A lipid zipper triggers bacterial invasion - PNAS

4 downloads 0 Views 2MB Size Report
Sep 2, 2014 - Edited by Kai Simons, Max Planck Institute of Molecular Cell Biology and ..... into our cell model, the human lung epithelial cell line H1299,.
A lipid zipper triggers bacterial invasion Thorsten Eierhoffa,b,1, Björn Bastianc, Roland Thuenauera,b, Josef Madla,b, Aymeric Audfrayd, Sahaja Aigala,b,e, Samuel Juillota,b,f, Gustaf E. Rydellg, Stefan Müllera,b, Sophie de Bentzmannh, Anne Imbertyd, Christian Fleckc,i,1, and Winfried Römera,b,f,1 a Faculty of Biology, bBIOSS Centre for Biological Signalling Studies, cCentre for Biological Systems Analysis, and fSpemann Graduate School of Biology and Medicine, Albert Ludwigs University Freiburg, 79104 Freiburg, Germany; dCentre de Recherches sur les Macromolécules Végétales, Université Grenoble Alpes and Centre National de la Recherche Scientifique, F38000 Grenoble, France; eInternational Max Planck Research School for Molecular and Cellular Biology, Max Planck Institute of Immunobiology and Epigenetics, 79108 Freiburg, Germany; gDepartment of Infectious Diseases, Section for Clinical Virology, University of Gothenburg, S-413 46 Gothenburg, Sweden; hLaboratoire d’Ingénierie des Systèmes Macromoléculaires, UMR7255 Centre National de la Recherche Scientifique and Aix Marseille University, Institut de Microbiologie de la Méditerranée, 13402 Marseille, France; and iLaboratory for Systems and Synthetic Biology, Wageningen University, 6703 HB, Wageningen, The Netherlands

Edited by Kai Simons, Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany, and approved July 23, 2014 (received for review February 11, 2014)

membrane curvature

identified as important molecules contributing to host specificity and adhesion of P. aeruginosa (19, 20). Recent observations suggest that GSLs might be of critical importance for the internalization of P. aeruginosa into nonphagocytic cells (9). The homotetrameric, galactophilic lectin LecA, which is localized to the outer bacterial membrane (21), belongs to the carbohydrate binding proteins expressed by P. aeruginosa that recognize GSLs (22, 23) and represents one of the virulence factors (24). The preferential binding of LecA to the GSL globotriaosylceramide (also known as Gb3/CD77 or Pk histo-blood group antigen) (22, 25) prompted us to investigate the role of LecA–Gb3 interaction in the cellular uptake of P. aeruginosa. In our study, we demonstrated that GSLs have a major impact on the bending of the host plasma membrane and engulfment of the pathogen. Binding of GSL by P. aeruginosa induced negative membrane curvature on synthetic, giant unilamellar vesicles (GUVs), solely driven by LecA–Gb3 interaction. This interaction resulted in a bacterium tightly engulfed by the lipid bilayer of the GUV. Because the glycolipid component plays the major role in this process, we termed this mechanism “lipid zipper” as a novel zipper-type mode of bacterial entry (26). Further experiments revealed that the lipid zipper has a significant impact on P. aeruginosa uptake in several epithelial cell lines. Interestingly, actin Significance Entry of bacteria into host cells critically depends on their proper engulfment by the plasma membrane. So far, actin polymerization has been described as a major driving force in this process. However, our study reveals that the interaction of the bacterial surface lectin LecA with the host cell glycosphingolipid Gb3 is fully sufficient to promote engulfment of Pseudomonas aeruginosa, whereas actin polymerization is dispensable. Hence, the formation of a “lipid zipper” represents a previously unidentified mechanism of bacterial uptake and demonstrates that bacterial pathogens have evolved lipidbased invasion strategies that may function in addition to protein receptor-based ones. Furthermore, by identifying the LecA/Gb3 interaction as the major invasion-promoting factor, our study provides new targets for drugs that may prevent bacterial invasion and dissemination.

| glycolipid | infection

G

lycosphingolipids (GSLs) are essential components of biological membranes that have significant impact on the organization and the molecular architecture of the membrane (1, 2). Preferential association of GSLs with cholesterol, various other lipid species, and proteins induces formation of micro- and nanodomains. Such domains, which are also termed “lipid rafts,” are involved in the signal transduction across the membrane (3). Several invasive bacterial pathogens hijack these membrane domains as signaling platforms for actin polymerization, which leads to the engulfment of the bacterium by the host plasma membrane (PM) (4–6). For example, Listeria monocytogenes exploits lipid raftdependent E-cadherin and HGF-R/Met signaling in the host cell to trigger actin polymerization for bacterial uptake (7). The opportunistic bacterium Pseudomonas aeruginosa can also invade and survive in diverse epithelial and endothelial cells (8–14), which significantly contributes to its pathogenicity (8, 10, 15). Numerous reports suggested various host cell factors, including cell surface receptors and signaling components, including α5β1 integrin, cystic fibrosis transmembrane conductance regulator, and Abelson tyrosine-protein kinase 1-dependent pathway (16–18), that are involved in the cellular uptake of P. aeruginosa. Host-cell GSLs have been www.pnas.org/cgi/doi/10.1073/pnas.1402637111

Author contributions: T.E., B.B., A.I., C.F., and W.R. designed research; T.E., B.B., J.M., G.E.R., S.M., C.F., and W.R. performed research; R.T., A.A., S.A., S.J., S.d.B., and A.I. contributed new reagents/analytic tools; T.E., B.B., C.F., and W.R. analyzed data; and T.E., C.F., and W.R. wrote the paper. The authors declare no conflict of interest. This article is a PNAS Direct Submission. 1

To whom correspondence may be addressed. Email: [email protected], [email protected], or [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1402637111/-/DCSupplemental.

PNAS | September 2, 2014 | vol. 111 | no. 35 | 12895–12900

MICROBIOLOGY

Glycosphingolipids are important structural constituents of cellular membranes. They are involved in the formation of nanodomains (“lipid rafts”), which serve as important signaling platforms. Invasive bacterial pathogens exploit these signaling domains to trigger actin polymerization for the bending of the plasma membrane and the engulfment of the bacterium—a key process in bacterial uptake. However, it is unknown whether glycosphingolipids directly take part in the membrane invagination process. Here, we demonstrate that a “lipid zipper,” which is formed by the interaction between the bacterial surface lectin LecA and its cellular receptor, the glycosphingolipid Gb3, triggers plasma membrane bending during host cell invasion of the bacterium Pseudomonas aeruginosa. In vitro experiments with Gb3-containing giant unilamellar vesicles revealed that LecA/Gb3-mediated lipid zippering was sufficient to achieve complete membrane engulfment of the bacterium. In addition, theoretical modeling elucidated that the adhesion energy of the LecA–Gb3 interaction is adequate to drive the engulfment process. In cellulo experiments demonstrated that inhibition of the LecA/Gb3 lipid zipper by either lecA knockout, Gb3 depletion, or application of soluble sugars that interfere with LecA binding to Gb3 significantly lowered P. aeruginosa uptake by host cells. Of note, membrane engulfment of P. aeruginosa occurred independently of actin polymerization, thus corroborating that lipid zippering alone is sufficient for this crucial first step of bacterial hostcell entry. Our study sheds new light on the impact of glycosphingolipids in the cellular invasion of bacterial pathogens and provides a mechanistic explication of the initial uptake processes.

polymerization, which was previously identified as a major driving force for membrane deformation and engulfment, was not required for the lipid-triggered process. Moreover, ectopic expression of LecA in Escherichia coli was sufficient to greatly increase its invasion efficiency in Gb3-positive cells. Results A LecA–Gb3 Lipid Zipper Is Sufficient for Membrane Engulfment of P. aeruginosa in a Synthetic Lipid Bilayer System. We used Gb3-

containing GUVs as a membrane model system to directly address the impact of LecA–Gb3 interactions on the curvature of a lipid bilayer without any potential interference of cellular factors, particularly actin. We incubated GUVs with the invasive P. aeruginosa PAO1 WT strain and a lecA deletion mutant (ΔlecA) that did not express LecA (Fig. S1, Inset). Both strains were chromosomally GFP-tagged and did not differ in their growth rates (Fig. S1). PAO1 WT not only bound to Gb3-containing GUVs but also induced a highly curved GUV membrane at the contact point, which finally led to a complete engulfment of the WT bacteria by the GUV membrane (Fig. 1A and Fig. S2 A–C). Remarkably, we also observed membrane-engulfed bacteria that were connected via tether-like structures to the surrounding GUV membrane (Fig. S2 A, arrowhead and B). Interestingly, we observed clustering of lipid material at sites where bacteria attached to the GUVs and were engulfed (Fig. 1A). In total, 82 ± 6.5% of the analyzed GUVs were bound by PAO1 WT and of these 45 ± 6.4% showed bacteria, which were engulfed by

the GUV membrane (Fig. 1B). In contrast, the ΔlecA mutant essentially did not induce membrane invaginations (Fig. S2 D–F). Although still 22 ± 7.7% of the GUVs were bound by the ΔlecA mutant (Fig. 1B), only 1 of 102 GUVs analyzed in total showed a membrane-engulfed bacterium. Lowering of the Gb3 concentration in GUVs from 5 to 0 mol% led to a dose-dependent decrease of GUVs containing membrane-engulfed bacteria (Fig. 1C), clearly indicating a crucial role of Gb3 in the process of bacterial membrane engulfment. Without Gb3, only 1 of the 106 GUVs analyzed showed a wrapped bacterium. In the next step we addressed the impact of the mechanical properties of the GUV membrane. When decreasing the surface tension by applying a hyperosmolar buffer solution from the outside (550 mOsm·L−1 outside vs. 290 mOsm·L−1 inside the GUVs), GUVs containing 0.1 mol% Gb3 showed a number of membrane-wrapped bacteria comparable to that obtained with 5 mol% Gb3 GUVs at isoosmolar conditions (Fig. S3 and Fig. 1C). Cholesterol plays a pivotal role in membrane physiology because it is implicated in the generation of lipid raft domains and can influence membrane rigidity, which might affect the formation of the lipid zipper. Therefore, we compared the efficiency of lipid zipper formation on GUVs in the presence or absence of cholesterol. The efficiency was reduced by 65% in the absence of cholesterol (45 ± 6.4% vs.15.7 ± 4.8%), which, apparently, was not due to a decrease in binding of bacteria to GUVs, which was just 10% less (70 ± 4.2%) compared with cholesterol-containing GUVs (81.7 ± 5.4%) (Fig. 1D). In summary, these data provide clear evidence that binding of LecA to Gb3 mediates a cholesterol-dependent, zipperlike engulfment of bacteria by the GUV membrane. The LecA–Gb3 Interaction Provides Enough Energy for Complete Engulfment of P. aeruginosa by a Lipid Membrane. We developed

