Structure of a putative BenF-like porin from Pseudomonas fluorescens Pf-5 at 2.6 Å resolution

Abstract

The X-ray structure of a putative BenF-like (gene name: PFL1329) protein from Pseudomonas fluorescens Pf-5 (PflBenF) has been determined at 2.6Å resolution. X-ray crystallography revealed a canonical 18-stranded β-barrel fold that forms a central pore with a diameter of ∼4.6Å, which is consistent with the size and physicochemical properties of the presumed aromatic acid substrate, benzoate. Detailed comparisons with the previously-determined structure of Pseudomonas aeruginosa OpdK, a vanillate influx channel, revealed an arginine-rich aromatic acid selectivity filter of nearly identical structure composed of seven highly conserved residues Arg∼Asp∼Arg∼Arg∼Ser∼Asp∼Arg (R∼D∼R∼R∼S∼D∼R sequence motif, where ∼ denotes intervening residues) that define the narrowest part of the pore.

Introduction

Gram-negative bacteria typically overcome poor permeability of outer membranes through general porins like OmpF and OmpC, which form water filled transmembrane pores permitting diffusion of hydrophilic molecules with no particular selectivity.¹ Many bacteria lacking such general porins utilize substrate-specific porins to overcome growth-limiting conditions and facilitate selective transport of metabolites. Exclusive reliance on substrate-specific porins yields lower membrane permeability to small molecules (<600 Da) versus that seen for Escherichia coli. In Pseudomonads, transit of most small molecules across the cell membrane is thought to be mediated by substrate-specific channels of the OprD superfamily. ² This property explains, at least in part, the high incidence of Pseudomonas aeruginosa antibiotic resistance. High-throughput DNA sequencing of the P. aeruginosa chromosome revealed the presence of 19 genes encoding structurally related, substrate-specific porins (with 30-45% pairwise amino acid sequence identity) that mediate transmembrane passage of small, water-soluble compounds. The OprD superfamily encompasses the eponymous OprD subfamily, which includes 9 P. aeruginosa proteins that convey basic amino acids and carbapenem antibiotics,³ and the OpdK subfamily, which includes 11 P. aeruginosa proteins that convey aromatic acids and other small aromatic compounds.⁴ Genome sequencing of other Gram-negative bacteria has revealed additional members of the OprD and OpdK subfamilies in various organisms, including other pseudomonads. Among the many bacteria in which OprD superfamily members have been identified are P. putida, P. fluorescens Pf-5, P. syringae, and Azotobacter vinelandii, all of which share closely related genes that encode the so-called BenF-like porins. In P. putida, benF is part of an operon involved in benzoate catabolism regulated by benR.⁵ Within this operon benK, benE, and benF genes have been suggested to contribute towards either influx or efflux of benzoate.^5,6 BLAST⁷ analysis of the amino acid sequence of P. fluorescens Pf-5 gene PFL1329 (Uniprot⁸ id: http://www.uniprot.org/uniprot/Q4KH25)⁹ against P. putida KT2440 strain¹⁰ identified twenty related porins. The top six hits include P. putida KT2440 genes PP1383 (annotated as BenF-like), PP2517 (annotated as BenF-like), and PP3168 (annotated as BenF), which share sequence identities of 76%, 66%, and 44% with PFL1329, respectively. The precise functions of these genes are not yet known. Therefore, we refer to the protein product of gene PFL1329 as PflBenF-like, which reflects its current annotation in the Unitprot database.

Crystal structures of OprD¹¹ and OpdK (vanillate specific porin)¹², both from P. aeruginosa (designated below as PaOprD and PaOpdK, respectively) have been determined. Herein, we report crystal structure of a putative BenF-like porin from P. fluorescens Pf-5 (PflBenF-like). For the sake of brevity, all subsequent references to the PflBenF-like will made using PflBenF. X-ray crystallography revealed a canonical 18-stranded β-barrel fold that forms a central pore with a diameter of ∼4.6Å. We describe detailed comparisons of the PflBenF structure with those of PaOprD and PaOpdK.

