.

Fig. 1 Purified OMPs trigger autophagy in macrophages and intestinal epithelial cells.

(A) THP1 cells were treated with control buffer or the indicated doses of E. coli OmpA, S. Typhimurium OmpF, E. coli K12FimH, S. Typhimurium FliC, corresponding volume of buffer containing detergent octyl-β-glucopyranoside, or previously boiled E. coli OmpA for 2 hours. In addition, control small interfering RNA (siRNA)– and SlamF8 siRNA–treated THP1 cells were treated with increasing doses of the indicated OMP or buffer. Cells were stained with LysoTracker and analyzed by flow cytometry. Means ± SD of the percentage of cells with OMP-mediated endosomal acidification are shown from three independent experiments. For control siRNA compared to SlamF8 siRNA upon VDAC1 treatment, P < 0.05, Wilcoxon test; for E. coli OmpA compared to boiled E. coli OmpA, P < 0.05, Wilcoxon test. (B) Representative flow cytometry images for (A) showing LysoTracker staining of THP1 cells treated with the indicated doses of E. coli OmpA and previously boiled E. coli OmpA for 2 hours, representative of three independent experiments. The gated population shows the percentage of cells with LysoTracker staining for each condition. (C) THP1 cells were transfected with control or Atg16L1 siRNA, treated with 250 nM P. aeruginosa OprF for 2 hours, and stained with LysoTracker. Means ± SD from three independent experiments are shown. **P < 0.005, t test. Right: Atg16L1 Western blot from a representative experiment showing the efficiency of RNA silencing. (D) Representative flow cytometry images for (C) showing LysoTracker staining of control and Atg16L1 siRNA–transfected THP1 cells treated with P. aeruginosa OprF for 2 hours. The gated population shows the percentage of cells with LysoTracker staining for each condition. (E) Representative Western blot for LC3B and p62 using cell lysates prepared from THP1 cells treated with 100 nM OmpA for 15 min, with or without bafilomycin before treatment for 2 hours. Blots are representative of three independent experiments. (F) Representative Western blots showing p62 degradation and LC3B lipidation in THP1 cells transfected with SlamF8, SlamF2, or control siRNAs and treated with 100 nM VDAC1. Blots are representative of three independent experiments. (G) Cleavage of GFP-LC3 in HEK293T cells treated with 100 nM E. coli OmpA or rapamycin for 15 min. Blots are representative of three independent experiments.

Because treatment of host cells with bacterial OMP induced endosomal acidification, we next asked whether other proteins with similar structural folds could induce the same response. VDAC1 is a mitochondrial OMP (7) with a similar β-barrel fold as bacterial OMPs. Similar to bacterial OMPs, treatment of THP1 cells with purified VDAC1 also induced dose-dependent endosomal acidification (fig. S2A). Together, these data indicate that the treatment of THP1 cells with specific bacterial and mitochondrial OMPs triggers the induction of Atg16L1-dependent endosomal acidification.

Bacterial and mitochondrial OMPs induce LC3B and p62 degradation in human monocyte–derived and epithelial cell lines

Because Atg16L1 facilitates LC3B conjugation to phosphatidylethanolamine in autophagy, we investigated the effects of OMP treatment on LC3B lipidation. Treatment of THP1 cells with low doses of purified E. coli OmpA induced the lipidation of LC3B, an effect that was increased upon inhibition of acidification of endosomal pH with bafilomycin (Fig. 1E). The autophagosome substrate p62 (also known as SQSTM1), a cargo for LC3B-mediated autophagic degradation (10), was also degraded upon OMP treatment, and this degradation was inhibited by bafilomycin treatment. Similarly, treatment of THP1 cells with recombinant VDAC1 induced the degradation of LC3B and p62 (Fig. 1F and fig. S3A). Because OMPs induced autophagy in monocyte-derived cell lines, we next asked whether OMPs could trigger similar phenotypes in epithelial cells. Both rapamycin and E. coli OmpA treatment triggered the cleavage of green fluorescent protein (GFP)–tagged LC3 to generate free GFP in an in vitro assay for autophagy in human embryonic kidney (HEK) 293T cells (Fig. 1G) (11). In addition, immunofluorescence studies in polarized epithelial colorectal Caco2/T7 cells demonstrated the induction of LC3B puncta in OmpA-treated cells (Fig. 2, A and B). Together, these data indicate that bacterial and mitochondrial OMPs trigger LC3B and p62 degradation in monocytes and epithelial cell lines.

