Oligomeric lipoprotein PelC guides Pel polysaccharideexport across the outer membrane ofPseudomonas aeruginosaLindsey S. Marmonta,b, Jacquelyn D. Richc, John C. Whitneya,b,2, Gregory B. Whitfielda,b, Henrik Almbladc,Howard Robinsond, Matthew R. Parseke, Joe J. Harrisonc, and P. Lynne Howella,b,1

aProgram in Molecular Structure & Function, The Hospital for Sick Children, Toronto, ON, Canada M5G 0A4; bDepartment of Biochemistry, University ofToronto, Toronto, ON, Canada M5S 1A8; cDepartment of Biological Sciences, University of Calgary, Calgary, AB, Canada T2N 1N4; dPhoton Science Division,Brookhaven National Laboratory, Upton, NY 11973-5000; and eDepartment of Microbiology, University of Washington, Seattle, WA 98195-7242

Edited by Scott J. Hultgren, Washington University School of Medicine, St. Louis, MO, and approved January 31, 2017 (received for review August 18, 2016)

Secreted polysaccharides are important functional and structuralcomponents of bacterial biofilms. The opportunistic pathogen Pseu-domonas aeruginosa produces the cationic exopolysaccharide Pel,which protects bacteria from aminoglycoside antibiotics and contrib-utes to biofilm architecture through ionic interactions with extracel-lular DNA. A bioinformatics analysis of genome databases suggeststhat gene clusters for Pel biosynthesis are present in >125 bacterialspecies, yet little is known about how this biofilm exopolysaccharideis synthesized and exported from the cell. In this work, we character-ize PelC, an outer membrane lipoprotein essential for Pel production.Crystal structures of PelC from Geobacter metallireducens and Para-burkholderia phytofirmans coupled with structure-guided disulfidecross-linking in P. aeruginosa suggest that PelC assembles into a12- subunit ring-shaped oligomer. In this arrangement, an aromaticbelt in proximity to its lipidation site positions the highly electroneg-ative surface of PelC toward the periplasm. PelC is structurally similarto the Escherichia coli amyloid exporter CsgG; however, unlike CsgG,PelC does not possess membrane-spanning segments required forpolymer export across the outer membrane. We show that the mul-tidomain protein PelB with a predicted C-terminal β-barrel porin lo-calizes to the outer membrane, and propose that PelC functions as anelectronegative funnel to guide the positively charged Pel polysaccha-ride toward an exit channel formed by PelB. Together, our findingsprovide insight into the unique molecular architecture and exportmechanism of the Pel apparatus, a widespread exopolysaccharidesecretion system found in environmental and pathogenic bacteria.

biofilms | exopolysaccharides | Pseudomonas aeruginosa | PEL |X-ray crystallography

Exopolysaccharides are crucial for bacterial biofilm develop-ment and pathogenesis. Studies have shown that deletion of

genes responsible for polysaccharide synthesis in Pseudomonasaeruginosa (1, 2), Escherichia coli (3, 4), Yersinia pestis (5), Staphy-lococcus aureus (6), Bordetella pertussis (7), and Burkholderiacepacia (8) abolish biofilm formation and/or significantly com-promise bacterial virulence (4, 9, 10). Although the importanceof exopolysaccharides is widely accepted for these and otherpathogens, the exact mechanisms underlying their biosynthesisand export remain poorly understood.The opportunistic pathogen P. aeruginosa is genetically capable

of synthesizing at least three distinct exopolysaccharides: alginate,PSL, and PEL (11). Each of these polysaccharides contributes tothe ability of the bacterium to cause chronic infections in the lungsof cystic fibrosis (CF) patients (12, 13). Within that environment,P. aeruginosa undergoes significant genetic diversification, leadingto phenotypic heterogeneity (14). For example, some P. aeruginosaCF isolates display a mucoid colony morphology, which is causedby the overproduction of alginate (15); others form small wrinklycolonies, which are characteristic of PSL and PEL overproduction(13, 16). PSL and PEL can serve redundant functions in the

biofilm structure, and together or individually act as structuralscaffolds to maintain biofilm integrity (17, 18).PEL is a cationic exopolysaccharide composed of partially

deacetylated 1,4-linked N-acetylglucosamine and N-acetylgalactos-amine (19), which has been shown to initiate and maintain bacterialcell–cell interactions and provide resistance to aminoglycoside an-tibiotics (17). The capacity for PEL to cross-link extracellular DNAwithin the biofilm and interact with the anionic host-derivedpolymers mucin and hyaluronan suggest an important role for thispolysaccharide in disease pathogenesis (19). Previous work de-fined a seven-gene operon, pelABCDEFG, essential for PELbiosynthesis in P. aeruginosa (9, 20). Bioinformatics and limitedstructural and biochemical analyses have begun to shed light onthe enzymes and chemistry underlying PEL polymerization anddeacetylation (21–24); however, details surrounding secretion ofthe mature Pel polysaccharide across the outer membrane (OM)remain largely unknown.Relative to better-studied exopolysaccharide biosynthetic systems,

one of the curious features of the PEL system is the OM-associatedprotein, PelC. In this study, we identified PelC orthologs in over 125bacterial species linked to other pel genes; this allowed us to solve

Significance

Pseudomonas aeruginosa biofilms are exceedingly difficult toeradicate once established. This resilience is facilitated, in part,by the secretion of polysaccharides that contribute to biofilmstructural integrity. The cationic exopolysaccharide PEL playsan important role in disease pathogenesis; however, the mec-hanisms underlying its biosynthesis are poorly understood. Inthis work, we identify the pel operon in more than 125 pro-teobacteria, demonstrating that its distribution was previouslyunderestimated. We show that the essential outer membrane-anchored protein PelC forms a 12-subunit ring with an elec-tronegative surface that we propose guides PEL toward themembrane-embedded secretion channel. Our work providesinsight into a widespread outer membrane infrastructure un-observed in any other currently identified polysaccharidebiosynthetic apparatus.

