A lipid zipper triggers bacterial invasionThorsten 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

aFaculty of Biology, bBIOSS Centre for Biological Signalling Studies, cCentre for Biological Systems Analysis, and fSpemann Graduate School of Biology andMedicine, Albert Ludwigs University Freiburg, 79104 Freiburg, Germany; dCentre de Recherches sur les Macromolécules Végétales, Université Grenoble Alpesand 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 laRecherche Scientifique and Aix Marseille University, Institut de Microbiologie de la Méditerranée, 13402 Marseille, France; and iLaboratory for Systems andSynthetic 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 forreview February 11, 2014)

Glycosphingolipids are important structural constituents of cellularmembranes. They are involved in the formation of nanodomains(“lipid rafts”), which serve as important signaling platforms. Inva-sive bacterial pathogens exploit these signaling domains to triggeractin polymerization for the bending of the plasma membrane andthe engulfment of the bacterium—a key process in bacterial up-take. However, it is unknown whether glycosphingolipids directlytake part in the membrane invagination process. Here, we dem-onstrate that a “lipid zipper,” which is formed by the interactionbetween the bacterial surface lectin LecA and its cellular receptor,the glycosphingolipid Gb3, triggers plasma membrane bendingduring host cell invasion of the bacterium Pseudomonas aerugi-nosa. In vitro experiments with Gb3-containing giant unilamellarvesicles revealed that LecA/Gb3-mediated lipid zippering was suf-ficient to achieve complete membrane engulfment of the bacte-rium. In addition, theoretical modeling elucidated that the adhe-sion energy of the LecA–Gb3 interaction is adequate to drive theengulfment process. In cellulo experiments demonstrated that inhi-bition of the LecA/Gb3 lipid zipper by either lecA knockout, Gb3depletion, or application of soluble sugars that interfere with LecAbinding to Gb3 significantly lowered P. aeruginosa uptake by hostcells. Of note, membrane engulfment of P. aeruginosa occurred in-dependently of actin polymerization, thus corroborating that lipidzippering alone is sufficient for this crucial first step of bacterial host-cell entry. Our study sheds new light on the impact of glycosphin-golipids in the cellular invasion of bacterial pathogens and providesa mechanistic explication of the initial uptake processes.

membrane curvature | glycolipid | infection

Glycosphingolipids (GSLs) are essential components of bio-logical membranes that have significant impact on the or-

ganization and the molecular architecture of the membrane (1, 2).Preferential association of GSLs with cholesterol, various otherlipid species, and proteins induces formation of micro- and nano-domains. Such domains, which are also termed “lipid rafts,” areinvolved in the signal transduction across the membrane (3). Sev-eral invasive bacterial pathogens hijack these membrane domainsas signaling platforms for actin polymerization, which leads to theengulfment of the bacterium by the host plasma membrane (PM)(4–6). For example, Listeria monocytogenes exploits lipid raft-dependent E-cadherin and HGF-R/Met signaling in the host cell totrigger actin polymerization for bacterial uptake (7). The oppor-tunistic bacterium Pseudomonas aeruginosa can also invade andsurvive in diverse epithelial and endothelial cells (8–14), whichsignificantly contributes to its pathogenicity (8, 10, 15). Numerousreports suggested various host cell factors, including cell surfacereceptors and signaling components, including α5β1 integrin, cysticfibrosis transmembrane conductance regulator, and Abelson tyro-sine-protein kinase 1-dependent pathway (16–18), that are involvedin the cellular uptake of P. aeruginosa. Host-cell GSLs have been

identified as important molecules contributing to host specificityand adhesion of P. aeruginosa (19, 20). Recent observations suggestthat GSLs might be of critical importance for the internalization ofP. aeruginosa into nonphagocytic cells (9). The homotetrameric,galactophilic lectin LecA, which is localized to the outer bacterialmembrane (21), belongs to the carbohydrate binding proteinsexpressed by P. aeruginosa that recognize GSLs (22, 23) and rep-resents one of the virulence factors (24). The preferential bindingof LecA to the GSL globotriaosylceramide (also known asGb3/CD77 or Pk histo-blood group antigen) (22, 25) promptedus to investigate the role of LecA–Gb3 interaction in the cellularuptake 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 thepathogen. Binding of GSL by P. aeruginosa induced negativemembrane curvature on synthetic, giant unilamellar vesicles(GUVs), solely driven by LecA–Gb3 interaction. This interactionresulted in a bacterium tightly engulfed by the lipid bilayer of theGUV. Because the glycolipid component plays the major role inthis process, we termed this mechanism “lipid zipper” as a novelzipper-type mode of bacterial entry (26). Further experiments re-vealed that the lipid zipper has a significant impact on P. aeruginosauptake in several epithelial cell lines. Interestingly, actin