Fig. 2 SlamF8 and XRCC6 are important for OMP-mediated autophagy in macrophage and epithelial cells.

(A and B) Representative immunofluorescence (A) showing LC3B puncta in XRCC6 or control siRNA–transfected Caco2 cells, treated with OmpA and/or chloroquine (CQ) as indicated. Images are representative of three independent experiments. Scale bar, 20 μm. LC3B puncta were counted from a total of 100 cells for each condition from three different experiments (B). ***P < 0.001, t test comparing effect of control and XRCC6 siRNA in OmpA-treated cells. (C) Percentage of LysoTracker-positive cells induced in response to E. coli OmpA treatment of BMDMs from SlamF8−/−, SlamF1,5,6−/−, SlamF4−/−, and SlamF1,4,5,6+/+ mice for 2 hours. Means ± SD of three independent experiments using freshly derived BMDMs are shown. F1.516,4.549 = 92.62, P < 0.005 for overall comparisons of matched data between SlamF8−/− and SlamF8+/+, repeated-measures one-way analysis of variance (ANOVA); F3,12 = 52.57, P <0.0001 for comparisons between SlamF8−/− and SlamF8+/+ BMDMs at individual OMP doses, repeated-measures one-way ANOVA. (D) Representative flow cytometry images for (C) showing LysoTracker staining of SlamF8+/+ and SlamF8−/− BMDMs treated with the indicated doses of E. coli OmpA for 2 hours. (E) Percentage of cells stained with LysoTracker upon rapamycin treatment of SlamF8+/+ and SlamF8−/− BMDMs. Data are shown from three independent experiments using freshly derived BMDMs. (F) Representative Western blot for p62 in control and SlamF8 siRNA–transfected THP1 cells upon OMP treatment. Data are representative of results from three independent experiments. (G) PKC activity from SlamF8+/+ and SlamF8−/− BMDMs treated with 100 nM E. coli OmpA or VDAC1 as indicated. Means ± SD from three independent experiments are shown. ***P < 0.005. (H) Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis of SlamF8 mRNA levels in THP1 and Caco2 cells after treatment with 10 nM OMP for 16 hours. Means ± SD of three independent experiments performed in technical triplicate are shown. *P < 0.05. (I) Representative Western blot for p62 and LC3B expression from Caco2 cells transfected with XRCC6 or control siRNA and treated with 100 nM OmpA. Data are representative of results from three independent experiments.

SlamF8 mediates the host response to bacterial and mitochondrial OMPs in macrophages

The signaling lymphocyte activation molecule (SLAM) family member SlamF4 is an inhibitor of cellular autophagy in response to rapamycin and serum starvation (9). In addition, family members SlamF1 and SlamF8 regulate Nox2 activity in macrophages (12, 13) that can generate reactive oxygen species (ROS) and that can precede LC3BII recruitment to the phagosome (14). Last, SlamF1 recognizes the E. coli porins OmpC and OmpF, and the proteins themselves have a β-barrel structure that can mediate homotypic interactions (8). We therefore examined the role of Slam family receptors in OMP-mediated host responses. Whereas bone marrow–derived macrophages (BMDMs) from SlamF1, SlamF4, SlamF5, and SlamF6 knockout mice showed an induction of OMP-mediated endosomal acidification, BMDMs from SlamF8 knockout mice showed reduced induction of endosomal acidification in response to OMP treatment, with only around 5% cells showing increased LysoTracker staining when exposed to 2 μM OmpA (Fig. 2, C and D). The loss of SlamF8 did not affect rapamycin-mediated induction of endosomal acidification in mouse BMDMs because acidification occurred to a similar extent as in wild-type BMDMs (Fig. 2E and fig. S3C), indicating that the machinery for endosomal acidification was intact. SlamF8 BMDMs also showed reduced basal endosomal acidification that may be a result of loss of homotypic SlamF8 interactions. However, the effects of SlamF8 on endosomal acidification were OMP specific. Furthermore, unlike wild-type BMDMs that showed degradation of LC3B with 15 min of OmpA treatment, SlamF8−/− BMDMs showed inhibited degradation of LC3B at more than 90 min of OmpA treatment (fig. S3B). Silencing of SlamF8 in THP1 cells also inhibited OmpA-mediated induction of endosomal acidification (fig. S2A) and LC3BII and p62 degradation (Fig. 2F and fig. S3A). Similar to OmpA, VDAC1 induced endosomal acidification in THP1 cells that was attenuated in SlamF8 siRNA–treated THP1 cells (Fig. 1A). In addition, silencing of SlamF8, but not SlamF2 expression abrogated VDAC1-induced LC3B and p62 degradation (Fig. 1F and fig. S3A).