Author contributions: L.S.M., J.D.R., J.C.W., G.B.W., M.R.P., J.J.H., and P.L.H. designed research;L.S.M., J.D.R., J.C.W., G.B.W., H.A., H.R., and J.J.H. performed research; L.S.M., J.D.R., J.C.W.,G.B.W., H.A., H.R., M.R.P., J.J.H., and P.L.H. analyzed data; and L.S.M., J.J.H., and P.L.H. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.pdb.org (PDB ID codes 5T11, 5T10, and 5T0Z).1To whom correspondence should be addressed. Email: howell@sickkids.ca.2Present address: Department of Biochemistry and Biomedical Sciences, McMaster Uni-versity, Hamilton, ON, Canada L8S 4K1.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1613606114/-/DCSupplemental.

2892–2897 | PNAS | March 14, 2017 | vol. 114 | no. 11 www.pnas.org/cgi/doi/10.1073/pnas.1613606114

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1

X-ray crystal structures of PelC orthologs to reveal a 12-subunit ringarrangement with a central opening of over 30 Å. Functional studiesreveal PelC is localized to the inner leaflet of the OM, positioningthe electronegative surface of PelC toward the periplasm. Wedemonstrate that the negative charge within the PelC pore is criticalfor biofilm formation and propose that PelC acts as a molecularfunnel to guide the positively charged PEL toward the β-barrelPelB pore.

ResultsThe pel Operon Is Widely Distributed in Proteobacteria. The Pel poly-saccharide was first identified in P. aeruginosa (25), and has beenstudied nearly exclusively using this model organism. Sequence-based searching has previously identified putative pel genes inGeobacter, Burkholderia, and Ralstonia species (9), and the Pseu-domonas Orthologous Group classification system (26) suggeststhat pel operons may be widespread in Pseudomonads. BecausePelC is short (172 residues), has no identifiable conserved domain,and is essential for PEL production in P. aeruginosa, we reasonedthat its amino acid sequence might be ideal for identifying pel geneclusters in other bacteria.BLASTP searching of the Pseudomonas, Burkholderia, and Na-

tional Center for Biotechnology Information (NCBI) databasesrevealed PelC orthologs in 128 bacterial species. A total of 126nonredundant PelC orthologs had >70% coverage and >35% iden-tity to P. aeruginosa PelC (PelCPa), and were linked to at least one ormore genes with similarities to those encoding the P. aeruginosa Pelapparatus. The majority (∼54%) were linked with operons that hadcomplete synteny with the archetypal P. aeruginosa pelABCDEFGoperon. A Bayesian analysis of operon-linked PelC orthologs sug-gests that the Pel secretion system is distributed in at least 45 bac-terial genera, but especially in beta- and gammaproteobacteria (Fig.S1). These results suggest that distribution of pel gene clusters hasbeen previously underestimated among bacteria. Because the ge-netic basis for biofilm formation in the majority of cultured bacteriahas not been defined, we can now begin to determine the contri-bution of PEL to biofilm formation in these bacteria, which includeenvironmental and opportunistically pathogenic species.

PelC Assembles into a Ring-Shaped Dodecamer. Previously, it wasproposed that the amphipathic C-terminal α-helix of PelC oligo-merizes to form a membrane-spanning channel in the OM in amanner analogous to the capsular polysaccharide export proteinWza from E. coli (24). However, more recent bioinformaticsanalysis of PelC suggests that the C-terminal helix is not membranespanning and instead is an integral part of its predicted α/β fold(11). To resolve these discrepancies and shed light on the functionof the protein, we undertook structural studies on PelC. BecausePelCPa was recalcitrant to crystallization, we determined thestructure of matureGeobacter metallireducens PelC minus its signalsequence (NHis6-PelCGm

18–179, hereafter referred to as PelCGm) to2.25-Å resolution (Table S1). PelCGm is 39% identical to PelCPa.PelCGm has a mixed α/β fold (Fig. 1A), with four β-strands

(β1 and β5–β7) forming an antiparallel β-sheet on one face of theprotein, whereas helix α1 and the C-terminal helix α3 form theopposite face (Fig. S2). PelCGm crystallizes with four molecules inthe asymmetric unit, with two molecules arranged in an antiparallelorientation, and two of these pairs packing together via theirβ-sheets (Fig. 1B). Because homodimerization of a pair of mole-cules is facilitated by strand β1, which is formed by residues of thevector-encoded thrombin cleavage site of the histidine-tag (Fig.S3A), we reasoned that this tetramer is unlikely to represent a bi-ologically relevant arrangement of PelC. Therefore, to determinethe structure of PelC in the absence of any nonnative amino acids,we removed the His6-tag from PelCGm. Because untagged PelCGmfailed to crystallize, we explored the use of alternative orthologs anddetermined the structure of untagged PelC from Paraburkholderiaphytofirmans (PelCPp

28-183 hereafter referred to as PelCPp) to 2.1-Åresolution. PelCPp shares 41% identity with PelCPa.PelCPp and a single amino acid variant PelCPp

L103M werecrystallized. Although the L103M variant was designed to facilitate

structure determination using the single wavelength anomalous dif-fraction technique, both structures were determined by molecularreplacement, using PelCGm as a search model. PelCPp

L103M crys-tallized in space group C2 with 12 molecules per asymmetric unit(Fig. S3B). Although the native PelCPp crystallized in space groupP6 with four molecules per asymmetric unit (Fig. S3C, gray),analysis of the crystal packing reveals that both the wild-type andL103M variant form the same ring-shaped oligomer. Each ring iscomprised of 12 PelC molecules with an overall diameter of 120 Åand a 32-Å diameter pore (Fig. 1C). The ring arrangement issimilar to a funnel without its stem, and as such, displays bothconvex and concave surfaces (Fig. 1E). Individual subunits ofPelCGm and PelCPp are very similar, with a rmsd of 0.881 Å over105 Cα between the two most complete chains (Fig. S2).To determine if the dodecameric arrangement observed might

represent a biologically relevant state, we examined the location ofconserved residues across the 126 nonredundant PelC orthologsequences using MUSCLE alignment (27). In support of the bi-ological relevance of our oligomeric structure, we found that themost highly conserved residues line the pore and form the pro-tomer–protomer interface (Fig. 1 D and E). In addition, a searchof the Protein Data Bank (www.rcsb.org) using DALI (28) revealsthat PelC most closely resembles the middle domain of Vibrioalginolyticus FlgT (FlgT-M), a protein required for anchoring theflagellum to the pole of the cell (Fig. S4A) (29, 30), and theperiplasmic segment of the curli amyloid export protein CsgGfrom E. coli (Fig. S4B). Both of these proteins form higher-orderoligomeric structures, providing structural precedents for the PelCarrangement observed in this study.