Fig. 1. P. aeruginosa induces a lipid zipper on giant unilamellar vesicles by binding of LecA to Gb3. (A) GUVs reconstituted with Texas Red-labeled 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine (DHPE-TR) and Gb3 were inoculated with GFP-tagged P. aeruginosa PAO1 WT (PAO1-GFP) for 15 min at 37 °C. Lipid clustering at the bacterial attachment site is visible (arrowhead). (Scale bar, 5 μm.) (B) Relative number of GUVs containing bound or membrane-wrapped P. aeruginosa PAO1 WT or PAO1 ΔlecA. Bound GUVs were normalized to the total number of GUVs analyzed and invaginated GUVs were normalized to GUVs bound by bacteria. Bars represent mean values ± SEM of n = 4 independent experiments with ≥100 GUVs analyzed in total. (C) Relative number of GUVs with indicated Gb3 concentrations containing membrane-engulfed P. aeruginosa PAO1 WT. Bars represent mean values ± SEM of n ≥ 3 independent experiments. For 0 mol% Gb3, 106 GUVs were analyzed; for 0.01 mol% Gb3, 71 GUVs; for 0.1 mol% Gb3, 105 GUVs; for 1 mol%, 54 GUVs; and for 5 mol% Gb3, 100 GUVs. (D) Relative number of GUVs with bound or invaginated bacteria as normalized in B with lipid mixtures containing 5 mol% Gb3 with or without 30 mol% cholesterol. Bars represent mean values ± SEM of n ≥ 3 independent experiments with ≥100 GUVs analyzed in total.

12896 | www.pnas.org/cgi/doi/10.1073/pnas.1402637111

a theoretical model to investigate whether the LecA–Gb3 interaction is sufficient to mediate membrane engulfment of bacteria. The GUV was described as an elastic surface consisting of mobile lipid constituents, and P. aeruginosa was modeled as a LecA-covered, rigid rod (Fig. 2A and SI Materials and Methods). The total energy of the system consists of three terms: (i) mechanical Helfrich energy (27), which favors flat membrane configurations; (ii) curvature-promoting adhesion energy owing to LecA–Gb3 interaction; and (iii) mixing entropy (28) of mobile lipids. The final configuration of the bacterium relative to the GUV membrane is the result of minimizing the total energy and depends solely on the physical properties of the system, that is, size of the bacterium, bending rigidity, and surface tension of the GUV membrane, as well as Gb3 density. In Fig. 2B we show the wrapping height h of a bacterium (L = 2.5 μm) in dependence of surface tension γ and Gb3 density σ for constant bending rigidity (SI Materials and Methods). For realistic values of GUV Gb3 densities (σ ≈ 105 μm−2), the model predicts a complete membrane engulfment of the bacterium. Similar effects are predicted for the host-cell PMs, for which Gb3 densities in the range of σ = 1,400−7,400 μm−2 (SI Materials and Methods) and surface tensions between 10 μJ·m−2 and 300 μJ·m−2 (29–31) are realistic. Furthermore, the model reveals that LecA-dependent clustering of mobile Gb3 lipids is crucial. When Gb3 is kept immobile, the global Gb3 concentration required to achieve full engulfment increased by two orders of magnitude (Fig. 2C; compare curves in color and black for mobile and immobile Gb3, respectively). Full engulfment occurred at local Gb3 densities well below critical values (i.e., close packing of Gb3) (compare Fig. 2C and Fig. S4C). To further explore the dependence of the wrapping on the global Gb3 density and the vesicle size, we incubated GUVs prepared from lipid mixtures containing different Gb3 concentrations (Materials and Methods) with P. aeruginosa PAO1 WT. By microscopic analysis we measured the equatorial diameter of GUVs in a random field of view. On the basis of our model Eierhoff et al.

wrapping (Materials and Methods). According to the model the critical vesicle size should decrease with decreasing surface tension (compare in Fig. 2D the solid line for γ = 800 μJ·m−2 with the dashed line for γ = 600 μJ·m−2). This should lead to an increased fraction of bacteria-wrapped GUV in the low-diameter range. The experimental results for GUVs with a diameter of ≤20 μm show that the fraction of bacteria-wrapped GUVs is increased by 25% at higher osmolarity compared with iso-osmolar conditions (compare in Fig. 2E GUVs with 0.1 mol% Gb3 at 290 vs. 550 mOsm·L−1). Taken together, our theoretical model shows that for typical Gb3 densities and surface tensions of GUVs and cells the energy provided by LecA–Gb3 binding and clustering is sufficient to induce total membrane engulfment of bacteria-sized objects. For engulfment of smaller particles such as viruses additional mechanisms were required (SI Materials and Methods), which is in agreement with previous experimental and theoretical results (32, 33).

Fig. 2. Gb3 binding and clustering predicts complete membrane engulfment of P. aeruginosa. (A) Rotationally symmetric geometry of a modeled rodshaped bacterium engulfed by GUV membrane. (B) Wrapping height in dependence of surface tension and Gb3 surface density (GUV diameter d = 16 μm). (C) Invagination depth and local Gb3 density as function of the global Gb3 density and different GUV diameters for intermediate surface tension (γ = 10−4 J·m−2). Clustering of mobile Gb3 reduces the global Gb3 concentration necessary for full wrapping (compare colored curves for mobile vs. black curve for immobile Gb3). Shaded area corresponds to Gb3 densities in the PM of host cells. (D) Critical GUV sizes. Circles represent observed vesicles of different size and with different Gb3 content (they are scattered around x-values for improved visibility). Each wrapping event is denoted by red padding. Three lines represent the critical GUV size for full enclosing of bacteria at likely surface tensions γ. The shaded area shows the effect of uncertainty in effective Gb3 content in a single vesicle at γ = 800 μJ·m−2 (SI Materials and Methods). For the right column, the osmolarity of the outside buffer was increased from 290 to 550 mOsm·L−1, whereas the inside buffer was kept at 290 mOsm·L−1; the thereby reduced surface tension enhances wrapping at moderate GUV sizes. The theoretical curves show qualitative agreement with the observed statistics.

(see above), we calculated the minimal vesicle size for which the bacteria would be fully wrapped (h = 2.5 μm) for the global Gb3 density σ on the vesicle at the applied Gb3 concentrations. Fig. 2D compares the experimental results with the theoretical prediction for three different surface tensions (solid line, γ = 800 μJ·m−2). The shaded area in Fig. 2D denotes the uncertainty range for the minimal vesicle size assuming that the bacterium is at least wrapped up to the spherical cap (h = 2.25 μm) and a 40% variation in the Gb3 content (SI Materials and Methods). Additionally, we experimentally tested our model by variation of the surface tension of GUVs as a critical parameter for bacterial Eierhoff et al.

a lipid zipper mechanism is implicated in bacterial invasion of living host cells. We first verified that cell entry of P. aeruginosa into our cell model, the human lung epithelial cell line H1299, indeed depends on GSL expression, as previously reported (9). Quantification of the cellular invasion of P. aeruginosa showed an 80% reduction of invasion upon inhibition of GSL synthesis by D-threo-l-phenyl-2-palmitoylarmino-3-morpholino-l-propanol (PPMP) treatment, whereas bacterial attachment was not affected (Fig. S5A). Disappearance of Gb3 expression was verified by a loss of Shiga toxin B-subunit (StxB) binding to cells (Fig. S5B). In the following we studied the entry process of P. aeruginosa into H1299 cells by confocal fluorescence microscopy. Shortly after binding to H1299 cells, bacteria induced negative PM curvature at the cellular entry site (Fig. S6A) that progressed into the engulfment of the bacterium by PM (Fig. 3A, arrowheads). For the ΔlecA mutant, the number of PM-surrounded bacteria was significantly lower (about 64.5%) at 1 h postinfection compared with the WT strain (Fig. 3B), suggesting that mainly LecA contributes to the zipper-like engulfment of bacteria by PM. In line with the GUV data, extraction of cholesterol by methyl-β-cyclodextrin treatment lowered the occurrence of zipper-like, PM-engulfed bacteria by about 50% (Fig. S7) and thereby to the same extent as observed for GUVs, with and without cholesterol (Fig. 1D). Interestingly, actin polymerization does not play a role for the PM engulfment of the bacteria, because actin-related protein 2/3 complex (Arp2) depletion (Fig. 3D and Fig. S6 B and D) (34) or latrunculin A treatment (Fig. S6E) (35) did not prevent initial envelopment of bacteria by PM. Nevertheless, accumulation of actin, as well as of the PM marker GPI-mCherry, was frequently seen close to invading bacteria even in Arp2-depleted cells (Fig. 3D). However, actin coverage seems not to be functionally linked to the formation of membrane invaginations, because we observed actin-covered, PMenveloped bacteria under all perturbations (Fig. 3D and Fig. S6E, arrowheads). Most likely, actin coverage around invading bacteria represents persistent, cortical actin filaments. To further test whether actin polymerization is redundant for the zipper process, we studied the interactions of WT bacteria with plasma membrane spheres (PMS) induced in H1299 cells that stably expressed the actin marker LifeAct-mCherry. In PMS, no endocytic machinery is present and (cortical) actin is excluded (36). Intriguingly, we observed bacteria that were tightly engulfed by the membrane of PMS, which demonstrates membrane wrapping even in the absence of actin (Fig. S8). These PM-enveloped bacteria closely resembled those wrapped by GUV membranes (Fig. S2 B and C). Therefore, we propose a mechanistic model in which actin polymerization is dispensable for membrane engulfment but required for completion of the invasion process and PNAS | September 2, 2014 | vol. 111 | no. 35 | 12897