SlamF8 regulates Nox2 activity by activating protein kinase C (PKC) (13, 15). We therefore investigated the effect of OMPs on SlamF8-mediated activation of PKC. BMDMs from wild-type mice showed robust phosphorylation of a PKC substrate peptide in response to VDAC1 and OmpA (Fig. 2G), whereas PKC activation was significantly impaired in BMDMs from SlamF8 null mice. Together, these data suggest that OMPs activate PKC through SlamF8, thereby mediating endosomal acidification and LC3B and p62 degradation in macrophages and monocytic cells.

SlamF8 expression is induced upon OMP treatment of monocytic cells, and XRCC6 mediates OMP responses in epithelial cells

SlamF8 expression in macrophages is induced upon activation of cells by interferon-γ (IFN-γ) and Gram-negative bacteria (12). We therefore examined induction of SlamF8 mRNA in THP1 cells that were stimulated with purified OMP or serum starved to induce conventional autophagy. OmpA treatment induced SlamF8 mRNA by 50-fold, whereas serum starvation did not affect SlamF8 mRNA expression in THP1 cells (Fig. 2H). Expression of SlamF8 was not detectable in epithelial Caco2 cells and was not induced by OMPs (Fig. 2H). Therefore, we rationalized that although specific OMP-mediated macrophage responses depended on SlamF8, similar responses in epithelial cells may be mediated by other receptors.

The Rickettsia OMP protein rOmpB is a ligand for the XRCC6 subunit of DNA-dependent protein kinase (DNA-PK), and OMP binding and recruitment of XRCC6 at the plasma membrane enable the invasion of Rickettsia conorii into nonphagocytic mammalian cells (16). Because SlamF8 expression was low in epithelial cells and not altered upon OmpA treatment, we investigated a role for XRCC6 in OMP-mediated host responses in epithelial cells. Silencing of XRCC6 in Caco2 epithelial cells abrogated OmpA-mediated LC3B lipidation and p62 degradation (Fig. 2I and fig. S3D). In addition, XRCC6 siRNA–treated Caco2 cells showed significantly fewer LC3B puncta upon chloroquine treatment (Fig. 2, A and B). Together, these data indicate that XRCC6 mediates autophagy responses to OmpA in polarized Caco2/TC7 human epithelial cells.

OMPs reduce the phosphorylation of mTOR2 and Akt at activating sites in Caco2/TC7 cells

Because OMP treatment affected endosomal acidification, LC3B recruitment, and p62 degradation in mammalian cells, we investigated the effects of purified OMPs on the activation of mammalian target of rapamycin (mTOR), a kinase that is a key regulator of cellular autophagy. OMP treatment of Caco2 cells reduced the phosphorylation of mTOR at Ser2481 but not Ser2448 (Fig. 3A). This selective effect on these phosphorylation sites indicated that OMP treatment preferentially inhibited mTOR associated with the mTORC2 complex (17), which promotes the activation of serine-threonine kinase Akt (18). We therefore examined the phosphorylation of Ser473 in Akt as an indicator of Akt activation in response to OMP treatment. Consistently, OMP treatment of Caco2 cells inhibited the phosphorylation of Ser473 in Akt in a dose-dependent manner (Fig. 3B). Together, these data indicate that OMP can inhibit the mTORC2/Akt signaling pathway.

Fig. 3 OMPs inhibit mTORC2 and Akt activation and inhibit intracellular bacterial replication.