Structure-Guided Disulfide Cross-Linking Stabilizes a PelC Multimer inP. aeruginosa. In vitro characterization of mature PelCPa, whichlacks the native lipidation site, indicates that the protein is mo-nomeric in solution (Fig. S5A). PelCGm and PelCPp similarly do notform a dodecamer in solution (Fig. S5 B and C). We hypothesizedthat PelC may require the presence of its lipid anchor and the OMfor oligomerization (31, 32). To address this question, structure-guided disulfide cross-linking was performed in P. aeruginosa todetermine whether the protein adopted a higher-order oligomericstate in vivo. For this and subsequent experiments, we used theP. aeruginosa PAO1 ΔwspF Δpsl PBADpel strain, which was engi-neered for arabinose-inducible expression of the Pel poly-saccharide (23) (Fig. S6A). To investigate structure–functionrelationships of PelC in vivo, an in-frame deletion of pelC was in-troduced into the ΔwspF Δpsl PBADpel strain. This mutationabolished arabinose-inducible biofilm and pellicle formation (Fig.S6A). Next, araC-PBAD was fused directly to the pelC ORF. Sub-sequently, the ΔwspF Δpsl PBADpel ΔpelC strain was complemented

Fig. 1. Overall structure of PelC. (A) Topology model of the PelC structure.The loops that face the center of the PelCPp pore are indicated in red.(B) Cartoon representation of the tetrameric arrangement of PelCGm in theasymmetric unit, with one protomer colored by secondary structure for clarity.(C) Cartoon representation of the dodecameric arrangement of PelCPp foundin the asymmetric unit. In A–C, N and C represent the N and C termini, re-spectively. (D and E) Surface representations of the convex and protomersurfaces of PelCPp colored by residue conservation. Conservation, as defined bythe ConSurf server, is displayed from magenta (conserved) to cyan (variable).

Marmont et al. PNAS | March 14, 2017 | vol. 114 | no. 11 | 2893

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1

with wild-type pelC on this miniTn7 vector, or with a panel ofderived double-cysteine mutants designed to allow for in-termolecular disulfide cross-linking between PelC protomers (SIMaterials and Methods). We reasoned that if PelC forms an olig-omer as suggested by our crystal structure, cysteine cross-linkingwould allow for covalent trapping of this species (18). We foundthat an N116C/L118C double-mutant (hereafter referred to asPelCDSB; Fig. 2A) resulted in the formation of a single oligomer asseen by Western blot (Fig. 2B and Fig. S6B). This oligomer wassensitive to treatment with a reducing agent, indicating that itsformation was due to the engineered cysteines. Importantly, com-plementation of our pelC deletion strain with pelCDSB restoredbiofilm formation, indicating that the crosslinked protein is func-tional (Fig. S6A). To examine the cross-linked PelC further, wepurified a C-terminal histidine-tagged PelCDSB from P. aeruginosa.Preliminary analysis using negative-stain electron microscopy re-veals a uniform ring-like complex ∼120 Å in diameter (Fig. S7),consistent with our crystal structures. Taken together, these datasuggest that PelC forms a functional multimer in vivo.

PelC Localizes to the Inner Leaflet of the Outer Membrane. Previously,PelC has been determined to localize to the OM (24); however,whether PelC resides on the periplasmic or extracellular leaflet ofthis membrane is unknown. We therefore examined the suscep-tibility of PelC in intact P. aeruginosa cells to proteinase K treat-ment, which would shear any extracellular regions from themembrane. Western blot analysis of whole cells revealed that PelClevels after treatment were unchanged relative to untreated cells(Fig. 3A). In contrast, treatment of lysed cells resulted in signifi-cant degradation of PelC. Because the lipoprotein PilF exhibited asimilar degradation profile, our data suggests that PelC, like PilF,localizes to the periplasmic leaflet of the OM (33).

Residues on the Membrane Proximal Convex Surface of PelC Are Essentialfor Function.The PelCPp structure suggests that the convex surface ofthe ring (Fig. 1E) is proximal to the membrane. This surface notonly exhibits an overall positive charge allowing the protein subunitsto interact with negatively charged phosphate groups of OM phos-pholipids (Fig. 3B), but would also contain the lipid anchor found atthe N terminus of each PelC protomer. The structure also revealsthat spatially adjacent to the N terminus is a conserved tryptophan(W149 in PelCPa), which in the PelC oligomer forms an aromaticbelt (Fig. 3C). Tryptophan residues are commonly found at mem-brane interfaces, further supporting our hypothesis that the convexsurface is positioned toward the membrane (34, 35). To test this, wegenerated lipidation site variants, C19S, C19A, and C19G, and al-anine point variants of W149, R151, and E121 (Fig. 3D) and ex-amined their effects on biofilm formation (Fig. 3E). R151 is in closeproximity to W149 and also extends away from the convex surface,whereas E121 is found on an adjacent loop and is not predicted toface the membrane. Although all of the C19 variants ablated biofilmformation, we are unable to attribute this to loss of lipidation as anymutation of this residue destabilizes the protein. This destabilization

may have occurred as a result of the inability of the localization oflipoproteins (Lol) machinery, a system for lipoprotein sorting andmembrane localization that is highly conserved in Gram-negativebacteria, to properly recognize and target PelC to the OM (36). Theresults for the C19S variant are shown in Fig. 3E, Bottom. TheW149, R151, and E121 alanine variants produce PelC protein atcomparable levels to wild type, but each mutation has a differenteffect on biofilm formation. The W149A variant was unable to forma biofilm, whereas R151A displayed an intermediate phenotype.The E121A variant had no effect on biofilm formation (Fig. 3E, Topand Middle). Comparable results were observed for site-directedmutations engineered into the chromosome of P. aeruginosa PA14, astrain that naturally produces a PEL-dominant biofilm matrix (Fig.S8). Together with our structural analysis, these data suggest that inaddition to the C19 lipid anchor, W149 may play a role in posi-tioning PelC at the OM, possibly acting as a second point of contactwith the membrane. We propose that in the absence of W149, PelCwould then rely on one anchor point, C19; this may not be enoughto maintain PelC in its functional state. To test this finding, we usedthe covalently linked functional PelCDSB oligomer (Fig. S6A), andgenerated the W149A mutation in this background. We found thatPelCDSB