MICROBIOLOGY

LecA Promotes a Lipid Zipper on Epithelial Cells Independent of Actin Polymerization. In further experiments, we directly assessed whether

PAO1 WT (Fig. 4A, “untreated”). Next, we depleted PM Gb3 levels by StxB-induced endocytosis of Gb3 (32) (Fig. S9 A and B). StxBmediated depletion of Gb3 reduced the invasion of the WT strain into H1299 cells by 69.0% (Fig. 4A, “StxB-treated”) compared with untreated cells. No significant difference in invasion efficiency was observed with the ΔlecA strain on Gb3-depleted (“StxB-treated”) and Gb3-containing (“untreated”) H1299 cells. Additionally, we confirmed by ectopic expression of Gb3 in MDCKII cells that the invasion of PAO1 WT correlated with the amount of PM-localized Gb3 (Fig. 4B). Moreover, pretreatment of the bacteria with the LecA-binder para-nitrophenyl-α-D-galactopyranoside (PNPG) decreased the invasiveness of PAO1 WT by 68.6% (Fig. 4A, “PNPGtreated”). PNPG selectively prevented LecA, but not StxB, binding to cells (Fig. S9 C and D). In contrast, cellular invasion by the ΔlecA strain was not affected by PNPG, because the invasion efficiency was at a similar level compared with the untreated ΔlecA strain. None of the tested conditions significantly affected the adhesion of P. aeruginosa PAO1 WT or ΔlecA strains to host cells (Fig. 4C). These findings strongly suggest a major role of LecA and Gb3 as internalization factors rather than adhesion factors. Moreover, the cholesterol dependency of the lipid zipper formation in GUVs and

Fig. 3. P. aeruginosa induces plasma membrane invaginations depending on LecA. (A) PAO1 (green)-induced membrane invaginations (red) colocalize with actin (blue) in H1299 cells during cellular entry. The white squared area is shown enlarged in the lower panel. (Scale bars, 5 μm and 2 μm, respectively.) (B) Relative numbers of PM-engulfed PAO1 WT and ΔlecA during cell entry stage as shown in A, n = 3 independent experiments; mean ± SEM P value was calculated by two-tailed, paired t test. (C) Intensity profile across a membrane invagination in A (yellow line). (D) Actin polymerization is not required for PAO1-induced membrane invaginations. Bacteria still induce membrane invaginations (arrowheads) and are surrounded by host PM, either covered (zoom 1) or uncovered (zoom 2, recorded 0.5 μm from above the focal plane shown in the overview) by actin in Arp2 knockdown cells (see Fig. S6D for control siRNA-transfected cells). The white squared areas are shown enlarged in the lower panels. (Scale bars, 10 μm and 2.5 μm, respectively.)

closing of the invaginated membrane cup to form an intracellular bacteria-containing vesicle. This hypothesis is supported by the fact that P. aeruginosa requires actin polymerization for complete entry into lung cells, as indicated by the reduced invasion into Arp2-knockdown cells (Fig. S6C). LecA/Gb3-Mediated Lipid Zipper Contributes Measurable Impact on the Cellular Invasion of P. aeruginosa. Because P. aeruginosa is capa-

ble of inducing a Gb3–LecA lipid zipper on GUVs (Fig. 1) and cells (Fig. 3) we quantified how the lipid zipper affects bacterial host-cell invasion efficiency. The P. aeruginosa PAO1 ΔlecA mutant strain showed a 61.3% reduced invasion into H1299 cells compared with 12898 | www.pnas.org/cgi/doi/10.1073/pnas.1402637111

Fig. 4. LecA in conjunction with Gb3 promotes efficient cellular invasion of P. aeruginosa. (A) Invasion of P. aeruginosa strains PAO1 WT and PAO1 ΔlecA into untreated or Gb3-depleted (StxB-treated) H1299 cells, and with LecA inhibition (PNPG-treated). Data represent mean values of n ≥ 3 independent experiments. The P value for WT vs. ΔlecA was calculated by a two-tailed, paired t test. (B) Ectopic expression of Gb3 in MDCKII cells enhances uptake of P. aeruginosa. Invasion of P. aeruginosa WT into untransfected MDCKII cells (wt) and Gb3-synthase-transfected MDCKII cells (clone 1 and clone 2) was evaluated. Both clones differ in their Gb3 expression, which was analyzed by FACS after incubation with StxB-Alexa488 (hatched bars). All data represent mean values ± SEM for n ≥ 3 independent experiments with normalized StxB fluorescence and invasion, to MDCKII clone 2. The P values were calculated by a two-tailed, paired t test. (C) Adhesion of PAO1 WT and ΔlecA to H1299 cells according to the treatments as in A. All data represent mean values ± SEM of n ≥ 4 independent experiments, normalized to the WT. (D, Upper) Induced expression of lecA by IPTG in E. coli BL21 (DE3) pET25pa1l detected by standard Western blot analysis. The parental, WT strain [BL21 (DE3)] serves as control (P, bacterial pellet; S, culture supernatant). (D, Lower) Invasion and adhesion of indicated E. coli strains to H1299 cells. Invasion upon IPTG-induced expression of lecA is strongly enhanced. Data represent mean values of normalized invasion ± SEM for n ≥ 4 independent experiments. For better visibility, adhesion was normalized to non-IPTG-treated E. coli BL21 (DE3).

Eierhoff et al.

Discussion By studying the interaction of P. aeruginosa with synthetic and cellular membranes we have identified a mechanism for bacterial entry that requires host GSLs and bacterial lectins but does not depend on actin polymerization for membrane engulfment. The lipid zipper highlights the particular importance of glycolipids in bacterial invasion processes. Numerous host-cell factors have been described to interact with P. aeruginosa to orchestrate cellular entry and infection of the bacterium. Therefore, it is intriguing that only two interacting factors, the host-cell GSL Gb3 and the bacterial, GSL-binding lectin LecA are sufficient to induce the initial processes of the cellular entry of P. aeruginosa. GSLs are key components of lipid rafts and as such are involved in establishing signal transduction platforms in biological membranes (1, 37). Cholesterol has been described to cluster GSLs in the external leaflet of the PM and thereby to stabilize lipid raft domains. We show that the lipid zipper mechanism depends on lipid raft components such as cholesterol and the GSL Gb3, whereas a nonraft lipid, 1,2-dioleoyl-sn-glycero-3phosphocholine (DOPC), does not directly promote the induction of the lipid zipper. Our results suggest that cholesterol stabilizes LecA-induced domains with a sufficiently high Gb3 density to trigger the lipid zipper, and finally the efficient cellular invasion. Therefore, our findings expand the lipid raft concept because (i) they introduce an autonomous role for GSLs (i.e., Gb3) in triggering an endocytic event, uncoupled from cytoplasmic signaling processes, and (ii) consider raft domains to serve as membrane areas providing sufficiently high GSL densities for lipid zippering bacterial pathogens. The effect of cholesterol on membrane stiffness is complex. Membrane stiffness is not necessarily decreased in cholesteroldepleted membranes. It could even be increased, at least in cholesterol-depleted cells (38). We found that cholesterol extraction on cells decreased the frequency of the lipid zipper, which could, in line with ref. 38, point to an increase in membrane stiffness, impairing the lipid zipper. Following a study on DOPC vesicles, cholesterol has no (significant) effect on bending rigidity (39). However, we found a significantly lower occurrence of the lipid zipper in cholesterol-free GUVs. Interpreting our results in the context of the above-mentioned report (39), cholesterol might affect the lipid zipper on GUVs by mechanisms other than membrane stiffness. Studies of SV40 entry engaging the GSL GM1 for cellular entry (33) revealed that the chemical structure of the GSL (i.e., fatty acid chain length and saturation) is critical for the uptake. Therefore, it will be interesting to investigate in future studies whether and how different Gb3 species affect the uptake of P. aeruginosa. Eierhoff et al.