(A) Representative Western blots probing the phosphorylation of mTOR at Ser2481 and Ser2448 in Caco2 cells in response to 100 nM OmpA treatment for 15 min. Data are representative of results from three independent experiments. (B) Representative Western blots for phosphorylation of Akt at Ser473 in Caco2 cells upon treatment with 0, 5, 10, 50, and 100 nM OmpA for 15 min. Data are representative of results from three independent experiments. (C) Representative Western blot showing time course of OMV-mediated p62 and LC3B degradation in Caco2 cells treated with B. theta OMVs (10 μg/ml). Data are representative of results from three independent experiments. (D) Representative Western blot for GFP showing cleavage of GFP-LC3 in HEK293T cells transfected with GFP-LC3 and treated with B. theta OMVs (0, 1, 5, and 10 μg/ml) for 15 min. Data are representative of results from three independent experiments. (E) Representative Western blot showing LC3B levels in BMDMs treated with B. theta OMVs (10 μg/ml) for the indicated times. Data are representative of results from three independent experiments using freshly derived BMDMs. (F) Representative Western blot showing Akt Ser473 phosphorylation in BMDMs treated with control (C), LPS (L), and OmpA (O) or cotreated with LPS + OmpA (LO) for 15 min or 1 hour after S. Typhimurium (S.Tm) infection at a multiplicity of infection (MOI) of 10. Data are representative of results from three independent experiments using freshly derived BMDMs. (G) Representative Western blot showing the time course of mTOR Ser2481 and Akt Ser473 phosphorylation in total lysates from S. Typhimurium–infected BMDMs at an MOI of 10. Data are representative of results from three independent experiments using freshly derived BMDMs. (H) Means ± SD of colony-forming units (CFUs) in BMDMs infected with S. Typhimurium at an MOI of 10, further treated with OMP or control buffer 1 hour after infection. (I) Ratio of CFUs at 16 hours/2 hours in control or OMP-treated S. Typhimurium–infected BMDMs. Means ± SD from three independent experiments using freshly derived BMDMs are shown. **P < 0.01. (J) Mechanistic model of action of OMPs on SlamF8 in macrophages and XRCC6 in epithelial cells.

OMVs from Bacteroides thetaiotaomicron trigger LC3B lipidation and p62 degradation in an intestinal epithelial cell line but not macrophages

Outer membrane lipid vesicles that contain OMPs from Neisseria meningitides (19) and S. Typhimurium (20) can induce adaptive immune responses by increasing the abundance of the major histocompatibility complex class II receptor human lymphocyte antigen (HLA)–DR on macrophages, independently of LPS (20). We therefore examined the induction of HLA-DR in OmpA-treated THP1 cells using flow cytometry. Purified OmpA and LPS, both of which are components of OMV, induced the surface expression of HLA-DR in these cells (fig. S4, A and B). These results illustrated the difficulty in separating the effects of LPS from OMPs when administering whole bacteria and outer membrane fragments in macrophages. Intestinal symbionts such as B. thetaiotaomicron (B. theta) physiologically produce OMVs within the intestinal tract (21). We therefore asked whether B. theta OMVs induced autophagy upon administration to differentiated epithelial cells, which lack the TLR4/MD2 (myeloid differentiation protein 2) complex. Treatment of polarized Caco2 cells with B. theta OMVs triggered LC3B lipidation and the degradation of p62 (Fig. 3C). OMV treatment of HEK293T cells expressing GFP-LC3 also triggered the generation of detectable-free GFP (Fig. 3D). Together, these results demonstrate that physiologically generated OMVs can trigger autophagy responses in highly differentiated intestinal epithelial cell lines. In support of this effect, OMVs from Porphyromonas gingivalis and enterohemorrhagic E. coli have been detected within the lysosomal compartments of host cells (22, 23), and OMV from Helicobacter pylori and P. aeruginosa can induce autophagosome formation in epithelial cells (24). However, OMV treatment of mouse BMDMs did not affect LC3B lipidation (Fig. 3E). Together, these data indicate that although treatment with both purified OMPs and OMVs can induce autophagy in epithelial cells, these responses are induced in macrophages only upon treatment with purified OMPs.

OMPs counteract the effects of LPS on the mTORC2/Akt pathway and inhibit intracellular bacterial replication