Significance

Entry of bacteria into host cells critically depends on theirproper engulfment by the plasma membrane. So far, actinpolymerization has been described as a major driving force inthis process. However, our study reveals that the interaction ofthe bacterial surface lectin LecA with the host cell glyco-sphingolipid Gb3 is fully sufficient to promote engulfment ofPseudomonas aeruginosa, whereas actin polymerization isdispensable. Hence, the formation of a “lipid zipper” repre-sents a previously unidentified mechanism of bacterial uptakeand demonstrates that bacterial pathogens have evolved lipid-based invasion strategies that may function in addition toprotein receptor-based ones. Furthermore, by identifying theLecA/Gb3 interaction as the major invasion-promoting factor,our study provides new targets for drugs that may preventbacterial invasion and dissemination.

Author contributions: T.E., B.B., A.I., C.F., andW.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. contributednew reagents/analytic tools; T.E., B.B., C.F., andW.R. analyzed data; and T.E., C.F., andW.R.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence may be addressed. Email: thorsten.eierhoff@bioss.uni-freiburg.de,christian.fleck@wur.nl, or winfried.roemer@bioss.uni-freiburg.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1402637111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1402637111 PNAS | September 2, 2014 | vol. 111 | no. 35 | 12895–12900

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

polymerization, which was previously identified as a majordriving force for membrane deformation and engulfment, wasnot required for the lipid-triggered process. Moreover, ectopicexpression of LecA in Escherichia coli was sufficient to greatlyincrease its invasion efficiency in Gb3-positive cells.

ResultsA LecA–Gb3 Lipid Zipper Is Sufficient for Membrane Engulfment ofP. aeruginosa in a Synthetic Lipid Bilayer System. We used Gb3-containing GUVs as a membrane model system to directly ad-dress the impact of LecA–Gb3 interactions on the curvature ofa lipid bilayer without any potential interference of cellular fac-tors, particularly actin. We incubated GUVs with the invasiveP. aeruginosa PAO1 WT strain and a lecA deletion mutant(ΔlecA) that did not express LecA (Fig. S1, Inset). Both strainswere chromosomally GFP-tagged and did not differ in theirgrowth rates (Fig. S1). PAO1 WT not only bound to Gb3-con-taining GUVs but also induced a highly curved GUV membraneat the contact point, which finally led to a complete engulfmentof the WT bacteria by the GUV membrane (Fig. 1A and Fig. S2A–C). Remarkably, we also observed membrane-engulfed bac-teria that were connected via tether-like structures to the sur-rounding GUV membrane (Fig. S2 A, arrowhead and B). Inter-estingly, we observed clustering of lipid material at sites wherebacteria attached to the GUVs and were engulfed (Fig. 1A). Intotal, 82 ± 6.5% of the analyzed GUVs were bound by PAO1WTand of these 45 ± 6.4% showed bacteria, which were engulfed by

the GUV membrane (Fig. 1B). In contrast, the ΔlecA mutantessentially did not induce membrane invaginations (Fig. S2 D–F).Although still 22 ± 7.7% of the GUVs were bound by the ΔlecAmutant (Fig. 1B), only 1 of 102 GUVs analyzed in total showeda membrane-engulfed bacterium. Lowering of the Gb3 concen-tration in GUVs from 5 to 0 mol% led to a dose-dependentdecrease of GUVs containing membrane-engulfed bacteria (Fig.1C), clearly indicating a crucial role of Gb3 in the process ofbacterial membrane engulfment. Without Gb3, only 1 of the 106GUVs analyzed showed a wrapped bacterium. In the next step weaddressed the impact of the mechanical properties of the GUVmembrane. When decreasing the surface tension by applying ahyperosmolar buffer solution from the outside (550 mOsm·L−1

outside vs. 290 mOsm·L−1 inside the GUVs), GUVs containing0.1 mol% Gb3 showed a number of membrane-wrapped bacteriacomparable to that obtained with 5 mol% Gb3 GUVs at iso-osmolar 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 in-fluence membrane rigidity, which might affect the formation of thelipid zipper. Therefore, we compared the efficiency of lipid zipperformation on GUVs in the presence or absence of cholesterol. Theefficiency was reduced by 65% in the absence of cholesterol (45 ±6.4% vs.15.7 ± 4.8%), which, apparently, was not due to a decreasein 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 thatbinding of LecA to Gb3 mediates a cholesterol-dependent, zipper-like engulfment of bacteria by the GUV membrane.