It is well established that the cellular uptake of Gram-negative bacterial pathogens (e.g., Salmonella and Shigella species) critically depends on secreted bacterial effector proteins that induce a reorganization of the host cytoskeleton (40, 41). The expression of a type-III secretion system (T3SS) by P. aeruginosa and T3SS- and T6SS-secreted effector proteins have been reported to be required for the cellular invasion and are supposed to influence the growthphase dependent cellular uptake of P. aeruginosa (11, 42–45). However, our data suggest that initial membrane curvature and wrapping of the bacterium by the host PM depends neither on a bacterial secretion system nor on secreted effector proteins. Furthermore, actin polymerization and an active endocytic machinery are not required for the formation of PM invaginations, as suggested by actin inhibition experiments. The GUV experiments and model calculations clearly demonstrate that the induction of membrane invaginations by P. aeruginosa through LecA–Gb3 interactions is a thermodynamically favored, ATP-independent process that does not depend on cytosolic factors or on bacterial effectors. However, other bacterial and cellular effector proteins might be necessary to accelerate and/or finalize the uptake. In analogy to a previous report indicating that actin dynamics drive membrane scission on StxB-induced membrane invaginations (46), actin could be actively involved in late stages of endocytic processing of PM invaginations. In addition, because P. aeruginosa induces membrane curvature, curvature-sensing Bin Amphiphysin Rvs. (BAR) domain-containing proteins (47) might be recruited to the entry site and activate signaling pathways to promote the cellular uptake of P. aeruginosa. Moreover, our work demonstrates a previously unidentified pathophysiological relevance for LecA. So far, LecA has been related to bacterial adhesion and biofilm formation (20, 48) and represents one of the virulence factors contributing to lung injury during infection (24). Our data identify LecA as an invasion factor for P. aeruginosa, which might contribute to the dissemination of the pathogen within its host organism during infection. This is strongly supported by the finding that ectopic lecA expression induced an invasive phenotype of an originally noninvasive E. coli strain. Moreover, this finding shows that LecA does not need to interact with specific P. aeruginosa factors to fulfill its invasive function. Inhibition of the LecA/Gb3-dependent pathway reduced hostcell invasion by P. aeruginosa. Therefore, strategies to interfere with the binding of LecA to Gb3 could be the basis for new drugs that efficiently prevent the dissemination and persistence of P. aeruginosa. Moreover, many bacterial pathogens express lectins on their cellular surface that recognize host GSLs (49). In the context of our findings, it is possible that these pathogens use GSL-dependent mechanisms to trigger their uptake by host cells, alternatively or synergistically to protein receptor-based entry pathways, such as the invasin-induced entry of Yersinia (50) or the internalin-triggered pathway of Listeria spp. (51, 52). Clearly, the lipid zipper mechanism that drives bacterial invasion shares some common features with the GSL-mediated endocytosis of toxins (32) and viruses (33, 53). Multivalent lectin– glycosphingolipid interactions trigger the formation of plasma membrane invaginations in the absence of cytosolic proteins (e.g., coat proteins and actin). However, for bacteria, in contrast to smaller particles such as toxins and viruses, the adhesion energy resulting from lectin-induced GSL clustering is already sufficient to overcome the energy penalty for membrane bending as an initial step for their cellular uptake. Materials and Methods Detailed materials, methods, and the theoretical model are described in SI Materials and Methods. Briefly, GUVs were prepared as described (54). H1299 and MDCKII cells stably expressing marker proteins or Gb3 synthase were cultured in RPMI or DMEM supplemented with Geneticin and 10% (vol/vol) (5% vol/vol) FBS. Recombinant LecA-, lecA-, and non-lecA-expressing E. coli BL21 (DE3) were produced as reported (25). GFP tags and LecA mutants of

PNAS | September 2, 2014 | vol. 111 | no. 35 | 12899

MICROBIOLOGY

cells correlated with the efficiency of cellular invasion by PAO1 WT. Invasion in cholesterol-depleted cells was significantly decreased by about 70% compared with control cells (Fig. S10). This suggests an overall critical role of cholesterol for the lipid zipper in vitro and in cellulo. To test the capacity of LecA as an invasion factor, we explored the invasiveness of a noninvasive E. coli strain upon induced expression of lecA by isopropyl β-D-1-thiogalactopyranoside (IPTG) (25). Remarkably, ectopic expression of lecA in E. coli BL21 (DE3) transduced this strain into an invasive one, evidenced by a significant increase of invasion into H1299 cells of about 340% compared with non-lecA-expressing E. coli strains (Fig. 4D). Interestingly, adherence of lecA-expressing E. coli to H1299 cells was drastically reduced (approximately 20-fold) compared with the non-lecA-expressing E. coli strains (Fig. 4D). Probably LecA functionally interferes with endogenous adherence factors of E. coli, which consequently prevents efficient adhesion to H1299 cells. However, these findings highlight the dominant, invasin-like function of LecA. In summary, these data demonstrate that Gb3 and LecA represent key factors for the cellular entry of P. aeruginosa into nonphagocytic cells.

P. aeruginosa PAO1 were produced as described in SI Materials and Methods. Invasion assays were performed as essentially described (11). GSL synthesis was inhibited by incubation of cells for 3 d with 5 μM PPMP. For LecA inhibition, soluble LecA (2 μg·mL−1) or bacteria were incubated for 15 min at 37 °C with 10 mM PNPG. Gb3 and cholesterol was depleted using 5 μg·mL−1 Cy3-labled StxB and 10 mM methyl-β-cyclodextrin, respectively. Actin polymerization was inhibited by latrunculin A treatment (0.1 μM) or knockdown of Arp2 by siRNA. PMS were induced as described (36) and visualized by FM464 dye. Cells, GUVs, and P. aeruginosa infection were imaged on a confocal microscope (Nikon Eclipse Ti-E with A1R confocal laser scanner, 60× oil objective, N.A. 1.49). Statistical testing was performed with Excel with data

of n ≥ 3 independent experiments using two-tailed, unpaired t test, unless stated otherwise.

1. Simons K, Ikonen E (1997) Functional rafts in cell membranes. Nature 387(6633):569–572. 2. Lingwood CA (2011) Glycosphingolipid functions. Cold Spring Harb Perspect Biol 3(7): a011874. 3. Lingwood D, Simons K (2010) Lipid rafts as a membrane-organizing principle. Science 327(5961):46–50. 4. Riethmüller J, Riehle A, Grassmé H, Gulbins E (2006) Membrane rafts in host-pathogen interactions. Biochim Biophys Acta 1758(12):2139–2147. 5. Lafont F, Abrami L, van der Goot FG (2004) Bacterial subversion of lipid rafts. Curr Opin Microbiol 7(1):4–10. 6. Pizarro-Cerdá J, Cossart P (2006) Bacterial adhesion and entry into host cells. Cell 124(4):715–727. 7. Seveau S, Bierne H, Giroux S, Prévost MC, Cossart P (2004) Role of lipid rafts in E-cadherin— and HGF-R/Met—mediated entry of Listeria monocytogenes into host cells. J Cell Biol 166(5):743–753. 8. Chi E, Mehl T, Nunn D, Lory S (1991) Interaction of Pseudomonas aeruginosa with A549 pneumocyte cells. Infect Immun 59(3):822–828. 9. Emam A, Carter WG, Lingwood C (2010) Glycolipid-dependent, protease sensitive internalization of Pseudomonas aeruginosa into cultured human respiratory epithelial cells. Open Microbiol J 4:106–115. 10. Fleiszig SM, Zaidi TS, Fletcher EL, Preston MJ, Pier GB (1994) Pseudomonas aeruginosa invades corneal epithelial cells during experimental infection. Infect Immun 62(8): 3485–3493. 11. Ha U, Jin S (2001) Growth phase-dependent invasion of Pseudomonas aeruginosa and its survival within HeLa cells. Infect Immun 69(7):4398–4406. 12. Pereira SH, Cervante MP, Bentzmann S, Plotkowski MC (1997) Pseudomonas aeruginosa entry into Caco-2 cells is enhanced in repairing wounded monolayers. Microb Pathog 23(4):249–255. 13. Plotkowski MC, Meirelles MN (1997) Concomitant endosome-phagosome fusion and lysis of endosomal membranes account for Pseudomonas aeruginosa survival in human endothelial cells. J Submicrosc Cytol Pathol 29(2):229–237. 14. Plotkowski MC, Saliba AM, Pereira SH, Cervante MP, Bajolet-Laudinat O (1994) Pseudomonas aeruginosa selective adherence to and entry into human endothelial cells. Infect Immun 62(12):5456–5463. 15. Fleiszig SM, et al. (1996) Relationship between cytotoxicity and corneal epithelial cell invasion by clinical isolates of Pseudomonas aeruginosa. Infect Immun 64(6):2288–2294. 16. Roger P, et al. (1999) Fibronectin and alpha5beta1 integrin mediate binding of Pseudomonas aeruginosa to repairing airway epithelium. Eur Respir J 13(6):1301–1309. 17. Pier GB, Grout M, Zaidi TS (1997) Cystic fibrosis transmembrane conductance regulator is an epithelial cell receptor for clearance of Pseudomonas aeruginosa from the lung. Proc Natl Acad Sci USA 94(22):12088–12093. 18. Pielage JF, Powell KR, Kalman D, Engel JN (2008) RNAi screen reveals an Abl kinasedependent host cell pathway involved in Pseudomonas aeruginosa internalization. PLoS Pathog 4(3):e1000031. 19. de Bentzmann S, et al. (1996) Asialo GM1 is a receptor for Pseudomonas aeruginosa adherence to regenerating respiratory epithelial cells. Infect Immun 64(5):1582–1588. 20. Gilboa-Garber N, Sudakevitz D, Sheffi M, Sela R, Levene C (1994) PA-I and PA-II lectin interactions with the ABO(H) and P blood group glycosphingolipid antigens may contribute to the broad spectrum adherence of Pseudomonas aeruginosa to human tissues in secondary infections. Glycoconj J 11(5):414–417. 21. Glick J, Garber N (1983) The intracellular localization of Pseudumonas aeruginosa lectins. J Gen Microbiol 129:3085–3090. 22. Lanne B, Cîopraga J, Bergström J, Motas C, Karlsson KA (1994) Binding of the galactose-specific Pseudomonas aeruginosa lectin, PA-I, to glycosphingolipids and other glycoconjugates. Glycoconj J 11(4):292–298. 23. Gilboa-Garber N (1972) Inhibition of broad spectrum hemagglutinin from Pseudomonas aeruginosa by D-galactose and its derivatives. FEBS Lett 20(2):242–244. 24. Chemani C, et al. (2009) Role of LecA and LecB lectins in Pseudomonas aeruginosa-induced lung injury and effect of carbohydrate ligands. Infect Immun 77(5):2065–2075. 25. Blanchard B, et al. (2008) Structural basis of the preferential binding for globo-series glycosphingolipids displayed by Pseudomonas aeruginosa lectin I. J Mol Biol 383(4): 837–853. 26. Cossart P, Sansonetti PJ (2004) Bacterial invasion: The paradigms of enteroinvasive pathogens. Science 304(5668):242–248. 27. Helfrich W (1973) Elastic properties of lipid bilayers: Theory and possible experiments. Z Naturforsch C 28(11):693–703.