Because LPS treatment of macrophages induced mTOR phosphorylation at Ser2481, we speculated that the effects of OMPs on macrophages could be countered by the effects of LPS, thereby providing an explanation for the lack of OMV responses in macrophages. Consistently, LPS treatment induced activating phosphorylation of Akt at Ser473 in BMDMs, whereas cotreatment of cells with LPS and OMP suppressed the LPS-mediated phosphorylation of Ser473 Akt (Fig. 3F). These data indicated that in the presence of LPS and other unknown molecules in OMVs, OMP-mediated effects on the mTORC2/Akt signaling pathway may be suppressed in macrophages. Gram-negative, intracellular pathogen, S. Typhimurium contain LPS, OMP, and a battery of effector proteins that can modulate host Akt pathways. S. Typhimurium–infected cells showed reduced Akt phosphorylation compared to control cells at 1 hour after infection (Fig. 3F). We therefore examined the changes in mTOR Ser2481 and Akt Ser473 phosphorylation over 120 min in BMDMs infected with S. Typhimurium in vitro. Infected BMDMs demonstrated reduced phosphorylation of mTOR Ser2481 and Akt Ser473 at 60 to 120 min after infection, suggesting inactivation of this pathway (Fig. 3G). We then elucidated the consequences of OMP treatment on intracellular bacterial replication in these cells (Fig. 3H). Further treatment of infected BMDMs with OmpA at 60 min after infection inhibited intracellular bacterial replication by greater than a log scale (Fig. 3I), indicating that OMP-mediated host cell responses may facilitate host bacterial clearance through inactivation of mTORC2/Akt signaling. Although bacterial clearance was impaired upon silencing Atg16L1 in THP1 cells, the ability of OmpA to facilitate intracellular bacterial clearance was also abrogated in Atg16L1-silenced cells, indicating a requirement for autophagy in this effect of OmpA (fig. S4C). Collectively, these data indicate the importance of OMP-mediated autophagy in host cells.


In this study, we demonstrated that OMP from bacteria and mitochondrial membranes triggered specific cellular responses, including LC3B lipidation, p62 degradation, and mTORC2 inactivation in macrophages and epithelial cell lines (Fig. 3J). We demonstrated that these OMP-mediated responses in macrophages depended on SlamF8, an IFN-γ and E. coli inducible gene (12), and were likely masked by countering effects of LPS when presented in OMVs or whole bacteria. When presented to macrophages, OMPs activated S. Typhimurium clearance and induced the surface expression of HLA-DR.

Collectively, these findings define a host response likely mediated upon recognition of a specific set of proteins that share a unique β-barrel structure (25) and are an integral component of Gram-negative bacterial and eukaryotic mitochondrial membranes. Like other innate immune ligands, this unique structure is a logical candidate for recognition of autonomous immune responses, which may affect autoimmune diseases in which antibodies to E. coli OMP are detected, in infectious diseases with Gram-negative bacteria, in autoimmune diseases with a bacterial component at mucosal surfaces such as inflammatory bowel diseases (IBDs), and in vaccine responses. Consistent with this possibility, we found that OMPs increased HLA surface expression and S. Typhimurium killing by macrophages.

Our data show that OMP-mediated autophagy required SlamF8 in macrophages. SlamF8 can inhibit PKC in macrophages, thereby inhibiting the NADPH oxidase Nox2 (12). We demonstrated that OMPs activated PKC through SlamF8 (Fig. 2G). PKC activation of Nox2 is important for the induction of antibacterial autophagy pathways (14). Furthermore, the Slam family members SlamF4, which is implicated as an inhibitor of autophagy in response to rapamycin and serum starvation (9), and SlamF1, which can bind and respond to bacterial OMPs (8), can both engage Beclin-1/Vp34 upon extracellular engagement of ligand, suggesting that this family of receptors is intimately associated with OMP recognition, autophagy, and other immune responses. OMPs are not large particles and therefore unlikely to undergo LC3-associated phagocytosis. Nonetheless, establishing a role for proteins like Rubicon, which can discriminate canonical autophagy from LC3-associated phagocytosis, will clarify the mechanism of action for OMPs (26). Last, SlamF8 contains a potential noncanonical LC3-interacting motif (YKDVLLVVVP), which is found in the LC3-interacting proteins NDP52 and Tax1bp1 (27, 28) that could mediate direct interaction with components of the autophagy machinery.

Our results showed that OMVs and purified OMPs induced autophagy in epithelial cells, but that this response was detected in cultured macrophages only with purified OMPs. The release of OMVs and the effects of OMPs could be moderated by LPS responses in macrophages, whereas in epithelial cells OMP responses may be more dominant and relevant for OMVs that are liberated from microbiota or pathogens at mucosal surfaces. The macrophage responses observed may be critical in organisms such as Yersinia pestis, the cause of plague, or Francisella tularensis, the cause of tularemia, and other Gram-negative infections in which LPS generated from the microbe is poorly recognized by human TLR4/MD2 (29). Antigen-presenting cells may also more robustly respond to vaccines containing OMPs that do not contain LPS or robust adjuvants. Last, during mitochondrial disorders, release of OMPs such as VDAC1 into the circulation or tissues could result in the induction of autophagy responses in cells or tissues not directly affected by the disease.