The LecA–Gb3 Interaction Provides Enough Energy for CompleteEngulfment of P. aeruginosa by a Lipid Membrane. We developeda theoretical model to investigate whether the LecA–Gb3 in-teraction is sufficient to mediate membrane engulfment of bac-teria. The GUV was described as an elastic surface consisting ofmobile lipid constituents, and P. aeruginosa was modeled as aLecA-covered, rigid rod (Fig. 2A and SI Materials and Methods).The total energy of the system consists of three terms: (i) me-chanical Helfrich energy (27), which favors flat membrane con-figurations; (ii) curvature-promoting adhesion energy owing toLecA–Gb3 interaction; and (iii) mixing entropy (28) of mobilelipids. The final configuration of the bacterium relative to theGUV membrane is the result of minimizing the total energy anddepends solely on the physical properties of the system, that is,size of the bacterium, bending rigidity, and surface tension of theGUV membrane, as well as Gb3 density. In Fig. 2B we show thewrapping height h of a bacterium (L = 2.5 μm) in dependence ofsurface tension γ and Gb3 density σ for constant bending rigidity(SI Materials and Methods). For realistic values of GUV Gb3densities (σ ≈ 105 μm−2), the model predicts a complete mem-brane engulfment of the bacterium. Similar effects are predictedfor 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 ten-sions between 10 μJ·m−2 and 300 μJ·m−2 (29–31) are realistic.Furthermore, the model reveals that LecA-dependent clusteringof mobile Gb3 lipids is crucial. When Gb3 is kept immobile, theglobal Gb3 concentration required to achieve full engulfmentincreased by two orders of magnitude (Fig. 2C; compare curvesin color and black for mobile and immobile Gb3, respectively).Full engulfment occurred at local Gb3 densities well belowcritical values (i.e., close packing of Gb3) (compare Fig. 2C andFig. S4C).To further explore the dependence of the wrapping on the

global Gb3 density and the vesicle size, we incubated GUVsprepared from lipid mixtures containing different Gb3 concen-trations (Materials and Methods) with P. aeruginosa PAO1 WT.By microscopic analysis we measured the equatorial diameterof GUVs in a random field of view. On the basis of our model

Fig. 1. P. aeruginosa induces a lipid zipper on giant unilamellar vesiclesby binding of LecA to Gb3. (A) GUVs reconstituted with Texas Red-labeled1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine (DHPE-TR) and Gb3were inoculated with GFP-tagged P. aeruginosa PAO1 WT (PAO1-GFP) for 15min at 37 °C. Lipid clustering at the bacterial attachment site is visible (ar-rowhead). (Scale bar, 5 μm.) (B) Relative number of GUVs containing boundor membrane-wrapped P. aeruginosa PAO1 WT or PAO1 ΔlecA. Bound GUVswere normalized to the total number of GUVs analyzed and invaginatedGUVs were normalized to GUVs bound by bacteria. Bars represent meanvalues ± SEM of n = 4 independent experiments with ≥100 GUVs analyzed intotal. (C) Relative number of GUVs with indicated Gb3 concentrations con-taining membrane-engulfed P. aeruginosa PAO1 WT. Bars represent meanvalues ± SEM of n ≥ 3 independent experiments. For 0 mol% Gb3, 106 GUVswere 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 ofGUVs with bound or invaginated bacteria as normalized in B with lipidmixtures containing 5 mol% Gb3 with or without 30 mol% cholesterol. Barsrepresent mean values ± SEM of n ≥ 3 independent experiments with ≥100GUVs analyzed in total.

12896 | www.pnas.org/cgi/doi/10.1073/pnas.1402637111 Eierhoff et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

(see above), we calculated the minimal vesicle size for which thebacteria would be fully wrapped (h = 2.5 μm) for the global Gb3density σ on the vesicle at the applied Gb3 concentrations. Fig.2D compares the experimental results with the theoretical pre-diction for three different surface tensions (solid line, γ =800 μJ·m−2). The shaded area in Fig. 2D denotes the uncertaintyrange for the minimal vesicle size assuming that the bacterium isat least wrapped up to the spherical cap (h = 2.25 μm) and a 40%variation in the Gb3 content (SI Materials and Methods). Addi-tionally, we experimentally tested our model by variation of thesurface tension of GUVs as a critical parameter for bacterial