28. Reichl L (2009) A Modern Course in Statistical Physics (Wiley-VCH, Weinheim, Germany). 29. Lieber AD, Yehudai-Resheff S, Barnhart EL, Theriot JA, Keren K (2013) Membrane tension in rapidly moving cells is determined by cytoskeletal forces. Curr Biol 23(15): 1409–1417. 30. Pietuch A, Brückner BR, Janshoff A (2013) Membrane tension homeostasis of epithelial cells through surface area regulation in response to osmotic stress. Biochim Biophys Acta 1833(3):712–722. 31. Dai J, Sheetz MP (1999) Membrane tether formation from blebbing cells. Biophys J 77(6):3363–3370. 32. Römer W, et al. (2007) Shiga toxin induces tubular membrane invaginations for its uptake into cells. Nature 450(7170):670–675. 33. Ewers H, et al. (2010) GM1 structure determines SV40-induced membrane invagination and infection. Nat Cell Biol 12(1):11–18; sup pp 11-12. 34. Pollard TD, Blanchoin L, Mullins RD (2000) Molecular mechanisms controlling actin filament dynamics in nonmuscle cells. Annu Rev Biophys Biomol Struct 29:545–576. 35. Coué M, Brenner SL, Spector I, Korn ED (1987) Inhibition of actin polymerization by latrunculin A. FEBS Lett 213(2):316–318. 36. Lingwood D, Ries J, Schwille P, Simons K (2008) Plasma membranes are poised for activation of raft phase coalescence at physiological temperature. Proc Natl Acad Sci USA 105(29):10005–10010. 37. Simons K, Toomre D (2000) Lipid rafts and signal transduction. Nat Rev Mol Cell Biol 1(1):31–39. 38. Byfield FJ, Aranda-Espinoza H, Romanenko VG, Rothblat GH, Levitan I (2004) Cholesterol depletion increases membrane stiffness of aortic endothelial cells. Biophys J 87(5):3336–3343. 39. Gracia RS, Bezlyepkina N, Knorr RL, Lipowsky R, Dimova R (2010) Effect of cholesterol on the rigidity of saturated and unsaturated membranes: Fluctuation and electrodeformation analysis of giant vesicles. Soft Matter 6(7):1472–1482. 40. McGhie EJ, Brawn LC, Hume PJ, Humphreys D, Koronakis V (2009) Salmonella takes control: Effector-driven manipulation of the host. Curr Opin Microbiol 12(1):117–124. 41. Ogawa M, Handa Y, Ashida H, Suzuki M, Sasakawa C (2008) The versatility of Shigella effectors. Nat Rev Microbiol 6(1):11–16. 42. Galán JE, Collmer A (1999) Type III secretion machines: Bacterial devices for protein delivery into host cells. Science 284(5418):1322–1328. 43. Jendrossek V, et al. (2003) Apoptotic response of Chang cells to infection with Pseudomonas aeruginosa strains PAK and PAO-I: Molecular ordering of the apoptosis signaling cascade and role of type IV pili. Infect Immun 71(5):2665–2673. 44. Barbieri JT, Sun J (2004) Pseudomonas aeruginosa ExoS and ExoT. Rev Physiol Biochem Pharmacol 152:79–92. 45. Sana TG, et al. (2012) The second type VI secretion system of Pseudomonas aeruginosa strain PAO1 is regulated by quorum sensing and Fur and modulates internalization in epithelial cells. J Biol Chem 287(32):27095–27105. 46. Römer W, et al. (2010) Actin dynamics drive membrane reorganization and scission in clathrin-independent endocytosis. Cell 140(4):540–553. 47. Peter BJ, et al. (2004) BAR domains as sensors of membrane curvature: The amphiphysin BAR structure. Science 303(5657):495–499. 48. Diggle SP, et al. (2006) The galactophilic lectin, LecA, contributes to biofilm development in Pseudomonas aeruginosa. Environ Microbiol 8(6):1095–1104. 49. Sharon N (2006) Carbohydrates as future anti-adhesion drugs for infectious diseases. Biochim Biophys Acta 1760(4):527–537. 50. Isberg RR, Leong JM (1990) Multiple beta 1 chain integrins are receptors for invasin, a protein that promotes bacterial penetration into mammalian cells. Cell 60(5):861–871. 51. Shen Y, Naujokas M, Park M, Ireton K (2000) InIB-dependent internalization of Listeria is mediated by the Met receptor tyrosine kinase. Cell 103(3):501–510. 52. Mengaud J, Ohayon H, Gounon P, Cossart P, Cossart P; Mege R-M (1996) E-cadherin is the receptor for internalin, a surface protein required for entry of L. monocytogenes into epithelial cells. Cell 84(6):923–932. 53. Rydell GE, Svensson L, Larson G, Johannes L, Römer W (2013) Human GII.4 norovirus VLP induces membrane invaginations on giant unilamellar vesicles containing secretor gene dependent α1,2-fucosylated glycosphingolipids. Biochim Biophys Acta 1828(8):1840–1845. 54. Mathivet L, Cribier S, Devaux PF (1996) Shape change and physical properties of giant phospholipid vesicles prepared in the presence of an AC electric field. Biophys J 70(3): 1112–1121.

12900 | www.pnas.org/cgi/doi/10.1073/pnas.1402637111

ACKNOWLEDGMENTS. W.R. acknowledges support by the Excellence Initiative of the German Research Foundation (EXC 294), the Ministry of Science, Research and the Arts of Baden-Württemberg (Az: 33-7532.20), a starting grant from the European Research Council (Programme “Ideas,” ERC-2011StG 282105), the Excellence Initiative of the German Research Foundation (GSC-4, Spemann Graduate School). A.I. and A.A. acknowledge support from Neolect (ANR-11-BSV5-002), COST action BM1003, and Labex ARCANE (ANR11-LABX-003). S.A. acknowledges support from the International Max Planck Research School for Molecular and Cellular Biology.

Eierhoff et al.

Supporting Information Eierhoff et al. 10.1073/pnas.1402637111 SI Materials and Methods Lectin A Incubation and Inhibition and Depletion of Glycosphingolipids.

Depletion of plasma membrane (PM)-localized globotriaosylceramide (Gb3) was achieved by 30-min preincubation of cells at 37 °C with 5 μg·mL−1 Cy3-labeled Shiga toxin B-subunit (StxB). For glycosphingolipid (GSL) depletion, cells were passaged for 3 d in the presence of 5 μM D-threo-l-phenyl-2-palmitoylarmino-3-morpholinol-propanol (PPMP; Santa Cruz), a substrate analog of the glucosylceramide synthase (1). For lectin LecA inhibition, purified, Alexa488-conjugated LecA (2 μg·mL−1) or bacteria were preincubated for 15 min at 37 °C with para-nitrophenyl-α-D-galactopyranoside (PNPG) (Sigma-Aldrich) at 10 mM final concentration and then incubated with cells for fluorescence microscopic analysis or invasion and adhesion assay, respectively. For confirmation of Gb3 depletion and LecA inhibition, cells were grown on coverslips incubated as described above with Cy3labeled StxB and PNPG, respectively. Afterward, cells were washed twice with Dulbecco’s phosphate-buffered saline (DPBS), incubated with 2 μg·mL−1 StxB-Alexa488 or LecA-Alexa488 for 30 min at 37 °C. After incubation all cells were washed, fixed with 4% paraformaldehyde and stained for actin by Phalloidin-ATTO647N (Sigma-Aldrich). Generation of Stable Cell Lines. The human lung epithelial cell line H1299 (American Type Culture Collection no. CRL-5803) was grown in RPMI medium, supplemented with 10% FCS and L-glutamine at 37 °C and 5% CO2. For cultivation of MDCKII cells we used DMEM supplemented with 5% FCS and L-glutamine. For the generation of stable cell lines H1299 cells were transfected with a GPI-mCherry or LifeAct-mCherry encoding plasmid and MDCKII cells were transfected with a Gb3-synthase encoding plasmid. Both plasmids additionally encode for a Geneticin resistance. For stable expression of GPI-mCherry and Gb3-synthase, positive clones were selected and continuously cultivated in RPMI or DMEM supplemented with 500 μg·mL−1 Geneticin, L-glutamine, and 10% FCS (for H1299 cells) or 5% FCS (for MDCKII cells). For the detection of Gb3, MDCKII cells were trypsinized, washed once with DPBS, incubated for 30 min at 37 °C with StxBAlexa488 (12 μg·mL−1), washed again twice with DPBS, and subjected to FACS. GFP Tagging and Deletion of lecA in Pseudomonas aeruginosa. GFPtagged P. aeruginosa PAO1 WT and PAO1 ΔlecA were constructed as followed: PCR was used to generate a 479-bp DNA fragment upstream of the lecA gene, with the (DellecAUp5ACCCCGTGCCGGTTCGACCCCGGC, DellecAUp3GGTTGGCAGGCCACCCCGTG) oligonucleotide pairs and a 468-bp DNA fragment downstream of the lecA using the (DellecADn5CAAGTTATCACCAAGCATGATTGATCTC, DellecADn3CATGGCTTGGTGATAACTTGTCTCGGAAAA) oligonucleotide pairs. The resulting DNA fragments were further used as templates for a second overlapping PCR run using a pair of external oligonucleotides (DellecAUp5 and DellecADn3), thus leading to a final approximate 1.1-kb DNA fragment that was cloned into the pCR2.1 vector. The resulting DNA fragment bearing appropriate sites, namely, BamHI/EcoRV, was further hydrolyzed and cloned into the suicide vector pKNG101. The recombinant plasmid was then mobilized into P. aeruginosa and the deletion mutant was selected on LB plates containing 6% sucrose and streptomycin. The P. aeruginosa strains, the Escherichia coli donor strain harboring a plasmid containing miniTn7Δgfp (Gm cassette), and the two E. coli helper strains harboring Eierhoff et al. www.pnas.org/cgi/content/short/1402637111