In a gastrointestinal epithelial cell line, we demonstrated that OMPs required the XRCC6 subunit of DNA-PK (16) for induction of cellular responses. These DNA-PKs colocalize with Akt at the plasma membrane (30), on lipid rafts (31), and in the cytosol (32) and translocate to the nucleus after activation. DNA-PKs can activate Akt (33) upon association with mTORC2 component SIN1, which is required for Akt phosphorylation (34). We showed that OMP inhibited phosphorylation of Ser473 in Akt, which is an important requisite for the induction of both macroautophagy and chaperone-mediated autophagy (35), a cell survival pathway induced in response to oxidative or genotoxic stress (36, 37). In addition, ROS can also regulate mTOR-independent autophagy (38). OMPs may mediate the deactivation of DNA-PK, leading to the deactivation of the mTORC2-Akt axis. In addition, OMPs may also promote autophagy flux by modulating cellular ROS levels that can induce both autophagy and lysosome biogenesis.

OMP-mediated autophagy could substantially affect mammalian physiology because autophagy is important for Paneth cell morphology and function (39), in the induction of T regulatory cells to suppress mucosal inflammation (40), and in antigen presentation (41). The response of differentiated, polarized, gastrointestinal epithelial cell lines to OMPs and commensal membrane vesicles may play a role in gastrointestinal inflammation in humans. During normal intestinal homeostasis, the epithelium would be protected from released OMPs by the mucous layer; however, an inflammatory response such as IBD could allow vesicles and OMP access to cells in the gastrointestinal tract. Some individuals with Crohn’s disease have relatively high concentrations of antibodies to E. coli OmpC, raising the possibility that autophagy-mediated antigen presentation may be altered in these individuals (42). In addition, the occurrence and magnitude of seroreactivity have been associated with complicated small bowel disease and the need for surgical intervention. Polymorphisms in autophagy genes Atg16L1 and IRGM are risk factors for IBD (43, 44). XRCC6 is important for colonic homeostasis, and XRCC6 null mice show a loss of goblet cells and lymphocyte infiltration (45). These conditions are frequently seen in individuals with IBD (46) and associated with defective colonic mucosa (47). In addition, there is predicted global variation in the human expression of XRCC6, based on noted frequency of ancestral and derived alleles for single-nucleotide polymorphism rs720825, a cis–expression quantitative trait locus for XRCC6, with Amerindians from South America showing the highest frequency of the derived allele (0.850), whereas Mbuti from Africa show the lowest frequency (0.030) (48). It is therefore tempting to speculate that differential expression of XRCC6 could likely contribute to differences in intestinal autophagy responses to OMPs and OMVs, which could affect immune responses in different individuals. Last, a missense mutation predicted to affect the protein function of SlamF8 is implicated as a causal variant for IBD (49), raising the intriguing possibility that this mutation may alter OMP responses in individuals.

In conclusion, we present evidence that OMPs from multiple Gram-negative bacteria and mitochondria triggered noncanonical mTORC2-dependent autophagic responses, likely through specific receptors in macrophages and epithelial cells. These responses were moderated in macrophages by specific responses to LPS. Consequently, common genetic variations in receptors that modulate human responses to LPS could also affect OMP-mediated autophagic responses to bacteria in macrophages. These responses may be important for the efficacy of many vaccines that contain OMP and could play a role in epithelial development and maturation, as well as mucosal immune responses that could affect various autoimmune diseases. Last, OMP or artificial peptides that contain relevant motifs to specifically manipulate mTORC2 may be useful as therapeutics to alter this pathway.