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

LecA Promotes a Lipid Zipper on Epithelial Cells Independent of ActinPolymerization. In further experiments, we directly assessed whethera lipid zipper mechanism is implicated in bacterial invasion ofliving host cells. We first verified that cell entry of P. aeruginosainto 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 showedan 80% reduction of invasion upon inhibition of GSL synthesisby 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 lossof Shiga toxin B-subunit (StxB) binding to cells (Fig. S5B). In thefollowing we studied the entry process of P. aeruginosa into H1299cells by confocal fluorescence microscopy.Shortly after binding to H1299 cells, bacteria induced negative

PM curvature at the cellular entry site (Fig. S6A) that progressedinto the engulfment of the bacterium by PM (Fig. 3A, arrow-heads). For the ΔlecA mutant, the number of PM-surroundedbacteria was significantly lower (about 64.5%) at 1 h postin-fection compared with the WT strain (Fig. 3B), suggesting thatmainly LecA contributes to the zipper-like engulfment of bac-teria by PM. In line with the GUV data, extraction of cholesterolby methyl-β-cyclodextrin treatment lowered the occurrence ofzipper-like, PM-engulfed bacteria by about 50% (Fig. S7) andthereby to the same extent as observed for GUVs, with andwithout cholesterol (Fig. 1D). Interestingly, actin polymerizationdoes 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 markerGPI-mCherry, was frequently seen close to invading bacteriaeven in Arp2-depleted cells (Fig. 3D). However, actin coverageseems not to be functionally linked to the formation of mem-brane invaginations, because we observed actin-covered, PM-enveloped bacteria under all perturbations (Fig. 3D and Fig.S6E, arrowheads). Most likely, actin coverage around invadingbacteria represents persistent, cortical actin filaments. To furthertest whether actin polymerization is redundant for the zipperprocess, we studied the interactions of WT bacteria with plasmamembrane spheres (PMS) induced in H1299 cells that stablyexpressed the actin marker LifeAct-mCherry. In PMS, no endo-cytic machinery is present and (cortical) actin is excluded (36).Intriguingly, we observed bacteria that were tightly engulfed bythe membrane of PMS, which demonstrates membrane wrappingeven in the absence of actin (Fig. S8). These PM-envelopedbacteria closely resembled those wrapped by GUV membranes(Fig. S2 B and C). Therefore, we propose a mechanistic model inwhich actin polymerization is dispensable for membrane engulf-ment but required for completion of the invasion process and

Fig. 2. Gb3 binding and clustering predicts complete membrane engulfmentof P. aeruginosa. (A) Rotationally symmetric geometry of a modeled rod-shaped bacterium engulfed by GUV membrane. (B) Wrapping height in de-pendence of surface tension and Gb3 surface density (GUV diameter d =16 μm). (C) Invagination depth and local Gb3 density as function of the globalGb3 density and different GUV diameters for intermediate surface tension(γ = 10−4 J·m−2). Clustering of mobile Gb3 reduces the global Gb3 concen-tration necessary for full wrapping (compare colored curves for mobile vs.black curve for immobile Gb3). Shaded area corresponds to Gb3 densities inthe PM of host cells. (D) Critical GUV sizes. Circles represent observed vesiclesof different size and with different Gb3 content (they are scattered aroundx-values for improved visibility). Each wrapping event is denoted by red pad-ding. Three lines represent the critical GUV size for full enclosing of bacteriaat likely surface tensions γ. The shaded area shows the effect of uncertaintyin effective Gb3 content in a single vesicle at γ = 800 μJ·m−2 (SI Materials andMethods). For the right column, the osmolarity of the outside buffer wasincreased from 290 to 550 mOsm·L−1, whereas the inside buffer was kept at290 mOsm·L−1; the thereby reduced surface tension enhances wrapping atmoderate GUV sizes. The theoretical curves show qualitative agreementwith the observed statistics.

Eierhoff et al. PNAS | September 2, 2014 | vol. 111 | no. 35 | 12897

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

closing of the invaginated membrane cup to form an intracellularbacteria-containing vesicle. This hypothesis is supported by thefact that P. aeruginosa requires actin polymerization for completeentry into lung cells, as indicated by the reduced invasion intoArp2-knockdown cells (Fig. S6C).