W149A was able to oligomerize, as well as form a biofilm;however, surface attachment was impaired (Fig. S9), which suggeststhat stabilizing the oligomeric state of the protein in the cross-linkedPelCDSB can partially compensate for the absence of this criticaltryptophan residue.

A Negatively Charged Pore Is Critical for PelC Function. Our datasuggests that the convex face of the PelC dodecamer is orientedtoward the membrane; this would position the highly electroneg-ative concave surface of PelC toward the periplasm (Fig. 4A). Totest the functional significance of this negatively charged surface,we generated a series of lysine variants (Fig. 4B) and examinedtheir effects on biofilm formation (Fig. 4C). Select solvent-exposed

Fig. 2. PelCPaDSB forms a multimer in vivo. (A) Cartoon representations depict-

ing the model of PelCPa and residues targeted for mutagenesis; each cysteinefrom a single protomer (light blue) was designed to interact with the corre-sponding cysteine partner from the neighboring molecule (red). (B) Western blotanalysis using PelC-specific antibodies of whole-cell lysates of the indicatedP. aeruginosa strains in the presence (+) or absence (−) of reducing agent (DTT).

Fig. 3. PelC localization and analysis of membrane proximal residues. (A) Thesusceptibility of PelC to proteinase K (PK) treatment in whole cells (intact) andlysed cells (lysed) was observed using PelC-specific antibodies. PilF is an innerleaflet OM lipoprotein. (+) and (−) indicate the presence or absence of PK.(B) Surface representation of PelCPp with charge distribution of the convex sur-face ranging from −12 kT/e (red) to +12 kT/e (blue). (C) Cartoon representation ofPelCPp with the aromatic tryptophan belt displayed in stick representation (red).(D) Cartoon representation of the PelCPa model with conserved residues targetedfor mutation shown as sticks (red). (E, Top) Surface attachment determined usingthe microtiter dish biofilm assay; error bars indicate the SEM of 10 independenttrials performed in triplicate. Statistical significance was evaluated using one-wayANOVA with Bonferroni correction; **P < 0.0001 compared with the wild-typestrain. (E, Middle) Standing biofilm assay; the black arrow indicates the locationof the biofilms. (E, Bottom) Western blot analysis of whole-cell lysates of theindicated P. aeruginosa strains using PelB- and PelC-specific antibodies.

2894 | www.pnas.org/cgi/doi/10.1073/pnas.1613606114 Marmont et al.

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1

residues were mutated to lysine, and glutamine residues were alsomutated to glutamate. Of the residues tested, only mutation ofD119, which is positioned in the central pore, abrogated biofilmformation (Fig. 4C, Top and Middle). This defect is not due todestabilization of PelC or disruption of oligomerization, becauseprotein levels are comparable to the parental strain (Fig. 4C,Bottom) and the oligomer still forms in the PelCDSB

D119K mutant(Fig. S10). Structural modeling suggests that a lysine in this posi-tion would not occlude the pore. In contrast, the biofilms of variantsdistal to the pore were indistinguishable from the parental strain. Thereduced surface attachment phenotypes of Q46K, Q46E, and Q49E,together with the biofilm-deficient D119K mutant, suggest that theoverall surface charge of PelC proximal to the pore is important, andthat the charge within pore is critical for PelC function. Collectively,our data suggests that the cationic PEL may be extruded through thenegatively charged pore formed by PelC. In this model, PEL is morelikely to encounter residues within the pore than at the extremity ofthe concave surface, which may account for the surface attachmentdefects for proximal but not distal variants.

PelB Is an Outer Membrane Protein with a Predicted Transmembraneβ-Barrel Domain. PelC shares structural similarity to the periplasmicportion of the E. coli amyloid export protein CsgG (Fig. S4B).Each CsgG monomer also possesses four β-strands, which multi-merize to form a single 36-stranded transmembrane β-barrel. Incontrast, our crystal structures suggest that PelC lacks membrane-spanning segments that could serve as a polysaccharide exportchannel. Therefore, we reasoned that another Pel protein, mostlikely PelB, encodes the integral OM exporter. PelB has beenshown to be essential for PEL synthesis, because deletion of pelBcauses a severe biofilm deficiency (17). Structure prediction algo-rithms indicate that the N terminus of PelB contains a tetra-tricopeptide repeat (TPR) domain, whereas the C terminus ispredicted to form an OM β-barrel porin (11, 37). To test this

prediction, we performed subcellular fractionation studies in ourPEL-inducible strain and found that PelB, like PelC, partitions tothe OM (Fig. 5A). Because our strain does not overexpress theassembly machinery for protein transport across the IM or forβ-barrel assembly, as anticipated some PelB was also found in thecytoplasm and IM. We next sought to test whether PelB possesses aβ-barrel porin, which would allow PEL to traverse the OM.Transmembrane β-barrel porins are generally resistant to SDS de-naturation, which results in aberrant migration of these proteinsduring SDS/PAGE. This resistance to denaturation can be over-come by heat treatment before electrophoresis, resulting in dena-tured protein that migrates at its expected molecular weight (38,39). Consistent with the hypothesis that PelB contains a trans-membrane β-barrel porin, we observed a difference in molecularweight between untreated and heat-treated samples (Fig. 5B).