Methods

Cloning and expression of PflBenF

The gene encoding putative P. fluorescens BenF-like porin was cloned from genomic DNA of the strain Pf-5 (American Type Culture Collection, USA). The desired truncation (encoding residues 30-420) was PCR amplified using GAGTCGGGCTTTCTCGAAGATGC and CCAGCAGGCTGAGGGGATAGC as forward and reverse primers, respectively. The purified PCR product was subsequently TOPO^® (Invitrogen, USA) cloned into pSGX3, a derivative of pET26b(+), giving rise to a fusion protein with a non-cleavable C-terminal hexa-histidine tag. Plasmids were transfected into BL21(DE3)-Condon+RIL (Invitrogen, USA) cells for overexpression. Expression was carried out at 22 °C in 4L of Terrific Broth¹³ supplemented with kanamycin (50 μg/ml) and chloramphenicol (35 μg/ml). Protein expression was induced by addition of 0.4 mM IPTG at an OD₆₀₀ of 1.0. Cells were harvested after 21 hours by centrifugation at 4 °C. Virtually all PflBenF protein was found in inclusion bodies.

Purification and refolding of PflBenF

The E. coli cell pellet was resuspended in 30mL of cold buffer (20 mM Tris HCl pH 8.0, 500 mM NaCl, 25 mM imidazole, and 0.1% Tween20) and cells were lysed via sonication. Inclusion bodies were pelleted by centrifugation at 4 °C and solubilized in 20 mM TrisHCl pH 8.0, 1 mM β-mercaptoethanol, and 6 M guanidinium chloride. Solubilized PflBenF was purified under denaturing conditions via immobilized metal ion affinity chromatography with NiNTA resin by washing the column with solubilization buffer plus 50 mM imidazole. Following elution with the solubilization buffer plus 400 mM imidazole, PflBenF was diluted to a concentration of ∼1 mg/mL and refolded by drop wise dilution into a buffer containing 20 mM TrisHCl pH 8.0, 150 mM NaCl, 0.5 mM TCEP, and 0.5% (w/v) lauryldimethylamineoxide (LDAO) to a final protein concentration of ∼250 μg/mL. The refolded protein was concentrated by immobilization on NiNTA resin followed by elution with 20 mM TrisHCl pH 8.0, 200 mM NaCl, 10% (v/v) glycerol, 0.5 mM TCEP, 5 mM LDAO, and 400 mM imidazole. Refolded PflBenF was further purified by size exclusion chromatography using a 120 mL Superdex S200 column equilibrated with 20 mM TrisHCl pH 8.0, 200 mM NaCl, 10% (v/v) glycerol, 0.5 mM TCEP, and 5 mM LDAO. SDS-PAGE analysis showed greater than 95% purity and protein fractions corresponding to the symmetric portion of the size exclusion chromatography profile were pooled and concentrated using spin filters. Size Exclusion Chromatography with Multi Angle Laser Light Scattering (SEC-MALLS) revealed PflBenF to be a monomer associated with one LDAO micelle (data not shown).

Crystallization, data collection, and structure determination

PflBenF (protein concentration∼9.0 mg/mL; 0.3 μL protein-containing solution + 0.3 μL reservoir solution) was subjected to crystallization screening with the Classics, Classics II, ComPAS, and PEG kits (Qiagen, USA) using a Phoenix Liquid Handling System (Art Robbins Instruments, USA) via sitting drop vapor diffusion at 4 and 21 °C. Several conditions yielded thin needles and microcrystals. Subsequent optimization was performed with additive and detergent screening. A single crystal diffracting to ∼2.6Å resolution was obtained with 21% PEG 3350, 100 mM citrate, and 10% (w/v) ANAPOE-X-114 (from the detergent screen) at 21 °C. Diffraction data were recorded under standard cryogenic conditions using the LRL-CAT 31-ID Beamline at the Advanced Photon Source (APS) and processed with MOSFLM¹⁴ and SCALA (Collaborative Computing Project Number 4, 1994).¹⁵ A poly-alanine model of PaOpdK was used for molecular replacement with PHASER¹⁶ as implemented in CCP4. The atomic model of PflBenF was built manually by visual inspection with COOT¹⁷ and refined to convergence using REFMAC5¹⁸ [Table 1]. Structural analyses were carried out using COOT and CCP4 and illustrations were prepared using PyMol (http://pymol.sourceforge.net).

Table 1.

Crystallographic data collection and refinement statistics for PflBenF.