Cells and reagents

THP1 cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS), 2 mM glutamine, penicillin G (100 U/ml), and streptomycin (100 μg/ml) and assayed in phenol red–free RPMI 1640 medium with 10% FBS, unless stated otherwise. Caco2/TC7 cells were seeded onto 12-well Transwell membranes with 3-μm pores and maintained for 21 days to allow polarization and the formation of tight junctions. Antibodies against SlamF8 (human AF1907, Novus; mouse AF4156, R&D Systems), GFP (A11122, Invitrogen), LC3B (51520 Abcam), p62 (SQSTM1) (5114S, Cell Signaling Technology), and β-actin–horseradish peroxidase (20272, Abcam) were purchased as indicated. The mycDDK-tagged hSlamF8 and VDAC1 plasmids were obtained from OriGene. TaqMan probes Hs00252301_m1 and Hs_02758991_g1 were used for SlamF8 and GADPH, respectively. Accell SlamF8, Atg16L1, and XRCC6 siRNA SMARTpool were purchased from Dharmacon/GE Life Sciences. A plasmid encoding the ompA gene of A. baumannii was obtained from H. Nikaido (PubMed Central identifier: PMC3416538).

Ethics statement

All mice were maintained according to the Guide for the Care and Use of Laboratory Animals (National Institutes of Health publication 1.0.0; revised 2011) and approved by the Institutional Animal Care and Use Committee at the University of Washington.

Plasmid preparation

Plasmids to express proteins from the oprF gene of P. aeruginosa (PA14), the ompA gene of E. coli, and the ompF gene of S. Typhimurium were created by designing primers to amplify those genes by PCR and to contain restriction sites for subsequent digestion and ligation into the pet29b vector, which generates a C-terminal His6 tag. PCR, digestion, and gel purification were all performed using standard methods. The plasmids were then transformed into BL21 E. coli for subsequent protein expression.

Outer membrane fraction preparation

BL21 cells containing expression plasmids were grown in LB containing kanamycin (50 μg/ml) to OD (optical density) 0.4, then induced with IPTG (isopropyl-β-d-thiogalactopyranoside), and grown to early stationary phase at 30°C. The membrane preparation protocol was adapted from Osborn et al. (50). Briefly, cells were pelleted, then resuspended in 25 ml of a cold hypertonic solution [10 mM tris (pH 7.8), 0.5 M sucrose, and lysozyme (300 μg/ml)], and incubated on ice for 5 min. Spheroplasts were formed by diluting while stirring on ice for 10 min with 20 ml of 1 mM EDTA and were subsequently stabilized by adding 5 ml of 1 M MgCl2. The cells were then pelleted and resuspended in 35 ml of 0.25 M sucrose in 10 mM tris containing protease inhibitors, deoxyribonuclease, ribonuclease, and 1.5 mM MgCl2. Cells were lysed using an Avestin high-pressure homogenizer, and the membranes were pelleted by ultracentrifugation. Membranes were resuspended in 20% sucrose and then separated by overnight ultracentrifugation in a sucrose gradient of 73, 53, and 20% sucrose. The following day, outer membrane fractions were collected and resolubilized in tris-buffered saline (TBS) with 10% glycerol and 1% n-dodecyl β-d-maltoside (DDM) or octyl β-glucopyranoside as indicated.

OMP purification

Solubilized outer membrane fractions were loaded onto a 5-ml HiTrap fast protein liquid chromatography (FPLC) column containing 0.1 M NiSO4. Nickel columns were purified with TBS with 10% glycerol and 0.1% DDM, and proteins were eluted using 300 mM imidazole. Appropriate fractions were collected and loaded onto a Superdex 200 column for size-exclusion chromatography using a buffer containing TBS with 10% glycerol and 0.02% DDM. FPLC was performed using the ÄKTA protein purification system. Protein was tested for endotoxin with the Limulus amebocyte lysate test, and all preparations were tested negative for LPS.

OMV purification

B. theta VPI-5482 were grown in brain heart infusion broth supplemented with hemin (5 μg/ml) and menadione (1 μg/ml) in an anaerobic atmosphere of 90% N2, 5% H2, and 5% CO2. The culture was centrifuged at 10,000 rpm at 4°C. The supernatant was spun down twice and then filtered sequentially through filters of 5, 1, 0.6, and 0.2 μm size to eliminate bacteria. The clarified supernatant was then centrifuged in a Ti45 rotor at 40,000 rpm for 1 hour. The OMV pellet was dissolved in 10 ml of 10 mM Hepes (pH 7.4) buffer and spun down in a Ti70 rotor at 50,000 rpm. The pellet was resuspended in 1 ml of 10 mM Hepes (pH 7.4) buffer, and the protein concentration was determined using Bradford protein reagent.