Fig. 4. A negatively charged PelC pore is required for biofilm formation.(A) Surface representation of PelCPp with charge distribution of the concavesurface ranging from −12 kT/e (red) to +12 kT/e (blue). (B) Cartoon represen-tation of the PelCPa model with charged residues targeted for mutation shownas sticks (red). (C, Top) Surface attachment determined using the microtiterdish biofilm assay; error bars indicate the SEM from 10 independent trialsperformed in triplicate. Statistical significance was evaluated using one-wayANOVAwith Bonferroni correction; *P < 0.01, **P < 0.0001 comparedwith thewild-type strain. (C, Middle) Standing biofilm assay; the black arrow indicatesthe location of the biofilm. (C, Bottom) Western blot analysis of whole-celllysates of the indicated P. aeruginosa strains using PelC-specific antibodies.NSB, nonspecific band.

Fig. 5. PelB is an outer membrane protein with a predicted β-barrel porindomain. (A) Subcellular fractionation of the cytoplasmic (C), inner membrane(IM), periplasmic (P), and outer membrane (OM) fractions of PAO1 ΔwspF ΔpslPBADpel. PilP and PilQ serve as inner and outer membrane controls, respectively.(B) Western blot analysis of the gel mobility assay in PAO1 ΔwspF Δpsl PBADpelusing PelB-specific antibodies. Samples of cell lysate were either untreated(25 °C) or heat treated (80 °C). The molecular weight is indicated on the left.

Fig. 6. Model of the Pel polysaccharide outer membrane secretion complex.(A) PEL becomes positively charged following deacetylation by PelA. Thepolymer is drawn toward the electronegatively charged concave surface ofPelC, guiding PEL toward the exit channel formed by PelB. The region betweenthe β-barrel and the TPR domains of PelB is proposed to thread through thePelC pore. Figure is not to scale. (B) Cartoon representations of PelCPp (Top,blue) modeled with 16-stranded β-barrel PgaA (purple; PDB ID code 4Y25) andCsgG (Bottom) (PDB ID code 4UV3). CsgG forms a 36-stranded transmembraneβ-barrel (purple) from the β-strands of nine protomers. The inner diameter ofeach β-barrel is indicated at the top, and the overall diameter is marked below.The diameter of the PelCPp aromatic ring is also indicated.

Marmont et al. PNAS | March 14, 2017 | vol. 114 | no. 11 | 2895

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1

DiscussionIn this study, our phylogenetic analysis using PelC revealed that thedistribution of the pel operon has been previously underestimated.The operon was identified in more than 125 bacterial species,suggesting that PEL may play an important role in biofilm forma-tion in a wide range of bacteria, especially in the beta- and gam-maproteobacteria. Herein we have determined the structures of twoPelC orthologs, PelCPp and PelCGm. PelCPp revealed a dodecamericring arrangement, which we have validated as biologically relevantusing structure-guided disulfide cross-linking in P. aeruginosa. Se-quence- and structure-based comparisons of PelC orthologs haveproven useful in exposing conserved features that are critical for thebiological function of this protein. Our in vivo mutagenesis studieshave not only identified residues that appear essential for the cor-rect positioning of the PelC oligomer against the OM (e.g., C19 andW149), but we have also shown that a negatively charged PelC poreis critical for protein function and biofilm formation.PelC is structurally similar to two other multimeric ring-forming

proteins. FlgT is a periplasmic protein whose middle domain,FlgT-M, closely resembles PelC (29). FlgT anchors the flagellainto the membrane in V. alginolyticus, because ΔflgT strains re-lease the fully formed flagella into the growth medium (30). Incontrast, the ability of FlgT to form a ring is lost when the FlgT-Mdomain is deleted. This ring assembly is required for interactionswith MotX and MotY, which form an additional ring structurethat surrounds FlgT and are required for incorporation of thestator unit into the basal body and its stabilization (29). Theperiplasmic region of the curli amyloid secretion channel formedby CsgG in E. coli also shares significant structural similarity toPelC (31, 40). Nonameric CsgE and CsgG interact to form achamber enclosing discrete polypeptide chains to prevent the ag-gregation of nascent curli amyloid (41). These structurally similaroligomeric proteins lend support to the ring forming capacity ofthe PelC fold, and suggest that the arrangement observed in thecrystal structures is essential for PelC function.Considering our structural and functional analyses, we propose

the following model. Once newly synthesized PEL enters theperiplasm, it is deacetylated by the periplasmic deacetylase PelA,which confers a net positive charge on the polymer (19, 23). Thiscationic polymer then traverses the periplasm to reach the OMsecretion apparatus. Like the AlgK and PgaA proteins found inthe alginate and PNAG secretion systems, respectively, we expectthat the N-terminal TPRs of PelB exist in the periplasm. In-terestingly, between the predicted N-terminal TPR and C-terminalβ-barrel domains of PelB, is a stretch of ∼120 residues, whichprediction algorithms can not confidently model as regular sec-ondary structure. In our model, PelC assembles around this regionof PelB, which would thread through the 30-Å pore created uponPelC oligomerization. We hypothesize that the highly electroneg-ative surface of the PelC ring attracts cationic PEL toward its pore,funneling the polymer toward the transmembrane β-barrel domainof PelB, which is predicted to contain 16 β-strands (Fig. 6A). Toassess whether PelC could accommodate a β-barrel of this sizewithin the dimensions of the aromatic belt, we modeled the crystalstructure of the 16-stranded β-barrel of PgaA at the convex surfaceof the PelCPp structure (42). PgaA has a similar domain archi-tecture to PelB and is responsible for the export of poly-β-1,6-N-acetylglucosamine in E. coli (4, 37). The diameter of PgaA is ∼40 Å,suggesting that PelC could multimerize around a β-barrel of thissize, because the diameter of the tryptophan ring on the convexface is ∼60 Å (Fig. 6B). Although we have been unable to dem-onstrate an interaction between PelC and PelB, this might be re-flective of the size of the interface. It may be that the interfacebetween the two proteins is mediated by only a few residues. It ispossible that W149 could be involved in stacking interactions with

aromatic residues in PelB located at the OM-periplasmic interface.PgaA has at least 12–14 aromatic residues at the OM–periplasmicinterface, and perhaps PelB displays a comparable number. An-other possibility includes the linker region of PelB, which mayinteract with PelC. At present, programs for β-barrel predictionare not accurate enough for us to be able to model PelB with anyconfidence, and without precise structural information we can onlyspeculate about the nature of the interaction surface.PelC is essential for PEL biosynthesis; however, analogous

protein folds are not found in any other currently identified exo-polysaccharide secretion systems. Indeed, our phylogenetic anal-ysis did not reveal any PelC orthologs outside of the context of apel gene cluster. Future studies to ascertain more about PELbiosynthesis and export, as well as exploring the protein interac-tions of this system, will benefit our understanding of other similarexopolysaccharide secretion systems and provide a platform forfuture therapeutic target development.