Data Collection
PDB code	3JTY
Space group:	P21
Unit-cell dimensions (Å, °):	a=62.1, b=210.6, c=84.1, β=97.8
Resolution (Å):	53.68-2.58 (2.72-2.58)*
Number of unique reflections:	67053 (9774)
Completeness (%):	100.0 (100.0)
Rsymm (%):	14.0 (52.7)
Multiplicity:	3.9 (3.8)
< I/σ(I) >:	6.1 (2.7)
Refinement
Resolution (Å):	35.11-2.58
R-factor (%):	22.0
Rfree (%):	27.5
Number of non-hydrogen atoms
Protein	11859
Ligands	32
Water molecules	25
Average B-factors (Å2)
Protein	41.9
Ligands	75.4
Water molecules	35.2
RMS deviations from ideal values
Bond length (Å):	0.018
Bond angles (°)	1.60
Ramachandran Plot25:
MolProbity26 Residues in
Favored region (%):	95.6
Allowed region (%):	99.9

^*Values in parenthesis correspond to highest-resolution shell

Results and Discussion

Overall structure of PflBenF-like porin

An N-terminal truncation of the putative BenF-like porin from P. fluorescens Pf-5 (PflBenF, residues 30-420) was over-expressed in E. coli, purified under denaturing conditions from inclusion bodies by immobilized metal ion affinity chromatography, refolded by dilution into a buffer containing LDAO, and further purified by size exclusion chromatography. PflBenF crystallized in the monoclinic space group P2₁ with four molecules in the asymmetric unit [Table 1]. The PflBenF polypeptide chain could be continuously traced for one protomer (denoted chain A and used for all illustrations and analyses herein). In each of chains B, C, and D, residues 112-115 were disordered in the experimental electron density map and could not be modeled. Otherwise, structures of the four protomers are very similar [pairwise root mean square deviations (r.m.s.d.s) range from 0.31 to 0.39Å for common α-carbon atomic pairs].

PflBenF adopts the canonical β-barrel porin fold consisting of 18 anti-parallel β-strands arranged in a cylinder [Figs. 1A and B]. Pairwise comparisons of the structure of PflBenF with those of PaOpdK and PaOprD using SSM¹⁹ revealed highly similar polypeptide chain folds (PflBenF versus PaOpdK: r.m.s.d. = 0.9Å, sequence identity = 53% for 369 α-carbon pairs; PflBenF versus PaOprD: r.m.s.d. = 1.3Å, sequence identity = 40% for 342 α-carbon pairs; PaOpdK versus PaOprD: r.m.s.d. = 1.3Å, sequence identity = 39% for 382 α-carbon pairs). The β-strands of PflBenF are connected by shorter turns T1-T8 and longer loops L1-L7 on the periplasmic and extracellular faces of the barrel, respectively. Loops L3 and L7 fold into the interior of the β-barrel, where they help define the geometry of the pore [Fig. 1B]. As for PaOpdK and PaOprD, PflBenF also possesses short S5 and S6 β-strands, which are characteristic of porins that form trimers in the outer membrane.²⁰ Notwithstanding this similarity, PflBenF protomers do not make extensive intermolecular interactions within the crystal lattice. Recombinant forms of both PaOpdK and PaOprD also appear monomeric in the crystal lattice. However, recomibant PaOpdK inserted into artificial lipid bilayer oligomerizes and behaves as a trimeric ion-conductor.¹¹ Given the short lengths of the S5 and S6 β-strands in the PflBenF-like porin and the PaOpdK precedent, there is no compelling reason to believe that PflBenF functions as a monomer.

Figure 1. Structure of PflBenF.

A: Rainbow-colored ribbon drawing of PflBenF with a dot surface representation of the pore calculated using HOLE²¹ viewed parallel to the cell membrane.

B: Stereodrawing of PflBenF with the pore viewed from the periplasmic face of the protein (using the same color coding as in Fig. 1A).

C: Sequence alignment of PlfBenF (gi|123657525; PDB Code 3JTY), P. putida BenF (gi|81441797), P. aeruginosa OpdK porin (gi|81539679; PDB Code 2QTK), P. putida outer membrane protein (gi|167033173), Acinetobacter baumannii VanP (gi|193076825), A. sp. RUH2654 OprD family porin (gi|260549927), and P. aeruginosa D2 porin (gi|417518; PDB Code 2ODJ) generated using ClustalW²³ and color coded with ESPRIPT²⁴; numbers after “_” denote sequence identity with respect to the PflBenF and aromatic acid selectivity pore residues are marked with .

D: Semitransparent representation of the solvent accessible periplasmic (upper panel) and extracellular (lower panel) surfaces of PlfBenF, color coded for qualitative electrostatic potential (generated using PyMol) shown with polypeptide chain as ribbon, using the same orientation as in Fig. 1B.