pRK600 and pUX-BF13 were grown overnight at 37 °C in LB, in the presence of the required antibiotics. This protocol is similar to the three-partner mating but with a longer period of incubation at 42 °C for the recipient strain and overnight contact. P. aeruginosa stocks were prepared by inoculating LB–Miller medium containing 60 μg·mL−1 Gentamicin with material of a single colony of P. aeruginosa grown on HiFluoro Pseudomonas Agar Base (Sigma-Aldrich), incubated overnight at 37 °C, mixed with glycerol (30% vol/vol), aliquoted, and stored at −80 °C. For experiments, LB–Miller medium containing 60 μg·mL−1 Gentamicin was inoculated with P. aeruginosa of a stock aliquot incubated at 37 °C on a Thermomixer (PeqLab) at 650 rpm for 16–20 h to ensure that LecA was efficiently expressed (2). The growth kinetic of P. aeruginosa was recorded in a 96-well plate containing LB medium by measurement of the OD at 600 nm in a Tecan Safire Plate Reader (Tecan). LecA expression was tested by standard Western blotting using LecA-specific, polyclonal rabbit antibody. Invasion and Adhesion Assay. Overnight cultures of bacteria were pelleted and resuspended in RPMI or DMEM for infection of H1299 and MDCKII cells, respectively, containing additionally 1 mM CaCl2 and MgCl2. Bacteria were incubated for 2 h at 37 °C with cells (70–80% confluent) at a multiplicity of infection (MOI) of 100. Cells were washed three times with DPBS. Extracellular bacteria were inactivated by a 2-h treatment of cells at 37 °C with 400 μg·mL−1 Amikacin sulfate (Sigma-Aldrich). Afterward, cells were washed two times with DPBS and lysed with 0.25% (vol/vol) Triton X-100 at 37 °C. Cell extracts were plated on Gentamicin-containing (60 μg·mL−1) LB–Miller Agar plates and incubated over night at 37 °C. The next day bacterial colonies were counted. Invasion was calculated as percentage of Amikacin-survived bacteria compared with the total number of bacteria associated with Amikacin-untreated cells. For comparison purposes mean values of n ≥ 3 independent experiments were normalized to the invasion of untreated, WT bacteria. Invasion and adhesion assays with lecA- and non-lecA-expressing E. coli strains were performed as described above. Instead of Amikacin, 100 μg·mL−1 Gentamicin was used for 1 h at 37 °C. For adhesion assays cells were inoculated with bacteria at 4 °C for 1 h, afterward washed three times with cold DPBS, lysed, and plated as described for the invasion assay. Inhibition of Actin Polymerization. Inhibition of actin polymerization was achieved by two complementary approaches: knockdown of actin-related protein 2 (Arp2) and treatment of H1299 cells with latrunculin A (Sigma-Aldrich), a compound that binds to monomeric actin, thereby preventing actin polymerization. Knockdown of Arp2 was achieved by transfecting 5 × 105 H1299 cells two times within 3 d with 200 pmol siRNA (Santa Cruz) using Lipofectamine 2000 (Invitrogen). The knockdown efficiency was tested by Western blot analysis using an Arp2-specific antibody (Santa Cruz). Three days posttransfection, cells were inoculated with P. aeruginosa PAO1 WT for fluorescence microscopy analysis or invasion assays as described. Latrunculin A was applied at 0.1 μM final concentration 15 min before and during inoculation with P. aeruginosa PAO1 WT. Plasma Membrane Spheres. For plasma membrane spheres (PMS) studies, we used H1299 cells stably expressing LifeAct-mCherry to visualize actin polymerization (3). PMS buffer contained 10 μM (final concentration) of MG132 (Sigma). Cells were 1 of 10

incubated with PMS buffer for 13–16 h at 37 °C. Subsequently, cells were incubated for 1 min at room temperature with FM4-64 Dye [N-(3-triethylammoniumpropyl)-4-(6-(4-(diethylamino) phenyl) hexatrienyl) pyridinium dibromide; Life Technologies]in PMS buffer (dilution: 1:10,000) to visualize the PM. After replacement by fresh PMS buffer cells were inoculated with bacteria to perform confocal microscopy. Cholesterol Depletion. Cellular cholesterol was depleted by preincubation of cells at 37 °C with 10 mM methyl-β-cyclodextrin (MCD) for 30 min. Afterward, cells were washed once with DPBS and inoculated with P. aeruginosa PAO1 WT to conduct invasion assays or microscopic analyses. Microscopic Imaging. Cells, GUVs, and P. aeruginosa infection were imaged on a confocal microscope (Nikon Eclipse Ti-E with A1R confocal laser scanner, 60× oil objective, N.A. 1.49). Image acquisition and analysis was performed with NIS-Elements (Nikon). For visualization of cellular infection, H1299 GPI-mCherry or H1299 LifeAct-mCherry cells grown on glass coverslips or in CELLview dishes (Greiner) were inoculated for 0.5 and 1 h at 37 °C, respectively, with P. aeruginosa with an MOI of 100. Afterward, cells were washed once with DPBS, fixed, and stained for actin by phalloidin-ATTO 647N (Sigma-Aldrich) and Arp2 using an Arp2-specific rabbit antibody (Santa Cruz) detected by an anti-rabbit Alexa405 (Life Technologies) secondary antibody. Live cell imaging was performed at 37 °C by using an incubator stage mounted onto the microscope (Okolab). For visualization of GUVs, images were recorded using the resonant scanning mode of the A1R confocal. GUV Preparation. GUVs were prepared by the electroformation technique at room temperature on indium-tin oxide (ITO)-coated slides. If not indicated otherwise all lipid preparations contain 30 mol% cholesterol (Avanti Polar Lipids), 0.25 mol% DHPE-TR (Life Technologies), and indicated concentrations of Gb3 (Matreya LLC). Additionally, lipid mixtures contain 1,2-dioleoyl-sn-glycero-3phosphocholine (DOPC) (Avanti Polar Lipids) of the following concentrations: 64.75 mol% for 5 mol% Gb3-mixtures, 68.75 mol% for 1 mol% Gb3, 69.65 mol% for 0.1 mol% Gb3, 69.25 mol% for 0.5 mol% Gb3, 69.74 mol% for 0.01 mol% Gb3, and 69.75 mol% for 0 mol% Gb3. Cholesterol-free GUVs contained 94.75 mol% DOPC, 5 mol% Gb3, and 0.25 mol% DHPE–Texas Red. Lipid mixtures were dissolved in chloroform (0.5 mg·mL−1) and 15 μL was spread on the conductive faces of the ITO slides. After at least 2 h of drying under vacuum, GUVs were grown in a 290 mOsm·L−1 sucrose solution by applying an alternating electric field from 20 mV to 1.1 V for 3 h. Surface tension of GUVs was decreased by increasing the osmolarity of the outer GUV buffer, which was adjusted to 550 mOsm·L−1. Bacteria were incubated with GUVs at room temperature and examined on the inverted confocal fluorescence microscope. Physical Model. The mechanical Helfrich energy associated with an elastic membrane is, per unit area,   1 dψ sinðψÞ e = γ + 2κH 2 ; H = cosðψÞ + ; [S1] 2 dr r

with surface tension γ, bending rigidity κ, mean curvature H (4), and angle ψ and radius r as defined in Fig. 2A. For a surface element 2πr(s) ds the Helfrich energy and adhesion energy, as opposed to the unbound plane membrane, are with the adhesion energy per Gb3: Eierhoff et al. www.pnas.org/cgi/content/short/1402637111

   2 2κHðsÞ2 · 2πrðsÞds = 2πκ 2 RHðsÞ ~r ð~sÞd~s ðbendingÞ   γð1− cos ψÞ·2πrðsÞds =2πκ ~γ ð1− cos ψÞ~rð~sÞd~s ðsurface tensionÞ   −eρ · 2πrðsÞds = 2πκ −~e~ ρ~r ð~sÞd~s ðadhesion energyÞ: Lipid «, the local Gb3 surface density ρ, the cylinder radius R, and the geometry are as defined in Fig. 2A. On the right-hand side we have rescaled the lengths by R, surfaces by 2πR2 (and surface densities by the inverse), γ by κR−2, and energies by 2πκ. The cylinder surface fraction where the membrane and the rodshaped particle are in contact is A = 2πRh and, after rescaling, ~ We replace A ~ and omit the tilde for rescaled ~ = h=R = h. ~ by h A quantities in the following. The rescaled free energy is Fm ðhÞ =

Z    2   2 RHðsÞ + γ 1 − cos ψðsÞ − eρðhÞ rðsÞds; [S2]

where the integral is carried out over the cylinder surface fraction h. The contribution of the free membrane part—the part not attached to the cylinder—cannot be treated analytically. However, the contribution of this part is always very small, as we confirmed by numerical integration (5). To make progress we neglect this part in what follows. In Fig. S4A we compare the full free energy with the approximation neglecting the free membrane part. After integration we obtain 8   1 > > for 0 ≤ h ≤ 1 > h 2 − eρðhÞ + γh2 > > 2 > >   > < 1 3 1 for 1 ≤ h ≤ L + 1 Fm ðhÞ = h − eρðhÞ + γ + − γ 2 2 2 > > > >   > >   > > h 2 − eρðhÞ + 1 γðh − LÞ2 + L γ − 3 for L + 1 ≤ h ≤ L + 2: : 2 2 [S3]

Considering N Gb3 lipids (M DOPC lipids) on the vesicle with surface V from which n (m) lipids lie within the surface fraction h we can write the relations between areas and lipid numbers V = Na + Mb;

h = na + mb

[S4]

with the surface areas a and b of Gb3 and DOPC, respectively. The entropy of the noninteracting particles is proportional to the logarithm of the number of characteristic microstates:   Sðh; ρÞ = ln gðn; m; N; MÞ ;

 g=

n+m n

   N −n+M −m × N −n [S5]

with the local and average Gb3 densities ρ = n/h and σ = N/V. The entropy loss is Fl ðh; ρÞ =

   kB T  SðρÞ − SðσÞ =: K SðρÞ − SðσÞ : 2πκ

[S6]

Note that it does not matter whether the mixing entropy is derived in the microcanonical ensemble, as we do it here, or in the canonical ensemble. Using the Stirling formula and Eq. S4 we obtain for the derivative     1 ∂Fl a bρ a bðV σ − AρÞ − log 1 + = log 1 + Kh ∂ρ b 1 − aρ b V − A − aðV σ − AρÞ     b−a 1 b−a V −A + + log + : − log b bρ b bðV σ − AρÞ [S7] 2 of 10

By minimizing Fl + Fm for fixed h we obtain the local density ρ0. With the simplification b = a we can obtain a closed expression ∂Fm ∂Fl + =0 ∂ρ ∂ρ



ρðσÞ =

νðσÞ −

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi νðσÞ2 − 4Vhaσαðα − 1Þ 2haðα − 1Þ [S8]

with α = exp («/K) and ν(σ) = V + (α − 1)(Vaσ + h). Further progress with these results turns out to be difficult (i.e., analytical progress would be blocked and one would have to refer to numerical methods). However, the steric repulsion between the particle is only important for high local densities, that is, as long as the local density is low one could treat the lipids as point particles. Of course the local density is unknown a priori because it results from the minimization of the free energy. A possible route out of this dilemma is to approximate the lipids as point particles, calculate h, and check subsequently the validity of the approximation. In the following we derive the mixing entropy of point-like particles in the canonical ensemble:   1 V N ZðT; V ; NÞ = ; N! l2

FðT; V ; NÞ = − 

 kB T  ln ZðT; V ; NÞ ; 2πκ [S9]

with some length l. Again, we use the Stirling formula, add the free energies for the surface fractions h and V − h, subtract the free energy for ρ = σ, and obtain    ρ hρ Fl ðh; ρÞ = Khρ ln + KV σ 1 − ln σ Vσ ⇒

σα : ρ= h 1 + ðα − 1Þ V

hρ Vσ h 1− V

!