Materials and MethodsDetailed methods are described in SI Materials and Methods.

Bacterial Strains, Plasmids, and Growth Conditions. All P. aeruginosa strainswere derived from PAO1 (43). P. aeruginosa mutant and complementedstrains were generated using allelic exchange and mini-Tn7 mutagenesis,respectively, as described previously (44, 45). A detailed description ofstrains, plasmids, and primers used in this study can be found in SI Materialsand Methods and Tables S2 and S3.

Crystallization and Structure Determination. The portion of each geneencoding mature pelC without its lipidation site from G. metallireducens andP. phytofirmans were cloned into pET-28a, which fuses an NHis6-tag to theprotein to facilitate affinity purification. The vector was transformed intoE. coli BL21 CodonPlus (DE3) for expression. Affinity-purified protein wassubjected to size-exclusion chromatography using a HiLoad 16/60 Superdex200 gel filtration column (GE Healthcare). Crystallization screens were set up byhand using the hanging-drop vapor-diffusion technique at room temperature.Crystals of PelCGm appeared in 10 d, and for PelCPp, crystals grew in less than5 d. Se-Met derivative crystals were prepared as described above for the nativecrystals, with the exception that they were expressed in E. coli B834 Met− cellsas described previously (46).

Biofilm Assays. Standing biofilm assays were performed as described previously(22). Briefly, overnight cultures were grown in no-salt lysogeny broth (NSLB),and normalized to an OD600 of 0.5. These cultures were diluted 1/100 in 3 mLNSLB containing 0.5% (wt/vol) L-(+)-ara and 30 μg/mL gentamycin where re-quired and left undisturbed at 25 °C for 48 h. Diluted cultures were also ali-quoted into a 96-well polypropylene plate in triplicate with 0.5% (wt/vol)L-(+)-ara and incubated statically for 24 h at 25 °C. Following incubation,nonattached cells were removed, and the plate was rinsed thoroughly withwater. The plates were stained with 150 μL of 0.1% (wt/vol) crystal violet for10 min. The plate was rinsed, and adhered crystal violet was solubilized in200 μL of 95% (vol/vol) ethanol for 10 min with gentle agitation, and theabsorbance subsequently measured at 595 nm.

ACKNOWLEDGMENTS. We thank S. D. Tammam, L. M. Riley, and D. J. Little forhelpful discussions, P. Yip and C. A. Stremick for technical assistance, andJ. Koo for assistance with negative stain electron microscopy data collectionand analysis. This work was supported in part by Canadian Institutes ofHealth Research Grant 43998 (to P.L.H.); NIH Grant 2R01AI077628 (to M.R.P.);and Natural Sciences and Engineering Research Council (NSERC) DiscoveryGrant 435631 (to J.J.H.). P.L.H. and J.J.H. are recipients of Canada ResearchChairs; L.S.M., G.B.W., and J.C.W. were supported by NSERC Canada GraduateScholarships; and L.S.M. and J.C.W. were supported by graduate scholarshipsfrom the Ontario Graduate Scholarship Program and the Hospital for SickChildren Foundation Student Scholarship Program. Beamline X29 at NationalSynchrotron Light Source is supported by the US Department of Energy andthe NIH National Center for Research Resources.

1. Lyczak JB, Cannon CL, Pier GB (2000) Establishment of Pseudomonas aeruginosa in-

fection: Lessons from a versatile opportunist. Microbes Infect 2(9):1051–1060.2. Mishra M, et al. (2012) Pseudomonas aeruginosa Psl polysaccharide reduces neutro-

phil phagocytosis and the oxidative response by limiting complement-mediated op-

sonization. Cell Microbiol 14(1):95–106.

3. Vuong C, et al. (2004) A crucial role for exopolysaccharide modification in bacterial

biofilm formation, immune evasion, and virulence. J Biol Chem 279(52):54881–54886.4. Wang X, Preston JF, 3rd, Romeo T (2004) The pgaABCD locus of Escherichia coli

promotes the synthesis of a polysaccharide adhesin required for biofilm formation.

J Bacteriol 186(9):2724–2734.

2896 | www.pnas.org/cgi/doi/10.1073/pnas.1613606114 Marmont et al.

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1

5. Bobrov AG, Kirillina O, Forman S, Mack D, Perry RD (2008) Insights into Yersinia pestisbiofilm development: Topology and co-interaction of Hms inner membrane proteinsinvolved in exopolysaccharide production. Environ Microbiol 10(6):1419–1432.

6. Cramton SE, Gerke C, Schnell NF, Nichols WW, Götz F (1999) The intercellular adhe-sion (ica) locus is present in Staphylococcus aureus and is required for biofilm for-mation. Infect Immun 67(10):5427–5433.

7. Parise G, Mishra M, Itoh Y, Romeo T, Deora R (2007) Role of a putative polysaccharidelocus in Bordetella biofilm development. J Bacteriol 189(3):750–760.

8. Ferreira AS, et al. (2010) Distribution of cepacian biosynthesis genes among envi-ronmental and clinical Burkholderia strains and role of cepacian exopolysaccharide inresistance to stress conditions. Appl Environ Microbiol 76(2):441–450.

9. Friedman L, Kolter R (2004) Genes involved in matrix formation in Pseudomonasaeruginosa PA14 biofilms. Mol Microbiol 51(3):675–690.