Structure of the PflBenF benzoate-specific channel

The structure of the PflBenF β-barrel revealed a quasi-circular pore [Fig. 1B] with a diameter of ∼4.6Å (as estimated with HOLE²¹). Sequence alignment of PflBenF, P. putida BenF, PaOpdK, PaOprD, and three other porins with various annotations (pairwise sequence identities with PflBenF ranging from 34 to 76%) revealed significant conservation for β-strand residues [Fig. 1C] and those within the loops L2, L3, and L7. The electrostatic properties of the periplasmic and extracellular faces of PflBenF are illustrated in Fig. 1D. On the periplasmic face, conserved basic and non-polar surface features are visible in the immediate vicinity of channel constriction. There is no corresponding concentration of polar surface features on the extracellular side of the channel immediately beyond the constriction. In fact, the extracellular face of the protein is largely devoid of polar features [Fig. 1D lower panel]. The entire periplasmic face of PflBenF is also considerably more polar than the extracellular face of the protein.

Evidence for an aromatic acid selectivity filter shared by PflBenF and P. aeruginosa OpdK

The pore constriction of PflBenF is lined by residues Asp151 and Arg154 from loop L3; Ser310, Asp317, Ser318, and Tyr319 from loop L7; Phe46, Phe48, and Arg50 from the S1 β-strand; Arg186 from loop L4; and Glu408 and Arg410 from the S18 β-strand [Fig. 2A]. These four Arg residues decorate the perpiplasmic lip of the pore giving rise to a discrete positively charged patch between 4 and 6 O'clock [Fig 1D, upper panel]. Of these 12 pore forming residues, four arginines and an aspartate at position 317 are invariant among the six divergent sequences included in Fig. 1C. In addition, Asp151 and Ser310 are highly conserved.

Figure 2. Structure of a putative aromatic acid selectivity filter.

A: Atomic stick figure representation of the PflBenF pore forming residues surrounding the dot surface of the pore²¹ viewed parallel to the cell membrane (carbon atoms color coded to correspond to the representation in Fig. 1A: blue denotes β1, green denotes L3 and L4, light brown denotes L7, pink denotes β10; nitrogen atoms-blue; oxygen atoms-red).

B: Stereoview overlay of selected pore forming residues of PlfBenF with residues defining the aromatic acid selectivity pore (color coded as in Fig. 2A, labeled with PflBenF sequence numbering) and P. aeruginosa OpdK (carbon atoms color coded cyan).

C: Stereoview overlay of selected pore forming residues of PlfBenF (color coded as in Fig. 2A) and P. aeruginosa OprD (carbon atoms color coded yellow).

Given that PlfBenF and PaOpdK are thought to transport chemically similar substrates (putative benzoate selectivity versus experimentally confirmed vanillate uptake), it is not unexpected that their respective pore constrictions are essentially identical in size (diameters of ∼4.6Å and ∼4.4Å, respectively) and shape [Fig. 2B]. What is surprising, however, is the finding that the seven pore forming residues with sidechains that contribute directly to the constriction (or selectivity filter) are identical between PflBenF and PaOpdK [Figs. 1C and 2B]. Specifically, Arg50 (Arg22, italics denotes PaOpdK residues), Asp151 (Asp123), Arg154 (Arg126), Arg186 (Arg158), Ser310 (Ser282), Asp317 (Asp289), and Arg410 (Arg381) are invariant between these two porins [Figs. 1C and 2B]. The only notable difference is the sidechain conformation of Arg410, which brings its guanidinium group closer to the constriction in the case of PflBenF. In contrast, the PaOprD pore constriction is smaller than that of PlfBenF and PaOpdK (∼3.7Å versus ∼4.5Å), differs in shape, and is defined by a different configuration of selectivity filter residues [Fig. 2C]. Thus, the selectivity determining residues of the vanillate and putative benzoate pores appear to be similar. Further characterization of the function of PflBenF and other porins sharing this sequence motif will be required to establish whether or not it is in fact a signature of aromatic acid channels.

One of the metrics guiding target selection in structural genomics is modeling leverage at the level of greater than 30% sequence identity (see Supplementary Methods).²² Searching of the UniProtKB database of all known protein sequences using the sequences of PlfBenF, PaOpdK, and PaOprD identified 221 unique protein sequences with sequence identity greater than 30% to at least one of these three proteins. The inaugural structure of PaOprD, enabled modeling of 165 related proteins (Supplementary Figure S1). Subsequent determination of the structure of PaOpdK enabled homology modeling of an additional 3 protein sequences. In contrast, determination of the PflBenF structure enabled homology modeling of 53 additional protein sequences (Supplementary Figure S1). Thus, the structure of PflBenF, reported herein, expands significantly the number of homology models of OprD and OpdK family of substrate-specific porins. Experimental structures of additional OprD / OpdK subfamily members should provide useful guides for planning experiments aimed at defining the mechanisms governing pore selectivity.