1−

[S10]

Substituting ρ into Eqs. S3 and S10 the wrapping height follows from ′ ðhÞ + Fl′ðhÞ 0 = Fm 8 Kσðα − 1Þ > > − + 2 + γh for > > 1 + ðα − 1Þh=V > > > > > < Kσðα − 1Þ 1 + +γ for = − > 1 + ðα − 1Þh=V 2 > > > > > > > Kσðα − 1Þ > :− + 2 + γðh − LÞ for 1 + ðα − 1Þh=V

0≤h≤1 1≤h≤L+1 L+1≤h≤L+2: [S11]

To verify the validity of this result, we numerically calculated the full free energy (including the numerically evaluated free membrane part in Eq. S2 and the free energy for the lattice gas in Eq. S6) for different surface densities σ. Fig. S4A depicts the good agreement between the simplified energies for point particles, the corrected energy for the lattice gas, and the full free energy. For small densities, the contribution of the free membrane part dominates the deviations, whereas the saturating density of the lattice gas (Fig. S4C) becomes important for increasing densities. It can be seen that the approximations done only introduce a minor error. Further, we show in Fig. S4B the relative error of the local density ρP, if one treats the lipids as point particles and ignores the steric repulsion, in comparEierhoff et al. www.pnas.org/cgi/content/short/1402637111

ison with the saturating local densities ρ for the lattice gas. The relative error is in the relevant regime always smaller than 5%, which justifies the point-particle approximation. Critical GUV Diameters for Full Wrapping. For larger vesicles, entropy loss becomes less important owing to the larger surface reservoir. By solving Eq. S11 with respect to the diameter of the vesicle, we achieve an expression for the minimal or critical diameter dc for which the bacteria is at least wrapped up to a height h as a function of the system parameters. For the case of full wrapping (h = L + 2) and of wrapping just up to the lower boundary of the spherical cap (h = L + 1), one obtains sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 18ðα − 1Þ [S12] dc ðh = L + 1Þ = R Kσðα − 1Þ=ð0:5 + γÞ − 1

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 20ðα − 1Þ : dc ðh = L + 2Þ = R Kσðα − 1Þ=ð2 + 2γÞ − 1

[S13]

Owing to the dynamics of the GUV system and resultant technical challenges we were unable to clearly resolve the extent of wrapping. In addition, the global Gb3 concentrations on the vesicles might vary owing to lipid heterogeneities during GUV preparation. The resulting uncertainty regime of the critical vesicle diameter is shown in Fig. 2D. The lower boundary of the uncertainty regime was calculated by using dc(h = L + 1) with 40% increased Gb3 concentration and the upper boundary by using dc(h = L + 2) with 40% decreased Gb3 concentration. The surface tension was fixed to γ = 800 μJ/m2 in both cases. Values. The calculations were carried out at room temperature (T = 293 K). The bending rigidity for typical phospholipid bilayers is κ = 20 kBT (see refs. 6 and 7). The surface tension of the GUVs lies within γ = 10−6 to 10−3 J/m2 (8). Microcalorimetry titration shows that the free energy gain owing to the binding of Gb3 (αGal14βGal1-4Glc) to LecA (PA-IL) is −ΔG = 5.6 kcal/mol (9), which is equivalent to an adhesion energy « ≈ 9.6 kBT per lipid. In GUV experiments, the bilayer contained 65 mol% DOPC and 5 mol% Gb3. The area per lipid of Gb3 and DOPC in a bilayer is a = 80 Å2 (10) and b = 67.4 Å2 (11), respectively. Thus, the surface concentration of Gb3 was σ ≈ 105 μm−2 on average and may not surpass 1.25 nm−2. The fraction of glycosphingolipids in the plasma membrane is believed to be ∼1–5% of all lipids. For Gb3, 0.1–0.5% seems plausible, so we expect the average surface density of Gb3 within the range σ = 1,400−7,400 μm−2. Clustering of Gb3 Lipids. During the wrapping process of the bacteria Gb3 lipids are recruited to the wrapping region and thereby enhance the adhesion energy and subsequently the wrapping height. The local density of Gb3 lipids is shown in Fig. S4C as a function of the global Gb3 density σ and the diameter of the GUV. Owing to the recruiting mechanism bacteria are fully wrapped at a lower global Gb3 density compared with the case of immobile lipids (Fig. 2C). This effect is more pronounced for larger vesicles, because the Gb3 lipid reservoir is larger compared with smaller vesicles. Although the clustering of Gb3 lipids is important for the wrapping process, the local density does not saturate before the bacteria are fully wrapped. This can be seen by comparing Fig. 2C and Fig. S4C. For example, a global Gb3 density of σ = 0.001 nm−2 results for a GUV of 50-μm diameter (red curve) in a fully wrapped bacteria, whereas the local density is significantly increased but still well below the saturation limit. 3 of 10

where « is the adhesion energy per glycolipid and ρ the local glycolipid concentration. The size R determines the strength

of the influence of the bending rigidity κ on the wrapping of the capsid. We calculate ρ by dividing the number of receptors [72 for the capsids (8)] by the surface 4πR2. The radii and adhesion energies (see ref. 8) for the receptor monomers, pentamers, and capsids as well as the derived quantities are given in Table S1. All densities are far below the upper boundary for Gb3, ρmax = 1.25 nm−2. Because Z < 2 for the monomers and pentamers, Eq. S14 has no positive solution for h, so without mechanisms other than described here no invagination would occur. The capsids will be just fully wrapped when ðZ − 2Þ~γ −1 = 2 ⇔ ~γ −1 ≈ 2:3 ⇔ γ = 3 · 10−5  J=m2 . Invaginations on GUVs induced by SV40 virus-like particles have been observed even with surface tensions in the order of 10−3 J/m2 (see ref. 8), so other mechanisms have to be considered for the capsids as well.

1. Abe A, et al. (1992) Improved inhibitors of glucosylceramide synthase. J Biochem 111(2):191–196. 2. Winzer K, et al. (2000) The Pseudomonas aeruginosa lectins PA-IL and PA-IIL are controlled by quorum sensing and by RpoS. J Bacteriol 182(22):6401–6411. 3. Riedl J, et al. (2008) Lifeact: A versatile marker to visualize F-actin. Nat Methods 5(7): 605–607. 4. Helfrich W (1973) Elastic properties of lipid bilayers: Theory and possible experiments. Z Naturforsch C 28(11):693–703. 5. Deserno M, Bickel T (2003) Wrapping of a spherical colloid by a fluid membrane. EPL 62:767. 6. Reynwar BJ, et al. (2007) Aggregation and vesiculation of membrane proteins by curvature-mediated interactions. Nature 447(7143):461–464.

7. Deserno M (2004) Elastic deformation of a fluid membrane upon colloid binding. Phys Rev E Stat Nonlin Soft Matter Phys 69(3 Pt 1):031903. 8. Ewers H, et al. (2010) GM1 structure determines SV40-induced membrane invagination and infection. Nat Cell Biol 12(1):11–18, 1–12. 9. Blanchard B, et al. (2008) Structural basis of the preferential binding for globo-series glycosphingolipids displayed by Pseudomonas aeruginosa lectin I. J Mol Biol 383(4): 837–853. 10. Mahfoud R, Mylvaganam M, Lingwood CA, Fantini J (2002) A novel soluble analog of the HIV-1 fusion cofactor, globotriaosylceramide (Gb(3)), eliminates the cholesterol requirement for high affinity gp120/Gb(3) interaction. J Lipid Res 43(10):1670–1679. 11. Kucerka N, et al. (2008) Lipid bilayer structure determined by the simultaneous analysis of neutron and X-ray scattering data. Biophys J 95(5):2356–2367.

Simian Virus 40. Although we treated the density of (mobile) LecA

on bacteria as an unknown but nonlimiting factor, the number of glycolipid receptors on the SV 40 virus capsids and its geometry are well known. To compare our results to the invagination of virus capsids, we assume for simplicity that the local concentration of Gb3 is as large as the receptor density, so the adhesion effect is maximal. We thus minimize Fl + Fm with respect to h for a constant local density ρ to calculate the ratio of wrapping height h to capsid radius R:    2  h 2κ −1 R κ −1 ρe − 2   γ =: ðZ − 2Þ~γ −1 ; = ρe − 2 γ = κ R R R2

[S14]

Fig. S1. Characteristics of P. aeruginosa PAO1 WT and ΔlecA mutant. Growth kinetics of P. aeruginosa PAO1 WT and ΔlecA cultures at 37 °C. PAO1 WT and ΔlecA exhibit the same growth kinetics. (Inset) Overnight cultures of P. aeruginosa PAO1 WT and ΔlecA were tested for LecA production by standard Western blot analysis, verifying the lack of LecA synthesis in the ΔlecA mutant.