10. Fong JCN, Syed KA, Klose KE, Yildiz FH (2010) Role of Vibrio polysaccharide (vps)genes in VPS production, biofilm formation and Vibrio cholerae pathogenesis.Microbiology 156(Pt 9):2757–2769.

11. Franklin MJ, Nivens DE, Weadge JT, Howell PL (2011) Biosynthesis of the Pseudo-monas aeruginosa extracellular polysaccharides, Alginate, Pel, and Psl. FrontMicrobiol 2:167.

12. Ramsey DM, Wozniak DJ (2005) Understanding the control of Pseudomonas aerugi-nosa alginate synthesis and the prospects for management of chronic infections incystic fibrosis. Mol Microbiol 56(2):309–322.

13. Starkey M, et al. (2009) Pseudomonas aeruginosa rugose small-colony variants haveadaptations that likely promote persistence in the cystic fibrosis lung. J Bacteriol191(11):3492–3503.

14. Jorth P, et al. (2015) Regional isolation drives bacterial diversification within cysticfibrosis lungs. Cell Host Microbe 18(3):307–319.

15. Govan JR, Deretic V (1996) Microbial pathogenesis in cystic fibrosis: Mucoid Pseudo-monas aeruginosa and Burkholderia cepacia. Microbiol Rev 60(3):539–574.

16. Häussler S, et al. (2003) Highly adherent small-colony variants of Pseudomonas aer-uginosa in cystic fibrosis lung infection. J Med Microbiol 52(Pt 4):295–301.

17. Colvin KM, et al. (2011) The pel polysaccharide can serve a structural and protectiverole in the biofilm matrix of Pseudomonas aeruginosa. PLoS Pathog 7(1):e1001264.

18. Colvin KM, et al. (2012) The Pel and Psl polysaccharides provide Pseudomonas aeru-ginosa structural redundancy within the biofilm matrix. Environ Microbiol 14(8):1913–1928.

19. Jennings LK, et al. (2015) Pel is a cationic exopolysaccharide that cross-links extra-cellular DNA in the Pseudomonas aeruginosa biofilm matrix. Proc Natl Acad Sci USA112(36):11353–11358.

20. Vasseur P, Vallet-Gely I, Soscia C, Genin S, Filloux A (2005) The pel genes of thePseudomonas aeruginosa PAK strain are involved at early and late stages of biofilmformation. Microbiology 151(Pt 3):985–997.

21. Ghafoor A, Jordens Z, Rehm BHA (2013) Role of PelF in pel polysaccharide bio-synthesis in Pseudomonas aeruginosa. Appl Environ Microbiol 79(9):2968–2978.

22. Whitney JC, et al. (2012) Structure of the cytoplasmic region of PelD, a degeneratediguanylate cyclase receptor that regulates exopolysaccharide production in Pseu-domonas aeruginosa. J Biol Chem 287(28):23582–23593.

23. Colvin KM, et al. (2013) PelA deacetylase activity is required for Pel polysaccharidesynthesis in Pseudomonas aeruginosa. J Bacteriol 195(10):2329–2339.

24. Vasseur P, Soscia C, Voulhoux R, Filloux A (2007) PelC is a Pseudomonas aeruginosaouter membrane lipoprotein of the OMA family of proteins involved in exopoly-saccharide transport. Biochimie 89(8):903–915.

25. Friedman L, Kolter R (2004) Two genetic loci produce distinct carbohydrate-richstructural components of the Pseudomonas aeruginosa biofilm matrix. J Bacteriol186(14):4457–4465.

26. Winsor GL, et al. (2011) Pseudomonas Genome Database: Improved comparativeanalysis and population genomics capability for Pseudomonas genomes. Nucleic AcidsRes 39(Database issue):D596–D600.

27. Edgar RC (2004) MUSCLE: Multiple sequence alignment with high accuracy and highthroughput. Nucleic Acids Res 32(5):1792–1797.

28. Hasegawa H, Holm L (2009) Advances and pitfalls of protein structural alignment.Curr Opin Struct Biol 19(3):341–348.

29. Terashima H, et al. (2013) Insight into the assembly mechanism in the supramolecularrings of the sodium-driven Vibrio flagellar motor from the structure of FlgT. Proc NatlAcad Sci USA 110(15):6133–6138.

30. Martinez RM, Jude BA, Kirn TJ, Skorupski K, Taylor RK (2010) Role of FlgT in an-choring the flagellum of Vibrio cholerae. J Bacteriol 192(8):2085–2092.

31. Goyal P, et al. (2014) Structural and mechanistic insights into the bacterial amyloidsecretion channel CsgG. Nature 516(7530):250–253.

32. Dong C, et al. (2006) Wza the translocon for E. coli capsular polysaccharides defines anew class of membrane protein. Nature 444(7116):226–229.

33. Koo J, et al. (2008) PilF is an outer membrane lipoprotein required for multi-merization and localization of the Pseudomonas aeruginosa type IV pilus secretin.J Bacteriol 190(21):6961–6969.

34. Macedo-Ribeiro S, et al. (1999) Crystal structures of the membrane-binding C2 do-main of human coagulation factor V. Nature 402(6760):434–439.

35. Yau WM, Wimley WC, Gawrisch K, White SH (1998) The preference of tryptophan formembrane interfaces. Biochemistry 37(42):14713–14718.

36. Okuda S, Tokuda H (2011) Lipoprotein sorting in bacteria. Annu Rev Microbiol 65:239–259.

37. Keiski C-L, et al. (2010) AlgK is a TPR-containing protein and the periplasmic com-ponent of a novel exopolysaccharide secretin. Structure 18(2):265–273.

38. Conlan S, Bayley H (2003) Folding of a monomeric porin, OmpG, in detergent solu-tion. Biochemistry 42(31):9453–9465.

39. Kleinschmidt JH, Wiener MC, Tamm LK (1999) Outer membrane protein A of E. colifolds into detergent micelles, but not in the presence of monomeric detergent.Protein Sci 8(10):2065–2071.