LecA/Gb3-Mediated Lipid Zipper Contributes Measurable Impact onthe 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-cellinvasion efficiency. The P. aeruginosa PAO1 ΔlecA mutant strainshowed a 61.3% reduced invasion into H1299 cells compared with

PAO1WT (Fig. 4A, “untreated”). Next, we depleted PMGb3 levelsby StxB-induced endocytosis of Gb3 (32) (Fig. S9 A and B). StxB-mediated depletion of Gb3 reduced the invasion of the WT straininto H1299 cells by 69.0% (Fig. 4A, “StxB-treated”) compared withuntreated cells. No significant difference in invasion efficiency wasobserved with the ΔlecA strain on Gb3-depleted (“StxB-treated”)and Gb3-containing (“untreated”) H1299 cells. Additionally, weconfirmed by ectopic expression of Gb3 in MDCKII cells that theinvasion of PAO1 WT correlated with the amount of PM-localizedGb3 (Fig. 4B). Moreover, pretreatment of the bacteria with theLecA-binder para-nitrophenyl-α-D-galactopyranoside (PNPG) de-creased the invasiveness of PAO1 WT by 68.6% (Fig. 4A, “PNPG-treated”). PNPG selectively prevented LecA, but not StxB, bindingto cells (Fig. S9 C andD). In contrast, cellular invasion by the ΔlecAstrain was not affected by PNPG, because the invasion efficiencywas at a similar level compared with the untreated ΔlecA strain.None of the tested conditions significantly affected the adhesion ofP. aeruginosa PAO1 WT or ΔlecA strains to host cells (Fig. 4C).These findings strongly suggest a major role of LecA and Gb3 asinternalization factors rather than adhesion factors. Moreover, thecholesterol dependency of the lipid zipper formation in GUVs and

Fig. 3. P. aeruginosa induces plasma membrane invaginations depending onLecA. (A) PAO1 (green)-induced membrane invaginations (red) colocalize withactin (blue) in H1299 cells during cellular entry. The white squared area isshown 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 entrystage as shown in A, n = 3 independent experiments; mean ± SEM P value wascalculated by two-tailed, paired t test. (C) Intensity profile across a membraneinvagination in A (yellow line). (D) Actin polymerization is not required forPAO1-induced membrane invaginations. Bacteria still induce membraneinvaginations (arrowheads) and are surrounded by host PM, either covered(zoom 1) or uncovered (zoom 2, recorded 0.5 μm from above the focal planeshown in the overview) by actin in Arp2 knockdown cells (see Fig. S6D forcontrol siRNA-transfected cells). The white squared areas are shown enlargedin the lower panels. (Scale bars, 10 μm and 2.5 μm, respectively.)

Fig. 4. LecA in conjunction with Gb3 promotes efficient cellular invasion ofP. aeruginosa. (A) Invasion of P. aeruginosa strains PAO1 WT and PAO1ΔlecA into untreated or Gb3-depleted (StxB-treated) H1299 cells, and withLecA inhibition (PNPG-treated). Data represent mean values of n ≥ 3 in-dependent experiments. The P value for WT vs. ΔlecA was calculated bya two-tailed, paired t test. (B) Ectopic expression of Gb3 in MDCKII cellsenhances uptake of P. aeruginosa. Invasion of P. aeruginosa WT intountransfected MDCKII cells (wt) and Gb3-synthase-transfected MDCKII cells(clone 1 and clone 2) was evaluated. Both clones differ in their Gb3 ex-pression, which was analyzed by FACS after incubation with StxB-Alexa488(hatched bars). All data represent mean values ± SEM for n ≥ 3 independentexperiments with normalized StxB fluorescence and invasion, to MDCKIIclone 2. The P values were calculated by a two-tailed, paired t test. (C) Ad-hesion of PAO1 WT and ΔlecA to H1299 cells according to the treatments asin A. All data represent mean values ± SEM of n ≥ 4 independent experi-ments, normalized to the WT. (D, Upper) Induced expression of lecA by IPTGin 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. colistrains to H1299 cells. Invasion upon IPTG-induced expression of lecA isstrongly enhanced. Data represent mean values of normalized invasion ±SEM for n ≥ 4 independent experiments. For better visibility, adhesion wasnormalized to non-IPTG-treated E. coli BL21 (DE3).

12898 | www.pnas.org/cgi/doi/10.1073/pnas.1402637111 Eierhoff et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

cells correlated with the efficiency of cellular invasion by PAO1WT. Invasion in cholesterol-depleted cells was significantly de-creased by about 70% compared with control cells (Fig. S10). Thissuggests an overall critical role of cholesterol for the lipid zipper invitro and in cellulo.To test the capacity of LecA as an invasion factor, we explored

the invasiveness of a noninvasive E. coli strain upon inducedexpression of lecA by isopropyl β-D-1-thiogalactopyranoside(IPTG) (25). Remarkably, ectopic expression of lecA in E. coliBL21 (DE3) transduced this strain into an invasive one, evi-denced by a significant increase of invasion into H1299 cells ofabout 340% compared with non-lecA-expressing E. coli strains(Fig. 4D). Interestingly, adherence of lecA-expressing E. coli toH1299 cells was drastically reduced (approximately 20-fold)compared with the non-lecA-expressing E. coli strains (Fig. 4D).Probably LecA functionally interferes with endogenous adher-ence factors of E. coli, which consequently prevents efficientadhesion to H1299 cells. However, these findings highlight thedominant, invasin-like function of LecA. In summary, these datademonstrate that Gb3 and LecA represent key factors for thecellular entry of P. aeruginosa into nonphagocytic cells.