Measurement of endosomal acidification by flow cytometry

Endosomal acidification in live cells was measured using LysoTracker Red DND-99 (Life Technologies). The autophagic state of the two distinct LysoTracker-stained populations of cells has been described previously (9).

Cell lysis and Western blots

Whole-cell lysates were collected in modified radioimmunoprecipitation assay (RIPA) buffer [150 mM NaCl, 1% NP-40, 0.25% sodium deoxycholate, and 50 mM tris-Cl (pH 7.8)] containing 1× cOmplete mini protease inhibitor tablet (Roche), 10 mM NaF, and 1 mM activated Na3VO4. Lysates were separated by SDS–polyacrylamide gel electrophoresis, transferred to Immobilon membranes, and subjected to Western blotting as indicated.

RNA interference experiments

Cells (2 × 105) were treated for 3 days in 500 μl of Accell medium (Dharmacon) with either 10 μM nontargeting Accell siRNA or a SMARTpool directed against a specific gene (human SlamF8, Atg16L1, or XRCC6 in this study). Cells were collected, counted, and checked for viability with a Guava easyCyte flow cytometer using ViaCount, then plated at 1 × 105 in 100 μl of assay medium without antibiotics in 96-well plates, and treated with designated stimulus or control as noted.

Real-time qPCR

RNA was extracted from 105 to 106 cells using the Qiagen RNeasy Plus Mini Kit, transcribed to complementary DNA with the Bio-Rad iCycler, and subjected to real-time PCR using the Applied Biosystems TaqMan Fast Universal PCR Master Mix and the indicated TaqMan probes. GADPH was used as an internal reference to normalize transcript levels. The relative gene mRNA levels were determined by the 2−ΔΔCt method.

Measurement of PKC activity in BMDMs

PKC kinase activity was measured using the ADI-EKS-420A Assay Kit (Enzo Life Sciences, Plymouth Meeting, PA). Briefly, SlamF8+/+ or SlamF8−/− BMDMs were stimulated with 100 nM E. coli OmpA or VDAC1 for 15 min and then lysed with RIPA buffer. Protein was estimated by Bradford assay, and PKC activity in lysates was measured according to the manufacturer’s protocol.

S. Typhimurium infection of BMDMs

Briefly, overnight bacterial cultures were subcultured 1:33 and grown for 2 hours at 37°C in LB. BMDMs were infected at an MOI of 10 for 1 hour followed by gentamicin (50 μg/ml) to kill extracellular bacteria and OmpA or control buffer where indicated. Cells were collected at 2 and 16 hours after infection and lysed, and intracellular bacteria were enumerated by colony counting of serially diluted lysate.

Statistical analysis

Cell culture experiments were independently repeated at least three times using matched controls. The data were pooled, and statistical analyses were performed. Results are expressed as means ± SD for all in vitro experiments. Statistical analysis of dose-response curves was performed using the Wilcoxon matched-pairs signed-rank test or repeated-measures ANOVA as indicated. Single-dose data were assessed by Student’s t test or by one-way ANOVA, and P values below 0.05 were considered statistically significant. All statistical tests were performed using Prism 7 for Mac OSX (GraphPad Software).


Fig. S1. Representative purification of an OMP.

Fig. S2. Purified OMPs trigger autophagy but not cell death in THP1 cells.

Fig. S3. SlamF8 and XRCC6 are required for OMP-mediated autophagy in macrophages and epithelial cells.

Fig. S4. Purified OmpA induces surface expression of HLA-DR and facilitates Atg16L1-dependent bacterial clearance in THP1 cells.


Funding: This work was supported by an award from the NIAID to the Northwest Regional Center of Excellence for Biodefense and Emerging Infectious Diseases Research (U54 AI057141), a grant from the Rainin Foundation to S.I.M., and NIH grant PO1-AI065687-01A1 to C.T. Author contributions: A.C. and S.I.M. contributed to study design, data analyses, and manuscript preparation. A.C. and M.L. performed cell-based experiments. L.K., C.K., H.K., and M.-P.B. cloned, expressed, and purified OMP. C.K. and H.K. generated B. theta OMV. M.A.A. performed immunofluorescence analysis. M.-P.B. derived BMDMs from femurs from wild-type C57/BL6 and BALB/c. G.W. and C.T. generated SlamF1, SlamF4, SlamF5, SlamF6, and SlamF8 null mice. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are in the paper or Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article