40. Barnhart MM, ChapmanMR (2006) Curli biogenesis and function. Annu Rev Microbiol60:131–147.

41. Nenninger AA, et al. (2011) CsgE is a curli secretion specificity factor that preventsamyloid fibre aggregation. Mol Microbiol 81(2):486–499.

42. Wang Y, et al. (2016) Structural basis for translocation of a biofilm-supportingexopolysaccharide across the bacterial outer membrane. J Biol Chem 291(19):10046–10057.

43. Stover CK, et al. (2000) Complete genome sequence of Pseudomonas aeruginosaPAO1, an opportunistic pathogen. Nature 406(6799):959–964.

44. Hmelo LR, et al. (2015) Precision-engineering the Pseudomonas aeruginosa genomewith two-step allelic exchange. Nat Protoc 10(11):1820–1841.

45. Choi KH, Schweizer HP (2006) mini-Tn7 insertion in bacteria with single attTn7 sites:Example Pseudomonas aeruginosa. Nat Protoc 1(1):153–161.

46. Lee JE, Cornell KA, Riscoe MK, Howell PL (2001) Structure of E. coli 5′-methyl-thioadenosine/S-adenosylhomocysteine nucleosidase reveals similarity to the purinenucleoside phosphorylases. Structure 9(10):941–953.

47. Green MR, Sambrook J (2012) Molecular Cloning: A Laboratory Manual (Cold SpringHarbor Laboratory Press, Cold Spring Harbor, NY), 4th Ed.

48. Choi KH, Kumar A, Schweizer HP (2006) A 10-min method for preparation of highlyelectrocompetent Pseudomonas aeruginosa cells: Application for DNA fragmenttransfer between chromosomes and plasmid transformation. J Microbiol Methods64(3):391–397.

49. de Lorenzo V, Timmis KN (1994) Analysis and construction of stable phenotypes ingram-negative bacteria with Tn5- and Tn10-derived minitransposons. MethodsEnzymol 235:386–405.

50. Zhao K, et al. (2013) Psl trails guide exploration and microcolony formation in Pseu-domonas aeruginosa biofilms. Nature 497(7449):388–391.

51. Almblad H, et al. (2015) The cyclic AMP-Vfr signaling pathway in Pseudomonas aer-uginosa is inhibited by cyclic Di-GMP. J Bacteriol 197(13):2190–2200.

52. Kearse M, et al. (2012) Geneious Basic: An integrated and extendable desktop soft-ware platform for the organization and analysis of sequence data. Bioinformatics28(12):1647–1649.

53. Winsor GL, et al. (2008) The Burkholderia Genome database: Facilitating flexiblequeries and comparative analyses. Bioinformatics 24(23):2803–2804.

54. Kelley LA, Sternberg MJE (2009) Protein structure prediction on the web: A case studyusing the Phyre server. Nat Protoc 4(3):363–371.

55. Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic inference undermixed models. Bioinformatics 19(12):1572–1574.

56. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in os-cillation mode. Methods Enzymol 276:307–326.

57. Morin A, et al. (2013) Collaboration gets the most out of software. eLife 2:e01456.58. Pape T, Schneider TR (2004) HKL2MAP: A graphical user interface for macromolecular

phasing with SHELX programs. J Appl Cryst 37:843–844.59. Terwilliger T (2004) SOLVE and RESOLVE: Automated structure solution, density

modification and model building. J Synchrotron Radiat 11(Pt 1):49–52.60. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. Acta

Crystallogr D Biol Crystallogr 60(Pt 12 Pt 1):2126–2132.61. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro-

molecular structure solution. Acta Crystallogr D Biol Crystallogr 66(Pt 2):213–221.62. Celniker G, et al. (2013) ConSurf: Using evolutionary data to raise testable hypotheses

about protein function. Isr J Chem 53:199–206.63. Ashkenazy H, Erez E, Martz E, Pupko T, Ben-Tal N (2010) ConSurf 2010: Calculating

evolutionary conservation in sequence and structure of proteins and nucleic acids.Nucleic Acids Res 38(Web Server issue):W529–W533.

64. Pettersen EF, et al. (2004) UCSF Chimera–a visualization system for exploratory re-search and analysis. J Comput Chem 25(13):1605–1612.

65. Salamitou S, et al. (1994) Subcellular localization of Clostridium thermocellum ORF3p,a protein carrying a receptor for the docking sequence borne by the catalytic com-ponents of the cellulosome. J Bacteriol 176(10):2828–2834.

66. O’Toole GA, et al. (1999) Genetic approaches to study of biofilms. Methods Enzymol310:91–109.

67. Merritt JH, Kadouri DE, O’Toole GA (2011) Growing and analyzing static biofilms. CurrProtoc in Micro, 10.1002/9780471729259.mc01b01s22.

68. Thein M, Sauer G, Paramasivam N, Grin I, Linke D (2010) Efficient subfractionation ofgram-negative bacteria for proteomics studies. J Proteome Res 9(12):6135–6147.

69. Chen VB, et al. (2010) MolProbity: All-atom structure validation for macromolecularcrystallography. Acta Crystallogr D Biol Crystallogr 66(Pt 1):12–21.

70. Hoang TT, Karkhoff-Schweizer RR, Kutchma AJ, Schweizer HP (1998) A broad-host-range Flp-FRT recombination system for site-specific excision of chromosomally-located DNA sequences: Application for isolation of unmarked Pseudomonas aeru-ginosa mutants. Gene 212(1):77–86.

71. Simon R, Priefer U, Pühler A (1983) A broad host range mobilization system for in vivogenetic engineering: Transposon mutagenesis in gram negative bacteria. NatBiotechnol 1:784–791.

72. Irie Y, Parsek MR (2014) LC/MS/MS-based quantitative assay for the secondary mes-senger molecule, c-di-GMP. Methods Mol Biol 1149:271–279.

73. Choi K-H, Schweizer HP (2005) An improved method for rapid generation of un-marked Pseudomonas aeruginosa deletion mutants. BMC Microbiol 5:30.

Marmont et al. PNAS | March 14, 2017 | vol. 114 | no. 11 | 2897

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 22

, 202

1