DiscussionBy studying the interaction of P. aeruginosa with synthetic andcellular membranes we have identified a mechanism for bacterialentry that requires host GSLs and bacterial lectins but does notdepend on actin polymerization for membrane engulfment. Thelipid zipper highlights the particular importance of glycolipids inbacterial invasion processes. Numerous host-cell factors havebeen described to interact with P. aeruginosa to orchestrate cel-lular entry and infection of the bacterium. Therefore, it is in-triguing that only two interacting factors, the host-cell GSL Gb3and the bacterial, GSL-binding lectin LecA are sufficient to in-duce the initial processes of the cellular entry of P. aeruginosa.GSLs are key components of lipid rafts and as such are in-

volved in establishing signal transduction platforms in biologicalmembranes (1, 37). Cholesterol has been described to clusterGSLs in the external leaflet of the PM and thereby to stabilizelipid raft domains. We show that the lipid zipper mechanismdepends on lipid raft components such as cholesterol and theGSL Gb3, whereas a nonraft lipid, 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), does not directly promote the in-duction of the lipid zipper. Our results suggest that cholesterolstabilizes LecA-induced domains with a sufficiently high Gb3 den-sity to trigger the lipid zipper, and finally the efficient cellular in-vasion. Therefore, our findings expand the lipid raft conceptbecause (i) they introduce an autonomous role for GSLs (i.e., Gb3)in triggering an endocytic event, uncoupled from cytoplasmic sig-naling processes, and (ii) consider raft domains to serve as mem-brane areas providing sufficiently high GSL densities for lipidzippering bacterial pathogens.The effect of cholesterol on membrane stiffness is complex.

Membrane stiffness is not necessarily decreased in cholesterol-depleted membranes. It could even be increased, at least in choles-terol-depleted cells (38). We found that cholesterol extraction oncells decreased the frequency of the lipid zipper, which could, in linewith ref. 38, point to an increase in membrane stiffness, impairingthe lipid zipper. Following a study on DOPC vesicles, cholesterolhas no (significant) effect on bending rigidity (39). However, wefound a significantly lower occurrence of the lipid zipper in cho-lesterol-free GUVs. Interpreting our results in the context of theabove-mentioned report (39), cholesterol might affect the lipidzipper 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 acidchain lengthand saturation) is critical for theuptake.Therefore, itwillbe interesting to investigate in future studies whether and how dif-ferent Gb3 species affect the uptake of P. aeruginosa.

It is well established that the cellular uptake of Gram-negativebacterial pathogens (e.g., Salmonella and Shigella species) criticallydepends on secreted bacterial effector proteins that induce a re-organization of the host cytoskeleton (40, 41). The expression ofa type-III secretion system (T3SS) by P. aeruginosa and T3SS- andT6SS-secreted effector proteins have been reported to be requiredfor the cellular invasion and are supposed to influence the growth-phase dependent cellular uptake of P. aeruginosa (11, 42–45).However, our data suggest that initial membrane curvature andwrapping of the bacterium by the host PM depends neither ona bacterial secretion system nor on secreted effector proteins.Furthermore, actin polymerization and an active endocytic ma-chinery are not required for the formation of PM invaginations, assuggested by actin inhibition experiments. The GUV experimentsand model calculations clearly demonstrate that the induction ofmembrane invaginations by P. aeruginosa through LecA–Gb3 in-teractions is a thermodynamically favored, ATP-independentprocess that does not depend on cytosolic factors or on bacterialeffectors. However, other bacterial and cellular effector proteinsmight be necessary to accelerate and/or finalize the uptake. Inanalogy to a previous report indicating that actin dynamics drivemembrane scission on StxB-induced membrane invaginations (46),actin could be actively involved in late stages of endocytic pro-cessing of PM invaginations. In addition, because P. aeruginosainduces membrane curvature, curvature-sensing Bin AmphiphysinRvs. (BAR) domain-containing proteins (47) might be recruited tothe entry site and activate signaling pathways to promote the cel-lular uptake of P. aeruginosa.Moreover, our work demonstrates a previously unidentified