Eierhoff et al. www.pnas.org/cgi/content/short/1402637111

4 of 10

Fig. S2. Representative images of GUVs inoculated with P. aeruginosa PAO1 WT or ΔlecA mutant. (A–C) Images of GUVs as shown in Fig. 1 containing 5 mol% Gb3 inoculated with P. aeruginosa PAO1 WT or ΔlecA. (D and E) In A and B, a tether-like structure is visible between the membrane-wrapped bacterium and the outer GUV membrane (white arrowhead). Although PAO1 ΔlecA bacteria bound either horizontally or vertically to Gb3-containing GUVs, they do not induce membrane invaginations, as observed frequently for the WT strain. (F) In most cases PAO1 ΔlecA bacteria did not bind to Gb3-containing GUVs. (Scale bars, A–C, Upper, D, and F, 5 μm; A–C, Lower, and E, 2.5 μm.)

Fig. S3. Decrease in membrane tension facilitates the lipid zipper. Membrane tension of GUVs containing indicated Gb3 concentrations was decreased by exposure of GUVs to a hyperosmolar, external buffer solution (550 mOsm·L−1 outside buffer vs. 290 mOsm·L−1 inside buffer). Decrease of surface tension leads to an increase of GUVs showing membrane-wrapped bacteria. Relative numbers of GUVs containing membrane-wrapped P. aeruginosa PAO1 WT are shown. Numbers of invaginated GUVs were normalized to number of GUVs bound by bacteria. Bars represent mean values ± SEM of n ≥ 3 independent experiments. In total ≥100 GUVs were analyzed (46 GUVs at 550 mOsm·L−1).

Eierhoff et al. www.pnas.org/cgi/content/short/1402637111

5 of 10

Dimensionless free energy

A

B

Point particles Lattice gas Full free energy Calculated minima

d = 16 µm Jm 0.5

m

1

1.5

2

2.5

1.2 1.0 0.8 0.6 0.4 0.2 0.0

C

GUV diameters 50 µm 16 µm 5 µm

1

10 100 1000 Average Gb3 surface density σ in µm

Local density in nm

Wrapping height h in µm

104

Fig. S4. (A) Numerically calculated full free energy (including the numerically evaluated free membrane part in Eq. S2 and the free energy for the lattice gas in Eq. S6) for different surface densities σ. The perpendicular black lines show the minimum of the free energy. At these points the difference between the numerically calculated free energy and the used approximations are always minute, justifying the validity of the simplifications made. (B) Relative error of the local density ρP, for point particles (ignoring steric repulsion) in comparison with the saturating local densities ρ for the lattice gas. The relative error is always smaller than 5% in the relevant regime, which justifies the point-particle approximation. (C) Local Gb3 density as a function of the global density σ on the vesicle for different GUV diameters. Although the mobility of the Gb3 density leads to significantly enhanced local densities, the bacteria are always fully wrapped before the local density goes into saturation (compare with Fig. 2C).

Fig. S5. Glycosphingolipid expression is a prerequisite for efficient P. aeruginosa uptake. (A, Left) Inhibition of glucosylceramide synthase by its substrate analog PPMP resulted in an 80% reduced invasiveness of P. aeruginosa. Internalization of P. aeruginosa PAO1 WT into untreated and PPMP-treated H1299 cells as measured by the invasion assay. (A, Right) Adhesion of P. aeruginosa PAO1 WT to untreated or PPMP-treated H1299 cells. All data represent mean values ± SEM for n ≥ 3 experiments normalized to the WT. (B) Inhibition of glucosylceramide synthase by PPMP prevents binding of StxB-Cy3, which selectively binds to the glucosylceramide-derived GSL Gb3, to H1299 cells, representing the inhibition of GSL synthesis in H1299 cells by PPMP (Lower). TM, transmission. (Scale bar, 10 μm.)

Eierhoff et al. www.pnas.org/cgi/content/short/1402637111

6 of 10

Fig. S6. P. aeruginosa induces negative PM curvature independently of actin polymerization. (A) GFP-tagged P. aeruginosa WT (PAO1-GFP) bound to an H1299 cell, which stably expresses GPI-mCherry as a PM marker. The lower panel represents a zoom of the squared area of the upper panel. Note that the PM is bent at the bacterial pole region at the cellular adhesion site, colocalizing with actin (arrowheads). Images were recorded from cells inoculated for 1 h with P. aeruginosa at MOI ∼100. For better visualization of the curved PM, the contrast of the images in the lower panel was adjusted. (Scale bars, 10 μm and 2.5 μm, respectively.) (B) RNAi-mediated knockdown of Arp2 by siRNA was confirmed by Western blotting of uninfected H1299 cells. In total, Arp2 was reduced by 72% as quantified by densiometric analysis of Western blots compared with the level of Akt as a loading control (n = 3, mean ± SEM). (C) Invasion of Arp2-depleted H1299 cells by P. aeruginosa WT was significantly reduced by about 66%. The reduction of invasion for the lecA mutant (ΔlecA) by 75% was even more pronounced (n = 4 independent experiments, mean ± SEM, P value calculated by two-tailed, paired t test). These observations show that actin polymerization is in general crucial for a subset of P. aeruginosa cells to efficiently enter host cells. However, the initial steps of membrane invagination (Fig. 3A) do not require actin polymerization-dependent processes. (D) Control cells, corresponding to Arp2-depleted cells shown in Fig. 3D. Cells were transfected with scrambled siRNA and recorded with the same laser power and gains as the Arp2 siRNA-transfected cells. The same type of membrane invaginations (arrowhead) induced by P. aeruginosa WT as observed in untransfected (Fig. 3A) and Arp2-depleted H1299 cells (Fig. 3D) is visible. Images were recorded from cells inoculated with P. aeruginosa for 1 h at MOI ∼100. (Scale bars, 10 μm and 2.5 μm, respectively.) (E) Complementary to the Arp2 knockdown approach we assessed the dependency of membrane invaginations on actin polymerization processes, which we inhibited by latrunculin A (0.1 μM). In latrunculin A-treated H1299 cells P. aeruginosa is localized in PM invaginations (arrowheads) at 1 h postinoculation (MOI ∼100). The same type of localization can be seen in untreated cells (Fig. 3A) where bacteria are also engulfed by the host cell membrane. Actin-covered (zoom 1) as well as uncovered invaginations (zoom 2) were observed. Therefore, actin polymerization is not essential for membrane wrapping and invaginations. (Scale bars, 10 μm and 2.5 μm, respectively.)

Eierhoff et al. www.pnas.org/cgi/content/short/1402637111

7 of 10

Fig. S7. Cholesterol depletion significantly reduces lipid zipper on H1299 cells. Cells expressing GPI-mCherry as a PM marker were cholesterol-depleted by treatment with 10 mM MCD for 30 min or left untreated. Afterward, cells were infected for 1 h at 37 °C with P. aeruginosa PAO1 WT. PM-engulfed bacteria per cell of untreated and MCD-treated cells were counted. Values represent mean values ± SEM of n = 4 independent experiments with >500 bacteria analyzed in total.

Eierhoff et al. www.pnas.org/cgi/content/short/1402637111

8 of 10

Fig. S8. P. aeruginosa induces membrane invaginations in PMS. (A) PMS were induced in H1299 cells stably expressing the actin polymerization marker LifeAct-mCherry. Actin was only visible in the cell body but was excluded in the PMS. PAO1 WT inoculated with PMS-containing cells induced a membrane invagination in the PMS, which was not covered by actin (arrowhead). Membrane was visualized by FM4-64 dye. (B) Example of another PMS induced from a H1299 cell, which shows a bacterial cell engulfed by the membrane of the PMS in the upper part of the cell body (“top”). Interestingly, the invaginated membrane is connected via a tether-like structure with the outer membrane of the PMS (arrowhead) as already observed in GUVs (Fig. S2B). The lower part (“bottom”) shows actin in the remaining cell body. (Scale bars, 10 μm and 5 μm.)

Eierhoff et al. www.pnas.org/cgi/content/short/1402637111

9 of 10

Fig. S9. Control of Gb3 depletion and LecA inhibition. (A) PM-localized Gb3 content of H1299 cells is significantly decreased by pretreatment with StxB-Cy3 as indicated by a strong reduction of StxB-Alexa488 uptake. (B) Decreased PM-localized Gb3 in H1299 cells results in significantly reduced LecA binding and uptake compared with untreated (control) cells. Furthermore, this indicates that Gb3 is a receptor for LecA. For comparison, untreated and StxB-Cy3–treated cells were imaged with equal laser power and gain settings for StxB-Alexa488 and LecA-Alexa488, respectively. (Scale bars, 20 μm.) (C) PNPG selectively inhibits LecA uptake. Binding and uptake of LecA-Alexa488 (C) and StxB-Alexa488 (D) after treatment of H1299 cells with PNPG was visualized. PNPG significantly impairs LecA but not StxB uptake. Images of Alexa488 fluorescence were recorded with equal laser power and gain settings. TM, transmission. (Scale bars, 10 μm.)

Fig. S10. Cholesterol depletion significantly reduces host cell invasion by P. aeruginosa. Invasion of PAO1 WT into cholesterol-depleted H1299 cells was quantified. Values represent mean values ± SEM of n = 4 independent experiments.

Table S1. Comparison to SV40 capsid proteins Parameter r, nm « ρ Z ~γ

Monomer

Pentamer

Capsid

Large spheres

5 2 kBT 1.6 × 10−3 nm−2 Z = 0.008

12 10 kBT 5.5 × 10−4 nm−2 Z = 0.04

50 10 kBT 2.3 × 10−3 nm−2 Z − 2 = 0.86 0.03−32

250 10 kBT 2.3 × 10−3 nm−2 Z − 2 = 70 0.001−1.3

Eierhoff et al. www.pnas.org/cgi/content/short/1402637111

10 of 10