pathophysiological relevance for LecA. So far, LecA has been re-lated to bacterial adhesion and biofilm formation (20, 48) andrepresents one of the virulence factors contributing to lung injuryduring infection (24). Our data identify LecA as an invasion factorfor P. aeruginosa, which might contribute to the dissemination of thepathogen within its host organism during infection. This is stronglysupported by the finding that ectopic lecA expression induced aninvasive phenotype of an originally noninvasive E. coli strain.Moreover, this finding shows that LecA does not need to interactwith specific P. aeruginosa factors to fulfill its invasive function.Inhibition of the LecA/Gb3-dependent pathway reduced host-

cell invasion by P. aeruginosa. Therefore, strategies to interferewith the binding of LecA to Gb3 could be the basis for new drugsthat efficiently prevent the dissemination and persistence ofP. aeruginosa. Moreover, many bacterial pathogens express lec-tins on their cellular surface that recognize host GSLs (49). Inthe context of our findings, it is possible that these pathogens useGSL-dependent mechanisms to trigger their uptake by host cells,alternatively or synergistically to protein receptor-based entrypathways, such as the invasin-induced entry of Yersinia (50) orthe internalin-triggered pathway of Listeria spp. (51, 52).Clearly, the lipid zipper mechanism that drives bacterial in-

vasion shares some common features with the GSL-mediatedendocytosis of toxins (32) and viruses (33, 53). Multivalent lectin–glycosphingolipid interactions trigger the formation of plasmamembrane invaginations in the absence of cytosolic proteins (e.g.,coat proteins and actin). However, for bacteria, in contrast tosmaller particles such as toxins and viruses, the adhesion energyresulting from lectin-induced GSL clustering is already sufficientto overcome the energy penalty for membrane bending as aninitial step for their cellular uptake.

Materials and MethodsDetailed materials, methods, and the theoretical model are described in SIMaterials and Methods. Briefly, GUVs were prepared as described (54).H1299 and MDCKII cells stably expressing marker proteins or Gb3 synthasewere 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

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

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

P. aeruginosa PAO1 were produced as described in SI Materials and Methods.Invasion assays were performed as essentially described (11). GSL synthesiswas inhibited by incubation of cells for 3 d with 5 μM PPMP. For LecA in-hibition, soluble LecA (2 μg·mL−1) or bacteria were incubated for 15 min at37 °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 poly-merization was inhibited by latrunculin A treatment (0.1 μM) or knockdownof Arp2 by siRNA. PMS were induced as described (36) and visualized by FM4-64 dye. Cells, GUVs, and P. aeruginosa infection were imaged on a confocalmicroscope (Nikon Eclipse Ti-E with A1R confocal laser scanner, 60× oil ob-jective, N.A. 1.49). Statistical testing was performed with Excel with data

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

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

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 Biol166(5):743–753.

8. Chi E, Mehl T, Nunn D, Lory S (1991) Interaction of Pseudomonas aeruginosa withA549 pneumocyte cells. Infect Immun 59(3):822–828.

9. Emam A, Carter WG, Lingwood C (2010) Glycolipid-dependent, protease sensitiveinternalization of Pseudomonas aeruginosa into cultured human respiratory epithe-lial cells. Open Microbiol J 4:106–115.

10. Fleiszig SM, Zaidi TS, Fletcher EL, Preston MJ, Pier GB (1994) Pseudomonas aeruginosainvades 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 andits survival within HeLa cells. Infect Immun 69(7):4398–4406.

12. Pereira SH, Cervante MP, Bentzmann S, Plotkowski MC (1997) Pseudomonas aerugi-nosa entry into Caco-2 cells is enhanced in repairing wounded monolayers. MicrobPathog 23(4):249–255.

13. Plotkowski MC, Meirelles MN (1997) Concomitant endosome-phagosome fusion andlysis of endosomal membranes account for Pseudomonas aeruginosa survival in hu-man 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 endothelialcells. Infect Immun 62(12):5456–5463.

15. Fleiszig SM, et al. (1996) Relationship between cytotoxicity and corneal epithelial cell in-vasion 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 Pseudo-monas aeruginosa to repairing airway epithelium. Eur Respir J 13(6):1301–1309.

17. Pier GB, Grout M, Zaidi TS (1997) Cystic fibrosis transmembrane conductance regu-lator is an epithelial cell receptor for clearance of Pseudomonas aeruginosa from thelung. Proc Natl Acad Sci USA 94(22):12088–12093.

18. Pielage JF, Powell KR, Kalman D, Engel JN (2008) RNAi screen reveals an Abl kinase-dependent 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 aeruginosaadherence 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 lectininteractions with the ABO(H) and P blood group glycosphingolipid antigens maycontribute to the broad spectrum adherence of Pseudomonas aeruginosa to humantissues in secondary infections. Glycoconj J 11(5):414–417.

21. Glick J, Garber N (1983) The intracellular localization of Pseudumonas aeruginosalectins. J Gen Microbiol 129:3085–3090.

22. Lanne B, Cîopraga J, Bergström J, Motas C, Karlsson KA (1994) Binding of thegalactose-specific Pseudomonas aeruginosa lectin, PA-I, to glycosphingolipids andother glycoconjugates. Glycoconj J 11(4):292–298.

23. Gilboa-Garber N (1972) Inhibition of broad spectrum hemagglutinin from Pseudo-monas 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-inducedlung 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-seriesglycosphingolipids displayed by Pseudomonas aeruginosa lectin I. J Mol Biol 383(4):837–853.

26. Cossart P, Sansonetti PJ (2004) Bacterial invasion: The paradigms of enteroinvasivepathogens. 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 epi-thelial cells through surface area regulation in response to osmotic stress. BiochimBiophys Acta 1833(3):712–722.

31. Dai J, Sheetz MP (1999) Membrane tether formation from blebbing cells. Biophys J77(6):3363–3370.

32. Römer W, et al. (2007) Shiga toxin induces tubular membrane invaginations for itsuptake into cells. Nature 450(7170):670–675.

33. Ewers H, et al. (2010) GM1 structure determines SV40-induced membrane in-vagination and infection. Nat Cell Biol 12(1):11–18; sup pp 11-12.

34. Pollard TD, Blanchoin L, Mullins RD (2000) Molecular mechanisms controlling actinfilament 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 bylatrunculin A. FEBS Lett 213(2):316–318.

36. Lingwood D, Ries J, Schwille P, Simons K (2008) Plasma membranes are poised foractivation of raft phase coalescence at physiological temperature. Proc Natl Acad SciUSA 105(29):10005–10010.

37. Simons K, Toomre D (2000) Lipid rafts and signal transduction. Nat Rev Mol Cell Biol1(1):31–39.

38. Byfield FJ, Aranda-Espinoza H, Romanenko VG, Rothblat GH, Levitan I (2004) Cho-lesterol depletion increases membrane stiffness of aortic endothelial cells. Biophys J87(5):3336–3343.

39. Gracia RS, Bezlyepkina N, Knorr RL, Lipowsky R, Dimova R (2010) Effect of cholesterolon the rigidity of saturated and unsaturated membranes: Fluctuation and electro-deformation analysis of giant vesicles. Soft Matter 6(7):1472–1482.

40. McGhie EJ, Brawn LC, Hume PJ, Humphreys D, Koronakis V (2009) Salmonella takescontrol: 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 Shigellaeffectors. Nat Rev Microbiol 6(1):11–16.

42. Galán JE, Collmer A (1999) Type III secretion machines: Bacterial devices for proteindelivery into host cells. Science 284(5418):1322–1328.

43. Jendrossek V, et al. (2003) Apoptotic response of Chang cells to infection withPseudomonas aeruginosa strains PAK and PAO-I: Molecular ordering of the apoptosissignaling 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 BiochemPharmacol 152:79–92.

45. Sana TG, et al. (2012) The second type VI secretion system of Pseudomonas aerugi-nosa strain PAO1 is regulated by quorum sensing and Fur and modulates in-ternalization in epithelial cells. J Biol Chem 287(32):27095–27105.

46. Römer W, et al. (2010) Actin dynamics drive membrane reorganization and scission inclathrin-independent endocytosis. Cell 140(4):540–553.

47. Peter BJ, et al. (2004) BAR domains as sensors of membrane curvature: The amphi-physin BAR structure. Science 303(5657):495–499.

48. Diggle SP, et al. (2006) The galactophilic lectin, LecA, contributes to biofilm de-velopment 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 Lis-teria 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 isthe receptor for internalin, a surface protein required for entry of L. monocytogenesinto epithelial cells. Cell 84(6):923–932.

53. Rydell GE, Svensson L, Larson G, Johannes L, Römer W (2013) Human GII.4 norovirus VLPinduces membrane invaginations on giant unilamellar vesicles containing secretor genedependent 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 giantphospholipid 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 Eierhoff et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021